Compare commits

..

17 Commits

Author SHA1 Message Date
lizzie 58123ef9c9 let openbsd be fucked then i guess (FOR NOW) 2026-08-09 16:12:22 +00:00
lizzie 63f2278c96 fx 2026-08-09 15:15:44 +00:00
lizzie d7300e2f6f qt diffs 2026-08-09 15:15:44 +00:00
lizzie 671fe8ca3f oops stray file 2026-08-09 15:15:44 +00:00
lizzie 4792e6705e oh, we were the ones doing it wrong, ok 2026-08-09 15:15:44 +00:00
lizzie 7fd12c3ae6 correct qt6 instructions 2026-08-09 15:15:44 +00:00
lizzie 68499cad1a ok x11 fine 2026-08-09 15:15:44 +00:00
lizzie cb92bc64b0 nvm not x11::x11 2026-08-09 15:15:44 +00:00
lizzie 89d03fc719 better x11 2026-08-09 15:15:44 +00:00
lizzie f34a73e5fe ok gcc14 2026-08-09 15:15:44 +00:00
lizzie eb511eacb0 [dynarmic] fix NetBSD builds on gcc14
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-08-09 15:15:44 +00:00
BoiledElectricity 13d39f39aa [qt] fix relative ROM paths passed on the CLI on macOS (#4105)
On macOS, convert relative paths to absolute paths before changing the
working directory. Otherwise, valid Switch formats may incorrectly fail
with a "The ROM format is not supported" error.

Fixes: https://github.com/eden-emulator/Issue-Reports/issues/404
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4105
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: Samuel <lizzie@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
2026-08-09 17:18:58 +02:00
lizzie 7731b5bc86 [core/hle/kernel/svc] demote ResetSignal log Debug->Trace (#4161)
I don't see why it's LOG_DEBUG if WaitSynchronization is LOG_TRACE, leaves incomplete data
may as well bundle it to trace with the others? :)

Signed-off-by: lizzie <lizzie@eden-emu.dev>

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4161
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>
Reviewed-by: Maufeat <sahyno1996@gmail.com>
2026-08-09 14:13:27 +02:00
lizzie c4ab2f6b9e [cmake] Add ENABLE_WERROR option (#4080)
Should warnings be errors? that's a philosophical debate which I can only answer with "no, but if you think otherwise here is an option for you"

Signed-off-by: lizzie <lizzie@eden-emu.dev>

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4080
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
2026-08-09 02:56:28 +02:00
xbzk dd12266c26 [input] added option to disable wgi/xinput to prevent SDL GUIDE hack (#4237)
- [x] I have read and followed the [Contribution Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/CONTRIBUTING.md#code-contributions).
- [x] I have read and followed the [AI Policy](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/AI.md)
- [x] I have read and followed the [Coding Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/Coding.md) to the best of my ability.

-------------------
So, many users hate the fact when they press START+SELECT, an additional HOME event is sent.

That is annoying coz for most users' default hotkey settings, it closes eden when not in game, and when in game it calls a NOT controller responsive prompt about leaving Eden.
There are several games with reasons to press start+select, also there are homebrews which relies on that combo to open internal menus.
Again. That is annoying.

Digging deep i've found that SDL implements something called guide hack: When Start+Select is pressed it sends a synthetic GUIDE(which code equals to HOME).
This was intended for controllers missing the HOME button, but SDL fails to track them all, and misses API to disabling it.

I've figured out a patch to SDL to disable it, but since it would be better to invest on a PR straight into SDL repo, on our side i thought it better to simply add an option to disable WGI (Windows.Gaming.Input) and Xinput, both paths which could lead into the hack enabling.

Setting is available in a remote place no one will bother.

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4237
Reviewed-by: Samuel <lizzie@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
2026-08-09 02:53:37 +02:00
crueter a3d32a18b8 [meta] Discourage profanity usage (#4239)
I'm not into millennial humor, personally.

Proactively removed as much profanity as I could grep.

Signed-off-by: crueter <crueter@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4239
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Samuel <lizzie@eden-emu.dev>
2026-08-09 02:53:16 +02:00
crueter c0ffc900cd [ci] Only update translations once a week (#4256)
They've slowed down enough at this point.

Signed-off-by: crueter <crueter@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4256
2026-08-08 05:46:52 +02:00
72 changed files with 222 additions and 1556 deletions
+2 -2
View File
@@ -1,9 +1,9 @@
name: tx-pull
on:
# tuesday, saturday at 2pm
# monday at 4pm
schedule:
- cron: '0 14 * * 2,6'
- cron: '0 16 * * 1'
workflow_dispatch:
jobs:
+19 -13
View File
@@ -23,20 +23,12 @@ include(CMakeDependentOption)
include(CTest)
include(CPMUtil)
if (NOT DEFINED ARCHITECTURE)
message(FATAL_ERROR "Architecture didn't make it out of scope, did you delete DetectArchitecture.cmake?")
if (CXX_GCC AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS "14.0")
message(FATAL_ERROR "You must use gcc14 (or newer) feel free to fix clang yourself otherwise.")
endif()
# Needed for FFmpeg w/ VAAPI and DRM
if (PLATFORM_OPENBSD)
# OpenBSD 7.8 broke libcxx when upgrading, so we must define the PSTL backend manually
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R6/include -D_LIBCPP_PSTL_BACKEND_SERIAL=1")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R6/include -D_LIBCPP_PSTL_BACKEND_SERIAL=1")
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -L${CMAKE_SYSROOT}/usr/X11R6/lib")
elseif (PLATFORM_NETBSD)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R7/include -I${CMAKE_SYSROOT}/usr/pkg/include/c++/v1")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${CMAKE_SYSROOT}/usr/X11R7/include -I${CMAKE_SYSROOT}/usr/pkg/include/c++/v1")
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -L${CMAKE_SYSROOT}/usr/X11R7/lib")
if (NOT DEFINED ARCHITECTURE)
message(FATAL_ERROR "Architecture didn't make it out of scope, did you delete DetectArchitecture.cmake?")
endif()
# NetBSD: Fun for the whole family!
@@ -79,6 +71,7 @@ set(YUZU_QT_MIRROR "" CACHE STRING "What mirror to use for downloading the bundl
cmake_dependent_option(YUZU_USE_BUNDLED_QT "Download bundled Qt binaries" "${MSVC}" "ENABLE_QT" OFF)
option(ENABLE_DEBUG_TOOLS "Enable debugging tools (maxwell disassembler, SPIRV translator, etc)" OFF)
option(ENABLE_WERROR "Enable -Werror diagnostics" ON)
# non-linux bundled qt are static
if (YUZU_USE_BUNDLED_QT AND (APPLE OR NOT UNIX))
@@ -157,6 +150,19 @@ if (CXX_CLANG_CL)
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-reserved-identifier>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-deprecated-declarations>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type-mismatch>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-extensions>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++17-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-reserved-user-defined-literal>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-deprecated-writable-strings>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat-pedantic>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++11-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++0x-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-c++11-compat-binary-literal>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-compat-pedantic>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c++98-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c98-compat>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-c99-extensions>
$<$<COMPILE_LANGUAGE:C,CXX>:/EHsc>)
# REQUIRED CPU features IN Windows-amd64
if (ARCHITECTURE_x86_64)
@@ -604,7 +610,7 @@ if (ENABLE_QT)
# Best practice is to ask for all components at once, so they are from the same version
set(YUZU_QT_COMPONENTS Core Widgets Charts Concurrent Gui)
if (PLATFORM_LINUX OR PLATFORM_FREEBSD)
if (NOT WIN32 AND NOT APPLE)
list(APPEND YUZU_QT_COMPONENTS DBus)
# yes Qt, we get it
set(QT_NO_PRIVATE_MODULE_WARNING ON)
+5 -1
View File
@@ -22,9 +22,13 @@ Eden is free, open-source, copyleft software, licensed under the terms of the [G
- No LLM or AI usage, *period*, for patches, pull requests, issues, comments, debugging, brainstorming, etc.
- For details on why, see the [detailed AI policy](docs/policies/AI.md).
- Usage of any form of profanity or otherwise unsavory language is generally discouraged.
- This is primarily because it rarely helps to actually understand what's going on.
- Remember that your comments should be focused and actually address what's happening. With very few exceptions, expletives are actively detrimental at best.
- New code must follow the same general style as the surrounding codebase. Exceptions may be granted in certain cases.
- Maintainers reserve the right to change your patches and pull requests at will. We will try to avoid this.
- You should respect all decisions made by the [code owners](docs/CODEOWNERS) in your particular subsystem. If you feel they are overstepping or are incorrect, don't be afraid to stand your ground!
- You should generally respect all decisions made by the [code owners](docs/CODEOWNERS) in your particular subsystem.
- However, if you feel they are overstepping or are incorrect, don't be afraid to stand your ground! Maintainers are not always correct.
- While we do *not* adhere to the terms of a formal code of conduct, you will generally be expected to respect other developers, contributors, and community members.
- You **must** have basic knowledge of [Git](https://git-scm.com/learn). Knowing how to manage your branches and follow proper fork policies is a necessity.
+3 -1
View File
@@ -71,7 +71,7 @@ export LIBGL_ALWAYS_SOFTWARE=1
Install `developer/gcc14` on OmniOS using pkgsrc.
Since so many dependencies are missing on `OmniOS`, you may wish to use `-DCPMUTIL_FORCE_BUNDLED=ON`
Since so many dependencies are missing on `OmniOS`, you may wish to use `-DCPMUTIL_FORCE_BUNDLED=ON` and `-DYUZU_USE_BUNDLED_OPENSSL=OFF`.
For OmniOS you are required to build glslang yourself:
```sh
@@ -92,6 +92,8 @@ You may also need to install `gmake` in order to properly build FFmpeg, this is
If it wasn't obvious already, you require a X11 server to properly run the emulator within OmniOS, [this guide](https://web.archive.org/web/20260424200928/https://geekblood.wordpress.com/2017/10/26/installing-x11-and-a-desktop-environment-on-omnios/) is a great starting point for that, the links to pkgsrc are outdated so follow [this exemplar](https://pkgsrc.smartos.org/install-on-illumos/) as well:
For Solaris based OSes, `${CMAKE_SYSTEM_NAME}` isn't properly set on CMake (it's set to i686 on AMD64), you may find issues when building OpenSSL from `openssl-cmake`.
## HaikuOS
It's recommended to do a `pkgman full-sync` before installing. See [HaikuOS: Installing applications](https://www.haiku-os.org/guides/daily-tasks/install-applications/). Sometimes the process may be interrupted by an error like "Interrupted syscall". Simply firing the command again fixes the issue. By default `g++` is included on the default installation.
+1 -1
View File
@@ -273,7 +273,7 @@ If using FreeBSD 12 or prior, use `devel/pkg-config` instead.
For NetBSD +10.1:
```sh
pkgin install git cmake boost fmtlib SDL3 catch2 libjwt spirv-headers spirv-tools ffmpeg7 libva nlohmann-json jq libopus qt6 cpp-httplib lz4 vulkan-headers nasm autoconf enet pkg-config libusb1 libcxx frozen
pkgin install git cmake boost fmtlib SDL3 catch2 libjwt spirv-headers spirv-tools ffmpeg7 libva nlohmann-json jq libopus qt6-qtbase qt6-qtcharts qt6-qtmultimedia qt6-qttools cpp-httplib lz4 vulkan-headers nasm autoconf enet pkg-config libusb1 libcxx frozen
```
[Caveats](./Caveats.md#netbsd).
+2
View File
@@ -46,6 +46,8 @@ These options control dependencies.
- `YUZU_INSTALL_UDEV_RULES` (OFF) Install udev rules to enable hidraw access
- Needed for gyroscopes
- Only available on Linux
- `ENABLE_DEBUG_TOOLS` (OFF) Enables debugging and development tools, see [tools](../tools/README.md).
- `ENABLE_WERROR` (ON) Enables warnings as errors (-Werror).
### Flavors
+1 -1
View File
@@ -34,7 +34,7 @@ Everyone has their own way of viewing good/bad C++ practices, my general outline
- The reason is because the project has `-fno-rtti` disabled by default, due to the costs of dynamic polymorphism.
- Always copy-on-value for objects with `sizeof(void *) >= sizeof(T) * 2`, i.e objects sized as 2 pointers or less, for bigger objects you can use ref/pointer as usual.
- Try using move semantics instead of references, whenever possible.
- Remember function parameters are extremelly cheap as fuck, don't be afraid to place upto 8 parameters on a given function.
- Function parameters are cheap. Don't be afraid to use as many as needed (API usability permitting).
- Don't save a reference in structures of a parent object, i.e:
```c++
+2 -2
View File
@@ -1177,7 +1177,7 @@ static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int re
#endif
#ifndef STBI_NO_TGA
// test tga last because it's a crappy test!
// test tga last because it's a bad test!
if (stbi__tga_test(s))
return stbi__tga_load(s,x,y,comp,req_comp, ri);
#endif
@@ -7662,7 +7662,7 @@ static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp)
if (stbi__hdr_info(s, x, y, comp)) return 1;
#endif
// test tga last because it's a crappy test!
// test tga last because it's a bad test!
#ifndef STBI_NO_TGA
if (stbi__tga_info(s, x, y, comp))
return 1;
+20 -8
View File
@@ -155,12 +155,24 @@ else()
$<$<COMPILE_LANGUAGE:CXX>:-fno-rtti>)
endif()
if (ENABLE_WERROR)
add_compile_options(
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=all>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=extra>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=missing-declarations>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unused>)
else()
add_compile_options(
$<$<COMPILE_LANGUAGE:C,CXX>:-Wall>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wextra>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wmissing-declarations>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wshadow>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wunused>
# Some compilers could particularly misbehave :)
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-error-all>)
endif()
add_compile_options(
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=all>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=extra>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=missing-declarations>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unused>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-attributes>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-invalid-offsetof>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-unused-parameter>
@@ -169,9 +181,9 @@ else()
if (CXX_CLANG OR CXX_ICC OR CXX_APPLE) # Clang, AppleClang, or Intel C++
if (NOT MSVC)
add_compile_options(
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=shadow-uncaptured-local>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=implicit-fallthrough>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=type-limits>)
$<$<COMPILE_LANGUAGE:C,CXX>:-Wshadow-uncaptured-local>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wimplicit-fallthrough>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wtype-limits>)
endif()
add_compile_options(
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-braced-scalar-init>
@@ -27,7 +27,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
RENDERER_ASYNCHRONOUS_GPU_EMULATION("use_asynchronous_gpu_emulation"),
RENDERER_ASYNC_PRESENTATION("async_presentation"),
RENDERER_ASYNCHRONOUS_SHADERS("use_asynchronous_shaders"),
RENDERER_UNIFIED_MEMORY("use_unified_memory"),
RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
ENABLE_BUFFER_HISTORY("enable_buffer_history"),
USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
@@ -685,13 +685,6 @@ abstract class SettingsItem(
descriptionId = R.string.renderer_asynchronous_shaders_description
)
)
put(
SwitchSetting(
BooleanSetting.RENDERER_UNIFIED_MEMORY,
titleId = R.string.renderer_unified_memory,
descriptionId = R.string.renderer_unified_memory_description
)
)
put(
SingleChoiceSetting(
IntSetting.FAST_GPU_TIME,
@@ -304,7 +304,6 @@ class SettingsFragmentPresenter(
add(BooleanSetting.EMULATE_BGR565.key)
add(BooleanSetting.RESCALE_HACK.key)
add(BooleanSetting.RENDERER_ASYNCHRONOUS_SHADERS.key)
add(BooleanSetting.RENDERER_UNIFIED_MEMORY.key)
add(IntSetting.ANDROID_PIPELINE_WORKERS.key)
add(BooleanSetting.RENDERER_ASYNCHRONOUS_GPU_EMULATION.key)
add(BooleanSetting.RENDERER_ASYNC_PRESENTATION.key)
@@ -524,8 +524,6 @@
<string name="rescale_hack_description">Enables a legacy handling for the rescale configuration pass for games by using a quick rescale path</string>
<string name="renderer_asynchronous_shaders">Use asynchronous shaders</string>
<string name="renderer_asynchronous_shaders_description">Compiles shaders asynchronously. This may reduce stutters but may also introduce glitches.</string>
<string name="renderer_unified_memory">Unified memory access</string>
<string name="renderer_unified_memory_description">Allows GPU write buffer readbacks directly into guest memory, skipping the CPU staging copy.</string>
<string name="gpu_unswizzle_settings">GPU Unswizzle Settings</string>
<string name="gpu_unswizzle_settings_description">Configure GPU-based texture unswizzling parameters or disable it entirely. Adjust these settings to balance performance and texture loading quality.</string>
<string name="gpu_unswizzle_enable">Enable GPU Unswizzle</string>
+1 -1
View File
@@ -222,7 +222,7 @@ if (MSVC)
)
else()
target_compile_options(audio_core PRIVATE
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>)
endif()
+1 -1
View File
@@ -220,7 +220,7 @@ endif()
if(CXX_CLANG)
target_compile_options(common PRIVATE
$<$<COMPILE_LANGUAGE:C,CXX>:-fsized-deallocation>
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=unreachable-code-aggressive>)
$<$<COMPILE_LANGUAGE:C,CXX>:-Wunreachable-code-aggressive>)
target_compile_definitions(
common
PRIVATE
+8 -21
View File
@@ -31,9 +31,7 @@ s64 GetMaxPermissibleResidentMapCount() {
} // namespace
HeapTracker::HeapTracker(Common::HostMemory& buffer)
: m_buffer(buffer),
m_has_hardware_buffer_backing(!buffer.BackingHardwareBuffers().empty()),
m_max_resident_map_count(GetMaxPermissibleResidentMapCount()) {}
: m_buffer(buffer), m_max_resident_map_count(GetMaxPermissibleResidentMapCount()) {}
HeapTracker::~HeapTracker() = default;
void HeapTracker::Map(size_t virtual_offset, size_t host_offset, size_t length,
@@ -87,8 +85,7 @@ void HeapTracker::Unmap(size_t virtual_offset, size_t size, bool is_separate_hea
// If resident, erase from resident map.
if (item->is_resident) {
m_resident_map_count -= this->HostMapCount(item->paddr, item->size);
ASSERT(m_resident_map_count >= 0);
ASSERT(--m_resident_map_count >= 0);
m_resident_mappings.erase(m_resident_mappings.iterator_to(*item));
}
@@ -194,7 +191,7 @@ bool HeapTracker::DeferredMapSeparateHeap(size_t virtual_offset) {
// This map is now resident.
it->is_resident = true;
m_resident_map_count += this->HostMapCount(it->paddr, it->size);
m_resident_map_count++;
m_resident_mappings.insert(*it);
}
@@ -216,17 +213,17 @@ void HeapTracker::RebuildSeparateHeapAddressSpace() {
// Despite being worse in theory, this has proven to be better in practice than more
// regularly dumping a smaller amount, because it significantly reduces average case
// lock contention.
s64 const desired_count = (std::min)(m_resident_map_count, m_max_resident_map_count) / 2;
std::size_t const desired_count = (std::min)(m_resident_map_count, m_max_resident_map_count) / 2;
std::size_t const evict_count = m_resident_map_count - desired_count;
auto it = m_resident_mappings.begin();
while (m_resident_map_count > desired_count && it != m_resident_mappings.end()) {
for (size_t i = 0; i < evict_count && it != m_resident_mappings.end(); i++) {
// Unmark and unmap.
it->is_resident = false;
m_buffer.Unmap(it->vaddr, it->size, false);
// Advance.
m_resident_map_count -= this->HostMapCount(it->paddr, it->size);
ASSERT(m_resident_map_count >= 0);
ASSERT(--m_resident_map_count >= 0);
it = m_resident_mappings.erase(it);
}
}
@@ -248,7 +245,6 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
// Cache the original values.
auto* const left = std::addressof(*it);
const size_t orig_size = left->size;
const s64 orig_host_map_count = this->HostMapCount(left->paddr, orig_size);
// Adjust the left map.
const size_t left_size = offset - left->vaddr;
@@ -270,20 +266,11 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
// If resident, also insert into resident map.
if (right->is_resident) {
m_resident_map_count += this->HostMapCount(left->paddr, left->size) +
this->HostMapCount(right->paddr, right->size) -
orig_host_map_count;
m_resident_map_count++;
m_resident_mappings.insert(*right);
}
}
s64 HeapTracker::HostMapCount(PAddr paddr, size_t size) const {
if (!m_has_hardware_buffer_backing) {
return size != 0 ? 1 : 0;
}
return static_cast<s64>(m_buffer.BackingMapCount(paddr, size));
}
HeapTracker::AddrTree::iterator HeapTracker::GetNearestHeapMapLocked(VAddr offset) {
const SeparateHeapMap key{
.vaddr = offset,
-6
View File
@@ -1,6 +1,3 @@
// 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
@@ -85,13 +82,10 @@ private:
AddrTree::iterator GetNearestHeapMapLocked(VAddr offset);
s64 HostMapCount(PAddr paddr, size_t size) const;
void RebuildSeparateHeapAddressSpace();
private:
Common::HostMemory& m_buffer;
const bool m_has_hardware_buffer_backing;
const s64 m_max_resident_map_count;
std::shared_mutex m_rebuild_lock{};
+6 -368
View File
@@ -51,45 +51,14 @@
#endif // ^^^ POSIX ^^^
#include <atomic>
#include <mutex>
#include <random>
#include <vector>
#include "common/alignment.h"
#include "common/assert.h"
#include "common/free_region_manager.h"
#include "common/host_memory.h"
#include "common/logging.h"
#include "common/memory_detect.h"
#include "common/settings.h"
#ifdef __ANDROID__
#include <dlfcn.h>
#include <android/hardware_buffer.h>
namespace {
struct NativeHandle {
int version;
int numFds;
int numInts;
int data[1];
};
using PFN_AHardwareBuffer_getNativeHandle = const NativeHandle* (*)(const AHardwareBuffer*);
PFN_AHardwareBuffer_getNativeHandle ResolveGetNativeHandle() {
void* const lib = dlopen("libnativewindow.so", RTLD_NOW);
if (lib == nullptr) {
return nullptr;
}
return reinterpret_cast<PFN_AHardwareBuffer_getNativeHandle>(
dlsym(lib, "AHardwareBuffer_getNativeHandle"));
}
} // namespace
#endif
#if defined(__ANDROID__) && __ANDROID_API__ < 30
#include <sys/syscall.h>
@@ -106,12 +75,6 @@ namespace Common {
[[maybe_unused]] constexpr size_t PageAlignment = 0x1000;
[[maybe_unused]] constexpr size_t HugePageSize = 0x200000;
static std::atomic<u64> committed_backing_size{};
u64 GetCommittedBackingSize() noexcept {
return committed_backing_size.load(std::memory_order_relaxed);
}
#ifdef _WIN32
// Manually imported for MinGW compatibility
@@ -160,7 +123,7 @@ static void GetFuncAddress(Common::DynamicLibrary& dll, const char* name, T& pfn
class HostMemory::Impl {
public:
explicit Impl(size_t backing_size_, size_t virtual_size_, size_t)
explicit Impl(size_t backing_size_, size_t virtual_size_)
: backing_size{backing_size_}
, virtual_size{virtual_size_}
, process{GetCurrentProcess()}
@@ -266,10 +229,6 @@ public:
UNREACHABLE();
}
bool IsBackingShared() const noexcept {
return true;
}
const size_t backing_size; ///< Size of the backing memory in bytes
const size_t virtual_size; ///< Size of the virtual address placeholder in bytes
@@ -542,10 +501,9 @@ static int shm_open_anon(int flags, mode_t mode) {
class HostMemory::Impl {
public:
explicit Impl(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_)
explicit Impl(size_t backing_size_, size_t virtual_size_)
: backing_size{backing_size_}
, virtual_size{virtual_size_}
, preferred_offset{preferred_offset_}
{}
bool Init() {
@@ -585,15 +543,10 @@ public:
LOG_WARNING(Common_Memory, "Using private mappings instead of shared ones");
backing_base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_READ | PROT_WRITE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0));
if (fd > 0) {
fd = -1;
close(fd);
}
fd = -1;
} else {
#ifdef __ANDROID__
if (InitAhbBacking()) {
return InitVirtual();
}
#endif
backing_base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0));
}
if (backing_base == MAP_FAILED) {
@@ -601,10 +554,7 @@ public:
return false;
}
return InitVirtual();
}
bool InitVirtual() {
// Virtual memory initialization
virtual_base = virtual_map_base = static_cast<u8*>(ChooseVirtualBase(virtual_size));
if (virtual_base == MAP_FAILED) {
LOG_CRITICAL(HW_Memory, "mmap failed: {}", strerror(errno));
@@ -617,242 +567,6 @@ public:
return true;
}
#ifdef __ANDROID__
static AHardwareBuffer_Desc MakeBlobDesc(size_t len) {
return AHardwareBuffer_Desc{
.width = static_cast<u32>(len),
.height = 1,
.layers = 1,
.format = AHARDWAREBUFFER_FORMAT_BLOB,
.usage = AHARDWAREBUFFER_USAGE_CPU_READ_OFTEN |
AHARDWAREBUFFER_USAGE_CPU_WRITE_OFTEN |
AHARDWAREBUFFER_USAGE_GPU_DATA_BUFFER,
.stride = 0,
.rfu0 = 0,
.rfu1 = 0,
};
}
static bool ProbeAhbBacking(PFN_AHardwareBuffer_getNativeHandle get_native_handle) {
const AHardwareBuffer_Desc desc = MakeBlobDesc(PageAlignment * 2);
AHardwareBuffer* buffer{};
if (AHardwareBuffer_allocate(&desc, &buffer) != 0 || buffer == nullptr) {
return false;
}
const NativeHandle* const handle = get_native_handle(buffer);
if (handle == nullptr || handle->numFds < 1) {
AHardwareBuffer_release(buffer);
return false;
}
const int probe_fd = handle->data[0];
bool ok = true;
const auto try_map = [&](int prot, off_t offset, const char* what) {
if (!ok) {
return;
}
void* const ptr = mmap(nullptr, PageAlignment, prot, MAP_SHARED, probe_fd, offset);
if (ptr == MAP_FAILED) {
ok = false;
return;
}
munmap(ptr, PageAlignment);
};
try_map(PROT_READ | PROT_WRITE, 0, "shared mappings");
try_map(PROT_READ | PROT_WRITE, static_cast<off_t>(PageAlignment), "mappings at an offset");
#ifdef ARCHITECTURE_arm64
try_map(PROT_READ | PROT_EXEC, 0, "executable mappings");
#endif
AHardwareBuffer_release(buffer);
return ok;
}
size_t ComputeAhbBudget(size_t window_size) const {
const u64 total_physical = Common::GetMemInfo().TotalPhysicalMemory;
constexpr u64 BaselineFootprint = 6ULL << 30;
if (total_physical <= BaselineFootprint) {
return 0;
}
const u64 max_map_count = Common::GetMaxMapCount();
constexpr u64 ReservedMaps = 24576;
if (max_map_count == 0 || max_map_count <= ReservedMaps) {
return 0;
}
u64 budget = (total_physical - BaselineFootprint) / 2;
constexpr u64 MapSlotsPerWindow = 4096;
const u64 affordable_windows = (max_map_count - ReservedMaps) / MapSlotsPerWindow;
budget = (std::min)(budget, affordable_windows * window_size);
const u64 available = Common::GetAvailablePhysicalMemory();
if (available != 0) {
budget = (std::min)(budget, available / 2);
}
budget = (std::min)(budget, static_cast<u64>(backing_size));
budget = Common::AlignDown(budget, window_size);
constexpr u64 MinimumBudget = 256ULL << 20;
if (budget < MinimumBudget) {
return 0;
}
return static_cast<size_t>(budget);
}
bool InitAhbBacking() {
if (!Settings::values.use_unified_memory.GetValue()) {
return false;
}
static const PFN_AHardwareBuffer_getNativeHandle get_native_handle =
ResolveGetNativeHandle();
if (get_native_handle == nullptr) {
return false;
}
constexpr size_t window_size = 256ULL << 20;
const size_t budget = ComputeAhbBudget(window_size);
if (budget == 0) {
return false;
}
if (!ProbeAhbBacking(get_native_handle)) {
return false;
}
const size_t aligned_backing = Common::AlignDown(backing_size, window_size);
const size_t max_windows = (std::min)(budget, aligned_backing) / window_size;
std::vector<AHardwareBuffer*> buffers;
std::vector<int> buffer_fds;
const auto cleanup = [&] {
for (AHardwareBuffer* buffer : buffers) {
AHardwareBuffer_release(buffer);
}
buffers.clear();
buffer_fds.clear();
};
for (size_t i = 0; i < max_windows; ++i) {
const AHardwareBuffer_Desc desc = MakeBlobDesc(window_size);
AHardwareBuffer* buffer{};
if (AHardwareBuffer_allocate(&desc, &buffer) != 0 || buffer == nullptr) {
break;
}
const NativeHandle* const handle = get_native_handle(buffer);
if (handle == nullptr || handle->numFds < 1) {
AHardwareBuffer_release(buffer);
break;
}
const int buffer_fd = handle->data[0];
const off_t buffer_len = lseek(buffer_fd, 0, SEEK_END);
if (buffer_len < static_cast<off_t>(window_size)) {
AHardwareBuffer_release(buffer);
break;
}
buffers.push_back(buffer);
buffer_fds.push_back(buffer_fd);
}
const size_t num_windows = buffers.size();
if (num_windows == 0) {
return false;
}
const size_t region_size = num_windows * window_size;
const size_t region_base = Common::AlignDown(
(std::min)(preferred_offset, aligned_backing - region_size), window_size);
u8* const base = static_cast<u8*>(mmap(nullptr, backing_size, PROT_NONE,
MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0));
if (base == MAP_FAILED) {
cleanup();
return false;
}
const auto map_over_reservation = [&](size_t offset, size_t len, int map_fd,
off_t map_offset) {
if (len == 0) {
return true;
}
if (mmap(base + offset, len, PROT_READ | PROT_WRITE, MAP_SHARED | MAP_FIXED, map_fd,
map_offset) == MAP_FAILED) {
munmap(base, backing_size);
cleanup();
return false;
}
return true;
};
if (!map_over_reservation(0, region_base, fd, 0)) {
return false;
}
for (size_t i = 0; i < num_windows; ++i) {
if (!map_over_reservation(region_base + i * window_size, window_size, buffer_fds[i],
0)) {
return false;
}
}
const size_t tail_offset = region_base + region_size;
if (!map_over_reservation(tail_offset, backing_size - tail_offset, fd,
static_cast<off_t>(tail_offset))) {
return false;
}
backing_base = base;
ahb_windows = std::move(buffers);
ahb_fds = std::move(buffer_fds);
ahb_window_size = window_size;
ahb_base = region_base;
ahb_bytes = region_size;
committed_backing_size.store(region_size, std::memory_order_relaxed);
return true;
}
void MapBackingRange(size_t virtual_offset, size_t host_offset, size_t length, int prot_flags) {
while (length > 0) {
int map_fd = fd;
off_t map_offset = static_cast<off_t>(host_offset);
size_t chunk = length;
if (host_offset < ahb_base) {
chunk = (std::min)(chunk, ahb_base - host_offset);
} else if (host_offset < ahb_base + ahb_bytes) {
const size_t relative = host_offset - ahb_base;
const size_t window = relative / ahb_window_size;
const size_t local = relative % ahb_window_size;
map_fd = ahb_fds[window];
map_offset = static_cast<off_t>(local);
chunk = (std::min)(chunk, ahb_window_size - local);
}
void* const ret = mmap(virtual_base + virtual_offset, chunk, prot_flags,
MAP_SHARED | MAP_FIXED, map_fd, map_offset);
ASSERT_MSG(ret != MAP_FAILED, "mmap: {}", strerror(errno));
virtual_offset += chunk;
host_offset += chunk;
length -= chunk;
}
}
size_t BackingMapCount(size_t host_offset, size_t length) const noexcept {
if (length == 0) {
return 0;
}
if (ahb_bytes == 0) {
return 1;
}
size_t count = 0;
while (length > 0) {
size_t chunk = length;
if (host_offset < ahb_base) {
chunk = (std::min)(chunk, ahb_base - host_offset);
} else if (host_offset < ahb_base + ahb_bytes) {
const size_t local = (host_offset - ahb_base) % ahb_window_size;
chunk = (std::min)(chunk, ahb_window_size - local);
}
host_offset += chunk;
length -= chunk;
++count;
}
return count;
}
std::span<AHardwareBuffer* const> AhbWindows() const noexcept {
return ahb_windows;
}
size_t AhbWindowSize() const noexcept {
return ahb_bytes != 0 ? ahb_window_size : 0;
}
size_t AhbBase() const noexcept {
return ahb_base;
}
#endif
~Impl() {
Release();
}
@@ -873,12 +587,6 @@ public:
#ifdef ARCHITECTURE_arm64
if (True(perms & MemoryPermission::Execute))
prot_flags |= PROT_EXEC;
#endif
#ifdef __ANDROID__
if (ahb_bytes != 0) {
MapBackingRange(virtual_offset, host_offset, length, prot_flags);
return;
}
#endif
int flags = (fd >= 0 ? MAP_SHARED : MAP_PRIVATE) | MAP_FIXED;
void* ret = mmap(virtual_base + virtual_offset, length, prot_flags, flags, fd, host_offset);
@@ -924,18 +632,8 @@ public:
virtual_base = nullptr;
}
bool IsBackingShared() const noexcept {
#ifdef __ANDROID__
if (ahb_bytes != 0) {
return true;
}
#endif
return fd >= 0;
}
const size_t backing_size; ///< Size of the backing memory in bytes
const size_t virtual_size; ///< Size of the virtual address placeholder in bytes
const size_t preferred_offset;
u8* backing_base{reinterpret_cast<u8*>(MAP_FAILED)};
u8* virtual_base{reinterpret_cast<u8*>(MAP_FAILED)};
@@ -958,18 +656,6 @@ private:
int ret = close(fd);
ASSERT_MSG(ret == 0, "close failed: {}", strerror(errno));
}
#ifdef __ANDROID__
for (AHardwareBuffer* buffer : ahb_windows) {
AHardwareBuffer_release(buffer);
}
ahb_windows.clear();
ahb_fds.clear();
if (ahb_bytes != 0) {
committed_backing_size.store(0, std::memory_order_relaxed);
ahb_bytes = 0;
}
#endif
}
void AdjustMap(size_t* virtual_offset, size_t* length) {
@@ -995,19 +681,11 @@ private:
int fd{-1}; // memfd file descriptor, -1 is the error value of memfd_create
FreeRegionManager free_manager{};
#ifdef __ANDROID__
std::vector<AHardwareBuffer*> ahb_windows;
std::vector<int> ahb_fds;
size_t ahb_window_size{};
size_t ahb_base{};
size_t ahb_bytes{};
#endif
};
#endif // ^^^ POSIX ^^^
HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_)
HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_)
: backing_size(backing_size_)
, virtual_size(virtual_size_)
{
@@ -1019,7 +697,7 @@ HostMemory::HostMemory(size_t backing_size_, size_t virtual_size_, size_t prefer
#else
// Try to allocate a fastmem arena.
// The implementation will fail with std::bad_alloc on errors.
impl = std::make_unique<HostMemory::Impl>(AlignUp(backing_size, PageAlignment), AlignUp(virtual_size, PageAlignment) + HugePageSize, preferred_offset_);
impl = std::make_unique<HostMemory::Impl>(AlignUp(backing_size, PageAlignment), AlignUp(virtual_size, PageAlignment) + HugePageSize);
if (impl->Init()) {
backing_base = impl->backing_base;
virtual_base = impl->virtual_base;
@@ -1089,46 +767,6 @@ void HostMemory::ClearBackingRegion(size_t physical_offset, size_t length, u32 f
std::memset(backing_base + physical_offset, fill_value, length);
}
std::span<AHardwareBuffer* const> HostMemory::BackingHardwareBuffers() const noexcept {
#ifdef __ANDROID__
return impl ? impl->AhbWindows() : std::span<AHardwareBuffer* const>{};
#else
return {};
#endif
}
size_t HostMemory::BackingMapCount(size_t host_offset, size_t length) const noexcept {
#ifdef __ANDROID__
return impl ? impl->BackingMapCount(host_offset, length) : (length != 0 ? 1 : 0);
#else
return length != 0 ? 1 : 0;
#endif
}
size_t HostMemory::BackingHardwareBufferWindowSize() const noexcept {
#ifdef __ANDROID__
return impl ? impl->AhbWindowSize() : 0;
#else
return 0;
#endif
}
bool HostMemory::IsBackingShared() const noexcept {
#if defined(__OPENORBIS__) || defined(__managarm__)
return false;
#else
return impl && impl->IsBackingShared();
#endif
}
size_t HostMemory::BackingHardwareBufferBase() const noexcept {
#ifdef __ANDROID__
return impl ? impl->AhbBase() : 0;
#else
return 0;
#endif
}
void HostMemory::EnableDirectMappedAddress() {
#if !(defined(__OPENORBIS__) || defined(__managarm__))
if (impl) {
+1 -20
View File
@@ -8,17 +8,12 @@
#include <memory>
#include <optional>
#include <span>
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "common/virtual_buffer.h"
struct AHardwareBuffer;
namespace Common {
[[nodiscard]] u64 GetCommittedBackingSize() noexcept;
enum class MemoryPermission : u32 {
Read = 1 << 0,
Write = 1 << 1,
@@ -33,7 +28,7 @@ DECLARE_ENUM_FLAG_OPERATORS(MemoryPermission)
*/
class HostMemory {
public:
explicit HostMemory(size_t backing_size_, size_t virtual_size_, size_t preferred_offset_ = 0);
explicit HostMemory(size_t backing_size_, size_t virtual_size_);
~HostMemory();
/**
@@ -67,20 +62,6 @@ public:
return backing_base;
}
[[nodiscard]] size_t BackingSize() const noexcept {
return backing_size;
}
[[nodiscard]] size_t BackingMapCount(size_t host_offset, size_t length) const noexcept;
[[nodiscard]] std::span<AHardwareBuffer* const> BackingHardwareBuffers() const noexcept;
[[nodiscard]] size_t BackingHardwareBufferWindowSize() const noexcept;
[[nodiscard]] size_t BackingHardwareBufferBase() const noexcept;
[[nodiscard]] bool IsBackingShared() const noexcept;
[[nodiscard]] u8* VirtualBasePointer() noexcept {
return virtual_base;
}
-58
View File
@@ -1,6 +1,3 @@
// 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
@@ -20,10 +17,6 @@
#endif
#endif
#include <cstdio>
#include <cstdlib>
#include <cstring>
#include "common/memory_detect.h"
namespace Common {
@@ -76,55 +69,4 @@ const MemoryInfo& GetMemInfo() {
return mem_info;
}
u64 GetAvailablePhysicalMemory() {
#ifdef _WIN32
MEMORYSTATUSEX memorystatus;
memorystatus.dwLength = sizeof(memorystatus);
if (GlobalMemoryStatusEx(&memorystatus)) {
return memorystatus.ullAvailPhys;
}
return 0;
#elif defined(__linux__)
if (std::FILE* const file = std::fopen("/proc/meminfo", "re")) {
char line[256];
u64 available = 0;
while (std::fgets(line, sizeof(line), file) != nullptr) {
if (std::strncmp(line, "MemAvailable:", 13) == 0) {
available = std::strtoull(line + 13, nullptr, 10) * 1024ULL;
break;
}
}
std::fclose(file);
if (available != 0) {
return available;
}
}
struct sysinfo info;
if (sysinfo(&info) == 0) {
const u64 unit = info.mem_unit != 0 ? info.mem_unit : 1ULL;
return (static_cast<u64>(info.freeram) + static_cast<u64>(info.bufferram)) * unit;
}
return 0;
#else
return 0;
#endif
}
u64 GetMaxMapCount() {
#ifdef __linux__
if (std::FILE* const file = std::fopen("/proc/sys/vm/max_map_count", "re")) {
char line[32];
u64 count = 0;
if (std::fgets(line, sizeof(line), file) != nullptr) {
count = std::strtoull(line, nullptr, 10);
}
std::fclose(file);
return count;
}
return 0;
#else
return 0;
#endif
}
} // namespace Common
-7
View File
@@ -1,6 +1,3 @@
// 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
@@ -21,8 +18,4 @@ struct MemoryInfo {
*/
[[nodiscard]] const MemoryInfo& GetMemInfo();
[[nodiscard]] u64 GetAvailablePhysicalMemory();
[[nodiscard]] u64 GetMaxMapCount();
} // namespace Common
+9 -4
View File
@@ -587,9 +587,6 @@ struct Values {
SwitchableSetting<bool> use_asynchronous_shaders{linkage, false, "use_asynchronous_shaders",
Category::RendererHacks};
SwitchableSetting<bool> use_unified_memory{linkage, false, "use_unified_memory",
Category::RendererHacks};
SwitchableSetting<GpuUnswizzleSize> gpu_unswizzle_texture_size{linkage,
GpuUnswizzleSize::Large,
"gpu_unswizzle_texture_size",
@@ -705,7 +702,15 @@ struct Values {
// Controls
InputSetting<std::array<PlayerInput, 10>> players;
Setting<bool> disable_wgi_xinput{
linkage, false, "disable_wgi_xinput", Category::Controls, Specialization::Default,
// Only read/write disable_wgi_xinput on Windows platforms
#ifdef _WIN32
true
#else
false
#endif
};
Setting<bool> enable_raw_input{
linkage, false, "enable_raw_input", Category::Controls, Specialization::Default,
// Only read/write enable_raw_input on Windows platforms
+1 -1
View File
@@ -1199,7 +1199,7 @@ if (MSVC)
)
else()
target_compile_options(core PRIVATE
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type>
$<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>)
+1 -5
View File
@@ -119,7 +119,6 @@ struct System::Impl {
is_multicore = Settings::values.use_multi_core.GetValue();
extended_memory_layout = Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_4Gb;
unified_memory = Settings::values.use_unified_memory.GetValue();
core_timing.SetMulticore(is_multicore);
core_timing.Initialize([&system]() { system.RegisterHostThread(); });
@@ -147,8 +146,7 @@ struct System::Impl {
!device_memory.has_value() ||
is_multicore != Settings::values.use_multi_core.GetValue() ||
extended_memory_layout != (Settings::values.memory_layout_mode.GetValue() !=
Settings::MemoryLayout::Memory_4Gb) ||
unified_memory != Settings::values.use_unified_memory.GetValue();
Settings::MemoryLayout::Memory_4Gb);
if (!must_reinitialize) {
return;
@@ -159,7 +157,6 @@ struct System::Impl {
is_multicore = Settings::values.use_multi_core.GetValue();
extended_memory_layout =
Settings::values.memory_layout_mode.GetValue() != Settings::MemoryLayout::Memory_4Gb;
unified_memory = Settings::values.use_unified_memory.GetValue();
Initialize(system);
}
@@ -506,7 +503,6 @@ struct System::Impl {
std::atomic_bool is_powered_on{};
bool is_multicore : 1 = false;
bool extended_memory_layout : 1 = false;
bool unified_memory : 1 = false;
bool exit_locked : 1 = false;
bool exit_requested : 1 = false;
bool nvdec_active : 1 = false;
+1 -13
View File
@@ -1,6 +1,3 @@
// 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
@@ -15,18 +12,9 @@ constexpr size_t VirtualReserveSize = 1ULL << 38;
constexpr size_t VirtualReserveSize = 1ULL << 39;
#endif
namespace {
size_t ApplicationPoolOffset() {
using Init = Kernel::Board::Nintendo::Nx::KSystemControl::Init;
const size_t dram_size = Init::GetIntendedMemorySize();
const size_t application_pool_size = Init::GetApplicationPoolSize();
return dram_size > application_pool_size ? dram_size - application_pool_size : 0;
}
}
DeviceMemory::DeviceMemory()
: buffer{Kernel::Board::Nintendo::Nx::KSystemControl::Init::GetIntendedMemorySize(),
VirtualReserveSize, ApplicationPoolOffset()} {}
VirtualReserveSize} {}
DeviceMemory::~DeviceMemory() = default;
-35
View File
@@ -20,8 +20,6 @@
#include "common/scratch_buffer.h"
#include "common/virtual_buffer.h"
struct AHardwareBuffer;
namespace Core {
constexpr size_t DEVICE_PAGEBITS = 12ULL;
@@ -97,34 +95,6 @@ public:
ApplyOpOnPAddr(address, buffer, operation);
}
u8* GetPhysicalBase() noexcept {
return reinterpret_cast<u8*>(physical_base);
}
const u8* GetPhysicalBase() const noexcept {
return reinterpret_cast<const u8*>(physical_base);
}
size_t GetPhysicalSize() const noexcept {
return physical_size;
}
std::span<AHardwareBuffer* const> GetBackingHardwareBuffers() const noexcept {
return ahb_windows;
}
size_t GetBackingHardwareBufferWindowSize() const noexcept {
return ahb_window_size;
}
size_t GetBackingHardwareBufferBase() const noexcept {
return ahb_base;
}
bool IsBackingShared() const noexcept {
return backing_is_shared;
}
PAddr GetPhysicalRawAddressFromDAddr(DAddr address) const {
PAddr subbits = PAddr(address & page_mask);
auto paddr = tracked_entries[(address >> page_bits)].compressed_physical_ptr;
@@ -201,11 +171,6 @@ private:
std::unique_ptr<DeviceMemoryManagerAllocator<Traits>> impl;
const uintptr_t physical_base;
const size_t physical_size;
const std::span<AHardwareBuffer* const> ahb_windows;
const size_t ahb_window_size;
const size_t ahb_base;
const bool backing_is_shared;
DeviceInterface* device_inter;
struct TrackedEntry {
-5
View File
@@ -171,11 +171,6 @@ struct DeviceMemoryManagerAllocator {
template <typename Traits>
DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memory_)
: physical_base{uintptr_t(device_memory_.buffer.BackingBasePointer())}
, physical_size{device_memory_.buffer.BackingSize()}
, ahb_windows{device_memory_.buffer.BackingHardwareBuffers()}
, ahb_window_size{device_memory_.buffer.BackingHardwareBufferWindowSize()}
, ahb_base{device_memory_.buffer.BackingHardwareBufferBase()}
, backing_is_shared{device_memory_.buffer.IsBackingShared()}
, device_inter{nullptr}
, compressed_device_addr(1ULL << ((Settings::values.memory_layout_mode.GetValue() == Settings::MemoryLayout::Memory_4Gb ? physical_min_bits : physical_max_bits) - Memory::YUZU_PAGEBITS))
, tracked_entries(device_as_size >> Memory::YUZU_PAGEBITS)
@@ -28,7 +28,7 @@ Result CloseHandle(Core::System& system, Handle handle) {
/// Clears the signaled state of an event or process.
Result ResetSignal(Core::System& system, Handle handle) {
LOG_DEBUG(Kernel_SVC, "called handle {:#08x}", handle);
LOG_TRACE(Kernel_SVC, "called handle {:#08x}", handle);
// Get the current handle table.
const auto& handle_table = GetCurrentProcess(system.Kernel()).GetHandleTable();
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator
@@ -84,7 +84,7 @@ private:
RestrictionSettings restriction_settings{};
std::array<char, 8> pin_code{};
Capability capability{};
// TODO: this is RAW as fuck
// TODO: this is raw
PlayTimerSettings raw_play_timer_settings{};
KernelHelpers::ServiceContext service_context;
@@ -1020,7 +1020,7 @@ Result ISystemSettingsServer::GetBatteryLot(Out<BatteryLot> out_battery_lot) {
c.lot_number[1] = 'H';
c.lot_number[2] = 'A';
c.lot_number[3] = 'C';
// TODO: I have no fucking idea what the letters mean
// TODO: what do the letters mean?
c.lot_number[4] = 'H';
c.lot_number[5] = 'Z';
c.lot_number[6] = 'Z';
+2 -1
View File
@@ -531,7 +531,8 @@ int TranslateTypeToNative(Type type) {
NETWORK_PROTOCOL_TRANSLATE_ELEM(MPLS) \
NETWORK_PROTOCOL_TRANSLATE_ELEM(PFSYNC)
#elif defined(__linux__)
// Other platforms get fucked
// Other platforms may not support some niche protocols.
// This is usually not an issue
#define NETWORK_PROTOCOL_TRANSLATE_LIST \
NETWORK_PROTOCOL_TRANSLATE_ELEM(IP) \
/*NETWORK_PROTOCOL_TRANSLATE_ELEM(HOPOPTS)*/ \
@@ -234,7 +234,7 @@ private:
BlockOfCode block_of_code;
A32EmitX64 emitter;
Optimization::PolyfillOptions polyfill_options;
// Keep it here, you don't wanna mess with the fuckery that's initializer lists
// Keep it here in order to not mess with initializer lists
const A32::UserConfig conf;
Jit* jit_interface;
@@ -2160,8 +2160,6 @@ void EmitFPVectorToFixed(BlockOfCode& code, EmitContext& ctx, IR::Inst* inst) {
ROUNDING_MODE_CASE(CASE, 0x3e) \
ROUNDING_MODE_CASE(CASE, 0x3f)
// FUCK YOU MSVC, FUCKING DEPTH CANT EVEN HANDLE 8+16+32+64 DEPTH OF A ELSE STATMENT YOU FUCKING STUPID
// BURN MSVC BURN IT STUPID COMPILER CAN'T EVEN COMPILE THE MOST BASIC C++
ROUNDING_MODE_SWITCH(ToNearest_TieEven)
ROUNDING_MODE_SWITCH(TowardsPlusInfinity)
ROUNDING_MODE_SWITCH(TowardsMinusInfinity)
+1 -3
View File
@@ -67,9 +67,7 @@
#else
# define CTX_SP (_UC_MACHINE_SP(ucontext))
#endif
#endif
#if defined(ARCHITECTURE_arm64)
#elif defined(ARCHITECTURE_arm64)
# if defined(__APPLE__)
# define CTX_PC (mctx->__ss.__pc)
# define CTX_SP (mctx->__ss.__sp)
+1 -1
View File
@@ -787,7 +787,7 @@ static void FoldCountLeadingZeros(IR::Inst& inst, bool is_32_bit) {
/// Folds division operations based on the following:
///
/// 1. x / 0 -> 0 (NOTE: This is an ARM-specific behavior defined in the architecture reference manual)
/// 2a. 0x8000_0000 / 0xFFFF_FFFF -> 0x8000_0000 (NOTE: More ARM bullshit)
/// 2a. 0x8000_0000 / 0xFFFF_FFFF -> 0x8000_0000 (NOTE: More ARM errata)
/// 2b. 0x8000_0000_0000_0000 / 0xFFFF_FFFF_FFFF_FFFF -> 0x8000_0000_0000_0000
/// 3. imm_x / imm_y -> result
/// 4. x / 1 -> x
+1 -1
View File
@@ -151,7 +151,7 @@ if (MSVC)
)
else()
target_compile_options(hid_core PRIVATE
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-cast-function-type>
$<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>)
+1 -1
View File
@@ -44,7 +44,7 @@ if (MSVC)
/we4800 # Implicit conversion from 'type' to bool. Possible information loss
)
else()
target_compile_options(input_common PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>)
target_compile_options(input_common PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>)
endif()
if (ANDROID)
+5 -2
View File
@@ -1,3 +1,6 @@
// 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
@@ -157,8 +160,8 @@ Common::ParamPackage Android::BuildButtonParamPackageForButton(PadIdentifier ide
bool Android::MatchVID(Common::UUID device, const std::vector<std::string>& vids) const {
for (size_t i = 0; i < vids.size(); ++i) {
auto fucker = device.RawString();
if (fucker.find(vids[i]) != std::string::npos) {
auto dev_str = device.RawString();
if (dev_str.find(vids[i]) != std::string::npos) {
return true;
}
}
+7
View File
@@ -648,6 +648,13 @@ SDLDriver::SDLDriver(std::string input_engine_) : InputEngine(std::move(input_en
// Disable raw input. When enabled this setting causes SDL to die when a web applet opens
SDL_SetHint(SDL_HINT_JOYSTICK_RAWINPUT, Settings::values.enable_raw_input ? "1" : "0");
#ifdef _WIN32
if (Settings::values.disable_wgi_xinput) {
SDL_SetHintWithPriority(SDL_HINT_JOYSTICK_RAWINPUT_CORRELATE_XINPUT, "0", SDL_HINT_OVERRIDE);
SDL_SetHintWithPriority(SDL_HINT_JOYSTICK_WGI, "0", SDL_HINT_OVERRIDE);
}
#endif
// SDL3 defaults Steam Controller Bluetooth HIDAPI support to off, which can disable gyro.
SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_STEAM, "1");
SDL_SetHint(SDL_HINT_GAMECONTROLLER_SENSOR_FUSION, "1");
+10
View File
@@ -62,6 +62,16 @@ if (USE_DISCORD_PRESENCE)
target_compile_definitions(qt_common PUBLIC USE_DISCORD_PRESENCE)
endif()
# Qt itself may use Vulkan stuffs so we'd need to account for that as well
# must match src/video_core/vulkan_common/vulkan.h logic
if (WIN32 OR APPLE OR ANDROID OR PLATFORM_HAIKU OR PLATFORM_PS4)
# ...
else()
find_package(X11 REQUIRED)
target_include_directories(video_core PUBLIC ${X11_INCLUDE_DIR})
target_link_libraries(video_core PUBLIC ${X11_LIBRARIES})
endif()
# TODO(crueter)
target_link_libraries(qt_common PRIVATE Qt6::Widgets)
@@ -227,8 +227,6 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
tr("Preserves GPU-modified data by reading it back before uploading.\nSome games require this to render certain effects properly."));
INSERT(Settings, use_asynchronous_shaders, tr("Enable asynchronous shader compilation"),
tr("May reduce shader stutter."));
INSERT(Settings, use_unified_memory, tr("Enable unified memory access"),
tr("Lets the GPU write buffer readbacks directly into guest memory."));
INSERT(Settings, gpu_clock, tr("GPU Clocks"),
tr("Makes the game believe GPU work finishes faster than it does, so it stops lowering "
"resolution and render distance to fit the Switch's clocks."));
+1 -1
View File
@@ -253,7 +253,7 @@ if (MSVC)
)
else()
target_compile_options(shader_recompiler PRIVATE
$<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>
# Bracket depth determines maximum size of a fold expression in Clang since 9c9974c3ccb6.
# And this in turns limits the size of a std::array.
$<$<CXX_COMPILER_ID:Clang>:-fbracket-depth=1024>
@@ -13,7 +13,6 @@
namespace Shader::Backend::SPIRV {
namespace {
Id SharedPointer(EmitContext& ctx, Id offset, u32 index_offset = 0) {
offset = ctx.BoundSharedOffset(offset, 4 + index_offset * 4);
const Id shift_id{ctx.Const(2U)};
Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
if (index_offset > 0) {
@@ -161,8 +160,7 @@ Id EmitSharedAtomicExchange32(EmitContext& ctx, Id offset, Id value) {
Id EmitSharedAtomicExchange64(EmitContext& ctx, Id offset, Id value) {
if (ctx.profile.support_shared_int64_atomics && ctx.uses_explicit_workgroup_layout) {
const Id shift_id{ctx.Const(3U)};
const Id index{
ctx.OpShiftRightArithmetic(ctx.U32[1], ctx.BoundSharedOffset(offset, 8), shift_id)};
const Id index{ctx.OpShiftRightArithmetic(ctx.U32[1], offset, shift_id)};
const Id pointer{
ctx.OpAccessChain(ctx.shared_u64, ctx.shared_memory_u64, ctx.u32_zero_value, index)};
const auto [scope, semantics]{AtomicArgs(ctx)};
@@ -31,7 +31,6 @@ std::pair<Id, Id> ExtractArgs(EmitContext& ctx, Id offset, u32 mask, u32 count)
} // Anonymous namespace
Id EmitLoadSharedU8(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 1);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{
ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -43,7 +42,6 @@ Id EmitLoadSharedU8(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedS8(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 1);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{
ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -55,7 +53,6 @@ Id EmitLoadSharedS8(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedU16(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 2);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
return ctx.OpUConvert(ctx.U32[1], ctx.OpLoad(ctx.U16, pointer));
@@ -66,7 +63,6 @@ Id EmitLoadSharedU16(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedS16(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 2);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
return ctx.OpSConvert(ctx.U32[1], ctx.OpLoad(ctx.U16, pointer));
@@ -77,7 +73,6 @@ Id EmitLoadSharedS16(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedU32(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 4);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u32, ctx.shared_memory_u32, offset, 2)};
return ctx.OpLoad(ctx.U32[1], pointer);
@@ -87,7 +82,6 @@ Id EmitLoadSharedU32(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedU64(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 8);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u32x2, ctx.shared_memory_u32x2, offset, 3)};
return ctx.OpLoad(ctx.U32[2], pointer);
@@ -103,7 +97,6 @@ Id EmitLoadSharedU64(EmitContext& ctx, Id offset) {
}
Id EmitLoadSharedU128(EmitContext& ctx, Id offset) {
offset = ctx.BoundSharedOffset(offset, 16);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u32x4, ctx.shared_memory_u32x4, offset, 4)};
return ctx.OpLoad(ctx.U32[4], pointer);
@@ -120,7 +113,6 @@ Id EmitLoadSharedU128(EmitContext& ctx, Id offset) {
}
void EmitWriteSharedU8(EmitContext& ctx, Id offset, Id value) {
offset = ctx.BoundSharedOffset(offset, 1);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{
ctx.OpAccessChain(ctx.shared_u8, ctx.shared_memory_u8, ctx.u32_zero_value, offset)};
@@ -131,7 +123,6 @@ void EmitWriteSharedU8(EmitContext& ctx, Id offset, Id value) {
}
void EmitWriteSharedU16(EmitContext& ctx, Id offset, Id value) {
offset = ctx.BoundSharedOffset(offset, 2);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u16, ctx.shared_memory_u16, offset, 1)};
ctx.OpStore(pointer, ctx.OpUConvert(ctx.U16, value));
@@ -141,7 +132,6 @@ void EmitWriteSharedU16(EmitContext& ctx, Id offset, Id value) {
}
void EmitWriteSharedU32(EmitContext& ctx, Id offset, Id value) {
offset = ctx.BoundSharedOffset(offset, 4);
Id pointer{};
if (ctx.uses_explicit_workgroup_layout) {
pointer = Pointer(ctx, ctx.shared_u32, ctx.shared_memory_u32, offset, 2);
@@ -154,7 +144,6 @@ void EmitWriteSharedU32(EmitContext& ctx, Id offset, Id value) {
}
void EmitWriteSharedU64(EmitContext& ctx, Id offset, Id value) {
offset = ctx.BoundSharedOffset(offset, 8);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u32x2, ctx.shared_memory_u32x2, offset, 3)};
ctx.OpStore(pointer, value);
@@ -170,7 +159,6 @@ void EmitWriteSharedU64(EmitContext& ctx, Id offset, Id value) {
}
void EmitWriteSharedU128(EmitContext& ctx, Id offset, Id value) {
offset = ctx.BoundSharedOffset(offset, 16);
if (ctx.uses_explicit_workgroup_layout) {
const Id pointer{Pointer(ctx, ctx.shared_u32x4, ctx.shared_memory_u32x4, offset, 4)};
ctx.OpStore(pointer, value);
@@ -600,16 +600,6 @@ void EmitContext::DefineLocalMemory(const IR::Program& program) {
}
}
Id EmitContext::BoundSharedOffset(Id offset, u32 access_bytes) {
if (shared_memory_declared_bytes == 0) {
return offset;
}
const u32 last_valid{shared_memory_declared_bytes > access_bytes
? shared_memory_declared_bytes - access_bytes
: 0U};
return OpUMin(U32[1], offset, Const(last_valid));
}
void EmitContext::DefineSharedMemory(const IR::Program& program) {
uses_explicit_workgroup_layout =
profile.support_explicit_workgroup_layout &&
@@ -618,15 +608,8 @@ void EmitContext::DefineSharedMemory(const IR::Program& program) {
if (program.shared_memory_size == 0) {
return;
}
const u32 device_limit{profile.max_shared_memory_size};
const u32 shared_memory_size{device_limit != 0 && program.shared_memory_size > device_limit
? device_limit
: program.shared_memory_size};
if (shared_memory_size != program.shared_memory_size) {
shared_memory_declared_bytes = shared_memory_size;
}
const auto make{[&](Id element_type, u32 element_size) {
const u32 num_elements{Common::DivCeil(shared_memory_size, element_size)};
const u32 num_elements{Common::DivCeil(program.shared_memory_size, element_size)};
const Id array_type{TypeArray(element_type, Const(num_elements))};
Decorate(array_type, spv::Decoration::ArrayStride, element_size);
@@ -661,7 +644,7 @@ void EmitContext::DefineSharedMemory(const IR::Program& program) {
std::tie(shared_memory_u32x4, shared_u32x4, std::ignore) = make(U32[4], 16);
return;
}
const u32 num_elements{Common::DivCeil(shared_memory_size, 4U)};
const u32 num_elements{Common::DivCeil(program.shared_memory_size, 4U)};
const Id type{TypeArray(U32[1], Const(num_elements))};
shared_memory_u32_type = TypePointer(spv::StorageClass::Workgroup, type);
@@ -312,8 +312,6 @@ public:
Id local_memory{};
bool uses_explicit_workgroup_layout{};
u32 shared_memory_declared_bytes{};
[[nodiscard]] Id BoundSharedOffset(Id offset, u32 access_bytes);
Id shared_memory_u8{};
Id shared_memory_u16{};
Id shared_memory_u32{};
-3
View File
@@ -101,9 +101,6 @@ struct Profile {
u32 gl_max_compute_smem_size{};
/// Largest workgroup shared memory allocation the device accepts, 0 when unconstrained
u32 max_shared_memory_size{};
/// Maxwell and earlier nVidia architectures have broken robust support
bool has_broken_robust{};
+10 -1
View File
@@ -326,6 +326,15 @@ if (ENABLE_OPENGL)
)
endif()
# Must match src/video_core/vulkan_common/vulkan.h X11 logic
if (WIN32 OR APPLE OR ANDROID OR PLATFORM_HAIKU OR PLATFORM_PS4)
# ...
else()
find_package(X11 REQUIRED)
target_include_directories(video_core PUBLIC ${X11_INCLUDE_DIR})
target_link_libraries(video_core PUBLIC ${X11_LIBRARIES})
endif()
target_link_libraries(video_core PUBLIC common core)
target_link_libraries(video_core PUBLIC shader_recompiler stb bc_decoder gpu_logging)
if (ENABLE_OPENGL)
@@ -373,7 +382,7 @@ else()
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-shadow>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-unused-local-typedef>)
else()
target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Werror=conversion>)
target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wconversion>)
endif()
target_compile_options(video_core PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-conversion>)
+35 -213
View File
@@ -571,11 +571,7 @@ void BufferCache<P>::AccumulateFlushes() {
template <class P>
bool BufferCache<P>::ShouldWaitAsyncFlushes() const noexcept {
if (async_buffers.empty()) {
return false;
}
return async_buffers.front().has_value() ||
!pending_downloads.front().unified_copies.empty();
return (!async_buffers.empty() && async_buffers.front().has_value());
}
template <class P>
@@ -583,7 +579,6 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
AccumulateFlushes();
if (committed_gpu_modified_ranges.empty()) {
pending_downloads.emplace_back();
async_buffers.emplace_back(std::optional<Async_Buffer>{});
return;
}
@@ -643,84 +638,27 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
}
committed_gpu_modified_ranges.clear();
if (downloads.empty()) {
pending_downloads.emplace_back();
async_buffers.emplace_back(std::optional<Async_Buffer>{});
return;
}
struct QueuedUnifiedCopy {
u64 window;
BufferId buffer_id;
boost::container::small_vector<BufferCopy, 16> copies;
};
AsyncDownloadBatch batch;
boost::container::small_vector<std::pair<BufferCopy, BufferId>, 16> staging_downloads;
boost::container::small_vector<QueuedUnifiedCopy, 4> unified_copy_queue;
boost::container::small_vector<u64, 4> window_ids;
UnifiedWindowGroups groups;
u64 staging_size_bytes = 0;
for (auto& [copy, buffer_id] : downloads) {
Buffer& buffer = slot_buffers[buffer_id];
const DAddr orig_device_addr = buffer.CpuAddr() + copy.src_offset;
bool unified = false;
if constexpr (USE_UNIFIED_MEMORY) {
if (runtime.HasUnifiedMemory()) {
window_ids.clear();
groups.clear();
unified = ResolveUnifiedWindows(orig_device_addr, copy.src_offset, copy.size,
window_ids, groups);
}
}
BufferCopy record{copy};
record.src_offset = static_cast<size_t>(orig_device_addr);
if (unified) {
async_downloads.Add(orig_device_addr, copy.size);
buffer.MarkUsage(copy.src_offset, copy.size);
for (size_t i = 0; i < window_ids.size(); ++i) {
unified_copy_queue.push_back(
QueuedUnifiedCopy{window_ids[i], buffer_id, std::move(groups[i])});
}
batch.unified_copies.push_back(record);
continue;
}
copy.dst_offset = staging_size_bytes;
constexpr u64 align = 64ULL;
staging_size_bytes += (copy.size + align - 1) & ~(align - 1ULL);
staging_downloads.push_back({copy, buffer_id});
}
std::optional<Async_Buffer> download_staging;
if (!staging_downloads.empty()) {
download_staging = runtime.DownloadStagingBuffer(staging_size_bytes, true);
}
auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes, true);
boost::container::small_vector<BufferCopy, 4> normalized_copies;
runtime.PreCopyBarrier();
for (auto& [copy, buffer_id] : staging_downloads) {
copy.dst_offset += download_staging->offset;
for (auto& [copy, buffer_id] : downloads) {
copy.dst_offset += download_staging.offset;
const std::array copies{copy};
BufferCopy second_copy{copy};
Buffer& buffer = slot_buffers[buffer_id];
BufferCopy record{copy};
record.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
const DAddr orig_device_addr = static_cast<DAddr>(record.src_offset);
second_copy.src_offset = static_cast<size_t>(buffer.CpuAddr()) + copy.src_offset;
const DAddr orig_device_addr = static_cast<DAddr>(second_copy.src_offset);
async_downloads.Add(orig_device_addr, copy.size);
buffer.MarkUsage(copy.src_offset, copy.size);
runtime.CopyBuffer(download_staging->buffer, buffer, copies, false);
batch.staging_copies.push_back(record);
}
if constexpr (USE_UNIFIED_MEMORY) {
for (const auto& queued : unified_copy_queue) {
const std::span<const BufferCopy> group_span(queued.copies.data(),
queued.copies.size());
runtime.CopyToUnifiedMemory(queued.window, slot_buffers[queued.buffer_id], group_span);
}
if (!unified_copy_queue.empty()) {
runtime.FlushUnifiedMemoryCopies();
runtime.UnifiedMemoryHostBarrier();
}
runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
normalized_copies.push_back(second_copy);
}
runtime.PostCopyBarrier();
pending_downloads.emplace_back(std::move(batch));
async_buffers.emplace_back(std::move(download_staging));
pending_downloads.emplace_back(std::move(normalized_copies));
async_buffers.emplace_back(download_staging);
}
template <class P>
@@ -735,49 +673,32 @@ void BufferCache<P>::PopAsyncFlushes() {
template <class P>
void BufferCache<P>::PopAsyncBuffers() {
struct Writeback {
DAddr addr;
const u8* src;
u64 size;
};
boost::container::small_vector<Writeback, 8> writebacks;
{
std::scoped_lock lock{mutex};
if (async_buffers.empty()) {
return;
}
auto& batch = pending_downloads.front();
auto& async_buffer = async_buffers.front();
if (async_buffer.has_value()) {
const u8* base = async_buffer->mapped_span.data();
const size_t base_offset = async_buffer->offset;
for (const auto& copy : batch.staging_copies) {
const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
const u64 dst_offset = copy.dst_offset - base_offset;
const u8* read_mapped_memory = base + dst_offset;
async_downloads.ForEachInRange(
device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
writebacks.push_back(
{start, &read_mapped_memory[start - device_addr], end - start});
});
async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
gpu_modified_ranges.Subtract(start, end - start);
});
}
async_buffers_death_ring.emplace_back(*async_buffer);
}
for (const auto& copy : batch.unified_copies) {
const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
gpu_modified_ranges.Subtract(start, end - start);
});
}
if (async_buffers.empty()) {
return;
}
if (!async_buffers.front().has_value()) {
async_buffers.pop_front();
pending_downloads.pop_front();
return;
}
for (const auto& wb : writebacks) {
device_memory.WriteBlockUnsafe(wb.addr, wb.src, wb.size);
auto& downloads = pending_downloads.front();
auto& async_buffer = async_buffers.front();
u8* base = async_buffer->mapped_span.data();
const size_t base_offset = async_buffer->offset;
for (const auto& copy : downloads) {
const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
const u64 dst_offset = copy.dst_offset - base_offset;
const u8* read_mapped_memory = base + dst_offset;
async_downloads.ForEachInRange(device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
device_memory.WriteBlockUnsafe(start, &read_mapped_memory[start - device_addr],
end - start);
});
async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
gpu_modified_ranges.Subtract(start, end - start);
});
}
async_buffers_death_ring.emplace_back(*async_buffer);
async_buffers.pop_front();
pending_downloads.pop_front();
}
template <class P>
@@ -1778,99 +1699,6 @@ void BufferCache<P>::ImmediateUploadMemory([[maybe_unused]] Buffer& buffer,
}
}
template <class P>
bool BufferCache<P>::ResolveUnifiedWindows(
[[maybe_unused]] DAddr device_addr, [[maybe_unused]] u64 buffer_offset,
[[maybe_unused]] u64 size, [[maybe_unused]] boost::container::small_vector<u64, 4>& window_ids,
[[maybe_unused]] UnifiedWindowGroups& groups) {
if constexpr (USE_UNIFIED_MEMORY) {
const u8* const physical_base = device_memory.GetPhysicalBase();
const u64 unified_base = runtime.UnifiedMemoryBase();
const u64 unified_size = runtime.UnifiedMemorySize();
const u64 window_size = runtime.UnifiedMemoryWindowSize();
if (window_size == 0) {
return false;
}
const auto group_for = [&](u64 window) -> boost::container::small_vector<BufferCopy, 16>& {
for (size_t i = 0; i < window_ids.size(); ++i) {
if (window_ids[i] == window) {
return groups[i];
}
}
window_ids.push_back(window);
groups.emplace_back();
return groups.back();
};
u64 downloaded = 0;
while (downloaded < size) {
const DAddr page_addr = device_addr + downloaded;
const u8* const ptr = device_memory.GetPointer<u8>(page_addr);
if (ptr == nullptr) {
return false;
}
const u64 page_offset = page_addr & Core::DEVICE_PAGEMASK;
u64 chunk = (std::min)(size - downloaded,
static_cast<u64>(Core::DEVICE_PAGESIZE) - page_offset);
const u64 phys_offset = static_cast<u64>(ptr - physical_base);
if (phys_offset < unified_base || phys_offset - unified_base + chunk > unified_size) {
return false;
}
const u64 relative = phys_offset - unified_base;
const u64 window = relative / window_size;
const u64 local_offset = relative % window_size;
chunk = (std::min)(chunk, window_size - local_offset);
auto& group = group_for(window);
if (!group.empty()) {
BufferCopy& last = group.back();
if (last.src_offset + last.size == buffer_offset + downloaded &&
last.dst_offset + last.size == local_offset) {
last.size += chunk;
downloaded += chunk;
continue;
}
}
group.push_back(BufferCopy{
.src_offset = buffer_offset + downloaded,
.dst_offset = local_offset,
.size = chunk,
});
downloaded += chunk;
}
return true;
} else {
return false;
}
}
template <class P>
bool BufferCache<P>::TryUnifiedDownloadMemory([[maybe_unused]] Buffer& buffer,
[[maybe_unused]] std::span<BufferCopy> copies) {
if constexpr (USE_UNIFIED_MEMORY) {
boost::container::small_vector<u64, 4> window_ids;
UnifiedWindowGroups groups;
for (const BufferCopy& copy : copies) {
if (!ResolveUnifiedWindows(buffer.CpuAddr() + copy.src_offset, copy.src_offset,
copy.size, window_ids, groups)) {
return false;
}
}
for (const BufferCopy& copy : copies) {
buffer.MarkUsage(copy.src_offset, copy.size);
}
runtime.PreCopyBarrier();
for (size_t i = 0; i < window_ids.size(); ++i) {
const std::span<const BufferCopy> group_span(groups[i].data(), groups[i].size());
runtime.CopyToUnifiedMemory(window_ids[i], buffer, group_span);
}
runtime.FlushUnifiedMemoryCopies();
runtime.UnifiedMemoryHostBarrier();
runtime.Finish();
return true;
} else {
return false;
}
}
template <class P>
void BufferCache<P>::MappedUploadMemory([[maybe_unused]] Buffer& buffer,
[[maybe_unused]] u64 total_size_bytes,
@@ -1974,12 +1802,6 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, DAddr device_addr, u64
}
if constexpr (USE_MEMORY_MAPS) {
if constexpr (USE_UNIFIED_MEMORY) {
if (runtime.HasUnifiedMemory() &&
TryUnifiedDownloadMemory(buffer, std::span(copies.data(), copies.size()))) {
return;
}
}
auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
const u8* const mapped_memory = download_staging.mapped_span.data();
const std::span<BufferCopy> copies_span(copies.data(), copies.data() + copies.size());
@@ -180,7 +180,6 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<BufferCacheChannelInf
static constexpr bool USE_MEMORY_MAPS = P::USE_MEMORY_MAPS;
static constexpr bool SEPARATE_IMAGE_BUFFERS_BINDINGS = P::SEPARATE_IMAGE_BUFFER_BINDINGS;
static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = P::USE_MEMORY_MAPS_FOR_UPLOADS;
static constexpr bool USE_UNIFIED_MEMORY = P::USE_UNIFIED_MEMORY;
#ifdef YUZU_LEGACY
static constexpr s64 TARGET_THRESHOLD = 3_GiB;
@@ -444,15 +443,6 @@ private:
void MappedUploadMemory(Buffer& buffer, u64 total_size_bytes, std::span<BufferCopy> copies);
bool TryUnifiedDownloadMemory(Buffer& buffer, std::span<BufferCopy> copies);
using UnifiedWindowGroups =
boost::container::small_vector<boost::container::small_vector<BufferCopy, 16>, 4>;
bool ResolveUnifiedWindows(DAddr device_addr, u64 buffer_offset, u64 size,
boost::container::small_vector<u64, 4>& window_ids,
UnifiedWindowGroups& groups);
void DownloadBufferMemory(Buffer& buffer_id);
void DownloadBufferMemory(Buffer& buffer_id, DAddr device_addr, u64 size);
@@ -508,14 +498,9 @@ private:
std::deque<Common::RangeSet<DAddr>> committed_gpu_modified_ranges;
// Async Buffers
struct AsyncDownloadBatch {
boost::container::small_vector<BufferCopy, 4> staging_copies;
boost::container::small_vector<BufferCopy, 4> unified_copies;
};
Common::OverlapRangeSet<DAddr> async_downloads;
std::deque<std::optional<Async_Buffer>> async_buffers;
std::deque<AsyncDownloadBatch> pending_downloads;
std::deque<boost::container::small_vector<BufferCopy, 4>> pending_downloads;
std::optional<Async_Buffer> current_buffer;
std::deque<Async_Buffer> async_buffers_death_ring;
+1 -1
View File
@@ -2257,7 +2257,7 @@ public:
/// Returns whether the vertex array specified by index is supposed to be
/// accessed per instance or not.
bool IsInstancingEnabled(std::size_t index) const {
return bool(is_instanced[index]); //FUCK YOU MSVC
return bool(is_instanced[index]);
}
};
-1
View File
@@ -40,7 +40,6 @@ public:
}
struct Regs {
// No fucking idea
INSERT_PADDING_BYTES_NOINIT(0x48);
} regs{};
private:
+1 -1
View File
@@ -963,7 +963,7 @@ void Vic::WriteABGR(const OutputSurfaceConfig& output_surface_config, VideoPixel
auto pixel1213 = _mm_load_si128((__m128i*)&inp[src + x + 12]);
auto pixel1415 = _mm_load_si128((__m128i*)&inp[src + x + 14]);
// Right-shift the channels by 16 to un-do the left shit on read and bring the range
// Right-shift the channels by 16 to un-do the left shift on read and bring the range
// back to 8-bit.
pixel01 = _mm_srli_epi16(pixel01, 2);
pixel23 = _mm_srli_epi16(pixel23, 2);
+1 -1
View File
@@ -441,7 +441,7 @@ void MacroInterpreterImpl::Reset() {
pc = 0;
delayed_pc = {};
method_address.raw = 0;
// Vector must hold its last indices otherwise wonky shit will happen
// Vector must hold its last indices otherwise chaos will ensue
// The next parameter index starts at 1, because $r1 already has the value of the first
// parameter.
next_parameter_index = 1;
@@ -261,7 +261,6 @@ struct BufferCacheParams {
// TODO: Investigate why OpenGL seems to perform worse with persistently mapped buffer uploads
static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = false;
static constexpr bool USE_UNIFIED_MEMORY = false;
};
using BufferCache = VideoCommon::BufferCache<BufferCacheParams>;
+1 -1
View File
@@ -59,7 +59,7 @@ std::vector<std::string> GetExtensions() {
std::vector<std::string> extensions;
for (GLint index = 0; index < num_extensions; ++index) {
auto const* p = reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, GLuint(index)));
if (p != nullptr) // Fuck you? - sincerely, buggy mesa drivers
if (p != nullptr)
extensions.push_back(std::string{p});
}
return extensions;
@@ -237,7 +237,6 @@ ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
.has_gl_bool_ref_bug = device.HasBoolRefBug(),
.ignore_nan_fp_comparisons = true,
.gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
.max_shared_memory_size = device.GetMaxComputeSharedMemorySize(),
.min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
// Use the host limit, but never more than the guest can produce. Maxwell exposes 8 clip
// distances and the SPIR-V output array is sized for at most 8, so clamping here keeps a
@@ -8,7 +8,6 @@
#include <array>
#include <cstring>
#include <span>
#include <utility>
#include <vector>
#include "video_core/buffer_cache/buffer_cache_base.h"
@@ -33,32 +32,6 @@ VkBufferCopy MakeBufferCopy(const VideoCommon::BufferCopy& copy) {
};
}
constexpr size_t MAX_WINDOW_BARRIER_RANGES = 8;
using WindowRange = std::pair<VkDeviceSize, VkDeviceSize>;
using WindowRanges = boost::container::small_vector<WindowRange, MAX_WINDOW_BARRIER_RANGES>;
void CoalesceWindowRanges(WindowRanges& ranges) {
if (ranges.size() < 2) {
return;
}
std::sort(ranges.begin(), ranges.end());
size_t merged = 0;
for (size_t index = 1; index < ranges.size(); ++index) {
if (ranges[index].first <= ranges[merged].second) {
ranges[merged].second = (std::max)(ranges[merged].second, ranges[index].second);
} else {
ranges[++merged] = ranges[index];
}
}
ranges.resize(merged + 1);
if (ranges.size() > MAX_WINDOW_BARRIER_RANGES) {
const WindowRange bounding{ranges.front().first, ranges.back().second};
ranges.clear();
ranges.push_back(bounding);
}
}
VkIndexType IndexTypeFromNumElements(const Device& device, u32 num_elements) {
if (num_elements <= 0xff && device.IsExtIndexTypeUint8Supported()) {
return VK_INDEX_TYPE_UINT8_EXT;
@@ -392,141 +365,6 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
scheduler_, staging_pool_);
}
void BufferCacheRuntime::TryEnableUnifiedMemory(void* base, size_t size,
std::span<AHardwareBuffer* const> hardware_buffers,
size_t hardware_buffer_window,
size_t hardware_buffer_base) {
unified_memory = std::make_unique<HostMemoryImport>(
device, base, size, hardware_buffers, hardware_buffer_window, hardware_buffer_base);
if (!unified_memory->IsValid()) {
unified_memory.reset();
}
}
void BufferCacheRuntime::CopyToUnifiedMemory(
size_t window_index, VkBuffer src_buffer,
std::span<const VideoCommon::BufferCopy> copies) {
if (!unified_memory || src_buffer == VK_NULL_HANDLE || copies.empty() ||
window_index >= unified_memory->GetWindowCount() ||
unified_memory->GetWindowBuffer(window_index) == VK_NULL_HANDLE) {
return;
}
PendingUnifiedCopy& pending = pending_unified_copies.emplace_back();
pending.window = window_index;
pending.buffer = src_buffer;
pending.copies.resize(copies.size());
std::ranges::transform(copies, pending.copies.begin(), MakeBufferCopy);
}
void BufferCacheRuntime::FlushUnifiedMemoryCopies() {
if (pending_unified_copies.empty()) {
return;
}
struct UnifiedCopyCommand {
VkBuffer buffer;
boost::container::small_vector<VkBufferCopy, 8> copies;
};
std::stable_sort(pending_unified_copies.begin(), pending_unified_copies.end(),
[](const PendingUnifiedCopy& lhs, const PendingUnifiedCopy& rhs) {
return lhs.window < rhs.window;
});
const bool foreign = unified_memory->NeedsForeignOwnershipTransfer();
const u32 queue_family = device.GetGraphicsFamily();
size_t group_begin = 0;
while (group_begin < pending_unified_copies.size()) {
const size_t window = pending_unified_copies[group_begin].window;
size_t group_end = group_begin;
while (group_end < pending_unified_copies.size() &&
pending_unified_copies[group_end].window == window) {
++group_end;
}
const VkBuffer window_buffer = unified_memory->GetWindowBuffer(window);
WindowRanges ranges;
for (size_t index = group_begin; index < group_end; ++index) {
for (const VkBufferCopy& copy : pending_unified_copies[index].copies) {
ranges.emplace_back(copy.dstOffset, copy.dstOffset + copy.size);
}
}
CoalesceWindowRanges(ranges);
boost::container::small_vector<VkBufferMemoryBarrier, MAX_WINDOW_BARRIER_RANGES> acquire;
boost::container::small_vector<VkBufferMemoryBarrier, MAX_WINDOW_BARRIER_RANGES> release;
if (foreign) {
for (const WindowRange& range : ranges) {
acquire.push_back(VkBufferMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = 0,
.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
.dstQueueFamilyIndex = queue_family,
.buffer = window_buffer,
.offset = range.first,
.size = range.second - range.first,
});
release.push_back(VkBufferMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = 0,
.srcQueueFamilyIndex = queue_family,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_FOREIGN_EXT,
.buffer = window_buffer,
.offset = range.first,
.size = range.second - range.first,
});
}
}
boost::container::small_vector<UnifiedCopyCommand, 4> commands;
commands.reserve(group_end - group_begin);
for (size_t index = group_begin; index < group_end; ++index) {
PendingUnifiedCopy& pending = pending_unified_copies[index];
commands.push_back(UnifiedCopyCommand{pending.buffer, std::move(pending.copies)});
}
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([window_buffer, acquire = std::move(acquire), release = std::move(release),
commands = std::move(commands)](vk::CommandBuffer cmdbuf) {
if (!acquire.empty()) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT,
VK_PIPELINE_STAGE_TRANSFER_BIT, 0, {},
VideoCommon::FixSmallVectorADL(acquire), {});
}
for (const UnifiedCopyCommand& command : commands) {
cmdbuf.CopyBuffer(command.buffer, window_buffer,
VideoCommon::FixSmallVectorADL(command.copies));
}
if (!release.empty()) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, 0, {},
VideoCommon::FixSmallVectorADL(release), {});
}
});
group_begin = group_end;
}
pending_unified_copies.clear();
}
void BufferCacheRuntime::UnifiedMemoryHostBarrier() {
static constexpr VkMemoryBarrier HOST_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_HOST_READ_BIT,
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([](vk::CommandBuffer cmdbuf) {
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_HOST_BIT, 0,
HOST_BARRIER);
});
}
StagingBufferRef BufferCacheRuntime::UploadStagingBuffer(size_t size) {
return staging_pool.Request(size, MemoryUsage::Upload);
}
@@ -564,7 +402,6 @@ u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
}
void BufferCacheRuntime::TickFrame(Common::SlotVector<Buffer>& slot_buffers) noexcept {
FlushUnifiedMemoryCopies();
for (auto it = slot_buffers.begin(); it != slot_buffers.end(); it++) {
if (scheduler.IsFree(it->LastUsageTick())) {
it->ResetUsageTracking();
@@ -7,10 +7,6 @@
#pragma once
#include <limits>
#include <memory>
#include <span>
#include <boost/container/small_vector.hpp>
#include "video_core/buffer_cache/buffer_cache_base.h"
#include "video_core/buffer_cache/memory_tracker_base.h"
@@ -101,33 +97,6 @@ public:
void TickFrame(Common::SlotVector<Buffer>& slot_buffers) noexcept;
void TryEnableUnifiedMemory(void* base, size_t size,
std::span<AHardwareBuffer* const> hardware_buffers,
size_t hardware_buffer_window, size_t hardware_buffer_base);
[[nodiscard]] bool HasUnifiedMemory() const noexcept {
return unified_memory != nullptr && unified_memory->IsValid();
}
[[nodiscard]] u64 UnifiedMemorySize() const noexcept {
return unified_memory ? unified_memory->GetSize() : 0;
}
[[nodiscard]] u64 UnifiedMemoryBase() const noexcept {
return unified_memory ? unified_memory->GetBaseOffset() : 0;
}
[[nodiscard]] u64 UnifiedMemoryWindowSize() const noexcept {
return unified_memory ? unified_memory->GetWindowSize() : 0;
}
void CopyToUnifiedMemory(size_t window_index, VkBuffer src_buffer,
std::span<const VideoCommon::BufferCopy> copies);
void FlushUnifiedMemoryCopies();
void UnifiedMemoryHostBarrier();
u64 CurrentTick();
u64 KnownGpuTick();
@@ -211,12 +180,6 @@ public:
}
private:
struct PendingUnifiedCopy {
size_t window;
VkBuffer buffer;
boost::container::small_vector<VkBufferCopy, 8> copies;
};
void BindBuffer(const Buffer& buffer, u32 offset, u32 size) {
const VkBuffer handle = buffer.Handle();
if (handle == VK_NULL_HANDLE) {
@@ -241,8 +204,6 @@ private:
std::shared_ptr<QuadStripIndexBuffer> quad_strip_index_buffer;
vk::Buffer null_buffer;
std::unique_ptr<HostMemoryImport> unified_memory;
boost::container::small_vector<PendingUnifiedCopy, 8> pending_unified_copies;
std::unique_ptr<Uint8Pass> uint8_pass;
QuadIndexedPass quad_index_pass;
@@ -265,7 +226,6 @@ struct BufferCacheParams {
static constexpr bool USE_MEMORY_MAPS = true;
static constexpr bool SEPARATE_IMAGE_BUFFER_BINDINGS = false;
static constexpr bool USE_MEMORY_MAPS_FOR_UPLOADS = true;
static constexpr bool USE_UNIFIED_MEMORY = true;
};
using BufferCache = VideoCommon::BufferCache<BufferCacheParams>;
@@ -446,7 +446,6 @@ PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
.has_broken_fp16_float_controls = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY,
.ignore_nan_fp_comparisons = false,
.has_broken_spirv_subgroup_mask_vector_extract_dynamic = false,
.max_shared_memory_size = device.GetMaxComputeSharedMemorySize(),
.has_broken_robust =
device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
.min_ssbo_alignment = device.GetStorageBufferAlignment(),
@@ -922,6 +921,19 @@ std::unique_ptr<ComputePipeline> PipelineCache::CreateComputePipeline(
}
auto program{TranslateProgram(pools.inst, pools.block, env, cfg, host_info)};
const VkDriverIdKHR driver_id = device.GetDriverID();
const bool needs_shared_mem_clamp =
driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
driver_id == VK_DRIVER_ID_ARM_PROPRIETARY;
const u32 max_shared_memory = device.GetMaxComputeSharedMemorySize();
if (needs_shared_mem_clamp && program.shared_memory_size > max_shared_memory) {
LOG_WARNING(Render_Vulkan,
"Compute shader {:#016x} requests {}KB shared memory but device max is {}KB - clamping",
key.unique_hash,
program.shared_memory_size / 1024,
max_shared_memory / 1024);
program.shared_memory_size = max_shared_memory;
}
const std::vector<u32> code{EmitSPIRV(profile, program)};
device.SaveShader(code);
vk::ShaderModule spv_module{BuildShader(device, code)};
@@ -225,13 +225,6 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache, device, scheduler),
wfi_event(device.GetLogical().CreateEvent()) {
scheduler.SetQueryCache(query_cache);
if (Settings::values.use_unified_memory.GetValue() && device_memory.IsBackingShared()) {
buffer_cache_runtime.TryEnableUnifiedMemory(
device_memory.GetPhysicalBase(), device_memory.GetPhysicalSize(),
device_memory.GetBackingHardwareBuffers(),
device_memory.GetBackingHardwareBufferWindowSize(),
device_memory.GetBackingHardwareBufferBase());
}
}
RasterizerVulkan::~RasterizerVulkan() {
@@ -16,7 +16,6 @@
#include <fmt/format.h>
#include "common/assert.h"
#include "common/host_memory.h"
#include "common/literals.h"
#include <ranges>
#include "common/settings.h"
@@ -975,7 +974,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
FOR_EACH_VK_FEATURE_EXT(FEATURE_EXTENSION);
FOR_EACH_VK_EXTENSION(EXTENSION);
FOR_EACH_VK_PLATFORM_EXTENSION(EXTENSION);
if (supported_extensions.contains(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME)) {
loaded_extensions.erase(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
@@ -988,13 +986,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
extensions.robustness_2 = false;
}
#ifdef __ANDROID__
if (extensions.external_memory_ahb && !extensions.queue_family_foreign) {
loaded_extensions.erase(VK_ANDROID_EXTERNAL_MEMORY_ANDROID_HARDWARE_BUFFER_EXTENSION_NAME);
extensions.external_memory_ahb = false;
}
#endif
#undef FEATURE_EXTENSION
#undef EXTENSION
@@ -1151,21 +1142,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_5_PROPERTIES_KHR;
SetNext(next, properties.maintenance5);
}
if (extensions.external_memory_host) {
properties.external_memory_host.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTERNAL_MEMORY_HOST_PROPERTIES_EXT;
SetNext(next, properties.external_memory_host);
}
if (extensions.maintenance3 || instance_version >= VK_API_VERSION_1_1) {
properties.maintenance3.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_3_PROPERTIES;
SetNext(next, properties.maintenance3);
}
if (extensions.maintenance4 || features.maintenance4.maintenance4) {
properties.maintenance4.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_4_PROPERTIES;
SetNext(next, properties.maintenance4);
}
// Perform the property fetch.
physical.GetProperties2(properties2);
@@ -1540,27 +1516,12 @@ void Device::CollectPhysicalMemoryInfo() {
device_access_memory = 0;
u64 device_initial_usage = 0;
u64 local_memory = 0;
const auto heap_has_usable_type = [&mem_properties](size_t heap) {
for (u32 index = 0; index < mem_properties.memoryTypeCount; ++index) {
if (mem_properties.memoryTypes[index].heapIndex != heap) {
continue;
}
if ((mem_properties.memoryTypes[index].propertyFlags &
VK_MEMORY_PROPERTY_PROTECTED_BIT) == 0) {
return true;
}
}
return false;
};
for (size_t element = 0; element < num_properties; ++element) {
const bool is_heap_local =
(mem_properties.memoryHeaps[element].flags & VK_MEMORY_HEAP_DEVICE_LOCAL_BIT) != 0;
if (!is_integrated && !is_heap_local) {
continue;
}
if (!heap_has_usable_type(element)) {
continue;
}
valid_heap_memory.push_back(element);
if (is_heap_local) {
local_memory += mem_properties.memoryHeaps[element].size;
@@ -1572,13 +1533,6 @@ void Device::CollectPhysicalMemoryInfo() {
}
device_access_memory += mem_properties.memoryHeaps[element].size;
}
const u64 committed_backing = Common::GetCommittedBackingSize();
if (committed_backing != 0) {
LOG_INFO(Render_Vulkan, "Discounting {} MiB of guest memory committed by the host",
committed_backing >> 20);
local_memory -= std::min(local_memory, committed_backing);
device_access_memory -= std::min(device_access_memory, committed_backing);
}
if (is_integrated) {
const s64 available_memory = static_cast<s64>(device_access_memory - device_initial_usage);
const u64 memory_size = Settings::values.vram_usage_mode.GetValue() == Settings::VramUsageMode::Aggressive ? 6_GiB : 4_GiB;
@@ -83,7 +83,6 @@ VK_DEFINE_HANDLE(VmaAllocator)
EXTENSION(EXT, CONDITIONAL_RENDERING, conditional_rendering) \
EXTENSION(EXT, CONSERVATIVE_RASTERIZATION, conservative_rasterization) \
EXTENSION(EXT, DEPTH_RANGE_UNRESTRICTED, depth_range_unrestricted) \
EXTENSION(EXT, EXTERNAL_MEMORY_HOST, external_memory_host) \
EXTENSION(EXT, MEMORY_BUDGET, memory_budget) \
EXTENSION(EXT, ROBUSTNESS_2, robustness_2) \
EXTENSION(EXT, SAMPLER_FILTER_MINMAX, sampler_filter_minmax) \
@@ -113,14 +112,6 @@ VK_DEFINE_HANDLE(VmaAllocator)
EXTENSION(IMG, FILTER_CUBIC, filter_cubic_img) \
EXTENSION(QCOM, FILTER_CUBIC_WEIGHTS, filter_cubic_weights)
#ifdef __ANDROID__
#define FOR_EACH_VK_PLATFORM_EXTENSION(EXTENSION) \
EXTENSION(EXT, QUEUE_FAMILY_FOREIGN, queue_family_foreign) \
EXTENSION(ANDROID, EXTERNAL_MEMORY_ANDROID_HARDWARE_BUFFER, external_memory_ahb)
#else
#define FOR_EACH_VK_PLATFORM_EXTENSION(EXTENSION)
#endif
// Define extensions which must be supported.
#define FOR_EACH_VK_MANDATORY_EXTENSION(EXTENSION_NAME) \
EXTENSION_NAME(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME) \
@@ -847,30 +838,6 @@ FN_MAX_LIMIT_LIST
return extensions.conditional_rendering;
}
bool IsExtExternalMemoryHostSupported() const {
return extensions.external_memory_host;
}
bool IsExtExternalMemoryAhbSupported() const {
#ifdef __ANDROID__
return extensions.external_memory_ahb && extensions.queue_family_foreign;
#else
return false;
#endif
}
u64 GetMinImportedHostPointerAlignment() const {
return properties.external_memory_host.minImportedHostPointerAlignment;
}
u64 GetMaxBufferSize() const {
return properties.maintenance4.maxBufferSize;
}
u64 GetMaxMemoryAllocationSize() const {
return properties.maintenance3.maxMemoryAllocationSize;
}
bool IsExtAstcDecodeModeSupported() const {
return extensions.astc_decode_mode;
}
@@ -1143,7 +1110,6 @@ private:
FOR_EACH_VK_FEATURE_1_4(FEATURE);
FOR_EACH_VK_FEATURE_EXT(FEATURE);
FOR_EACH_VK_EXTENSION(EXTENSION);
FOR_EACH_VK_PLATFORM_EXTENSION(EXTENSION);
#undef EXTENSION
#undef FEATURE
@@ -1175,10 +1141,7 @@ private:
VkPhysicalDeviceDescriptorBufferPropertiesEXT descriptor_buffer{};
VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{};
VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{};
VkPhysicalDeviceMaintenance3Properties maintenance3{};
VkPhysicalDeviceMaintenance4Properties maintenance4{};
VkPhysicalDeviceMaintenance5PropertiesKHR maintenance5{};
VkPhysicalDeviceExternalMemoryHostPropertiesEXT external_memory_host{};
VkPhysicalDeviceProperties properties{};
};
@@ -25,34 +25,9 @@
#include "video_core/gpu_logging/gpu_logging.h"
#include "common/settings.h"
#ifdef __ANDROID__
#include <android/hardware_buffer.h>
#endif
namespace Vulkan {
namespace {
[[nodiscard]] std::optional<u32> FindImportMemoryType(
const VkPhysicalDeviceMemoryProperties &props, u32 type_mask) {
const auto find = [&](VkMemoryPropertyFlags wanted) -> std::optional<u32> {
for (u32 i = 0; i < props.memoryTypeCount; ++i) {
if (((type_mask >> i) & 1u) != 0 &&
(props.memoryTypes[i].propertyFlags & wanted) == wanted) {
return i;
}
}
return std::nullopt;
};
auto type_index = find(VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
VK_MEMORY_PROPERTY_HOST_CACHED_BIT);
if (!type_index) {
type_index = find(VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT);
}
return type_index;
}
// Helpers translating MemoryUsage to flags/usage
[[maybe_unused]] VkMemoryPropertyFlags MemoryUsagePropertyFlags(MemoryUsage usage) {
@@ -225,263 +200,6 @@ namespace Vulkan {
size = 0;
}
HostMemoryImport::HostMemoryImport(const Device &device_, void *base, size_t size,
std::span<AHardwareBuffer *const> hardware_buffers,
size_t hardware_buffer_window, size_t hardware_buffer_base)
: device{device_} {
if (ImportHardwareBuffers(hardware_buffers, hardware_buffer_window, hardware_buffer_base,
size)) {
return;
}
if (device.IsTiler()) {
return;
}
if (!hardware_buffers.empty()) {
return;
}
if (ImportHostPointer(base, size)) {
return;
}
LOG_INFO(Render_Vulkan, "Unified memory disabled, no host memory import path");
}
bool HostMemoryImport::ImportHostPointer(void *base, size_t size) {
if (!device.IsExtExternalMemoryHostSupported()) {
return false;
}
const u64 alignment = device.GetMinImportedHostPointerAlignment();
if (alignment == 0 || !Common::IsAligned(reinterpret_cast<uintptr_t>(base), alignment) ||
!Common::IsAligned(size, alignment)) {
return false;
}
using namespace Common::Literals;
constexpr VkDeviceSize DesktopWindowSize = 4_GiB;
VkDeviceSize candidate_window = DesktopWindowSize;
const u64 max_buffer_size = device.GetMaxBufferSize();
if (max_buffer_size != 0 && max_buffer_size < candidate_window) {
candidate_window = max_buffer_size;
}
const u64 max_allocation_size = device.GetMaxMemoryAllocationSize();
if (max_allocation_size != 0 && max_allocation_size < candidate_window) {
candidate_window = max_allocation_size;
}
candidate_window = Common::AlignDown(candidate_window, alignment);
if (candidate_window == 0) {
return false;
}
window_size = candidate_window;
const auto &logical = device.GetLogical();
const auto memory_props = device.GetPhysical().GetMemoryProperties().memoryProperties;
for (size_t offset = 0; offset < size; offset += window_size) {
u8 *const window_base = static_cast<u8 *>(base) + offset;
const VkDeviceSize window_len =
(std::min)(static_cast<VkDeviceSize>(size - offset), window_size);
VkMemoryHostPointerPropertiesEXT host_props{
.sType = VK_STRUCTURE_TYPE_MEMORY_HOST_POINTER_PROPERTIES_EXT,
.pNext = nullptr,
.memoryTypeBits = 0,
};
if (logical.GetMemoryHostPointerPropertiesEXT(
VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT, window_base,
&host_props) != VK_SUCCESS ||
host_props.memoryTypeBits == 0) {
break;
}
const VkExternalMemoryBufferCreateInfo external_info{
.sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_BUFFER_CREATE_INFO,
.pNext = nullptr,
.handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
};
const VkBufferCreateInfo buffer_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = &external_info,
.flags = 0,
.size = window_len,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer new_buffer{};
if (logical.CreateBufferRaw(buffer_ci, &new_buffer) != VK_SUCCESS) {
break;
}
const VkMemoryRequirements requirements =
logical.GetBufferMemoryRequirements(new_buffer);
const u32 type_mask = requirements.memoryTypeBits & host_props.memoryTypeBits;
if (type_mask == 0 || requirements.size > window_len) {
logical.DestroyBufferRaw(new_buffer);
break;
}
const auto type_index = FindImportMemoryType(memory_props, type_mask);
if (!type_index) {
logical.DestroyBufferRaw(new_buffer);
break;
}
const u32 heap_index = memory_props.memoryTypes[*type_index].heapIndex;
const VkDeviceSize heap_size = memory_props.memoryHeaps[heap_index].size;
if (imported_size + window_len > heap_size / 2) {
logical.DestroyBufferRaw(new_buffer);
break;
}
const VkImportMemoryHostPointerInfoEXT import_info{
.sType = VK_STRUCTURE_TYPE_IMPORT_MEMORY_HOST_POINTER_INFO_EXT,
.pNext = nullptr,
.handleType = VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT,
.pHostPointer = window_base,
};
const VkMemoryAllocateInfo alloc_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = &import_info,
.allocationSize = window_len,
.memoryTypeIndex = *type_index,
};
vk::DeviceMemory memory = logical.TryAllocateMemory(alloc_info);
if (!memory) {
logical.DestroyBufferRaw(new_buffer);
break;
}
if (logical.BindBufferMemory(new_buffer, *memory, 0) != VK_SUCCESS) {
logical.DestroyBufferRaw(new_buffer);
break;
}
windows.push_back(Window{
.memory = std::move(memory),
.buffer = new_buffer,
});
imported_size += static_cast<size_t>(window_len);
}
if (windows.empty()) {
return false;
}
return true;
}
bool HostMemoryImport::ImportHardwareBuffers(
[[maybe_unused]] std::span<AHardwareBuffer *const> hardware_buffers,
[[maybe_unused]] size_t hardware_buffer_window,
[[maybe_unused]] size_t hardware_buffer_base, [[maybe_unused]] size_t size) {
#ifdef __ANDROID__
if (hardware_buffers.empty() || hardware_buffer_window == 0 ||
!device.IsExtExternalMemoryAhbSupported()) {
return false;
}
const u64 max_allocation_size = device.GetMaxMemoryAllocationSize();
if (max_allocation_size != 0 && hardware_buffer_window > max_allocation_size) {
return false;
}
if (hardware_buffer_base >= size) {
return false;
}
const auto &logical = device.GetLogical();
const auto memory_props = device.GetPhysical().GetMemoryProperties().memoryProperties;
window_size = hardware_buffer_window;
base_offset = hardware_buffer_base;
for (size_t i = 0; i < hardware_buffers.size(); ++i) {
const size_t offset = hardware_buffer_base + i * hardware_buffer_window;
if (offset >= size) {
break;
}
const VkDeviceSize window_len = (std::min)(
static_cast<VkDeviceSize>(size - offset),
static_cast<VkDeviceSize>(hardware_buffer_window));
VkAndroidHardwareBufferPropertiesANDROID ahb_props{
.sType = VK_STRUCTURE_TYPE_ANDROID_HARDWARE_BUFFER_PROPERTIES_ANDROID,
.pNext = nullptr,
.allocationSize = 0,
.memoryTypeBits = 0,
};
if (logical.GetAndroidHardwareBufferPropertiesANDROID(hardware_buffers[i],
&ahb_props) != VK_SUCCESS ||
ahb_props.memoryTypeBits == 0 || ahb_props.allocationSize < window_len) {
break;
}
const VkExternalMemoryBufferCreateInfo external_info{
.sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_BUFFER_CREATE_INFO,
.pNext = nullptr,
.handleTypes =
VK_EXTERNAL_MEMORY_HANDLE_TYPE_ANDROID_HARDWARE_BUFFER_BIT_ANDROID,
};
const VkBufferCreateInfo buffer_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = &external_info,
.flags = 0,
.size = window_len,
.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer new_buffer{};
if (logical.CreateBufferRaw(buffer_ci, &new_buffer) != VK_SUCCESS) {
break;
}
const VkMemoryRequirements requirements =
logical.GetBufferMemoryRequirements(new_buffer);
const u32 type_mask = requirements.memoryTypeBits & ahb_props.memoryTypeBits;
if (type_mask == 0 || requirements.size > ahb_props.allocationSize) {
logical.DestroyBufferRaw(new_buffer);
break;
}
const auto type_index = FindImportMemoryType(memory_props, type_mask);
if (!type_index) {
logical.DestroyBufferRaw(new_buffer);
break;
}
const VkImportAndroidHardwareBufferInfoANDROID import_info{
.sType = VK_STRUCTURE_TYPE_IMPORT_ANDROID_HARDWARE_BUFFER_INFO_ANDROID,
.pNext = nullptr,
.buffer = hardware_buffers[i],
};
const VkMemoryDedicatedAllocateInfo dedicated_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_DEDICATED_ALLOCATE_INFO,
.pNext = &import_info,
.image = VK_NULL_HANDLE,
.buffer = new_buffer,
};
const VkMemoryAllocateInfo alloc_info{
.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
.pNext = &dedicated_info,
.allocationSize = ahb_props.allocationSize,
.memoryTypeIndex = *type_index,
};
vk::DeviceMemory memory = logical.TryAllocateMemory(alloc_info);
if (!memory) {
logical.DestroyBufferRaw(new_buffer);
break;
}
if (logical.BindBufferMemory(new_buffer, *memory, 0) != VK_SUCCESS) {
logical.DestroyBufferRaw(new_buffer);
break;
}
windows.push_back(Window{
.memory = std::move(memory),
.buffer = new_buffer,
});
imported_size += static_cast<size_t>(window_len);
}
if (windows.empty()) {
window_size = 0;
base_offset = 0;
return false;
}
foreign_ownership = true;
return true;
#else
return false;
#endif
}
HostMemoryImport::~HostMemoryImport() {
for (Window &window : windows) {
if (window.buffer != VK_NULL_HANDLE) {
device.GetLogical().DestroyBufferRaw(window.buffer);
}
}
}
MemoryAllocator::MemoryAllocator(const Device &device_)
: device{device_}, allocator{device.GetAllocator()},
properties{device_.GetPhysical().GetMemoryProperties().memoryProperties},
@@ -15,8 +15,6 @@
#include "video_core/vulkan_common/vulkan_wrapper.h"
#include "video_core/vulkan_common/vma.h"
struct AHardwareBuffer;
namespace Vulkan {
class Device;
@@ -86,66 +84,6 @@ namespace Vulkan {
void *mapped_ptr{}; ///< Optional persistent mapped pointer
};
class HostMemoryImport {
public:
explicit HostMemoryImport(const Device &device_, void *base, size_t size,
std::span<AHardwareBuffer *const> hardware_buffers,
size_t hardware_buffer_window, size_t hardware_buffer_base);
~HostMemoryImport();
HostMemoryImport(const HostMemoryImport &) = delete;
HostMemoryImport &operator=(const HostMemoryImport &) = delete;
[[nodiscard]] bool IsValid() const noexcept {
return !windows.empty();
}
[[nodiscard]] size_t GetSize() const noexcept {
return imported_size;
}
[[nodiscard]] size_t GetBaseOffset() const noexcept {
return base_offset;
}
[[nodiscard]] bool NeedsForeignOwnershipTransfer() const noexcept {
return foreign_ownership;
}
[[nodiscard]] VkDeviceSize GetWindowSize() const noexcept {
return window_size;
}
[[nodiscard]] VkBuffer GetWindowBuffer(size_t index) const noexcept {
return windows[index].buffer;
}
[[nodiscard]] size_t GetWindowCount() const noexcept {
return windows.size();
}
private:
struct Window {
vk::DeviceMemory memory;
VkBuffer buffer{};
};
bool ImportHostPointer(void *base, size_t size);
bool ImportHardwareBuffers(std::span<AHardwareBuffer *const> hardware_buffers,
size_t hardware_buffer_window, size_t hardware_buffer_base,
size_t size);
const Device &device;
std::vector<Window> windows;
VkDeviceSize window_size{};
size_t imported_size{};
size_t base_offset{};
bool foreign_ownership{};
};
/// Memory allocator container.
/// Allocates and releases memory allocations on demand.
class MemoryAllocator {
@@ -216,16 +216,12 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkGetBufferMemoryRequirements2);
X(vkGetDeviceQueue);
X(vkGetEventStatus);
X(vkGetMemoryHostPointerPropertiesEXT);
X(vkGetFenceStatus);
X(vkGetImageMemoryRequirements);
X(vkGetPipelineCacheData);
X(vkGetMemoryFdKHR);
#ifdef _WIN32
X(vkGetMemoryWin32HandleKHR);
#endif
#ifdef __ANDROID__
X(vkGetAndroidHardwareBufferPropertiesANDROID);
#endif
X(vkGetQueryPoolResults);
X(vkGetPipelineExecutablePropertiesKHR);
@@ -332,16 +332,12 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkGetBufferMemoryRequirements2 vkGetBufferMemoryRequirements2{};
PFN_vkGetDeviceQueue vkGetDeviceQueue{};
PFN_vkGetEventStatus vkGetEventStatus{};
PFN_vkGetMemoryHostPointerPropertiesEXT vkGetMemoryHostPointerPropertiesEXT{};
PFN_vkGetFenceStatus vkGetFenceStatus{};
PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements{};
PFN_vkGetPipelineCacheData vkGetPipelineCacheData{};
PFN_vkGetMemoryFdKHR vkGetMemoryFdKHR{};
#ifdef _WIN32
PFN_vkGetMemoryWin32HandleKHR vkGetMemoryWin32HandleKHR{};
#endif
#ifdef __ANDROID__
PFN_vkGetAndroidHardwareBufferPropertiesANDROID vkGetAndroidHardwareBufferPropertiesANDROID{};
#endif
PFN_vkGetPipelineExecutablePropertiesKHR vkGetPipelineExecutablePropertiesKHR{};
PFN_vkGetPipelineExecutableStatisticsKHR vkGetPipelineExecutableStatisticsKHR{};
@@ -1086,34 +1082,6 @@ public:
VkMemoryRequirements GetBufferMemoryRequirements(VkBuffer buffer,
void* pnext = nullptr) const noexcept;
VkResult GetMemoryHostPointerPropertiesEXT(
VkExternalMemoryHandleTypeFlagBits handle_type, const void* host_pointer,
VkMemoryHostPointerPropertiesEXT* out_properties) const noexcept {
return dld->vkGetMemoryHostPointerPropertiesEXT(handle, handle_type, host_pointer,
out_properties);
}
#ifdef __ANDROID__
VkResult GetAndroidHardwareBufferPropertiesANDROID(
const struct AHardwareBuffer* buffer,
VkAndroidHardwareBufferPropertiesANDROID* out_properties) const noexcept {
return dld->vkGetAndroidHardwareBufferPropertiesANDROID(handle, buffer, out_properties);
}
#endif
VkResult CreateBufferRaw(const VkBufferCreateInfo& ci, VkBuffer* out_buffer) const noexcept {
return dld->vkCreateBuffer(handle, &ci, nullptr, out_buffer);
}
void DestroyBufferRaw(VkBuffer buffer) const noexcept {
dld->vkDestroyBuffer(handle, buffer, nullptr);
}
VkResult BindBufferMemory(VkBuffer buffer, VkDeviceMemory memory,
VkDeviceSize offset) const noexcept {
return dld->vkBindBufferMemory(handle, buffer, memory, offset);
}
VkMemoryRequirements GetImageMemoryRequirements(VkImage image) const noexcept;
std::vector<VkPipelineExecutablePropertiesKHR> GetPipelineExecutablePropertiesKHR(
@@ -99,6 +99,7 @@ ConfigureInputAdvanced::ConfigureInputAdvanced(Core::HID::HIDCore& hid_core_, QW
#ifndef _WIN32
ui->enable_raw_input->setVisible(false);
ui->disable_wgi_xinput->setVisible(false);
#endif
LoadConfiguration();
@@ -139,6 +140,7 @@ void ConfigureInputAdvanced::ApplyConfiguration() {
Settings::values.emulate_analog_keyboard = ui->emulate_analog_keyboard->isChecked();
Settings::values.touchscreen.enabled = ui->touchscreen_enabled->isChecked();
Settings::values.enable_raw_input = ui->enable_raw_input->isChecked();
Settings::values.disable_wgi_xinput = ui->disable_wgi_xinput->isChecked();
Settings::values.enable_udp_controller = ui->enable_udp_controller->isChecked();
Settings::values.controller_navigation = ui->controller_navigation->isChecked();
Settings::values.enable_ring_controller = ui->enable_ring_controller->isChecked();
@@ -174,6 +176,7 @@ void ConfigureInputAdvanced::LoadConfiguration() {
ui->emulate_analog_keyboard->setChecked(Settings::values.emulate_analog_keyboard.GetValue());
ui->touchscreen_enabled->setChecked(Settings::values.touchscreen.enabled);
ui->enable_raw_input->setChecked(Settings::values.enable_raw_input.GetValue());
ui->disable_wgi_xinput->setChecked(Settings::values.disable_wgi_xinput.GetValue());
ui->enable_udp_controller->setChecked(Settings::values.enable_udp_controller.GetValue());
ui->controller_navigation->setChecked(Settings::values.controller_navigation.GetValue());
ui->enable_ring_controller->setChecked(Settings::values.enable_ring_controller.GetValue());
@@ -2757,6 +2757,22 @@
</property>
</widget>
</item>
<item row="9" column="0">
<widget class="QCheckBox" name="disable_wgi_xinput">
<property name="toolTip">
<string>Aimed to disable SDL GUIDE button hack: synthetic GUIDE(HOME) event when SELECT(MINUS) + START(PLUS) pressed. May impact Win related trigger/rumble/etc stuff</string>
</property>
<property name="minimumSize">
<size>
<width>0</width>
<height>23</height>
</size>
</property>
<property name="text">
<string>Disable SDL WGI/XInput (Requires restart)</string>
</property>
</widget>
</item>
</layout>
</widget>
</item>
+16
View File
@@ -16,6 +16,12 @@
#include <sys/resource.h>
#endif
#if defined(__APPLE__)
#include <climits>
#include <cstdlib>
#include <cstring>
#endif
#include "main_window.h"
#ifdef _WIN32
@@ -126,6 +132,16 @@ int main(int argc, char* argv[]) {
#endif // _WIN32
#if defined(__APPLE__)
// Convert the relative path to an absolute path before the chdir
char resolved[PATH_MAX];
for (int i = 1; i < argc; ++i) {
if (strcmp(argv[i], "-u") == 0 || strcmp(argv[i], "-input-profile") == 0) {
++i;
} else if (argv[i][0] != '-' && argv[i][0] != '/' && realpath(argv[i], resolved)) {
argv[i] = strdup(resolved);
}
}
// If you start a bundle (binary) on OSX without the Terminal, the working directory is "/".
// But since we require the working directory to be the executable path for the location of
// the user folder in the Qt Frontend, we need to cd into that working directory
+2 -1
View File
@@ -1,7 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// Qt on macOS doesn't define VMA shit
// Static Qt on macOS doesn't use Vulkan
// Other platforms do, and thus have conflicting VulkanMemoryAllocator symbols
#if defined(QT_STATICPLUGIN) && !defined(__APPLE__)
#undef VMA_IMPLEMENTATION
#endif
+1 -1
View File
@@ -38,7 +38,7 @@ void MigrationWorker::process() {
try {
fs::remove_all(eden_dir);
} catch (fs::filesystem_error& _) {
// ignore because linux does stupid crap sometimes
// directory may not exist at this point, that's fine
}
switch (strategy) {