[dynarmic] inlined pool in block + slab-like for each block

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

README.md

@@ -57,13 +57,7 @@ If you would like to contribute, we are open to new developers and pull requests
 
 ## Building
 
-* **Windows**: [Windows Building Guide](./docs/build/Windows.md)
-* **Linux**: [Linux Building Guide](./docs/build/Linux.md)
-* **Android**: [Android Building Guide](./docs/build/Android.md)
-* **Solaris**: [Solaris Building Guide](./docs/build/Solaris.md)
-* **FreeBSD**: [FreeBSD Building Guide](./docs/build/FreeBSD.md)
-* **macOS**: [macOS Building Guide](./docs/build/macOS.md)
-* **OpenBSD**: [OpenBSD Building Guide](./docs/build/OpenBSD.md)
+See the [General Build Guide](docs/Build.md)
 
 ## Download
 

docs/Build.md

@@ -0,0 +1,160 @@
+# Building Eden
+
+> [!WARNING]
+> This guide is intended for developers ONLY. If you are not a developer or packager, you are unlikely to receive support.
+
+This is a full-fledged guide to build Eden on all supported platforms.
+
+## Dependencies
+First, you must [install some dependencies](Deps.md).
+
+## Clone
+Next, you will want to clone Eden via the terminal:
+
+```sh
+git clone https://git.eden-emu.dev/eden-emu/eden.git
+cd eden
+```
+
+Or use Qt Creator (Create Project -> Import Project -> Git Clone).
+
+## Android
+
+Android has a completely different build process than other platforms. See its [dedicated page](build/Android.md).
+
+## Initial Configuration
+
+If the configure phase fails, see the `Troubleshooting` section below. Usually, as long as you followed the dependencies guide, the defaults *should* successfully configure and build.
+
+### Qt Creator
+
+This is the recommended GUI method for Linux, macOS, and Windows.
+
+<details>
+<summary>Click to Open</summary>
+
+> [!WARNING]
+> On MSYS2, to use Qt Creator you are recommended to *also* install Qt from the online installer, ensuring to select the "MinGW" version.
+
+Open the CMakeLists.txt file in your cloned directory via File -> Open File or Project (Ctrl+O), if you didn't clone Eden via the project import tool.
+
+Select your desired "kit" (usually, the default is okay). RelWithDebInfo or Release is recommended:
+
+![Qt Creator kits](img/creator-1.png)
+
+Hit "Configure Project", then wait for CMake to finish configuring (may take a while on Windows).
+
+</details>
+
+### Command Line
+
+This is recommended for *BSD, Solaris, Linux, and MSYS2. MSVC is possible, but not recommended.
+
+<details>
+<summary>Click to Open</summary>
+
+Note that CMake must be in your PATH, and you must be in the cloned Eden directory. On Windows, you must also set up a Visual C++ development environment. This can be done by running `C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat` in the same terminal.
+
+Recommended generators:
+
+- MSYS2: `MSYS Makefiles`
+- MSVC: Install **[ninja](https://ninja-build.org/)** and use `Ninja`, OR use `Visual Studio 17 2022`
+- macOS: `Ninja` (preferred) or `Xcode`
+- Others: `Ninja` (preferred) or `UNIX Makefiles`
+
+BUILD_TYPE should usually be `Release` or `RelWithDebInfo` (debug symbols--compiled executable will be large). If you are using a debugger and annoyed with stuff getting optimized out, try `Debug`.
+
+Also see the [Options](Options.md) page for additional CMake options.
+
+```sh
+cmake -S . -B build -G "GENERATOR" -DCMAKE_BUILD_TYPE=<BUILD_TYPE> -DYUZU_TESTS=OFF
+```
+
+If you are on Windows and prefer to use Clang:
+
+```sh
+cmake -S . -B build -G "GENERATOR" -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl
+```
+
+</details>
+
+### [CLion](https://www.jetbrains.com/clion/)
+
+<details>
+<summary>Click to Open</summary>
+
+* Clone the Repository:
+
+<img src="https://user-images.githubusercontent.com/42481638/216899046-0d41d7d6-8e4d-4ed2-9587-b57088af5214.png" width="500">
+<img src="https://user-images.githubusercontent.com/42481638/216899061-b2ea274a-e88c-40ae-bf0b-4450b46e9fea.png" width="500">
+<img src="https://user-images.githubusercontent.com/42481638/216899076-0e5988c4-d431-4284-a5ff-9ecff973db76.png" width="500">
+
+---
+
+### Building & Setup
+
+* Once Cloned, You will be taken to a prompt like the image below:
+
+<img src="https://user-images.githubusercontent.com/42481638/216899092-3fe4cec6-a540-44e3-9e1e-3de9c2fffc2f.png" width="500">
+
+* Set the settings to the image below:
+* Change `Build type: Release`
+* Change `Name: Release`
+* Change `Toolchain Visual Studio`
+* Change `Generator: Let CMake decide`
+* Change `Build directory: build`
+
+<img src="https://user-images.githubusercontent.com/42481638/216899164-6cee8482-3d59-428f-b1bc-e6dc793c9b20.png" width="500">
+
+* Click OK; now Clion will build a directory and index your code to allow for IntelliSense. Please be patient.
+* Once this process has been completed (No loading bar bottom right), you can now build eden
+* In the top right, click on the drop-down menu, select all configurations, then select eden
+
+<img src="https://user-images.githubusercontent.com/42481638/216899226-975048e9-bc6d-4ec1-bc2d-bd8a1e15ed04.png" height="500" >
+
+* Now run by clicking the play button or pressing Shift+F10, and eden will auto-launch once built.
+
+<img src="https://user-images.githubusercontent.com/42481638/216899275-d514ec6a-e563-470e-81e2-3e04f0429b68.png" width="500">
+</details>
+
+## Troubleshooting
+
+If your initial configure failed:
+- *Carefully* re-read the [dependencies guide](Deps.md)
+- Clear the CPM cache (`.cache/cpm`) and CMake cache (`<build directory>/CMakeCache.txt`)
+- Evaluate the error and find any related settings
+- See the [CPM docs](CPM.md) to see if you may need to forcefully bundle any packages
+
+Otherwise, feel free to ask for help in Revolt or Discord.
+
+## Caveats
+
+Many platforms have quirks, bugs, and other fun stuff that may cause issues when building OR running. See the [Caveats page](Caveats.md) before continuing.
+
+## Building & Running
+
+### Qt Creator
+
+Simply hit Ctrl+B, or the "hammer" icon in the bottom left. To run, hit the "play" icon, or Ctrl+R.
+
+### Command Line
+
+If you are not on Windows and are using the `UNIX Makefiles` generator, you must also add `-j$(nproc)` to this command.
+
+```
+cmake --build build
+```
+
+Your compiled executable will be in:
+- `build/bin/eden.exe` for Windows,
+- `build/bin/eden.app/Contents/MacOS/eden` for macOS,
+- and `build/bin/eden` for others.
+
+## Scripts
+
+Some platforms have convenience scripts provided for building.
+
+- **[Linux](scripts/Linux.md)**
+- **[Windows](scripts/Windows.md)**
+
+macOS scripts will come soon.

docs/CODEOWNERS

@@ -0,0 +1,26 @@
+# ui stuff
+/src/android         @AleksandrPopovich @nyxynx @Producdevity
+/src/yuzu            @crueter
+/src/eden            @crueter
+/src/frontend_common @crueter
+/src/qt_common       @crueter
+
+# docs, meta
+/docs                @Lizzie @crueter
+/.ci                 @crueter
+
+# cmake
+*.cmake             @crueter
+*/CMakeLists.txt    @crueter
+*.in                @crueter
+
+# individual stuff
+/src/web_service     @AleksandrPopovich
+/src/dynarmic        @Lizzie
+/src/core            @Lizzie @Maufeat @PavelBARABANOV @MrPurple666
+/src/core/hle        @Maufeat @PavelBARABANOV @SDK-Chan
+/src/*_room          @AleksandrPopovich
+/src/video_core      @CamilleLaVey @MaranBr @Wildcard @weakboson
+
+# Global owners/triage
+*                    @CamilleLaVey @Maufeat @crueter @MrPurple666 @MaranBr @Lizzie

docs/CPM.md

@@ -177,7 +177,7 @@ If `ci` is `true`:
 
 ### Examples
 
-In order: OpenSSL CI, Boost (tag + artifact), discord-rpc (sha + options + patches), Opus (options + find_args)
+In order: OpenSSL CI, Boost (tag + artifact), Opus (options + find_args), discord-rpc (sha + options + patches)
 
 ```json
 {

docs/Caveats.md

@@ -0,0 +1,52 @@
+# Caveats
+
+## Arch Linux
+
+- httplib AUR package is broken. Set `httplib_FORCE_BUNDLED=ON` if you have it installed.
+- Eden is also available as an [AUR package](https://aur.archlinux.org/packages/eden-git). If you are unable to build, either use that or compare your process to the PKGBUILD.
+
+## Gentoo Linux
+
+Do not use the system sirit or xbyak packages.
+
+## macOS
+
+macOS is largely untested. Expect crashes, significant Vulkan issues, and other fun stuff.
+
+## Solaris
+
+Qt Widgets appears to be broken. For now, add `-DENABLE_QT=OFF` to your configure command. In the meantime, a Qt Quick frontend is in the works--check back later!
+
+This is needed for some dependencies that call cc directly (tz):
+
+```sh
+echo '#!/bin/sh' >cc
+echo 'gcc $@' >>cc
+chmod +x cc
+export PATH="$PATH:$PWD"
+```
+
+Default MESA is a bit outdated, the following environment variables should be set for a smoother experience:
+```sh
+export MESA_GL_VERSION_OVERRIDE=4.6
+export MESA_GLSL_VERSION_OVERRIDE=460
+export MESA_EXTENSION_MAX_YEAR=2025
+export MESA_DEBUG=1
+export MESA_VK_VERSION_OVERRIDE=1.3
+# Only if nvidia/intel drm drivers cause crashes, will severely hinder performance
+export LIBGL_ALWAYS_SOFTWARE=1
+```
+
+- Modify the generated ffmpeg.make (in build dir) if using multiple threads (base system `make` doesn't use `-j4`, so change for `gmake`).
+- If using OpenIndiana, due to a bug in SDL2's CMake configuration, audio driver defaults to SunOS `<sys/audioio.h>`, which does not exist on OpenIndiana. Using external or bundled SDL2 may solve this.
+- System OpenSSL generally does not work. Instead, use `-DYUZU_USE_BUNDLED_OPENSSL=ON` to use a bundled static OpenSSL, or build a system dependency from source.
+
+## OpenBSD
+
+After configuration, you may need to modify `externals/ffmpeg/CMakeFiles/ffmpeg-build/build.make` to use `-j$(nproc)` instead of just `-j`.
+
+## FreeBSD
+
+Eden is not currently available as a port on FreeBSD, though it is in the works. For now, the recommended method of usage is to compile it yourself.
+
+The available OpenSSL port (3.0.17) is out-of-date, and using a bundled static library instead is recommended; to do so, add `-DYUZU_USE_BUNDLED_OPENSSL=ON` to your CMake configure command.

docs/Deps.md

@@ -0,0 +1,214 @@
+# Dependencies
+
+To build Eden, you MUST have a C++ compiler.
+* On Linux, this is usually [GCC](https://gcc.gnu.org/) 11+ or [Clang](https://clang.llvm.org/) v14+
+  - GCC 12 also requires Clang 14+
+* On Windows, this is either:
+  - **[MSVC](https://visualstudio.microsoft.com/downloads/)**,
+    * *A convenience script to install the **minimal** version (Visual Build Tools) is provided in `.ci/windows/install-msvc.ps1`*
+  - clang-cl - can be downloaded from the MSVC installer,
+  - or **[MSYS2](https://www.msys2.org)**
+* On macOS, this is Apple Clang
+  - This can be installed with `xcode-select --install`
+
+The following additional tools are also required:
+
+* **[CMake](https://www.cmake.org/)** 3.22+ - already included with the Android SDK
+* **[Git](https://git-scm.com/)** for version control
+  - **[Windows installer](https://gitforwindows.org)**
+* On Windows, you must install the **[Vulkan SDK](https://vulkan.lunarg.com/sdk/home#windows)** as well
+  - *A convenience script to install the latest SDK is provided in `.ci/windows/install-vulkan-sdk.ps1`*
+
+If you are on desktop and plan to use the Qt frontend, you *must* install Qt 6, and optionally Qt Creator (the recommended IDE for building)
+* On Linux, *BSD and macOS, this can be done by the package manager
+  - If you wish to use Qt Creator, append `qtcreator` or `qt-creator` to the commands seen below.
+* MSVC/clang-cl users on Windows must install through the [official installer](https://www.qt.io/download-qt-installer-oss)
+* Linux and macOS users may choose to use the installer as well.
+* MSYS2 can also install Qt 6 via the package manager
+
+If you are on Windows and NOT building with MSYS2, you may go [back home](Build.md) and continue.
+
+## Externals
+The following are handled by Eden's externals:
+
+* [FFmpeg](https://ffmpeg.org/) (should use `-DYUZU_USE_EXTERNAL_FFMPEG=ON`)
+* [SDL2](https://www.libsdl.org/download-2.0.php) 2.0.18+ (should use `-DYUZU_USE_EXTERNAL_SDL2=ON` OR `-DYUZU_USE_BUNDLED_SDL2=ON` to reduce compile time)
+
+All other dependencies will be downloaded and built by [CPM](https://github.com/cpm-cmake/CPM.cmake/) if `YUZU_USE_CPM` is on, but will always use system dependencies if available (UNIX-like only):
+
+* [Boost](https://www.boost.org/users/download/) 1.57.0+
+* [Catch2](https://github.com/catchorg/Catch2) 3.0.1 if `YUZU_TESTS` or `DYNARMIC_TESTS` are on
+* [fmt](https://fmt.dev/) 8.0.1+
+* [lz4](http://www.lz4.org)
+* [nlohmann\_json](https://github.com/nlohmann/json) 3.8+
+* [OpenSSL](https://www.openssl.org/source/) 1.1.1+
+* [ZLIB](https://www.zlib.net/) 1.2+
+* [zstd](https://facebook.github.io/zstd/) 1.5+
+* [enet](http://enet.bespin.org/) 1.3+
+* [Opus](https://opus-codec.org/) 1.3+
+* [MbedTLS](https://github.com/Mbed-TLS/mbedtls) 3+
+
+Vulkan 1.3.274+ is also needed:
+* [VulkanUtilityLibraries](https://github.com/KhronosGroup/Vulkan-Utility-Libraries)
+* [VulkanHeaders](https://github.com/KhronosGroup/Vulkan-Headers)
+* [SPIRV-Tools](https://github.com/KhronosGroup/SPIRV-Tools)
+* [SPIRV-Headers](https://github.com/KhronosGroup/SPIRV-Headers)
+
+Certain other dependencies will be fetched by CPM regardless. System packages *can* be used for these libraries, but many are either not packaged by most distributions OR have issues when used by the system:
+
+* [SimpleIni](https://github.com/brofield/simpleini)
+* [DiscordRPC](https://github.com/eden-emulator/discord-rpc)
+* [cubeb](https://github.com/mozilla/cubeb)
+* [libusb](https://github.com/libusb/libusb)
+* [VulkanMemoryAllocator](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator)
+* [sirit](https://github.com/eden-emulator/sirit)
+* [httplib](https://github.com/yhirose/cpp-httplib) - if `ENABLE_QT_UPDATE_CHECKER` or `ENABLE_WEB_SERVICE` are on
+* [cpp-jwt](https://github.com/arun11299/cpp-jwt) 1.4+ - if `ENABLE_WEB_SERVICE` is on
+* [unordered-dense](https://github.com/martinus/unordered_dense)
+* [mcl](https://github.com/azahar-emu/mcl) - subject to removal
+
+On amd64:
+* [xbyak](https://github.com/herumi/xbyak) - 7.22 or earlier is recommended
+* [zycore](https://github.com/zyantific/zycore-c)
+* [zydis](https://github.com/zyantific/zydis) 4+
+* Note: zydis and zycore-c MUST match. Using one as a system dependency and the other as a bundled dependency WILL break things
+
+On aarch64 OR if `DYNARMIC_TESTS` is on:
+* [oaknut](https://github.com/merryhime/oaknut) 2.0.1+
+
+On riscv64:
+* [biscuit](https://github.com/lioncash/biscuit) 0.9.1+
+
+## Commands
+
+These are commands to install all necessary dependencies on various Linux and BSD distributions, as well as macOS. Always review what you're running before you hit Enter!
+
+Click on the arrows to expand.
+
+<details>
+<summary>Arch Linux</summary>
+
+```sh
+sudo pacman -Syu --needed base-devel boost catch2 cmake enet ffmpeg fmt git glslang libzip lz4 mbedtls ninja nlohmann-json openssl opus qt6-base qt6-multimedia sdl2 zlib zstd zip unzip zydis zycore vulkan-headers vulkan-utility-libraries libusb spirv-tools spirv-headers
+```
+
+* Building with QT Web Engine requires `qt6-webengine` as well.
+* Proper Wayland support requires `qt6-wayland`
+* GCC 11 or later is required.
+</details>
+
+<details>
+<summary>Ubuntu, Debian, Mint Linux</summary>
+
+```sh
+sudo apt-get install autoconf cmake g++ gcc git glslang-tools libasound2 libboost-context-dev libglu1-mesa-dev libhidapi-dev libpulse-dev libtool libudev-dev libxcb-icccm4 libxcb-image0 libxcb-keysyms1 libxcb-render-util0 libxcb-xinerama0 libxcb-xkb1 libxext-dev libxkbcommon-x11-0 mesa-common-dev nasm ninja-build qt6-base-private-dev libmbedtls-dev catch2 libfmt-dev liblz4-dev nlohmann-json3-dev libzstd-dev libssl-dev libavfilter-dev libavcodec-dev libswscale-dev pkg-config zlib1g-dev libva-dev libvdpau-dev qt6-tools-dev libzydis-dev zydis-tools libzycore-dev
+```
+
+* Ubuntu 22.04, Linux Mint 20, or Debian 12 or later is required.
+* To enable QT Web Engine, add `-DYUZU_USE_QT_WEB_ENGINE=ON` when running CMake.
+</details>
+
+<details>
+<summary>Fedora Linux</summary>
+
+```sh
+sudo dnf install autoconf ccache cmake fmt-devel gcc{,-c++} glslang hidapi-devel json-devel libtool libusb1-devel libzstd-devel lz4-devel nasm ninja-build openssl-devel pulseaudio-libs-devel qt6-linguist qt6-qtbase{-private,}-devel qt6-qtwebengine-devel qt6-qtmultimedia-devel speexdsp-devel wayland-devel zlib-devel ffmpeg-devel libXext-devel
+```
+
+* Force system libraries via CMake arguments:
+  * SDL2: `-DYUZU_USE_BUNDLED_SDL2=OFF -DYUZU_USE_EXTERNAL_SDL2=OFF`
+  * FFmpeg: `-DYUZU_USE_EXTERNAL_FFMPEG=OFF`
+* [RPM Fusion](https://rpmfusion.org/) is required for `ffmpeg-devel`
+* Fedora 32 or later is required.
+* Fedora 36+ users with GCC 12 need Clang and should configure CMake with:
+</details>
+
+<details>
+<summary>macOS</summary>
+
+Install dependencies from **[Homebrew](https://brew.sh/)**
+
+```sh
+brew install autoconf automake boost ffmpeg fmt glslang hidapi libtool libusb lz4 ninja nlohmann-json openssl pkg-config qt@6 sdl2 speexdsp zlib zstd cmake Catch2 molten-vk vulkan-loader spirv-tools
+```
+
+If you are compiling on Intel Mac, or are using a Rosetta Homebrew installation, you must replace all references of `/opt/homebrew` with `/usr/local`.
+
+To run with MoltenVK, install additional dependencies:
+```sh
+brew install molten-vk vulkan-loader
+```
+
+</details>
+
+<details>
+<summary>FreeBSD</summary>
+
+```
+devel/cmake
+devel/sdl20
+devel/boost-libs
+devel/catch2
+devel/libfmt
+devel/nlohmann-json
+devel/ninja
+devel/nasm
+devel/autoconf
+devel/pkgconf
+devel/qt6-base
+
+net/enet
+
+multimedia/ffnvcodec-headers
+multimedia/ffmpeg
+
+audio/opus
+
+archivers/liblz4
+
+lang/gcc12
+
+graphics/glslang
+graphics/vulkan-utility-libraries
+```
+
+If using FreeBSD 12 or prior, use `devel/pkg-config` instead.
+</details>
+
+<details>
+<summary>OpenBSD</summary>
+
+```sh
+pkg_add -u
+pkg_add cmake nasm git boost unzip--iconv autoconf-2.72p0 bash ffmpeg glslang gmake llvm-19.1.7p3 qt6 jq fmt nlohmann-json enet boost vulkan-utility-libraries vulkan-headers spirv-headers spirv-tools catch2 sdl2 libusb1.1.0.27
+```
+</details>
+
+<details>
+<summary>Solaris / OpenIndiana</summary>
+
+Always consult [the OpenIndiana package list](https://pkg.openindiana.org/hipster/en/index.shtml) to cross-verify availability.
+
+Run the usual update + install of essential toolings: `sudo pkg update && sudo pkg install git cmake`.
+
+- **gcc**: `sudo pkg install developer/gcc-14`.
+- **clang**: Version 20 is broken, use `sudo pkg install developer/clang-19`.
+
+Then install the libraries: `sudo pkg install qt6 boost glslang libzip library/lz4 nlohmann-json openssl opus sdl2 zlib compress/zstd unzip pkg-config nasm autoconf mesa library/libdrm header-drm developer/fmt`.
+</details>
+
+<details>
+<summary>MSYS2</summary>
+
+* Open the `MSYS2 MinGW 64-bit` shell (`mingw64.exe`)
+* Download and install all dependencies using:
+  * `pacman -Syu git make mingw-w64-x86_64-SDL2 mingw-w64-x86_64-cmake mingw-w64-x86_64-python-pip mingw-w64-x86_64-qt6 mingw-w64-x86_64-toolchain autoconf libtool automake-wrapper`
+* Add MinGW binaries to the PATH:
+  * `echo 'PATH=/mingw64/bin:$PATH' >> ~/.bashrc`
+* Add VulkanSDK to the PATH:
+  * `echo 'PATH=$(readlink -e /c/VulkanSDK/*/Bin/):$PATH' >> ~/.bashrc`
+</details>
+
+## All Done
+
+You may now return to the **[root build guide](Build.md)**.

docs/Development.md

@@ -1,23 +1,3 @@
-# Development
-
-* **Windows**: [Windows Building Guide](./build/Windows.md)
-* **Linux**: [Linux Building Guide](./build/Linux.md)
-* **Android**: [Android Building Guide](./build/Android.md)
-* **Solaris**: [Solaris Building Guide](./build/Solaris.md)
-* **FreeBSD**: [FreeBSD Building Guide](./build/FreeBSD.md)
-* **macOS**: [macOS Building Guide](./build/macOS.md)
-* **OpenBSD**: [OpenBSD Building Guide](./build/OpenBSD.md)
-
-# CPM
-
-CPM (CMake Package Manager) is the preferred method of managing dependencies within Eden. Documentation on adding dependencies/using CPMUtil is in the works.
-
-Notes:
-- `YUZU_USE_CPM` is set by default on MSVC and Android. Other platforms should use this if certain "required" system dependencies (e.g. OpenSSL) are broken or missing
-- `CPMUTIL_DEFAULT_SYSTEM` can be set to `OFF` to force the usage of bundled dependencies. This can marginally decrease the final package size.
-- When adding new prebuilt dependencies a la OpenSSL, SDL2, or FFmpeg, there *must* be a CMake option made available to forcefully download this bundle. See the OpenSSL implementation in the root CMakeLists for an example.
-  * This is necessary to allow for creation of fully-qualified source packs that allow for offline builds after download (some package managers and distros enforce this)
-	
 # Guidelines
 
 ## License Headers
@@ -37,6 +17,8 @@ FIX=true .ci/license-header.sh
 git commit --amend -a --no-edit
 ```
 
+If the work is licensed/vendored from other people or projects, you may omit the license headers. Additionally, if you wish to retain authorship over a piece of code, you may attribute it to yourself; however, the code may be changed at any given point and brought under the attribution of Eden.
+
 ## Pull Requests
 Pull requests are only to be merged by core developers when properly tested and discussions conclude on Discord or other communication channels. Labels are recommended but not required. However, all PRs MUST be namespaced and optionally typed:
 ```
@@ -49,7 +31,7 @@ Pull requests are only to be merged by core developers when properly tested and
 
 - The level of namespacing is generally left to the committer's choice.
 - However, we never recommend going more than two levels *except* in `hle`, in which case you may go as many as four levels depending on the specificity of your changes.
-- Ocassionally, up to two namespaces may be provided for more clarity.
+- Ocassionally, up to two additional namespaces may be provided for more clarity.
   * Changes that affect the entire project (sans CMake changes) should be namespaced as `meta`.
 - Maintainers are permitted to change namespaces at will.
 - Commits within PRs are not required to be namespaced, but it is highly recommended.

docs/Options.md

@@ -0,0 +1,69 @@
+# CMake Options
+
+To change these options, add `-DOPTION_NAME=NEWVALUE` to the command line.
+
+- On Qt Creator, go to Project -> Current Configuration
+
+Notes:
+- Defaults are marked per-platform.
+- "Non-UNIX" just means Windows/MSVC and Android (yes, macOS is UNIX
+- Android generally doesn't need to change anything; if you do, go to `src/android/app/build.gradle.kts`
+- To set a boolean variable to on, use `ON` for the value; to turn it off, use `OFF`
+- If a variable is mentioned as being e.g. "ON" for a specific platform(s), that means it is defaulted to OFF on others
+- TYPE is always boolean unless otherwise specified
+- Format:
+  * `OPTION_NAME` (TYPE DEFAULT) DESCRIPTION
+
+## Options
+
+- `YUZU_USE_CPM` (ON for non-UNIX) Use CPM to fetch system dependencies (fmt, boost, etc) if needed. Externals will still be fetched. See the [CPM](CPM.md) and [Deps](Deps.md) docs for more info.
+- `ENABLE_WEB_SERVICE` (ON) Enable multiplayer service
+- `ENABLE_WIFI_SCAN` (OFF) Enable WiFi scanning (requires iw on Linux) - experimental
+- `YUZU_USE_BUNDLED_FFMPEG` (ON for non-UNIX) Download (Windows, Android) or build (UNIX) bundled FFmpeg
+- `ENABLE_CUBEB` (ON) Enables the cubeb audio backend
+- `YUZU_TESTS` (ON) Compile tests - requires Catch2
+- `YUZU_USE_PRECOMPILED_HEADERS` (ON for non-UNIX) Use precompiled headers
+- `YUZU_DOWNLOAD_ANDROID_VVL` (ON) Download validation layer binary for Android
+- `YUZU_ENABLE_LTO` (OFF) Enable link-time optimization
+  * Not recommended on Windows
+  * UNIX may be better off appending `-flto=thin` to compiler args
+- `YUZU_DOWNLOAD_TIME_ZONE_DATA` (ON) Always download time zone binaries
+  * Currently, build fails without this
+- `YUZU_USE_FASTER_LD` (ON) Check if a faster linker is available
+  * Only available on UNIX
+- `USE_SYSTEM_MOLTENVK` (OFF, macOS only) Use the system MoltenVK lib (instead of the bundled one)
+- `YUZU_TZDB_PATH` (string) Path to a pre-downloaded timezone database (useful for nixOS)
+- `ENABLE_OPENSSL` (ON for Linux and *BSD) Enable OpenSSL backend for the ssl service
+  * Always enabled if the web service is enabled
+- `YUZU_USE_BUNDLED_OPENSSL` (ON for MSVC) Download bundled OpenSSL build
+  * Always on for Android
+  * Unavailable on OpenBSD
+
+The following options are desktop only:
+- `ENABLE_SDL2` (ON) Enable the SDL2 desktop, audio, and input frontend (HIGHLY RECOMMENDED!)
+  * Unavailable on Android
+- `YUZU_USE_EXTERNAL_SDL2` (ON for non-UNIX) Compiles SDL2 from source
+- `YUZU_USE_BUNDLED_SDL2` (ON for MSVC) Download a prebuilt SDL2
+  * Unavailable on OpenBSD
+  * Only enabled if YUZU_USE_CPM and ENABLE_SDL2 are both ON
+- `ENABLE_LIBUSB` (ON) Enable the use of the libusb input frontend (HIGHLY RECOMMENDED)
+- `ENABLE_OPENGL` (ON) Enable the OpenGL graphics frontend
+  * Unavailable on Windows/ARM64 and Android
+- `ENABLE_QT` (ON) Enable the Qt frontend (recommended)
+- `ENABLE_QT_TRANSLATION` (OFF) Enable translations for the Qt frontend
+- `ENABLE_QT_UPDATE_CHECKER` (OFF) Enable update checker for the Qt frontend
+- `YUZU_USE_BUNDLED_QT` (ON for MSVC) Download bundled Qt binaries
+  * Note that using **system Qt** requires you to include the Qt CMake directory in `CMAKE_PREFIX_PATH`, e.g:
+    * `-DCMAKE_PREFIX_PATH=C:/Qt/6.9.0/msvc2022_64/lib/cmake/Qt6`
+- `YUZU_QT_MIRROR` (string) What mirror to use for downloading the bundled Qt libraries
+- `YUZU_USE_QT_MULTIMEDIA` (OFF) Use QtMultimedia for camera support
+- `YUZU_USE_QT_WEB_ENGINE` (OFF) Use QtWebEngine for web applet implementation (requires the huge QtWebEngine dependency; not recommended)
+- `USE_DISCORD_PRESENCE` (OFF) Enables Discord Rich Presence (Qt frontend only)
+- `YUZU_ROOM` (ON) Enable dedicated room functionality
+- `YUZU_ROOM_STANDALONE` (ON) Enable standalone room executable (eden-room)
+  * Requires `YUZU_ROOM`
+- `YUZU_CMD` (ON) Compile the SDL2 frontend (eden-cli) - requires SDL2
+- `YUZU_CRASH_DUMPS` Compile crash dump (Minidump) support"
+  * Currently only available on Windows and Linux
+
+See `src/dynarmic/CMakeLists.txt` for additional options--usually, these don't need changed

docs/README.md

@@ -0,0 +1,10 @@
+# Eden Build Documentation
+
+This contains documentation created by developers. This contains build instructions, guidelines, instructions/layouts for [cool stuff we made](CPM.md), and more.
+
+- **[General Build Instructions](Build.md)**
+- **[Development Guidelines](Development.md)**
+- **[Dependencies](Deps.md)**
+- **[CPM - CMake Package Manager](CPM.md)**
+- **[Platform-Specific Caveats](Caveats.md)**
+- **[User Directory Handling](User.md)**

docs/build/Android.md

@@ -2,7 +2,7 @@
 
 ## Dependencies
 * [Android Studio](https://developer.android.com/studio)
-* [NDK 25.2.9519653 and CMake 3.22.1](https://developer.android.com/studio/projects/install-ndk#default-version)
+* [NDK 27+ and CMake 3.22.1](https://developer.android.com/studio/projects/install-ndk#default-version)
 * [Git](https://git-scm.com/download)
 
 ### WINDOWS ONLY - Additional Dependencies
@@ -16,8 +16,7 @@ git clone --recursive https://git.eden-emu.dev/eden-emu/eden.git
 ```
 Eden by default will be cloned into -
 * `C:\Users\<user-name>\eden` on Windows
-* `~/eden` on Linux
-* And wherever on macOS
+* `~/eden` on Linux and macOS
 
 ## Building
 1. Start Android Studio, on the startup dialog select `Open`.
@@ -32,7 +31,7 @@ Eden by default will be cloned into -
 `export ANDROID_SDK_ROOT=path/to/sdk`
 `export ANDROID_NDK_ROOT=path/to/ndk`.
 4. Navigate to `eden/src/android`.
-5. Then Build with `./gradlew assemblerelWithDebInfo`.
+5. Then Build with `./gradlew assembleRelWithDebInfo`.
 6. To build the optimised build use `./gradlew assembleGenshinSpoofRelWithDebInfo`.
 
 ### Script

docs/build/FreeBSD.md

@@ -1,81 +0,0 @@
-Eden is not currently available as a port on FreeBSD, though it is in the works. For now, the recommended method of usage is to compile it yourself. Check back often, as the build process frequently changes.
-
-## Dependencies.
-Eden needs the following dependencies:
-
-```
-devel/cmake
-devel/sdl20
-devel/boost-libs
-devel/catch2
-devel/libfmt
-devel/nlohmann-json
-devel/ninja
-devel/nasm
-devel/autoconf
-devel/pkgconf
-devel/qt6-base
-
-net/enet
-
-multimedia/ffnvcodec-headers
-multimedia/ffmpeg
-
-audio/opus
-
-archivers/liblz4
-
-lang/gcc12
-
-graphics/glslang
-graphics/vulkan-utility-libraries
-```
-
-If using FreeBSD 12 or prior, use `devel/pkg-config` instead.
-
----
-
-### Build preparations:
-Run the following command to clone eden with git:
-```sh
-git clone --recursive https://git.eden-emu.dev/eden-emu/eden
-```
-You usually want to add the `--recursive` parameter as it also takes care of the external dependencies for you.
-
-Now change into the eden directory and create a build directory there:
-```sh
-cd eden
-mkdir build
-```
-
-Change into that build directory:
-```sh
-cd build
-```
-
-#### 1. Building in Release Mode (usually preferred and the most performant choice):
-```sh
-cmake .. -GNinja -DYUZU_TESTS=OFF
-```
-
-#### 2. Building in Release Mode with debugging symbols (useful if you want to debug errors for a eventual fix):
-```sh
-cmake .. -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU_TESTS=ON
-```
-
-Build the emulator locally:
-```sh
-ninja
-```
-
-Optional: If you wish to install eden globally onto your system issue the following command:
-```sh
-sudo ninja install
-```
-OR
-```sh
-doas -- ninja install
-```
-
-## OpenSSL
-The available OpenSSL port (3.0.17) is out-of-date, and using a bundled static library instead is recommended; to do so, add `-DYUZU_USE_CPM=ON` to your CMake configure command.

docs/build/Linux.md

@@ -1,138 +0,0 @@
-### Dependencies
-
-You'll need to download and install the following to build Eden:
-
-  * [GCC](https://gcc.gnu.org/) v11+ (for C++20 support) & misc
-  * If GCC 12 is installed, [Clang](https://clang.llvm.org/) v14+ is required for compiling
-  * [CMake](https://www.cmake.org/) 3.22+
-
-The following are handled by Eden's externals:
-
-  * [FFmpeg](https://ffmpeg.org/)
-  * [SDL2](https://www.libsdl.org/download-2.0.php) 2.0.18+
-  * [opus](https://opus-codec.org/downloads/) 1.3+
-  
-All other dependencies will be downloaded and built by [CPM](https://github.com/cpm-cmake/CPM.cmake/) if `YUZU_USE_CPM` is on, but will always use system dependencies if available:
-
-  * [Boost](https://www.boost.org/users/download/) 1.79.0+
-  * [Catch2](https://github.com/catchorg/Catch2) 2.13.7 - 2.13.9
-  * [fmt](https://fmt.dev/) 8.0.1+
-  * [lz4](http://www.lz4.org) 1.8+
-  * [nlohmann_json](https://github.com/nlohmann/json) 3.8+
-  * [OpenSSL](https://www.openssl.org/source/) 1.1.1+
-  * [ZLIB](https://www.zlib.net/) 1.2+
-  * [zstd](https://facebook.github.io/zstd/) 1.5+
-  * [enet](http://enet.bespin.org/) 1.3+
-  * [cubeb](https://github.com/mozilla/cubeb)
-  * [SimpleIni](https://github.com/brofield/simpleini)
-
-Certain other dependencies (httplib, jwt, sirit, etc.) will be fetched by CPM regardless. System packages *can* be used for these libraries but this is generally not recommended.
-
-Dependencies are listed here as commands that can be copied/pasted. Of course, they should be inspected before being run.
-
-- Arch / Manjaro:
-  - `sudo pacman -Syu --needed base-devel boost catch2 cmake enet ffmpeg fmt git glslang libzip lz4 mbedtls ninja nlohmann-json openssl opus qt6-base qt6-multimedia sdl2 zlib zstd zip unzip`
-  - Building with QT Web Engine requires `qt6-webengine` as well.
-  - Proper wayland support requires `qt6-wayland`
-  - GCC 11 or later is required.
-  
-- Ubuntu / Linux Mint / Debian:
-  - `sudo apt-get install autoconf cmake g++ gcc git glslang-tools libasound2 libboost-context-dev libglu1-mesa-dev libhidapi-dev libpulse-dev libtool libudev-dev libxcb-icccm4 libxcb-image0 libxcb-keysyms1 libxcb-render-util0 libxcb-xinerama0 libxcb-xkb1 libxext-dev libxkbcommon-x11-0 mesa-common-dev nasm ninja-build qt6-base-private-dev libmbedtls-dev catch2 libfmt-dev liblz4-dev nlohmann-json3-dev libzstd-dev libssl-dev libavfilter-dev libavcodec-dev libswscale-dev pkg-config zlib1g-dev libva-dev libvdpau-dev`
-  - Ubuntu 22.04, Linux Mint 20, or Debian 12 or later is required.
-  - Users need to manually specify building with QT Web Engine enabled.  This is done using the parameter `-DYUZU_USE_QT_WEB_ENGINE=ON` when running CMake.
-  - Users need to manually disable building SDL2 from externals if they intend to use the version provided by their system by adding the parameters `-DYUZU_USE_EXTERNAL_SDL2=OFF`
-
-```sh
-git submodule update --init --recursive
-cmake .. -GNinja -DCMAKE_C_COMPILER=gcc-11 -DCMAKE_CXX_COMPILER=g++-11
-```
-
-- Fedora:
-  - `sudo dnf install autoconf ccache cmake fmt-devel gcc{,-c++} glslang hidapi-devel json-devel libtool libusb1-devel libzstd-devel lz4-devel nasm ninja-build openssl-devel pulseaudio-libs-devel qt6-linguist qt6-qtbase{-private,}-devel qt6-qtwebengine-devel qt6-qtmultimedia-devel speexdsp-devel wayland-devel zlib-devel ffmpeg-devel libXext-devel`
-  - Fedora 32 or later is required.
-  - Due to GCC 12, Fedora 36 or later users need to install `clang`, and configure CMake to use it via `-DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_COMPILER=clang`
-  - CMake arguments to force system libraries:
-    - SDL2: `-DYUZU_USE_BUNDLED_SDL2=OFF -DYUZU_USE_EXTERNAL_SDL2=OFF`
-    - FFmpeg: `-DYUZU_USE_EXTERNAL_FFMPEG=OFF`
-  - [RPM Fusion](https://rpmfusion.org/) (free) is required to install `ffmpeg-devel`
-
-### Cloning Eden with Git
-
-**Master:**
-
-```bash
-git clone --recursive https://git.eden-emu.dev/eden-emu/eden
-cd eden
-```
-
-The `--recursive` option automatically clones the required Git submodules.
-
-### Building Eden in Release Mode (Optimised)
-
-If you need to run ctests, you can disable `-DYUZU_TESTS=OFF` and install Catch2.
-
-```bash
-mkdir build && cd build
-cmake .. -GNinja -DYUZU_TESTS=OFF
-ninja
-sudo ninja install 
-```
-You may also want to include support for Discord Rich Presence by adding `-DUSE_DISCORD_PRESENCE=ON` after `cmake ..`
-
-`-DYUZU_USE_EXTERNAL_VULKAN_SPIRV_TOOLS=OFF` might be needed if ninja command failed with `undefined reference to symbol 'spvOptimizerOptionsCreate`, reason currently unknown
-
-Optionally, you can use `cmake-gui ..` to adjust various options (e.g. disable the Qt GUI).
-
-### Building Eden in Debug Mode (Slow)
-
-```bash
-mkdir build && cd build
-cmake .. -GNinja -DCMAKE_BUILD_TYPE=Debug -DYUZU_TESTS=OFF
-ninja
-```
-
-### Building with debug symbols
-
-```bash
-mkdir build && cd build
-cmake .. -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU -DYUZU_TESTS=OFF
-ninja
-```
-
-### Building with Scripts
-A convenience script for building is provided in `.ci/linux/build.sh`. You must provide an arch target for optimization, e.g. `.ci/linux/build.sh amd64`. Valid targets:
-- `legacy`: x86_64 generic, only needed for CPUs older than 2013 or so
-- `amd64`: x86_64-v3, for CPUs newer than 2013 or so
-- `steamdeck` / `zen2`: For Steam Deck or Zen >= 2 AMD CPUs (untested on Intel)
-- `rog-ally` / `allyx` / `zen4`: For ROG Ally X or Zen >= 4 AMD CPUs (untested on Intel)
-- `aarch64`: For armv8-a CPUs, older than mid-2021 or so
-- `armv9`: For armv9-a CPUs, newer than mid-2021 or so
-- `native`: Optimize to your native host architecture
-
-Extra flags to pass to CMake should be passed after the arch target.
-
-Additional environment variables can be used to control building:
-- `NPROC`: Number of threads to use for compilation (defaults to all)
-- `TARGET`: Set to `appimage` to disable standalone `eden-cli` and `eden-room` executables
-- `BUILD_TYPE`: Sets the build type to use. Defaults to `Release`
-
-The following environment variables are boolean flags. Set to `true` to enable or `false` to disable:
-- `DEVEL` (default FALSE): Disable Qt update checker
-- `USE_WEBENGINE` (default FALSE): Enable Qt WebEngine
-- `USE_MULTIMEDIA` (default TRUE): Enable Qt Multimedia
-
-After building, an AppImage can be packaged via `.ci/linux/package.sh`. This script takes the same arch targets as the build script. If the build was created in a different directory, you can specify its path relative to the source directory, e.g. `.ci/linux/package.sh amd64 build-appimage`. Additionally, set the `DEVEL` environment variable to `true` to change the app name to `Eden Nightly`.
-
-### Running without installing
-
-After building, the binaries `eden` and `eden-cmd` (depending on your build options) will end up in `build/bin/`.
-
-```bash
-# SDL
-cd build/bin/
-./eden-cmd
-
-# Qt
-cd build/bin/
-./eden
-```

docs/build/OpenBSD.md

@@ -1,10 +0,0 @@
-# Building for OpenBSD
-
-```sh
-pkg_add -u
-pkg_add cmake nasm git boost unzip--iconv autoconf-2.72p0 bash ffmpeg glslang gmake llvm-19.1.7p3 qt6 jq
-git --recursive https://git.eden-emu.dev/eden-emu/eden
-cmake -DCMAKE_C_COMPILER=clang-19 -DCMAKE_CXX_COMPILER=clang++-19 -DDYNARMIC_USE_PRECOMPILED_HEADERS=OFF -DCMAKE_BUILD_TYPE=Debug -DENABLE_QT=OFF -DENABLE_OPENSSL=OFF -DENABLE_WEB_SERVICE=OFF -B /usr/obj/eden
-```
-
-- Modify `externals/ffmpeg/CMakeFiles/ffmpeg-build/build.make` to use `-j$(nproc)` instead of just `-j`.

docs/build/Solaris.md

@@ -1,51 +0,0 @@
-# Building for Solaris
-
-## Dependencies.  
-Always consult [the OpenIndiana package list](https://pkg.openindiana.org/hipster/en/index.shtml) to cross-verify availability.
-
-Run the usual update + install of essential toolings: `sudo pkg update && sudo pkg install git cmake`.
-
-- **gcc**: `sudo pkg install developer/gcc-14`.
-- **clang**: Version 20 is broken, use `sudo pkg install developer/clang-19`.
-
-Then install the libraies: `sudo pkg install qt6 boost glslang libzip library/lz4 nlohmann-json openssl opus sdl2 zlib compress/zstd unzip pkg-config nasm autoconf mesa library/libdrm header-drm developer/fmt`.
-
-### Building
-
-Clone eden with git `git clone --recursive https://git.eden-emu.dev/eden-emu/eden`
-
-```sh
-# Needed for some dependencies that call cc directly (tz)
-echo '#!/bin/sh' >cc
-echo 'gcc $@' >>cc
-chmod +x cc
-export PATH="$PATH:$PWD"
-```
-
-Patch for FFmpeg:
-```sh
-sed -i 's/ make / gmake /' externals/ffmpeg/CMakeFiles/ffmpeg-build.dir/build.make
-```
-
-- **Configure**: `cmake -B build -DYUZU_USE_CPM=ON -DCMAKE_CXX_FLAGS="-I/usr/include/SDL2" -DCMAKE_C_FLAGS="-I/usr/include/SDL2"`.
-- **Build**: `cmake --build build`.
-- **Installing**: `sudo cmake --install build`.
-
-### Running
-
-Default Mesa is a bit outdated, the following environment variables should be set for a smoother experience:
-```sh
-export MESA_GL_VERSION_OVERRIDE=4.6
-export MESA_GLSL_VERSION_OVERRIDE=460
-export MESA_EXTENSION_MAX_YEAR=2025
-export MESA_DEBUG=1
-export MESA_VK_VERSION_OVERRIDE=1.3
-# Only if nvidia/intel drm drivers cause crashes, will severely hinder performance
-export LIBGL_ALWAYS_SOFTWARE=1
-```
-
-### Notes
-
-- Modify the generated ffmpeg.make (in build dir) if using multiple threads (base system `make` doesn't use `-j4`, so change for `gmake`).
-- If using OpenIndiana, due to a bug in SDL2 cmake configuration; Audio driver defaults to SunOS `<sys/audioio.h>`, which does not exist on OpenIndiana.
-- System OpenSSL generally does not work. Instead, use `-DYUZU_USE_CPM=ON` to use a bundled static OpenSSL, or build a system dependency from source.

docs/build/Windows.md

@@ -1,259 +0,0 @@
-# ⚠️ This guide is for developers ONLY! Support will be provided to developers ONLY.
-
-## 📋 Current building methods:
-
-* [ Minimal Dependencies](#minimal-dependencies)
-* [⚡ Method I: MSVC Build for Windows](#method-i-msvc-build-for-windows)
-* [🐧 Method II: MinGW-w64 Build with MSYS2](#method-ii-mingw-w64-build-with-msys2)
-* [🖥️ Method III: CLion Environment Setup](#method-iii-clion-environment-setup)
-* [💻 Building from the command line with MSVC](#building-from-the-command-line-with-msvc)
-* [📜 Building with Scripts](#building-with-scripts)
-
----
-
-
-## Minimal Dependencies
-
-On Windows, **all** library dependencies are **automatically included** within the `externals` folder.
-
-You still need to install:
-
-* **[CMake](https://cmake.org/download/)** - Used to generate Visual Studio project files.
-* **[Vulkan SDK](https://vulkan.lunarg.com/sdk/home#windows)** - Make sure to select **Latest SDK**.
-
-  * *A convenience script to install the latest SDK is provided in `.ci/windows/install-vulkan-sdk.ps1`*
-* **[Git for Windows](https://gitforwindows.org)** - We recommend installing Git for command line use and version control integration.
-
-  <img src="https://i.imgur.com/x0rRs1t.png" width="500">
-
-  * *While installing Git Bash, select "Git from the command line and also from 3rd-party software". If missed, manually set `git.exe` path in CMake.*
-
----
-
-## ⚡ Method I: MSVC Build for Windows
-
-### a. Prerequisites to MSVC Build
-
-* **[Visual Studio 2022 Community](https://visualstudio.microsoft.com/downloads/)** - Make sure to **select C++ support** in the installer, or **update to the latest version** if already installed.
-
-  * *A convenience script to install the **minimal** version (Visual Build Tools) is provided in `.ci/windows/install-msvc.ps1`*
-
----
-
-### b. Clone the eden repository with Git
-
-Open Terminal on 
-
-```cmd
-git clone https://git.eden-emu.dev/eden-emu/eden
-cd eden
-```
-
-* *By default `eden` downloads to `C:\Users\<user-name>\eden`*
-
----
-
-### c. Building
-
-* Open the CMake GUI application and point it to the `eden`
-
-  <img src="https://i.imgur.com/qOslIWv.png" width="500">
-
-* For the build directory, use a `build/` subdirectory inside the source directory or some other directory of your choice. (Tell CMake to create it.)
-
-* Click the "Configure" button and choose `Visual Studio 17 2022`, with `x64` for the optional platform.
-
-  <img src="https://i.imgur.com/DKiREaK.png" width="500">
-
-  * *(You may also want to disable `YUZU_TESTS` in this case since Catch2 is not yet supported with this.)*
-
-  <img src="https://user-images.githubusercontent.com/22451773/180585999-07316d6e-9751-4d11-b957-1cf57cd7cd58.png" width="500">
-
-* Click "Generate" to create the project files.
-
-  <img src="https://i.imgur.com/5LKg92k.png" width="500">
-
-* Open the solution file `yuzu.sln` in Visual Studio 2022, which is located in the build folder.
-
-  <img src="https://i.imgur.com/208yMml.png" width="500">
-
-* * Depending on whether you want a graphical user interface or not, select in the Solution Explorer:
-    * `eden` (GUI)
-    * `eden-cmd` (command-line only)
-  * Then **right-click** and choose `Set as StartUp Project`.
-
-  <img src="https://i.imgur.com/nPMajnn.png" height="500">
-  <img src="https://i.imgur.com/BDMLzRZ.png" height="500">
-
-* Select the appropriate build type, `Debug` for debug purposes or `Release` for performance (in case of doubt choose `Release`).
-
-  <img src="https://i.imgur.com/qxg4roC.png" width="500">
-
-* **Right-click** the project you want to build and press **Build** in the submenu or press `F5`.
-
-  <img src="https://i.imgur.com/CkQgOFW.png" height="500">
-
----
-
-## 🐧 Method II: MinGW-w64 Build with MSYS2
-
-### a. Prerequisites to MinGW-w64
-
-* **[MSYS2](https://www.msys2.org)** - A versatile and up-to-date development environment for Windows, providing a Unix-like shell, package manager, and toolchain.
-
----
-
-### b. Install eden dependencies for MinGW-w64
-
-* Open the `MSYS2 MinGW 64-bit` shell (`mingw64.exe`)
-* Download and install all dependencies using:
-  * `pacman -Syu git make mingw-w64-x86_64-SDL2 mingw-w64-x86_64-cmake mingw-w64-x86_64-python-pip mingw-w64-x86_64-qt6 mingw-w64-x86_64-toolchain autoconf libtool automake-wrapper`
-* Add MinGW binaries to the PATH:
-  * `echo 'PATH=/mingw64/bin:$PATH' >> ~/.bashrc`
-* Add VulkanSDK to the PATH:
-  * `echo 'PATH=$(readlink -e /c/VulkanSDK/*/Bin/):$PATH' >> ~/.bashrc`
-
----
-
-### c. Clone the eden repository with Git
-
-```cmd
-git clone https://git.eden-emu.dev/eden-emu/eden
-cd eden
-```
-
----
-
-### d. Building dynamically-linked eden
-
-* This process will generate a *dynamically* linked build
-
-```bash
-# Make build dir and enter
-mkdir build && cd build
-
-# Generate CMake Makefiles
-cmake .. -G "MSYS Makefiles" -DYUZU_TESTS=OFF
-
-# Build
-make -j$(nproc)
-
-# Run eden!
-./bin/eden.exe
-```
-
-* *Warning: This build is not a **static** build meaning that you **need** to include all of the DLLs with the .exe in order to use it or other systems!*
-
----
-
-### Additional notes
-
-
-* Eden doesn't require the rather large Qt dependency, but you will lack a GUI frontend
-
-```bash
-# ...
-
-# Generate CMake Makefiles (withou QT)
-cmake .. -G "MSYS Makefiles" -DYUZU_TESTS=OFF -DENABLE_QT=no
-
-$ ...
-```
-
-* Running programs from inside `MSYS2 MinGW x64` shell has a different `%PATH%` than directly from explorer.
-  * This different `%PATH%` has the locations of the other DLLs required.
-
-    <img src="https://user-images.githubusercontent.com/190571/165000848-005e8428-8a82-41b1-bb4d-4ce7797cdac8.png" width="500">
-
----
-
-## 🖥️ Method III: CLion Environment Setup
-
-### a. Prerequisites to CLion
-
-* [CLion](https://www.jetbrains.com/clion/) - This IDE is not free; for a free alternative, check Method I
-
----
-
-### b. Cloning eden with CLion
-
-* Clone the Repository:
-
-<img src="https://user-images.githubusercontent.com/42481638/216899046-0d41d7d6-8e4d-4ed2-9587-b57088af5214.png" width="500">
-<img src="https://user-images.githubusercontent.com/42481638/216899061-b2ea274a-e88c-40ae-bf0b-4450b46e9fea.png" width="500">
-<img src="https://user-images.githubusercontent.com/42481638/216899076-0e5988c4-d431-4284-a5ff-9ecff973db76.png" width="500">
-
----
-
-### c. Building & Setup
-
-* Once Cloned, You will be taken to a prompt like the image below:
-
-<img src="https://user-images.githubusercontent.com/42481638/216899092-3fe4cec6-a540-44e3-9e1e-3de9c2fffc2f.png" width="500">
-
-* Set the settings to the image below:
-* Change `Build type: Release`
-* Change `Name: Release`
-* Change `Toolchain Visual Studio`
-* Change `Generator: Let CMake decide`
-* Change `Build directory: build`
-
-<img src="https://user-images.githubusercontent.com/42481638/216899164-6cee8482-3d59-428f-b1bc-e6dc793c9b20.png" width="500">
-
-* Click OK; now Clion will build a directory and index your code to allow for IntelliSense. Please be patient.
-* Once this process has been completed (No loading bar bottom right), you can now build eden
-* In the top right, click on the drop-down menu, select all configurations, then select eden
-
-<img src="https://user-images.githubusercontent.com/42481638/216899226-975048e9-bc6d-4ec1-bc2d-bd8a1e15ed04.png" height="500" >
-
-* Now run by clicking the play button or pressing Shift+F10, and eden will auto-launch once built.
-
-<img src="https://user-images.githubusercontent.com/42481638/216899275-d514ec6a-e563-470e-81e2-3e04f0429b68.png" width="500">
-
----
-
-## 💻 Building from the command line with MSVC
-
-```cmd
-# Clone eden and enter
-git clone https://git.eden-emu.dev/eden-emu/eden
-cd eden
-
-# Make build dir and enter
-mkdir build && cd build
-
-# Generate CMake Makefiles
-cmake .. -G "Visual Studio 17 2022" -A x64 -DYUZU_TESTS=OFF
-
-# Build
-cmake --build . --config Release
-```
-
-## 📜 Building with Scripts
-
-* A convenience script for building is provided in `.ci/windows/build.sh`.
-* You must run this with Bash, e.g. Git Bash or MinGW TTY.
-* To use this script, you must have `windeployqt` installed (usually bundled with Qt) and set the `WINDEPLOYQT` environment variable to its canonical Bash location:
-  * `WINDEPLOYQT="/c/Qt/6.9.1/msvc2022_64/bin/windeployqt6.exe" .ci/windows/build.sh`.
-* You can use `aqtinstall`, more info on <https://github.com/miurahr/aqtinstall> and <https://ddalcino.github.io/aqt-list-server/>
-
-
-* Extra CMake flags should be placed in the arguments of the script.
-
-#### Additional environment variables can be used to control building:
-
-* `BUILD_TYPE` (default `Release`): Sets the build type to use.
-
-* The following environment variables are boolean flags. Set to `true` to enable or `false` to disable:
-
-  * `DEVEL` (default FALSE): Disable Qt update checker
-  * `USE_WEBENGINE` (default FALSE): Enable Qt WebEngine
-  * `USE_MULTIMEDIA` (default TRUE): Enable Qt Multimedia
-  * `BUNDLE_QT` (default FALSE): Use bundled Qt
-
-  * Note that using **system Qt** requires you to include the Qt CMake directory in `CMAKE_PREFIX_PATH`
-    * `.ci/windows/build.sh -DCMAKE_PREFIX_PATH=C:/Qt/6.9.0/msvc2022_64/lib/cmake/Qt6`
-
-* After building, a zip can be packaged via `.ci/windows/package.sh`. You must have 7-zip installed and in your PATH.
-  * The resulting zip will be placed into `artifacts` in the source directory.
-

docs/build/macOS.md

@@ -1,78 +0,0 @@
-Please note this article is intended for development, and Eden on macOS is not currently ready for regular use.
-
-This article was written for developers. Eden support for macOS is not ready for casual use.
-
-## Dependencies
-Install dependencies from Homebrew:
-```sh
-brew install autoconf automake boost ffmpeg fmt glslang hidapi libtool libusb lz4 ninja nlohmann-json openssl pkg-config qt@6 sdl2 speexdsp zlib zstd cmake Catch2 molten-vk vulkan-loader spirv-tools
-```
-
-If you are compiling on Intel Mac, or are using a Rosetta Homebrew installation, you must replace all references of `/opt/homebrew` with `/usr/local`.
-
-Now, clone the repo:
-```sh
-git clone --recursive https://git.eden-emu.dev/eden-emu/eden
-cd eden
-```
-
-## Method I: ninja
-
----
-Build for release
-```sh
-export Qt6_DIR="/opt/homebrew/opt/qt@6/lib/cmake"
-export LIBVULKAN_PATH=/opt/homebrew/lib/libvulkan.dylib
-cmake -B build -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU_TESTS=OFF -DENABLE_WEB_SERVICE=ON -DENABLE_LIBUSB=OFF -DCLANG_FORMAT=ON -DSDL2_DISABLE_INSTALL=ON -DSDL_ALTIVEC=ON
-ninja
-```
-
-You may also want to include support for Discord Rich Presence by adding `-DUSE_DISCORD_PRESENCE=ON`
-```sh
-export Qt6_DIR="/opt/homebrew/opt/qt@6/lib/cmake"
-cmake -B build -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU_TESTS=OFF -DENABLE_WEB_SERVICE=OFF -DENABLE_LIBUSB=OFF
-ninja
-```
-
-Run the output:
-```
-bin/eden.app/Contents/MacOS/eden
-```
-
-## Method II: Xcode
-
----
-Build for release
-```sh
-export Qt6_DIR="/opt/homebrew/opt/qt@6/lib/cmake"
-export LIBVULKAN_PATH=/opt/homebrew/lib/libvulkan.dylib
-# Only if having errors about Xcode 15.0
-sudo /usr/bin/xcode-select --switch /Users/admin/Downloads/Xcode.ap
-cmake -B build -GXcode -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU_TESTS=OFF -DENABLE_WEB_SERVICE=ON -DENABLE_LIBUSB=OFF -DCLANG_FORMAT=ON -DSDL2_DISABLE_INSTALL=ON -DSDL_ALTIVEC=ON
-xcodebuild build -project yuzu.xcodeproj -scheme "yuzu" -configuration "RelWithDebInfo"
-```
-
-Build with debug symbols:
-```sh
-export Qt6_DIR="/opt/homebrew/opt/qt@6/lib/cmake"
-cmake -B build -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo -DYUZU_TESTS=OFF -DENABLE_WEB_SERVICE=OFF -DENABLE_LIBUSB=OFF
-ninja
-```
-
-Run the output:
-```
-bin/eden.app/Contents/MacOS/eden
-```
-
----
-
-To run with MoltenVK, install additional dependencies:
-```sh
-brew install molten-vk vulkan-loader
-```
-
-Run with Vulkan loader path:
-```sh
-export LIBVULKAN_PATH=/opt/homebrew/lib/libvulkan.dylib
-bin/eden.app/Contents/MacOS/eden
-```

docs/scripts/Linux.md

@@ -0,0 +1,31 @@
+# Linux Build Scripts
+
+* Provided script: `.ci/linux/build.sh`
+* Must specify arch target, e.g.: `.ci/linux/build.sh amd64`
+* Valid targets:
+  * `native`: Optimize to your native host architecture
+  * `legacy`: x86\_64 generic, only needed for CPUs older than 2013 or so
+  * `amd64`: x86\_64-v3, for CPUs newer than 2013 or so
+  * `steamdeck` / `zen2`: For Steam Deck or Zen >= 2 AMD CPUs (untested on Intel)
+  * `rog-ally` / `allyx` / `zen4`: For ROG Ally X or Zen >= 4 AMD CPUs (untested on Intel)
+  * `aarch64`: For armv8-a CPUs, older than mid-2021 or so
+  * `armv9`: For armv9-a CPUs, newer than mid-2021 or so
+* Extra CMake flags go after the arch target.
+
+### Environment Variables
+
+* `NPROC`: Number of compilation threads (default: all cores)
+* `TARGET`: Set `appimage` to disable standalone `eden-cli` and `eden-room`
+* `BUILD_TYPE`: Build type (default: `Release`)
+
+Boolean flags (set `true` to enable, `false` to disable):
+
+* `DEVEL` (default `FALSE`): Disable Qt update checker
+* `USE_WEBENGINE` (default `FALSE`): Enable Qt WebEngine
+* `USE_MULTIMEDIA` (default `FALSE`): Enable Qt Multimedia
+
+* AppImage packaging script: `.ci/linux/package.sh`
+
+  * Accepts same arch targets as build script
+  * Use `DEVEL=true` to rename app to `Eden Nightly`
+  * This should generally not be used unless in a tailor-made packaging environment (e.g. Actions/CI)

docs/scripts/Windows.md

@@ -0,0 +1,29 @@
+# Windows Build Scripts
+
+* A convenience script for building is provided in `.ci/windows/build.sh`.
+* You must run this with Bash, e.g. Git Bash or the MinGW TTY.
+* To use this script, you must have `windeployqt` installed (usually bundled with Qt) and set the `WINDEPLOYQT` environment variable to its canonical Bash location:
+  * `WINDEPLOYQT="/c/Qt/6.9.1/msvc2022_64/bin/windeployqt6.exe" .ci/windows/build.sh`.
+* You can use `aqtinstall`, more info on <https://github.com/miurahr/aqtinstall> and <https://ddalcino.github.io/aqt-list-server/>
+
+
+* Extra CMake flags should be placed in the arguments of the script.
+
+#### Additional environment variables can be used to control building:
+
+* `BUILD_TYPE` (default `Release`): Sets the build type to use.
+
+* The following environment variables are boolean flags. Set to `true` to enable or `false` to disable:
+
+  * `DEVEL` (default FALSE): Disable Qt update checker
+  * `USE_WEBENGINE` (default FALSE): Enable Qt WebEngine
+  * `USE_MULTIMEDIA` (default FALSE): Enable Qt Multimedia
+  * `BUNDLE_QT` (default FALSE): Use bundled Qt
+
+  * Note that using **system Qt** requires you to include the Qt CMake directory in `CMAKE_PREFIX_PATH`
+    * `.ci/windows/build.sh -DCMAKE_PREFIX_PATH=C:/Qt/6.9.0/msvc2022_64/lib/cmake/Qt6`
+
+* After building, a zip can be packaged via `.ci/windows/package.sh`. You must have 7-zip installed and in your PATH.
+  * The resulting zip will be placed into `artifacts` in the source directory.
+
+

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -210,12 +210,8 @@ std::shared_ptr<Dynarmic::A32::Jit> ArmDynarmic32::MakeJit(Common::PageTable* pa
     config.wall_clock_cntpct = m_uses_wall_clock;
     config.enable_cycle_counting = !m_uses_wall_clock;
 
-    // Code cache size
-#ifdef ARCHITECTURE_arm64
+    // Code cache size - max in ARM is 128MiB, max in x86_64 is 2GiB
     config.code_cache_size = std::uint32_t(128_MiB);
-#else
-    config.code_cache_size = std::uint32_t(512_MiB);
-#endif
 
     // Allow memory fault handling to work
     if (m_system.DebuggerEnabled()) {

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -269,12 +269,8 @@ std::shared_ptr<Dynarmic::A64::Jit> ArmDynarmic64::MakeJit(Common::PageTable* pa
     config.wall_clock_cntpct = m_uses_wall_clock;
     config.enable_cycle_counting = !m_uses_wall_clock;
 
-    // Code cache size
-#ifdef ARCHITECTURE_arm64
+    // Code cache size - max in ARM is 128MiB, max in x86_64 is 2GiB
     config.code_cache_size = std::uint32_t(128_MiB);
-#else
-    config.code_cache_size = std::uint32_t(512_MiB);
-#endif
 
     // Allow memory fault handling to work
     if (m_system.DebuggerEnabled()) {

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp

@@ -256,61 +256,103 @@ NvResult nvhost_ctrl_gpu::ZCullGetInfo(IoctlNvgpuGpuZcullGetInfoArgs& params) {
 }
 
 NvResult nvhost_ctrl_gpu::ZBCSetTable(IoctlZbcSetTable& params) {
-    LOG_DEBUG(Service_NVDRV, "called");
-    ZbcEntry entry = {};
-    std::memset(&entry, 0, sizeof(entry));
-    // TODO(ogniK): What does this even actually do?
-    // TODO(myself): This thing I guess
-    if (params.type == 1) {
-        for (auto i = 0; i < 4; ++i) {
-            entry.color_ds[i] = params.color_ds[i];
-            entry.color_l2[i] = params.color_l2[i];
-        }
-        ASSERT(this->max_color_entries < 16);
-        this->color_entries[this->max_color_entries] = entry;
-        ++this->max_color_entries;
-    } else if (params.type == 2) {
-        entry.depth = params.depth;
-        ASSERT(this->max_depth_entries < 16);
-        this->depth_entries[this->max_depth_entries] = entry;
-        ++this->max_depth_entries;
+    if (params.type > supported_types) {
+        LOG_ERROR(Service_NVDRV, "ZBCSetTable: invalid type {:#X}", params.type);
+        return NvResult::BadParameter;
     }
+
+    std::scoped_lock lk(zbc_mutex);
+
+    switch (static_cast<ZBCTypes>(params.type)) {
+        case ZBCTypes::color: {
+            ZbcColorEntry color_entry{};
+            std::copy_n(std::begin(params.color_ds), color_entry.color_ds.size(), color_entry.color_ds.begin());
+            std::copy_n(std::begin(params.color_l2), color_entry.color_l2.size(), color_entry.color_l2.begin());
+            color_entry.format = params.format;
+            color_entry.ref_cnt = 1u;
+
+            auto color_it = std::ranges::find_if(zbc_colors,
+                                                 [&](const ZbcColorEntry& color_in_question) {
+                                                     return color_entry.format == color_in_question.format &&
+                                                            color_entry.color_ds == color_in_question.color_ds &&
+                                                            color_entry.color_l2 == color_in_question.color_l2;
+                                                 });
+
+            if (color_it != zbc_colors.end()) {
+                ++color_it->ref_cnt;
+                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: reused color entry fmt={:#X}, ref_cnt={:#X}",
+                          params.format, color_it->ref_cnt);
+            } else {
+                zbc_colors.push_back(color_entry);
+                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added color entry fmt={:#X}, index={:#X}",
+                          params.format, zbc_colors.size() - 1);
+            }
+            break;
+        }
+        case ZBCTypes::depth: {
+            ZbcDepthEntry depth_entry{params.depth, params.format, 1u};
+
+            auto depth_it = std::ranges::find_if(zbc_depths,
+                                                 [&](const ZbcDepthEntry& depth_entry_in_question) {
+                                                     return depth_entry.format == depth_entry_in_question.format &&
+                                                            depth_entry.depth == depth_entry_in_question.depth;
+                                                 });
+
+            if (depth_it != zbc_depths.end()) {
+                ++depth_it->ref_cnt;
+                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: reused depth entry fmt={:#X}, ref_cnt={:#X}",
+                          depth_entry.format, depth_it->ref_cnt);
+            } else {
+                zbc_depths.push_back(depth_entry);
+                LOG_DEBUG(Service_NVDRV, "ZBCSetTable: added depth entry fmt={:#X}, index={:#X}",
+                          depth_entry.format, zbc_depths.size() - 1);
+            }
+        }
+    }
+
     return NvResult::Success;
 }
 
 NvResult nvhost_ctrl_gpu::ZBCQueryTable(IoctlZbcQueryTable& params) {
-    LOG_DEBUG(Service_NVDRV, "called");
-    struct ZbcQueryParams {
-        u32_le color_ds[4];
-        u32_le color_l2[4];
-        u32_le depth;
-        u32_le ref_cnt;
-        u32_le format;
-        u32_le type;
-        u32_le index_size;
-    } entry = {};
-    std::memset(&entry, 0, sizeof(entry));
-    auto const index = params.index_size;
-    if (params.type == 0) { //no
-        entry.index_size = 15;
-    } else if (params.type == 1) { //color
-        ASSERT(index < 16);
-        for (auto i = 0; i < 4; ++i) {
-            params.color_ds[i] = this->color_entries[index].color_ds[i];
-            params.color_l2[i] = this->color_entries[index].color_l2[i];
-        }
-        // TODO: Only if no error thrown (otherwise dont modify)
-        params.format = this->color_entries[index].format;
-        //params.ref_cnt = this->color_entries[index].ref_cnt;
-    } else if (params.type == 2) { //depth
-        ASSERT(index < 16);
-        params.depth = this->depth_entries[index].depth;
-        // TODO: Only if no error thrown (otherwise dont modify)
-        params.format = this->depth_entries[index].format;
-        //params.ref_cnt = this->depth_entries[index].ref_cnt;
-    } else {
-        UNREACHABLE();
+    if (params.type > supported_types) {
+        LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid type {:#X}", params.type);
+        return NvResult::BadParameter;
     }
+
+    std::scoped_lock lk(zbc_mutex);
+
+    switch (static_cast<ZBCTypes>(params.type)) {
+        case ZBCTypes::color: {
+            if (params.index_size >= zbc_colors.size()) {
+                LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid color index {:#X}", params.index_size);
+                return NvResult::BadParameter;
+            }
+
+            const auto& colors = zbc_colors[params.index_size];
+            std::copy_n(colors.color_ds.begin(), colors.color_ds.size(), std::begin(params.color_ds));
+            std::copy_n(colors.color_l2.begin(), colors.color_l2.size(), std::begin(params.color_l2));
+            params.depth = 0;
+            params.ref_cnt = colors.ref_cnt;
+            params.format = colors.format;
+            params.index_size = static_cast<u32>(zbc_colors.size());
+            break;
+        }
+        case ZBCTypes::depth: {
+            if (params.index_size >= zbc_depths.size()) {
+                LOG_ERROR(Service_NVDRV, "ZBCQueryTable: invalid depth index {:#X}", params.index_size);
+                return NvResult::BadParameter;
+            }
+
+            const auto& depth_entry = zbc_depths[params.index_size];
+            std::fill(std::begin(params.color_ds), std::end(params.color_ds), 0);
+            std::fill(std::begin(params.color_l2), std::end(params.color_l2), 0);
+            params.depth = depth_entry.depth;
+            params.ref_cnt = depth_entry.ref_cnt;
+            params.format = depth_entry.format;
+            params.index_size = static_cast<u32>(zbc_depths.size());
+        }
+    }
+
     return NvResult::Success;
 }
 

src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -34,6 +37,11 @@ public:
     Kernel::KEvent* QueryEvent(u32 event_id) override;
 
 private:
+    enum class ZBCTypes {
+        color = 1,
+        depth = 2,
+    };
+
     struct IoctlGpuCharacteristics {
         u32_le arch;                       // 0x120 (NVGPU_GPU_ARCH_GM200)
         u32_le impl;                       // 0xB (NVGPU_GPU_IMPL_GM20B)
@@ -139,6 +147,21 @@ private:
     };
     static_assert(sizeof(IoctlZbcQueryTable) == 52, "IoctlZbcQueryTable is incorrect size");
 
+    struct ZbcColorEntry {
+        std::array<u32, 4> color_ds{};
+        std::array<u32, 4> color_l2{};
+        u32 format{};
+        u32 ref_cnt{};
+    };
+    static_assert(sizeof(ZbcColorEntry) == 40, "ZbcColorEntry is incorrect size");
+
+    struct ZbcDepthEntry {
+        u32 depth{};
+        u32 format{};
+        u32 ref_cnt{};
+    };
+    static_assert(sizeof(ZbcDepthEntry) == 12, "ZbcDepthEntry is incorrect size");
+
     struct IoctlFlushL2 {
         u32_le flush; // l2_flush | l2_invalidate << 1 | fb_flush << 2
         u32_le reserved;
@@ -182,17 +205,11 @@ private:
     Kernel::KEvent* error_notifier_event;
     Kernel::KEvent* unknown_event;
 
-    struct ZbcEntry {
-        u32_le color_ds[4];
-        u32_le color_l2[4];
-        u32_le depth;
-        u32_le type;
-        u32_le format;
-    };
-    std::array<ZbcEntry, 16> color_entries;
-    std::array<ZbcEntry, 16> depth_entries;
-    u8 max_color_entries;
-    u8 max_depth_entries;
+    // ZBC Tables
+    std::mutex zbc_mutex{};
+    std::vector<ZbcColorEntry> zbc_colors{};
+    std::vector<ZbcDepthEntry> zbc_depths{};
+    const u32 supported_types = 2u;
 };
 
 } // namespace Service::Nvidia::Devices

src/dynarmic/CMakeLists.txt

@@ -39,9 +39,6 @@ option(DYNARMIC_INSTALL "Install dynarmic headers and CMake files" OFF)
 option(DYNARMIC_USE_BUNDLED_EXTERNALS "Use all bundled externals (useful when e.g. cross-compiling)" OFF)
 option(DYNARMIC_WARNINGS_AS_ERRORS "Warnings as errors" ${MASTER_PROJECT})
 option(DYNARMIC_ENABLE_LTO "Enable LTO" OFF)
-if (NOT DEFINED DYNARMIC_FRONTENDS)
-    set(DYNARMIC_FRONTENDS "A32;A64" CACHE STRING "Selects which frontends to enable")
-endif()
 
 # Default to a Release build
 if (NOT CMAKE_BUILD_TYPE)

src/dynarmic/docs/Design.md

@@ -273,52 +273,73 @@ Exclusive OR (i.e.: XOR)
 
 ### Callback: {Read,Write}Memory{8,16,32,64}
 
-    <u8> ReadMemory8(<u32> vaddr)
-    <u8> ReadMemory16(<u32> vaddr)
-    <u8> ReadMemory32(<u32> vaddr)
-    <u8> ReadMemory64(<u32> vaddr)
-    <void> WriteMemory8(<u32> vaddr, <u8> value_to_store)
-    <void> WriteMemory16(<u32> vaddr, <u16> value_to_store)
-    <void> WriteMemory32(<u32> vaddr, <u32> value_to_store)
-    <void> WriteMemory64(<u32> vaddr, <u64> value_to_store)
+```c++
+<u8> ReadMemory8(<u32> vaddr)
+<u8> ReadMemory16(<u32> vaddr)
+<u8> ReadMemory32(<u32> vaddr)
+<u8> ReadMemory64(<u32> vaddr)
+<void> WriteMemory8(<u32> vaddr, <u8> value_to_store)
+<void> WriteMemory16(<u32> vaddr, <u16> value_to_store)
+<void> WriteMemory32(<u32> vaddr, <u32> value_to_store)
+<void> WriteMemory64(<u32> vaddr, <u64> value_to_store)
+```
 
 Memory access.
 
 ### Terminal: Interpret
 
-    SetTerm(IR::Term::Interpret{next})
+```c++
+SetTerm(IR::Term::Interpret{next})
+```
 
 This terminal instruction calls the interpreter, starting at `next`.
 The interpreter must interpret exactly one instruction.
 
 ### Terminal: ReturnToDispatch
 
-    SetTerm(IR::Term::ReturnToDispatch{})
+```c++
+SetTerm(IR::Term::ReturnToDispatch{})
+```
 
 This terminal instruction returns control to the dispatcher.
 The dispatcher will use the value in R15 to determine what comes next.
 
 ### Terminal: LinkBlock
 
-    SetTerm(IR::Term::LinkBlock{next})
+```c++
+SetTerm(IR::Term::LinkBlock{next})
+```
 
 This terminal instruction jumps to the basic block described by `next` if we have enough
 cycles remaining. If we do not have enough cycles remaining, we return to the
 dispatcher, which will return control to the host.
 
+### Terminal: LinkBlockFast
+
+```c++
+SetTerm(IR::Term::LinkBlockFast{next})
+```
+
+This terminal instruction jumps to the basic block described by `next` unconditionally.
+This promises guarantees that must be held at runtime - i.e that the program wont hang,
+
 ### Terminal: PopRSBHint
 
-    SetTerm(IR::Term::PopRSBHint{})
+```c++
+SetTerm(IR::Term::PopRSBHint{})
+```
 
 This terminal instruction checks the top of the Return Stack Buffer against R15.
 If RSB lookup fails, control is returned to the dispatcher.
 This is an optimization for faster function calls. A backend that doesn't support
 this optimization or doesn't have a RSB may choose to implement this exactly as
-ReturnToDispatch.
+`ReturnToDispatch`.
 
 ### Terminal: If
 
-    SetTerm(IR::Term::If{cond, term_then, term_else})
+```c++
+SetTerm(IR::Term::If{cond, term_then, term_else})
+```
 
 This terminal instruction conditionally executes one terminal or another depending
 on the run-time state of the ARM flags.

src/dynarmic/docs/FastMemory.md

@@ -0,0 +1,19 @@
+# Fast memory (Fastmem)
+
+The main way of accessing memory in JITed programs is via an invoked function, say "Read()" and "Write()". On our translator, such functions usually take a sizable amounts of code space (push + call + pop). Trash the i-cache (due to an indirect call) and overall make code emission more bloated.
+
+The solution? Delegate invalid accesses to a dedicated arena, similar to a swap. The main idea behind such mechanism is to allow the OS to transmit page faults from invalid accesses into the JIT translator directly, bypassing address space calls, while this sacrifices i-cache coherency, it allows for smaller code-size and "faster" throguhput.
+
+Many kernels however, do not support fast signal dispatching (Solaris, OpenBSD, FreeBSD). Only Linux and Windows support relatively "fast" signal dispatching. Hence this feature is better suited for them only.
+
+![Host to guest translation](./HostToGuest.svg)
+
+![Fastmem translation](./Fastmem.svg)
+
+In x86_64 for example, when a page fault occurs, the CPU will transmit via control registers and the stack (see `IRETQ`) the appropriate arguments for a page fault handler, the OS then will transform that into something that can be sent into userspace.
+
+Most modern OSes implement kernel-page-table-isolation, which means a set of system calls will invoke a context switch (not often used syscalls), whereas others are handled by the same process address space (the smaller kernel portion, often used syscalls) without needing a context switch. This effect can be negated on systems with PCID (up to 4096 unique IDs).
+
+Signal dispatching takes a performance hit from reloading `%cr3` - but Linux does something more clever to avoid reloads:  VDSO will take care of the entire thing in the same address space. Making dispatching as costly as an indirect call - without the hazards of increased code size.
+
+The main downside from this is the constant i-cache trashing and pipeline hazards introduced by the VDSO signal handlers. However on most benchmarks fastmem does perform faster than without (Linux only). This also abuses the fact of continous address space emulation by using an arena - which can then be potentially transparently mapped into a hugepage, reducing TLB walk times.

src/dynarmic/docs/RegisterAllocator.md

@@ -16,19 +16,34 @@ Note that `Use`ing a value decrements its `use_count` by one. When the `use_coun
 
 The member functions on `RegAlloc` are just a combination of the above concepts.
 
+The following registers are reserved for internal use and should NOT participate in register allocation:
+- `%xmm0`, `%xmm1`, `%xmm2`: Used as scratch in exclusive memory access.
+- `%rsp`: Stack pointer.
+- `%r15`: JIT pointer
+- `%r14`: Page table pointer.
+- `%r13`: Fastmem pointer.
+
+The layout convenes `%r15` as the JIT state pointer - while it may be tempting to turn it into a synthetic pointer, keeping an entire register (out of 12 available) is preferable over inlining a directly computed immediate.
+
+Do NEVER modify `%r15`, we must make it clear that this register is "immutable" for the entirety of the JIT block duration.
+
 ### `Scratch`
 
-    Xbyak::Reg64 ScratchGpr(HostLocList desired_locations = any_gpr)
-    Xbyak::Xmm ScratchXmm(HostLocList desired_locations = any_xmm)
+```c++
+Xbyak::Reg64 ScratchGpr(HostLocList desired_locations = any_gpr);
+Xbyak::Xmm ScratchXmm(HostLocList desired_locations = any_xmm);
+```
 
 At runtime, allocate one of the registers in `desired_locations`. You are free to modify the register. The register is discarded at the end of the allocation scope.
 
 ### Pure `Use`
 
-    Xbyak::Reg64 UseGpr(Argument& arg);
-    Xbyak::Xmm UseXmm(Argument& arg);
-    OpArg UseOpArg(Argument& arg);
-    void Use(Argument& arg, HostLoc host_loc);
+```c++
+Xbyak::Reg64 UseGpr(Argument& arg);
+Xbyak::Xmm UseXmm(Argument& arg);
+OpArg UseOpArg(Argument& arg);
+void Use(Argument& arg, HostLoc host_loc);
+```
 
 At runtime, the value corresponding to `arg` will be placed a register. The actual register is determined by
 which one of the above functions is called. `UseGpr` places it in an unused GPR, `UseXmm` places it
@@ -39,9 +54,11 @@ This register **must not** have it's value changed.
 
 ### `UseScratch`
 
-    Xbyak::Reg64 UseScratchGpr(Argument& arg);
-    Xbyak::Xmm UseScratchXmm(Argument& arg);
-    void UseScratch(Argument& arg, HostLoc host_loc);
+```c++
+Xbyak::Reg64 UseScratchGpr(Argument& arg);
+Xbyak::Xmm UseScratchXmm(Argument& arg);
+void UseScratch(Argument& arg, HostLoc host_loc);
+```
 
 At runtime, the value corresponding to `arg` will be placed a register. The actual register is determined by
 which one of the above functions is called. `UseScratchGpr` places it in an unused GPR, `UseScratchXmm` places it
@@ -55,7 +72,9 @@ You are free to modify the value in the register. The register is discarded at t
 
 A `Define` is the defintion of a value. This is the only time when a value may be set.
 
-    void DefineValue(IR::Inst* inst, const Xbyak::Reg& reg);
+```c++
+void DefineValue(IR::Inst* inst, const Xbyak::Reg& reg);
+```
 
 By calling `DefineValue`, you are stating that you wish to define the value for `inst`, and you have written the
 value to the specified register `reg`.
@@ -64,7 +83,9 @@ value to the specified register `reg`.
 
 Adding a `Define` to an existing value.
 
-    void DefineValue(IR::Inst* inst, Argument& arg);
+```c++
+void DefineValue(IR::Inst* inst, Argument& arg);
+```
 
 You are declaring that the value for `inst` is the same as the value for `arg`. No host machine instructions are
 emitted.

src/dynarmic/docs/ReturnStackBufferOptimization.md

@@ -23,15 +23,17 @@ One complication dynarmic has is that a compiled block is not uniquely identifia
 the PC alone, but bits in the FPSCR and CPSR are also relevant. We resolve this by
 computing a 64-bit `UniqueHash` that is guaranteed to uniquely identify a block.
 
-    u64 LocationDescriptor::UniqueHash() const {
-        // This value MUST BE UNIQUE.
-        // This calculation has to match up with EmitX64::EmitTerminalPopRSBHint
-        u64 pc_u64 = u64(arm_pc) << 32;
-        u64 fpscr_u64 = u64(fpscr.Value());
-        u64 t_u64 = cpsr.T() ? 1 : 0;
-        u64 e_u64 = cpsr.E() ? 2 : 0;
-        return pc_u64 | fpscr_u64 | t_u64 | e_u64;
-    }
+```c++
+u64 LocationDescriptor::UniqueHash() const {
+    // This value MUST BE UNIQUE.
+    // This calculation has to match up with EmitX64::EmitTerminalPopRSBHint
+    u64 pc_u64 = u64(arm_pc) << 32;
+    u64 fpscr_u64 = u64(fpscr.Value());
+    u64 t_u64 = cpsr.T() ? 1 : 0;
+    u64 e_u64 = cpsr.E() ? 2 : 0;
+    return pc_u64 | fpscr_u64 | t_u64 | e_u64;
+}
+```
 
 ## Our implementation isn't actually a stack
 
@@ -49,97 +51,107 @@ host addresses for the corresponding the compiled blocks.
 size of the real RSB in hardware (which has 3 entries). Larger RSBs than 8
 showed degraded performance.
 
-    struct JitState {
-        // ...
+```c++
+struct JitState {
+    // ...
 
-        static constexpr size_t RSBSize = 8; // MUST be a power of 2.
-        u32 rsb_ptr = 0;
-        std::array<u64, RSBSize> rsb_location_descriptors;
-        std::array<u64, RSBSize> rsb_codeptrs;
-        void ResetRSB();
+    static constexpr size_t RSBSize = 8; // MUST be a power of 2.
+    u32 rsb_ptr = 0;
+    std::array<u64, RSBSize> rsb_location_descriptors;
+    std::array<u64, RSBSize> rsb_codeptrs;
+    void ResetRSB();
 
-        // ...
-    };
+    // ...
+};
+```
 
 ### RSB Push
 
 We insert our prediction at the insertion point iff the RSB doesn't already
 contain a prediction with the same `UniqueHash`.
 
-    void EmitX64::EmitPushRSB(IR::Block&, IR::Inst* inst) {
-        using namespace Xbyak::util;
+```c++
+void EmitX64::EmitPushRSB(IR::Block&, IR::Inst* inst) {
+    using namespace Xbyak::util;
 
-        ASSERT(inst->GetArg(0).IsImmediate());
-        u64 imm64 = inst->GetArg(0).GetU64();
+    ASSERT(inst->GetArg(0).IsImmediate());
+    u64 imm64 = inst->GetArg(0).GetU64();
 
-        Xbyak::Reg64 code_ptr_reg = reg_alloc.ScratchGpr({HostLoc::RCX});
-        Xbyak::Reg64 loc_desc_reg = reg_alloc.ScratchGpr();
-        Xbyak::Reg32 index_reg = reg_alloc.ScratchGpr().cvt32();
-        u64 code_ptr = unique_hash_to_code_ptr.find(imm64) != unique_hash_to_code_ptr.end()
-                        ? u64(unique_hash_to_code_ptr[imm64])
-                        : u64(code->GetReturnFromRunCodeAddress());
+    Xbyak::Reg64 code_ptr_reg = reg_alloc.ScratchGpr({HostLoc::RCX});
+    Xbyak::Reg64 loc_desc_reg = reg_alloc.ScratchGpr();
+    Xbyak::Reg32 index_reg = reg_alloc.ScratchGpr().cvt32();
+    u64 code_ptr = unique_hash_to_code_ptr.find(imm64) != unique_hash_to_code_ptr.end()
+                    ? u64(unique_hash_to_code_ptr[imm64])
+                    : u64(code->GetReturnFromRunCodeAddress());
 
-        code->mov(index_reg, dword[code.ABI_JIT_PTR + offsetof(JitState, rsb_ptr)]);
-        code->add(index_reg, 1);
-        code->and_(index_reg, u32(JitState::RSBSize - 1));
+    code->mov(index_reg, dword[code.ABI_JIT_PTR + offsetof(JitState, rsb_ptr)]);
+    code->add(index_reg, 1);
+    code->and_(index_reg, u32(JitState::RSBSize - 1));
 
-        code->mov(loc_desc_reg, u64(imm64));
-        CodePtr patch_location = code->getCurr<CodePtr>();
-        patch_unique_hash_locations[imm64].emplace_back(patch_location);
-        code->mov(code_ptr_reg, u64(code_ptr)); // This line has to match up with EmitX64::Patch.
-        code->EnsurePatchLocationSize(patch_location, 10);
+    code->mov(loc_desc_reg, u64(imm64));
+    CodePtr patch_location = code->getCurr<CodePtr>();
+    patch_unique_hash_locations[imm64].emplace_back(patch_location);
+    code->mov(code_ptr_reg, u64(code_ptr)); // This line has to match up with EmitX64::Patch.
+    code->EnsurePatchLocationSize(patch_location, 10);
 
-        Xbyak::Label label;
-        for (size_t i = 0; i < JitState::RSBSize; ++i) {
-            code->cmp(loc_desc_reg, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_location_descriptors) + i * sizeof(u64)]);
-            code->je(label, code->T_SHORT);
-        }
-
-        code->mov(dword[code.ABI_JIT_PTR + offsetof(JitState, rsb_ptr)], index_reg);
-        code->mov(qword[code.ABI_JIT_PTR + index_reg.cvt64() * 8 + offsetof(JitState, rsb_location_descriptors)], loc_desc_reg);
-        code->mov(qword[code.ABI_JIT_PTR + index_reg.cvt64() * 8 + offsetof(JitState, rsb_codeptrs)], code_ptr_reg);
-        code->L(label);
+    Xbyak::Label label;
+    for (size_t i = 0; i < JitState::RSBSize; ++i) {
+        code->cmp(loc_desc_reg, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_location_descriptors) + i * sizeof(u64)]);
+        code->je(label, code->T_SHORT);
     }
 
+    code->mov(dword[code.ABI_JIT_PTR + offsetof(JitState, rsb_ptr)], index_reg);
+    code->mov(qword[code.ABI_JIT_PTR + index_reg.cvt64() * 8 + offsetof(JitState, rsb_location_descriptors)], loc_desc_reg);
+    code->mov(qword[code.ABI_JIT_PTR + index_reg.cvt64() * 8 + offsetof(JitState, rsb_codeptrs)], code_ptr_reg);
+    code->L(label);
+}
+```
+
 In pseudocode:
 
-      for (i := 0 .. RSBSize-1)
-          if (rsb_location_descriptors[i] == imm64)
-            goto label;
-      rsb_ptr++;
-      rsb_ptr %= RSBSize;
-      rsb_location_desciptors[rsb_ptr] = imm64; //< The UniqueHash
-      rsb_codeptr[rsb_ptr] = /* codeptr corresponding to the UniqueHash */;
-    label:
+```c++
+    for (i := 0 .. RSBSize-1)
+        if (rsb_location_descriptors[i] == imm64)
+        goto label;
+    rsb_ptr++;
+    rsb_ptr %= RSBSize;
+    rsb_location_desciptors[rsb_ptr] = imm64; //< The UniqueHash
+    rsb_codeptr[rsb_ptr] = /* codeptr corresponding to the UniqueHash */;
+label:
+```
 
 ## RSB Pop
 
 To check if a predicition is in the RSB, we linearly scan the RSB.
 
-    void EmitX64::EmitTerminalPopRSBHint(IR::Term::PopRSBHint, IR::LocationDescriptor initial_location) {
-        using namespace Xbyak::util;
+```c++
+void EmitX64::EmitTerminalPopRSBHint(IR::Term::PopRSBHint, IR::LocationDescriptor initial_location) {
+    using namespace Xbyak::util;
 
-        // This calculation has to match up with IREmitter::PushRSB
-        code->mov(ecx, MJitStateReg(Arm::Reg::PC));
-        code->shl(rcx, 32);
-        code->mov(ebx, dword[code.ABI_JIT_PTR + offsetof(JitState, FPSCR_mode)]);
-        code->or_(ebx, dword[code.ABI_JIT_PTR + offsetof(JitState, CPSR_et)]);
-        code->or_(rbx, rcx);
+    // This calculation has to match up with IREmitter::PushRSB
+    code->mov(ecx, MJitStateReg(Arm::Reg::PC));
+    code->shl(rcx, 32);
+    code->mov(ebx, dword[code.ABI_JIT_PTR + offsetof(JitState, FPSCR_mode)]);
+    code->or_(ebx, dword[code.ABI_JIT_PTR + offsetof(JitState, CPSR_et)]);
+    code->or_(rbx, rcx);
 
-        code->mov(rax, u64(code->GetReturnFromRunCodeAddress()));
-        for (size_t i = 0; i < JitState::RSBSize; ++i) {
-            code->cmp(rbx, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_location_descriptors) + i * sizeof(u64)]);
-            code->cmove(rax, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_codeptrs) + i * sizeof(u64)]);
-        }
-
-        code->jmp(rax);
+    code->mov(rax, u64(code->GetReturnFromRunCodeAddress()));
+    for (size_t i = 0; i < JitState::RSBSize; ++i) {
+        code->cmp(rbx, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_location_descriptors) + i * sizeof(u64)]);
+        code->cmove(rax, qword[code.ABI_JIT_PTR + offsetof(JitState, rsb_codeptrs) + i * sizeof(u64)]);
     }
 
+    code->jmp(rax);
+}
+```
+
 In pseudocode:
 
-    rbx := ComputeUniqueHash()
-    rax := ReturnToDispatch
-    for (i := 0 .. RSBSize-1)
-       if (rbx == rsb_location_descriptors[i])
-          rax = rsb_codeptrs[i]
-    goto rax
+```c++
+rbx := ComputeUniqueHash()
+rax := ReturnToDispatch
+for (i := 0 .. RSBSize-1)
+    if (rbx == rsb_location_descriptors[i])
+        rax = rsb_codeptrs[i]
+goto rax
+```

src/dynarmic/src/dynarmic/CMakeLists.txt

@@ -56,14 +56,11 @@ add_library(dynarmic
     common/lut_from_list.h
     common/math_util.cpp
     common/math_util.h
-    common/memory_pool.cpp
-    common/memory_pool.h
     common/safe_ops.h
     common/spin_lock.h
     common/string_util.h
     common/u128.cpp
     common/u128.h
-    common/variant_util.h
     frontend/A32/a32_types.cpp
     frontend/A32/a32_types.h
     frontend/A64/a64_types.cpp
@@ -78,7 +75,6 @@ add_library(dynarmic
     ir/basic_block.cpp
     ir/basic_block.h
     ir/cond.h
-    ir/ir_emitter.cpp
     ir/ir_emitter.h
     ir/location_descriptor.cpp
     ir/location_descriptor.h
@@ -87,78 +83,59 @@ add_library(dynarmic
     ir/opcodes.cpp
     ir/opcodes.h
     ir/opcodes.inc
-    ir/opt/constant_propagation_pass.cpp
-    ir/opt/dead_code_elimination_pass.cpp
-    ir/opt/identity_removal_pass.cpp
-    ir/opt/ir_matcher.h
-    ir/opt/naming_pass.cpp
-    ir/opt/passes.h
-    ir/opt/polyfill_pass.cpp
-    ir/opt/verification_pass.cpp
+    ir/opt_passes.cpp
+    ir/opt_passes.h
     ir/terminal.h
     ir/type.cpp
     ir/type.h
     ir/value.cpp
     ir/value.h
+    # A32
+    frontend/A32/a32_ir_emitter.cpp
+    frontend/A32/a32_ir_emitter.h
+    frontend/A32/a32_location_descriptor.cpp
+    frontend/A32/a32_location_descriptor.h
+    frontend/A32/decoder/arm.h
+    frontend/A32/decoder/arm.inc
+    frontend/A32/decoder/asimd.h
+    frontend/A32/decoder/asimd.inc
+    frontend/A32/decoder/thumb16.h
+    frontend/A32/decoder/thumb16.inc
+    frontend/A32/decoder/thumb32.h
+    frontend/A32/decoder/thumb32.inc
+    frontend/A32/decoder/vfp.h
+    frontend/A32/decoder/vfp.inc
+    frontend/A32/disassembler/disassembler.h
+    frontend/A32/disassembler/disassembler_arm.cpp
+    frontend/A32/disassembler/disassembler_thumb.cpp
+    frontend/A32/FPSCR.h
+    frontend/A32/ITState.h
+    frontend/A32/PSR.h
+    frontend/A32/translate/a32_translate.cpp
+    frontend/A32/translate/a32_translate.h
+    frontend/A32/translate/conditional_state.cpp
+    frontend/A32/translate/conditional_state.h
+    frontend/A32/translate/translate_arm.cpp
+    frontend/A32/translate/translate_thumb.cpp
+    interface/A32/a32.h
+    interface/A32/arch_version.h
+    interface/A32/config.h
+    interface/A32/coprocessor.h
+    interface/A32/coprocessor_util.h
+    interface/A32/disassembler.h
+    # A64
+    frontend/A64/a64_ir_emitter.cpp
+    frontend/A64/a64_ir_emitter.h
+    frontend/A64/a64_location_descriptor.cpp
+    frontend/A64/a64_location_descriptor.h
+    frontend/A64/decoder/a64.h
+    frontend/A64/decoder/a64.inc
+    frontend/A64/translate/a64_translate.cpp
+    frontend/A64/translate/a64_translate.h
+    interface/A64/a64.h
+    interface/A64/config.h
 )
 
-if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-    target_sources(dynarmic PRIVATE
-        frontend/A32/a32_ir_emitter.cpp
-        frontend/A32/a32_ir_emitter.h
-        frontend/A32/a32_location_descriptor.cpp
-        frontend/A32/a32_location_descriptor.h
-        frontend/A32/decoder/arm.h
-        frontend/A32/decoder/arm.inc
-        frontend/A32/decoder/asimd.h
-        frontend/A32/decoder/asimd.inc
-        frontend/A32/decoder/thumb16.h
-        frontend/A32/decoder/thumb16.inc
-        frontend/A32/decoder/thumb32.h
-        frontend/A32/decoder/thumb32.inc
-        frontend/A32/decoder/vfp.h
-        frontend/A32/decoder/vfp.inc
-        frontend/A32/disassembler/disassembler.h
-        frontend/A32/disassembler/disassembler_arm.cpp
-        frontend/A32/disassembler/disassembler_thumb.cpp
-        frontend/A32/FPSCR.h
-        frontend/A32/ITState.h
-        frontend/A32/PSR.h
-        frontend/A32/translate/a32_translate.cpp
-        frontend/A32/translate/a32_translate.h
-        frontend/A32/translate/conditional_state.cpp
-        frontend/A32/translate/conditional_state.h
-        frontend/A32/translate/translate_arm.cpp
-        frontend/A32/translate/translate_thumb.cpp
-        interface/A32/a32.h
-        interface/A32/arch_version.h
-        interface/A32/config.h
-        interface/A32/coprocessor.h
-        interface/A32/coprocessor_util.h
-        interface/A32/disassembler.h
-        ir/opt/a32_constant_memory_reads_pass.cpp
-        ir/opt/a32_get_set_elimination_pass.cpp
-    )
-endif()
-
-if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-    target_sources(dynarmic PRIVATE
-        frontend/A64/a64_ir_emitter.cpp
-        frontend/A64/a64_ir_emitter.h
-        frontend/A64/a64_location_descriptor.cpp
-        frontend/A64/a64_location_descriptor.h
-        frontend/A64/decoder/a64.h
-        frontend/A64/decoder/a64.inc
-        frontend/A64/translate/a64_translate.cpp
-        frontend/A64/translate/a64_translate.h
-        interface/A64/a64.h
-        interface/A64/config.h
-        ir/opt/a64_callback_config_pass.cpp
-        ir/opt/a64_get_set_elimination_pass.cpp
-        ir/opt/a64_merge_interpret_blocks.cpp
-    )
-endif()
-
 if ("x86_64" IN_LIST ARCHITECTURE)
     # Newer versions of xbyak (>= 7.25.0) have stricter checks that currently
     # fail in dynarmic
@@ -215,29 +192,21 @@ if ("x86_64" IN_LIST ARCHITECTURE)
         common/spin_lock_x64.h
         common/x64_disassemble.cpp
         common/x64_disassemble.h
+        # A32
+        backend/x64/a32_emit_x64.cpp
+        backend/x64/a32_emit_x64.h
+        backend/x64/a32_emit_x64_memory.cpp
+        backend/x64/a32_interface.cpp
+        backend/x64/a32_jitstate.cpp
+        backend/x64/a32_jitstate.h
+        # A64
+        backend/x64/a64_emit_x64.cpp
+        backend/x64/a64_emit_x64.h
+        backend/x64/a64_emit_x64_memory.cpp
+        backend/x64/a64_interface.cpp
+        backend/x64/a64_jitstate.cpp
+        backend/x64/a64_jitstate.h
     )
-
-    if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-        target_architecture_specific_sources(dynarmic "x86_64"
-            backend/x64/a32_emit_x64.cpp
-            backend/x64/a32_emit_x64.h
-            backend/x64/a32_emit_x64_memory.cpp
-            backend/x64/a32_interface.cpp
-            backend/x64/a32_jitstate.cpp
-            backend/x64/a32_jitstate.h
-        )
-    endif()
-
-    if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-        target_architecture_specific_sources(dynarmic "x86_64"
-            backend/x64/a64_emit_x64.cpp
-            backend/x64/a64_emit_x64.h
-            backend/x64/a64_emit_x64_memory.cpp
-            backend/x64/a64_interface.cpp
-            backend/x64/a64_jitstate.cpp
-            backend/x64/a64_jitstate.h
-        )
-    endif()
 endif()
 
 if ("arm64" IN_LIST ARCHITECTURE)
@@ -281,25 +250,17 @@ if ("arm64" IN_LIST ARCHITECTURE)
         backend/arm64/verbose_debugging_output.h
         common/spin_lock_arm64.cpp
         common/spin_lock_arm64.h
+        # A32
+        backend/arm64/a32_address_space.cpp
+        backend/arm64/a32_address_space.h
+        backend/arm64/a32_core.h
+        backend/arm64/a32_interface.cpp
+        # A64
+        backend/arm64/a64_address_space.cpp
+        backend/arm64/a64_address_space.h
+        backend/arm64/a64_core.h
+        backend/arm64/a64_interface.cpp
     )
-
-    if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-        target_architecture_specific_sources(dynarmic "arm64"
-            backend/arm64/a32_address_space.cpp
-            backend/arm64/a32_address_space.h
-            backend/arm64/a32_core.h
-            backend/arm64/a32_interface.cpp
-        )
-    endif()
-
-    if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-        target_architecture_specific_sources(dynarmic "arm64"
-            backend/arm64/a64_address_space.cpp
-            backend/arm64/a64_address_space.h
-            backend/arm64/a64_core.h
-            backend/arm64/a64_interface.cpp
-        )
-    endif()
 endif()
 
 if ("riscv" IN_LIST ARCHITECTURE)
@@ -328,21 +289,14 @@ if ("riscv" IN_LIST ARCHITECTURE)
         backend/riscv64/reg_alloc.cpp
         backend/riscv64/reg_alloc.h
         backend/riscv64/stack_layout.h
+        # A32
+        backend/riscv64/a32_address_space.cpp
+        backend/riscv64/a32_address_space.h
+        backend/riscv64/a32_core.h
+        backend/riscv64/a32_interface.cpp
+        backend/riscv64/code_block.h
     )
-
-    if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-        target_sources(dynarmic PRIVATE
-            backend/riscv64/a32_address_space.cpp
-            backend/riscv64/a32_address_space.h
-            backend/riscv64/a32_core.h
-            backend/riscv64/a32_interface.cpp
-            backend/riscv64/code_block.h
-        )
-    endif()
-
-    if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-        message(FATAL_ERROR "TODO: Unimplemented frontend for this host architecture")
-    endif()
+    message(FATAL_ERROR "TODO: Unimplemented frontend for this host architecture")
 endif()
 
 if (WIN32)
@@ -420,7 +374,7 @@ target_link_libraries(dynarmic
 )
 
 if (BOOST_NO_HEADERS)
-target_link_libraries(dynarmic PRIVATE Boost::variant Boost::icl Boost::pool)
+    target_link_libraries(dynarmic PRIVATE Boost::variant Boost::icl Boost::pool)
 else()
     target_link_libraries(dynarmic PRIVATE Boost::headers)
 endif()

src/dynarmic/src/dynarmic/backend/arm64/a32_address_space.cpp

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 /* This file is part of the dynarmic project.
  * Copyright (c) 2022 MerryMage
  * SPDX-License-Identifier: 0BSD
@@ -16,7 +19,7 @@
 #include "dynarmic/frontend/A32/translate/a32_translate.h"
 #include "dynarmic/interface/A32/config.h"
 #include "dynarmic/interface/exclusive_monitor.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 namespace Dynarmic::Backend::Arm64 {
 
@@ -163,21 +166,7 @@ A32AddressSpace::A32AddressSpace(const A32::UserConfig& conf)
 
 IR::Block A32AddressSpace::GenerateIR(IR::LocationDescriptor descriptor) const {
     IR::Block ir_block = A32::Translate(A32::LocationDescriptor{descriptor}, conf.callbacks, {conf.arch_version, conf.define_unpredictable_behaviour, conf.hook_hint_instructions});
-
-    Optimization::PolyfillPass(ir_block, {});
-    Optimization::NamingPass(ir_block);
-    if (conf.HasOptimization(OptimizationFlag::GetSetElimination)) {
-        Optimization::A32GetSetElimination(ir_block, {.convert_nzc_to_nz = true});
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    if (conf.HasOptimization(OptimizationFlag::ConstProp)) {
-        Optimization::A32ConstantMemoryReads(ir_block, conf.callbacks);
-        Optimization::ConstantPropagation(ir_block);
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    Optimization::IdentityRemovalPass(ir_block);
-    Optimization::VerificationPass(ir_block);
-
+    Optimization::Optimize(ir_block, conf, {});
     return ir_block;
 }
 

src/dynarmic/src/dynarmic/backend/arm64/a64_address_space.cpp

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 /* This file is part of the dynarmic project.
  * Copyright (c) 2022 MerryMage
  * SPDX-License-Identifier: 0BSD
@@ -15,7 +18,7 @@
 #include "dynarmic/frontend/A64/translate/a64_translate.h"
 #include "dynarmic/interface/A64/config.h"
 #include "dynarmic/interface/exclusive_monitor.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 namespace Dynarmic::Backend::Arm64 {
 
@@ -331,22 +334,7 @@ IR::Block A64AddressSpace::GenerateIR(IR::LocationDescriptor descriptor) const {
     const auto get_code = [this](u64 vaddr) { return conf.callbacks->MemoryReadCode(vaddr); };
     IR::Block ir_block = A64::Translate(A64::LocationDescriptor{descriptor}, get_code,
                                         {conf.define_unpredictable_behaviour, conf.wall_clock_cntpct});
-
-    Optimization::A64CallbackConfigPass(ir_block, conf);
-    Optimization::NamingPass(ir_block);
-    if (conf.HasOptimization(OptimizationFlag::GetSetElimination) && !conf.check_halt_on_memory_access) {
-        Optimization::A64GetSetElimination(ir_block);
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    if (conf.HasOptimization(OptimizationFlag::ConstProp)) {
-        Optimization::ConstantPropagation(ir_block);
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    if (conf.HasOptimization(OptimizationFlag::MiscIROpt)) {
-        Optimization::A64MergeInterpretBlocksPass(ir_block, conf.callbacks);
-    }
-    Optimization::VerificationPass(ir_block);
-
+    Optimization::Optimize(ir_block, conf, {});
     return ir_block;
 }
 

src/dynarmic/src/dynarmic/backend/exception_handler.h

@@ -15,15 +15,15 @@
 #include <mcl/macro/architecture.hpp>
 #include "dynarmic/common/common_types.h"
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 namespace Dynarmic::Backend::X64 {
 class BlockOfCode;
 }  // namespace Dynarmic::Backend::X64
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 namespace oaknut {
 class CodeBlock;
 }  // namespace oaknut
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
 namespace Dynarmic::Backend::RV64 {
 class CodeBlock;
 }  // namespace Dynarmic::Backend::RV64
@@ -33,16 +33,16 @@ class CodeBlock;
 
 namespace Dynarmic::Backend {
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 struct FakeCall {
     u64 call_rip;
     u64 ret_rip;
 };
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 struct FakeCall {
     u64 call_pc;
 };
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
 struct FakeCall {
 };
 #else
@@ -54,11 +54,11 @@ public:
     ExceptionHandler();
     ~ExceptionHandler();
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
     void Register(X64::BlockOfCode& code);
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
     void Register(oaknut::CodeBlock& mem, std::size_t mem_size);
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
     void Register(RV64::CodeBlock& mem, std::size_t mem_size);
 #else
 #    error "Invalid architecture"

src/dynarmic/src/dynarmic/backend/exception_handler_generic.cpp

@@ -13,15 +13,15 @@ struct ExceptionHandler::Impl final {
 ExceptionHandler::ExceptionHandler() = default;
 ExceptionHandler::~ExceptionHandler() = default;
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 void ExceptionHandler::Register(X64::BlockOfCode&) {
     // Do nothing
 }
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 void ExceptionHandler::Register(oaknut::CodeBlock&, std::size_t) {
     // Do nothing
 }
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
 void ExceptionHandler::Register(RV64::CodeBlock&, std::size_t) {
     // Do nothing
 }

src/dynarmic/src/dynarmic/backend/exception_handler_macos.cpp

@@ -25,7 +25,7 @@
 
 #include "dynarmic/backend/exception_handler.h"
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 
 #    include "dynarmic/backend/x64/block_of_code.h"
 #    define mig_external extern "C"
@@ -36,7 +36,7 @@
 
 using dynarmic_thread_state_t = x86_thread_state64_t;
 
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 
 #    include <oaknut/code_block.hpp>
 #    define mig_external extern "C"
@@ -133,7 +133,7 @@ void MachHandler::MessagePump() {
     }
 }
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 kern_return_t MachHandler::HandleRequest(x86_thread_state64_t* ts) {
     std::lock_guard<std::mutex> guard(code_block_infos_mutex);
 
@@ -151,7 +151,7 @@ kern_return_t MachHandler::HandleRequest(x86_thread_state64_t* ts) {
 
     return KERN_SUCCESS;
 }
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 kern_return_t MachHandler::HandleRequest(arm_thread_state64_t* ts) {
     std::lock_guard<std::mutex> guard(code_block_infos_mutex);
 
@@ -269,13 +269,13 @@ private:
 ExceptionHandler::ExceptionHandler() = default;
 ExceptionHandler::~ExceptionHandler() = default;
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 void ExceptionHandler::Register(X64::BlockOfCode& code) {
     const u64 code_begin = mcl::bit_cast<u64>(code.getCode());
     const u64 code_end = code_begin + code.GetTotalCodeSize();
     impl = std::make_unique<Impl>(code_begin, code_end);
 }
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 void ExceptionHandler::Register(oaknut::CodeBlock& mem, std::size_t size) {
     const u64 code_begin = mcl::bit_cast<u64>(mem.ptr());
     const u64 code_end = code_begin + size;

src/dynarmic/src/dynarmic/backend/exception_handler_macos_mig.c

@@ -5,9 +5,9 @@
 
 #include <mcl/macro/architecture.hpp>
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 #    include "dynarmic/backend/x64/mig/mach_exc_server.c"
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 #    include "dynarmic/backend/arm64/mig/mach_exc_server.c"
 #else
 #    error "Invalid architecture"

src/dynarmic/src/dynarmic/backend/exception_handler_posix.cpp

@@ -12,31 +12,17 @@
 #include <mutex>
 #include <shared_mutex>
 #include <optional>
-#include <sys/mman.h>
-#ifdef __APPLE__
-#    include <signal.h>
-#    include <sys/ucontext.h>
-#else
-#    include <signal.h>
-#    ifndef __OpenBSD__
-#        include <ucontext.h>
-#    endif
-#    ifdef __sun__
-#        include <sys/regset.h>
-#    endif
-#endif
-
 #include <ankerl/unordered_dense.h>
-
 #include "dynarmic/backend/exception_handler.h"
+#include "dynarmic/common/assert.h"
+#include "dynarmic/common/context.h"
 #include "dynarmic/common/common_types.h"
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 #    include "dynarmic/backend/x64/block_of_code.h"
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 #    include <oaknut/code_block.hpp>
-
 #    include "dynarmic/backend/arm64/abi.h"
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
 #    include "dynarmic/backend/riscv64/code_block.h"
 #else
 #    error "Invalid architecture"
@@ -129,35 +115,8 @@ void RegisterHandler() {
 
 void SigHandler::SigAction(int sig, siginfo_t* info, void* raw_context) {
     DEBUG_ASSERT(sig == SIGSEGV || sig == SIGBUS);
-#ifndef MCL_ARCHITECTURE_RISCV
-    ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(raw_context);
-#ifndef __OpenBSD__
-    auto& mctx = ucontext->uc_mcontext;
-#endif
-#endif
-
-#if defined(MCL_ARCHITECTURE_X86_64)
-#    if defined(__APPLE__)
-#        define CTX_RIP (mctx->__ss.__rip)
-#        define CTX_RSP (mctx->__ss.__rsp)
-#    elif defined(__linux__)
-#        define CTX_RIP (mctx.gregs[REG_RIP])
-#        define CTX_RSP (mctx.gregs[REG_RSP])
-#    elif defined(__FreeBSD__)
-#        define CTX_RIP (mctx.mc_rip)
-#        define CTX_RSP (mctx.mc_rsp)
-#    elif defined(__NetBSD__)
-#        define CTX_RIP (mctx.__gregs[_REG_RIP])
-#        define CTX_RSP (mctx.__gregs[_REG_RSP])
-#    elif defined(__OpenBSD__)
-#        define CTX_RIP (ucontext->sc_rip)
-#        define CTX_RSP (ucontext->sc_rsp)
-#    elif defined(__sun__)
-#        define CTX_RIP (mctx.gregs[REG_RIP])
-#        define CTX_RSP (mctx.gregs[REG_RSP])
-#    else
-#        error "Unknown platform"
-#    endif
+    CTX_DECLARE(raw_context);
+#if defined(ARCHITECTURE_x86_64)
     {
         std::shared_lock guard(sig_handler->code_block_infos_mutex);
         if (auto const iter = sig_handler->FindCodeBlockInfo(CTX_RIP); iter != sig_handler->code_block_infos.end()) {
@@ -169,48 +128,7 @@ void SigHandler::SigAction(int sig, siginfo_t* info, void* raw_context) {
         }
     }
     fmt::print(stderr, "Unhandled {} at rip {:#018x}\n", sig == SIGSEGV ? "SIGSEGV" : "SIGBUS", CTX_RIP);
-#elif defined(MCL_ARCHITECTURE_ARM64)
-#    if defined(__APPLE__)
-#        define CTX_PC (mctx->__ss.__pc)
-#        define CTX_SP (mctx->__ss.__sp)
-#        define CTX_LR (mctx->__ss.__lr)
-#        define CTX_X(i) (mctx->__ss.__x[i])
-#        define CTX_Q(i) (mctx->__ns.__v[i])
-#    elif defined(__linux__)
-#        define CTX_PC (mctx.pc)
-#        define CTX_SP (mctx.sp)
-#        define CTX_LR (mctx.regs[30])
-#        define CTX_X(i) (mctx.regs[i])
-#        define CTX_Q(i) (fpctx->vregs[i])
-    [[maybe_unused]] const auto fpctx = [&mctx] {
-        _aarch64_ctx* header = (_aarch64_ctx*)&mctx.__reserved;
-        while (header->magic != FPSIMD_MAGIC) {
-            ASSERT(header->magic && header->size);
-            header = (_aarch64_ctx*)((char*)header + header->size);
-        }
-        return (fpsimd_context*)header;
-    }();
-#    elif defined(__FreeBSD__)
-#        define CTX_PC (mctx.mc_gpregs.gp_elr)
-#        define CTX_SP (mctx.mc_gpregs.gp_sp)
-#        define CTX_LR (mctx.mc_gpregs.gp_lr)
-#        define CTX_X(i) (mctx.mc_gpregs.gp_x[i])
-#        define CTX_Q(i) (mctx.mc_fpregs.fp_q[i])
-#    elif defined(__NetBSD__)
-#        define CTX_PC (mctx.mc_gpregs.gp_elr)
-#        define CTX_SP (mctx.mc_gpregs.gp_sp)
-#        define CTX_LR (mctx.mc_gpregs.gp_lr)
-#        define CTX_X(i) (mctx.mc_gpregs.gp_x[i])
-#        define CTX_Q(i) (mctx.mc_fpregs.fp_q[i])
-#    elif defined(__OpenBSD__)
-#        define CTX_PC (ucontext->sc_elr)
-#        define CTX_SP (ucontext->sc_sp)
-#        define CTX_LR (ucontext->sc_lr)
-#        define CTX_X(i) (ucontext->sc_x[i])
-#        define CTX_Q(i) (ucontext->sc_q[i])
-#    else
-#        error "Unknown platform"
-#    endif
+#elif defined(ARCHITECTURE_arm64)
     {
         std::shared_lock guard(sig_handler->code_block_infos_mutex);
         if (const auto iter = sig_handler->FindCodeBlockInfo(CTX_PC); iter != sig_handler->code_block_infos.end()) {
@@ -220,7 +138,7 @@ void SigHandler::SigAction(int sig, siginfo_t* info, void* raw_context) {
         }
     }
     fmt::print(stderr, "Unhandled {} at pc {:#018x}\n", sig == SIGSEGV ? "SIGSEGV" : "SIGBUS", CTX_PC);
-#elif defined(MCL_ARCHITECTURE_RISCV)
+#elif defined(ARCHITECTURE_riscv64)
     ASSERT_FALSE("Unimplemented");
 #else
 #    error "Invalid architecture"

src/dynarmic/src/dynarmic/backend/exception_handler_windows.cpp

@@ -5,9 +5,9 @@
 
 #include <mcl/macro/architecture.hpp>
 
-#if defined(MCL_ARCHITECTURE_X86_64)
+#if defined(ARCHITECTURE_x86_64)
 #    include "dynarmic/backend/x64/exception_handler_windows.cpp"
-#elif defined(MCL_ARCHITECTURE_ARM64)
+#elif defined(ARCHITECTURE_arm64)
 #    include "dynarmic/backend/exception_handler_generic.cpp"
 #else
 #    error "Invalid architecture"

src/dynarmic/src/dynarmic/backend/riscv64/a32_address_space.cpp

@@ -15,7 +15,7 @@
 #include "dynarmic/backend/riscv64/stack_layout.h"
 #include "dynarmic/frontend/A32/a32_location_descriptor.h"
 #include "dynarmic/frontend/A32/translate/a32_translate.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 namespace Dynarmic::Backend::RV64 {
 
@@ -28,19 +28,7 @@ A32AddressSpace::A32AddressSpace(const A32::UserConfig& conf)
 
 IR::Block A32AddressSpace::GenerateIR(IR::LocationDescriptor descriptor) const {
     IR::Block ir_block = A32::Translate(A32::LocationDescriptor{descriptor}, conf.callbacks, {conf.arch_version, conf.define_unpredictable_behaviour, conf.hook_hint_instructions});
-
-    Optimization::PolyfillPass(ir_block, {});
-    if (conf.HasOptimization(OptimizationFlag::GetSetElimination)) {
-        Optimization::A32GetSetElimination(ir_block, {.convert_nzc_to_nz = true});
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    if (conf.HasOptimization(OptimizationFlag::ConstProp)) {
-        Optimization::A32ConstantMemoryReads(ir_block, conf.callbacks);
-        Optimization::ConstantPropagation(ir_block);
-        Optimization::DeadCodeElimination(ir_block);
-    }
-    Optimization::VerificationPass(ir_block);
-
+    Optimization::Optimize(ir_block, conf, {});
     return ir_block;
 }
 

src/dynarmic/src/dynarmic/backend/x64/a32_emit_x64.cpp

@@ -28,7 +28,6 @@
 #include "dynarmic/backend/x64/nzcv_util.h"
 #include "dynarmic/backend/x64/perf_map.h"
 #include "dynarmic/backend/x64/stack_layout.h"
-#include "dynarmic/common/variant_util.h"
 #include "dynarmic/frontend/A32/a32_location_descriptor.h"
 #include "dynarmic/frontend/A32/a32_types.h"
 #include "dynarmic/interface/A32/coprocessor.h"

src/dynarmic/src/dynarmic/backend/x64/a32_interface.cpp

@@ -29,7 +29,7 @@
 #include "dynarmic/interface/A32/a32.h"
 #include "dynarmic/ir/basic_block.h"
 #include "dynarmic/ir/location_descriptor.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 namespace Dynarmic::A32 {
 
@@ -217,19 +217,7 @@ private:
         block_of_code.EnsureMemoryCommitted(MINIMUM_REMAINING_CODESIZE);
 
         IR::Block ir_block = A32::Translate(A32::LocationDescriptor{descriptor}, conf.callbacks, {conf.arch_version, conf.define_unpredictable_behaviour, conf.hook_hint_instructions});
-        Optimization::PolyfillPass(ir_block, polyfill_options);
-        Optimization::NamingPass(ir_block);
-        if (conf.HasOptimization(OptimizationFlag::GetSetElimination) && !conf.check_halt_on_memory_access) {
-            Optimization::A32GetSetElimination(ir_block, {.convert_nz_to_nzc = true});
-            Optimization::DeadCodeElimination(ir_block);
-        }
-        if (conf.HasOptimization(OptimizationFlag::ConstProp)) {
-            Optimization::A32ConstantMemoryReads(ir_block, conf.callbacks);
-            Optimization::ConstantPropagation(ir_block);
-            Optimization::DeadCodeElimination(ir_block);
-        }
-        Optimization::IdentityRemovalPass(ir_block);
-        Optimization::VerificationPass(ir_block);
+        Optimization::Optimize(ir_block, conf, polyfill_options);
         return emitter.Emit(ir_block);
     }
 

src/dynarmic/src/dynarmic/backend/x64/a64_emit_x64.cpp

@@ -122,9 +122,9 @@ A64EmitX64::BlockDescriptor A64EmitX64::Emit(IR::Block& block) noexcept {
         auto const opcode = inst.GetOpcode();
         // Call the relevant Emit* member function.
         switch (opcode) {
-#define OPCODE(name, type, ...) [[likely]] case IR::Opcode::name: goto opcode_branch;
+#define OPCODE(name, type, ...) case IR::Opcode::name: goto opcode_branch;
 #define A32OPC(name, type, ...)
-#define A64OPC(name, type, ...) [[likely]] case IR::Opcode::A64##name: goto a64_branch;
+#define A64OPC(name, type, ...) case IR::Opcode::A64##name: goto a64_branch;
 #include "dynarmic/ir/opcodes.inc"
 #undef OPCODE
 #undef A32OPC
@@ -764,7 +764,7 @@ void A64EmitX64::EmitPatchMovRcx(CodePtr target_code_ptr) {
         target_code_ptr = code.GetReturnFromRunCodeAddress();
     }
     const CodePtr patch_location = code.getCurr();
-    code.mov(code.rcx, reinterpret_cast<u64>(target_code_ptr));
+    code.mov(code.rcx, u64(target_code_ptr));
     code.EnsurePatchLocationSize(patch_location, 10);
 }
 

src/dynarmic/src/dynarmic/backend/x64/a64_interface.cpp

@@ -25,7 +25,7 @@
 #include "dynarmic/frontend/A64/translate/a64_translate.h"
 #include "dynarmic/interface/A64/a64.h"
 #include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 namespace Dynarmic::A64 {
 
@@ -80,16 +80,16 @@ public:
         };
 
         // TODO: Check code alignment
-
-        const CodePtr current_code_ptr = [this] {
+        const CodePtr aligned_code_ptr = CodePtr((uintptr_t(GetCurrentBlock()) + 15) & ~uintptr_t(15));
+        const CodePtr current_code_ptr = [this, aligned_code_ptr] {
             // RSB optimization
             const u32 new_rsb_ptr = (jit_state.rsb_ptr - 1) & A64JitState::RSBPtrMask;
             if (jit_state.GetUniqueHash() == jit_state.rsb_location_descriptors[new_rsb_ptr]) {
                 jit_state.rsb_ptr = new_rsb_ptr;
-                return reinterpret_cast<CodePtr>(jit_state.rsb_codeptrs[new_rsb_ptr]);
+                return CodePtr(jit_state.rsb_codeptrs[new_rsb_ptr]);
             }
-
-            return GetCurrentBlock();
+            return aligned_code_ptr;
+            //return GetCurrentBlock();
         }();
 
         const HaltReason hr = block_of_code.RunCode(&jit_state, current_code_ptr);
@@ -275,21 +275,7 @@ private:
         const auto get_code = [this](u64 vaddr) { return conf.callbacks->MemoryReadCode(vaddr); };
         IR::Block ir_block = A64::Translate(A64::LocationDescriptor{current_location}, get_code,
                                             {conf.define_unpredictable_behaviour, conf.wall_clock_cntpct});
-        Optimization::PolyfillPass(ir_block, polyfill_options);
-        Optimization::A64CallbackConfigPass(ir_block, conf);
-        Optimization::NamingPass(ir_block);
-        if (conf.HasOptimization(OptimizationFlag::GetSetElimination) && !conf.check_halt_on_memory_access) {
-            Optimization::A64GetSetElimination(ir_block);
-            Optimization::DeadCodeElimination(ir_block);
-        }
-        if (conf.HasOptimization(OptimizationFlag::ConstProp)) {
-            Optimization::ConstantPropagation(ir_block);
-            Optimization::DeadCodeElimination(ir_block);
-        }
-        if (conf.HasOptimization(OptimizationFlag::MiscIROpt)) {
-            Optimization::A64MergeInterpretBlocksPass(ir_block, conf.callbacks);
-        }
-        Optimization::VerificationPass(ir_block);
+        Optimization::Optimize(ir_block, conf, polyfill_options);
         return emitter.Emit(ir_block).entrypoint;
     }
 

src/dynarmic/src/dynarmic/backend/x64/abi.cpp

@@ -10,7 +10,6 @@
 
 #include <algorithm>
 
-#include <mcl/iterator/reverse.hpp>
 #include "dynarmic/common/common_types.h"
 #include <xbyak/xbyak.h>
 
@@ -76,7 +75,8 @@ void ABI_PopRegistersAndAdjustStack(BlockOfCode& code, const size_t frame_size,
     const FrameInfo frame_info = CalculateFrameInfo(num_gprs, num_xmms, frame_size);
 
     size_t xmm_offset = frame_info.xmm_offset + (num_xmms * XMM_SIZE);
-    for (auto const xmm : mcl::iterator::reverse(regs)) {
+    for (auto it = regs.rbegin(); it != regs.rend(); ++it) {
+        auto const xmm = *it;
         if (HostLocIsXMM(xmm)) {
             xmm_offset -= XMM_SIZE;
             if (code.HasHostFeature(HostFeature::AVX)) {
@@ -88,9 +88,11 @@ void ABI_PopRegistersAndAdjustStack(BlockOfCode& code, const size_t frame_size,
     }
     if (frame_info.stack_subtraction != 0)
         code.add(rsp, u32(frame_info.stack_subtraction));
-    for (auto const gpr : mcl::iterator::reverse(regs))
+    for (auto it = regs.rbegin(); it != regs.rend(); ++it) {
+        auto const gpr = *it;
         if (HostLocIsGPR(gpr))
             code.pop(HostLocToReg64(gpr));
+    }
 }
 
 void ABI_PushCalleeSaveRegistersAndAdjustStack(BlockOfCode& code, const std::size_t frame_size) {

src/dynarmic/src/dynarmic/backend/x64/block_of_code.cpp

@@ -364,8 +364,7 @@ void BlockOfCode::GenRunCode(std::function<void(BlockOfCode&)> rcp) {
 
     cmp(dword[ABI_JIT_PTR + jsi.offsetof_halt_reason], 0);
     jne(return_to_caller_mxcsr_already_exited, T_NEAR);
-    lock();
-    or_(dword[ABI_JIT_PTR + jsi.offsetof_halt_reason], static_cast<u32>(HaltReason::Step));
+    lock(); or_(dword[ABI_JIT_PTR + jsi.offsetof_halt_reason], static_cast<u32>(HaltReason::Step));
 
     SwitchMxcsrOnEntry();
     jmp(ABI_PARAM2);
@@ -415,7 +414,6 @@ void BlockOfCode::GenRunCode(std::function<void(BlockOfCode&)> rcp) {
     }
 
     xor_(eax, eax);
-    lock();
     xchg(dword[ABI_JIT_PTR + jsi.offsetof_halt_reason], eax);
 
     ABI_PopCalleeSaveRegistersAndAdjustStack(*this, sizeof(StackLayout));

src/dynarmic/src/dynarmic/backend/x64/emit_x64.cpp

@@ -11,6 +11,7 @@
 #include <iterator>
 
 #include "dynarmic/common/assert.h"
+#include <boost/variant/detail/apply_visitor_binary.hpp>
 #include <mcl/bit/bit_field.hpp>
 #include <mcl/scope_exit.hpp>
 #include "dynarmic/common/common_types.h"
@@ -21,7 +22,6 @@
 #include "dynarmic/backend/x64/perf_map.h"
 #include "dynarmic/backend/x64/stack_layout.h"
 #include "dynarmic/backend/x64/verbose_debugging_output.h"
-#include "dynarmic/common/variant_util.h"
 #include "dynarmic/ir/basic_block.h"
 #include "dynarmic/ir/microinstruction.h"
 #include "dynarmic/ir/opcodes.h"
@@ -347,14 +347,14 @@ EmitX64::BlockDescriptor EmitX64::RegisterBlock(const IR::LocationDescriptor& de
 }
 
 void EmitX64::EmitTerminal(IR::Terminal terminal, IR::LocationDescriptor initial_location, bool is_single_step) {
-    Common::VisitVariant<void>(terminal, [this, initial_location, is_single_step](auto x) {
+    boost::apply_visitor([this, initial_location, is_single_step](auto x) {
         using T = std::decay_t<decltype(x)>;
         if constexpr (!std::is_same_v<T, IR::Term::Invalid>) {
             this->EmitTerminalImpl(x, initial_location, is_single_step);
         } else {
             ASSERT_MSG(false, "Invalid terminal");
         }
-    });
+    }, terminal);
 }
 
 void EmitX64::Patch(const IR::LocationDescriptor& target_desc, CodePtr target_code_ptr) {

src/dynarmic/src/dynarmic/backend/x64/emit_x64_floating_point.cpp

@@ -92,13 +92,10 @@ void ForceDenormalsToZero(BlockOfCode& code, std::initializer_list<Xbyak::Xmm> t
             FpFixup::Norm_Src,
             FpFixup::Norm_Src,
             FpFixup::Norm_Src);
-
-        const Xbyak::Xmm tmp = xmm16;
+        const Xbyak::Xmm tmp = xmm0;
         FCODE(vmovap)(tmp, code.BConst<fsize>(xword, denormal_to_zero));
-
-        for (const Xbyak::Xmm& xmm : to_daz) {
+        for (const Xbyak::Xmm& xmm : to_daz)
             FCODE(vfixupimms)(xmm, xmm, tmp, u8(0));
-        }
         return;
     }
 

src/dynarmic/src/dynarmic/backend/x64/emit_x64_vector.cpp

@@ -609,8 +609,8 @@ void EmitX64::EmitVectorArithmeticVShift16(EmitContext& ctx, IR::Inst* inst) {
 
         const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
         const Xbyak::Xmm left_shift = ctx.reg_alloc.UseScratchXmm(args[1]);
-        const Xbyak::Xmm right_shift = xmm16;
-        const Xbyak::Xmm tmp = xmm17;
+        const Xbyak::Xmm right_shift = ctx.reg_alloc.ScratchXmm();
+        const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
 
         code.vmovdqa32(tmp, code.Const(xword, 0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF));
         code.vpxord(right_shift, right_shift, right_shift);
@@ -674,8 +674,8 @@ void EmitX64::EmitVectorArithmeticVShift64(EmitContext& ctx, IR::Inst* inst) {
 
         const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
         const Xbyak::Xmm left_shift = ctx.reg_alloc.UseScratchXmm(args[1]);
-        const Xbyak::Xmm right_shift = xmm16;
-        const Xbyak::Xmm tmp = xmm17;
+        const Xbyak::Xmm right_shift = ctx.reg_alloc.ScratchXmm();
+        const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
 
         code.vmovdqa32(tmp, code.Const(xword, 0x00000000000000FF, 0x00000000000000FF));
         code.vpxorq(right_shift, right_shift, right_shift);
@@ -1955,8 +1955,8 @@ void EmitX64::EmitVectorLogicalVShift16(EmitContext& ctx, IR::Inst* inst) {
 
         const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
         const Xbyak::Xmm left_shift = ctx.reg_alloc.UseScratchXmm(args[1]);
-        const Xbyak::Xmm right_shift = xmm16;
-        const Xbyak::Xmm tmp = xmm17;
+        const Xbyak::Xmm right_shift = ctx.reg_alloc.ScratchXmm();
+        const Xbyak::Xmm tmp = ctx.reg_alloc.ScratchXmm();
 
         code.vmovdqa32(tmp, code.Const(xword, 0x00FF00FF00FF00FF, 0x00FF00FF00FF00FF));
         code.vpxord(right_shift, right_shift, right_shift);
@@ -2737,7 +2737,7 @@ void EmitX64::EmitVectorPairedAddSignedWiden32(EmitContext& ctx, IR::Inst* inst)
     const Xbyak::Xmm a = ctx.reg_alloc.UseScratchXmm(args[0]);
 
     if (code.HasHostFeature(HostFeature::AVX512_Ortho)) {
-        const Xbyak::Xmm c = xmm16;
+        const Xbyak::Xmm c = ctx.reg_alloc.ScratchXmm();
         code.vpsraq(c, a, 32);
         code.vpsllq(a, a, 32);
         code.vpsraq(a, a, 32);
@@ -5461,7 +5461,7 @@ void EmitX64::EmitVectorTableLookup128(EmitContext& ctx, IR::Inst* inst) {
     if (code.HasHostFeature(HostFeature::AVX512_Ortho | HostFeature::AVX512BW)) {
         const Xbyak::Xmm indicies = ctx.reg_alloc.UseXmm(args[2]);
         const Xbyak::Xmm result = ctx.reg_alloc.UseScratchXmm(args[0]);
-        const Xbyak::Xmm masked = xmm16;
+        const Xbyak::Xmm masked = ctx.reg_alloc.ScratchXmm();
 
         code.vpandd(masked, indicies, code.Const(xword_b, 0xF0F0F0F0F0F0F0F0, 0xF0F0F0F0F0F0F0F0));
 

src/dynarmic/src/dynarmic/backend/x64/reg_alloc.cpp

@@ -9,6 +9,7 @@
 #include "dynarmic/backend/x64/reg_alloc.h"
 
 #include <algorithm>
+#include <limits>
 #include <numeric>
 #include <utility>
 
@@ -118,7 +119,7 @@ void HostLocInfo::AddValue(IR::Inst* inst) noexcept {
     values.push_back(inst);
     ASSERT(size_t(total_uses) + inst->UseCount() < (std::numeric_limits<uint16_t>::max)());
     total_uses += inst->UseCount();
-    max_bit_width = std::max<uint8_t>(max_bit_width, GetBitWidth(inst->GetType()));
+    max_bit_width = std::max<uint8_t>(max_bit_width, std::countr_zero(GetBitWidth(inst->GetType())));
 }
 
 void HostLocInfo::EmitVerboseDebuggingOutput(BlockOfCode* code, size_t host_loc_index) const noexcept {
@@ -152,19 +153,19 @@ bool Argument::GetImmediateU1() const noexcept {
 
 u8 Argument::GetImmediateU8() const noexcept {
     const u64 imm = value.GetImmediateAsU64();
-    ASSERT(imm < 0x100);
+    ASSERT(imm <= u64(std::numeric_limits<u8>::max()));
     return u8(imm);
 }
 
 u16 Argument::GetImmediateU16() const noexcept {
     const u64 imm = value.GetImmediateAsU64();
-    ASSERT(imm < 0x10000);
+    ASSERT(imm <= u64(std::numeric_limits<u16>::max()));
     return u16(imm);
 }
 
 u32 Argument::GetImmediateU32() const noexcept {
     const u64 imm = value.GetImmediateAsU64();
-    ASSERT(imm < 0x100000000);
+    ASSERT(imm <= u64(std::numeric_limits<u32>::max()));
     return u32(imm);
 }
 
@@ -366,10 +367,20 @@ void RegAlloc::HostCall(IR::Inst* result_def,
     if (result_def) {
         DefineValueImpl(result_def, ABI_RETURN);
     }
-
+    for (size_t i = 0; i < args.size(); i++) {
+        if (args[i]) {
+            UseScratch(*args[i], args_hostloc[i]);
+        } else {
+            ScratchGpr(args_hostloc[i]); // TODO: Force spill
+        }
+    }
+    // Must match with with ScratchImpl
+    for (auto const gpr : other_caller_save) {
+        MoveOutOfTheWay(gpr);
+        LocInfo(gpr).WriteLock();
+    }
     for (size_t i = 0; i < args.size(); i++) {
         if (args[i] && !args[i]->get().IsVoid()) {
-            UseScratch(*args[i], args_hostloc[i]);
             // LLVM puts the burden of zero-extension of 8 and 16 bit values on the caller instead of the callee
             const Xbyak::Reg64 reg = HostLocToReg64(args_hostloc[i]);
             switch (args[i]->get().GetType()) {
@@ -389,14 +400,6 @@ void RegAlloc::HostCall(IR::Inst* result_def,
             }
         }
     }
-
-    for (size_t i = 0; i < args.size(); i++)
-        if (!args[i]) {
-            // TODO: Force spill
-            ScratchGpr(args_hostloc[i]);
-        }
-    for (auto const caller_saved : other_caller_save)
-        ScratchImpl({caller_saved});
 }
 
 void RegAlloc::AllocStackSpace(const size_t stack_space) noexcept {
@@ -559,13 +562,12 @@ void RegAlloc::SpillRegister(HostLoc loc) noexcept {
 }
 
 HostLoc RegAlloc::FindFreeSpill(bool is_xmm) const noexcept {
-#if 0
     // TODO(lizzie): Ok, Windows hates XMM spills, this means less perf for windows
     // but it's fine anyways. We can find other ways to cheat it later - but which?!?!
     // we should NOT save xmm each block entering... MAYBE xbyak has a bug on start/end?
     // TODO(lizzie): This needs to be investigated further later.
     // Do not spill XMM into other XMM silly
-    if (!is_xmm) {
+    /*if (!is_xmm) {
         // TODO(lizzie): Using lower (xmm0 and such) registers results in issues/crashes - INVESTIGATE WHY
         // Intel recommends to spill GPR onto XMM registers IF POSSIBLE
         // TODO(lizzie): Issues on DBZ, theory: Scratch XMM not properly restored after a function call?
@@ -573,8 +575,9 @@ HostLoc RegAlloc::FindFreeSpill(bool is_xmm) const noexcept {
         for (size_t i = size_t(HostLoc::XMM15); i >= size_t(HostLoc::XMM3); --i)
             if (const auto loc = HostLoc(i); LocInfo(loc).IsEmpty())
                 return loc;
-    }
-#endif
+    }*/
+    // TODO: Doing this would mean saving XMM on each call... need to benchmark the benefits
+    // of spilling on XMM versus the potential cost of using XMM registers.....
     // Otherwise go to stack spilling
     for (size_t i = size_t(HostLoc::FirstSpill); i < hostloc_info.size(); ++i)
         if (const auto loc = HostLoc(i); LocInfo(loc).IsEmpty())

src/dynarmic/src/dynarmic/backend/x64/reg_alloc.h

@@ -12,6 +12,7 @@
 #include <functional>
 #include <optional>
 
+#include "boost/container/small_vector.hpp"
 #include "dynarmic/common/common_types.h"
 #include <xbyak/xbyak.h>
 #include <boost/container/static_vector.hpp>
@@ -77,13 +78,13 @@ public:
         return std::find(values.begin(), values.end(), inst) != values.end();
     }
     inline size_t GetMaxBitWidth() const noexcept {
-        return max_bit_width;
+        return 1 << max_bit_width;
     }
     void AddValue(IR::Inst* inst) noexcept;
     void EmitVerboseDebuggingOutput(BlockOfCode* code, size_t host_loc_index) const noexcept;
 private:
 //non trivial
-    std::vector<IR::Inst*> values; //24
+    boost::container::small_vector<IR::Inst*, 3> values; //24
     // Block state
     uint16_t total_uses = 0; //8
     //sometimes zeroed
@@ -93,10 +94,10 @@ private:
     uint16_t is_being_used_count = 0; //8
     uint16_t current_references = 0; //8
     // Value state
-    uint8_t max_bit_width = 0; //Valid values: 1,2,4,8,16,32,128
+    uint8_t max_bit_width : 4 = 0; //Valid values: log2(1,2,4,8,16,32,128) = (0, 1, 2, 3, 4, 5, 6)
+    uint8_t lru_counter : 2 = 0; //1
     bool is_scratch : 1 = false; //1
     bool is_set_last_use : 1 = false; //1
-    alignas(16) uint8_t lru_counter = 0; //1
     friend class RegAlloc;
 };
 static_assert(sizeof(HostLocInfo) == 64);

src/dynarmic/src/dynarmic/common/context.h

@@ -0,0 +1,120 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#ifdef __APPLE__
+#    include <signal.h>
+#    include <sys/ucontext.h>
+#else
+#    include <signal.h>
+#    ifndef __OpenBSD__
+#        include <ucontext.h>
+#    endif
+#    ifdef __sun__
+#        include <sys/regset.h>
+#    endif
+#    ifdef __linux__
+#       include <sys/syscall.h>
+#    endif
+#endif
+
+#ifdef ARCHITECTURE_x86_64
+#   ifdef __OpenBSD__
+#       define CTX_DECLARE(raw_context) ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(raw_context);
+#   else
+#       define CTX_DECLARE(raw_context) \
+        ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(raw_context);  \
+        [[maybe_unused]] auto& mctx = ucontext->uc_mcontext;
+#   endif
+#elif defined(ARCHITECTURE_arm64)
+#   ifdef __OpenBSD__
+#       define CTX_DECLARE(raw_context) ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(raw_context);
+#   else
+#       define CTX_DECLARE(raw_context) \
+        ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(raw_context);  \
+        [[maybe_unused]] auto& mctx = ucontext->uc_mcontext; \
+        [[maybe_unused]] const auto fpctx = GetFloatingPointState(mctx);
+#   endif
+#endif
+
+#if defined(ARCHITECTURE_x86_64)
+#    if defined(__APPLE__)
+#        define CTX_RIP (mctx->__ss.__rip)
+#        define CTX_RSP (mctx->__ss.__rsp)
+#    elif defined(__linux__)
+#        define CTX_RIP (mctx.gregs[REG_RIP])
+#        define CTX_RSP (mctx.gregs[REG_RSP])
+#    elif defined(__FreeBSD__)
+#        define CTX_RIP (mctx.mc_rip)
+#        define CTX_RSP (mctx.mc_rsp)
+#    elif defined(__NetBSD__)
+#        define CTX_RIP (mctx.__gregs[_REG_RIP])
+#        define CTX_RSP (mctx.__gregs[_REG_RSP])
+#    elif defined(__OpenBSD__)
+#        define CTX_RIP (ucontext->sc_rip)
+#        define CTX_RSP (ucontext->sc_rsp)
+#    elif defined(__sun__)
+#        define CTX_RIP (mctx.gregs[REG_RIP])
+#        define CTX_RSP (mctx.gregs[REG_RSP])
+#    else
+#        error "Unknown platform"
+#    endif
+#elif defined(ARCHITECTURE_arm64)
+#    if defined(__APPLE__)
+#        define CTX_PC (mctx->__ss.__pc)
+#        define CTX_SP (mctx->__ss.__sp)
+#        define CTX_LR (mctx->__ss.__lr)
+#        define CTX_PSTATE (mctx->__ss.__cpsr)
+#        define CTX_X(i) (mctx->__ss.__x[i])
+#        define CTX_Q(i) (mctx->__ns.__v[i])
+#        define CTX_FPSR (mctx->__ns.__fpsr)
+#        define CTX_FPCR (mctx->__ns.__fpcr)
+#    elif defined(__linux__)
+#        define CTX_PC (mctx.pc)
+#        define CTX_SP (mctx.sp)
+#        define CTX_LR (mctx.regs[30])
+#        define CTX_PSTATE (mctx.pstate)
+#        define CTX_X(i) (mctx.regs[i])
+#        define CTX_Q(i) (fpctx->vregs[i])
+#        define CTX_FPSR (fpctx->fpsr)
+#        define CTX_FPCR (fpctx->fpcr)
+#    elif defined(__FreeBSD__)
+#        define CTX_PC (mctx.mc_gpregs.gp_elr)
+#        define CTX_SP (mctx.mc_gpregs.gp_sp)
+#        define CTX_LR (mctx.mc_gpregs.gp_lr)
+#        define CTX_X(i) (mctx.mc_gpregs.gp_x[i])
+#        define CTX_Q(i) (mctx.mc_fpregs.fp_q[i])
+#    elif defined(__NetBSD__)
+#        define CTX_PC (mctx.mc_gpregs.gp_elr)
+#        define CTX_SP (mctx.mc_gpregs.gp_sp)
+#        define CTX_LR (mctx.mc_gpregs.gp_lr)
+#        define CTX_X(i) (mctx.mc_gpregs.gp_x[i])
+#        define CTX_Q(i) (mctx.mc_fpregs.fp_q[i])
+#    elif defined(__OpenBSD__)
+#        define CTX_PC (ucontext->sc_elr)
+#        define CTX_SP (ucontext->sc_sp)
+#        define CTX_LR (ucontext->sc_lr)
+#        define CTX_X(i) (ucontext->sc_x[i])
+#        define CTX_Q(i) (ucontext->sc_q[i])
+#    else
+#        error "Unknown platform"
+#    endif
+#else
+#    error "unimplemented"
+#endif
+
+#ifdef ARCHITECTURE_arm64
+#ifdef __APPLE__
+inline _STRUCT_ARM_NEON_STATE64* GetFloatingPointState(mcontext_t& host_ctx) {
+    return &(host_ctx->__ns);
+}
+#elif defined(__linux__)
+inline fpsimd_context* GetFloatingPointState(mcontext_t& host_ctx) {
+    _aarch64_ctx* header = reinterpret_cast<_aarch64_ctx*>(&host_ctx.__reserved);
+    while (header->magic != FPSIMD_MAGIC)
+        header = reinterpret_cast<_aarch64_ctx*>(reinterpret_cast<char*>(header) + header->size);
+    return reinterpret_cast<fpsimd_context*>(header);
+}
+#endif
+#endif

src/dynarmic/src/dynarmic/common/memory_pool.cpp

@@ -1,13 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/common/memory_pool.h"
-
-#include <cstdlib>
-
-namespace Dynarmic::Common {
-
-
-}  // namespace Dynarmic::Common

src/dynarmic/src/dynarmic/common/memory_pool.h

@@ -1,61 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#pragma once
-
-#include <cstddef>
-#include <vector>
-
-namespace Dynarmic::Common {
-
-/// @tparam object_size Byte-size of objects to construct
-/// @tparam slab_size Number of objects to have per slab
-template<size_t object_size, size_t slab_size>
-class Pool {
-public:
-    inline Pool() noexcept {
-        AllocateNewSlab();
-    }
-    inline ~Pool() noexcept {
-        std::free(current_slab);
-        for (char* slab : slabs) {
-            std::free(slab);
-        }
-    }
-
-    Pool(const Pool&) = delete;
-    Pool(Pool&&) = delete;
-
-    Pool& operator=(const Pool&) = delete;
-    Pool& operator=(Pool&&) = delete;
-
-    /// @brief Returns a pointer to an `object_size`-bytes block of memory.
-    [[nodiscard]] void* Alloc() noexcept {
-        if (remaining == 0) {
-            slabs.push_back(current_slab);
-            AllocateNewSlab();
-        }
-        void* ret = static_cast<void*>(current_ptr);
-        current_ptr += object_size;
-        remaining--;
-        return ret;
-    }
-private:
-    /// @brief Allocates a completely new memory slab.
-    /// Used when an entirely new slab is needed
-    /// due the current one running out of usable space.
-    void AllocateNewSlab() noexcept {
-        current_slab = static_cast<char*>(std::malloc(object_size * slab_size));
-        current_ptr = current_slab;
-        remaining = slab_size;
-    }
-
-    std::vector<char*> slabs;
-    char* current_slab = nullptr;
-    char* current_ptr = nullptr;
-    size_t remaining = 0;
-};
-
-}  // namespace Dynarmic::Common

src/dynarmic/src/dynarmic/common/variant_util.h

@@ -1,29 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#pragma once
-
-#include <boost/variant.hpp>
-
-namespace Dynarmic::Common {
-namespace detail {
-
-template<typename ReturnT, typename Lambda>
-struct VariantVisitor : boost::static_visitor<ReturnT>
-        , Lambda {
-    VariantVisitor(Lambda&& lambda)
-            : Lambda(std::move(lambda)) {}
-
-    using Lambda::operator();
-};
-
-}  // namespace detail
-
-template<typename ReturnT, typename Variant, typename Lambda>
-inline ReturnT VisitVariant(Variant&& variant, Lambda&& lambda) {
-    return boost::apply_visitor(detail::VariantVisitor<ReturnT, Lambda>(std::move(lambda)), variant);
-}
-
-}  // namespace Dynarmic::Common

src/dynarmic/src/dynarmic/frontend/A32/a32_types.h

@@ -9,12 +9,9 @@
 #pragma once
 
 #include <string>
-#include <utility>
-
 #include <fmt/format.h>
 #include "dynarmic/common/assert.h"
 #include "dynarmic/common/common_types.h"
-
 #include "dynarmic/interface/A32/coprocessor_util.h"
 #include "dynarmic/ir/cond.h"
 
@@ -89,24 +86,17 @@ constexpr bool IsQuadExtReg(ExtReg reg) {
 
 inline size_t RegNumber(Reg reg) {
     ASSERT(reg != Reg::INVALID_REG);
-    return static_cast<size_t>(reg);
+    return size_t(reg);
 }
 
 inline size_t RegNumber(ExtReg reg) {
     if (IsSingleExtReg(reg)) {
-        return static_cast<size_t>(reg) - static_cast<size_t>(ExtReg::S0);
+        return size_t(reg) - size_t(ExtReg::S0);
+    } else if (IsDoubleExtReg(reg)) {
+        return size_t(reg) - size_t(ExtReg::D0);
     }
-
-    if (IsDoubleExtReg(reg)) {
-        return static_cast<size_t>(reg) - static_cast<size_t>(ExtReg::D0);
-    }
-
-    if (IsQuadExtReg(reg)) {
-        return static_cast<size_t>(reg) - static_cast<size_t>(ExtReg::Q0);
-    }
-
-    ASSERT_MSG(false, "Invalid extended register");
-    return 0;
+    ASSERT(IsQuadExtReg(reg));
+    return size_t(reg) - size_t(ExtReg::Q0);
 }
 
 inline Reg operator+(Reg reg, size_t number) {

src/dynarmic/src/dynarmic/frontend/A32/decoder/arm.h

@@ -30,13 +30,13 @@ template<typename Visitor>
 using ArmDecodeTable = std::array<std::vector<ArmMatcher<Visitor>>, 0x1000>;
 
 namespace detail {
-inline size_t ToFastLookupIndexArm(u32 instruction) {
+inline size_t ToFastLookupIndexArm(u32 instruction) noexcept {
     return ((instruction >> 4) & 0x00F) | ((instruction >> 16) & 0xFF0);
 }
 }  // namespace detail
 
 template<typename V>
-constexpr ArmDecodeTable<V> GetArmDecodeTable() {
+constexpr ArmDecodeTable<V> GetArmDecodeTable() noexcept {
     std::vector<ArmMatcher<V>> list = {
 #define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(ArmMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)),
 #include "./arm.inc"
@@ -62,15 +62,27 @@ constexpr ArmDecodeTable<V> GetArmDecodeTable() {
 }
 
 template<typename V>
-std::optional<std::reference_wrapper<const ArmMatcher<V>>> DecodeArm(u32 instruction) {
+std::optional<std::reference_wrapper<const ArmMatcher<V>>> DecodeArm(u32 instruction) noexcept {
     alignas(64) static const auto table = GetArmDecodeTable<V>();
     const auto matches_instruction = [instruction](const auto& matcher) {
         return matcher.Matches(instruction);
     };
-
     const auto& subtable = table[detail::ToFastLookupIndexArm(instruction)];
     auto iter = std::find_if(subtable.begin(), subtable.end(), matches_instruction);
     return iter != subtable.end() ? std::optional<std::reference_wrapper<const ArmMatcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetNameARM(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, ArmMatcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(ArmMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
+#include "./arm.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A32

src/dynarmic/src/dynarmic/frontend/A32/decoder/asimd.h

@@ -26,15 +26,12 @@ template<typename Visitor>
 using ASIMDMatcher = Decoder::Matcher<Visitor, u32>;
 
 template<typename V>
-std::vector<ASIMDMatcher<V>> GetASIMDDecodeTable() {
-    std::vector<ASIMDMatcher<V>> table = {
-
-#define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(ASIMDMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)),
+std::vector<ASIMDMatcher<V>> GetASIMDDecodeTable() noexcept {
+    std::vector<std::pair<const char*, ASIMDMatcher<V>>> table = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(ASIMDMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
 #include "./asimd.inc"
 #undef INST
-
     };
-
     // Exceptions to the rule of thumb.
     const std::set<std::string> comes_first{
         "VBIC, VMOV, VMVN, VORR (immediate)",
@@ -53,29 +50,43 @@ std::vector<ASIMDMatcher<V>> GetASIMDDecodeTable() {
         "VQDMULH (scalar)",
         "VQRDMULH (scalar)",
     };
-    const auto sort_begin = std::stable_partition(table.begin(), table.end(), [&](const auto& matcher) {
-        return comes_first.count(matcher.GetName()) > 0;
+    const auto sort_begin = std::stable_partition(table.begin(), table.end(), [&](const auto& e) {
+        return comes_first.count(e.first) > 0;
     });
-    const auto sort_end = std::stable_partition(table.begin(), table.end(), [&](const auto& matcher) {
-        return comes_last.count(matcher.GetName()) == 0;
+    const auto sort_end = std::stable_partition(table.begin(), table.end(), [&](const auto& e) {
+        return comes_last.count(e.first) == 0;
     });
-
     // If a matcher has more bits in its mask it is more specific, so it should come first.
-    std::stable_sort(sort_begin, sort_end, [](const auto& matcher1, const auto& matcher2) {
-        return mcl::bit::count_ones(matcher1.GetMask()) > mcl::bit::count_ones(matcher2.GetMask());
+    std::stable_sort(sort_begin, sort_end, [](const auto& a, const auto& b) {
+        return mcl::bit::count_ones(a.second.GetMask()) > mcl::bit::count_ones(b.second.GetMask());
     });
-
-    return table;
+    std::vector<ASIMDMatcher<V>> final_table;
+    std::transform(table.cbegin(), table.cend(), std::back_inserter(final_table), [](auto const& e) {
+        return e.second;
+    });
+    return final_table;
 }
 
 template<typename V>
-std::optional<std::reference_wrapper<const ASIMDMatcher<V>>> DecodeASIMD(u32 instruction) {
-    static const auto table = GetASIMDDecodeTable<V>();
-
-    const auto matches_instruction = [instruction](const auto& matcher) { return matcher.Matches(instruction); };
-
-    auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
+std::optional<std::reference_wrapper<const ASIMDMatcher<V>>> DecodeASIMD(u32 instruction) noexcept {
+    alignas(64) static const auto table = GetASIMDDecodeTable<V>();
+    auto iter = std::find_if(table.begin(), table.end(), [instruction](const auto& matcher) {
+        return matcher.Matches(instruction);
+    });
     return iter != table.end() ? std::optional<std::reference_wrapper<const ASIMDMatcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetNameASIMD(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, ASIMDMatcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(ASIMDMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
+#include "./asimd.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A32

src/dynarmic/src/dynarmic/frontend/A32/decoder/thumb16.h

@@ -25,18 +25,28 @@ using Thumb16Matcher = Decoder::Matcher<Visitor, u16>;
 
 template<typename V>
 std::optional<std::reference_wrapper<const Thumb16Matcher<V>>> DecodeThumb16(u16 instruction) {
-    static const std::vector<Thumb16Matcher<V>> table = {
-
+    alignas(64) static const std::vector<Thumb16Matcher<V>> table = {
 #define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(Thumb16Matcher, fn, name, Decoder::detail::StringToArray<16>(bitstring)),
 #include "./thumb16.inc"
 #undef INST
-
     };
-
-    const auto matches_instruction = [instruction](const auto& matcher) { return matcher.Matches(instruction); };
-
-    auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
+    auto iter = std::find_if(table.begin(), table.end(), [instruction](const auto& matcher) {
+        return matcher.Matches(instruction);
+    });
     return iter != table.end() ? std::optional<std::reference_wrapper<const Thumb16Matcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetNameThumb16(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, Thumb16Matcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(Thumb16Matcher, fn, name, Decoder::detail::StringToArray<16>(bitstring)) },
+#include "./thumb16.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A32

src/dynarmic/src/dynarmic/frontend/A32/decoder/thumb32.h

@@ -24,18 +24,28 @@ using Thumb32Matcher = Decoder::Matcher<Visitor, u32>;
 
 template<typename V>
 std::optional<std::reference_wrapper<const Thumb32Matcher<V>>> DecodeThumb32(u32 instruction) {
-    static const std::vector<Thumb32Matcher<V>> table = {
-
+    alignas(64) static const std::vector<Thumb32Matcher<V>> table = {
 #define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(Thumb32Matcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)),
 #include "./thumb32.inc"
 #undef INST
-
     };
-
-    const auto matches_instruction = [instruction](const auto& matcher) { return matcher.Matches(instruction); };
-
-    auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
+    auto iter = std::find_if(table.begin(), table.end(), [instruction](const auto& matcher) {
+        return matcher.Matches(instruction);
+    });
     return iter != table.end() ? std::optional<std::reference_wrapper<const Thumb32Matcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetNameThumb32(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, Thumb32Matcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(Thumb32Matcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
+#include "./thumb32.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A32

src/dynarmic/src/dynarmic/frontend/A32/decoder/vfp.h

@@ -26,36 +26,42 @@ using VFPMatcher = Decoder::Matcher<Visitor, u32>;
 template<typename V>
 std::optional<std::reference_wrapper<const VFPMatcher<V>>> DecodeVFP(u32 instruction) {
     using Table = std::vector<VFPMatcher<V>>;
-
-    static const struct Tables {
+    alignas(64) static const struct Tables {
         Table unconditional;
         Table conditional;
-    } tables = [] {
+    } tables = []() {
         Table list = {
-
 #define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(VFPMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)),
 #include "./vfp.inc"
 #undef INST
-
         };
-
-        const auto division = std::stable_partition(list.begin(), list.end(), [&](const auto& matcher) {
+        auto const it = std::stable_partition(list.begin(), list.end(), [&](const auto& matcher) {
             return (matcher.GetMask() & 0xF0000000) == 0xF0000000;
         });
-
         return Tables{
-            Table{list.begin(), division},
-            Table{division, list.end()},
+            Table{list.begin(), it},
+            Table{it, list.end()},
         };
     }();
-
     const bool is_unconditional = (instruction & 0xF0000000) == 0xF0000000;
     const Table& table = is_unconditional ? tables.unconditional : tables.conditional;
-
-    const auto matches_instruction = [instruction](const auto& matcher) { return matcher.Matches(instruction); };
-
-    auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
+    auto iter = std::find_if(table.begin(), table.end(), [instruction](const auto& matcher) {
+        return matcher.Matches(instruction);
+    });
     return iter != table.end() ? std::optional<std::reference_wrapper<const VFPMatcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetNameVFP(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, VFPMatcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(VFPMatcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
+#include "./vfp.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A32

src/dynarmic/src/dynarmic/frontend/A32/translate/translate_thumb.cpp

@@ -97,7 +97,7 @@ u32 ConvertASIMDInstruction(u32 thumb_instruction) {
     return 0xF7F0A000;  // UDF
 }
 
-bool MaybeVFPOrASIMDInstruction(u32 thumb_instruction) {
+inline bool MaybeVFPOrASIMDInstruction(u32 thumb_instruction) noexcept {
     return (thumb_instruction & 0xEC000000) == 0xEC000000 || (thumb_instruction & 0xFF100000) == 0xF9000000;
 }
 

src/dynarmic/src/dynarmic/frontend/A64/decoder/a64.h

@@ -37,34 +37,31 @@ inline size_t ToFastLookupIndex(u32 instruction) {
 
 template<typename V>
 constexpr DecodeTable<V> GetDecodeTable() {
-    std::vector<Matcher<V>> list = {
-#define INST(fn, name, bitstring) DYNARMIC_DECODER_GET_MATCHER(Matcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)),
+    std::vector<std::pair<const char*, Matcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(Matcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
 #include "./a64.inc"
 #undef INST
     };
-
     // If a matcher has more bits in its mask it is more specific, so it should come first.
-    std::stable_sort(list.begin(), list.end(), [](const auto& matcher1, const auto& matcher2) {
+    std::stable_sort(list.begin(), list.end(), [](const auto& a, const auto& b) {
         // If a matcher has more bits in its mask it is more specific, so it should come first.
-        return mcl::bit::count_ones(matcher1.GetMask()) > mcl::bit::count_ones(matcher2.GetMask());
+        return mcl::bit::count_ones(a.second.GetMask()) > mcl::bit::count_ones(b.second.GetMask());
     });
-
     // Exceptions to the above rule of thumb.
-    std::stable_partition(list.begin(), list.end(), [&](const auto& matcher) {
+    std::stable_partition(list.begin(), list.end(), [&](const auto& e) {
         return std::set<std::string>{
             "MOVI, MVNI, ORR, BIC (vector, immediate)",
             "FMOV (vector, immediate)",
             "Unallocated SIMD modified immediate",
-        }.count(matcher.GetName()) > 0;
+        }.count(e.first) > 0;
     });
-
     DecodeTable<V> table{};
     for (size_t i = 0; i < table.size(); ++i) {
-        for (auto matcher : list) {
-            const auto expect = detail::ToFastLookupIndex(matcher.GetExpected());
-            const auto mask = detail::ToFastLookupIndex(matcher.GetMask());
+        for (auto const& e : list) {
+            const auto expect = detail::ToFastLookupIndex(e.second.GetExpected());
+            const auto mask = detail::ToFastLookupIndex(e.second.GetMask());
             if ((i & mask) == expect) {
-                table[i].push_back(matcher);
+                table[i].push_back(e.second);
             }
         }
     }
@@ -74,12 +71,24 @@ constexpr DecodeTable<V> GetDecodeTable() {
 template<typename V>
 std::optional<std::reference_wrapper<const Matcher<V>>> Decode(u32 instruction) {
     alignas(64) static const auto table = GetDecodeTable<V>();
-    const auto matches_instruction = [instruction](const auto& matcher) {
-        return matcher.Matches(instruction);
-    };
     const auto& subtable = table[detail::ToFastLookupIndex(instruction)];
-    auto iter = std::find_if(subtable.begin(), subtable.end(), matches_instruction);
+    auto iter = std::find_if(subtable.begin(), subtable.end(), [instruction](const auto& matcher) {
+        return matcher.Matches(instruction);
+    });
     return iter != subtable.end() ? std::optional<std::reference_wrapper<const Matcher<V>>>(*iter) : std::nullopt;
 }
 
+template<typename V>
+std::optional<std::string_view> GetName(u32 inst) noexcept {
+    std::vector<std::pair<std::string_view, Matcher<V>>> list = {
+#define INST(fn, name, bitstring) { name, DYNARMIC_DECODER_GET_MATCHER(Matcher, fn, name, Decoder::detail::StringToArray<32>(bitstring)) },
+#include "./a64.inc"
+#undef INST
+    };
+    auto const iter = std::find_if(list.cbegin(), list.cend(), [inst](auto const& m) {
+        return m.second.Matches(inst);
+    });
+    return iter != list.cend() ? std::optional{iter->first} : std::nullopt;
+}
+
 }  // namespace Dynarmic::A64

src/dynarmic/src/dynarmic/frontend/A64/translate/impl/a64_branch.cpp

@@ -20,9 +20,12 @@ bool TranslatorVisitor::B_cond(Imm<19> imm19, Cond cond) {
 bool TranslatorVisitor::B_uncond(Imm<26> imm26) {
     const s64 offset = concatenate(imm26, Imm<2>{0}).SignExtend<s64>();
     const u64 target = ir.PC() + offset;
-
-    //ir.SetTerm(IR::Term::LinkBlockFast{ir.current_location->SetPC(target)});
-    ir.SetTerm(IR::Term::LinkBlock{ir.current_location->SetPC(target)});
+    // Pattern to halt execution (B .)
+    if (target == ir.PC()) {
+        ir.SetTerm(IR::Term::LinkBlock{ir.current_location->SetPC(target)});
+        return false;
+    }
+    ir.SetTerm(IR::Term::LinkBlockFast{ir.current_location->SetPC(target)});
     return false;
 }
 

src/dynarmic/src/dynarmic/frontend/decoder/decoder_detail.h

@@ -70,11 +70,9 @@ struct detail {
         return std::make_tuple(mask, expect);
     }
 
-    /**
-     * Generates the masks and shifts for each argument.
-     * A '-' in a bitstring indicates that we don't care about that value.
-     * An argument is specified by a continuous string of the same character.
-     */
+    /// @brief Generates the masks and shifts for each argument.
+    /// A '-' in a bitstring indicates that we don't care about that value.
+    /// An argument is specified by a continuous string of the same character.
     template<size_t N>
     static consteval auto GetArgInfo(std::array<char, opcode_bitsize> bitstring) {
         std::array<opcode_type, N> masks = {};
@@ -98,7 +96,6 @@ struct detail {
 
                 if constexpr (N > 0) {
                     const size_t bit_position = opcode_bitsize - i - 1;
-
                     if (arg_index >= N)
                         throw std::out_of_range("Unexpected field");
 
@@ -109,20 +106,16 @@ struct detail {
                 }
             }
         }
-
 #if !defined(DYNARMIC_IGNORE_ASSERTS) && !defined(__ANDROID__)
         // Avoids a MSVC ICE, and avoids Android NDK issue.
         ASSERT(std::all_of(masks.begin(), masks.end(), [](auto m) { return m != 0; }));
 #endif
-
         return std::make_tuple(masks, shifts);
     }
 
-    /**
-     * This struct's Make member function generates a lambda which decodes an instruction based on
-     * the provided arg_masks and arg_shifts. The Visitor member function to call is provided as a
-     * template argument.
-     */
+    /// @brief This struct's Make member function generates a lambda which decodes an instruction
+    /// based on the provided arg_masks and arg_shifts. The Visitor member function to call is
+    /// provided as a template argument.
     template<typename FnT>
     struct VisitorCaller;
 
@@ -130,36 +123,36 @@ struct detail {
 #    pragma warning(push)
 #    pragma warning(disable : 4800)  // forcing value to bool 'true' or 'false' (performance warning)
 #endif
-    template<typename Visitor, typename... Args, typename CallRetT>
-    struct VisitorCaller<CallRetT (Visitor::*)(Args...)> {
+    template<typename V, typename... Args, typename ReturnType>
+    struct VisitorCaller<ReturnType (V::*)(Args...)> {
         template<size_t... iota>
-        static auto Make(std::integer_sequence<size_t, iota...>,
-                         CallRetT (Visitor::*const fn)(Args...),
+        static constexpr auto Make(std::integer_sequence<size_t, iota...>,
+                         ReturnType (V::*const fn)(Args...),
                          const std::array<opcode_type, sizeof...(iota)> arg_masks,
                          const std::array<size_t, sizeof...(iota)> arg_shifts) {
-            static_assert(std::is_same_v<visitor_type, Visitor>, "Member function is not from Matcher's Visitor");
-            return [fn, arg_masks, arg_shifts](Visitor& v, opcode_type instruction) {
+            static_assert(std::is_same_v<visitor_type, V>, "Member function is not from Matcher's Visitor");
+            return [fn, arg_masks, arg_shifts](V& v, opcode_type instruction) {
                 (void)instruction;
                 (void)arg_masks;
                 (void)arg_shifts;
-                return (v.*fn)(static_cast<Args>((instruction & arg_masks[iota]) >> arg_shifts[iota])...);
+                return (v.*fn)(Args((instruction & arg_masks[iota]) >> arg_shifts[iota])...);
             };
         }
     };
 
-    template<typename Visitor, typename... Args, typename CallRetT>
-    struct VisitorCaller<CallRetT (Visitor::*)(Args...) const> {
+    template<typename V, typename... Args, typename ReturnType>
+    struct VisitorCaller<ReturnType (V::*)(Args...) const> {
         template<size_t... iota>
-        static auto Make(std::integer_sequence<size_t, iota...>,
-                         CallRetT (Visitor::*const fn)(Args...) const,
+        static constexpr auto Make(std::integer_sequence<size_t, iota...>,
+                         ReturnType (V::*const fn)(Args...) const,
                          const std::array<opcode_type, sizeof...(iota)> arg_masks,
                          const std::array<size_t, sizeof...(iota)> arg_shifts) {
-            static_assert(std::is_same_v<visitor_type, const Visitor>, "Member function is not from Matcher's Visitor");
-            return [fn, arg_masks, arg_shifts](const Visitor& v, opcode_type instruction) {
+            static_assert(std::is_same_v<visitor_type, const V>, "Member function is not from Matcher's Visitor");
+            return [fn, arg_masks, arg_shifts](const V& v, opcode_type instruction) {
                 (void)instruction;
                 (void)arg_masks;
                 (void)arg_shifts;
-                return (v.*fn)(static_cast<Args>((instruction & arg_masks[iota]) >> arg_shifts[iota])...);
+                return (v.*fn)(Args((instruction & arg_masks[iota]) >> arg_shifts[iota])...);
             };
         }
     };
@@ -167,27 +160,21 @@ struct detail {
 #    pragma warning(pop)
 #endif
 
-    /**
-     * Creates a matcher that can match and parse instructions based on bitstring.
-     * See also: GetMaskAndExpect and GetArgInfo for format of bitstring.
-     */
-    template<auto bitstring, typename FnT>
-    static auto GetMatcher(FnT fn, const char* const name) {
-        constexpr size_t args_count = mcl::parameter_count_v<FnT>;
-
+    /// @brief Creates a matcher that can match and parse instructions based on bitstring.
+    /// See also: GetMaskAndExpect and GetArgInfo for format of bitstring.
+    template<auto bitstring, typename F>
+    static constexpr auto GetMatcher(F fn) {
+        constexpr size_t args_count = mcl::parameter_count_v<F>;
         constexpr auto mask = std::get<0>(GetMaskAndExpect(bitstring));
         constexpr auto expect = std::get<1>(GetMaskAndExpect(bitstring));
         constexpr auto arg_masks = std::get<0>(GetArgInfo<args_count>(bitstring));
         constexpr auto arg_shifts = std::get<1>(GetArgInfo<args_count>(bitstring));
-
-        using Iota = std::make_index_sequence<args_count>;
-
-        const auto proxy_fn = VisitorCaller<FnT>::Make(Iota(), fn, arg_masks, arg_shifts);
-        return MatcherT(name, mask, expect, proxy_fn);
+        const auto proxy_fn = VisitorCaller<F>::Make(std::make_index_sequence<args_count>(), fn, arg_masks, arg_shifts);
+        return MatcherT(mask, expect, proxy_fn);
     }
 };
 
-#define DYNARMIC_DECODER_GET_MATCHER(MatcherT, fn, name, bitstring) Decoder::detail::detail<MatcherT<V>>::template GetMatcher<bitstring>(&V::fn, name)
+#define DYNARMIC_DECODER_GET_MATCHER(MatcherT, fn, name, bitstring) Decoder::detail::detail<MatcherT<V>>::template GetMatcher<bitstring>(&V::fn)
 
 }  // namespace detail
 }  // namespace Dynarmic::Decoder

src/dynarmic/src/dynarmic/frontend/decoder/matcher.h

@@ -14,16 +14,12 @@
 
 namespace Dynarmic::Decoder {
 
-/**
- * Generic instruction handling construct.
- *
- * @tparam Visitor An arbitrary visitor type that will be passed through
- *                 to the function being handled. This type must be the
- *                 type of the first parameter in a handler function.
- *
- * @tparam OpcodeType Type representing an opcode. This must be the
- *                    type of the second parameter in a handler function.
- */
+/// Generic instruction handling construct.
+/// @tparam Visitor An arbitrary visitor type that will be passed through
+///                 to the function being handled. This type must be the
+///                 type of the first parameter in a handler function.
+/// @tparam OpcodeType Type representing an opcode. This must be the
+///                    type of the second parameter in a handler function.
 template<typename Visitor, typename OpcodeType>
 class Matcher {
 public:
@@ -31,46 +27,35 @@ public:
     using visitor_type = Visitor;
     using handler_return_type = typename Visitor::instruction_return_type;
     using handler_function = std::function<handler_return_type(Visitor&, opcode_type)>;
-
-    Matcher(const char* const name, opcode_type mask, opcode_type expected, handler_function func)
-            : name{name}, mask{mask}, expected{expected}, fn{std::move(func)} {}
-
-    /// Gets the name of this type of instruction.
-    const char* GetName() const {
-        return name;
-    }
+    Matcher(opcode_type mask, opcode_type expected, handler_function func)
+            : mask{mask}, expected{expected}, fn{std::move(func)} {}
 
     /// Gets the mask for this instruction.
-    opcode_type GetMask() const {
+    inline opcode_type GetMask() const noexcept {
         return mask;
     }
 
     /// Gets the expected value after masking for this instruction.
-    opcode_type GetExpected() const {
+    inline opcode_type GetExpected() const noexcept {
         return expected;
     }
 
-    /**
-     * Tests to see if the given instruction is the instruction this matcher represents.
-     * @param instruction The instruction to test
-     * @returns true if the given instruction matches.
-     */
-    bool Matches(opcode_type instruction) const {
+    /// Tests to see if the given instruction is the instruction this matcher represents.
+    /// @param instruction The instruction to test
+    /// @returns true if the given instruction matches.
+    inline bool Matches(opcode_type instruction) const noexcept {
         return (instruction & mask) == expected;
     }
 
-    /**
-     * Calls the corresponding instruction handler on visitor for this type of instruction.
-     * @param v The visitor to use
-     * @param instruction The instruction to decode.
-     */
-    handler_return_type call(Visitor& v, opcode_type instruction) const {
+    /// Calls the corresponding instruction handler on visitor for this type of instruction.
+    /// @param v The visitor to use
+    /// @param instruction The instruction to decode.
+    inline handler_return_type call(Visitor& v, opcode_type instruction) const noexcept {
         ASSERT(Matches(instruction));
         return fn(v, instruction);
     }
 
 private:
-    const char* name;
     opcode_type mask;
     opcode_type expected;
     handler_function fn;

src/dynarmic/src/dynarmic/interface/optimization_flags.h

@@ -34,6 +34,8 @@ enum class OptimizationFlag : std::uint32_t {
     MiscIROpt = 0x00000020,
     /// Optimize for code speed rather than for code size (this serves well for tight loops)
     CodeSpeed = 0x00000040,
+    /// Disable verification passes
+    DisableVerification = 0x00000080,
 
     /// This is an UNSAFE optimization that reduces accuracy of fused multiply-add operations.
     /// This unfuses fused instructions to improve performance on host CPUs without FMA support.

src/dynarmic/src/dynarmic/ir/basic_block.cpp

@@ -15,8 +15,6 @@
 
 #include <fmt/format.h>
 #include "dynarmic/common/assert.h"
-
-#include "dynarmic/common/memory_pool.h"
 #include "dynarmic/frontend/A32/a32_types.h"
 #include "dynarmic/frontend/A64/a64_types.h"
 #include "dynarmic/ir/cond.h"
@@ -27,8 +25,7 @@ namespace Dynarmic::IR {
 Block::Block(const LocationDescriptor& location)
     : location{location},
     end_location{location},
-    cond{Cond::AL},
-    instruction_alloc_pool{std::make_unique<std::remove_reference_t<decltype(*instruction_alloc_pool)>>()}
+    cond{Cond::AL}
 {
 
 }
@@ -40,7 +37,21 @@ Block::Block(const LocationDescriptor& location)
 /// @param args            A sequence of Value instances used as arguments for the instruction.
 /// @returns Iterator to the newly created instruction.
 Block::iterator Block::PrependNewInst(iterator insertion_point, Opcode opcode, std::initializer_list<Value> args) noexcept {
-    IR::Inst* inst = new (instruction_alloc_pool->Alloc()) IR::Inst(opcode);
+    // First try using the "inline" buffer, otherwise fallback to a slower slab-like allocation scheme
+    // purpouse is to avoid many calls to new/delete which invoke malloc which invokes mmap
+    // just pool it!!! - reason why there is an inline buffer is because many small blocks are created
+    // with few instructions due to subpar optimisations on other passes... plus branch-heavy code will
+    // hugely benefit from the coherency of faster allocations...
+    IR::Inst* inst;
+    if (inlined_inst.size() < inlined_inst.max_size()) {
+        inst = &inlined_inst[inlined_inst.size()];
+        inlined_inst.emplace_back(opcode);
+    } else {
+        if (pooled_inst.empty() || pooled_inst.back().size() == pooled_inst.back().max_size())
+            pooled_inst.emplace_back();
+        inst = &pooled_inst.back()[pooled_inst.back().size()];
+        pooled_inst.back().emplace_back(opcode);
+    }
     DEBUG_ASSERT(args.size() == inst->NumArgs());
     std::for_each(args.begin(), args.end(), [&inst, index = size_t(0)](const auto& arg) mutable {
         inst->SetArg(index, arg);

src/dynarmic/src/dynarmic/ir/basic_block.h

@@ -13,6 +13,9 @@
 #include <optional>
 #include <string>
 
+#include <boost/container/container_fwd.hpp>
+#include <boost/container/static_vector.hpp>
+#include <boost/container/stable_vector.hpp>
 #include <mcl/container/intrusive_list.hpp>
 #include "dynarmic/common/common_types.h"
 
@@ -21,7 +24,6 @@
 #include "dynarmic/ir/terminal.h"
 #include "dynarmic/ir/value.h"
 #include "dynarmic/ir/dense_list.h"
-#include "dynarmic/common/memory_pool.h"
 
 namespace Dynarmic::IR {
 
@@ -164,8 +166,12 @@ public:
         return cycle_count;
     }
 private:
+    /// "Hot cache" for small blocks so we don't call global allocator
+    boost::container::static_vector<Inst, 14> inlined_inst;
     /// List of instructions in this block.
     instruction_list_type instructions;
+    /// "Long/far" memory pool
+    boost::container::stable_vector<boost::container::static_vector<Inst, 32>> pooled_inst;
     /// Block to execute next if `cond` did not pass.
     std::optional<LocationDescriptor> cond_failed = {};
     /// Description of the starting location of this block
@@ -174,8 +180,6 @@ private:
     LocationDescriptor end_location;
     /// Conditional to pass in order to execute this block
     Cond cond;
-    /// Memory pool for instruction list
-    std::unique_ptr<Common::Pool<sizeof(Inst), 2097152UL / sizeof(Inst)>> instruction_alloc_pool;
     /// Terminal instruction of this block.
     Terminal terminal = Term::Invalid{};
     /// Number of cycles this block takes to execute if the conditional fails.
@@ -183,6 +187,7 @@ private:
     /// Number of cycles this block takes to execute.
     size_t cycle_count = 0;
 };
+static_assert(sizeof(Block) == 2048);
 
 /// Returns a string representation of the contents of block. Intended for debugging.
 std::string DumpBlock(const IR::Block& block) noexcept;

src/dynarmic/src/dynarmic/ir/ir_emitter.cpp

@@ -1,21 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/ir/ir_emitter.h"
-
-#include <vector>
-
-#include "dynarmic/common/assert.h"
-#include <mcl/bit_cast.hpp>
-
-#include "dynarmic/ir/opcodes.h"
-
-namespace Dynarmic::IR {
-
-
-}  // namespace Dynarmic::IR

src/dynarmic/src/dynarmic/ir/opt/a32_constant_memory_reads_pass.cpp

@@ -1,70 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/interface/A32/config.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-void A32ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb) {
-    for (auto& inst : block) {
-        switch (inst.GetOpcode()) {
-        case IR::Opcode::A32ReadMemory8: {
-            if (!inst.AreAllArgsImmediates()) {
-                break;
-            }
-
-            const u32 vaddr = inst.GetArg(1).GetU32();
-            if (cb->IsReadOnlyMemory(vaddr)) {
-                const u8 value_from_memory = cb->MemoryRead8(vaddr);
-                inst.ReplaceUsesWith(IR::Value{value_from_memory});
-            }
-            break;
-        }
-        case IR::Opcode::A32ReadMemory16: {
-            if (!inst.AreAllArgsImmediates()) {
-                break;
-            }
-
-            const u32 vaddr = inst.GetArg(1).GetU32();
-            if (cb->IsReadOnlyMemory(vaddr)) {
-                const u16 value_from_memory = cb->MemoryRead16(vaddr);
-                inst.ReplaceUsesWith(IR::Value{value_from_memory});
-            }
-            break;
-        }
-        case IR::Opcode::A32ReadMemory32: {
-            if (!inst.AreAllArgsImmediates()) {
-                break;
-            }
-
-            const u32 vaddr = inst.GetArg(1).GetU32();
-            if (cb->IsReadOnlyMemory(vaddr)) {
-                const u32 value_from_memory = cb->MemoryRead32(vaddr);
-                inst.ReplaceUsesWith(IR::Value{value_from_memory});
-            }
-            break;
-        }
-        case IR::Opcode::A32ReadMemory64: {
-            if (!inst.AreAllArgsImmediates()) {
-                break;
-            }
-
-            const u32 vaddr = inst.GetArg(1).GetU32();
-            if (cb->IsReadOnlyMemory(vaddr)) {
-                const u64 value_from_memory = cb->MemoryRead64(vaddr);
-                inst.ReplaceUsesWith(IR::Value{value_from_memory});
-            }
-            break;
-        }
-        default:
-            break;
-        }
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/a32_get_set_elimination_pass.cpp

@@ -1,382 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <algorithm>
-#include <array>
-#include <functional>
-
-#include "dynarmic/common/assert.h"
-#include "dynarmic/common/common_types.h"
-
-#include "dynarmic/frontend/A32/a32_ir_emitter.h"
-#include "dynarmic/frontend/A32/a32_types.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-#include "dynarmic/ir/value.h"
-
-namespace Dynarmic::Optimization {
-
-namespace {
-
-void FlagsPass(IR::Block& block) {
-    using Iterator = std::reverse_iterator<IR::Block::iterator>;
-
-    struct FlagInfo {
-        bool set_not_required = false;
-        bool has_value_request = false;
-        Iterator value_request = {};
-    };
-    struct ValuelessFlagInfo {
-        bool set_not_required = false;
-    };
-    ValuelessFlagInfo nzcvq;
-    ValuelessFlagInfo nzcv;
-    ValuelessFlagInfo nz;
-    FlagInfo c_flag;
-    FlagInfo ge;
-
-    auto do_set = [&](FlagInfo& info, IR::Value value, Iterator inst) {
-        if (info.has_value_request) {
-            info.value_request->ReplaceUsesWith(value);
-        }
-        info.has_value_request = false;
-
-        if (info.set_not_required) {
-            inst->Invalidate();
-        }
-        info.set_not_required = true;
-    };
-
-    auto do_set_valueless = [&](ValuelessFlagInfo& info, Iterator inst) {
-        if (info.set_not_required) {
-            inst->Invalidate();
-        }
-        info.set_not_required = true;
-    };
-
-    auto do_get = [](FlagInfo& info, Iterator inst) {
-        if (info.has_value_request) {
-            info.value_request->ReplaceUsesWith(IR::Value{&*inst});
-        }
-        info.has_value_request = true;
-        info.value_request = inst;
-    };
-
-    A32::IREmitter ir{block, A32::LocationDescriptor{block.Location()}, {}};
-
-    for (auto inst = block.rbegin(); inst != block.rend(); ++inst) {
-        auto const opcode = inst->GetOpcode();
-        switch (opcode) {
-        case IR::Opcode::A32GetCFlag: {
-            do_get(c_flag, inst);
-            break;
-        }
-        case IR::Opcode::A32SetCpsrNZCV: {
-            if (c_flag.has_value_request) {
-                ir.SetInsertionPointBefore(inst.base());  // base is one ahead
-                IR::U1 c = ir.GetCFlagFromNZCV(IR::NZCV{inst->GetArg(0)});
-                c_flag.value_request->ReplaceUsesWith(c);
-                c_flag.has_value_request = false;
-                break;  // This case will be executed again because of the above
-            }
-
-            do_set_valueless(nzcv, inst);
-
-            nz = {.set_not_required = true};
-            c_flag = {.set_not_required = true};
-            break;
-        }
-        case IR::Opcode::A32SetCpsrNZCVRaw: {
-            if (c_flag.has_value_request) {
-                nzcv.set_not_required = false;
-            }
-
-            do_set_valueless(nzcv, inst);
-
-            nzcvq = {};
-            nz = {.set_not_required = true};
-            c_flag = {.set_not_required = true};
-            break;
-        }
-        case IR::Opcode::A32SetCpsrNZCVQ: {
-            if (c_flag.has_value_request) {
-                nzcvq.set_not_required = false;
-            }
-
-            do_set_valueless(nzcvq, inst);
-
-            nzcv = {.set_not_required = true};
-            nz = {.set_not_required = true};
-            c_flag = {.set_not_required = true};
-            break;
-        }
-        case IR::Opcode::A32SetCpsrNZ: {
-            do_set_valueless(nz, inst);
-
-            nzcvq = {};
-            nzcv = {};
-            break;
-        }
-        case IR::Opcode::A32SetCpsrNZC: {
-            if (c_flag.has_value_request) {
-                c_flag.value_request->ReplaceUsesWith(inst->GetArg(1));
-                c_flag.has_value_request = false;
-            }
-
-            if (!inst->GetArg(1).IsImmediate() && inst->GetArg(1).GetInstRecursive()->GetOpcode() == IR::Opcode::A32GetCFlag) {
-                const auto nz_value = inst->GetArg(0);
-
-                inst->Invalidate();
-
-                ir.SetInsertionPointBefore(inst.base());
-                ir.SetCpsrNZ(IR::NZCV{nz_value});
-
-                nzcvq = {};
-                nzcv = {};
-                nz = {.set_not_required = true};
-                break;
-            }
-
-            if (nz.set_not_required && c_flag.set_not_required) {
-                inst->Invalidate();
-            } else if (nz.set_not_required) {
-                inst->SetArg(0, IR::Value::EmptyNZCVImmediateMarker());
-            }
-            nz.set_not_required = true;
-            c_flag.set_not_required = true;
-
-            nzcv = {};
-            nzcvq = {};
-            break;
-        }
-        case IR::Opcode::A32SetGEFlags: {
-            do_set(ge, inst->GetArg(0), inst);
-            break;
-        }
-        case IR::Opcode::A32GetGEFlags: {
-            do_get(ge, inst);
-            break;
-        }
-        case IR::Opcode::A32SetGEFlagsCompressed: {
-            ge = {.set_not_required = true};
-            break;
-        }
-        case IR::Opcode::A32OrQFlag: {
-            break;
-        }
-        default: {
-            if (ReadsFromCPSR(opcode) || WritesToCPSR(opcode)) {
-                nzcvq = {};
-                nzcv = {};
-                nz = {};
-                c_flag = {};
-                ge = {};
-            }
-            break;
-        }
-        }
-    }
-}
-
-void RegisterPass(IR::Block& block) {
-    using Iterator = IR::Block::iterator;
-
-    struct RegInfo {
-        IR::Value register_value;
-        std::optional<Iterator> last_set_instruction;
-    };
-    std::array<RegInfo, 15> reg_info;
-
-    const auto do_get = [](RegInfo& info, Iterator get_inst) {
-        if (info.register_value.IsEmpty()) {
-            info.register_value = IR::Value(&*get_inst);
-            return;
-        }
-        get_inst->ReplaceUsesWith(info.register_value);
-    };
-
-    const auto do_set = [](RegInfo& info, IR::Value value, Iterator set_inst) {
-        if (info.last_set_instruction) {
-            (*info.last_set_instruction)->Invalidate();
-        }
-        info = {
-            .register_value = value,
-            .last_set_instruction = set_inst,
-        };
-    };
-
-    enum class ExtValueType {
-        Empty,
-        Single,
-        Double,
-        VectorDouble,
-        VectorQuad,
-    };
-    struct ExtRegInfo {
-        ExtValueType value_type = {};
-        IR::Value register_value;
-        std::optional<Iterator> last_set_instruction;
-    };
-    std::array<ExtRegInfo, 64> ext_reg_info;
-
-    const auto do_ext_get = [](ExtValueType type, std::initializer_list<std::reference_wrapper<ExtRegInfo>> infos, Iterator get_inst) {
-        if (!std::all_of(infos.begin(), infos.end(), [type](const auto& info) { return info.get().value_type == type; })) {
-            for (auto& info : infos) {
-                info.get() = {
-                    .value_type = type,
-                    .register_value = IR::Value(&*get_inst),
-                    .last_set_instruction = std::nullopt,
-                };
-            }
-            return;
-        }
-        get_inst->ReplaceUsesWith(std::data(infos)[0].get().register_value);
-    };
-
-    const auto do_ext_set = [](ExtValueType type, std::initializer_list<std::reference_wrapper<ExtRegInfo>> infos, IR::Value value, Iterator set_inst) {
-        if (std::all_of(infos.begin(), infos.end(), [type](const auto& info) { return info.get().value_type == type; })) {
-            if (std::data(infos)[0].get().last_set_instruction) {
-                (*std::data(infos)[0].get().last_set_instruction)->Invalidate();
-            }
-        }
-        for (auto& info : infos) {
-            info.get() = {
-                .value_type = type,
-                .register_value = value,
-                .last_set_instruction = set_inst,
-            };
-        }
-    };
-
-    // Location and version don't matter here.
-    A32::IREmitter ir{block, A32::LocationDescriptor{block.Location()}, {}};
-
-    for (auto inst = block.begin(); inst != block.end(); ++inst) {
-        auto const opcode = inst->GetOpcode();
-        switch (opcode) {
-        case IR::Opcode::A32GetRegister: {
-            const A32::Reg reg = inst->GetArg(0).GetA32RegRef();
-            ASSERT(reg != A32::Reg::PC);
-            const size_t reg_index = static_cast<size_t>(reg);
-            do_get(reg_info[reg_index], inst);
-            break;
-        }
-        case IR::Opcode::A32SetRegister: {
-            const A32::Reg reg = inst->GetArg(0).GetA32RegRef();
-            if (reg == A32::Reg::PC) {
-                break;
-            }
-            const auto reg_index = static_cast<size_t>(reg);
-            do_set(reg_info[reg_index], inst->GetArg(1), inst);
-            break;
-        }
-        case IR::Opcode::A32GetExtendedRegister32: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            do_ext_get(ExtValueType::Single, {ext_reg_info[reg_index]}, inst);
-            break;
-        }
-        case IR::Opcode::A32SetExtendedRegister32: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            do_ext_set(ExtValueType::Single, {ext_reg_info[reg_index]}, inst->GetArg(1), inst);
-            break;
-        }
-        case IR::Opcode::A32GetExtendedRegister64: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            do_ext_get(ExtValueType::Double,
-                       {
-                           ext_reg_info[reg_index * 2 + 0],
-                           ext_reg_info[reg_index * 2 + 1],
-                       },
-                       inst);
-            break;
-        }
-        case IR::Opcode::A32SetExtendedRegister64: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            do_ext_set(ExtValueType::Double,
-                       {
-                           ext_reg_info[reg_index * 2 + 0],
-                           ext_reg_info[reg_index * 2 + 1],
-                       },
-                       inst->GetArg(1),
-                       inst);
-            break;
-        }
-        case IR::Opcode::A32GetVector: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            if (A32::IsDoubleExtReg(reg)) {
-                do_ext_get(ExtValueType::VectorDouble,
-                           {
-                               ext_reg_info[reg_index * 2 + 0],
-                               ext_reg_info[reg_index * 2 + 1],
-                           },
-                           inst);
-            } else {
-                DEBUG_ASSERT(A32::IsQuadExtReg(reg));
-                do_ext_get(ExtValueType::VectorQuad,
-                           {
-                               ext_reg_info[reg_index * 4 + 0],
-                               ext_reg_info[reg_index * 4 + 1],
-                               ext_reg_info[reg_index * 4 + 2],
-                               ext_reg_info[reg_index * 4 + 3],
-                           },
-                           inst);
-            }
-            break;
-        }
-        case IR::Opcode::A32SetVector: {
-            const A32::ExtReg reg = inst->GetArg(0).GetA32ExtRegRef();
-            const size_t reg_index = A32::RegNumber(reg);
-            if (A32::IsDoubleExtReg(reg)) {
-                ir.SetInsertionPointAfter(inst);
-                const IR::U128 stored_value = ir.VectorZeroUpper(IR::U128{inst->GetArg(1)});
-                do_ext_set(ExtValueType::VectorDouble,
-                           {
-                               ext_reg_info[reg_index * 2 + 0],
-                               ext_reg_info[reg_index * 2 + 1],
-                           },
-                           stored_value,
-                           inst);
-            } else {
-                DEBUG_ASSERT(A32::IsQuadExtReg(reg));
-                do_ext_set(ExtValueType::VectorQuad,
-                           {
-                               ext_reg_info[reg_index * 4 + 0],
-                               ext_reg_info[reg_index * 4 + 1],
-                               ext_reg_info[reg_index * 4 + 2],
-                               ext_reg_info[reg_index * 4 + 3],
-                           },
-                           inst->GetArg(1),
-                           inst);
-            }
-            break;
-        }
-        default: {
-            if (ReadsFromCoreRegister(opcode) || WritesToCoreRegister(opcode)) {
-                reg_info = {};
-                ext_reg_info = {};
-            }
-            break;
-        }
-        }
-    }
-}
-
-}  // namespace
-
-void A32GetSetElimination(IR::Block& block, A32GetSetEliminationOptions) {
-    FlagsPass(block);
-    RegisterPass(block);
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/a64_callback_config_pass.cpp

@@ -1,57 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2018 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/frontend/A64/a64_ir_emitter.h"
-#include "dynarmic/interface/A64/config.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/microinstruction.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-void A64CallbackConfigPass(IR::Block& block, const A64::UserConfig& conf) {
-    if (conf.hook_data_cache_operations) {
-        return;
-    }
-
-    for (auto& inst : block) {
-        if (inst.GetOpcode() != IR::Opcode::A64DataCacheOperationRaised) {
-            continue;
-        }
-
-        const auto op = static_cast<A64::DataCacheOperation>(inst.GetArg(1).GetU64());
-        if (op == A64::DataCacheOperation::ZeroByVA) {
-            A64::IREmitter ir{block};
-            ir.current_location = A64::LocationDescriptor{IR::LocationDescriptor{inst.GetArg(0).GetU64()}};
-            ir.SetInsertionPointBefore(&inst);
-
-            size_t bytes = 4 << static_cast<size_t>(conf.dczid_el0 & 0b1111);
-            IR::U64 addr{inst.GetArg(2)};
-
-            const IR::U128 zero_u128 = ir.ZeroExtendToQuad(ir.Imm64(0));
-            while (bytes >= 16) {
-                ir.WriteMemory128(addr, zero_u128, IR::AccType::DCZVA);
-                addr = ir.Add(addr, ir.Imm64(16));
-                bytes -= 16;
-            }
-
-            while (bytes >= 8) {
-                ir.WriteMemory64(addr, ir.Imm64(0), IR::AccType::DCZVA);
-                addr = ir.Add(addr, ir.Imm64(8));
-                bytes -= 8;
-            }
-
-            while (bytes >= 4) {
-                ir.WriteMemory32(addr, ir.Imm32(0), IR::AccType::DCZVA);
-                addr = ir.Add(addr, ir.Imm64(4));
-                bytes -= 4;
-            }
-        }
-        inst.Invalidate();
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/a64_get_set_elimination_pass.cpp

@@ -1,165 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <array>
-
-#include "dynarmic/common/common_types.h"
-
-#include "dynarmic/frontend/A64/a64_types.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-#include "dynarmic/ir/value.h"
-
-namespace Dynarmic::Optimization {
-
-void A64GetSetElimination(IR::Block& block) {
-    using Iterator = IR::Block::iterator;
-
-    enum class TrackingType {
-        W,
-        X,
-        S,
-        D,
-        Q,
-        SP,
-        NZCV,
-        NZCVRaw,
-    };
-    struct RegisterInfo {
-        IR::Value register_value;
-        TrackingType tracking_type;
-        bool set_instruction_present = false;
-        Iterator last_set_instruction;
-    };
-    std::array<RegisterInfo, 31> reg_info;
-    std::array<RegisterInfo, 32> vec_info;
-    RegisterInfo sp_info;
-    RegisterInfo nzcv_info;
-
-    const auto do_set = [&block](RegisterInfo& info, IR::Value value, Iterator set_inst, TrackingType tracking_type) {
-        if (info.set_instruction_present) {
-            info.last_set_instruction->Invalidate();
-            block.Instructions().erase(info.last_set_instruction);
-        }
-
-        info.register_value = value;
-        info.tracking_type = tracking_type;
-        info.set_instruction_present = true;
-        info.last_set_instruction = set_inst;
-    };
-
-    const auto do_get = [](RegisterInfo& info, Iterator get_inst, TrackingType tracking_type) {
-        const auto do_nothing = [&] {
-            info = {};
-            info.register_value = IR::Value(&*get_inst);
-            info.tracking_type = tracking_type;
-        };
-
-        if (info.register_value.IsEmpty()) {
-            do_nothing();
-            return;
-        }
-
-        if (info.tracking_type == tracking_type) {
-            get_inst->ReplaceUsesWith(info.register_value);
-            return;
-        }
-
-        do_nothing();
-    };
-
-    for (auto inst = block.begin(); inst != block.end(); ++inst) {
-        auto const opcode = inst->GetOpcode();
-        switch (opcode) {
-        case IR::Opcode::A64GetW: {
-            const size_t index = A64::RegNumber(inst->GetArg(0).GetA64RegRef());
-            do_get(reg_info.at(index), inst, TrackingType::W);
-            break;
-        }
-        case IR::Opcode::A64GetX: {
-            const size_t index = A64::RegNumber(inst->GetArg(0).GetA64RegRef());
-            do_get(reg_info.at(index), inst, TrackingType::X);
-            break;
-        }
-        case IR::Opcode::A64GetS: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_get(vec_info.at(index), inst, TrackingType::S);
-            break;
-        }
-        case IR::Opcode::A64GetD: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_get(vec_info.at(index), inst, TrackingType::D);
-            break;
-        }
-        case IR::Opcode::A64GetQ: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_get(vec_info.at(index), inst, TrackingType::Q);
-            break;
-        }
-        case IR::Opcode::A64GetSP: {
-            do_get(sp_info, inst, TrackingType::SP);
-            break;
-        }
-        case IR::Opcode::A64GetNZCVRaw: {
-            do_get(nzcv_info, inst, TrackingType::NZCVRaw);
-            break;
-        }
-        case IR::Opcode::A64SetW: {
-            const size_t index = A64::RegNumber(inst->GetArg(0).GetA64RegRef());
-            do_set(reg_info.at(index), inst->GetArg(1), inst, TrackingType::W);
-            break;
-        }
-        case IR::Opcode::A64SetX: {
-            const size_t index = A64::RegNumber(inst->GetArg(0).GetA64RegRef());
-            do_set(reg_info.at(index), inst->GetArg(1), inst, TrackingType::X);
-            break;
-        }
-        case IR::Opcode::A64SetS: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_set(vec_info.at(index), inst->GetArg(1), inst, TrackingType::S);
-            break;
-        }
-        case IR::Opcode::A64SetD: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_set(vec_info.at(index), inst->GetArg(1), inst, TrackingType::D);
-            break;
-        }
-        case IR::Opcode::A64SetQ: {
-            const size_t index = A64::VecNumber(inst->GetArg(0).GetA64VecRef());
-            do_set(vec_info.at(index), inst->GetArg(1), inst, TrackingType::Q);
-            break;
-        }
-        case IR::Opcode::A64SetSP: {
-            do_set(sp_info, inst->GetArg(0), inst, TrackingType::SP);
-            break;
-        }
-        case IR::Opcode::A64SetNZCV: {
-            do_set(nzcv_info, inst->GetArg(0), inst, TrackingType::NZCV);
-            break;
-        }
-        case IR::Opcode::A64SetNZCVRaw: {
-            do_set(nzcv_info, inst->GetArg(0), inst, TrackingType::NZCVRaw);
-            break;
-        }
-        default: {
-            if (ReadsFromCPSR(opcode) || WritesToCPSR(opcode)) {
-                nzcv_info = {};
-            }
-            if (ReadsFromCoreRegister(opcode) || WritesToCoreRegister(opcode)) {
-                reg_info = {};
-                vec_info = {};
-                sp_info = {};
-            }
-            break;
-        }
-        }
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/a64_merge_interpret_blocks.cpp

@@ -1,57 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2018 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <boost/variant/get.hpp>
-#include "dynarmic/common/common_types.h"
-
-#include "dynarmic/frontend/A64/a64_location_descriptor.h"
-#include "dynarmic/frontend/A64/translate/a64_translate.h"
-#include "dynarmic/interface/A64/config.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-void A64MergeInterpretBlocksPass(IR::Block& block, A64::UserCallbacks* cb) {
-    const auto is_interpret_instruction = [cb](A64::LocationDescriptor location) {
-        const auto instruction = cb->MemoryReadCode(location.PC());
-        if (!instruction)
-            return false;
-
-        IR::Block new_block{location};
-        A64::TranslateSingleInstruction(new_block, location, *instruction);
-
-        if (!new_block.Instructions().empty())
-            return false;
-
-        const IR::Terminal terminal = new_block.GetTerminal();
-        if (auto term = boost::get<IR::Term::Interpret>(&terminal)) {
-            return term->next == location;
-        }
-
-        return false;
-    };
-
-    IR::Terminal terminal = block.GetTerminal();
-    auto term = boost::get<IR::Term::Interpret>(&terminal);
-    if (!term)
-        return;
-
-    A64::LocationDescriptor location{term->next};
-    size_t num_instructions = 1;
-
-    while (is_interpret_instruction(location.AdvancePC(static_cast<int>(num_instructions * 4)))) {
-        num_instructions++;
-    }
-
-    term->num_instructions = num_instructions;
-    block.ReplaceTerminal(terminal);
-    block.CycleCount() += num_instructions - 1;
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/constant_propagation_pass.cpp

@@ -1,559 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <optional>
-
-#include "dynarmic/common/assert.h"
-#include <mcl/bit/rotate.hpp>
-#include <mcl/bit/swap.hpp>
-#include "dynarmic/common/common_types.h"
-
-#include "dynarmic/common/safe_ops.h"
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/ir_emitter.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-using Op = Dynarmic::IR::Opcode;
-
-namespace {
-
-// Tiny helper to avoid the need to store based off the opcode
-// bit size all over the place within folding functions.
-void ReplaceUsesWith(IR::Inst& inst, bool is_32_bit, u64 value) {
-    if (is_32_bit) {
-        inst.ReplaceUsesWith(IR::Value{static_cast<u32>(value)});
-    } else {
-        inst.ReplaceUsesWith(IR::Value{value});
-    }
-}
-
-IR::Value Value(bool is_32_bit, u64 value) {
-    return is_32_bit ? IR::Value{static_cast<u32>(value)} : IR::Value{value};
-}
-
-template<typename ImmFn>
-bool FoldCommutative(IR::Inst& inst, bool is_32_bit, ImmFn imm_fn) {
-    const auto lhs = inst.GetArg(0);
-    const auto rhs = inst.GetArg(1);
-
-    const bool is_lhs_immediate = lhs.IsImmediate();
-    const bool is_rhs_immediate = rhs.IsImmediate();
-
-    if (is_lhs_immediate && is_rhs_immediate) {
-        const u64 result = imm_fn(lhs.GetImmediateAsU64(), rhs.GetImmediateAsU64());
-        ReplaceUsesWith(inst, is_32_bit, result);
-        return false;
-    }
-
-    if (is_lhs_immediate && !is_rhs_immediate) {
-        const IR::Inst* rhs_inst = rhs.GetInstRecursive();
-        if (rhs_inst->GetOpcode() == inst.GetOpcode() && rhs_inst->GetArg(1).IsImmediate()) {
-            const u64 combined = imm_fn(lhs.GetImmediateAsU64(), rhs_inst->GetArg(1).GetImmediateAsU64());
-            inst.SetArg(0, rhs_inst->GetArg(0));
-            inst.SetArg(1, Value(is_32_bit, combined));
-        } else {
-            // Normalize
-            inst.SetArg(0, rhs);
-            inst.SetArg(1, lhs);
-        }
-    }
-
-    if (!is_lhs_immediate && is_rhs_immediate) {
-        const IR::Inst* lhs_inst = lhs.GetInstRecursive();
-        if (lhs_inst->GetOpcode() == inst.GetOpcode() && lhs_inst->GetArg(1).IsImmediate()) {
-            const u64 combined = imm_fn(rhs.GetImmediateAsU64(), lhs_inst->GetArg(1).GetImmediateAsU64());
-            inst.SetArg(0, lhs_inst->GetArg(0));
-            inst.SetArg(1, Value(is_32_bit, combined));
-        }
-    }
-
-    return true;
-}
-
-void FoldAdd(IR::Inst& inst, bool is_32_bit) {
-    const auto lhs = inst.GetArg(0);
-    const auto rhs = inst.GetArg(1);
-    const auto carry = inst.GetArg(2);
-
-    if (lhs.IsImmediate() && !rhs.IsImmediate()) {
-        // Normalize
-        inst.SetArg(0, rhs);
-        inst.SetArg(1, lhs);
-        FoldAdd(inst, is_32_bit);
-        return;
-    }
-
-    if (inst.HasAssociatedPseudoOperation()) {
-        return;
-    }
-
-    if (!lhs.IsImmediate() && rhs.IsImmediate()) {
-        const IR::Inst* lhs_inst = lhs.GetInstRecursive();
-        if (lhs_inst->GetOpcode() == inst.GetOpcode() && lhs_inst->GetArg(1).IsImmediate() && lhs_inst->GetArg(2).IsImmediate()) {
-            const u64 combined = rhs.GetImmediateAsU64() + lhs_inst->GetArg(1).GetImmediateAsU64() + lhs_inst->GetArg(2).GetU1();
-            if (combined == 0) {
-                inst.ReplaceUsesWith(lhs_inst->GetArg(0));
-                return;
-            }
-            inst.SetArg(0, lhs_inst->GetArg(0));
-            inst.SetArg(1, Value(is_32_bit, combined));
-            return;
-        }
-        if (rhs.IsZero() && carry.IsZero()) {
-            inst.ReplaceUsesWith(lhs);
-            return;
-        }
-    }
-
-    if (inst.AreAllArgsImmediates()) {
-        const u64 result = lhs.GetImmediateAsU64() + rhs.GetImmediateAsU64() + carry.GetU1();
-        ReplaceUsesWith(inst, is_32_bit, result);
-        return;
-    }
-}
-
-/// Folds AND operations based on the following:
-///
-/// 1. imm_x & imm_y -> result
-/// 2. x & 0 -> 0
-/// 3. 0 & y -> 0
-/// 4. x & y -> y (where x has all bits set to 1)
-/// 5. x & y -> x (where y has all bits set to 1)
-///
-void FoldAND(IR::Inst& inst, bool is_32_bit) {
-    if (FoldCommutative(inst, is_32_bit, [](u64 a, u64 b) { return a & b; })) {
-        const auto rhs = inst.GetArg(1);
-        if (rhs.IsZero()) {
-            ReplaceUsesWith(inst, is_32_bit, 0);
-        } else if (rhs.HasAllBitsSet()) {
-            inst.ReplaceUsesWith(inst.GetArg(0));
-        }
-    }
-}
-
-/// Folds byte reversal opcodes based on the following:
-///
-/// 1. imm -> swap(imm)
-///
-void FoldByteReverse(IR::Inst& inst, Op op) {
-    const auto operand = inst.GetArg(0);
-
-    if (!operand.IsImmediate()) {
-        return;
-    }
-
-    if (op == Op::ByteReverseWord) {
-        const u32 result = mcl::bit::swap_bytes_32(static_cast<u32>(operand.GetImmediateAsU64()));
-        inst.ReplaceUsesWith(IR::Value{result});
-    } else if (op == Op::ByteReverseHalf) {
-        const u16 result = mcl::bit::swap_bytes_16(static_cast<u16>(operand.GetImmediateAsU64()));
-        inst.ReplaceUsesWith(IR::Value{result});
-    } else {
-        const u64 result = mcl::bit::swap_bytes_64(operand.GetImmediateAsU64());
-        inst.ReplaceUsesWith(IR::Value{result});
-    }
-}
-
-/// Folds division operations based on the following:
-///
-/// 1. x / 0 -> 0 (NOTE: This is an ARM-specific behavior defined in the architecture reference manual)
-/// 2. imm_x / imm_y -> result
-/// 3. x / 1 -> x
-///
-void FoldDivide(IR::Inst& inst, bool is_32_bit, bool is_signed) {
-    const auto rhs = inst.GetArg(1);
-
-    if (rhs.IsZero()) {
-        ReplaceUsesWith(inst, is_32_bit, 0);
-        return;
-    }
-
-    const auto lhs = inst.GetArg(0);
-    if (lhs.IsImmediate() && rhs.IsImmediate()) {
-        if (is_signed) {
-            const s64 result = lhs.GetImmediateAsS64() / rhs.GetImmediateAsS64();
-            ReplaceUsesWith(inst, is_32_bit, static_cast<u64>(result));
-        } else {
-            const u64 result = lhs.GetImmediateAsU64() / rhs.GetImmediateAsU64();
-            ReplaceUsesWith(inst, is_32_bit, result);
-        }
-    } else if (rhs.IsUnsignedImmediate(1)) {
-        inst.ReplaceUsesWith(IR::Value{lhs});
-    }
-}
-
-// Folds EOR operations based on the following:
-//
-// 1. imm_x ^ imm_y -> result
-// 2. x ^ 0 -> x
-// 3. 0 ^ y -> y
-//
-void FoldEOR(IR::Inst& inst, bool is_32_bit) {
-    if (FoldCommutative(inst, is_32_bit, [](u64 a, u64 b) { return a ^ b; })) {
-        const auto rhs = inst.GetArg(1);
-        if (rhs.IsZero()) {
-            inst.ReplaceUsesWith(inst.GetArg(0));
-        }
-    }
-}
-
-void FoldLeastSignificantByte(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const auto operand = inst.GetArg(0);
-    inst.ReplaceUsesWith(IR::Value{static_cast<u8>(operand.GetImmediateAsU64())});
-}
-
-void FoldLeastSignificantHalf(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const auto operand = inst.GetArg(0);
-    inst.ReplaceUsesWith(IR::Value{static_cast<u16>(operand.GetImmediateAsU64())});
-}
-
-void FoldLeastSignificantWord(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const auto operand = inst.GetArg(0);
-    inst.ReplaceUsesWith(IR::Value{static_cast<u32>(operand.GetImmediateAsU64())});
-}
-
-void FoldMostSignificantBit(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const auto operand = inst.GetArg(0);
-    inst.ReplaceUsesWith(IR::Value{(operand.GetImmediateAsU64() >> 31) != 0});
-}
-
-void FoldMostSignificantWord(IR::Inst& inst) {
-    IR::Inst* carry_inst = inst.GetAssociatedPseudoOperation(Op::GetCarryFromOp);
-
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const auto operand = inst.GetArg(0);
-    if (carry_inst) {
-        carry_inst->ReplaceUsesWith(IR::Value{mcl::bit::get_bit<31>(operand.GetImmediateAsU64())});
-    }
-    inst.ReplaceUsesWith(IR::Value{static_cast<u32>(operand.GetImmediateAsU64() >> 32)});
-}
-
-// Folds multiplication operations based on the following:
-//
-// 1. imm_x * imm_y -> result
-// 2. x * 0 -> 0
-// 3. 0 * y -> 0
-// 4. x * 1 -> x
-// 5. 1 * y -> y
-//
-void FoldMultiply(IR::Inst& inst, bool is_32_bit) {
-    if (FoldCommutative(inst, is_32_bit, [](u64 a, u64 b) { return a * b; })) {
-        const auto rhs = inst.GetArg(1);
-        if (rhs.IsZero()) {
-            ReplaceUsesWith(inst, is_32_bit, 0);
-        } else if (rhs.IsUnsignedImmediate(1)) {
-            inst.ReplaceUsesWith(inst.GetArg(0));
-        }
-    }
-}
-
-// Folds NOT operations if the contained value is an immediate.
-void FoldNOT(IR::Inst& inst, bool is_32_bit) {
-    const auto operand = inst.GetArg(0);
-
-    if (!operand.IsImmediate()) {
-        return;
-    }
-
-    const u64 result = ~operand.GetImmediateAsU64();
-    ReplaceUsesWith(inst, is_32_bit, result);
-}
-
-// Folds OR operations based on the following:
-//
-// 1. imm_x | imm_y -> result
-// 2. x | 0 -> x
-// 3. 0 | y -> y
-//
-void FoldOR(IR::Inst& inst, bool is_32_bit) {
-    if (FoldCommutative(inst, is_32_bit, [](u64 a, u64 b) { return a | b; })) {
-        const auto rhs = inst.GetArg(1);
-        if (rhs.IsZero()) {
-            inst.ReplaceUsesWith(inst.GetArg(0));
-        }
-    }
-}
-
-bool FoldShifts(IR::Inst& inst) {
-    IR::Inst* carry_inst = inst.GetAssociatedPseudoOperation(Op::GetCarryFromOp);
-
-    // The 32-bit variants can contain 3 arguments, while the
-    // 64-bit variants only contain 2.
-    if (inst.NumArgs() == 3 && !carry_inst) {
-        inst.SetArg(2, IR::Value(false));
-    }
-
-    const auto shift_amount = inst.GetArg(1);
-
-    if (shift_amount.IsZero()) {
-        if (carry_inst) {
-            carry_inst->ReplaceUsesWith(inst.GetArg(2));
-        }
-        inst.ReplaceUsesWith(inst.GetArg(0));
-        return false;
-    }
-
-    if (inst.NumArgs() == 3 && shift_amount.IsImmediate() && !shift_amount.IsZero()) {
-        inst.SetArg(2, IR::Value(false));
-    }
-
-    if (!inst.AreAllArgsImmediates() || carry_inst) {
-        return false;
-    }
-
-    return true;
-}
-
-void FoldSignExtendXToWord(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const s64 value = inst.GetArg(0).GetImmediateAsS64();
-    inst.ReplaceUsesWith(IR::Value{static_cast<u32>(value)});
-}
-
-void FoldSignExtendXToLong(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const s64 value = inst.GetArg(0).GetImmediateAsS64();
-    inst.ReplaceUsesWith(IR::Value{static_cast<u64>(value)});
-}
-
-void FoldSub(IR::Inst& inst, bool is_32_bit) {
-    if (!inst.AreAllArgsImmediates() || inst.HasAssociatedPseudoOperation()) {
-        return;
-    }
-
-    const auto lhs = inst.GetArg(0);
-    const auto rhs = inst.GetArg(1);
-    const auto carry = inst.GetArg(2);
-
-    const u64 result = lhs.GetImmediateAsU64() + (~rhs.GetImmediateAsU64()) + carry.GetU1();
-    ReplaceUsesWith(inst, is_32_bit, result);
-}
-
-void FoldZeroExtendXToWord(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const u64 value = inst.GetArg(0).GetImmediateAsU64();
-    inst.ReplaceUsesWith(IR::Value{static_cast<u32>(value)});
-}
-
-void FoldZeroExtendXToLong(IR::Inst& inst) {
-    if (!inst.AreAllArgsImmediates()) {
-        return;
-    }
-
-    const u64 value = inst.GetArg(0).GetImmediateAsU64();
-    inst.ReplaceUsesWith(IR::Value{value});
-}
-}  // Anonymous namespace
-
-void ConstantPropagation(IR::Block& block) {
-    for (auto& inst : block) {
-        const auto opcode = inst.GetOpcode();
-
-        switch (opcode) {
-        case Op::LeastSignificantWord:
-            FoldLeastSignificantWord(inst);
-            break;
-        case Op::MostSignificantWord:
-            FoldMostSignificantWord(inst);
-            break;
-        case Op::LeastSignificantHalf:
-            FoldLeastSignificantHalf(inst);
-            break;
-        case Op::LeastSignificantByte:
-            FoldLeastSignificantByte(inst);
-            break;
-        case Op::MostSignificantBit:
-            FoldMostSignificantBit(inst);
-            break;
-        case Op::IsZero32:
-            if (inst.AreAllArgsImmediates()) {
-                inst.ReplaceUsesWith(IR::Value{inst.GetArg(0).GetU32() == 0});
-            }
-            break;
-        case Op::IsZero64:
-            if (inst.AreAllArgsImmediates()) {
-                inst.ReplaceUsesWith(IR::Value{inst.GetArg(0).GetU64() == 0});
-            }
-            break;
-        case Op::LogicalShiftLeft32:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, true, Safe::LogicalShiftLeft<u32>(inst.GetArg(0).GetU32(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::LogicalShiftLeft64:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, false, Safe::LogicalShiftLeft<u64>(inst.GetArg(0).GetU64(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::LogicalShiftRight32:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, true, Safe::LogicalShiftRight<u32>(inst.GetArg(0).GetU32(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::LogicalShiftRight64:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, false, Safe::LogicalShiftRight<u64>(inst.GetArg(0).GetU64(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::ArithmeticShiftRight32:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, true, Safe::ArithmeticShiftRight<u32>(inst.GetArg(0).GetU32(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::ArithmeticShiftRight64:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, false, Safe::ArithmeticShiftRight<u64>(inst.GetArg(0).GetU64(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::RotateRight32:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, true, mcl::bit::rotate_right<u32>(inst.GetArg(0).GetU32(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::RotateRight64:
-            if (FoldShifts(inst)) {
-                ReplaceUsesWith(inst, false, mcl::bit::rotate_right<u64>(inst.GetArg(0).GetU64(), inst.GetArg(1).GetU8()));
-            }
-            break;
-        case Op::LogicalShiftLeftMasked32:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, true, inst.GetArg(0).GetU32() << (inst.GetArg(1).GetU32() & 0x1f));
-            }
-            break;
-        case Op::LogicalShiftLeftMasked64:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, false, inst.GetArg(0).GetU64() << (inst.GetArg(1).GetU64() & 0x3f));
-            }
-            break;
-        case Op::LogicalShiftRightMasked32:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, true, inst.GetArg(0).GetU32() >> (inst.GetArg(1).GetU32() & 0x1f));
-            }
-            break;
-        case Op::LogicalShiftRightMasked64:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, false, inst.GetArg(0).GetU64() >> (inst.GetArg(1).GetU64() & 0x3f));
-            }
-            break;
-        case Op::ArithmeticShiftRightMasked32:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, true, static_cast<s32>(inst.GetArg(0).GetU32()) >> (inst.GetArg(1).GetU32() & 0x1f));
-            }
-            break;
-        case Op::ArithmeticShiftRightMasked64:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, false, static_cast<s64>(inst.GetArg(0).GetU64()) >> (inst.GetArg(1).GetU64() & 0x3f));
-            }
-            break;
-        case Op::RotateRightMasked32:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, true, mcl::bit::rotate_right<u32>(inst.GetArg(0).GetU32(), inst.GetArg(1).GetU32()));
-            }
-            break;
-        case Op::RotateRightMasked64:
-            if (inst.AreAllArgsImmediates()) {
-                ReplaceUsesWith(inst, false, mcl::bit::rotate_right<u64>(inst.GetArg(0).GetU64(), inst.GetArg(1).GetU64()));
-            }
-            break;
-        case Op::Add32:
-        case Op::Add64:
-            FoldAdd(inst, opcode == Op::Add32);
-            break;
-        case Op::Sub32:
-        case Op::Sub64:
-            FoldSub(inst, opcode == Op::Sub32);
-            break;
-        case Op::Mul32:
-        case Op::Mul64:
-            FoldMultiply(inst, opcode == Op::Mul32);
-            break;
-        case Op::SignedDiv32:
-        case Op::SignedDiv64:
-            FoldDivide(inst, opcode == Op::SignedDiv32, true);
-            break;
-        case Op::UnsignedDiv32:
-        case Op::UnsignedDiv64:
-            FoldDivide(inst, opcode == Op::UnsignedDiv32, false);
-            break;
-        case Op::And32:
-        case Op::And64:
-            FoldAND(inst, opcode == Op::And32);
-            break;
-        case Op::Eor32:
-        case Op::Eor64:
-            FoldEOR(inst, opcode == Op::Eor32);
-            break;
-        case Op::Or32:
-        case Op::Or64:
-            FoldOR(inst, opcode == Op::Or32);
-            break;
-        case Op::Not32:
-        case Op::Not64:
-            FoldNOT(inst, opcode == Op::Not32);
-            break;
-        case Op::SignExtendByteToWord:
-        case Op::SignExtendHalfToWord:
-            FoldSignExtendXToWord(inst);
-            break;
-        case Op::SignExtendByteToLong:
-        case Op::SignExtendHalfToLong:
-        case Op::SignExtendWordToLong:
-            FoldSignExtendXToLong(inst);
-            break;
-        case Op::ZeroExtendByteToWord:
-        case Op::ZeroExtendHalfToWord:
-            FoldZeroExtendXToWord(inst);
-            break;
-        case Op::ZeroExtendByteToLong:
-        case Op::ZeroExtendHalfToLong:
-        case Op::ZeroExtendWordToLong:
-            FoldZeroExtendXToLong(inst);
-            break;
-        case Op::ByteReverseWord:
-        case Op::ByteReverseHalf:
-        case Op::ByteReverseDual:
-            FoldByteReverse(inst, opcode);
-            break;
-        default:
-            break;
-        }
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/dead_code_elimination_pass.cpp

@@ -1,23 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <mcl/iterator/reverse.hpp>
-
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-void DeadCodeElimination(IR::Block& block) {
-    // We iterate over the instructions in reverse order.
-    // This is because removing an instruction reduces the number of uses for earlier instructions.
-    for (auto& inst : mcl::iterator::reverse(block)) {
-        if (!inst.HasUses() && !MayHaveSideEffects(inst.GetOpcode())) {
-            inst.Invalidate();
-        }
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/identity_removal_pass.cpp

@@ -1,44 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2020 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <vector>
-
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-void IdentityRemovalPass(IR::Block& block) {
-    std::vector<IR::Inst*> to_invalidate;
-
-    auto iter = block.begin();
-    while (iter != block.end()) {
-        IR::Inst& inst = *iter;
-
-        const size_t num_args = inst.NumArgs();
-        for (size_t i = 0; i < num_args; i++) {
-            while (true) {
-                IR::Value arg = inst.GetArg(i);
-                if (!arg.IsIdentity())
-                    break;
-                inst.SetArg(i, arg.GetInst()->GetArg(0));
-            }
-        }
-
-        if (inst.GetOpcode() == IR::Opcode::Identity || inst.GetOpcode() == IR::Opcode::Void) {
-            iter = block.Instructions().erase(inst);
-            to_invalidate.push_back(&inst);
-        } else {
-            ++iter;
-        }
-    }
-
-    for (IR::Inst* inst : to_invalidate) {
-        inst->Invalidate();
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/ir_matcher.h

@@ -1,127 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2020 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#pragma once
-
-#include <optional>
-#include <tuple>
-
-#include <mp/metafunction/apply.h>
-#include <mp/typelist/concat.h>
-#include <mp/typelist/drop.h>
-#include <mp/typelist/get.h>
-#include <mp/typelist/head.h>
-#include <mp/typelist/list.h>
-#include <mp/typelist/prepend.h>
-
-#include "dynarmic/ir/microinstruction.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/value.h"
-
-namespace Dynarmic::Optimization::IRMatcher {
-
-struct CaptureValue {
-    using ReturnType = std::tuple<IR::Value>;
-
-    static std::optional<ReturnType> Match(IR::Value value) {
-        return std::tuple(value);
-    }
-};
-
-struct CaptureInst {
-    using ReturnType = std::tuple<IR::Inst*>;
-
-    static std::optional<ReturnType> Match(IR::Value value) {
-        if (value.IsImmediate())
-            return std::nullopt;
-        return std::tuple(value.GetInstRecursive());
-    }
-};
-
-struct CaptureUImm {
-    using ReturnType = std::tuple<u64>;
-
-    static std::optional<ReturnType> Match(IR::Value value) {
-        return std::tuple(value.GetImmediateAsU64());
-    }
-};
-
-struct CaptureSImm {
-    using ReturnType = std::tuple<s64>;
-
-    static std::optional<ReturnType> Match(IR::Value value) {
-        return std::tuple(value.GetImmediateAsS64());
-    }
-};
-
-template<u64 Value>
-struct UImm {
-    using ReturnType = std::tuple<>;
-
-    static std::optional<std::tuple<>> Match(IR::Value value) {
-        if (value.GetImmediateAsU64() == Value)
-            return std::tuple();
-        return std::nullopt;
-    }
-};
-
-template<s64 Value>
-struct SImm {
-    using ReturnType = std::tuple<>;
-
-    static std::optional<std::tuple<>> Match(IR::Value value) {
-        if (value.GetImmediateAsS64() == Value)
-            return std::tuple();
-        return std::nullopt;
-    }
-};
-
-template<IR::Opcode Opcode, typename... Args>
-struct Inst {
-public:
-    using ReturnType = mp::concat<std::tuple<>, typename Args::ReturnType...>;
-
-    static std::optional<ReturnType> Match(const IR::Inst& inst) {
-        if (inst.GetOpcode() != Opcode)
-            return std::nullopt;
-        if (inst.HasAssociatedPseudoOperation())
-            return std::nullopt;
-        return MatchArgs<0>(inst);
-    }
-
-    static std::optional<ReturnType> Match(IR::Value value) {
-        if (value.IsImmediate())
-            return std::nullopt;
-        return Match(*value.GetInstRecursive());
-    }
-
-private:
-    template<size_t I>
-    static auto MatchArgs(const IR::Inst& inst) -> std::optional<mp::apply<mp::concat, mp::prepend<mp::drop<I, mp::list<typename Args::ReturnType...>>, std::tuple<>>>> {
-        if constexpr (I >= sizeof...(Args)) {
-            return std::tuple();
-        } else {
-            using Arg = mp::get<I, mp::list<Args...>>;
-
-            if (const auto arg = Arg::Match(inst.GetArg(I))) {
-                if (const auto rest = MatchArgs<I + 1>(inst)) {
-                    return std::tuple_cat(*arg, *rest);
-                }
-            }
-
-            return std::nullopt;
-        }
-    }
-};
-
-inline bool IsSameInst(std::tuple<IR::Inst*, IR::Inst*> t) {
-    return std::get<0>(t) == std::get<1>(t);
-}
-
-inline bool IsSameInst(std::tuple<IR::Inst*, IR::Inst*, IR::Inst*> t) {
-    return std::get<0>(t) == std::get<1>(t) && std::get<0>(t) == std::get<2>(t);
-}
-
-}  // namespace Dynarmic::Optimization::IRMatcher

src/dynarmic/src/dynarmic/ir/opt/naming_pass.cpp

@@ -1,18 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2023 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/microinstruction.h"
-
-namespace Dynarmic::Optimization {
-
-void NamingPass(IR::Block& block) {
-    unsigned name = 1;
-    for (auto& inst : block) {
-        inst.SetName(name++);
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/passes.h

@@ -1,47 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#pragma once
-
-namespace Dynarmic::A32 {
-struct UserCallbacks;
-}
-
-namespace Dynarmic::A64 {
-struct UserCallbacks;
-struct UserConfig;
-}  // namespace Dynarmic::A64
-
-namespace Dynarmic::IR {
-class Block;
-}
-
-namespace Dynarmic::Optimization {
-
-struct PolyfillOptions {
-    bool sha256 = false;
-    bool vector_multiply_widen = false;
-
-    bool operator==(const PolyfillOptions&) const = default;
-};
-
-struct A32GetSetEliminationOptions {
-    bool convert_nzc_to_nz = false;
-    bool convert_nz_to_nzc = false;
-};
-
-void PolyfillPass(IR::Block& block, const PolyfillOptions& opt);
-void A32ConstantMemoryReads(IR::Block& block, A32::UserCallbacks* cb);
-void A32GetSetElimination(IR::Block& block, A32GetSetEliminationOptions opt);
-void A64CallbackConfigPass(IR::Block& block, const A64::UserConfig& conf);
-void A64GetSetElimination(IR::Block& block);
-void A64MergeInterpretBlocksPass(IR::Block& block, A64::UserCallbacks* cb);
-void ConstantPropagation(IR::Block& block);
-void DeadCodeElimination(IR::Block& block);
-void IdentityRemovalPass(IR::Block& block);
-void VerificationPass(const IR::Block& block);
-void NamingPass(IR::Block& block);
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/polyfill_pass.cpp

@@ -1,218 +0,0 @@
-/* This file is part of the dynarmic project.
- * Copyright (c) 2022 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/ir_emitter.h"
-#include "dynarmic/ir/microinstruction.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-
-namespace Dynarmic::Optimization {
-
-namespace {
-
-void PolyfillSHA256MessageSchedule0(IR::IREmitter& ir, IR::Inst& inst) {
-    const IR::U128 x = (IR::U128)inst.GetArg(0);
-    const IR::U128 y = (IR::U128)inst.GetArg(1);
-
-    const IR::U128 t = ir.VectorExtract(x, y, 32);
-
-    IR::U128 result = ir.ZeroVector();
-    for (size_t i = 0; i < 4; i++) {
-        const IR::U32 modified_element = [&] {
-            const IR::U32 element = ir.VectorGetElement(32, t, i);
-            const IR::U32 tmp1 = ir.RotateRight(element, ir.Imm8(7));
-            const IR::U32 tmp2 = ir.RotateRight(element, ir.Imm8(18));
-            const IR::U32 tmp3 = ir.LogicalShiftRight(element, ir.Imm8(3));
-
-            return ir.Eor(tmp1, ir.Eor(tmp2, tmp3));
-        }();
-
-        result = ir.VectorSetElement(32, result, i, modified_element);
-    }
-    result = ir.VectorAdd(32, result, x);
-
-    inst.ReplaceUsesWith(result);
-}
-
-void PolyfillSHA256MessageSchedule1(IR::IREmitter& ir, IR::Inst& inst) {
-    const IR::U128 x = (IR::U128)inst.GetArg(0);
-    const IR::U128 y = (IR::U128)inst.GetArg(1);
-    const IR::U128 z = (IR::U128)inst.GetArg(2);
-
-    const IR::U128 T0 = ir.VectorExtract(y, z, 32);
-
-    const IR::U128 lower_half = [&] {
-        const IR::U128 T = ir.VectorRotateWholeVectorRight(z, 64);
-        const IR::U128 tmp1 = ir.VectorRotateRight(32, T, 17);
-        const IR::U128 tmp2 = ir.VectorRotateRight(32, T, 19);
-        const IR::U128 tmp3 = ir.VectorLogicalShiftRight(32, T, 10);
-        const IR::U128 tmp4 = ir.VectorEor(tmp1, ir.VectorEor(tmp2, tmp3));
-        const IR::U128 tmp5 = ir.VectorAdd(32, tmp4, ir.VectorAdd(32, x, T0));
-        return ir.VectorZeroUpper(tmp5);
-    }();
-
-    const IR::U64 upper_half = [&] {
-        const IR::U128 tmp1 = ir.VectorRotateRight(32, lower_half, 17);
-        const IR::U128 tmp2 = ir.VectorRotateRight(32, lower_half, 19);
-        const IR::U128 tmp3 = ir.VectorLogicalShiftRight(32, lower_half, 10);
-        const IR::U128 tmp4 = ir.VectorEor(tmp1, ir.VectorEor(tmp2, tmp3));
-
-        // Shuffle the top two 32-bit elements downwards [3, 2, 1, 0] -> [1, 0, 3, 2]
-        const IR::U128 shuffled_d = ir.VectorRotateWholeVectorRight(x, 64);
-        const IR::U128 shuffled_T0 = ir.VectorRotateWholeVectorRight(T0, 64);
-
-        const IR::U128 tmp5 = ir.VectorAdd(32, tmp4, ir.VectorAdd(32, shuffled_d, shuffled_T0));
-        return ir.VectorGetElement(64, tmp5, 0);
-    }();
-
-    const IR::U128 result = ir.VectorSetElement(64, lower_half, 1, upper_half);
-
-    inst.ReplaceUsesWith(result);
-}
-
-IR::U32 SHAchoose(IR::IREmitter& ir, IR::U32 x, IR::U32 y, IR::U32 z) {
-    return ir.Eor(ir.And(ir.Eor(y, z), x), z);
-}
-
-IR::U32 SHAmajority(IR::IREmitter& ir, IR::U32 x, IR::U32 y, IR::U32 z) {
-    return ir.Or(ir.And(x, y), ir.And(ir.Or(x, y), z));
-}
-
-IR::U32 SHAhashSIGMA0(IR::IREmitter& ir, IR::U32 x) {
-    const IR::U32 tmp1 = ir.RotateRight(x, ir.Imm8(2));
-    const IR::U32 tmp2 = ir.RotateRight(x, ir.Imm8(13));
-    const IR::U32 tmp3 = ir.RotateRight(x, ir.Imm8(22));
-
-    return ir.Eor(tmp1, ir.Eor(tmp2, tmp3));
-}
-
-IR::U32 SHAhashSIGMA1(IR::IREmitter& ir, IR::U32 x) {
-    const IR::U32 tmp1 = ir.RotateRight(x, ir.Imm8(6));
-    const IR::U32 tmp2 = ir.RotateRight(x, ir.Imm8(11));
-    const IR::U32 tmp3 = ir.RotateRight(x, ir.Imm8(25));
-
-    return ir.Eor(tmp1, ir.Eor(tmp2, tmp3));
-}
-
-void PolyfillSHA256Hash(IR::IREmitter& ir, IR::Inst& inst) {
-    IR::U128 x = (IR::U128)inst.GetArg(0);
-    IR::U128 y = (IR::U128)inst.GetArg(1);
-    const IR::U128 w = (IR::U128)inst.GetArg(2);
-    const bool part1 = inst.GetArg(3).GetU1();
-
-    for (size_t i = 0; i < 4; i++) {
-        const IR::U32 low_x = ir.VectorGetElement(32, x, 0);
-        const IR::U32 after_low_x = ir.VectorGetElement(32, x, 1);
-        const IR::U32 before_high_x = ir.VectorGetElement(32, x, 2);
-        const IR::U32 high_x = ir.VectorGetElement(32, x, 3);
-
-        const IR::U32 low_y = ir.VectorGetElement(32, y, 0);
-        const IR::U32 after_low_y = ir.VectorGetElement(32, y, 1);
-        const IR::U32 before_high_y = ir.VectorGetElement(32, y, 2);
-        const IR::U32 high_y = ir.VectorGetElement(32, y, 3);
-
-        const IR::U32 choice = SHAchoose(ir, low_y, after_low_y, before_high_y);
-        const IR::U32 majority = SHAmajority(ir, low_x, after_low_x, before_high_x);
-
-        const IR::U32 t = [&] {
-            const IR::U32 w_element = ir.VectorGetElement(32, w, i);
-            const IR::U32 sig = SHAhashSIGMA1(ir, low_y);
-
-            return ir.Add(high_y, ir.Add(sig, ir.Add(choice, w_element)));
-        }();
-
-        const IR::U32 new_low_x = ir.Add(t, ir.Add(SHAhashSIGMA0(ir, low_x), majority));
-        const IR::U32 new_low_y = ir.Add(t, high_x);
-
-        // Shuffle all words left by 1 element: [3, 2, 1, 0] -> [2, 1, 0, 3]
-        const IR::U128 shuffled_x = ir.VectorRotateWholeVectorRight(x, 96);
-        const IR::U128 shuffled_y = ir.VectorRotateWholeVectorRight(y, 96);
-
-        x = ir.VectorSetElement(32, shuffled_x, 0, new_low_x);
-        y = ir.VectorSetElement(32, shuffled_y, 0, new_low_y);
-    }
-
-    inst.ReplaceUsesWith(part1 ? x : y);
-}
-
-template<size_t esize, bool is_signed>
-void PolyfillVectorMultiplyWiden(IR::IREmitter& ir, IR::Inst& inst) {
-    IR::U128 n = (IR::U128)inst.GetArg(0);
-    IR::U128 m = (IR::U128)inst.GetArg(1);
-
-    const IR::U128 wide_n = is_signed ? ir.VectorSignExtend(esize, n) : ir.VectorZeroExtend(esize, n);
-    const IR::U128 wide_m = is_signed ? ir.VectorSignExtend(esize, m) : ir.VectorZeroExtend(esize, m);
-
-    const IR::U128 result = ir.VectorMultiply(esize * 2, wide_n, wide_m);
-
-    inst.ReplaceUsesWith(result);
-}
-
-}  // namespace
-
-void PolyfillPass(IR::Block& block, const PolyfillOptions& polyfill) {
-    if (polyfill == PolyfillOptions{}) {
-        return;
-    }
-
-    IR::IREmitter ir{block};
-
-    for (auto& inst : block) {
-        ir.SetInsertionPointBefore(&inst);
-
-        switch (inst.GetOpcode()) {
-        case IR::Opcode::SHA256MessageSchedule0:
-            if (polyfill.sha256) {
-                PolyfillSHA256MessageSchedule0(ir, inst);
-            }
-            break;
-        case IR::Opcode::SHA256MessageSchedule1:
-            if (polyfill.sha256) {
-                PolyfillSHA256MessageSchedule1(ir, inst);
-            }
-            break;
-        case IR::Opcode::SHA256Hash:
-            if (polyfill.sha256) {
-                PolyfillSHA256Hash(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplySignedWiden8:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<8, true>(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplySignedWiden16:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<16, true>(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplySignedWiden32:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<32, true>(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplyUnsignedWiden8:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<8, false>(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplyUnsignedWiden16:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<16, false>(ir, inst);
-            }
-            break;
-        case IR::Opcode::VectorMultiplyUnsignedWiden32:
-            if (polyfill.vector_multiply_widen) {
-                PolyfillVectorMultiplyWiden<32, false>(ir, inst);
-            }
-            break;
-        default:
-            break;
-        }
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt/verification_pass.cpp

@@ -1,51 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
-/* This file is part of the dynarmic project.
- * Copyright (c) 2016 MerryMage
- * SPDX-License-Identifier: 0BSD
- */
-
-#include <cstdio>
-#include <map>
-
-#include "dynarmic/common/assert.h"
-#include "dynarmic/common/common_types.h"
-#include <ankerl/unordered_dense.h>
-
-#include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/microinstruction.h"
-#include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
-#include "dynarmic/ir/type.h"
-
-namespace Dynarmic::Optimization {
-
-void VerificationPass(const IR::Block& block) {
-    for (const auto& inst : block) {
-        for (size_t i = 0; i < inst.NumArgs(); i++) {
-            const IR::Type t1 = inst.GetArg(i).GetType();
-            const IR::Type t2 = IR::GetArgTypeOf(inst.GetOpcode(), i);
-            if (!IR::AreTypesCompatible(t1, t2)) {
-                std::puts(IR::DumpBlock(block).c_str());
-                ASSERT_FALSE("above block failed validation");
-            }
-        }
-    }
-
-    ankerl::unordered_dense::map<IR::Inst*, size_t> actual_uses;
-    for (const auto& inst : block) {
-        for (size_t i = 0; i < inst.NumArgs(); i++) {
-            const auto arg = inst.GetArg(i);
-            if (!arg.IsImmediate()) {
-                actual_uses[arg.GetInst()]++;
-            }
-        }
-    }
-
-    for (const auto& pair : actual_uses) {
-        ASSERT(pair.first->UseCount() == pair.second);
-    }
-}
-
-}  // namespace Dynarmic::Optimization

src/dynarmic/src/dynarmic/ir/opt_passes.h

@@ -0,0 +1,37 @@
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+/* This file is part of the dynarmic project.
+ * Copyright (c) 2016 MerryMage
+ * SPDX-License-Identifier: 0BSD
+ */
+
+#pragma once
+
+namespace Dynarmic::A32 {
+struct UserCallbacks;
+struct UserConfig;
+}
+
+namespace Dynarmic::A64 {
+struct UserCallbacks;
+struct UserConfig;
+}
+
+namespace Dynarmic::IR {
+class Block;
+}
+
+namespace Dynarmic::Optimization {
+
+struct PolyfillOptions {
+    bool sha256 = false;
+    bool vector_multiply_widen = false;
+
+    bool operator==(const PolyfillOptions&) const = default;
+};
+
+void Optimize(IR::Block& block, const A32::UserConfig& conf, const Optimization::PolyfillOptions& polyfill_options);
+void Optimize(IR::Block& block, const A64::UserConfig& conf, const Optimization::PolyfillOptions& polyfill_options);
+
+}  // namespace Dynarmic::Optimization

src/dynarmic/tests/A32/fuzz_arm.cpp

@@ -24,6 +24,7 @@
 #include "../rand_int.h"
 #include "../unicorn_emu/a32_unicorn.h"
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/common/fp/fpcr.h"
 #include "dynarmic/common/fp/fpsr.h"
 #include "dynarmic/common/llvm_disassemble.h"
@@ -46,7 +47,7 @@ using namespace Dynarmic;
 
 template<typename Fn>
 bool AnyLocationDescriptorForTerminalHas(IR::Terminal terminal, Fn fn) {
-    return Common::VisitVariant<bool>(terminal, [&](auto t) -> bool {
+    return boost::apply_visitor([&](auto t) -> bool {
         using T = std::decay_t<decltype(t)>;
         if constexpr (std::is_same_v<T, IR::Term::Invalid>) {
             return false;
@@ -72,7 +73,7 @@ bool AnyLocationDescriptorForTerminalHas(IR::Terminal terminal, Fn fn) {
             ASSERT_MSG(false, "Invalid terminal type");
             return false;
         }
-    });
+    }, terminal);
 }
 
 bool ShouldTestInst(u32 instruction, u32 pc, bool is_thumb, bool is_last_inst, A32::ITState it_state = {}) {

src/dynarmic/tests/A32/fuzz_thumb.cpp

@@ -22,6 +22,7 @@
 #include "../rand_int.h"
 #include "../unicorn_emu/a32_unicorn.h"
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/frontend/A32/FPSCR.h"
 #include "dynarmic/frontend/A32/PSR.h"
 #include "dynarmic/frontend/A32/a32_location_descriptor.h"
@@ -29,7 +30,7 @@
 #include "dynarmic/frontend/A32/translate/a32_translate.h"
 #include "dynarmic/interface/A32/a32.h"
 #include "dynarmic/ir/basic_block.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 using namespace Dynarmic;
 
@@ -179,13 +180,7 @@ static void RunInstance(size_t run_number, ThumbTestEnv& test_env, A32Unicorn<Th
         while (num_insts < instructions_to_execute_count) {
             A32::LocationDescriptor descriptor = {u32(num_insts * 4), cpsr, A32::FPSCR{}};
             IR::Block ir_block = A32::Translate(descriptor, &test_env, {});
-            Optimization::NamingPass(ir_block);
-            Optimization::A32GetSetElimination(ir_block, {.convert_nz_to_nzc = true});
-            Optimization::DeadCodeElimination(ir_block);
-            Optimization::A32ConstantMemoryReads(ir_block, &test_env);
-            Optimization::ConstantPropagation(ir_block);
-            Optimization::DeadCodeElimination(ir_block);
-            Optimization::VerificationPass(ir_block);
+            Optimization::Optimize(ir_block, &test_env, {});
             printf("\n\nIR:\n%s", IR::DumpBlock(ir_block).c_str());
             printf("\n\nx86_64:\n");
             jit.DumpDisassembly();

src/dynarmic/tests/A32/test_arm_instructions.cpp

@@ -6,6 +6,7 @@
 #include <catch2/catch_test_macros.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/frontend/A32/a32_location_descriptor.h"
 #include "dynarmic/interface/A32/a32.h"
 

src/dynarmic/tests/A32/test_coprocessor.cpp

@@ -8,6 +8,7 @@
 #include <catch2/catch_test_macros.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/frontend/A32/a32_location_descriptor.h"
 #include "dynarmic/interface/A32/a32.h"
 #include "dynarmic/interface/A32/coprocessor.h"

src/dynarmic/tests/A32/test_svc.cpp

@@ -8,6 +8,7 @@
 #include <catch2/catch_test_macros.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 
 using namespace Dynarmic;
 

src/dynarmic/tests/A32/test_thumb_instructions.cpp

@@ -10,6 +10,7 @@
 #include "dynarmic/common/common_types.h"
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/interface/A32/a32.h"
 
 static Dynarmic::A32::UserConfig GetUserConfig(ThumbTestEnv* testenv) {

src/dynarmic/tests/A32/testenv.h

@@ -17,7 +17,6 @@
 #include "dynarmic/common/assert.h"
 #include "dynarmic/common/common_types.h"
 #include "dynarmic/interface/A32/a32.h"
-#include "../native/testenv.h"
 
 template<typename InstructionType_, u32 infinite_loop_u32>
 class A32TestEnv : public Dynarmic::A32::UserCallbacks {

src/dynarmic/tests/A64/a64.cpp

@@ -7,6 +7,7 @@
 #include <oaknut/oaknut.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/common/fp/fpsr.h"
 #include "dynarmic/interface/exclusive_monitor.h"
 

src/dynarmic/tests/A64/fp_min_max.cpp

@@ -12,6 +12,7 @@
 #include "dynarmic/common/common_types.h"
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 
 using namespace Dynarmic;
 

src/dynarmic/tests/A64/fuzz_with_unicorn.cpp

@@ -19,6 +19,7 @@
 #include "../rand_int.h"
 #include "../unicorn_emu/a64_unicorn.h"
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/common/fp/fpcr.h"
 #include "dynarmic/common/fp/fpsr.h"
 #include "dynarmic/common/llvm_disassemble.h"
@@ -28,7 +29,7 @@
 #include "dynarmic/frontend/A64/translate/a64_translate.h"
 #include "dynarmic/ir/basic_block.h"
 #include "dynarmic/ir/opcodes.h"
-#include "dynarmic/ir/opt/passes.h"
+#include "dynarmic/ir/opt_passes.h"
 
 // Must be declared last for all necessary operator<< to be declared prior to this.
 #include <fmt/format.h>

src/dynarmic/tests/A64/misaligned_page_table.cpp

@@ -6,6 +6,7 @@
 #include <catch2/catch_test_macros.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/interface/A64/a64.h"
 
 TEST_CASE("misaligned load/store do not use page_table when detect_misaligned_access_via_page_table is set", "[a64]") {

src/dynarmic/tests/A64/real_world.cpp

@@ -5,6 +5,7 @@
 #include <oaknut/oaknut.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/interface/A64/a64.h"
 
 using namespace Dynarmic;

src/dynarmic/tests/A64/test_invalidation.cpp

@@ -6,6 +6,7 @@
 #include <catch2/catch_test_macros.hpp>
 
 #include "./testenv.h"
+#include "../native/testenv.h"
 #include "dynarmic/interface/A64/a64.h"
 
 using namespace Dynarmic;

src/dynarmic/tests/A64/testenv.h

@@ -12,7 +12,6 @@
 #include "dynarmic/common/assert.h"
 #include "dynarmic/common/common_types.h"
 #include "dynarmic/interface/A64/a64.h"
-#include "../native/testenv.h"
 
 using Vector = Dynarmic::A64::Vector;
 

src/dynarmic/tests/CMakeLists.txt

@@ -6,33 +6,24 @@ add_executable(dynarmic_tests
     fp/mantissa_util_tests.cpp
     fp/unpacked_tests.cpp
     rand_int.h
+    # A32
+    A32/test_arm_disassembler.cpp
+    A32/test_arm_instructions.cpp
+    A32/test_coprocessor.cpp
+    A32/test_svc.cpp
+    A32/test_thumb_instructions.cpp
+    A32/testenv.h
+    decoder_tests.cpp
+    # A64
+    A64/a64.cpp
+    A64/fibonacci.cpp
+    A64/fp_min_max.cpp
+    A64/misaligned_page_table.cpp
+    A64/test_invalidation.cpp
+    A64/real_world.cpp
+    A64/testenv.h
 )
-
-if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-    target_sources(dynarmic_tests PRIVATE
-        A32/test_arm_disassembler.cpp
-        A32/test_arm_instructions.cpp
-        A32/test_coprocessor.cpp
-        A32/test_svc.cpp
-        A32/test_thumb_instructions.cpp
-        A32/testenv.h
-        decoder_tests.cpp
-    )
-endif()
-
-if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-    target_link_libraries(dynarmic_tests PRIVATE merry::oaknut)
-
-    target_sources(dynarmic_tests PRIVATE
-        A64/a64.cpp
-        A64/fibonacci.cpp
-        A64/fp_min_max.cpp
-        A64/misaligned_page_table.cpp
-        A64/test_invalidation.cpp
-        A64/real_world.cpp
-        A64/testenv.h
-    )
-endif()
+target_link_libraries(dynarmic_tests PRIVATE merry::oaknut)
 
 if (DYNARMIC_TESTS_USE_UNICORN)
     target_link_libraries(dynarmic_tests PRIVATE Unicorn::Unicorn)
@@ -40,25 +31,17 @@ if (DYNARMIC_TESTS_USE_UNICORN)
     target_sources(dynarmic_tests PRIVATE
         fuzz_util.cpp
         fuzz_util.h
+        # A32
+        A32/fuzz_arm.cpp
+        A32/fuzz_thumb.cpp
+        unicorn_emu/a32_unicorn.cpp
+        unicorn_emu/a32_unicorn.h
+        # A64
+        A64/fuzz_with_unicorn.cpp
+        A64/verify_unicorn.cpp
+        unicorn_emu/a64_unicorn.cpp
+        unicorn_emu/a64_unicorn.h
     )
-
-    if ("A32" IN_LIST DYNARMIC_FRONTENDS)
-        target_sources(dynarmic_tests PRIVATE
-            A32/fuzz_arm.cpp
-            A32/fuzz_thumb.cpp
-            unicorn_emu/a32_unicorn.cpp
-            unicorn_emu/a32_unicorn.h
-        )
-    endif()
-
-    if ("A64" IN_LIST DYNARMIC_FRONTENDS)
-        target_sources(dynarmic_tests PRIVATE
-            A64/fuzz_with_unicorn.cpp
-            A64/verify_unicorn.cpp
-            unicorn_emu/a64_unicorn.cpp
-            unicorn_emu/a64_unicorn.h
-        )
-    endif()
 endif()
 
 if ("riscv" IN_LIST ARCHITECTURE)
@@ -69,9 +52,6 @@ if ("x86_64" IN_LIST ARCHITECTURE)
     target_link_libraries(dynarmic_tests PRIVATE xbyak::xbyak)
     target_architecture_specific_sources(dynarmic_tests "x86_64"
         x64_cpu_info.cpp
-    )
-
-    target_architecture_specific_sources(dynarmic_tests "x86_64"
         native/preserve_xmm.cpp
     )
 
@@ -85,50 +65,70 @@ endif()
 
 include(CreateDirectoryGroups)
 
-if (("A32" IN_LIST DYNARMIC_FRONTENDS) AND ("A64" IN_LIST DYNARMIC_FRONTENDS))
-    add_executable(dynarmic_print_info
-        print_info.cpp
-    )
-
-    create_target_directory_groups(dynarmic_print_info)
-
-    target_link_libraries(dynarmic_print_info PRIVATE dynarmic Boost::headers fmt::fmt merry::mcl)
-    target_include_directories(dynarmic_print_info PRIVATE . ../src)
-    target_compile_options(dynarmic_print_info PRIVATE ${DYNARMIC_CXX_FLAGS})
-    target_compile_definitions(dynarmic_print_info PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
+#
+# dynarmic_print_info
+#
+add_executable(dynarmic_print_info
+    print_info.cpp
+)
+create_target_directory_groups(dynarmic_print_info)
+target_link_libraries(dynarmic_print_info PRIVATE dynarmic fmt::fmt merry::mcl)
+if (BOOST_NO_HEADERS)
+    target_link_libraries(dynarmic_print_info PRIVATE Boost::variant Boost::icl Boost::pool)
+else()
+    target_link_libraries(dynarmic_print_info PRIVATE Boost::headers)
 endif()
+target_include_directories(dynarmic_print_info PRIVATE . ../src)
+target_compile_options(dynarmic_print_info PRIVATE ${DYNARMIC_CXX_FLAGS})
+target_compile_definitions(dynarmic_print_info PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
 
-if (("A32" IN_LIST DYNARMIC_FRONTENDS) AND ("A64" IN_LIST DYNARMIC_FRONTENDS))
-    add_executable(dynarmic_test_generator
-        fuzz_util.cpp
-        fuzz_util.h
-        test_generator.cpp
-    )
+#
+# dynarmic_test_generator
+#
+add_executable(dynarmic_test_generator
+    fuzz_util.cpp
+    fuzz_util.h
+    test_generator.cpp
+)
 
-    create_target_directory_groups(dynarmic_test_generator)
+create_target_directory_groups(dynarmic_test_generator)
 
-    target_link_libraries(dynarmic_test_generator PRIVATE dynarmic Boost::headers fmt::fmt merry::mcl)
-    target_include_directories(dynarmic_test_generator PRIVATE . ../src)
-    target_compile_options(dynarmic_test_generator PRIVATE ${DYNARMIC_CXX_FLAGS})
-    target_compile_definitions(dynarmic_test_generator PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
+target_link_libraries(dynarmic_test_generator PRIVATE dynarmic fmt::fmt merry::mcl)
+if (BOOST_NO_HEADERS)
+    target_link_libraries(dynarmic_test_generator PRIVATE Boost::variant Boost::icl Boost::pool)
+else()
+    target_link_libraries(dynarmic_test_generator PRIVATE Boost::headers)
 endif()
+target_include_directories(dynarmic_test_generator PRIVATE . ../src)
+target_compile_options(dynarmic_test_generator PRIVATE ${DYNARMIC_CXX_FLAGS})
+target_compile_definitions(dynarmic_test_generator PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
 
-if (("A32" IN_LIST DYNARMIC_FRONTENDS) AND ("A64" IN_LIST DYNARMIC_FRONTENDS))
-    add_executable(dynarmic_test_reader
-        test_reader.cpp
-    )
-
-    create_target_directory_groups(dynarmic_test_reader)
-
-    target_link_libraries(dynarmic_test_reader PRIVATE dynarmic Boost::headers fmt::fmt merry::mcl)
-    target_include_directories(dynarmic_test_reader PRIVATE . ../src)
-    target_compile_options(dynarmic_test_reader PRIVATE ${DYNARMIC_CXX_FLAGS})
-    target_compile_definitions(dynarmic_test_reader PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
+#
+# dynarmic_test_reader
+#
+add_executable(dynarmic_test_reader
+    test_reader.cpp
+)
+create_target_directory_groups(dynarmic_test_reader)
+target_link_libraries(dynarmic_test_reader PRIVATE dynarmic fmt::fmt merry::mcl)
+if (BOOST_NO_HEADERS)
+    target_link_libraries(dynarmic_test_reader PRIVATE Boost::variant Boost::icl Boost::pool)
+else()
+    target_link_libraries(dynarmic_test_reader PRIVATE Boost::headers)
 endif()
+target_include_directories(dynarmic_test_reader PRIVATE . ../src)
+target_compile_options(dynarmic_test_reader PRIVATE ${DYNARMIC_CXX_FLAGS})
+target_compile_definitions(dynarmic_test_reader PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)
 
+#
 create_target_directory_groups(dynarmic_tests)
 
-target_link_libraries(dynarmic_tests PRIVATE dynarmic Boost::headers Catch2::Catch2WithMain fmt::fmt merry::mcl)
+target_link_libraries(dynarmic_tests PRIVATE dynarmic Catch2::Catch2WithMain fmt::fmt merry::mcl)
+if (BOOST_NO_HEADERS)
+    target_link_libraries(dynarmic_tests PRIVATE Boost::variant Boost::icl Boost::pool)
+else()
+    target_link_libraries(dynarmic_tests PRIVATE Boost::headers)
+endif()
 target_include_directories(dynarmic_tests PRIVATE . ../src)
 target_compile_options(dynarmic_tests PRIVATE ${DYNARMIC_CXX_FLAGS})
 target_compile_definitions(dynarmic_tests PRIVATE FMT_USE_USER_DEFINED_LITERALS=1)