Docs: Update for release FEX-2609.1

InstcountCI: Update
FEXCore/OpcodeDispatcher: Fixes syscall instruction on Linux
2026-10-17 00:00:39 +02:00 · 2026-09-21 11:44:20 -07:00 · 2026-09-21 11:43:34 -07:00 · 2026-09-21 11:43:28 -07:00 · 2026-09-07 23:42:35 -07:00 · 2026-09-07 23:38:31 -07:00
1953 changed files with 223423 additions and 254432 deletions
@@ -32,7 +32,7 @@ AttributeMacros:
 BinPackArguments: true
 BinPackParameters: true
 BitFieldColonSpacing: Both
-BreakAfterAttributes: Always # clang 16 required
+BreakAfterAttributes: Leave
 BreakBeforeBraces: Attach
 BreakBeforeBinaryOperators: None
 BreakBeforeInlineASMColon: OnlyMultiline # clang 16 required
@@ -60,7 +60,7 @@ IndentRequires: false
 IndentWidth: 2
 InsertBraces: true
 KeepEmptyLinesAtTheStartOfBlocks: true
-LambdaBodyIndentation: OuterScope
+LambdaBodyIndentation: Signature
 LineEnding: LF # clang 16 required
 MaxEmptyLinesToKeep: 2
 NamespaceIndentation: Inner
@@ -1,12 +1,12 @@
 # This file is used to ignore files and directories from clang-format
-
-# Ignore all files in the External directory
-External/*
-
-# SoftFloat-3e code doesn't belong to us 
-FEXCore/Source/Common/SoftFloat-3e/*
 Source/Common/cpp-optparse/*

 # Files with human-indented tables for readability - don't mess with these
-FEXCore/Source/Interface/Core/X86Tables/*
+FEXCore/Source/Interface/Core/X86Tables/*.cpp

+# Inline headers with list-like content that can't be processed individually
+Source/Tools/LinuxEmulation/LinuxSyscalls/x*/SyscallsNames.inl
+Source/Tools/LinuxEmulation/LinuxSyscalls/x*/Ioctl/*.inl
+
+# Include files in unittests
+unittests/*ASM/Includes/*.inc
@@ -13,3 +13,15 @@

 # Second reformat to find fixed point PR#3577
 905aa935f5ce344a48ef4d5edab3c31efa8d793e
+
+# Reformat of CodeEmitter inl files
+8760c593ece92d7e9fa94c40da0368fd367c9cad
+
+# Whole-tree reformat with clang-format-19
+5267cde60e7642852d18f20ae8568643bb5293d5
+
+# Minor reformat with clang-format-19
+9fdd96af61c969cb5732471223f00eda64b7a069
+
+# Reformat of X86Tables.h
+ba2b0ef809f66f1a6d334f000798fa2ceafab26f
@@ -13,6 +13,7 @@ env:
  BUILD_TYPE: Release
  CC: clang
  CXX: clang++
+  FEX_PORTABLE: 1

 jobs:
  build_plus_test:
@@ -23,236 +24,139 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

-    - name: Set runner label
-      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
+    - name: Set runner info
      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+        echo "runner_name=$(hostname)" >> $GITHUB_ENV

-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
-
-    - name : submodule checkout
-      # Need to update submodules
-      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DBUILD_FEX_LINUX_TESTS=True -DBUILD_THUNKS=True -DCMAKE_INSTALL_PREFIX=${{runner.workspace}}/build/install
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True \
+          -DENABLE_X86_HOST_DEBUG=True -DBUILD_FEX_LINUX_TESTS=True -DBUILD_THUNKS=True \
+          -DCMAKE_INSTALL_PREFIX="$PWD"/build/install
+
+    # These steps make a lot of noise but rarely fail.
+    # Put them in a separate step to make normal build logs easier to parse
+    - name: Noisy Build Targets
+      run: cmake --build build --target asm_files 32bit_asm_files JemallocLibs Catch2 vixl cephes_128bit

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE
+      id: build
+      run: cmake --build build

    - name: Install
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target install
+      run: cmake --build build --target install

-    - name: gcc target tests 64
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the gvisor tests
-      run: cmake --build . --config $BUILD_TYPE --target gcc_target_tests_64
+    # GCC tests
+    - name: GCC64 Target Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: gcc_target_tests_64

-    - name: GCC64 Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_GCC64.log || true
+    - name: GCC32 Target Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: gcc_target_tests_32

-    - name: gcc target tests 32
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the gvisor tests
-      run: cmake --build . --config $BUILD_TYPE --target gcc_target_tests_32
+    # API tests
+    - name: API Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: api_tests

-    - name: GCC32 Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_GCC32.log || true
+    - name: FEXCore API Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: fexcore_apitests

-    - name: APITest tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target api_tests
+    # ARM emission tests
+    - name: ARM Emitter Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: emitter_tests

-    - name: APITest Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_APITests.log || true
-
-    - name: FEXCore APITest tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target fexcore_apitests
-
-    - name: FEXCore APITest Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_FEXCoreAPITests.log || true
-
-    - name: ARMEmitter tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target emitter_tests
-
-    - name: ARMEmitter Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ARMEmitterTests.log || true
-
-    - name: FEXLinuxTests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target fex_linux_tests_all
-
-    - name: FEXLinuxTests Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_FEXLinuxTests.log || true
+    # Linux  tests
+    - name: FEX Linux Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: fex_linux_tests_all
+      env:
+        FEX_PORTABLE: 0

+    # Thunking
    - name: Thunkgen tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target thunkgen_tests
-
-    - name: Thunkgen Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ThunkgenTests.log || true
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: thunkgen_tests

    - name: Test GL No-Thunks
-      if: matrix.arch[1] == 'x64'
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
+      if: ${{ steps.build.outcome == 'success' && matrix.arch[1] == 'x64' }}
+      uses: ./.github/workflows/test
+      with:
+        target: thunk_functional_tests_nothunks
      env:
-        DISPLAY: ":0"
-      run: cmake --build . --config $BUILD_TYPE --target thunk_functional_tests_nothunks
-
-    - name: No thunks Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_NoThunkResults.log || true
+        DISPLAY: ':0'

    - name: Test GL Thunks
-      if: matrix.arch[1] == 'x64'
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
+      if: ${{ steps.build.outcome == 'success' && matrix.arch[1] == 'x64' }}
+      uses: ./.github/workflows/test
+      with:
+        target: thunk_functional_tests_thunks
      env:
-        DISPLAY: ":0"
-      run: cmake --build . --config $BUILD_TYPE --target thunk_functional_tests_thunks
-
-    - name: Thunks Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ThunkResults.log || true
+        DISPLAY: ':0'

+    # ASM tests
    - name: ASM Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: asm_tests

-    - name: ASM Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM.log || true
+    # POSIX tests
+    - name: POSIX Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: posix_tests

-    - name: Posix Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the posixtest
-      run: cmake --build . --config $BUILD_TYPE --target posix_tests
-
-    - name: Posix Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_Posix.log || true
-
-    - name: gvisor tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the gvisor tests
-      run: cmake --build . --config $BUILD_TYPE --target gvisor_tests
-
-    - name: GVisor Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_GVisor.log || true
+    # GVisor tests
+    - name: GVisor Tests
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: gvisor_tests

+    # Struct verifier tests
    - name: Struct verifier tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target struct_verifier
-
-    - name: Struct verifier Test Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_StructVerifier.log || true
-
-    - name: Truncate test results
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Cap out the log files at 20M in case something crash spins and dumps fault text
-      # ASM tests get quite close to 10MB
-      run: truncate --size="<20M" ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log || true
+      if: steps.build.outcome == 'success'
+      uses: ./.github/workflows/test
+      with:
+        target: struct_verifier

    - name: Remove old SHM regions
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: cmake --build . --config $BUILD_TYPE --target remove_old_shm_regions
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
+      run: cmake --build build --target remove_old_shm_regions

    - name: Upload results
      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
+      uses: actions/upload-artifact@v6
      timeout-minutes: 1
      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        name: Results-${{ env.runner_name }}-${{ env.runner_label }}
+        path: results/*.log
        retention-days: 3
@@ -20,6 +20,7 @@ env:
  BUILD_TYPE: Release
  CC: clang
  CXX: clang++
+  FEX_PORTABLE: 1

 jobs:
  glibc_fault_test:
@@ -30,164 +31,94 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

-    - name: Set runner label
-      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
+    - name: Set runner info
      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+        echo "runner_name=$(hostname)" >> $GITHUB_ENV

-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
-
-    - name : submodule checkout
-      # Need to update submodules
-      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DBUILD_FEX_LINUX_TESTS=True -DENABLE_GLIBC_ALLOCATOR_HOOK_FAULT=True -DENABLE_JEMALLOC_GLIBC_ALLOC=False -DCMAKE_INSTALL_PREFIX=${{runner.workspace}}/build/install
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False \
+          -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DBUILD_FEX_LINUX_TESTS=True \
+          -DENABLE_GLIBC_ALLOCATOR_HOOK_FAULT=True -DENABLE_JEMALLOC_GLIBC_ALLOC=False \
+          -DCMAKE_INSTALL_PREFIX="$PWD"/build/install
+
+    # These steps make a lot of noise but rarely fail.
+    # Put them in a separate step to make normal build logs easier to parse
+    - name: Noisy Build Targets
+      run: cmake --build build --target asm_files 32bit_asm_files JemallocLibs Catch2 vixl cephes_128bit

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE
+      run: cmake --build build

    - name: Install
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target install
+      run: cmake --build build --target install

-    - name: gcc target tests 64
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the gvisor tests
-      run: cmake --build . --config $BUILD_TYPE --target gcc_target_tests_64
-
-    - name: GCC64 Test Results move
+    # GCC tests
+    - name: GCC64 Target Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_GCC64.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: gcc_target_tests_64

-    - name: gcc target tests 32
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the gvisor tests
-      run: cmake --build . --config $BUILD_TYPE --target gcc_target_tests_32
-
-    - name: GCC32 Test Results move
+    - name: GCC32 Target Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_GCC32.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: gcc_target_tests_32

-    - name: APITest tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target api_tests
-
-    - name: APITest Test Results move
+    # API Tests
+    - name: API Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_APITests.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: api_tests

-    - name: FEXCore APITest tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target fexcore_apitests
-
-    - name: FEXCore APITest Test Results move
+    - name: FEXCore API Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_FEXCoreAPITests.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: fexcore_apitests

-    - name: FEXLinuxTests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      run: cmake --build . --config $BUILD_TYPE --target fex_linux_tests_all
-
-    - name: FEXLinuxTests Results move
+    # Linux tests
+    - name: FEX Linux Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_FEXLinuxTests.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: fex_linux_tests_all

+    # ASM Tests
    - name: ASM Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
-
-    - name: ASM Test Results move
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: asm_tests

-    - name: Posix Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the posixtest
-      run: cmake --build . --config $BUILD_TYPE --target posix_tests
-
-    - name: Posix Test Results move
+    # POSIX Tests
+    - name: POSIX Tests
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_Posix.log || true
-
-    - name: Truncate test results
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Cap out the log files at 20M in case something crash spins and dumps fault text
-      # ASM tests get quite close to 10MB
-      run: truncate --size="<20M" ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: posix_tests

    - name: Remove old SHM regions
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: cmake --build . --config $BUILD_TYPE --target remove_old_shm_regions
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
+      run: cmake --build build --target remove_old_shm_regions

    - name: Upload results
      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
+      uses: actions/upload-artifact@v6
      timeout-minutes: 1
      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        name: Results-${{ env.runner_name }}-${{ env.runner_label }}
+        path: results/*.log
        retention-days: 3

@@ -13,6 +13,7 @@ env:
  BUILD_TYPE: Release
  CC: clang
  CXX: clang++
+  FEX_PORTABLE: 1

 jobs:
  hostrunner_tests:
@@ -23,84 +24,45 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

-    - name: Set runner label
-      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
+    - name: Set runner info
      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+        echo "runner_name=$(hostname)" >> $GITHUB_ENV

-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
-
-    - name : submodule checkout
-      # Need to update submodules
-      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False \
+          -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+
+    # These steps make a lot of noise but rarely fail.
+    # Put them in a separate step to make normal build logs easier to parse
+    - name: Noisy Build Targets
+      run: cmake --build build --target asm_files 32bit_asm_files JemallocLibs Catch2 vixl cephes_128bit

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE
+      run: cmake --build build

+    # ASM tests
    - name: ASM Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
-
-    - name: ASM Test Results move
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM.log || true
-
-    - name: Truncate test results
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Cap out the log files at 20M in case something crash spins and dumps fault text
-      # ASM tests get quite close to 10MB
-      run: truncate --size="<20M" ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log || true
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
+      uses: ./.github/workflows/test
+      with:
+        target: asm_tests

    - name: Upload results
      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
+      uses: actions/upload-artifact@v6
      timeout-minutes: 1
      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        name: Results-${{ env.runner_name }}-${{ env.runner_label }}
+        path: results/*.log
        retention-days: 3

@@ -23,124 +23,56 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

-    - name: Set runner label
-      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
+    - name: Set runner info
      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+        echo "runner_name=$(hostname)" >> $GITHUB_ENV

-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

-    - name : submodule checkout
-      # Need to update submodules
+    - name: Set VIXL_SIM_ENABLED
      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
-
-    - name: Set vixl_sim x86
-      if: matrix.arch[1] == 'x64'
-      run: |
-        echo "VIXL_SIM_ENABLED=True" >> $GITHUB_ENV
-
-    - name: Set vixl_sim Arm64
-      if: matrix.arch[1] == 'ARM64'
-      run: |
-        echo "VIXL_SIM_ENABLED=False" >> $GITHUB_ENV
+        case '${{ matrix.arch[1] }}' in
+          x64) _sim=True ;;
+          ARM64) _sim=False ;;
+        esac
+        echo "VIXL_SIM_ENABLED=$_sim" >> $GITHUB_ENV

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_VIXL_SIMULATOR=$VIXL_SIM_ENABLED -DENABLE_VIXL_DISASSEMBLER=True -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_VIXL_SIMULATOR=$VIXL_SIM_ENABLED \
+          -DENABLE_VIXL_DISASSEMBLER=True -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
      env:
        FEX_DISABLETELEMETRY: 1
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE --target CodeSizeValidation instcountci_test_files
+      run: cmake --build build --target CodeSizeValidation instcountci_test_files

    - name: Instruction Count Tests
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target instcountci_tests
-
-    - name: Instruction Count Test Results move
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_InstCountCI.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: instcountci_tests

    - name: Update local repo instcount
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: cmake --build . --config $BUILD_TYPE --target instcountci_update_tests
+      run: cmake --build build --target instcountci_update_tests

-    - name: Get instcountCI diff
+    - name: Check InstCountCI diff
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{github.workspace}}/
-      run: git diff --output=${{runner.workspace}}/build/InstCountCI.diff
-
-    - name: Check if InstCountCI Diff exists
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{github.workspace}}/
-      # Check if the file is empty
-      run: sh -c "! test -s ${{runner.workspace}}/build/InstCountCI.diff"
-
-    - name: Truncate test results
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Cap out the log files at 20M in case something crash spins and dumps fault text
-      # ASM tests get quite close to 10MB
-      run: truncate --size="<20M" ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log || true
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
+      run: git --no-pager diff --exit-code HEAD

    - name: Upload results
      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
+      uses: actions/upload-artifact@v6
      timeout-minutes: 1
      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        name: Results-${{ env.runner_name }}-${{ env.runner_label }}
+        path: results/*.log
        retention-days: 3
-
-    - name: Upload results InstCountCI
-      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
-      timeout-minutes: 1
-      with:
-        name: Results-${{ env.runner_name }}-instcountci
-        path: ${{runner.workspace}}/build/InstCountCI.diff
-        retention-days: 3
-
@@ -20,7 +20,10 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

    - name: Set runner label
      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
@@ -28,74 +31,23 @@ jobs:
    - name: Add MingGW to PATH
      run: echo "$HOME/llvm-mingw/build/bin/" >> $GITHUB_PATH

-    - name: Set CC x86
-      if: matrix.arch[1] == 'x64'
+    - name: Set CC
      run: |
-        echo "MINGW_TRIPLE=x86_64-w64-mingw32" >> $GITHUB_ENV
+        case '${{ matrix.arch[1] }}' in
+          x64) _cpu=x86_64 ;;
+          ARM64) _cpu=aarch64 ;;
+          ARM64EC) _cpu=arm64ec ;;
+        esac
+        echo "MINGW_TRIPLE=${_cpu}-w64-mingw32" >> $GITHUB_ENV

-    - name: Set CC Arm64
-      if: matrix.arch[1] == 'ARM64'
-      run: |
-        echo "MINGW_TRIPLE=aarch64-w64-mingw32" >> $GITHUB_ENV
-
-    - name: Set CC Arm64EC
-      if: matrix.arch[1] == 'ARM64EC'
-      run: |
-        echo "MINGW_TRIPLE=arm64ec-w64-mingw32" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
-      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-
-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
-
-    - name : submodule checkout
-      # Need to update submodules
-      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DCMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/toolchain_mingw.cmake -DMINGW_TRIPLE=$MINGW_TRIPLE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DBUILD_TESTS=False -DCMAKE_INSTALL_PREFIX=${{runner.workspace}}/build/install
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DCMAKE_TOOLCHAIN_FILE=$GITHUB_WORKSPACE/Data/CMake/toolchain_mingw.cmake \
+        -DMINGW_TRIPLE=$MINGW_TRIPLE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DBUILD_TESTING=False \
+        -DCMAKE_INSTALL_PREFIX="$PWD"/build/install

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
-
-    - name: Upload results
-      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
-      timeout-minutes: 1
-      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
-        retention-days: 3
-
+      run: cmake --build build
@@ -1,7 +1,7 @@
-# Inspired by LLVM's pr-code-format.yml at 
+# Inspired by LLVM's pr-code-format.yml at
 # https://github.com/llvm/llvm-project/blob/main/.github/workflows/pr-code-format.yml

-name: "Check code formatting"
+name: Check code formatting
 on:
  pull_request:
    branches:
@@ -13,7 +13,7 @@ jobs:
    if: github.repository == 'FEX-Emu/FEX'

    steps:
-      - name: Fetch FEX sources
+      - name: Checkout
        uses: actions/checkout@v4
        with:
          ref: ${{ github.event.pull_request.head.sha }}
@@ -27,50 +27,37 @@ jobs:
          deepen_length: 500

      - name: Get changed files
-        id: changed-files
-        uses: tj-actions/changed-files@v39
-        with:
-          separator: ","
-          skip_initial_fetch: true
-
-      - name: "Listed files"
-        env:
-          CHANGED_FILES: ${{ steps.changed-files.outputs.all_changed_files }}
        run: |
-          echo "Formatting files:"
-          echo "$CHANGED_FILES"
+          BASE=$(git merge-base main HEAD)
+          FILES=$(git diff --name-only "$BASE" | tr '\n' ',' | sed 's/,$//')
+          echo "CHANGED_FILES=$FILES" >> $GITHUB_ENV

-      - name: Check for correct clang-format version
-        run: clang-format --version | grep -qF '16.0.6'
+          echo "Changed files:"
+          echo "$FILES"

-      - name: Check git-clang-format-16 exists
-        run: which git-clang-format-16
+      - name: Check git-clang-format-19 exists
+        run: which git-clang-format-19

      - name: Setup Python env
        uses: actions/setup-python@v4
        with:
-          python-version: '3.11'
-          cache: 'pip'
-          cache-dependency-path: './External/code-format-helper/requirements_formatting.txt'
+          python-version: 3.11
+          cache: pip
+          cache-dependency-path: ./External/code-format-helper/requirements_formatting.txt

      - name: Install python dependencies
        run: pip install -r ./External/code-format-helper/requirements_formatting.txt

      - name: Run code formatter
        env:
-          CLANG_FORMAT_PATH: 'git-clang-format-16'
+          CLANG_FORMAT_PATH: git-clang-format-19
          GITHUB_PR_NUMBER: ${{ github.event.pull_request.number }}
          START_REV: ${{ github.event.pull_request.base.sha }}
          END_REV: ${{ github.event.pull_request.head.sha }}
-          CHANGED_FILES: ${{ steps.changed-files.outputs.all_changed_files }}
-        # TODO(pmatos): Once we adopt v18, we should be able
-        # to take advantage of the new --diff_from_common_commit option
-        # explicitly in code-format-helper.py and not have to diff starting at
-        # the merge base.
        run: |
          python ./External/code-format-helper/code-format-helper.py \
-            --repo "FEX-emu/FEX" \
-            --issue-number $GITHUB_PR_NUMBER \
-            --start-rev $(git merge-base $START_REV $END_REV) \
-            --end-rev $END_REV \
+            --repo "FEX-Emu/FEX" \
+            --issue-number "$GITHUB_PR_NUMBER" \
+            --start-rev "$START_REV" \
+            --end-rev "$END_REV" \
            --changed-files "$CHANGED_FILES"
@@ -0,0 +1,33 @@
+name: Setup Build Environment
+description: Setup RootFS and build environment
+
+inputs:
+  setup-rootfs:
+    description: 'Whether or not to set up the rootfs'
+    default: true
+
+runs:
+  using: composite
+  steps:
+    - name: Set rootfs paths
+      if: ${{ inputs.setup-rootfs == 'true' }}
+      shell: bash
+      run: |
+        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
+        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+
+    - name: Update RootFS cache
+      if: ${{ inputs.setup-rootfs == 'true' }}
+      shell: bash
+      run: python3 Scripts/CI_FetchRootFS.py
+
+    - name: Checkout Submodules
+      shell: bash
+      run: |
+        git submodule sync --recursive
+        git submodule update --init --depth 1
+
+    - name: Clean Build Environment
+      shell: bash
+      run: rm -Rf build
@@ -0,0 +1,72 @@
+name: steamrt4 build
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - main
+
+env:
+  DEBIAN_FRONTEND: noninteractive
+  BUILD_TYPE: Release
+  CC: clang
+  CXX: clang++
+
+jobs:
+  steamrt4_build:
+    runs-on: ${{ matrix.arch }}
+    strategy:
+      matrix:
+        arch: [[self-hosted, ARM64, distrobox]]
+      fail-fast: false
+
+    steps:
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'
+
+    - name: Set runner label
+      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env
+      with:
+        setup-rootfs: false
+
+    # Setup everything required.
+    - name : distrobox setup
+      run: |
+        distrobox create -Y -i registry.gitlab.steamos.cloud/steamrt/steamrt4/sdk/arm64:4.0.20251117.183306 steamrt4 || true
+        distrobox upgrade steamrt4
+        distrobox enter --name steamrt4 -- sudo apt-get install -y \
+          git cmake ninja-build ccache \
+          lld clang clang-tools \
+          libclang-dev llvm-dev \
+          libstdc++-14-dev-i386-cross libgcc-14-dev-i386-cross \
+          libstdc++-14-dev-amd64-cross libgcc-14-dev-amd64-cross
+
+    - name: Configure CMake
+      run: |
+        distrobox enter --name steamrt4 -- cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE \
+          -G Ninja -DBUILD_STEAM_SUPPORT=True -DENABLE_LTO=True -DENABLE_ASSERTIONS=False -DBUILD_THUNKS=True \
+          -DBUILD_FEXCONFIG=False -DBUILD_TESTING=False -DENABLE_CLANG_THUNKS=True -DUSE_LINKER=lld \
+          -DCMAKE_INSTALL_PREFIX=/usr
+
+    - name: Build
+      run: distrobox enter --name steamrt4 -- cmake --build build
+
+    - name: install
+      run: DESTDIR="$PWD"/install distrobox enter --name steamrt4 -- cmake --build build -t install
+
+    - name: Upload libraries
+      uses: actions/upload-artifact@v6
+      timeout-minutes: 1
+      with:
+        overwrite: true
+        name: steamrt4_steampipe_depot
+        path: ${{ github.workspace }}/install/*
+        retention-days: 60
+        compression-level: 9
@@ -0,0 +1,21 @@
+name: Run Test and Store Logs
+description: Run a test and store the log.
+inputs:
+  target:
+    description: 'The test target to run'
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Run Tests
+      shell: bash
+      run: cmake --build build --target ${{ inputs.target }}
+
+    - name: Move and Truncate Results
+      if: ${{ always() }}
+      shell: bash
+      run: |
+        mkdir -p results
+        mv build/Testing/Temporary/LastTest.log results/${{ inputs.target }}.log || true
+        truncate --size="<20M" results/${{ inputs.target }}.log || true
@@ -13,6 +13,7 @@ env:
  BUILD_TYPE: Release
  CC: clang
  CXX: clang++
+  FEX_PORTABLE: 1

 jobs:
  vixl_simulator:
@@ -24,112 +25,59 @@ jobs:
      fail-fast: false

    steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'

-    - name: Set runner label
-      run: echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
-
-    - name: Set rootfs paths
+    - name: Set runner info
      run: |
-        echo "FEX_ROOTFS_MOUNT=/mnt/AutoNFS/rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS_PATH=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "FEX_ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
-        echo "ROOTFS=$HOME/Rootfs/" >> $GITHUB_ENV
+        echo "runner_label=${{ matrix.arch[1] }}" >> $GITHUB_ENV
+        echo "runner_name=$(hostname)" >> $GITHUB_ENV

-    - name: Update RootFS cache
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py
-
-    - name : submodule checkout
-      # Need to update submodules
-      run: |
-        git submodule sync --recursive
-        git submodule update --init --depth 1
-
-    - name: Clean Build Environment
-      run: rm -Rf ${{runner.workspace}}/build
-
-    - name: Create Build Environment
-      # Some projects don't allow in-source building, so create a separate build directory
-      # We'll use this as our working directory for all subsequent commands
-      run: cmake -E make_directory ${{runner.workspace}}/build
+    - name: Setup Build Environment
+      uses: ./.github/workflows/setup-env

    - name: Configure CMake
-      # Use a bash shell so we can use the same syntax for environment variable
-      # access regardless of the host operating system
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Note the current convention is to use the -S and -B options here to specify source
-      # and build directories, but this is only available with CMake 3.13 and higher.
-      # The CMake binaries on the Github Actions machines are (as of this writing) 3.12
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_VIXL_SIMULATOR=True -DENABLE_VIXL_DISASSEMBLER=True -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+      run: |
+        cmake -S . -B build -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_VIXL_SIMULATOR=True -DENABLE_LTO=False \
+          -DENABLE_VIXL_DISASSEMBLER=True -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+
+    # These steps make a lot of noise but rarely fail.
+    # Put them in a separate step to make normal build logs easier to parse
+    - name: Noisy Build Targets
+      run: cmake --build build --target asm_files 32bit_asm_files JemallocLibs Catch2 vixl cephes_128bit

    - name: Build
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the build.  You can specify a specific target with "--target <NAME>"
-      run: cmake --build . --config $BUILD_TYPE
+      run: cmake --build build

    - name: ASM Tests - SVE256
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
-
-    - name: ASM Test SVE256 Results move
      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM_SVE256Bit.log || true
+      uses: ./.github/workflows/test
+      with:
+        target: asm_tests

    - name: ASM Tests - SVE128
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
+      if: ${{ always() }}
+      uses: ./.github/workflows/test
      env:
        FEX_FORCESVEWIDTH: "128"
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
-
-    - name: ASM Test 128-bit Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM_SVE128Bit.log || true
+      with:
+        target: asm_tests

    - name: ASM Tests - ASIMD
-      working-directory: ${{runner.workspace}}/build
-      shell: bash
+      if: ${{ always() }}
+      uses: ./.github/workflows/test
      env:
        FEX_HOSTFEATURES: "disablesve"
-      # Execute the unit tests
-      run: cmake --build . --config $BUILD_TYPE --target asm_tests
-
-    - name: ASM Test ASIMD Results move
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_ASM_ASIMD.log || true
-
-    - name: Truncate test results
-      if: ${{ always() }}
-      shell: bash
-      working-directory: ${{runner.workspace}}/build
-      # Cap out the log files at 20M in case something crash spins and dumps fault text
-      # ASM tests get quite close to 10MB
-      run: truncate --size="<20M" ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log || true
-
-    - name: Set runner name
-      if: ${{ always() }}
-      run: echo "runner_name=$(hostname)" >> $GITHUB_ENV
+      with:
+        target: asm_tests

    - name: Upload results
      if: ${{ always() }}
-      uses: 'actions/upload-artifact@v3'
+      uses: actions/upload-artifact@v6
      timeout-minutes: 1
      with:
-        name: Results-${{ env.runner_name }}
-        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        name: Results-${{ env.runner_name }}-${{ env.runner_label }}
+        path: results/*.log
        retention-days: 3
-
@@ -0,0 +1,49 @@
+name: Wine DLL Build
+description: Build a wow64 or arm64ec Wine DLL
+
+inputs:
+  target:
+    description: 'The target (arm64ec or wow64)'
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Clean Build Environment
+      shell: bash
+      run: rm -Rf build_${{ inputs.target }}
+
+    - name: Configure CMake
+      shell: bash
+      run: |
+        case "${{ inputs.target }}" in
+          wow64) _cc=aarch64 ;;
+          arm64ec) _cc=arm64ec ;;
+        esac
+
+        cmake -S . -B build_${{ inputs.target }} -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DCMAKE_TOOLCHAIN_FILE=Data/CMake/toolchain_mingw.cmake \
+          -DMINGW_TRIPLE=${_cc}-w64-mingw32 -DCMAKE_INSTALL_LIBDIR=/usr/lib/wine/aarch64-windows -G Ninja \
+          -DENABLE_LTO=False -DENABLE_ASSERTIONS=False -DENABLE_JEMALLOC_GLIBC_ALLOC=False \
+          -DBUILD_TESTING=False -DCMAKE_INSTALL_PREFIX=/usr -DTUNE_ARCH=generic -DTUNE_CPU=none -DRANGES_NATIVE=OFF
+
+    - name: Build
+      shell: bash
+      run: cmake --build build_${{ inputs.target }}
+
+    - name: Install
+      shell: bash
+      run: DESTDIR="$PWD"/install cmake --build build_${{ inputs.target }} -t install
+
+    - name: Configure UnixLib
+      shell: bash
+      run: |
+        cmake -S Source/Windows/UnixLib -B build_unixlib_${{ inputs.target }} -DCMAKE_BUILD_TYPE=$BUILD_TYPE \
+          -G Ninja -DCMAKE_INSTALL_LIBDIR=/usr/lib/wine/aarch64-unix -DCMAKE_INSTALL_PREFIX=/usr
+
+    - name: Build UnixLib
+      shell: bash
+      run: cmake --build build_unixlib_${{ inputs.target }}
+
+    - name: Install UnixLib
+      shell: bash
+      run: DESTDIR="$PWD"/install cmake --build build_unixlib_${{ inputs.target }} -t install
@@ -0,0 +1,57 @@
+name: Wine DLL artifacts
+
+on:
+  push:
+    branches:
+      - main
+
+env:
+  BUILD_TYPE: Release
+
+jobs:
+  wine_dll_artifacts:
+    runs-on: ${{ matrix.arch }}
+    strategy:
+      matrix:
+        arch: [[self-hosted, ARM64, mingw]]
+      fail-fast: false
+
+    steps:
+    - uses: actions/checkout@v6
+      with:
+        fetch-depth: '0'
+        fetch-tags: 'true'
+
+    - name: Add MingGW to PATH
+      run: echo "$HOME/llvm-mingw/build/bin/" >> $GITHUB_PATH
+
+    - name: Checkout Submodules
+      # Need to update submodules
+      run: |
+        git submodule sync --recursive
+        git submodule update --init --depth 1
+
+    - name: Clean install directory
+      run: rm -Rf install
+
+    - name: Build (wow64)
+      uses: ./.github/workflows/wine_build
+      with:
+        target: wow64
+
+    - name: Build (arm64ec)
+      uses: ./.github/workflows/wine_build
+      with:
+        target: arm64ec
+
+    - name: Upload libraries
+      uses: actions/upload-artifact@v6
+      timeout-minutes: 1
+      with:
+        overwrite: true
+        name: wine_dll_artifacts
+        path: |
+          ${{ github.workspace }}/install/usr/lib/wine/aarch64-windows/lib*.dll
+          ${{ github.workspace }}/install/usr/lib/wine/aarch64-unix/lib*.so
+        retention-days: 60
+        compression-level: 9
@@ -11,3 +11,5 @@ out/
 .vs/
 *.pyc
 .cache
+.idea/
+CMakeLists.txt.user
@@ -0,0 +1,71 @@
+spec:
+  inputs:
+    PROMOTE_BRANCH:
+      description: "Branch to promote the build to. Empty means no promotion."
+      default: "bleeding-edge"
+
+---
+
+workflow:
+  rules:
+    - when: always
+      variables:
+        PROMOTE_BRANCH: $[[ inputs.PROMOTE_BRANCH ]]
+
+variables:
+    DEBIAN_FRONTEND: noninteractive
+    GIT_SUBMODULE_STRATEGY: recursive
+    GIT_DEPTH: 0
+    CC: clang
+    CXX: clang++
+
+build:
+  stage: build
+  image: registry.gitlab.steamos.cloud/steamrt/steamrt4/sdk/arm64:4.0.20251117.183306
+  tags:
+    - docker
+    - linux
+    - arm64
+    - aarch64
+  script:
+    - apt-get -y update
+    - apt-get install -y
+        git cmake ninja-build ccache
+        lld clang clang-tools
+        libclang-dev llvm-dev
+        libstdc++-14-dev-i386-cross libgcc-14-dev-i386-cross
+        libstdc++-14-dev-amd64-cross libgcc-14-dev-amd64-cross
+    - cmake -E make_directory build/
+    - cmake -DCMAKE_BUILD_TYPE=Release -G Ninja -DBUILD_STEAM_SUPPORT=True -DENABLE_LTO=True -DENABLE_ASSERTIONS=False -DBUILD_THUNKS=True -DBUILD_FEXCONFIG=False -DBUILD_TESTING=False -DENABLE_CLANG_THUNKS=True -DUSE_LINKER=lld -DCMAKE_INSTALL_PREFIX=/usr -DTUNE_ARCH=armv8.2-a -DTUNE_CPU=none -DRANGES_NATIVE=OFF . -B build/
+    - cmake --build build/ --config Release
+    - DESTDIR=$(pwd)/install/ cmake --build build/ --config Release -t install
+
+  artifacts:
+    name: "steamrt artifacts"
+    untracked: false
+    paths:
+      - install/
+
+promote:
+  stage: deploy
+  variables:
+    GIT_STRATEGY: none
+  image: registry.gitlab.steamos.cloud/steamrt/steamrt4/sdk/arm64:4.0.20251117.183306
+  tags:
+    - docker
+    - linux
+    - arm64
+    - aarch64
+  rules:
+    - if: '$PROMOTE_BRANCH'
+  before_script:
+    - apt-get -y update
+    - apt-get install -y tmux curl
+  script:
+    # comment out to debug: SSH in via GCP, go down the container and attach to the session (with `tmux attach -t debug`)
+#    - tmux new-session -d -s debug
+#    - while tmux has-session -t debug 2>/dev/null; do sleep 1; done
+
+    # ref controls which fex-depot code runs the pipeline, while VERSION_PARAM controls which fex branch's artifacts that pipeline downloads.
+    - >
+      curl --fail --location --request POST --form token=${FEX_DEPOT_TRIGGER_TOKEN} --form ref=master --form "variables[PROMOTE_BRANCH]=${PROMOTE_BRANCH}" --form "variables[VERSION_PARAM]=${CI_COMMIT_REF_NAME}" "${CI_API_V4_URL}/projects/fex%2Ffex-depot/trigger/pipeline"
@@ -5,10 +5,6 @@
 [submodule "External/cpp-optparse"]
 	path = Source/Common/cpp-optparse
 	url = https://github.com/Sonicadvance1/cpp-optparse
-[submodule "External/xbyak"]
-  shallow = true
-	path = External/xbyak
-	url = https://github.com/herumi/xbyak.git
 [submodule "External/fex-posixtest-bins"]
  shallow = true
 	path = External/fex-posixtest-bins
@@ -21,9 +17,6 @@
  shallow = true
 	path = External/fex-gcc-target-tests-bins
 	url = https://github.com/FEX-Emu/fex-gcc-target-tests-bins.git
-[submodule "External/jemalloc"]
-	path = External/jemalloc
-	url = https://github.com/FEX-Emu/jemalloc.git
 [submodule "External/fmt"]
 	path = External/fmt
 	url = https://github.com/fmtlib/fmt.git
@@ -36,10 +29,6 @@
 [submodule "External/Catch2"]
 	path = External/Catch2
 	url = https://github.com/catchorg/Catch2.git
-[submodule "External/robin-map"]
-	shallow = true
-	path = External/robin-map
-	url = https://github.com/FEX-Emu/robin-map.git
 [submodule "External/Vulkan-Headers"]
 	shallow = true
 	path = External/Vulkan-Headers
@@ -47,3 +36,19 @@
 [submodule "External/jemalloc_glibc"]
 	path = External/jemalloc_glibc
 	url = https://github.com/FEX-Emu/jemalloc.git
+[submodule "External/tracy"]
+	path = External/tracy
+	url = https://github.com/wolfpld/tracy
+[submodule "External/range-v3"]
+	path = External/range-v3
+	url = https://github.com/ericniebler/range-v3.git
+[submodule "External/zydis"]
+	shallow = true
+	path = External/zydis
+	url = https://github.com/zyantific/zydis.git
+[submodule "External/unordered_dense"]
+	path = External/unordered_dense
+	url = https://github.com/martinus/unordered_dense.git
+[submodule "External/rpmalloc"]
+	path = External/rpmalloc
+	url = https://github.com/FEX-Emu/rpmalloc.git
@@ -0,0 +1 @@
+AI must not be used to generate code for contributions to this project.
@@ -0,0 +1 @@
+AI must not be used to generate code for contributions to this project.
@@ -1,66 +1,147 @@
 cmake_minimum_required(VERSION 3.14)
 project(FEX C CXX ASM)

-INCLUDE (CheckIncludeFiles)
-CHECK_INCLUDE_FILES ("gdb/jit-reader.h" HAVE_GDB_JIT_READER_H)
+include(CheckIncludeFiles)
+check_include_files("gdb/jit-reader.h" HAVE_GDB_JIT_READER_H)

-option(BUILD_TESTS "Build unit tests to ensure sanity" TRUE)
-option(BUILD_FEX_LINUX_TESTS "Build FEXLinuxTests, requires x86 compiler" FALSE)
+option(BUILD_FEX_LINUX_TESTS "Build FEXLinuxTests (requires x86 compiler)" FALSE)
 option(BUILD_THUNKS "Build thunks" FALSE)
 option(BUILD_FEXCONFIG "Build FEXConfig" TRUE)
-option(ENABLE_CLANG_THUNKS "Build thunks with clang" FALSE)
-option(ENABLE_IWYU "Enables include what you use program" FALSE)
+option(ENABLE_CLANG_THUNKS "Build thunks with clang" TRUE)
+option(ENABLE_IWYU "Enable the Include What You Use sanitizer" FALSE)
 option(ENABLE_LTO "Enable LTO with compilation" TRUE)
 option(ENABLE_XRAY "Enable building with LLVM X-Ray" FALSE)
-set(USE_LINKER "" CACHE STRING "Allow overriding the linker path directly")
-option(ENABLE_ASAN "Enables Clang ASAN" FALSE)
-option(ENABLE_TSAN "Enables Clang TSAN" FALSE)
-option(ENABLE_COVERAGE "Enables Coverage" FALSE)
-option(ENABLE_ASSERTIONS "Enables assertions in build" FALSE)
-option(ENABLE_GDB_SYMBOLS "Enables GDBSymbols integration support" ${HAVE_GDB_JIT_READER_H})
-option(ENABLE_STRICT_WERROR "Enables stricter -Werror for CI" FALSE)
-option(ENABLE_WERROR "Enables -Werror" FALSE)
-option(ENABLE_JEMALLOC "Enables jemalloc allocator" TRUE)
-option(ENABLE_JEMALLOC_GLIBC_ALLOC "Enables jemalloc glibc allocator" TRUE)
-option(ENABLE_OFFLINE_TELEMETRY "Enables FEX offline telemetry" TRUE)
-option(ENABLE_COMPILE_TIME_TRACE "Enables time trace compile option" FALSE)
-option(ENABLE_LIBCXX "Enables LLVM libc++" FALSE)
-option(ENABLE_CCACHE "Enables ccache for compile caching" TRUE)
-option(ENABLE_VIXL_SIMULATOR "Forces the FEX JIT to use the VIXL simulator" FALSE)
-option(ENABLE_VIXL_DISASSEMBLER "Enables debug disassembler output with VIXL" FALSE)
-option(USE_LEGACY_BINFMTMISC "Uses legacy method of setting up binfmt_misc" FALSE)
-option(COMPILE_VIXL_DISASSEMBLER "Compiles the vixl disassembler in to vixl" FALSE)
-option(ENABLE_FEXCORE_PROFILER "Enables use of the FEXCore timeline profiling capabilities" FALSE)
-set (FEXCORE_PROFILER_BACKEND "gpuvis" CACHE STRING "Set which backend you want to use for the FEXCore profiler")
+set(USE_LINKER "" CACHE STRING "Path to a custom linker program")
+option(ENABLE_UBSAN "Enable the Clang Undefined Behavior Sanitizer" FALSE)
+option(ENABLE_ASAN "Enable the Clang Address Sanitizer" FALSE)
+option(ENABLE_TSAN "Enable the Clang Thread Sanitizer" FALSE)
+option(ENABLE_COVERAGE "Enable Code Coverage" FALSE)
+option(ENABLE_ASSERTIONS "Enable debug assertions" FALSE)
+option(ENABLE_GDB_SYMBOLS "Enable GDBSymbols integration support" ${HAVE_GDB_JIT_READER_H})
+option(ENABLE_STRICT_WERROR "Enable stricter -Werror" FALSE)
+option(ENABLE_WERROR "Enable -Werror" FALSE)
+option(ENABLE_FEX_ALLOCATOR "Enable allocator for FEX" TRUE)
+option(ENABLE_JEMALLOC_GLIBC_ALLOC "Enable jemalloc glibc allocator" TRUE)
+option(ENABLE_OFFLINE_TELEMETRY "Enable FEX offline telemetry" TRUE)
+option(ENABLE_COMPILE_TIME_TRACE "Enable time trace compile option" FALSE)
+option(ENABLE_LIBCXX "Use LLVM's libc++ instead of the GNU libstdc++" FALSE)
+option(ENABLE_CCACHE "Enable ccache for build caching" TRUE)
+option(ENABLE_VIXL_SIMULATOR "Use the VIXL simulator for emulation (only useful for CI testing)" FALSE)
+option(ENABLE_VIXL_DISASSEMBLER "Enable debug disassembler output with VIXL" FALSE)
+option(ENABLE_ZYDIS "Enable x86/x86-64 guest disassembler output with Zydis" FALSE)
+option(USE_LEGACY_BINFMTMISC "Use legacy method of setting up binfmt_misc" FALSE)
+option(ENABLE_FEXCORE_PROFILER "Enable FEXCore's timeline profiling capabilities" FALSE)
+set(FEXCORE_PROFILER_BACKEND "gpuvis" CACHE STRING "Set which backend to use for FEXCore's profiler")
+set_property(CACHE FEXCORE_PROFILER_BACKEND PROPERTY STRINGS gpuvis tracy)
 option(ENABLE_GLIBC_ALLOCATOR_HOOK_FAULT "Enables glibc memory allocation hooking with fault for CI testing")
-option(USE_PDB_DEBUGINFO "Builds debug info in PDB format" FALSE)
+option(USE_PDB_DEBUGINFO "Build debug info in PDB format" FALSE)
+option(BUILD_STEAM_SUPPORT "Enable Steam integration" FALSE)

-set (X86_32_TOOLCHAIN_FILE "${CMAKE_CURRENT_SOURCE_DIR}/toolchain_x86_32.cmake" CACHE FILEPATH "Toolchain file for the (cross-)compiler targeting i686")
-set (X86_64_TOOLCHAIN_FILE "${CMAKE_CURRENT_SOURCE_DIR}/toolchain_x86_64.cmake" CACHE FILEPATH "Toolchain file for the (cross-)compiler targeting x86_64")
-set (DATA_DIRECTORY "${CMAKE_INSTALL_PREFIX}/share/fex-emu" CACHE PATH "global data directory")
+set(X86_32_TOOLCHAIN_FILE "${CMAKE_CURRENT_SOURCE_DIR}/Data/CMake/toolchain_x86_32.cmake" CACHE FILEPATH "Toolchain file for the (cross-)compiler targeting i686")
+set(X86_64_TOOLCHAIN_FILE "${CMAKE_CURRENT_SOURCE_DIR}/Data/CMake/toolchain_x86_64.cmake" CACHE FILEPATH "Toolchain file for the (cross-)compiler targeting x86_64")
+set(X86_DEV_ROOTFS "/" CACHE FILEPATH "Path to the sysroot used for cross-compiling for i686 and x86_64")
+set(DATA_DIRECTORY "" CACHE PATH "Global data directory (override)")
+set(HOSTLIBS_DATA_DIRECTORY "" CACHE PATH "Global data directory (override)")

-string(FIND ${CMAKE_BASE_NAME} mingw CONTAINS_MINGW)
-if (NOT CONTAINS_MINGW EQUAL -1)
-  message (STATUS "Mingw build")
-  set (MINGW_BUILD TRUE)
-  set (ENABLE_JEMALLOC TRUE)
-  set (ENABLE_JEMALLOC_GLIBC_ALLOC FALSE)
+if (NOT DATA_DIRECTORY)
+  set(DATA_DIRECTORY "${CMAKE_INSTALL_PREFIX}/share/fex-emu")
 endif()

-if (NOT MINGW_BUILD)
-  message (STATUS "Clang version ${CMAKE_CXX_COMPILER_VERSION}")
-  set (CLANG_MINIMUM_VERSION 13.0)
+include(GNUInstallDirs)
+if (NOT HOSTLIBS_DATA_DIRECTORY)
+  set(HOSTLIBS_DATA_DIRECTORY "${CMAKE_INSTALL_FULL_LIBDIR}/fex-emu")
+endif()
+
+## Platform Checks ##
+# Only 64-bit Linux and Windows are supported
+
+# NB: SIZEOF_VOID_P is in bytes, not bits
+# On 32-bit systems this is set to 4
+if (NOT CMAKE_SIZEOF_VOID_P EQUAL 8)
+  message(FATAL_ERROR "Unsupported pointer size ${CMAKE_SIZEOF_VOID_P}."
+    " FEX only supports 64-bit (8-byte pointer) systems."
+    " If you believe this is in error, file an issue.")
+elseif (NOT (WIN32 OR CMAKE_SYSTEM_NAME STREQUAL "Linux"))
+  message(FATAL_ERROR "Unsupported system type ${CMAKE_SYSTEM_NAME}."
+    " FEX only supports Linux and Windows."
+    " If you believe this is in error, file an issue.")
+endif()
+
+## Compiler Checks ##
+# GCC and MSVC are unsupported
+if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+  message(FATAL_ERROR "FEX doesn't support GCC! Use Clang instead.")
+elseif (MSVC)
+  message(FATAL_ERROR "FEX doesn't support MSVC! Use Clang on MinGW instead.")
+elseif (MINGW)
+  message(STATUS "Building for MinGW")
+  set(ENABLE_FEX_ALLOCATOR TRUE)
+  set(ENABLE_JEMALLOC_GLIBC_ALLOC FALSE)
+else ()
+  message(STATUS "Clang version ${CMAKE_CXX_COMPILER_VERSION}")
+  set(CLANG_MINIMUM_VERSION 13.0)
  if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS ${CLANG_MINIMUM_VERSION})
-    message (FATAL_ERROR "Clang version too old for FEX. Need at least ${CLANG_MINIMUM_VERSION} but has ${CMAKE_CXX_COMPILER_VERSION}")
+    message(FATAL_ERROR "Clang version too old for FEX. Need at least ${CLANG_MINIMUM_VERSION} but has ${CMAKE_CXX_COMPILER_VERSION}")
  endif()
 endif()

+## Architecture Handling ##
+string(TOLOWER ${CMAKE_SYSTEM_PROCESSOR} processor)
+if (processor MATCHES "x86|amd64")
+  option(ENABLE_X86_HOST_DEBUG "Enables compiling on x86_64 host" FALSE)
+  if (NOT ENABLE_X86_HOST_DEBUG)
+    message(FATAL_ERROR
+      " FEX doesn't support compiling for x86-64 hosts!"
+      " This is /only/ a supported configuration for FEX CI and nothing else!")
+  else()
+    message(STATUS "x86_64 debug build")
+  endif()
+
+  set(ARCHITECTURE_x86_64 1)
+  add_compile_definitions(ARCHITECTURE_x86_64=1)
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcx16")
+elseif (processor MATCHES "^aarch64|^arm64|^armv8\.*")
+  set(ARCHITECTURE_arm64 1)
+  add_compile_definitions(ARCHITECTURE_arm64=1)
+
+  # arm64ec needs to define both arm64 and arm64ec
+  if (processor MATCHES "^arm64ec")
+    set(ARCHITECTURE_arm64ec 1)
+    add_compile_definitions(ARCHITECTURE_arm64ec=1)
+  endif()
+endif()
+
+if (NOT (ARCHITECTURE_arm64 OR ARCHITECTURE_arm64ec OR ARCHITECTURE_x86_64))
+  message(FATAL_ERROR "Unsupported processor type ${processor}."
+    " If you believe this is in error, file an issue.")
+endif()
+
+if (BUILD_STEAM_SUPPORT)
+  add_compile_definitions(FEX_STEAM_SUPPORT=1)
+endif()
+
 if (ENABLE_FEXCORE_PROFILER)
-  add_definitions(-DENABLE_FEXCORE_PROFILER=1)
+  add_compile_definitions(ENABLE_FEXCORE_PROFILER=1)
  string(TOUPPER "${FEXCORE_PROFILER_BACKEND}" FEXCORE_PROFILER_BACKEND)

  if (FEXCORE_PROFILER_BACKEND STREQUAL "GPUVIS")
-    add_definitions(-DFEXCORE_PROFILER_BACKEND=1)
+    add_compile_definitions(FEXCORE_PROFILER_BACKEND=1)
+  elseif (FEXCORE_PROFILER_BACKEND STREQUAL "TRACY")
+    add_compile_definitions(FEXCORE_PROFILER_BACKEND=2)
+    add_compile_definitions(TRACY_ENABLE=1)
+    # Required so that Tracy will only start in the selected guest application
+    add_compile_definitions(TRACY_MANUAL_LIFETIME=1)
+    add_compile_definitions(TRACY_DELAYED_INIT=1)
+    # This interferes with FEX's signal handling
+    add_compile_definitions(TRACY_NO_CRASH_HANDLER=1)
+    # Tracy can gather call stack samples in regular intervals, but this
+    # isn't useful for us since it would usually sample opaque JIT code
+    add_compile_definitions(TRACY_NO_SAMPLING=1)
+    # This pulls in libbacktrace which allocators in global constructors (before FEX can set up its allocator hooks)
+    add_compile_definitions(TRACY_NO_CALLSTACK=1)
+    if (MINGW)
+      message(FATAL_ERROR "Tracy profiler not supported on MinGW")
+    endif()
  else()
    message(FATAL_ERROR "Unknown FEXCore profiler backend ${FEXCORE_PROFILER_BACKEND}")
  endif()
@@ -71,13 +152,13 @@ if (ENABLE_JEMALLOC_GLIBC_ALLOC AND ENABLE_GLIBC_ALLOCATOR_HOOK_FAULT)
 endif()

 if (ENABLE_GLIBC_ALLOCATOR_HOOK_FAULT)
-  add_definitions(-DGLIBC_ALLOCATOR_FAULT=1)
+  add_compile_definitions(GLIBC_ALLOCATOR_FAULT=1)
 endif()

 # uninstall target
 if(NOT TARGET uninstall)
  configure_file(
-    "${CMAKE_CURRENT_SOURCE_DIR}/CMakeFiles/cmake_uninstall.cmake.in"
+    "${CMAKE_CURRENT_SOURCE_DIR}/Data/CMake/cmake_uninstall.cmake.in"
    "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/cmake_uninstall.cmake"
    IMMEDIATE @ONLY)

@@ -86,9 +167,17 @@ if(NOT TARGET uninstall)
 endif()

 # These options are meant for package management
-set (TUNE_CPU "native" CACHE STRING "Override the CPU the build is tuned for")
-set (TUNE_ARCH "generic" CACHE STRING "Override the Arch the build is tuned for")
-set (OVERRIDE_VERSION "detect" CACHE STRING "Override the FEX version in the format of <MMYY>{.<REV>}")
+set(TUNE_CPU "native" CACHE STRING "Override the CPU the build is tuned for")
+set(TUNE_ARCH "generic" CACHE STRING "Override the Arch the build is tuned for")
+set(OVERRIDE_VERSION "detect" CACHE STRING "Override the FEX version")
+set(OVERRIDE_HASH "detect" CACHE STRING "Override the FEX git hash")
+
+get_property(IS_MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
+if (NOT IS_MULTI_CONFIG AND NOT CMAKE_BUILD_TYPE)
+    set(CMAKE_BUILD_TYPE Release
+        CACHE STRING "Choose the type of build." FORCE)
+    message(STATUS "No build type set, defaulting to a Release build")
+endif()

 string(TOUPPER "${CMAKE_BUILD_TYPE}" CMAKE_BUILD_TYPE)
 if (CMAKE_BUILD_TYPE MATCHES "DEBUG")
@@ -97,14 +186,18 @@ endif()

 if (ENABLE_ASSERTIONS)
  message(STATUS "Assertions enabled")
-  add_definitions(-DASSERTIONS_ENABLED=1)
+  add_compile_definitions(ASSERTIONS_ENABLED=1)
 endif()

 if (ENABLE_GDB_SYMBOLS)
  message(STATUS "GDBSymbols support enabled")
-  add_definitions(-DGDB_SYMBOLS_ENABLED=1)
+  add_compile_definitions(GDB_SYMBOLS_ENABLED=1)
 endif()

+add_compile_definitions(_LARGEFILE64_SOURCE)
+if (WIN32)
+  add_compile_definitions(UNICODE _UNICODE)
+endif()

 set(CMAKE_CXX_STANDARD 20)
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
@@ -115,33 +208,7 @@ cmake_policy(SET CMP0083 NEW) # Follow new PIE policy
 include(CheckPIESupported)
 check_pie_supported()

-if (ENABLE_LTO)
-  set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE)
-else()
-  set(CMAKE_INTERPROCEDURAL_OPTIMIZATION FALSE)
-endif()
-
-if (CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64")
-  option(ENABLE_X86_HOST_DEBUG "Enables compiling on x86_64 host" FALSE)
-  if (NOT ENABLE_X86_HOST_DEBUG)
-    message(FATAL_ERROR
-    " FEX-Emu doesn't support compiling for x86-64 hosts!"
-    " This is /only/ a supported configuration for FEX CI and nothing else!")
-  endif()
-  set(_M_X86_64 1)
-  add_definitions(-D_M_X86_64=1)
-  set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcx16")
-endif()
-
-if (CMAKE_SYSTEM_PROCESSOR MATCHES "^aarch64|^arm64|^armv8\.*")
-  set(_M_ARM_64 1)
-  add_definitions(-D_M_ARM_64=1)
-endif()
-
-if (CMAKE_SYSTEM_PROCESSOR MATCHES "^arm64ec")
-  set(_M_ARM_64EC 1)
-  add_definitions(-D_M_ARM_64EC=1)
-endif()
+set(CMAKE_INTERPROCEDURAL_OPTIMIZATION ${ENABLE_LTO})

 include(CheckCXXSourceCompiles)
 set(CMAKE_REQUIRED_FLAGS "-std=c++11 -Wattributes -Werror=attributes")
@@ -157,30 +224,47 @@ check_cxx_source_compiles(
  HAS_CLANG_PRESERVE_ALL)
 unset(CMAKE_REQUIRED_FLAGS)
 if (HAS_CLANG_PRESERVE_ALL)
-  if (MINGW_BUILD)
+  if (MINGW)
    message(STATUS "Ignoring broken clang::preserve_all support")
    set(HAS_CLANG_PRESERVE_ALL FALSE)
  else()
    message(STATUS "Has clang::preserve_all")
  endif()
-endif ()
-
-if (_M_ARM_64 AND HAS_CLANG_PRESERVE_ALL)
-  add_definitions("-DFEX_PRESERVE_ALL_ATTR=__attribute__((preserve_all))" "-DFEX_HAS_PRESERVE_ALL_ATTR=1")
-else()
-  add_definitions("-DFEX_PRESERVE_ALL_ATTR=" "-DFEX_HAS_PRESERVE_ALL_ATTR=0")
 endif()

+if (ARCHITECTURE_arm64 AND HAS_CLANG_PRESERVE_ALL)
+  add_compile_definitions("FEX_PRESERVE_ALL_ATTR=__attribute__((preserve_all))" "FEX_HAS_PRESERVE_ALL_ATTR=1")
+else()
+  add_compile_definitions("FEX_PRESERVE_ALL_ATTR=" "FEX_HAS_PRESERVE_ALL_ATTR=0")
+endif()
+
+check_cxx_source_compiles(
+  "
+  #define _GNU_SOURCE
+  #include <errno.h>
+  int main() {
+  return program_invocation_name == nullptr;
+  }"
+  HAS_PROGRAM_INVOCATION_NAME)
+add_compile_definitions("HAS_PROGRAM_INVOCATION_NAME=${HAS_PROGRAM_INVOCATION_NAME}")
+
 if (ENABLE_VIXL_SIMULATOR)
  # We can run the simulator on both x86-64 or AArch64 hosts
-  add_definitions(-DVIXL_SIMULATOR=1 -DVIXL_INCLUDE_SIMULATOR_AARCH64=1)
+  add_compile_definitions(VIXL_SIMULATOR=1 VIXL_INCLUDE_SIMULATOR_AARCH64=1)
 endif()

 if (ENABLE_CCACHE)
  find_program(CCACHE_PROGRAM ccache)
  if(CCACHE_PROGRAM)
-    message(STATUS "CCache enabled")
-    set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "${CCACHE_PROGRAM}")
+    execute_process(COMMAND "${CCACHE_PROGRAM}" --print-version
+     OUTPUT_VARIABLE CCACHE_VERSION OUTPUT_STRIP_TRAILING_WHITESPACE)
+    message(STATUS "Enabling ccache ${CCACHE_VERSION}")
+    if (CCACHE_VERSION VERSION_GREATER_EQUAL "4.8")
+      # Set sloppiness to enable caching even for files that use __DATE__/__TIME__ macros
+      set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "${CCACHE_PROGRAM} sloppiness=time_macros")
+    else()
+      set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE "${CCACHE_PROGRAM}")
+    endif()
  endif()
 endif()

@@ -194,7 +278,7 @@ if (ENABLE_COMPILE_TIME_TRACE)
  link_libraries(-ftime-trace)
 endif()

-set (PTHREAD_LIB pthread)
+set(PTHREAD_LIB pthread)

 if (USE_LINKER)
  message(STATUS "Overriding linker to: ${USE_LINKER}")
@@ -209,11 +293,24 @@ endif()

 if (NOT ENABLE_OFFLINE_TELEMETRY)
  # Disable FEX offline telemetry entirely if asked
-  add_definitions(-DFEX_DISABLE_TELEMETRY=1)
+  add_compile_definitions(FEX_DISABLE_TELEMETRY=1)
+endif()
+
+if (ENABLE_UBSAN)
+  # See https://github.com/FEX-Emu/FEX/pull/4494#issuecomment-2800608944
+  # and related discussion for the use of -fno-sanitize=alignment -fno-sanitize=function
+  # with UBSAN.
+  # alignment: we don't follow a strict alignment policy, for example IR uses packed structs
+  # that are regularly access unaligned.
+  # function: syscalls cast function pointers to void (*)(unsigned long...), causing warnings
+  # related to this access.
+  add_compile_definitions(ENABLE_UBSAN=1)
+  add_compile_options(-fno-omit-frame-pointer -fsanitize=undefined -fno-sanitize=alignment -fno-sanitize=function -fno-sanitize-recover=undefined)
+  link_libraries(-fno-omit-frame-pointer -fsanitize=undefined -fno-sanitize=alignment -fno-sanitize=function -fno-sanitize-recover=undefined)
 endif()

 if (ENABLE_ASAN)
-  add_definitions(-DENABLE_ASAN=1)
+  add_compile_definitions(ENABLE_ASAN=1)
  add_compile_options(-fno-omit-frame-pointer -fsanitize=address -fsanitize-address-use-after-scope)
  link_libraries(-fno-omit-frame-pointer -fsanitize=address -fsanitize-address-use-after-scope)
 endif()
@@ -233,20 +330,20 @@ if (ENABLE_JEMALLOC_GLIBC_ALLOC)
  # Required for thunks to work.
  # All host native libraries will use this allocator, while *most* other FEX internal allocations will use the other jemalloc allocator.
  add_subdirectory(External/jemalloc_glibc/)
-elseif (NOT MINGW_BUILD)
-  message (STATUS
+elseif (NOT MINGW)
+  message(STATUS
    " jemalloc glibc allocator disabled!\n"
    " This is not a recommended configuration!\n"
    " This will very explicitly break thunk execution!\n"
    " Use at your own risk!")
 endif()

-if (ENABLE_JEMALLOC)
-  # The jemalloc subproject that all FEXCore fextl objects allocate through.
-  add_subdirectory(External/jemalloc/)
-elseif (NOT MINGW_BUILD)
+if (ENABLE_FEX_ALLOCATOR)
+  # The rpmalloc subproject that all FEXCore fextl objects allocate through.
+  add_subdirectory(External/rpmalloc/)
+elseif (NOT MINGW)
  message (STATUS
-    " jemalloc disabled!\n"
+    " FEX allocator is disabled!\n"
    " This is not a recommended configuration!\n"
    " This will very explicitly break 32-bit application execution!\n"
    " Use at your own risk!")
@@ -257,48 +354,67 @@ if (USE_PDB_DEBUGINFO)
  add_link_options(-g -Wl,--pdb=)
 endif()

-set (CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} -fno-omit-frame-pointer")
-set (CMAKE_LINKER_FLAGS_RELWITHDEBINFO "${CMAKE_LINKER_FLAGS_RELWITHDEBINFO} -fno-omit-frame-pointer")
+set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELWITHDEBINFO} -fno-omit-frame-pointer")
+set(CMAKE_LINKER_FLAGS_RELWITHDEBINFO "${CMAKE_LINKER_FLAGS_RELWITHDEBINFO} -fno-omit-frame-pointer")

-set (CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fomit-frame-pointer")
-set (CMAKE_LINKER_FLAGS_RELEASE "${CMAKE_LINKER_FLAGS_RELEASE} -fomit-frame-pointer")
+set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fomit-frame-pointer")
+set(CMAKE_LINKER_FLAGS_RELEASE "${CMAKE_LINKER_FLAGS_RELEASE} -fomit-frame-pointer")

-include_directories(External/robin-map/include/)
+## Modules ##
+list(APPEND CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/Data/CMake/)

-if (BUILD_TESTS)
-  # Enable vixl disassembler if tests are enabled.
-  set(COMPILE_VIXL_DISASSEMBLER TRUE)
+include(LinkerGC)
+
+## Externals ##
+
+find_package(unordered_dense QUIET CONFIG)
+if (NOT unordered_dense_FOUND)
+  add_subdirectory(External/unordered_dense)
 endif()

-if (COMPILE_VIXL_DISASSEMBLER OR ENABLE_VIXL_SIMULATOR)
+include(CTest)
+if (BUILD_TESTING OR ENABLE_VIXL_DISASSEMBLER OR ENABLE_VIXL_SIMULATOR)
  add_subdirectory(External/vixl/)
-  include_directories(SYSTEM External/vixl/src/)
 endif()

-if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-  # This means we were attempted to get compiled with GCC
-  message(FATAL_ERROR "FEX doesn't support getting compiled with GCC!")
+if (ENABLE_ZYDIS)
+  find_package(Zycore 1.5 MODULE QUIET)
+  find_package(Zydis 4.0 MODULE QUIET)
+
+  if (TARGET Zydis::Zydis AND TARGET Zycore::Zycore)
+    message(STATUS "Using system Zydis")
+  else()
+    set(ZYDIS_BUILD_TOOLS OFF CACHE BOOL "" FORCE)
+    set(ZYDIS_BUILD_EXAMPLES OFF CACHE BOOL "" FORCE)
+
+    message(STATUS "Using bundled Zydis")
+    add_subdirectory(External/zydis/)
+  endif()
 endif()

-find_package(PkgConfig REQUIRED)
-find_package(Python 3.0 REQUIRED COMPONENTS Interpreter)
+if (ENABLE_FEXCORE_PROFILER AND FEXCORE_PROFILER_BACKEND STREQUAL "TRACY")
+  add_subdirectory(External/tracy)
+endif()
+
+find_package(Python 3.9 REQUIRED COMPONENTS Interpreter)

 set(BUILD_SHARED_LIBS OFF)

-pkg_search_module(xxhash IMPORTED_TARGET xxhash libxxhash)
-if (TARGET PkgConfig::xxhash AND NOT CMAKE_CROSSCOMPILING)
-  add_library(xxHash::xxhash ALIAS PkgConfig::xxhash)
-else()
+if (NOT CMAKE_CROSSCOMPILING)
+  find_package(xxhash MODULE QUIET)
+endif()
+
+if (NOT TARGET xxHash::xxhash)
  set(XXHASH_BUNDLED_MODE TRUE)
  set(XXHASH_BUILD_XXHSUM FALSE)
  add_subdirectory(External/xxhash/cmake_unofficial/)
 endif()

-add_definitions(-Wno-trigraphs)
-add_definitions(-DGLOBAL_DATA_DIRECTORY="${DATA_DIRECTORY}/")
+add_compile_options(-Wno-trigraphs)
+add_compile_definitions(GLOBAL_DATA_DIRECTORY="${DATA_DIRECTORY}/")

-if (BUILD_TESTS)
-  find_package(Catch2 QUIET)
+if (BUILD_TESTING)
+  find_package(Catch2 3 QUIET)
  if (NOT Catch2_FOUND)
    add_subdirectory(External/Catch2/)

@@ -307,17 +423,30 @@ if (BUILD_TESTS)
  endif()

  include(Catch)
+else ()
+  # Override any previously generated test list to avoid running stale test binaries
+  file(GENERATE OUTPUT CTestTestfile.cmake CONTENT "# No tests since BUILD_TESTING is disabled")
 endif()

-find_package(fmt QUIET)
-if (NOT fmt_FOUND)
-  # Disable fmt install
+if (MINGW)
  set(FMT_INSTALL OFF)
  add_subdirectory(External/fmt/)
+else()
+  find_package(fmt QUIET)
+  if (NOT fmt_FOUND)
+    # Disable fmt install
+    set(FMT_INSTALL OFF)
+    add_subdirectory(External/fmt/)
+  endif()
+endif()
+
+find_package(range-v3 QUIET)
+if (NOT range-v3_FOUND)
+  add_subdirectory(External/range-v3/)
+  target_compile_definitions(range-v3 INTERFACE RANGES_DISABLE_DEPRECATED_WARNINGS)
 endif()

 add_subdirectory(External/tiny-json/)
-include_directories(External/tiny-json/)

 include_directories(Source/)
 include_directories("${CMAKE_BINARY_DIR}/Source/")
@@ -341,6 +470,11 @@ if(ENUM_ENUM_WARNING)
  add_compile_options(-Wno-deprecated-enum-enum-conversion)
 endif()

+# GCC enables -Wchanges-meaning by default and treats some cases as an error
+if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+  add_compile_options(-Wno-error=changes-meaning)
+endif()
+
 if(ENABLE_WERROR OR ENABLE_STRICT_WERROR)
  add_compile_options(-Werror)
  if (NOT ENABLE_STRICT_WERROR)
@@ -351,16 +485,24 @@ endif()

 set(FEX_TUNE_COMPILE_FLAGS)
 if (NOT TUNE_ARCH STREQUAL "generic")
-  check_cxx_compiler_flag("-march=${TUNE_ARCH}" COMPILER_SUPPORTS_ARCH_TYPE)
-  if(COMPILER_SUPPORTS_ARCH_TYPE)
-    list(APPEND FEX_TUNE_COMPILE_FLAGS "-march=${TUNE_ARCH}")
-  else()
-    message(FATAL_ERROR "Trying to compile arch type '${TUNE_ARCH}' but the compiler doesn't support this")
+  set(TUNE_ARCH_STRING "${TUNE_ARCH}")
+  if(ARCHITECTURE_arm64)
+    set(TUNE_ARCH_STRING "${TUNE_ARCH}+crc")
  endif()
+  check_cxx_compiler_flag("-march=${TUNE_ARCH_STRING}" COMPILER_SUPPORTS_ARCH_TYPE)
+  if(COMPILER_SUPPORTS_ARCH_TYPE)
+    list(APPEND FEX_TUNE_COMPILE_FLAGS "-march=${TUNE_ARCH_STRING}")
+  else()
+    message(FATAL_ERROR "Trying to compile arch type '${TUNE_ARCH_STRING}' but the compiler doesn't support this")
+  endif()
+elseif(ARCHITECTURE_arm64)
+  # Need to always append crc
+  check_cxx_compiler_flag("-march=armv8-a+crc" COMPILER_SUPPORTS_ARCH_TYPE)
+  list(APPEND FEX_TUNE_COMPILE_FLAGS "-march=armv8-a+crc")
 endif()

 if (TUNE_CPU STREQUAL "native")
-  if(_M_ARM_64)
+  if(ARCHITECTURE_arm64)
    if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 999999.0)
      # Clang 12.0 fixed the -mcpu=native bug with mixed big.little implementers
      # Clang can not currently check for native Apple M1 type in hypervisor. Currently disabled
@@ -369,15 +511,18 @@ if (TUNE_CPU STREQUAL "native")
        list(APPEND FEX_TUNE_COMPILE_FLAGS "-mcpu=native")
      endif()
    else()
-      # Due to an oversight in llvm, it declares any reasonably new Kryo CPU to only be ARMv8.0
-      # Manually detect newer CPU revisions until clang and llvm fixes their bug
-      # This script will either provide a supported CPU or 'native'
-      # Additionally -march doesn't work under AArch64+Clang, so you have to use -mcpu or -mtune
      execute_process(COMMAND python3 "${PROJECT_SOURCE_DIR}/Scripts/aarch64_fit_native.py" "/proc/cpuinfo" "${CMAKE_CXX_COMPILER_VERSION}"
        OUTPUT_VARIABLE AARCH64_CPU)

      string(STRIP ${AARCH64_CPU} AARCH64_CPU)

+      execute_process(COMMAND python3 "${PROJECT_SOURCE_DIR}/Scripts/NeedDisabledSVE.py"
+        RESULT_VARIABLE NEEDS_SVE_DISABLED)
+      if (NEEDS_SVE_DISABLED)
+        message(STATUS "Platform has bugged SVE. Disabling")
+        set(AARCH64_CPU "cortex-a78")
+      endif()
+
      check_cxx_compiler_flag("-mcpu=${AARCH64_CPU}" COMPILER_SUPPORTS_CPU_TYPE)
      if(COMPILER_SUPPORTS_CPU_TYPE)
        list(APPEND FEX_TUNE_COMPILE_FLAGS "-mcpu=${AARCH64_CPU}")
@@ -389,7 +534,7 @@ if (TUNE_CPU STREQUAL "native")
      list(APPEND FEX_TUNE_COMPILE_FLAGS "-march=native")
    endif()
  endif()
-else()
+elseif (NOT TUNE_CPU STREQUAL "none")
  check_cxx_compiler_flag("-mcpu=${TUNE_CPU}" COMPILER_SUPPORTS_CPU_TYPE)
  if(COMPILER_SUPPORTS_CPU_TYPE)
    list(APPEND FEX_TUNE_COMPILE_FLAGS "-mcpu=${TUNE_CPU}")
@@ -398,8 +543,54 @@ else()
  endif()
 endif()

+set(GIT_DESCRIBE_STRING "FEX-Unknown")
+
+if (OVERRIDE_VERSION STREQUAL "detect")
+  find_package(Git)
+
+  if (GIT_FOUND)
+    execute_process(
+      COMMAND ${GIT_EXECUTABLE} describe --abbrev=7
+      WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
+      OUTPUT_VARIABLE GIT_DESCRIBE_STRING
+      ERROR_QUIET
+      OUTPUT_STRIP_TRAILING_WHITESPACE)
+  endif()
+else()
+  set(GIT_DESCRIBE_STRING "${OVERRIDE_VERSION}")
+endif()
+
+set(GIT_HASH "Unknown")
+
+if (OVERRIDE_HASH STREQUAL "detect")
+  find_package(Git)
+
+  if (GIT_FOUND)
+    execute_process(
+      COMMAND ${GIT_EXECUTABLE} rev-parse HEAD
+      WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
+      OUTPUT_VARIABLE GIT_HASH
+      ERROR_QUIET
+      OUTPUT_STRIP_TRAILING_WHITESPACE)
+  endif()
+else()
+  set(GIT_HASH "${OVERRIDE_HASH}")
+endif()
+
+message(STATUS "FEX version: ${GIT_DESCRIBE_STRING}")
+message(STATUS "FEX commit: ${GIT_HASH}")
+
+# Prepends 0x to every two-character sequence in the hash,
+# OR the final character of the hash, to plumb it for C++ usage. e.g.:
+# -DOVERRIDE_HASH=123456aa => 0x12, 0x34, 0x56, 0xaa,
+# -DOVERRIDE_HASH=12345678a => 0x12, 0x34, 0x56, 0x78, 0xa,
+string(REGEX
+  REPLACE "(..|.$)" "0x\\1, "
+  GIT_HASH_ARRAY "${GIT_HASH}")
+
 if (ENABLE_IWYU)
-  find_program(IWYU_EXE "iwyu")
+  find_program(IWYU_EXE
+    NAMES iwyu include-what-you-use)
  if (IWYU_EXE)
    message(STATUS "IWYU enabled")
    set(CMAKE_CXX_INCLUDE_WHAT_YOU_USE "${IWYU_EXE}")
@@ -408,19 +599,10 @@ endif()

 add_compile_options(-Wall)

-configure_file(
-  ${CMAKE_CURRENT_SOURCE_DIR}/include/Config.h.in
-  ${CMAKE_BINARY_DIR}/generated/ConfigDefines.h)
-
-include(CTest)
-if (BUILD_TESTS)
+if (BUILD_TESTING)
  message(STATUS "Unit tests are enabled")
-  if (NOT BUILD_TESTING)
-    # CMake checks this variable before generating CTestTestfile.cmake
-    message(SEND_ERROR "Unit tests require BUILD_TESTING to be enabled")
-  endif()

-  set (TEST_JOB_COUNT "" CACHE STRING "Override number of parallel jobs to use while running tests")
+  set(TEST_JOB_COUNT "" CACHE STRING "Override number of parallel jobs to use while running tests")
  if (TEST_JOB_COUNT)
    message(STATUS "Running tests with ${TEST_JOB_COUNT} jobs")
  elseif(CMAKE_VERSION VERSION_LESS "3.29")
@@ -429,17 +611,22 @@ if (BUILD_TESTS)
  set(TEST_JOB_FLAG "-j${TEST_JOB_COUNT}")
 endif()

+add_subdirectory(External/SoftFloat-3e/)
+add_subdirectory(External/cephes/)
 add_subdirectory(FEXHeaderUtils/)
 add_subdirectory(CodeEmitter/)
 add_subdirectory(FEXCore/)

-if (_M_ARM_64 AND NOT MINGW_BUILD)
+if (ARCHITECTURE_arm64 AND NOT MINGW AND NOT BUILD_STEAM_SUPPORT)
  # Binfmt_misc files must be installed prior to Source/ installs
  add_subdirectory(Data/binfmts/)
 endif()

 add_subdirectory(Source/)
-add_subdirectory(Data/AppConfig/)
+
+if (NOT BUILD_STEAM_SUPPORT)
+  add_subdirectory(Data/AppConfig/)
+endif()

 # Install the ThunksDB file
 file(GLOB CONFIG_SOURCES CONFIGURE_DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/Data/*.json)
@@ -447,15 +634,16 @@ file(GLOB CONFIG_SOURCES CONFIGURE_DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/Data/*.js
 # Any application configuration json file gets installed
 foreach(CONFIG_SRC ${CONFIG_SOURCES})
  install(FILES ${CONFIG_SRC}
-    DESTINATION ${DATA_DIRECTORY}/)
+    DESTINATION ${DATA_DIRECTORY}/
+    COMPONENT Runtime)
 endforeach()

-if (BUILD_TESTS)
+if (BUILD_TESTING)
  add_subdirectory(unittests/)
 endif()

 if (BUILD_THUNKS)
-  set (FEX_PROJECT_SOURCE_DIR ${PROJECT_SOURCE_DIR})
+  set(FEX_PROJECT_SOURCE_DIR ${PROJECT_SOURCE_DIR})
  add_subdirectory(ThunkLibs/Generator)

  # Thunk targets for both host libraries and IDE integration
@@ -479,10 +667,10 @@ if (BUILD_THUNKS)
      "-DCMAKE_INSTALL_PREFIX=${CMAKE_INSTALL_PREFIX}"
      "-DFEX_PROJECT_SOURCE_DIR=${FEX_PROJECT_SOURCE_DIR}"
      "-DGENERATOR_EXE=$<TARGET_FILE:thunkgen>"
+      "-DX86_DEV_ROOTFS=${X86_DEV_ROOTFS}"
    INSTALL_COMMAND ""
    BUILD_ALWAYS ON
-    DEPENDS thunkgen
-  )
+    DEPENDS thunkgen)

  ExternalProject_Add(guest-libs-32
    PREFIX guest-libs-32
@@ -497,108 +685,39 @@ if (BUILD_THUNKS)
      "-DCMAKE_INSTALL_PREFIX=${CMAKE_INSTALL_PREFIX}"
      "-DFEX_PROJECT_SOURCE_DIR=${FEX_PROJECT_SOURCE_DIR}"
      "-DGENERATOR_EXE=$<TARGET_FILE:thunkgen>"
+      "-DX86_DEV_ROOTFS=${X86_DEV_ROOTFS}"
    INSTALL_COMMAND ""
    BUILD_ALWAYS ON
-    DEPENDS thunkgen
-  )
+    DEPENDS thunkgen)

  install(
-    CODE "MESSAGE(\"-- Installing: guest-libs\")"
+    CODE "message(\"-- Installing: guest-libs\")"
    CODE "
-    EXECUTE_PROCESS(COMMAND ${CMAKE_COMMAND} --build . --target install
-    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest
-    )"
+      execute_process(COMMAND ${CMAKE_COMMAND} --build . --target install
+        WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest)"
    DEPENDS guest-libs
-  )
+    COMPONENT Runtime)

  install(
-    CODE "MESSAGE(\"-- Installing: guest-libs-32\")"
+    CODE "message(\"-- Installing: guest-libs-32\")"
    CODE "
-    EXECUTE_PROCESS(COMMAND ${CMAKE_COMMAND} --build . --target install
-    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest_32
-    )"
+      execute_process(COMMAND ${CMAKE_COMMAND} --build . --target install
+        WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest_32)"
    DEPENDS guest-libs-32
-  )
+    COMPONENT Runtime)

  add_custom_target(uninstall_guest-libs
    COMMAND ${CMAKE_COMMAND} "--build" "." "--target" "uninstall"
-    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest
-  )
+    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest)

  add_custom_target(uninstall_guest-libs-32
    COMMAND ${CMAKE_COMMAND} "--build" "." "--target" "uninstall"
-    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest_32
-  )
+    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest_32)

  add_dependencies(uninstall uninstall_guest-libs)
  add_dependencies(uninstall uninstall_guest-libs-32)
 endif()

-set(FEX_VERSION_MAJOR "0")
-set(FEX_VERSION_MINOR "0")
-set(FEX_VERSION_PATCH "0")
-
-if (OVERRIDE_VERSION STREQUAL "detect")
-  find_package(Git)
-  if (GIT_FOUND)
-    execute_process(
-      COMMAND ${GIT_EXECUTABLE} describe --abbrev=0
-      WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
-      OUTPUT_VARIABLE GIT_DESCRIBE_STRING
-      RESULT_VARIABLE GIT_ERROR
-      ERROR_QUIET
-      OUTPUT_STRIP_TRAILING_WHITESPACE
-    )
-
-    if (NOT ${GIT_ERROR} EQUAL 0)
-      # Likely built in a way that doesn't have tags
-      # Setup a version tag that is unknown
-      set(GIT_DESCRIBE_STRING "FEX-0000")
-    endif()
-  endif()
-else()
-  set(GIT_DESCRIBE_STRING "FEX-${OVERRIDE_VERSION}")
+if (NOT MINGW AND BUILD_STEAM_SUPPORT)
+  add_subdirectory(Source/Steam/)
 endif()
-
-# Parse the version here
-# Change something like `FEX-2106.1-76-<hash>` in to a list
-string(REPLACE "-" ";" DESCRIBE_LIST ${GIT_DESCRIBE_STRING})
-
-# Extract the `2106.1` element
-list(GET DESCRIBE_LIST 1 DESCRIBE_LIST)
-
-# Change `2106.1` in to a list
-string(REPLACE "." ";" DESCRIBE_LIST ${DESCRIBE_LIST})
-
-# Calculate list size
-list(LENGTH DESCRIBE_LIST LIST_SIZE)
-
-# Pull out the major version
-list(GET DESCRIBE_LIST 0 FEX_VERSION_MAJOR)
-
-# Minor version only exists if there is a .1 at the end
-# eg: 2106 versus 2106.1
-if (LIST_SIZE GREATER 1)
-  list(GET DESCRIBE_LIST 1 FEX_VERSION_MINOR)
-endif()
-
-# Package creation
-set (CPACK_GENERATOR "DEB")
-set (CPACK_PACKAGE_NAME fex-emu)
-set (CPACK_PACKAGE_FILE_NAME "${CPACK_PACKAGE_NAME}-${GIT_DESCRIBE_STRING}_${CMAKE_SYSTEM_PROCESSOR}")
-set (CPACK_PACKAGE_CONTACT "FEX-Emu Maintainers <team@fex-emu.com>")
-set (CPACK_PACKAGE_VERSION_MAJOR "${FEX_VERSION_MAJOR}")
-set (CPACK_PACKAGE_VERSION_MINOR "${FEX_VERSION_MINOR}")
-set (CPACK_PACKAGE_VERSION_PATCH "${FEX_VERSION_PATCH}")
-set (CPACK_PACKAGE_DESCRIPTION_FILE "${CMAKE_CURRENT_SOURCE_DIR}/CPack/Description.txt")
-
-# Debian defines
-set (CPACK_DEBIAN_PACKAGE_DEPENDS "libc6, libstdc++6, libepoxy0, libsdl2-2.0-0, libegl1, libx11-6, squashfuse")
-set (CPACK_DEBIAN_PACKAGE_CONTROL_EXTRA
-  "${CMAKE_CURRENT_SOURCE_DIR}/CPack/postinst;${CMAKE_CURRENT_SOURCE_DIR}/CPack/prerm;${CMAKE_CURRENT_SOURCE_DIR}/CPack/triggers")
-if (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
-  # binfmt_misc conflicts with qemu-user-static
-  # We also only install binfmt_misc on aarch64 hosts
-  set (CPACK_DEBIAN_PACKAGE_CONFLICTS "${CPACK_DEBIAN_PACKAGE_CONFLICTS}, qemu-user-static")
-endif()
-include (CPack)
@@ -1,132 +0,0 @@
-{
-	"environments": [
-		{
-			"BuildPath": "${projectDir}\\out\\build\\${name}",
-			"InstallPath": "${projectDir}\\out\\install\\${name}",
-			"clangcl": "clang-cl.exe",
-			"cc": "clang",
-			"cxx": "clang++"
-		}
-	],
-	"configurations": [
-		{
-			"name": "WSL-Clang-Debug",
-			"generator": "Ninja",
-			"configurationType": "Debug",
-			"buildRoot": "${env.BuildPath}",
-			"installRoot": "${env.InstallPath}",
-			"cmakeExecutable": "/usr/bin/cmake",
-			"cmakeCommandArgs": "",
-			"buildCommandArgs": "-v",
-			"ctestCommandArgs": "",
-			"wslPath": "${defaultWSLPath}",
-			"inheritEnvironments": [ "linux_clang_x64" ],
-			"addressSanitizerRuntimeFlags": "detect_leaks=0",
-			"variables": [
-				{
-					"name": "WSL",
-					"value": "TRUE",
-					"type": "BOOL"
-				}
-			]
-		},
-		{
-			"name": "WSL-Clang-Release",
-			"generator": "Ninja",
-			"configurationType": "RelWithDebInfo",
-			"buildRoot": "${env.BuildPath}",
-			"installRoot": "${env.InstallPath}",
-			"cmakeExecutable": "/usr/bin/cmake",
-			"cmakeCommandArgs": "",
-			"buildCommandArgs": "-v",
-			"ctestCommandArgs": "",
-			"wslPath": "${defaultWSLPath}",
-			"inheritEnvironments": [ "linux_clang_x64" ],
-			"addressSanitizerRuntimeFlags": "detect_leaks=0",
-			"variables": [
-				{
-					"name": "WSL",
-					"value": "TRUE",
-					"type": "BOOL"
-				}
-			]
-		},
-		{
-			"name": "x86-Clang-Cross-Debug",
-			"generator": "Ninja",
-			"configurationType": "Debug",
-			"buildRoot": "${env.BuildPath}",
-			"installRoot": "${env.InstallPath}",
-			"cmakeCommandArgs": "",
-			"buildCommandArgs": "-v",
-			"ctestCommandArgs": "",
-			"inheritEnvironments": [ "clang_cl_x86" ],
-			"variables": [
-				{
-					"name": "CMAKE_C_COMPILER",
-					"value": "${env.cc}",
-					"type": "STRING"
-				},
-				{
-					"name": "CMAKE_CXX_COMPILER",
-					"value": "${env.cxx}",
-					"type": "STRING"
-				},
-				{
-					"name": "CMAKE_SYSROOT",
-					"value": "${env.fexsysroot}",
-					"type": "STRING"
-				}
-			]
-		},
-		{
-			"name": "x64-Clang-Cross-Release",
-			"generator": "Ninja",
-			"configurationType": "RelWithDebInfo",
-			"buildRoot": "${env.BuildPath}",
-			"installRoot": "${env.InstallPath}",
-			"cmakeCommandArgs": "",
-			"buildCommandArgs": "-v",
-			"ctestCommandArgs": "",
-			"inheritEnvironments": [ "clang_cl_x86" ],
-			"variables": [
-				{
-					"name": "CMAKE_C_COMPILER",
-					"value": "${env.cc}",
-					"type": "STRING"
-				},
-				{
-					"name": "CMAKE_CXX_COMPILER",
-					"value": "${env.cxx}",
-					"type": "STRING"
-				},
-				{
-					"name": "CMAKE_SYSROOT",
-					"value": "${env.fexsysroot}",
-					"type": "STRING"
-				}
-			]
-		},
-		{
-			"name": "Linux-Clang-Remote-Debug",
-			"generator": "Ninja",
-			"configurationType": "Debug",
-			"cmakeExecutable": "/usr/bin/cmake",
-			"remoteCopySourcesExclusionList": [ ".vs", ".vscode", ".git", ".github", "build", "out", "bin" ],
-			"cmakeCommandArgs": "",
-			"buildCommandArgs": "-v",
-			"ctestCommandArgs": "",
-			"inheritEnvironments": [ "linux_clang_x64" ],
-			"remoteMachineName": "${env.fexremote}",
-			"remoteCMakeListsRoot": "$HOME/projects/.vs/${projectDirName}/src",
-			"remoteBuildRoot": "$HOME/projects/.vs/${projectDirName}/build/${name}",
-			"remoteInstallRoot": "$HOME/projects/.vs/${projectDirName}/install/${name}",
-			"remoteCopySources": true,
-			"rsyncCommandArgs": "-t --delete --delete-excluded",
-			"remoteCopyBuildOutput": false,
-			"remoteCopySourcesMethod": "rsync",
-			"addressSanitizerRuntimeFlags": "detect_leaks=0",
-			"variables": []
-		}
-	]
-}
@@ -0,0 +1 @@
+No AI/ML/LLM/etc code contributions.
@@ -1,3 +0,0 @@
-x86 and x86-64 Linux emulator
-
-FEX is very much work in progress, so expect things to change.
@@ -1,18 +0,0 @@
-#!/bin/sh
-set -e
-update_binfmt() {
-  # Check for update-binfmts
-  command -v update-binfmts >/dev/null || return 0
-
-  # Setup binfmt_misc
-  update-binfmts --import FEX-x86
-  update-binfmts --import FEX-x86_64
-}
-
-# Install FEXInterpreter hardlink
-# Needs to be done before setting up binfmt_misc
-ln -f /usr/bin/FEXLoader /usr/bin/FEXInterpreter
-
-if [ $(uname -m) = 'aarch64' ]; then
-  update_binfmt
-fi
@@ -1,17 +0,0 @@
-#!/bin/sh
-set -e
-update_binfmt() {
-  # Check for update-binfmts
-  command -v update-binfmts >/dev/null || return 0
-
-  # Uninstall
-  update-binfmts --unimport FEX-x86
-  update-binfmts --unimport FEX-x86_64
-}
-
-if [ $(uname -m) = 'aarch64' ]; then
-  update_binfmt
-fi
-
-# Remove FEXInterpreter hardlink
-unlink /usr/bin/FEXInterpreter
@@ -1 +0,0 @@
-activate-noawait ldconfig
@@ -11,6 +11,14 @@
 * FEX-Emu ALU operations usually have a 32-bit or 64-bit operating size encoded in the IR operation,
 * This allows FEX to use a single helper function which decodes to both handlers.
 */
+
+#pragma once
+#ifndef INCLUDED_BY_EMITTER
+#include <CodeEmitter/Emitter.h>
+namespace ARMEmitter {
+struct EmitterOps : Emitter {
+#endif
+
 private:
  static bool IsADRRange(int64_t Imm) {
    return Imm >= -1048576 && Imm <= 1048575;
@@ -28,27 +36,31 @@ public:
    DataProcessing_PCRel_Imm(Op, rd, Imm);
  }

-  void adr(ARMEmitter::Register rd, BackwardLabel const* Label) {
+  [[nodiscard]] BranchEncodeSucceeded adr(ARMEmitter::Register rd, const BackwardLabel* Label) {
    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-    LOGMAN_THROW_A_FMT(IsADRRange(Imm), "Unscaled offset too large");
+    if (IsADRRange(Imm)) {
+      constexpr uint32_t Op = 0b0001'0000 << 24;
+      DataProcessing_PCRel_Imm(Op, rd, Imm);
+      return BranchEncodeSucceeded::Success;
+    }

-    constexpr uint32_t Op = 0b0001'0000 << 24;
-    DataProcessing_PCRel_Imm(Op, rd, Imm);
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
  }
-  template<typename LabelType>
-  requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-  void adr(ARMEmitter::Register rd, LabelType *Label) {
-    AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::ADR });
+  [[nodiscard]] BranchEncodeSucceeded adr(ARMEmitter::Register rd, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::ADR});
    constexpr uint32_t Op = 0b0001'0000 << 24;
    DataProcessing_PCRel_Imm(Op, rd, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
  }

-  void adr(ARMEmitter::Register rd, BiDirectionalLabel *Label) {
+  [[nodiscard]] BranchEncodeSucceeded adr(ARMEmitter::Register rd, BiDirectionalLabel* Label) {
    if (Label->Backward.Location) {
-      adr(rd, &Label->Backward);
-    }
-    else {
-      adr(rd, &Label->Forward);
+      return adr(rd, &Label->Backward);
+    } else {
+      return adr(rd, &Label->Forward);
    }
  }

@@ -57,43 +69,53 @@ public:
    DataProcessing_PCRel_Imm(Op, rd, Imm);
  }

-  void adrp(ARMEmitter::Register rd, BackwardLabel const* Label) {
+  [[nodiscard]] BranchEncodeSucceeded adrp(ARMEmitter::Register rd, const BackwardLabel* Label) {
    int64_t Imm = reinterpret_cast<int64_t>(Label->Location) - (GetCursorAddress<int64_t>() & ~0xFFFLL);
-    LOGMAN_THROW_A_FMT(IsADRPRange(Imm) && IsADRPAligned(Imm), "Unscaled offset too large");

-    constexpr uint32_t Op = 0b1001'0000 << 24;
-    DataProcessing_PCRel_Imm(Op, rd, Imm);
+    if (IsADRPRange(Imm) && IsADRPAligned(Imm)) {
+      constexpr uint32_t Op = 0b1001'0000 << 24;
+      DataProcessing_PCRel_Imm(Op, rd, Imm);
+      return BranchEncodeSucceeded::Success;
+    }
+
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
  }
-  template<typename LabelType>
-  requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-  void adrp(ARMEmitter::Register rd, LabelType *Label) {
-    AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::ADRP });
+
+  [[nodiscard]] BranchEncodeSucceeded adrp(ARMEmitter::Register rd, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::ADRP});
    constexpr uint32_t Op = 0b1001'0000 << 24;
    DataProcessing_PCRel_Imm(Op, rd, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
  }

-  void adrp(ARMEmitter::Register rd, BiDirectionalLabel *Label) {
+  [[nodiscard]] BranchEncodeSucceeded adrp(ARMEmitter::Register rd, BiDirectionalLabel* Label) {
    if (Label->Backward.Location) {
-      adrp(rd, &Label->Backward);
-    }
-    else {
-      adrp(rd, &Label->Forward);
+      return adrp(rd, &Label->Backward);
+    } else {
+      return adrp(rd, &Label->Forward);
    }
  }

-  void LongAddressGen(ARMEmitter::Register rd, BackwardLabel const* Label) {
-    int64_t Imm = reinterpret_cast<int64_t>(Label->Location) - (GetCursorAddress<int64_t>());
+  [[nodiscard]] BranchEncodeSucceeded LongAddressGen(ARMEmitter::Register rd, const BackwardLabel* Label) {
+    const auto SLocation = reinterpret_cast<int64_t>(Label->Location);
+    const auto ULocation = std::bit_cast<uint64_t>(SLocation);
+
+    const int64_t Imm = SLocation - (GetCursorAddress<int64_t>());
+    const auto UImm = std::bit_cast<uint64_t>(Imm);
+
    if (IsADRRange(Imm)) {
      // If the range is in ADR range then we can just use ADR.
-      adr(rd, Label);
+      return adr(rd, Label);
    }
-    else if (IsADRPRange(Imm)) {
-      int64_t ADRPImm = (reinterpret_cast<int64_t>(Label->Location) & ~0xFFFLL)
-        - (GetCursorAddress<int64_t>() & ~0xFFFLL);
+    if (IsADRPRange(Imm)) {
+      const int64_t ADRPImm = (SLocation & ~0xFFFLL) - (GetCursorAddress<int64_t>() & ~0xFFFLL);

      // If the range is in the ADRP range then we can use ADRP.
-      bool NeedsOffset = !IsADRPAligned(reinterpret_cast<uint64_t>(Label->Location));
-      uint64_t AlignedOffset = reinterpret_cast<uint64_t>(Label->Location) & 0xFFFULL;
+      const bool NeedsOffset = !IsADRPAligned(ULocation);
+      const uint64_t AlignedOffset = ULocation & 0xFFFULL;

      // First emit ADRP
      adrp(rd, ADRPImm >> 12);
@@ -102,25 +124,33 @@ public:
        // Now even an add
        add(ARMEmitter::Size::i64Bit, rd, rd, AlignedOffset);
      }
+
+      return BranchEncodeSucceeded::Success;
    }
-    else {
-      LOGMAN_MSG_A_FMT("Unscaled offset too large");
-      FEX_UNREACHABLE;
-    }
+
+    // Stinky path, we need to load the address as a sequence of movz+movk+movk
+    movz(ARMEmitter::Size::i64Bit, rd, (UImm >> 32) & 0xFFFF, 32);
+    movk(ARMEmitter::Size::i64Bit, rd, (UImm >> 16) & 0xFFFF, 16);
+    movk(ARMEmitter::Size::i64Bit, rd, UImm & 0xFFFF);
+
+    return BranchEncodeSucceeded::Success;
  }
-  void LongAddressGen(ARMEmitter::Register rd, ForwardLabel* Label) {
-    Label->Insts.emplace_back(SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::LONG_ADDRESS_GEN });
-    // Emit a register index and a nop. These will be backpatched.
+  [[nodiscard]] BranchEncodeSucceeded LongAddressGen(ARMEmitter::Register rd, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::LONG_ADDRESS_GEN});
+    // Emit a register index and two nops. These will be backpatched.
    dc32(rd.Idx());
    nop();
+    nop();
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
  }

-  void LongAddressGen(ARMEmitter::Register rd, BiDirectionalLabel *Label) {
+  [[nodiscard]] BranchEncodeSucceeded LongAddressGen(ARMEmitter::Register rd, BiDirectionalLabel* Label) {
    if (Label->Backward.Location) {
-      LongAddressGen(rd, &Label->Backward);
-    }
-    else {
-      LongAddressGen(rd, &Label->Forward);
+      return LongAddressGen(rd, &Label->Backward);
+    } else {
+      return LongAddressGen(rd, &Label->Forward);
    }
  }

@@ -176,11 +206,7 @@ public:
  // Logical immediate
  void and_(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint64_t Imm) {
    uint32_t n, immr, imms;
-    [[maybe_unused]] const auto IsImm = IsImmLogical(Imm,
-                           RegSizeInBits(s),
-                           &n,
-                           &imms,
-                           &immr);
+    const auto IsImm = IsImmLogical(Imm, RegSizeInBits(s), &n, &imms, &immr);
    LOGMAN_THROW_A_FMT(IsImm, "Couldn't encode immediate to logical op");
    and_(s, rd, rn, n, immr, imms);
  }
@@ -191,11 +217,7 @@ public:

  void ands(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint64_t Imm) {
    uint32_t n, immr, imms;
-    [[maybe_unused]] const auto IsImm = IsImmLogical(Imm,
-                           RegSizeInBits(s),
-                           &n,
-                           &imms,
-                           &immr);
+    const auto IsImm = IsImmLogical(Imm, RegSizeInBits(s), &n, &imms, &immr);
    LOGMAN_THROW_A_FMT(IsImm, "Couldn't encode immediate to logical op");
    ands(s, rd, rn, n, immr, imms);
  }
@@ -206,22 +228,14 @@ public:

  void orr(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint64_t Imm) {
    uint32_t n, immr, imms;
-    [[maybe_unused]] const auto IsImm = IsImmLogical(Imm,
-                           RegSizeInBits(s),
-                           &n,
-                           &imms,
-                           &immr);
+    const auto IsImm = IsImmLogical(Imm, RegSizeInBits(s), &n, &imms, &immr);
    LOGMAN_THROW_A_FMT(IsImm, "Couldn't encode immediate to logical op");
    orr(s, rd, rn, n, immr, imms);
  }

  void eor(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint64_t Imm) {
    uint32_t n, immr, imms;
-    [[maybe_unused]] const auto IsImm = IsImmLogical(Imm,
-                           RegSizeInBits(s),
-                           &n,
-                           &imms,
-                           &immr);
+    const auto IsImm = IsImmLogical(Imm, RegSizeInBits(s), &n, &imms, &immr);
    LOGMAN_THROW_A_FMT(IsImm, "Couldn't encode immediate to logical op");
    eor(s, rd, rn, n, immr, imms);
  }
@@ -351,12 +365,12 @@ public:
    bfi(s, rd, Reg::zr, lsb, width);
  }
  void bfxil(ARMEmitter::Size s, Register rd, Register rn, uint32_t lsb, uint32_t width) {
-    [[maybe_unused]] const auto reg_size_bits = RegSizeInBits(s);
+    const auto reg_size_bits = RegSizeInBits(s);
    const auto lsb_p_width = lsb + width;

    LOGMAN_THROW_A_FMT(width >= 1, "bfxil needs width >= 1");
-    LOGMAN_THROW_A_FMT(lsb_p_width <= reg_size_bits, "bfxil lsb + width ({}) must be <= {}. lsb={}, width={}",
-                       lsb_p_width, reg_size_bits, lsb, width);
+    LOGMAN_THROW_A_FMT(lsb_p_width <= reg_size_bits, "bfxil lsb + width ({}) must be <= {}. lsb={}, width={}", lsb_p_width, reg_size_bits,
+                       lsb, width);

    bfm(s, rd, rn, lsb, lsb_p_width - 1);
  }
@@ -375,188 +389,142 @@ public:

  // Data processing - 2 source
  void udiv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0000'10U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0000'10U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void sdiv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0000'11U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0000'11U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }

  void lslv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0010'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0010'00U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void lsrv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0010'01U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0010'01U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void asrv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0010'10U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0010'10U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void rorv(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0010'11U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0010'11U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void crc32b(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0100'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0100'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void crc32h(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0100'01U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0100'01U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void crc32w(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0100'10U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0100'10U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void crc32cb(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0101'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0101'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void crc32ch(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0101'01U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0101'01U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void crc32cw(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0101'10U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0101'10U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i32Bit, rd, rn, rm);
  }
  void smax(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0110'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0110'00U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void umax(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0110'01U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0110'01U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void smin(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0110'10U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0110'10U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void umin(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0110'11U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0110'11U << 10);
    DataProcessing_2Source(Op, s, rd, rn, rm);
  }
  void subp(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0000'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0000'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void irg(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0001'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0001'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void gmi(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0001'01U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0001'01U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void pacga(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0011'00U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0011'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void crc32x(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0100'11U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0100'11U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void crc32cx(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b001'1010'110U << 21) |
-                        (0b0101'11U << 10);
+    constexpr uint32_t Op = (0b001'1010'110U << 21) | (0b0101'11U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }
  void subps(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm) {
-    constexpr uint32_t Op = (0b011'1010'110U << 21) |
-                        (0b0000'00U << 10);
+    constexpr uint32_t Op = (0b011'1010'110U << 21) | (0b0000'00U << 10);
    DataProcessing_2Source(Op, ARMEmitter::Size::i64Bit, rd, rn, rm);
  }

  // Data processing - 1 source
  void rbit(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'00U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'00U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void rev16(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'01U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'01U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void rev(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'10U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'10U << 10);
    DataProcessing_1Source(Op, ARMEmitter::Size::i32Bit, rd, rn);
  }
  void rev32(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'10U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'10U << 10);
    DataProcessing_1Source(Op, ARMEmitter::Size::i64Bit, rd, rn);
  }
  void clz(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0001'00U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0001'00U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void cls(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0001'01U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0001'01U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void rev(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'11U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'11U << 10);
    DataProcessing_1Source(Op, ARMEmitter::Size::i64Bit, rd, rn);
  }
  void rev(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0000'10U << 10) |
-                        (s == ARMEmitter::Size::i64Bit ? (1U << 10) : 0);
+    uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0000'10U << 10) | (s == ARMEmitter::Size::i64Bit ? (1U << 10) : 0);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void ctz(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0001'10U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0001'10U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void cnt(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0001'11U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0001'11U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }
  void abs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn) {
-    constexpr uint32_t Op = (0b101'1010'110U << 21) |
-                        (0b0'0000U << 16) |
-                        (0b0010'00U << 10);
+    constexpr uint32_t Op = (0b101'1010'110U << 21) | (0b0'0000U << 16) | (0b0010'00U << 10);
    DataProcessing_1Source(Op, s, rd, rn);
  }

@@ -573,27 +541,33 @@ public:
    orr(ARMEmitter::Size::i32Bit, rd.R(), ARMEmitter::Reg::zr, rn.R(), ARMEmitter::ShiftType::LSL, 0);
  }

-  void mvn(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void mvn(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL,
+           uint32_t amt = 0) {
    orn(s, rd, ARMEmitter::Reg::zr, rn, Shift, amt);
  }

-  void and_(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void and_(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b000'1010'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void ands(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void ands(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b110'1010'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void bic(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void bic(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b000'1010'001U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void bics(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void bics(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b110'1010'001U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void orr(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void orr(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b010'1010'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
@@ -601,30 +575,36 @@ public:
    ands(s, Reg::zr, rn, rm, shift, amt);
  }

-  void orn(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void orn(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b010'1010'001U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void eor(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void eor(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b100'1010'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void eon(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void eon(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    constexpr uint32_t Op = 0b100'1010'001U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }

  // AddSub - shifted register
-  void add(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void add(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    add(ARMEmitter::Size::i64Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
-  void adds(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void adds(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    adds(ARMEmitter::Size::i64Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void cmn(ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    adds(ARMEmitter::Size::i64Bit, ARMEmitter::XReg::zr, rn.R(), rm.R(), Shift, amt);
  }
-  void sub(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void sub(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    sub(ARMEmitter::Size::i64Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void neg(ARMEmitter::XRegister rd, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
@@ -633,23 +613,27 @@ public:
  void cmp(ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(ARMEmitter::Size::i64Bit, ARMEmitter::Reg::rsp, rn.R(), rm.R(), Shift, amt);
  }
-  void subs(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void subs(ARMEmitter::XRegister rd, ARMEmitter::XRegister rn, ARMEmitter::XRegister rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(ARMEmitter::Size::i64Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void negs(ARMEmitter::XRegister rd, ARMEmitter::XRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(rd, ARMEmitter::XReg::zr, rm, Shift, amt);
  }

-  void add(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void add(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    add(ARMEmitter::Size::i32Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
-  void adds(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void adds(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    adds(ARMEmitter::Size::i32Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void cmn(ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    adds(ARMEmitter::Size::i32Bit, ARMEmitter::WReg::zr, rn.R(), rm.R(), Shift, amt);
  }
-  void sub(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void sub(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    sub(ARMEmitter::Size::i32Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void neg(ARMEmitter::WRegister rd, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
@@ -658,65 +642,78 @@ public:
  void cmp(ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(ARMEmitter::Size::i32Bit, ARMEmitter::Reg::rsp, rn.R(), rm.R(), Shift, amt);
  }
-  void subs(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void subs(ARMEmitter::WRegister rd, ARMEmitter::WRegister rn, ARMEmitter::WRegister rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(ARMEmitter::Size::i32Bit, rd.R(), rn.R(), rm.R(), Shift, amt);
  }
  void negs(ARMEmitter::WRegister rd, ARMEmitter::WRegister rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
    subs(rd, ARMEmitter::WReg::zr, rm, Shift, amt);
  }

-  void add(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
-    LOGMAN_THROW_AA_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
+  void add(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+    LOGMAN_THROW_A_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
    constexpr uint32_t Op = 0b000'1011'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void adds(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
-    LOGMAN_THROW_AA_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
+  void adds(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+    LOGMAN_THROW_A_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
    constexpr uint32_t Op = 0b010'1011'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void cmn(ARMEmitter::Size s, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void cmn(ARMEmitter::Size s, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL,
+           uint32_t amt = 0) {
    adds(s, ARMEmitter::Reg::zr, rn, rm, Shift, amt);
  }
-  void sub(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
-    LOGMAN_THROW_AA_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
+  void sub(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+           ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+    LOGMAN_THROW_A_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
    constexpr uint32_t Op = 0b100'1011'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void neg(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void neg(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL,
+           uint32_t amt = 0) {
    sub(s, rd, ARMEmitter::Reg::zr, rm, Shift, amt);
  }
-  void cmp(ARMEmitter::Size s, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void cmp(ARMEmitter::Size s, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL,
+           uint32_t amt = 0) {
    subs(s, ARMEmitter::Reg::zr, rn, rm, Shift, amt);
  }

-  void subs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
-    LOGMAN_THROW_AA_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
+  void subs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+            ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+    LOGMAN_THROW_A_FMT(Shift != ARMEmitter::ShiftType::ROR, "Doesn't support ROR");
    constexpr uint32_t Op = 0b110'1011'000U << 21;
    DataProcessing_Shifted_Reg(Op, s, rd, rn, rm, Shift, amt);
  }
-  void negs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL, uint32_t amt = 0) {
+  void negs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift = ARMEmitter::ShiftType::LSL,
+            uint32_t amt = 0) {
    subs(s, rd, ARMEmitter::Reg::zr, rm, Shift, amt);
  }

  // AddSub - extended register
-  void add(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift = 0) {
-    LOGMAN_THROW_AA_FMT(Shift <= 4, "Shift amount is too large");
+  void add(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option,
+           uint32_t Shift = 0) {
+    LOGMAN_THROW_A_FMT(Shift <= 4, "Shift amount is too large");
    constexpr uint32_t Op = 0b000'1011'001U << 21;
    DataProcessing_Extended_Reg(Op, s, rd, rn, rm, Option, Shift);
  }
-  void adds(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift = 0) {
+  void adds(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option,
+            uint32_t Shift = 0) {
    constexpr uint32_t Op = 0b010'1011'001U << 21;
    DataProcessing_Extended_Reg(Op, s, rd, rn, rm, Option, Shift);
  }
  void cmn(ARMEmitter::Size s, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift = 0) {
    adds(s, ARMEmitter::Reg::zr, rn, rm, Option, Shift);
  }
-  void sub(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift = 0) {
+  void sub(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option,
+           uint32_t Shift = 0) {
    constexpr uint32_t Op = 0b100'1011'001U << 21;
    DataProcessing_Extended_Reg(Op, s, rd, rn, rm, Option, Shift);
  }
-  void subs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift = 0) {
+  void subs(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option,
+            uint32_t Shift = 0) {
    constexpr uint32_t Op = 0b110'1011'001U << 21;
    DataProcessing_Extended_Reg(Op, s, rd, rn, rm, Option, Shift);
  }
@@ -751,8 +748,8 @@ public:

  // Rotate right into flags
  void rmif(XRegister rn, uint32_t shift, uint32_t mask) {
-    LOGMAN_THROW_AA_FMT(shift <= 63, "Shift must be within 0-63. Shift: {}", shift);
-    LOGMAN_THROW_AA_FMT(mask <= 15, "Mask must be within 0-15. Mask: {}", mask);
+    LOGMAN_THROW_A_FMT(shift <= 63, "Shift must be within 0-63. Shift: {}", shift);
+    LOGMAN_THROW_A_FMT(mask <= 15, "Mask must be within 0-15. Mask: {}", mask);

    uint32_t Op = 0b1011'1010'0000'0000'0000'0100'0000'0000;
    Op |= rn.Idx() << 5;
@@ -816,7 +813,8 @@ public:
  }
  void cset(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Condition Cond) {
    constexpr uint32_t Op = 0b0001'1010'100 << 21;
-    ConditionalCompare(Op, 0, 0b01, s, rd, ARMEmitter::Reg::zr, ARMEmitter::Reg::zr, static_cast<ARMEmitter::Condition>(FEXCore::ToUnderlying(Cond) ^ FEXCore::ToUnderlying(ARMEmitter::Condition::CC_NE)));
+    ConditionalCompare(Op, 0, 0b01, s, rd, ARMEmitter::Reg::zr, ARMEmitter::Reg::zr,
+                       static_cast<ARMEmitter::Condition>(FEXCore::ToUnderlying(Cond) ^ FEXCore::ToUnderlying(ARMEmitter::Condition::CC_NE)));
  }
  void csinc(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::Condition Cond) {
    constexpr uint32_t Op = 0b0001'1010'100 << 21;
@@ -896,13 +894,6 @@ public:
  }

 private:
-  static constexpr Condition InvertCondition(Condition cond) {
-    // These behave as always, so it makes no sense to allow inverting these.
-    LOGMAN_THROW_AA_FMT(cond != Condition::CC_AL && cond != Condition::CC_NV,
-                        "Cannot invert CC_AL or CC_NV");
-    return static_cast<Condition>(FEXCore::ToUnderlying(cond) ^ 1);
-  }
-
  void and_(ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint32_t n, uint32_t immr, uint32_t imms) {
    constexpr uint32_t Op = 0b001'0010'00 << 22;
    DataProcessing_Logical_Imm(Op, s, rd, rn, n, immr, imms);
@@ -950,7 +941,7 @@ private:
      LSL12 = true;
      Imm >>= 12;
    }
-    LOGMAN_THROW_AA_FMT(TooLarge == false, "Imm amount too large: 0x{:x}", Imm);
+    LOGMAN_THROW_A_FMT(TooLarge == false, "Imm amount too large: 0x{:x}", Imm);

    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

@@ -995,7 +986,8 @@ private:
  }

  // Logical immediate
-  void DataProcessing_Logical_Imm(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint32_t n, uint32_t immr, uint32_t imms) {
+  void DataProcessing_Logical_Imm(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, uint32_t n,
+                                  uint32_t immr, uint32_t imms) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    uint32_t Instr = Op;
@@ -1011,12 +1003,11 @@ private:
  }

  void xbfiz_helper(bool is_signed, ARMEmitter::Size s, Register rd, Register rn, uint32_t lsb, uint32_t width) {
-    [[maybe_unused]] const auto lsb_p_width = lsb + width;
+    const auto lsb_p_width = lsb + width;
    const auto reg_size_bits = RegSizeInBits(s);

-    LOGMAN_THROW_AA_FMT(lsb_p_width <= reg_size_bits, "lsb + width ({}) must be <= {}. lsb={}, width={}",
-                        lsb_p_width, reg_size_bits, lsb, width);
-    LOGMAN_THROW_AA_FMT(width >= 1, "xbfiz width must be >= 1");
+    LOGMAN_THROW_A_FMT(lsb_p_width <= reg_size_bits, "lsb + width ({}) must be <= {}. lsb={}, width={}", lsb_p_width, reg_size_bits, lsb, width);
+    LOGMAN_THROW_A_FMT(width >= 1, "xbfiz width must be >= 1");

    const auto immr = (reg_size_bits - lsb) & (reg_size_bits - 1);
    const auto imms = width - 1;
@@ -1028,12 +1019,13 @@ private:
    }
  }

-  void DataProcessing_Extract(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, uint32_t Imm) {
+  void DataProcessing_Extract(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm,
+                              uint32_t Imm) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    // Current ARMv8 spec hardcodes SF == N for this class of instructions.
    // Anythign else is undefined behaviour.
-    const uint32_t N  = s == ARMEmitter::Size::i64Bit ? (1U << 22) : 0;
+    const uint32_t N = s == ARMEmitter::Size::i64Bit ? (1U << 22) : 0;

    uint32_t Instr = Op;

@@ -1076,10 +1068,11 @@ private:
  }

  // AddSub - shifted register
-  void DataProcessing_Shifted_Reg(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ShiftType Shift, uint32_t amt) {
-    LOGMAN_THROW_AA_FMT((amt & ~0b11'1111U) == 0, "Shift amount too large");
+  void DataProcessing_Shifted_Reg(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn,
+                                  ARMEmitter::Register rm, ARMEmitter::ShiftType Shift, uint32_t amt) {
+    LOGMAN_THROW_A_FMT((amt & ~0b11'1111U) == 0, "Shift amount too large");
    if (s == ARMEmitter::Size::i32Bit) {
-      LOGMAN_THROW_AA_FMT(amt < 32, "Shift amount for 32-bit must be below 32");
+      LOGMAN_THROW_A_FMT(amt < 32, "Shift amount for 32-bit must be below 32");
    }

    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;
@@ -1097,7 +1090,8 @@ private:
  }

  // AddSub - extended register
-  void DataProcessing_Extended_Reg(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift) {
+  void DataProcessing_Extended_Reg(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn,
+                                   ARMEmitter::Register rm, ARMEmitter::ExtendedType Option, uint32_t Shift) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    uint32_t Instr = Op;
@@ -1113,7 +1107,8 @@ private:
  }
  // Conditional compare - register
  template<typename T>
-  void ConditionalCompare(uint32_t Op, uint32_t o1, uint32_t o2, uint32_t o3, ARMEmitter::Size s, ARMEmitter::Register rn, T rm, ARMEmitter::StatusFlags flags, ARMEmitter::Condition Cond) {
+  void ConditionalCompare(uint32_t Op, uint32_t o1, uint32_t o2, uint32_t o3, ARMEmitter::Size s, ARMEmitter::Register rn, T rm,
+                          ARMEmitter::StatusFlags flags, ARMEmitter::Condition Cond) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    uint32_t Instr = Op;
@@ -1131,7 +1126,8 @@ private:
  }

  template<typename T>
-  void ConditionalCompare(uint32_t Op, uint32_t o1, uint32_t o2, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, T rm, ARMEmitter::Condition Cond) {
+  void ConditionalCompare(uint32_t Op, uint32_t o1, uint32_t o2, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, T rm,
+                          ARMEmitter::Condition Cond) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    uint32_t Instr = Op;
@@ -1148,7 +1144,8 @@ private:
  }

  // Data-processing - 3 source
-  void DataProcessing_3Source(uint32_t Op, uint32_t Op0, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn, ARMEmitter::Register rm, ARMEmitter::Register ra) {
+  void DataProcessing_3Source(uint32_t Op, uint32_t Op0, ARMEmitter::Size s, ARMEmitter::Register rd, ARMEmitter::Register rn,
+                              ARMEmitter::Register rm, ARMEmitter::Register ra) {
    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

    uint32_t Instr = Op;
@@ -1170,4 +1167,7 @@ private:
    dc32(Instr);
  }

-
+#ifndef INCLUDED_BY_EMITTER
+}; // struct LoadstoreEmitterOps
+} // namespace ARMEmitter
+#endif
@@ -3,339 +3,384 @@
 *
 * Most of these instructions will use `BackwardLabel`, `ForwardLabel`, or `BiDirectionLabel` to determine where a branch targets.
 */
+
+#pragma once
+#ifndef INCLUDED_BY_EMITTER
+#include <CodeEmitter/Emitter.h>
+namespace ARMEmitter {
+struct EmitterOps : Emitter {
+#endif
+
 public:
  // Branches, Exception Generating and System instructions
-  public:
-    // Conditional branch immediate
-    ///< Branch conditional
-    void b(ARMEmitter::Condition Cond, uint32_t Imm) {
-      constexpr uint32_t Op = 0b0101'010 << 25;
-      Branch_Conditional(Op, 0, 0, Cond, Imm);
-    }
-    void b(ARMEmitter::Condition Cond, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0), "Unscaled offset too large");
+public:
+  // Conditional branch immediate
+  ///< Branch conditional
+  void b(ARMEmitter::Condition Cond, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0101'010 << 25;
+    Branch_Conditional(Op, 0, 0, Cond, Imm);
+  }
+  [[nodiscard]] BranchEncodeSucceeded b(ARMEmitter::Condition Cond, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+    if (Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0101'010 << 25;
      Branch_Conditional(Op, 0, 0, Cond, Imm >> 2);
-    }
-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void b(ARMEmitter::Condition Cond, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::BC });
-      constexpr uint32_t Op = 0b0101'010 << 25;
-      Branch_Conditional(Op, 0, 0, Cond, 0);
+      return BranchEncodeSucceeded::Success;
    }

-    void b(ARMEmitter::Condition Cond, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        b(Cond, &Label->Backward);
-      }
-      else {
-        b(Cond, &Label->Forward);
-      }
-    }
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }
+  [[nodiscard]] BranchEncodeSucceeded b(ARMEmitter::Condition Cond, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::BC});
+    constexpr uint32_t Op = 0b0101'010 << 25;
+    Branch_Conditional(Op, 0, 0, Cond, 0);

-    ///< Branch consistent conditional
-    void bc(ARMEmitter::Condition Cond, uint32_t Imm) {
-      constexpr uint32_t Op = 0b0101'010 << 25;
-      Branch_Conditional(Op, 0, 1, Cond, Imm);
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded b(ARMEmitter::Condition Cond, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return b(Cond, &Label->Backward);
+    } else {
+      return b(Cond, &Label->Forward);
    }
-    void bc(ARMEmitter::Condition Cond, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0), "Unscaled offset too large");
+  }
+
+  ///< Branch consistent conditional
+  void bc(ARMEmitter::Condition Cond, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0101'010 << 25;
+    Branch_Conditional(Op, 0, 1, Cond, Imm);
+  }
+  [[nodiscard]] BranchEncodeSucceeded bc(ARMEmitter::Condition Cond, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+    if (Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0101'010 << 25;
      Branch_Conditional(Op, 0, 1, Cond, Imm >> 2);
+      return BranchEncodeSucceeded::Success;
    }

-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void bc(ARMEmitter::Condition Cond, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::BC });
-      constexpr uint32_t Op = 0b0101'010 << 25;
-      Branch_Conditional(Op, 0, 1, Cond, 0);
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded bc(ARMEmitter::Condition Cond, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::BC});
+    constexpr uint32_t Op = 0b0101'010 << 25;
+    Branch_Conditional(Op, 0, 1, Cond, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded bc(ARMEmitter::Condition Cond, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return bc(Cond, &Label->Backward);
+    } else {
+      return bc(Cond, &Label->Forward);
    }
+  }

-    void bc(ARMEmitter::Condition Cond, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        bc(Cond, &Label->Backward);
-      }
-      else {
-        bc(Cond, &Label->Forward);
-      }
-    }
+  // Unconditional branch register
+  void br(ARMEmitter::Register rn) {
+    constexpr uint32_t Op = 0b1101011 << 25 | 0b0'000 << 21 | // opc
+                            0b1'1111 << 16 |                  // op2
+                            0b0000'00 << 10 |                 // op3
+                            0b0'0000;                         // op4

-    // Unconditional branch register
-    void br(ARMEmitter::Register rn) {
-      constexpr uint32_t Op = 0b1101011 << 25 |
-                              0b0'000 << 21 |   // opc
-                              0b1'1111 << 16 |  // op2
-                              0b0000'00 << 10 | // op3
-                              0b0'0000;         // op4
+    UnconditionalBranch(Op, rn);
+  }
+  void blr(ARMEmitter::Register rn) {
+    constexpr uint32_t Op = 0b1101011 << 25 | 0b0'001 << 21 | // opc
+                            0b1'1111 << 16 |                  // op2
+                            0b0000'00 << 10 |                 // op3
+                            0b0'0000;                         // op4

-      UnconditionalBranch(Op, rn);
-    }
-    void blr(ARMEmitter::Register rn) {
-      constexpr uint32_t Op = 0b1101011 << 25 |
-                              0b0'001 << 21 |   // opc
-                              0b1'1111 << 16 |  // op2
-                              0b0000'00 << 10 | // op3
-                              0b0'0000;         // op4
+    UnconditionalBranch(Op, rn);
+  }
+  void ret(ARMEmitter::Register rn = ARMEmitter::Reg::r30) {
+    constexpr uint32_t Op = 0b1101011 << 25 | 0b0'010 << 21 | // opc
+                            0b1'1111 << 16 |                  // op2
+                            0b0000'00 << 10 |                 // op3
+                            0b0'0000;                         // op4

-      UnconditionalBranch(Op, rn);
-    }
-    void ret(ARMEmitter::Register rn = ARMEmitter::Reg::r30) {
-      constexpr uint32_t Op = 0b1101011 << 25 |
-                              0b0'010 << 21 |   // opc
-                              0b1'1111 << 16 |  // op2
-                              0b0000'00 << 10 | // op3
-                              0b0'0000;         // op4
+    UnconditionalBranch(Op, rn);
+  }

-      UnconditionalBranch(Op, rn);
-    }
+  // Unconditional branch immediate
+  void b(uint32_t Imm) {
+    constexpr uint32_t Op = 0b0001'01 << 26;

-    // Unconditional branch immediate
-    void b(uint32_t Imm) {
+    UnconditionalBranch(Op, Imm);
+  }
+  [[nodiscard]] BranchEncodeSucceeded b(const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+    if (Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0001'01 << 26;
-
-      UnconditionalBranch(Op, Imm);
-    }
-    void b(BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-      constexpr uint32_t Op = 0b0001'01 << 26;
-
      UnconditionalBranch(Op, Imm >> 2);
-    }
-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void b(LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::B });
-      constexpr uint32_t Op = 0b0001'01 << 26;
-
-      UnconditionalBranch(Op, 0);
+      return BranchEncodeSucceeded::Success;
    }

-    void b(BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        b(&Label->Backward);
-      }
-      else {
-        b(&Label->Forward);
-      }
-    }
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }
+  [[nodiscard]] BranchEncodeSucceeded b(ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::B});
+    constexpr uint32_t Op = 0b0001'01 << 26;

-    void bl(uint32_t Imm) {
+    UnconditionalBranch(Op, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded b(BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return b(&Label->Backward);
+    } else {
+      return b(&Label->Forward);
+    }
+  }
+
+  void bl(uint32_t Imm) {
+    constexpr uint32_t Op = 0b1001'01 << 26;
+
+    UnconditionalBranch(Op, Imm);
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded bl(const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+    if (Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b1001'01 << 26;
-
-      UnconditionalBranch(Op, Imm);
-    }
-
-    void bl(BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-      constexpr uint32_t Op = 0b1001'01 << 26;
-
      UnconditionalBranch(Op, Imm >> 2);
-    }
-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void bl(LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::B });
-      constexpr uint32_t Op = 0b1001'01 << 26;

-      UnconditionalBranch(Op, 0);
+      return BranchEncodeSucceeded::Success;
    }

-    void bl(BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        bl(&Label->Backward);
-      }
-      else {
-        bl(&Label->Forward);
-      }
-    }
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }
+  [[nodiscard]] BranchEncodeSucceeded bl(ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::B});
+    constexpr uint32_t Op = 0b1001'01 << 26;

-    // Compare and branch
-    void cbz(ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
+    UnconditionalBranch(Op, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded bl(BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return bl(&Label->Backward);
+    } else {
+      return bl(&Label->Forward);
+    }
+  }
+
+  // Compare and branch
+  void cbz(ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0011'0100 << 24;
+
+    CompareAndBranch(Op, s, rt, Imm);
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded cbz(ARMEmitter::Size s, ARMEmitter::Register rt, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+
+    if (Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0011'0100 << 24;
-
-      CompareAndBranch(Op, s, rt, Imm);
-    }
-
-    void cbz(ARMEmitter::Size s, ARMEmitter::Register rt, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-
-      constexpr uint32_t Op = 0b0011'0100 << 24;
-
      CompareAndBranch(Op, s, rt, Imm >> 2);
+      return BranchEncodeSucceeded::Success;
    }

-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void cbz(ARMEmitter::Size s, ARMEmitter::Register rt, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::BC });
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }

-      constexpr uint32_t Op = 0b0011'0100 << 24;
+  [[nodiscard]] BranchEncodeSucceeded cbz(ARMEmitter::Size s, ARMEmitter::Register rt, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::BC});

-      CompareAndBranch(Op, s, rt, 0);
+    constexpr uint32_t Op = 0b0011'0100 << 24;
+
+    CompareAndBranch(Op, s, rt, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded cbz(ARMEmitter::Size s, ARMEmitter::Register rt, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return cbz(s, rt, &Label->Backward);
+    } else {
+      return cbz(s, rt, &Label->Forward);
    }
+  }

-    void cbz(ARMEmitter::Size s, ARMEmitter::Register rt, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        cbz(s, rt, &Label->Backward);
-      }
-      else {
-        cbz(s, rt, &Label->Forward);
-      }
-    }
+  void cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0011'0101 << 24;

-    void cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
+    CompareAndBranch(Op, s, rt, Imm);
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+
+    if (Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0011'0101 << 24;
-
-      CompareAndBranch(Op, s, rt, Imm);
-    }
-
-    void cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-
-      constexpr uint32_t Op = 0b0011'0101 << 24;
-
      CompareAndBranch(Op, s, rt, Imm >> 2);
+      return BranchEncodeSucceeded::Success;
    }

-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::BC });
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }

-      constexpr uint32_t Op = 0b0011'0101 << 24;
+  [[nodiscard]] BranchEncodeSucceeded cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::BC});

-      CompareAndBranch(Op, s, rt, 0);
+    constexpr uint32_t Op = 0b0011'0101 << 24;
+
+    CompareAndBranch(Op, s, rt, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return cbnz(s, rt, &Label->Backward);
+    } else {
+      return cbnz(s, rt, &Label->Forward);
    }
+  }

-    void cbnz(ARMEmitter::Size s, ARMEmitter::Register rt, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        cbnz(s, rt, &Label->Backward);
-      }
-      else {
-        cbnz(s, rt, &Label->Forward);
-      }
-    }
+  // Test and branch immediate
+  void tbz(ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0011'0110 << 24;

-    // Test and branch immediate
-    void tbz(ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
+    TestAndBranch(Op, rt, Bit, Imm);
+  }
+  [[nodiscard]] BranchEncodeSucceeded tbz(ARMEmitter::Register rt, uint32_t Bit, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+
+    if (Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0011'0110 << 24;
-
-      TestAndBranch(Op, rt, Bit, Imm);
-    }
-    void tbz(ARMEmitter::Register rt, uint32_t Bit, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-
-      constexpr uint32_t Op = 0b0011'0110 << 24;
-
      TestAndBranch(Op, rt, Bit, Imm >> 2);
+      return BranchEncodeSucceeded::Success;
    }

-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void tbz(ARMEmitter::Register rt, uint32_t Bit, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::TEST_BRANCH });
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }

-      constexpr uint32_t Op = 0b0011'0110 << 24;
+  [[nodiscard]] BranchEncodeSucceeded tbz(ARMEmitter::Register rt, uint32_t Bit, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::TEST_BRANCH});

-      TestAndBranch(Op, rt, Bit, 0);
+    constexpr uint32_t Op = 0b0011'0110 << 24;
+
+    TestAndBranch(Op, rt, Bit, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded tbz(ARMEmitter::Register rt, uint32_t Bit, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return tbz(rt, Bit, &Label->Backward);
+    } else {
+      return tbz(rt, Bit, &Label->Forward);
    }
+  }

-    void tbz(ARMEmitter::Register rt, uint32_t Bit, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        tbz(rt, Bit, &Label->Backward);
-      }
-      else {
-        tbz(rt, Bit, &Label->Forward);
-      }
-    }
+  void tbnz(ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
+    constexpr uint32_t Op = 0b0011'0111 << 24;

-    void tbnz(ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
+    TestAndBranch(Op, rt, Bit, Imm);
+  }
+  [[nodiscard]] BranchEncodeSucceeded tbnz(ARMEmitter::Register rt, uint32_t Bit, const BackwardLabel* Label) {
+    int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
+
+    if (Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0)) {
      constexpr uint32_t Op = 0b0011'0111 << 24;
-
-      TestAndBranch(Op, rt, Bit, Imm);
-    }
-    void tbnz(ARMEmitter::Register rt, uint32_t Bit, BackwardLabel const* Label) {
-      int32_t Imm = static_cast<int32_t>(Label->Location - GetCursorAddress<uint8_t*>());
-      LOGMAN_THROW_A_FMT(Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0), "Unscaled offset too large");
-
-      constexpr uint32_t Op = 0b0011'0111 << 24;
-
      TestAndBranch(Op, rt, Bit, Imm >> 2);
+      return BranchEncodeSucceeded::Success;
    }

-    template<typename LabelType>
-    requires (std::is_same_v<LabelType, ForwardLabel> || std::is_same_v<LabelType, SingleUseForwardLabel>)
-    void tbnz(ARMEmitter::Register rt, uint32_t Bit, LabelType *Label) {
-      AddLocationToLabel(Label, SingleUseForwardLabel{ .Location = GetCursorAddress<uint8_t*>(), .Type = SingleUseForwardLabel::InstType::TEST_BRANCH });
-      constexpr uint32_t Op = 0b0011'0111 << 24;
+    // Can't encode.
+    return BranchEncodeSucceeded::Failure;
+  }

-      TestAndBranch(Op, rt, Bit, 0);
-    }
+  [[nodiscard]] BranchEncodeSucceeded tbnz(ARMEmitter::Register rt, uint32_t Bit, ForwardLabel* Label) {
+    AddLocationToLabel(Label, ForwardLabel::Reference {.Location = GetCursorAddress<uint8_t*>(), .Type = ForwardLabel::InstType::TEST_BRANCH});
+    constexpr uint32_t Op = 0b0011'0111 << 24;

-    void tbnz(ARMEmitter::Register rt, uint32_t Bit, BiDirectionalLabel *Label) {
-      if (Label->Backward.Location) {
-        tbnz(rt, Bit, &Label->Backward);
-      }
-      else {
-        tbnz(rt, Bit, &Label->Forward);
-      }
+    TestAndBranch(Op, rt, Bit, 0);
+
+    // Forward label doesn't know if it can encode until Bind.
+    return BranchEncodeSucceeded::Success;
+  }
+
+  [[nodiscard]] BranchEncodeSucceeded tbnz(ARMEmitter::Register rt, uint32_t Bit, BiDirectionalLabel* Label) {
+    if (Label->Backward.Location) {
+      return tbnz(rt, Bit, &Label->Backward);
+    } else {
+      return tbnz(rt, Bit, &Label->Forward);
    }
+  }

 private:
-    // Conditional branch immediate
-    void Branch_Conditional(uint32_t Op, uint32_t Op1, uint32_t Op0, ARMEmitter::Condition Cond, uint32_t Imm) {
-      uint32_t Instr = Op;
+  // Conditional branch immediate
+  void Branch_Conditional(uint32_t Op, uint32_t Op1, uint32_t Op0, ARMEmitter::Condition Cond, uint32_t Imm) {
+    uint32_t Instr = Op;

-      Instr |= Op1 << 24;
-      Instr |= (Imm & 0x7'FFFF) << 5;
-      Instr |= Op0 << 4;
-      Instr |= FEXCore::ToUnderlying(Cond);
+    Instr |= Op1 << 24;
+    Instr |= (Imm & 0x7'FFFF) << 5;
+    Instr |= Op0 << 4;
+    Instr |= FEXCore::ToUnderlying(Cond);

-      dc32(Instr);
-    }
+    dc32(Instr);
+  }

-    // Unconditional branch register
-    void UnconditionalBranch(uint32_t Op, ARMEmitter::Register rn) {
-      uint32_t Instr = Op;
-      Instr |= Encode_rn(rn);
-      dc32(Instr);
-    }
+  // Unconditional branch register
+  void UnconditionalBranch(uint32_t Op, ARMEmitter::Register rn) {
+    uint32_t Instr = Op;
+    Instr |= Encode_rn(rn);
+    dc32(Instr);
+  }

-    // Unconditional branch - immediate
-    void UnconditionalBranch(uint32_t Op, uint32_t Imm) {
-      uint32_t Instr = Op;
-      Instr |= Imm & 0x3FF'FFFF;
-      dc32(Instr);
-    }
+  // Unconditional branch - immediate
+  void UnconditionalBranch(uint32_t Op, uint32_t Imm) {
+    uint32_t Instr = Op;
+    Instr |= Imm & 0x3FF'FFFF;
+    dc32(Instr);
+  }

-    // Compare and branch
-    void CompareAndBranch(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
-      const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;
+  // Compare and branch
+  void CompareAndBranch(uint32_t Op, ARMEmitter::Size s, ARMEmitter::Register rt, uint32_t Imm) {
+    const uint32_t SF = s == ARMEmitter::Size::i64Bit ? (1U << 31) : 0;

-      uint32_t Instr = Op;
+    uint32_t Instr = Op;

-      Instr |= SF;
-      Instr |= (Imm & 0x7'FFFF) << 5;
-      Instr |= Encode_rt(rt);
-      dc32(Instr);
-    }
+    Instr |= SF;
+    Instr |= (Imm & 0x7'FFFF) << 5;
+    Instr |= Encode_rt(rt);
+    dc32(Instr);
+  }

-    // Test and branch - immediate
-    void TestAndBranch(uint32_t Op, ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
-      uint32_t Instr = Op;
+  // Test and branch - immediate
+  void TestAndBranch(uint32_t Op, ARMEmitter::Register rt, uint32_t Bit, uint32_t Imm) {
+    uint32_t Instr = Op;

-      Instr |= (Bit >> 5) << 31;
-      Instr |= (Bit & 0b1'1111) << 19;
-      Instr |= (Imm & 0x3FFF) << 5;
-      Instr |= Encode_rt(rt);
-      dc32(Instr);
-    }
+    Instr |= (Bit >> 5) << 31;
+    Instr |= (Bit & 0b1'1111) << 19;
+    Instr |= (Imm & 0x3FFF) << 5;
+    Instr |= Encode_rt(rt);
+    dc32(Instr);
+  }
+
+#ifndef INCLUDED_BY_EMITTER
+}; // struct LoadstoreEmitterOps
+} // namespace ARMEmitter
+#endif
@@ -3,6 +3,7 @@
 #include <cstddef>
 #include <cstdint>
 #include <cstring>
+#include <type_traits>

 namespace ARMEmitter {
 class Buffer {
@@ -21,42 +22,39 @@ public:
    Size = BaseSize;
  }

+  template<typename T>
+  requires (std::is_trivially_copyable_v<T>)
+  void dcn(const T& Data) {
+    std::memcpy(CurrentOffset, &Data, sizeof(Data));
+    CurrentOffset += sizeof(Data);
+  }
  void dc8(uint8_t Data) {
-    decltype(Data)* Memory = reinterpret_cast<decltype(Data)*>(CurrentOffset);
-    *Memory = Data;
-    CurrentOffset += sizeof(Data);
+    dcn(Data);
  }
-
  void dc16(uint16_t Data) {
-    decltype(Data)* Memory = reinterpret_cast<decltype(Data)*>(CurrentOffset);
-    *Memory = Data;
-    CurrentOffset += sizeof(Data);
+    dcn(Data);
  }
-
  void dc32(uint32_t Data) {
-    decltype(Data)* Memory = reinterpret_cast<decltype(Data)*>(CurrentOffset);
-    *Memory = Data;
-    CurrentOffset += sizeof(Data);
+    dcn(Data);
+  }
+  void dc64(uint64_t Data) {
+    dcn(Data);
  }

-  void dc64(uint64_t Data) {
-    decltype(Data)* Memory = reinterpret_cast<decltype(Data)*>(CurrentOffset);
-    *Memory = Data;
-    CurrentOffset += sizeof(Data);
-  }
  void EmitString(const char* String) {
    const auto StringLength = strlen(String);
    memcpy(CurrentOffset, String, StringLength);
    CurrentOffset += StringLength;
  }

-  void Align() {
-    // Align the buffer to instruction size
-    auto CurrentAlignment = reinterpret_cast<uint64_t>(CurrentOffset) & 0b11;
+  void Align(size_t Size = 4) {
+    // Align the buffer to provided size.
+    auto CurrentAlignment = reinterpret_cast<uint64_t>(CurrentOffset) & (Size - 1);
    if (!CurrentAlignment) {
      return;
    }
-    CurrentOffset += 4 - CurrentAlignment;
+    std::memset(CurrentOffset, 0, Size - CurrentAlignment);
+    CurrentOffset += Size - CurrentAlignment;
  }

  template<typename T>
@@ -95,10 +93,6 @@ public:

 protected:

-  void ResetBuffer() {
-    CurrentOffset = BufferBase;
-  }
-
  uint8_t* BufferBase;
  uint8_t* CurrentOffset;
  uint64_t Size;
@@ -12,6 +12,7 @@
 #include <CodeEmitter/Registers.h>

 #include <array>
+#include <bit>
 #include <cstdint>
 #include <utility>
 #include <type_traits>
@@ -86,6 +87,14 @@ constexpr size_t SubRegSizeInBits(SubRegSize size) {
  return size_t {8} << FEXCore::ToUnderlying(size);
 }

+// Many floating point operations constrain their element sizes to the
+// main three float sizes half, single, and double precision. This just
+// combines all the checks together for brevity.
+[[nodiscard]]
+constexpr bool IsStandardFloatSize(SubRegSize size) {
+  return size == SubRegSize::i16Bit || size == SubRegSize::i32Bit || size == SubRegSize::i64Bit;
+}
+
 /* This `ScalarRegSize` enum is used for most scalar float
 * operations.
 *
@@ -302,7 +311,7 @@ class ExtendedMemOperand final {
 public:
  ExtendedMemOperand(XRegister rn, XRegister rm = XReg::zr, ExtendedType Option = ExtendedType::LSL_64, uint32_t Shift = 0)
    : rn {rn}
-    , MetaType {.ExtendedType {
+    , MetaType {.Extended {
        .Header = {.MemType = TYPE_EXTENDED},
        .rm = rm,
        .Option = Option,
@@ -331,7 +340,7 @@ public:
      Register rm;
      ExtendedType Option;
      uint32_t Shift;
-    } ExtendedType;
+    } Extended;
    struct {
      HeaderStruct Header;
      IndexType Index;
@@ -341,94 +350,89 @@ public:
 };

 template<uint32_t op0, uint32_t op1, uint32_t CRn, uint32_t CRm, uint32_t op2>
-constexpr uint32_t GenSystemReg() {
-  return op0 << 19 | op1 << 16 | CRn << 12 | CRm << 8 | op2 << 5;
-};
+inline constexpr uint32_t GenSystemReg = op0 << 19 | op1 << 16 | CRn << 12 | CRm << 8 | op2 << 5;

 // This `SystemRegister` enum is used for the mrs/msr instructions.
 enum class SystemRegister : uint32_t {
-  CTR_EL0 = GenSystemReg<0b11, 0b011, 0b0000, 0b0000, 0b001>(),
-  DCZID_EL0 = GenSystemReg<0b11, 0b011, 0b0000, 0b0000, 0b111>(),
-  TPIDR_EL0 = GenSystemReg<0b11, 0b011, 0b1101, 0b0000, 0b010>(),
-  RNDR = GenSystemReg<0b11, 0b011, 0b0010, 0b0100, 0b000>(),
-  RNDRRS = GenSystemReg<0b11, 0b011, 0b0010, 0b0100, 0b001>(),
-  NZCV = GenSystemReg<0b11, 0b011, 0b0100, 0b0010, 0b000>(),
-  FPCR = GenSystemReg<0b11, 0b011, 0b0100, 0b0100, 0b000>(),
-  TPIDRRO_EL0 = GenSystemReg<0b11, 0b011, 0b1101, 0b0000, 0b011>(),
-  CNTFRQ_EL0 = GenSystemReg<0b11, 0b011, 0b1110, 0b0000, 0b000>(),
-  CNTVCT_EL0 = GenSystemReg<0b11, 0b011, 0b1110, 0b0000, 0b010>(),
+  CTR_EL0 = GenSystemReg<0b11, 0b011, 0b0000, 0b0000, 0b001>,
+  DCZID_EL0 = GenSystemReg<0b11, 0b011, 0b0000, 0b0000, 0b111>,
+  TPIDR_EL0 = GenSystemReg<0b11, 0b011, 0b1101, 0b0000, 0b010>,
+  RNDR = GenSystemReg<0b11, 0b011, 0b0010, 0b0100, 0b000>,
+  RNDRRS = GenSystemReg<0b11, 0b011, 0b0010, 0b0100, 0b001>,
+  NZCV = GenSystemReg<0b11, 0b011, 0b0100, 0b0010, 0b000>,
+  FPCR = GenSystemReg<0b11, 0b011, 0b0100, 0b0100, 0b000>,
+  TPIDRRO_EL0 = GenSystemReg<0b11, 0b011, 0b1101, 0b0000, 0b011>,
+  CNTFRQ_EL0 = GenSystemReg<0b11, 0b011, 0b1110, 0b0000, 0b000>,
+  CNTVCT_EL0 = GenSystemReg<0b11, 0b011, 0b1110, 0b0000, 0b010>,
+  CNTVCTSS_EL0 = GenSystemReg<0b11, 0b011, 0b1110, 0b0000, 0b110>,
 };

 template<uint32_t op1, uint32_t CRm, uint32_t op2>
-constexpr uint32_t GenDCReg() {
-  return op1 << 16 | CRm << 8 | op2 << 5;
-};
+inline constexpr uint32_t GenDCReg = op1 << 16 | CRm << 8 | op2 << 5;

 // This `DataCacheOperation` enum is used for the dc instruction.
 enum class DataCacheOperation : uint32_t {
-  IVAC = GenDCReg<0b000, 0b0110, 0b001>(),
-  ISW = GenDCReg<0b000, 0b0110, 0b010>(),
-  CSW = GenDCReg<0b000, 0b1010, 0b010>(),
-  CISW = GenDCReg<0b000, 0b1110, 0b010>(),
-  ZVA = GenDCReg<0b011, 0b0100, 0b001>(),
-  CVAC = GenDCReg<0b011, 0b1010, 0b001>(),
-  CVAU = GenDCReg<0b011, 0b1011, 0b001>(),
-  CIVAC = GenDCReg<0b011, 0b1110, 0b001>(),
+  IVAC = GenDCReg<0b000, 0b0110, 0b001>,
+  ISW = GenDCReg<0b000, 0b0110, 0b010>,
+  CSW = GenDCReg<0b000, 0b1010, 0b010>,
+  CISW = GenDCReg<0b000, 0b1110, 0b010>,
+  ZVA = GenDCReg<0b011, 0b0100, 0b001>,
+  CVAC = GenDCReg<0b011, 0b1010, 0b001>,
+  CVAU = GenDCReg<0b011, 0b1011, 0b001>,
+  CIVAC = GenDCReg<0b011, 0b1110, 0b001>,

  // MTE2
-  IGVAC = GenDCReg<0b000, 0b0110, 0b011>(),
-  IGSW = GenDCReg<0b000, 0b0110, 0b100>(),
-  IGDVAC = GenDCReg<0b000, 0b0110, 0b101>(),
-  IGDSW = GenDCReg<0b000, 0b0110, 0b110>(),
-  CGSW = GenDCReg<0b000, 0b1010, 0b100>(),
-  CGDSW = GenDCReg<0b000, 0b1010, 0b110>(),
-  CIGSW = GenDCReg<0b000, 0b1110, 0b100>(),
-  CIGDSW = GenDCReg<0b000, 0b1110, 0b110>(),
+  IGVAC = GenDCReg<0b000, 0b0110, 0b011>,
+  IGSW = GenDCReg<0b000, 0b0110, 0b100>,
+  IGDVAC = GenDCReg<0b000, 0b0110, 0b101>,
+  IGDSW = GenDCReg<0b000, 0b0110, 0b110>,
+  CGSW = GenDCReg<0b000, 0b1010, 0b100>,
+  CGDSW = GenDCReg<0b000, 0b1010, 0b110>,
+  CIGSW = GenDCReg<0b000, 0b1110, 0b100>,
+  CIGDSW = GenDCReg<0b000, 0b1110, 0b110>,

  // MTE
-  GVA = GenDCReg<0b011, 0b0100, 0b011>(),
-  GZVA = GenDCReg<0b011, 0b0100, 0b100>(),
-  CGVAC = GenDCReg<0b011, 0b1010, 0b011>(),
-  CGDVAC = GenDCReg<0b011, 0b1010, 0b101>(),
-  CGVAP = GenDCReg<0b011, 0b1100, 0b011>(),
-  CGDVAP = GenDCReg<0b011, 0b1100, 0b101>(),
-  CGVADP = GenDCReg<0b011, 0b1101, 0b011>(),
-  CGDVADP = GenDCReg<0b011, 0b1101, 0b101>(),
-  CIGVAC = GenDCReg<0b011, 0b1110, 0b011>(),
-  CIGDVAC = GenDCReg<0b011, 0b1110, 0b101>(),
+  GVA = GenDCReg<0b011, 0b0100, 0b011>,
+  GZVA = GenDCReg<0b011, 0b0100, 0b100>,
+  CGVAC = GenDCReg<0b011, 0b1010, 0b011>,
+  CGDVAC = GenDCReg<0b011, 0b1010, 0b101>,
+  CGVAP = GenDCReg<0b011, 0b1100, 0b011>,
+  CGDVAP = GenDCReg<0b011, 0b1100, 0b101>,
+  CGVADP = GenDCReg<0b011, 0b1101, 0b011>,
+  CGDVADP = GenDCReg<0b011, 0b1101, 0b101>,
+  CIGVAC = GenDCReg<0b011, 0b1110, 0b011>,
+  CIGDVAC = GenDCReg<0b011, 0b1110, 0b101>,

  // DPB
-  CVAP = GenDCReg<0b011, 0b1100, 0b001>(),
+  CVAP = GenDCReg<0b011, 0b1100, 0b001>,

  // DPB2
-  CVADP = GenDCReg<0b011, 0b1101, 0b001>(),
+  CVADP = GenDCReg<0b011, 0b1101, 0b001>,
 };

 template<uint32_t CRm, uint32_t op2>
-constexpr uint32_t GenHintBarrierReg() {
-  return CRm << 8 | op2 << 5;
-}
+inline constexpr uint32_t GenHintBarrierReg = CRm << 8 | op2 << 5;

 // This `HintRegister` enum is used for the hint instruction.
 enum class HintRegister : uint32_t {
-  NOP = GenHintBarrierReg<0b0000, 0b000>(),
-  YIELD = GenHintBarrierReg<0b0000, 0b001>(),
-  WFE = GenHintBarrierReg<0b0000, 0b010>(),
-  WFI = GenHintBarrierReg<0b0000, 0b011>(),
-  SEV = GenHintBarrierReg<0b0000, 0b100>(),
-  SEVL = GenHintBarrierReg<0b0000, 0b101>(),
-  DGH = GenHintBarrierReg<0b0000, 0b110>(),
-  CSDB = GenHintBarrierReg<0b0010, 0b100>(),
+  NOP = GenHintBarrierReg<0b0000, 0b000>,
+  YIELD = GenHintBarrierReg<0b0000, 0b001>,
+  WFE = GenHintBarrierReg<0b0000, 0b010>,
+  WFI = GenHintBarrierReg<0b0000, 0b011>,
+  SEV = GenHintBarrierReg<0b0000, 0b100>,
+  SEVL = GenHintBarrierReg<0b0000, 0b101>,
+  DGH = GenHintBarrierReg<0b0000, 0b110>,
+  CSDB = GenHintBarrierReg<0b0010, 0b100>,
 };

 // This `BarrierRegister` enum is used for the various barrier instructions.
 enum class BarrierRegister : uint32_t {
-  CLREX = GenHintBarrierReg<0b0000, 0b010>(),
-  TCOMMIT = GenHintBarrierReg<0b0000, 0b011>(),
-  DSB = GenHintBarrierReg<0b0000, 0b100>(),
-  DMB = GenHintBarrierReg<0b0000, 0b101>(),
-  ISB = GenHintBarrierReg<0b0000, 0b110>(),
-  SB = GenHintBarrierReg<0b0000, 0b111>(),
+  CLREX = GenHintBarrierReg<0b0000, 0b010>,
+  TCOMMIT = GenHintBarrierReg<0b0000, 0b011>,
+  DSB = GenHintBarrierReg<0b0000, 0b100>,
+  DMB = GenHintBarrierReg<0b0000, 0b101>,
+  ISB = GenHintBarrierReg<0b0000, 0b110>,
+  SB = GenHintBarrierReg<0b0000, 0b111>,
 };

 // This `BarrierScope` enum is used for the dsb/dmb instructions.
@@ -513,7 +517,7 @@ enum class SVEFMaxMinImm : uint32_t {
  _1_0,
 };

-/* This `BackwardLabel` struct used for retaining a location for PC-Relative instructions.
+/* This `BackwardLabel` struct is used for retaining a location for PC-Relative instructions.
 * This is specifically a label for a target that is logically `below` an instruction that uses it.
 * Which means that a branch would jump backwards.
 */
@@ -521,13 +525,11 @@ struct BackwardLabel {
  uint8_t* Location {};
 };

-/* This `SingleUseForwardLabel` struct used for retaining a location for PC-Relative instructions.
+/* This `ForwardLabel` struct is used for retaining a location for PC-Relative instructions.
 * This is specifically a label for a target that is logically `above` an instruction that uses it.
 * Which means that a branch would jump forwards.
- *
- * The `ForwardLabel` struct can be bound to multiple instructions, so it needs a vector for each bind instruction type.
 */
-struct SingleUseForwardLabel {
+struct ForwardLabel {
  enum class InstType {
    UNKNOWN,
    ADR,
@@ -538,12 +540,16 @@ struct SingleUseForwardLabel {
    RELATIVE_LOAD,
    LONG_ADDRESS_GEN,
  };
-  uint8_t* Location {};
-  InstType Type = InstType::UNKNOWN;
-};

-struct ForwardLabel {
-  fextl::vector<SingleUseForwardLabel> Insts {};
+  struct Reference {
+    uint8_t* Location {};
+    InstType Type = InstType::UNKNOWN;
+  };
+
+  // The first element is stored separately to avoid allocations for simple cases
+  Reference FirstInst;
+
+  fextl::vector<Reference> Insts;
 };

 /* This `BiDirectionalLabel` struct used for retaining a location for PC-Relative instructions.
@@ -555,14 +561,12 @@ struct BiDirectionalLabel {
  ForwardLabel Forward;
 };

-static inline void AddLocationToLabel(SingleUseForwardLabel* Label, SingleUseForwardLabel&& Location) {
-  LOGMAN_THROW_A_FMT(Label->Type == SingleUseForwardLabel::InstType::UNKNOWN, "Trying to bind a SingleUseForwardLabel to multiple "
-                                                                              "locations. Use ForwardLabel instead.");
-  *Label = std::move(Location);
-}
-
-static inline void AddLocationToLabel(ForwardLabel* Label, SingleUseForwardLabel&& Location) {
-  Label->Insts.emplace_back(std::move(Location));
+static inline void AddLocationToLabel(ForwardLabel* Label, ForwardLabel::Reference&& Location) {
+  if (Label->FirstInst.Location == nullptr) {
+    Label->FirstInst = Location;
+  } else {
+    Label->Insts.push_back(Location);
+  }
 }

 // Some FCMA ASIMD instructions support a rotation argument.
@@ -578,6 +582,19 @@ enum class Rotation : uint32_t {
 template<typename T>
 concept IsXOrWRegister = std::is_same_v<T, XRegister> || std::is_same_v<T, WRegister>;

+// Concept for contraining some instructions to accept only a QRegister or DRegister.
+template<typename T>
+concept IsQOrDRegister = std::is_same_v<T, QRegister> || std::is_same_v<T, DRegister>;
+
+template<typename T>
+concept IsLabel = std::is_same_v<T, ARMEmitter::ForwardLabel> || std::is_same_v<T, ARMEmitter::BackwardLabel> ||
+                  std::is_same_v<T, ARMEmitter::BiDirectionalLabel> || std::is_same_v<T, ARMEmitter::ForwardLabel::Reference>;
+
+enum class BranchEncodeSucceeded {
+  Success,
+  Failure,
+};
+
 // Whether or not a given set of vector registers are sequential
 // in increasing order as far as the register file is concerned (modulo its size)
 //
@@ -630,19 +647,25 @@ public:

  // Bind a backward label to an address.
  // Address that is bound is the current emitter location.
-  void Bind(BackwardLabel* Label) {
-    LOGMAN_THROW_AA_FMT(Label->Location == nullptr, "Trying to bind a label twice");
+  [[nodiscard]] bool Bind(BackwardLabel* Label) {
+    LOGMAN_THROW_A_FMT(Label->Location == nullptr, "Trying to bind a label twice");
    Label->Location = GetCursorAddress<uint8_t*>();
+
+    // Always binds because it is only storing a location.
+    return true;
  }

-  void Bind(const SingleUseForwardLabel* Label) {
+  [[nodiscard]] bool Bind(const ForwardLabel::Reference* Label) {
    uint8_t* CurrentAddress = GetCursorAddress<uint8_t*>();
    // Patch up the instructions
    switch (Label->Type) {
-    case SingleUseForwardLabel::InstType::ADR: {
+    case ForwardLabel::InstType::ADR: {
      uint32_t* Instruction = reinterpret_cast<uint32_t*>(Label->Location);
      int64_t Imm = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(Instruction);
-      LOGMAN_THROW_A_FMT(IsADRRange(Imm), "Unscaled offset too large");
+      if (!IsADRRange(Imm)) {
+        // Can't bind.
+        return false;
+      }
      uint32_t InstMask = 0b11 << 29 | 0b1111'1111'1111'1111'111 << 5;
      uint32_t Offset = static_cast<uint32_t>(Imm) & 0x3F'FFFF;
      uint32_t Inst = *Instruction & ~InstMask;
@@ -651,10 +674,15 @@ public:
      *Instruction = Inst;
      break;
    }
-    case SingleUseForwardLabel::InstType::ADRP: {
+    case ForwardLabel::InstType::ADRP: {
      uint32_t* Instruction = reinterpret_cast<uint32_t*>(Label->Location);
      int64_t Imm = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(Instruction);
-      LOGMAN_THROW_A_FMT(IsADRPRange(Imm) && IsADRPAligned(Imm), "Unscaled offset too large");
+
+      if (!(IsADRPRange(Imm) && IsADRPAligned(Imm))) {
+        // Can't bind.
+        return false;
+      }
+
      Imm >>= 12;
      uint32_t InstMask = 0b11 << 29 | 0b1111'1111'1111'1111'111 << 5;
      uint32_t Offset = static_cast<uint32_t>(Imm) & 0x3F'FFFF;
@@ -664,11 +692,13 @@ public:
      *Instruction = Inst;
      break;
    }
-
-    case SingleUseForwardLabel::InstType::B: {
+    case ForwardLabel::InstType::B: {
      uint32_t* Instruction = reinterpret_cast<uint32_t*>(Label->Location);
      int64_t Imm = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(Instruction);
-      LOGMAN_THROW_A_FMT(Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0), "Unscaled offset too large");
+      if (!(Imm >= -134217728 && Imm <= 134217724 && ((Imm & 0b11) == 0))) {
+        // Can't bind.
+        return false;
+      }
      Imm >>= 2;
      uint32_t InstMask = 0x3FF'FFFF;
      uint32_t Offset = static_cast<uint32_t>(Imm) & InstMask;
@@ -678,11 +708,13 @@ public:

      break;
    }
-
-    case SingleUseForwardLabel::InstType::TEST_BRANCH: {
+    case ForwardLabel::InstType::TEST_BRANCH: {
      uint32_t* Instruction = reinterpret_cast<uint32_t*>(Label->Location);
      int64_t Imm = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(Instruction);
-      LOGMAN_THROW_A_FMT(Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0), "Unscaled offset too large");
+      if (!(Imm >= -32768 && Imm <= 32764 && ((Imm & 0b11) == 0))) {
+        // Can't bind.
+        return false;
+      }
      Imm >>= 2;
      uint32_t InstMask = 0x3FFF;
      uint32_t Offset = static_cast<uint32_t>(Imm) & InstMask;
@@ -692,11 +724,14 @@ public:

      break;
    }
-    case SingleUseForwardLabel::InstType::BC:
-    case SingleUseForwardLabel::InstType::RELATIVE_LOAD: {
+    case ForwardLabel::InstType::BC:
+    case ForwardLabel::InstType::RELATIVE_LOAD: {
      uint32_t* Instruction = reinterpret_cast<uint32_t*>(Label->Location);
      int64_t Imm = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(Instruction);
-      LOGMAN_THROW_A_FMT(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0), "Unscaled offset too large");
+      if (!(Imm >= -1048576 && Imm <= 1048575 && ((Imm & 0b11) == 0))) {
+        // Can't bind.
+        return false;
+      }
      Imm >>= 2;
      uint32_t InstMask = 0x7'FFFF;
      uint32_t Offset = static_cast<uint32_t>(Imm) & InstMask;
@@ -705,39 +740,45 @@ public:
      *Instruction = Inst;
      break;
    }
-    case SingleUseForwardLabel::InstType::LONG_ADDRESS_GEN: {
-      uint32_t* Instructions = reinterpret_cast<uint32_t*>(Label->Location);
-      int64_t ImmInstOne = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(&Instructions[0]);
-      int64_t ImmInstTwo = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(&Instructions[1]);
-      auto OriginalOffset = GetCursorOffset();
+    case ForwardLabel::InstType::LONG_ADDRESS_GEN: {
+      const auto* Instructions = reinterpret_cast<uint32_t*>(Label->Location);
+      const auto ImmInstOne = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(&Instructions[0]);
+      const auto ImmInstTwo = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(&Instructions[1]);
+      const auto ImmInstThree = reinterpret_cast<int64_t>(CurrentAddress) - reinterpret_cast<int64_t>(&Instructions[2]);
+      const auto OriginalOffset = GetCursorOffset();

-      auto InstOffset = GetCursorOffsetFromAddress(Instructions);
+      const auto InstOffset = GetCursorOffsetFromAddress(Instructions);
      SetCursorOffset(InstOffset);

      // We encoded the destination register in to the first instruction space.
      // Read it back.
      ARMEmitter::Register DestReg(Instructions[0]);

-      if (IsADRRange(ImmInstTwo)) {
-        // If within ADR range from the second instruction, then we can emit NOP+ADR
+      if (IsADRRange(ImmInstThree)) {
+        // If within ADR range from the third instruction, then we can emit NOP+NOP+ADR
        nop();
-        adr(DestReg, static_cast<uint32_t>(ImmInstTwo) & 0x7FFF);
-      } else if (IsADRPRange(ImmInstOne)) {
+        nop();
+        adr(DestReg, static_cast<uint32_t>(ImmInstThree) & 0x7FFF);
+      } else if (IsADRPRange(ImmInstTwo)) {

        // If within ADRP range from the first instruction, then we are /definitely/ in range for the second instruction.
        // First check if we are in non-offset range for second instruction.
        if (IsADRPAligned(reinterpret_cast<uint64_t>(CurrentAddress))) {
-          // We can emit nop + adrp
+          // We can emit nop + nop + adrp
+          nop();
+          nop();
+          adrp(DestReg, static_cast<uint32_t>(ImmInstThree >> 12) & 0x7FFF);
+        } else {
+          // Not aligned, need nop + adrp + add
          nop();
          adrp(DestReg, static_cast<uint32_t>(ImmInstTwo >> 12) & 0x7FFF);
-        } else {
-          // Not aligned, need adrp + add
-          adrp(DestReg, static_cast<uint32_t>(ImmInstOne >> 12) & 0x7FFF);
-          add(ARMEmitter::Size::i64Bit, DestReg, DestReg, ImmInstOne & 0xFFF);
+          add(ARMEmitter::Size::i64Bit, DestReg, DestReg, ImmInstTwo & 0xFFF);
        }
      } else {
-        LOGMAN_MSG_A_FMT("Unscaled offset is too large");
-        FEX_UNREACHABLE;
+        // Stinky path, we need to emit a movz+movk+movk sequence.
+        movz(ARMEmitter::Size::i64Bit, DestReg, uint32_t(ImmInstOne >> 32) & 0x7FFF, 32);
+        movk(ARMEmitter::Size::i64Bit, DestReg, uint32_t(ImmInstOne >> 16) & 0xFFFF, 16);
+        movk(ARMEmitter::Size::i64Bit, DestReg, uint32_t(ImmInstOne) & 0xFFFF);
      }

      SetCursorOffset(OriginalOffset);
@@ -745,33 +786,52 @@ public:
    }
    default: LOGMAN_MSG_A_FMT("Unexpected inst type in label fixup");
    }
+
+    return true;
  }

  // Bind a forward label to a location.
  // This walks all the instructions in the label's vector.
  // Then backpatching all instructions that have used the label.
-  template<bool WarnAboutEmpty = false>
-  void Bind(ForwardLabel* Label) {
-    if constexpr (WarnAboutEmpty) {
-      LOGMAN_THROW_A_FMT(Label->Insts.empty() == false, "Binding forward label that didn't have any instructions using it");
+  [[nodiscard]] bool Bind(ForwardLabel* Label) {
+    bool Bound = true;
+    if (Label->FirstInst.Location) {
+      Bound &= Bind(&Label->FirstInst);
    }
    for (auto& Inst : Label->Insts) {
-      Bind(&Inst);
+      Bound &= Bind(&Inst);
    }
+
+    return Bound;
  }

  // Bind a bidirectional location to a location.
  // Binds both forwards and backwards depending on how the label was used.
-  void Bind(BiDirectionalLabel* Label) {
+  [[nodiscard]] bool Bind(BiDirectionalLabel* Label) {
+    bool Bound = true;
    if (!Label->Backward.Location) {
-      Bind(&Label->Backward);
+      Bound &= Bind(&Label->Backward);
    }
-    Bind<false>(&Label->Forward);
+    Bound &= Bind(&Label->Forward);
+
+    return Bound;
+  }
+
+  static constexpr Condition InvertCondition(Condition cond) {
+    // These behave as always, so it makes no sense to allow inverting these.
+    LOGMAN_THROW_A_FMT(cond != Condition::CC_AL && cond != Condition::CC_NV, "Cannot invert CC_AL or CC_NV");
+    return static_cast<Condition>(FEXCore::ToUnderlying(cond) ^ 1);
  }

 #include <CodeEmitter/VixlUtils.inl>

 public:
+
+// This symbol is used to allow external tooling (IDEs, clang-format, ...) to process the included files individually:
+// If defined, the files will inject member functions into this class.
+// If not, the files will wrap the member functions in a class so that tooling will process them properly.
+#define INCLUDED_BY_EMITTER
+
  // TODO: Implement SME when it matters.
 #include <CodeEmitter/ALUOps.inl>
 #include <CodeEmitter/BranchOps.inl>
@@ -781,7 +841,9 @@ public:
 #include <CodeEmitter/ASIMDOps.inl>
 #include <CodeEmitter/SVEOps.inl>

-private:
+#undef INCLUDED_BY_EMITTER
+
+protected:
  template<typename T>
  uint32_t Encode_ra(T Reg) const {
    return Reg.Idx() << 10;
@@ -793,7 +855,6 @@ private:
  uint32_t Encode_rt2(T Reg) const {
    return Reg.Idx() << 10;
  }
-  template<>
  uint32_t Encode_rt2(uint32_t Reg) const {
    return Reg << 10;
  }
@@ -829,7 +890,6 @@ private:
  uint32_t Encode_rt(T Reg) const {
    return Reg.Idx();
  }
-  template<>
  uint32_t Encode_rt(Prefetch Reg) const {
    return FEXCore::ToUnderlying(Reg);
  }
@@ -30,9 +30,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(Register) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<Register>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<Register>, "Needs to be standard");
+static_assert(sizeof(Register) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<Register>);
+static_assert(std::is_standard_layout_v<Register>);

 /* 32-bit GPR register class.
 * This class will imply a 32-bit register size being used.
@@ -58,9 +58,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(Register) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<Register>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<Register>, "Needs to be standard");
+static_assert(sizeof(WRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<WRegister>);
+static_assert(std::is_standard_layout_v<WRegister>);

 /* 64-bit GPR register class.
 * This class will imply a 64-bit register size being used.
@@ -86,9 +86,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(Register) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<Register>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<Register>, "Needs to be standard");
+static_assert(sizeof(XRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<XRegister>);
+static_assert(std::is_standard_layout_v<XRegister>);

 inline constexpr WRegister Register::W() const {
  return WRegister {Index};
@@ -283,9 +283,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(VRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<VRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<VRegister>, "Needs to be standard");
+static_assert(sizeof(VRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<VRegister>);
+static_assert(std::is_standard_layout_v<VRegister>);

 /* 8-bit ASIMD register class
 * This class implies 8-bit scalar register.
@@ -315,9 +315,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(BRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<BRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<BRegister>, "Needs to be standard");
+static_assert(sizeof(BRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<BRegister>);
+static_assert(std::is_standard_layout_v<BRegister>);

 /* 16-bit ASIMD register class
 * This class implies 16-bit scalar register.
@@ -347,9 +347,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(HRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<HRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<HRegister>, "Needs to be standard");
+static_assert(sizeof(HRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<HRegister>);
+static_assert(std::is_standard_layout_v<HRegister>);

 /* 32-bit ASIMD register class
 * This class implies 32-bit scalar register.
@@ -379,9 +379,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(SRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<SRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<SRegister>, "Needs to be standard");
+static_assert(sizeof(SRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<SRegister>);
+static_assert(std::is_standard_layout_v<SRegister>);

 /* 64-bit ASIMD register class
 * This class doesn't imply Vector or Scalar.
@@ -412,9 +412,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(DRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<DRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<DRegister>, "Needs to be standard");
+static_assert(sizeof(DRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<DRegister>);
+static_assert(std::is_standard_layout_v<DRegister>);

 /* 128-bit ASIMD register class
 * This class doesn't imply Vector or Scalar.
@@ -445,9 +445,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(QRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<QRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<QRegister>, "Needs to be standard");
+static_assert(sizeof(QRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<QRegister>);
+static_assert(std::is_standard_layout_v<QRegister>);

 /* Unsized SVE register class.
 * This class explicitly implies the instruction will operate using SVE.
@@ -474,9 +474,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(ZRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<ZRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<ZRegister>, "Needs to be standard");
+static_assert(sizeof(ZRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<ZRegister>);
+static_assert(std::is_standard_layout_v<ZRegister>);

 // VRegister
 inline constexpr BRegister VRegister::B() const {
@@ -919,9 +919,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(PRegister) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<PRegister>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<PRegister>, "Needs to be standard");
+static_assert(sizeof(PRegister) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<PRegister>);
+static_assert(std::is_standard_layout_v<PRegister>);

 // Unsized predicate register for SVE with zeroing semantics.
 class PRegisterZero {
@@ -947,9 +947,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(PRegisterZero) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<PRegisterZero>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<PRegisterZero>, "Needs to be standard");
+static_assert(sizeof(PRegisterZero) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<PRegisterZero>);
+static_assert(std::is_standard_layout_v<PRegisterZero>);

 // Unsized predicate register for SVE with merging semantics.
 class PRegisterMerge {
@@ -975,9 +975,9 @@ public:
 private:
  uint32_t Index;
 };
-static_assert(sizeof(PRegisterZero) == sizeof(uint32_t), "Needs to be uint32_t");
-static_assert(std::is_trivial_v<PRegisterZero>, "Needs to be trivial");
-static_assert(std::is_standard_layout_v<PRegisterZero>, "Needs to be standard");
+static_assert(sizeof(PRegisterMerge) == sizeof(uint32_t));
+static_assert(std::is_trivially_copyable_v<PRegisterMerge>);
+static_assert(std::is_standard_layout_v<PRegisterMerge>);

 // PRegister
 inline constexpr PRegisterZero PRegister::Zeroing() const {
@@ -16,28 +16,34 @@
 * Exceptions to this rule will have asserts in the emitter implementation when misused.
 *
 */
-public:
-// Advanced SIMD scalar copy
-  void dup(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Index) {
-    constexpr uint32_t Op = 0b0101'1110'0000'0000'0000'01 << 10;

+#pragma once
+#ifndef INCLUDED_BY_EMITTER
+#include <CodeEmitter/Emitter.h>
+namespace ARMEmitter {
+struct EmitterOps : Emitter {
+#endif
+
+public:
+  // Advanced SIMD scalar copy
+  void dup(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Index) {
    const uint32_t SizeImm = FEXCore::ToUnderlying(size);
    const uint32_t IndexShift = SizeImm + 1;
    const uint32_t ElementSize = 1U << SizeImm;
    const uint32_t MaxIndex = 128U / (ElementSize * 8);

-    LOGMAN_THROW_AA_FMT(Index < MaxIndex, "Index too large. Index={}, Max Index: {}", Index, MaxIndex);
+    LOGMAN_THROW_A_FMT(Index < MaxIndex, "Index too large. Index={}, Max Index: {}", Index, MaxIndex);

    const uint32_t imm5 = (Index << IndexShift) | ElementSize;

-    ASIMDScalarCopy(Op, 1, imm5, 0b0000, rd, rn);
+    ASIMDScalarCopy(1, 1, imm5, 0b0000, rd, rn);
  }

-  void mov(ARMEmitter::ScalarRegSize size, ARMEmitter::VRegister rd, ARMEmitter::VRegister rn, uint32_t Index) {
+  void mov(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Index) {
    dup(size, rd, rn, Index);
  }

-// Advanced SIMD scalar three same FP16
+  // Advanced SIMD scalar three same FP16
  void fmulx(HRegister rd, HRegister rn, HRegister rm) {
    ASIMDScalarThreeSameFP16(0, 0, 0b011, rm, rn, rd);
  }
@@ -66,7 +72,7 @@ public:
    ASIMDScalarThreeSameFP16(1, 1, 0b101, rm, rn, rd);
  }

-// Advanced SIMD scalar two-register miscellaneous FP16
+  // Advanced SIMD scalar two-register miscellaneous FP16
  void fcvtns(HRegister rd, HRegister rn) {
    ASIMDScalarTwoRegMiscFP16(0, 0, 0b11010, rn, rd);
  }
@@ -128,9 +134,17 @@ public:
    ASIMDScalarTwoRegMiscFP16(1, 1, 0b11101, rn, rd);
  }

-// Advanced SIMD scalar three same extra
-// XXX:
-// Advanced SIMD scalar two-register miscellaneous
+  // Advanced SIMD scalar three same extra
+  void sqrdmlah(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i32Bit, "Only supports 16/32-bit");
+    ASIMDScalarThreeSameExtra(1, size, 0b0000, rm, rn, rd);
+  }
+  void sqrdmlsh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i32Bit, "Only supports 16/32-bit");
+    ASIMDScalarThreeSameExtra(1, size, 0b0001, rm, rn, rd);
+  }
+
+  // Advanced SIMD scalar two-register miscellaneous
  void suqadd(ScalarRegSize size, VRegister rd, VRegister rn) {
    ASIMDScalar2RegMisc(0, 0, size, 0b00011, rd, rn);
  }
@@ -140,67 +154,55 @@ public:

  ///< Comparison against 0.0
  void cmgt(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 0, size, 0b01000, rd, rn);
  }
  ///< Comparison against 0.0
  void cmeq(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 0, size, 0b01001, rd, rn);
  }

  ///< Comparison against 0.0
  void cmlt(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 0, size, 0b01010, rd, rn);
  }
  void abs(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 0, size, 0b01011, rd, rn);
  }
  ///< size is destination size.
  void sqxtn(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
    ASIMDScalar2RegMisc(0, 0, size, 0b10100, rd, rn);
  }

  void fcvtns(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 0, ConvertedSize, 0b11010, rd, rn);
  }
  void fcvtms(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 0, ConvertedSize, 0b11011, rd, rn);
  }
  void fcvtas(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 0, ConvertedSize, 0b11100, rd, rn);
  }
  void scvtf(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 0, ConvertedSize, 0b11101, rd, rn);
  }
@@ -249,70 +251,58 @@ public:
  }
  ///< Comparison against 0.0
  void cmge(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 1, size, 0b01000, rd, rn);
  }
  ///< Comparison against 0.0
  void cmle(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 1, size, 0b01001, rd, rn);
  }
  void neg(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMDScalar2RegMisc(0, 1, size, 0b01011, rd, rn);
  }
  ///< size is destination.
  void sqxtun(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
    ASIMDScalar2RegMisc(0, 1, size, 0b10010, rd, rn);
  }
  ///< size is destination.
  void uqxtn(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "64-bit destination not supported");
    ASIMDScalar2RegMisc(0, 1, size, 0b10100, rd, rn);
  }
  ///< size is destination.
  void fcvtxn(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMDScalar2RegMisc(0, 1, ScalarRegSize::i16Bit, 0b10110, rd, rn);
  }
  void fcvtnu(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 1, ConvertedSize, 0b11010, rd, rn);
  }
  void fcvtmu(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 1, ConvertedSize, 0b11011, rd, rn);
  }
  void fcvtau(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 1, ConvertedSize, 0b11100, rd, rn);
  }
  void ucvtf(ScalarRegSize size, VRegister rd, VRegister rn) {
    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(0, 1, ConvertedSize, 0b11101, rd, rn);
  }
@@ -366,73 +356,55 @@ public:
  }

  void fmaxnmp(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(1, 1, ConvertedSize, 0b01100, rd, rn);
  }
  void faddp(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(1, 1, ConvertedSize, 0b01101, rd, rn);
  }
  void fmaxp(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMDScalar2RegMisc(1, 1, ConvertedSize, 0b01111, rd, rn);
  }
  void fminnmp(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMDScalar2RegMisc(1, 1, size, 0b01100, rd, rn);
  }
  void fminp(ScalarRegSize size, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMDScalar2RegMisc(1, 1, size, 0b01111, rd, rn);
  }
-// Advanced SIMD scalar three different
+  // Advanced SIMD scalar three different
  ///< size is destination.
  void sqdmlal(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i32Bit :
-        ScalarRegSize::i16Bit;
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i32Bit : ScalarRegSize::i16Bit;
    ASIMD3RegDifferent(0, ConvertedSize, 0b1001, rd, rn, rm);
  }
  ///< size is destination.
  void sqdmlsl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i32Bit :
-        ScalarRegSize::i16Bit;
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i32Bit : ScalarRegSize::i16Bit;
    ASIMD3RegDifferent(0, ConvertedSize, 0b1011, rd, rn, rm);
  }

  ///< size is destination.
  void sqdmull(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i32Bit :
-        ScalarRegSize::i16Bit;
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i32Bit : ScalarRegSize::i16Bit;
    ASIMD3RegDifferent(0, ConvertedSize, 0b1101, rd, rn, rm);
  }
-// Advanced SIMD scalar three same
+  // Advanced SIMD scalar three same
  void sqadd(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    ASIMD3RegSame(0, size, 0b00001, rd, rn, rm);
  }
@@ -440,71 +412,62 @@ public:
    ASIMD3RegSame(0, size, 0b00101, rd, rn, rm);
  }
  void cmgt(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b00110, rd, rn, rm);
  }
  void cmge(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b00111, rd, rn, rm);
  }
  void sshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b01000, rd, rn, rm);
  }
  void sqshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    ASIMD3RegSame(0, size, 0b01001, rd, rn, rm);
  }
  void srshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b01010, rd, rn, rm);
  }
  void sqrshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    ASIMD3RegSame(0, size, 0b01011, rd, rn, rm);
  }
  void add(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b10000, rd, rn, rm);
  }
  void cmtst(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(0, size, 0b10001, rd, rn, rm);
  }
  void sqdmulh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i32Bit || size == ScalarRegSize::i16Bit, "Invalid size");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i32Bit || size == ScalarRegSize::i16Bit, "Invalid size");
    ASIMD3RegSame(0, size, 0b10110, rd, rn, rm);
  }
  void fmulx(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMD3RegSame(0, ConvertedSize, 0b11011, rd, rn, rm);
  }
  void fcmeq(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMD3RegSame(0, ConvertedSize, 0b11100, rd, rn, rm);
  }
  void frecps(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMD3RegSame(0, ConvertedSize, 0b11111, rd, rn, rm);
  }
  void frsqrts(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMD3RegSame(0, size, 0b11111, rd, rn, rm);
  }
  void uqadd(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
@@ -514,75 +477,69 @@ public:
    ASIMD3RegSame(1, size, 0b00101, rd, rn, rm);
  }
  void cmhi(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b00110, rd, rn, rm);
  }
  void cmhs(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b00111, rd, rn, rm);
  }
  void ushl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b01000, rd, rn, rm);
  }
  void uqshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    ASIMD3RegSame(1, size, 0b01001, rd, rn, rm);
  }
  void urshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b01010, rd, rn, rm);
  }
  void uqrshl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    ASIMD3RegSame(1, size, 0b01011, rd, rn, rm);
  }
  void sub(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b10000, rd, rn, rm);
  }
  void cmeq(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit, "Only supports 64-bit");
    ASIMD3RegSame(1, size, 0b10001, rd, rn, rm);
  }
  void sqrdmulh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i32Bit || size == ScalarRegSize::i16Bit, "Invalid size");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i32Bit || size == ScalarRegSize::i16Bit, "Invalid size");
    ASIMD3RegSame(1, size, 0b10110, rd, rn, rm);
  }
  void fcmge(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMD3RegSame(1, ConvertedSize, 0b11100, rd, rn, rm);
  }
  void facge(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");

-    const ScalarRegSize ConvertedSize =
-      size == ScalarRegSize::i64Bit ?
-        ScalarRegSize::i16Bit :
-        ScalarRegSize::i8Bit;
+    const ScalarRegSize ConvertedSize = size == ScalarRegSize::i64Bit ? ScalarRegSize::i16Bit : ScalarRegSize::i8Bit;

    ASIMD3RegSame(1, ConvertedSize, 0b11101, rd, rn, rm);
  }
  void fabd(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMD3RegSame(1, size, 0b11010, rd, rn, rm);
  }
  void fcmgt(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMD3RegSame(1, size, 0b11100, rd, rn, rm);
  }
  void facgt(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for float convert");
    ASIMD3RegSame(1, size, 0b11101, rd, rn, rm);
  }
-// Advanced SIMD scalar shift by immediate
+  // Advanced SIMD scalar shift by immediate
  void sshr(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -592,8 +549,8 @@ public:
    ASIMDScalarShiftByImm(0, immh, immb, 0b00000, rd, rn);
  }
  void ssra(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -603,8 +560,8 @@ public:
    ASIMDScalarShiftByImm(0, immh, immb, 0b00010, rd, rn);
  }
  void srshr(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -614,8 +571,8 @@ public:
    ASIMDScalarShiftByImm(0, immh, immb, 0b00100, rd, rn);
  }
  void srsra(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -625,8 +582,8 @@ public:
    ASIMDScalarShiftByImm(0, immh, immb, 0b00110, rd, rn);
  }
  void shl(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    // Shift encoded a bit weirdly.
    // shift = immh:immb - elementsize but immh is /also/ used for element size.
    const uint32_t immh = 1 << FEXCore::ToUnderlying(size) | (Shift >> 3);
@@ -644,7 +601,7 @@ public:
  ///< size is destination
  void sqshrn(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrn");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrn");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -655,7 +612,7 @@ public:
  }
  void sqrshrn(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrn");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrn");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -666,8 +623,8 @@ public:
  }
  // TODO: SCVTF, FCVTZS
  void ushr(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -677,8 +634,8 @@ public:
    ASIMDScalarShiftByImm(1, immh, immb, 0b00000, rd, rn);
  }
  void usra(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -688,8 +645,8 @@ public:
    ASIMDScalarShiftByImm(1, immh, immb, 0b00010, rd, rn);
  }
  void urshr(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -699,8 +656,8 @@ public:
    ASIMDScalarShiftByImm(1, immh, immb, 0b00100, rd, rn);
  }
  void ursra(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -710,8 +667,8 @@ public:
    ASIMDScalarShiftByImm(1, immh, immb, 0b00110, rd, rn);
  }
  void sri(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -721,8 +678,8 @@ public:
    ASIMDScalarShiftByImm(1, immh, immb, 0b01000, rd, rn);
  }
  void sli(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
-    LOGMAN_THROW_AA_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
+    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < 64, "Invalid shift for sshr");
+    LOGMAN_THROW_A_FMT(size == ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sshr");
    // Shift encoded a bit weirdly.
    // shift = immh:immb - elementsize but immh is /also/ used for element size.
    const uint32_t immh = 1 << FEXCore::ToUnderlying(size) | (Shift >> 3);
@@ -748,7 +705,7 @@ public:
  ///< size is destination.
  void sqshrun(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrun");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqshrun");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -760,7 +717,7 @@ public:
  ///< size is destination.
  void sqrshrun(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -772,7 +729,7 @@ public:
  ///< size is destination.
  void uqshrn(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -784,7 +741,7 @@ public:
  ///< size is destination.
  void uqrshrn(ScalarRegSize size, VRegister rd, VRegister rn, uint32_t Shift) {
    LOGMAN_THROW_A_FMT(Shift > 0 && Shift < ScalarRegSizeInBits(size), "Invalid shift for sshr");
-    LOGMAN_THROW_AA_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
+    LOGMAN_THROW_A_FMT(size != ARMEmitter::ScalarRegSize::i64Bit, "Invalid size selected for sqrshrun");
    const size_t SubregSizeInBits = ScalarRegSizeInBits(size);
    // Shift encoded in immh:immb, but inverted with 128-bit source
    // shift = (esize * 2) - immh:immb
@@ -793,11 +750,52 @@ public:
    const uint32_t immb = InvertedShift & 0b111;
    ASIMDScalarShiftByImm(1, immh, immb, 0b10011, rd, rn);
  }
+
  // TODO: UCVTF, FCVTZU
-// Advanced SIMD scalar x indexed element
-// XXX:
-//
-// Floating-point data-processing (1 source)
+
+  // Advanced SIMD scalar x indexed element
+  void sqdmlal(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(0, size, 0b0011, rm, rn, rd, index);
+  }
+  void sqdmlsl(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(0, size, 0b0111, rm, rn, rd, index);
+  }
+  void sqdmull(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(0, size, 0b1011, rm, rn, rd, index);
+  }
+  void sqdmulh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(0, size, 0b1100, rm, rn, rd, index);
+  }
+  void sqrdmulh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(0, size, 0b1101, rm, rn, rd, index);
+  }
+  void fmla(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    ASIMDScalarXIndexedElement(0, size, 0b0001, rm, rn, rd, index);
+  }
+  void fmls(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    ASIMDScalarXIndexedElement(0, size, 0b0101, rm, rn, rd, index);
+  }
+  void fmul(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    ASIMDScalarXIndexedElement(0, size, 0b1001, rm, rn, rd, index);
+  }
+  void sqrdmlah(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(1, size, 0b1101, rm, rn, rd, index);
+  }
+  void sqrdmlsh(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i64Bit, "Scalar size must not be 64-bit");
+    ASIMDScalarXIndexedElement(1, size, 0b1111, rm, rn, rd, index);
+  }
+  void fmulx(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, uint32_t index) {
+    ASIMDScalarXIndexedElement(1, size, 0b1001, rm, rn, rd, index);
+  }
+
+  // Floating-point data-processing (1 source)
  void fmov(ScalarRegSize size, VRegister rd, VRegister rn) {
    Float1Source(size, 0, 0, 0b000000, rd, rn);
  }
@@ -991,14 +989,14 @@ public:
    Float1Source(0, 0, 0b11, 0b001111, rd.V(), rn.V());
  }

-// Floating-point compare
+  // Floating-point compare
  void fcmp(ScalarRegSize Size, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(Size != ScalarRegSize::i8Bit, "8-bit destination not supported");
+    LOGMAN_THROW_A_FMT(Size != ScalarRegSize::i8Bit, "8-bit destination not supported");

-    const auto ConvertedSize =
-      Size == ARMEmitter::ScalarRegSize::i64Bit ? 0b01 :
-      Size == ARMEmitter::ScalarRegSize::i32Bit ? 0b00 :
-      Size == ARMEmitter::ScalarRegSize::i16Bit ? 0b11 : 0;
+    const auto ConvertedSize = Size == ARMEmitter::ScalarRegSize::i64Bit ? 0b01 :
+                               Size == ARMEmitter::ScalarRegSize::i32Bit ? 0b00 :
+                               Size == ARMEmitter::ScalarRegSize::i16Bit ? 0b11 :
+                                                                           0;

    FloatCompare(0, 0, ConvertedSize, 0b00, 0b00000, rn, rm);
  }
@@ -1051,7 +1049,7 @@ public:
    FloatCompare(0, 0, 0b11, 0b00, 0b11000, rn.V(), VReg::v0);
  }

-// Floating-point immediate
+  // Floating-point immediate
  void fmov(ARMEmitter::ScalarRegSize size, ARMEmitter::VRegister rd, float Value) {
    uint32_t M = 0;
    uint32_t S = 0;
@@ -1061,16 +1059,13 @@ public:
    if (size == ARMEmitter::ScalarRegSize::i16Bit) {
      LOGMAN_MSG_A_FMT("Unsupported");
      FEX_UNREACHABLE;
-    }
-    else if (size == ARMEmitter::ScalarRegSize::i32Bit) {
+    } else if (size == ARMEmitter::ScalarRegSize::i32Bit) {
      ptype = 0b00;
      imm8 = FP32ToImm8(Value);
-    }
-    else if (size == ARMEmitter::ScalarRegSize::i64Bit) {
+    } else if (size == ARMEmitter::ScalarRegSize::i64Bit) {
      ptype = 0b01;
      imm8 = FP64ToImm8(Value);
-    }
-    else {
+    } else {
      FEX_UNREACHABLE;
    }

@@ -1090,7 +1085,7 @@ public:
    dc32(Instr);
  }

-// Floating-point conditional compare
+  // Floating-point conditional compare
  void fccmp(SRegister rn, SRegister rm, StatusFlags flags, Condition Cond) {
    FloatConditionalCompare(0, 0, 0b00, 0b0, rn.V(), rm.V(), flags, Cond);
  }
@@ -1110,7 +1105,7 @@ public:
    FloatConditionalCompare(0, 0, 0b11, 0b1, rn.V(), rm.V(), flags, Cond);
  }

-// Floating-point data-processing (2 source)
+  // Floating-point data-processing (2 source)
  void fmul(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm) {
    Float2Source(size, 0, 0, 0b0000, rd, rn, rm);
  }
@@ -1225,11 +1220,10 @@ public:

  // Floating-point conditional select
  void fcsel(ScalarRegSize size, VRegister rd, VRegister rn, VRegister rm, Condition Cond) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for {}", __func__);
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit,
+                       "Invalid size selected for {}", __func__);

-    const uint32_t ConvertedSize =
-      size == ScalarRegSize::i64Bit ? 0b01 :
-      size == ScalarRegSize::i32Bit ? 0b00 : 0b11;
+    const uint32_t ConvertedSize = size == ScalarRegSize::i64Bit ? 0b01 : size == ScalarRegSize::i32Bit ? 0b00 : 0b11;

    FloatConditionalSelect(0, 0, ConvertedSize, rd, rn, rm, Cond);
  }
@@ -1244,7 +1238,7 @@ public:
    FloatConditionalSelect(0, 0, 0b11, rd.V(), rn.V(), rm.V(), Cond);
  }

-// Floating-point data-processing (3 source)
+  // Floating-point data-processing (3 source)
  void fmadd(SRegister rd, SRegister rn, SRegister rm, SRegister ra) {
    Float3Source(0, 0, 0b00, 0, 0, rd.V(), rn.V(), rm.V(), ra.V());
  }
@@ -1285,11 +1279,11 @@ public:
  }

 private:
-// Advanced SIMD scalar copy
-  void ASIMDScalarCopy(uint32_t Op, uint32_t Q, uint32_t imm5, uint32_t imm4, ARMEmitter::VRegister rd, ARMEmitter::VRegister rn) {
-    uint32_t Instr = Op;
-
+  // Advanced SIMD scalar copy
+  void ASIMDScalarCopy(uint32_t Q, uint32_t b28, uint32_t imm5, uint32_t imm4, VRegister rd, VRegister rn) {
+    uint32_t Instr = 0b0000'1110'0000'0000'0000'01U << 10;
    Instr |= Q << 30;
+    Instr |= b28 << 28;
    Instr |= imm5 << 16;
    Instr |= imm4 << 11;
    Instr |= Encode_rn(rn);
@@ -1297,7 +1291,7 @@ private:
    dc32(Instr);
  }

-// Advanced SIMD scalar three same FP16
+  // Advanced SIMD scalar three same FP16
  void ASIMDScalarThreeSameFP16(uint32_t U, uint32_t a, uint32_t opcode, HRegister rm, HRegister rn, HRegister rd) {
    uint32_t Instr = 0b0101'1110'0100'0000'0000'0100'0000'0000;

@@ -1309,7 +1303,7 @@ private:
    Instr |= rd.Idx();
    dc32(Instr);
  }
-// Advanced SIMD scalar two-register miscellaneous FP16
+  // Advanced SIMD scalar two-register miscellaneous FP16
  void ASIMDScalarTwoRegMiscFP16(uint32_t U, uint32_t a, uint32_t opcode, HRegister rn, HRegister rd) {
    uint32_t Instr = 0b0101'1110'0111'1000'0000'1000'0000'0000;

@@ -1321,9 +1315,19 @@ private:
    dc32(Instr);
  }

-// Advanced SIMD scalar three same extra
-// XXX:
-// Advanced SIMD scalar two-register miscellaneous
+  // Advanced SIMD scalar three same extra
+  void ASIMDScalarThreeSameExtra(uint32_t U, ScalarRegSize size, uint32_t opcode, VRegister rm, VRegister rn, VRegister rd) {
+    uint32_t Instr = 0b0101'1110'0000'0000'1000'0100'0000'0000;
+    Instr |= U << 29;
+    Instr |= FEXCore::ToUnderlying(size) << 22;
+    Instr |= rm.Idx() << 16;
+    Instr |= opcode << 11;
+    Instr |= rn.Idx() << 5;
+    Instr |= rd.Idx();
+    dc32(Instr);
+  }
+
+  // Advanced SIMD scalar two-register miscellaneous
  void ASIMDScalar2RegMisc(uint32_t b20, uint32_t U, ScalarRegSize size, uint32_t opcode, VRegister rd, VRegister rn) {
    uint32_t Instr = 0b0101'1110'0010'0000'0000'1000'0000'0000;

@@ -1336,9 +1340,7 @@ private:
    dc32(Instr);
  }

-// Advanced SIMD scalar pairwise
-// XXX:
-// Advanced SIMD scalar three different
+  // Advanced SIMD scalar three different
  void ASIMD3RegDifferent(uint32_t U, ScalarRegSize size, uint32_t opcode, VRegister rd, VRegister rn, VRegister rm) {
    uint32_t Instr = 0b0101'1110'0010'0000'0000'0000'0000'0000;

@@ -1350,7 +1352,7 @@ private:
    Instr |= Encode_rd(rd);
    dc32(Instr);
  }
-// Advanced SIMD scalar three same
+  // Advanced SIMD scalar three same
  void ASIMD3RegSame(uint32_t U, ScalarRegSize size, uint32_t opcode, VRegister rd, VRegister rn, VRegister rm) {
    uint32_t Instr = 0b0101'1110'0010'0000'0000'0100'0000'0000;

@@ -1362,7 +1364,7 @@ private:
    Instr |= Encode_rd(rd);
    dc32(Instr);
  }
-// Advanced SIMD scalar shift by immediate
+  // Advanced SIMD scalar shift by immediate
  void ASIMDScalarShiftByImm(uint32_t U, uint32_t immh, uint32_t immb, uint32_t opcode, VRegister rd, VRegister rn) {
    uint32_t Instr = 0b0101'1111'0000'0000'0000'0100'0000'0000;

@@ -1374,9 +1376,51 @@ private:
    Instr |= Encode_rd(rd);
    dc32(Instr);
  }
-// Advanced SIMD scalar x indexed element
-// XXX:
-// Floating-point data-processing (1 source)
+
+  // Advanced SIMD scalar x indexed element
+  void ASIMDScalarXIndexedElement(uint32_t U, ScalarRegSize size, uint32_t opcode, VRegister rm, VRegister rn, VRegister rd, uint32_t index) {
+    LOGMAN_THROW_A_FMT(size != ScalarRegSize::i8Bit, "Scalar size must not be 8-bit");
+
+    const auto invalid_bound = 16U >> FEXCore::ToUnderlying(size);
+    LOGMAN_THROW_A_FMT(index < invalid_bound, "Index ({}) must be within [0-{}]", index, invalid_bound - 1);
+
+    uint32_t Instr = 0b0101'1111'0000'0000'0000'0000'0000'0000;
+
+    // FMUL/FMLA/FMLS indexed variants deal with size differently.
+    if (opcode == 0b0001 || opcode == 0b0101 || opcode == 0b1001) {
+      // Unlike other instructions in the group, 16-bit is encoded as zero
+      // and 32/64-bit are encoded with the top bit always set to one.
+      if (size != ScalarRegSize::i16Bit) {
+        Instr |= (0b10 | (FEXCore::ToUnderlying(size) & 1)) << 22;
+      }
+    } else {
+      Instr |= FEXCore::ToUnderlying(size) << 22;
+    }
+
+    uint32_t H = 0;
+    uint32_t LM = 0;
+    if (size == ScalarRegSize::i16Bit) {
+      LOGMAN_THROW_A_FMT(rm <= VReg::v15, "rm ({}) must be within [v0-v15]", rm.Idx());
+      H = (index >> 2) & 1;
+      LM = index & 0b11;
+    } else if (size == ScalarRegSize::i32Bit) {
+      H = (index >> 1) & 1;
+      LM = (index & 0b01) << 1;
+    } else {
+      H = index & 1;
+    }
+
+    Instr |= U << 29;
+    Instr |= LM << 20;
+    Instr |= rm.Idx() << 16;
+    Instr |= opcode << 12;
+    Instr |= H << 11;
+    Instr |= rn.Idx() << 5;
+    Instr |= rd.Idx();
+    dc32(Instr);
+  }
+
+  // Floating-point data-processing (1 source)
  void Float1Source(uint32_t M, uint32_t S, uint32_t ptype, uint32_t opcode, VRegister rd, VRegister rn) {
    uint32_t Instr = 0b0001'1110'0010'0000'0100'0000'0000'0000;

@@ -1390,16 +1434,15 @@ private:
    dc32(Instr);
  }
  void Float1Source(ScalarRegSize size, uint32_t M, uint32_t S, uint32_t opcode, VRegister rd, VRegister rn) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for {}", __func__);
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit,
+                       "Invalid size selected for {}", __func__);

-    const uint32_t ConvertedSize =
-      size == ScalarRegSize::i64Bit ? 0b01 :
-      size == ScalarRegSize::i32Bit ? 0b00 : 0b11;
+    const uint32_t ConvertedSize = size == ScalarRegSize::i64Bit ? 0b01 : size == ScalarRegSize::i32Bit ? 0b00 : 0b11;

    Float1Source(M, S, ConvertedSize, opcode, rd, rn);
  }

-// Floating-point compare
+  // Floating-point compare
  void FloatCompare(uint32_t M, uint32_t S, uint32_t ftype, uint32_t op, uint32_t opcode2, VRegister rn, VRegister rm) {
    uint32_t Instr = 0b0001'1110'0010'0000'0010'0000'0000'0000;

@@ -1413,9 +1456,9 @@ private:

    dc32(Instr);
  }
-// Floating-point immediate
-// XXX:
-// Floating-point conditional compare
+  // Floating-point immediate
+  // XXX:
+  // Floating-point conditional compare
  void FloatConditionalCompare(uint32_t M, uint32_t S, uint32_t ptype, uint32_t op, VRegister rn, VRegister rm, StatusFlags flags, Condition Cond) {
    uint32_t Instr = 0b0001'1110'0010'0000'0000'0100'0000'0000;

@@ -1430,7 +1473,7 @@ private:

    dc32(Instr);
  }
-// Floating-point data-processing (2 source)
+  // Floating-point data-processing (2 source)

  void Float2Source(uint32_t M, uint32_t S, uint32_t ptype, uint32_t opcode, VRegister rd, VRegister rn, VRegister rm) {
    uint32_t Instr = 0b0001'1110'0010'0000'0000'1000'0000'0000;
@@ -1447,16 +1490,15 @@ private:
  }

  void Float2Source(ScalarRegSize size, uint32_t M, uint32_t S, uint32_t opcode, VRegister rd, VRegister rn, VRegister rm) {
-    LOGMAN_THROW_AA_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit, "Invalid size selected for {}", __func__);
+    LOGMAN_THROW_A_FMT(size == ScalarRegSize::i16Bit || size == ScalarRegSize::i64Bit || size == ScalarRegSize::i32Bit,
+                       "Invalid size selected for {}", __func__);

-    const uint32_t ConvertedSize =
-      size == ScalarRegSize::i64Bit ? 0b01 :
-      size == ScalarRegSize::i32Bit ? 0b00 : 0b11;
+    const uint32_t ConvertedSize = size == ScalarRegSize::i64Bit ? 0b01 : size == ScalarRegSize::i32Bit ? 0b00 : 0b11;

    Float2Source(M, S, ConvertedSize, opcode, rd, rn, rm);
  }

-// Floating-point conditional select
+  // Floating-point conditional select
  void FloatConditionalSelect(uint32_t M, uint32_t S, uint32_t ptype, VRegister rd, VRegister rn, VRegister rm, Condition Cond) {
    uint32_t Instr = 0b0001'1110'0010'0000'0000'1100'0000'0000;

@@ -1470,7 +1512,7 @@ private:
    dc32(Instr);
  }

-// Floating-point data-processing (3 source)
+  // Floating-point data-processing (3 source)
  void Float3Source(uint32_t M, uint32_t S, uint32_t ptype, uint32_t o1, uint32_t o0, VRegister rd, VRegister rn, VRegister rm, VRegister ra) {
    uint32_t Instr = 0b0001'1111'0000'0000'0000'0000'0000'0000;

@@ -1485,3 +1527,8 @@ private:
    Instr |= Encode_rd(rd);
    dc32(Instr);
  }
+
+#ifndef INCLUDED_BY_EMITTER
+}; // struct LoadstoreEmitterOps
+} // namespace ARMEmitter
+#endif
@@ -4,173 +4,191 @@
 * This is mostly a mashup of various instruction types.
 * Nothing follows an explicit pattern since they are mostly different.
 */
+
+#pragma once
+#ifndef INCLUDED_BY_EMITTER
+#include <CodeEmitter/Emitter.h>
+namespace ARMEmitter {
+struct EmitterOps : Emitter {
+#endif
+
 public:
-    // System with result
-    // TODO: SYSL
-    // System Instruction
-    // TODO: AT
-    // TODO: CFP
-    // TODO: CPP
-    void dc(ARMEmitter::DataCacheOperation DCOp, ARMEmitter::Register rt) {
-      constexpr uint32_t Op = 0b1101'0101'0000'1000'0111 << 12;
-      SystemInstruction(Op, 0, FEXCore::ToUnderlying(DCOp), rt);
-    }
-    // TODO: DVP
-    // TODO: IC
-    // TODO: TLBI
+  // Reserved
+  void udf(uint32_t Imm) {
+    LOGMAN_THROW_A_FMT(Imm < 0x1'0000, "Immediate needs to be 16-bit");
+    dc32(Imm);
+  }

-    // Exception generation
-    void svc(uint32_t Imm) {
-      ExceptionGeneration(0b000, 0b000, 0b01, Imm);
-    }
-    void hvc(uint32_t Imm) {
-      ExceptionGeneration(0b000, 0b000, 0b10, Imm);
-    }
-    void smc(uint32_t Imm) {
-      ExceptionGeneration(0b000, 0b000, 0b11, Imm);
-    }
-    void brk(uint32_t Imm) {
-      ExceptionGeneration(0b001, 0b000, 0b00, Imm);
-    }
-    void hlt(uint32_t Imm) {
-      ExceptionGeneration(0b010, 0b000, 0b00, Imm);
-    }
-    void tcancel(uint32_t Imm) {
-      ExceptionGeneration(0b011, 0b000, 0b00, Imm);
-    }
-    void dcps1(uint32_t Imm) {
-      ExceptionGeneration(0b101, 0b000, 0b01, Imm);
-    }
-    void dcps2(uint32_t Imm) {
-      ExceptionGeneration(0b101, 0b000, 0b10, Imm);
-    }
-    void dcps3(uint32_t Imm) {
-      ExceptionGeneration(0b101, 0b000, 0b11, Imm);
-    }
-    // System instructions with register argument
-    void wfet(ARMEmitter::Register rt) {
-      SystemInstructionWithReg(0b0000, 0b000, rt);
-    }
-    void wfit(ARMEmitter::Register rt) {
-      SystemInstructionWithReg(0b0000, 0b001, rt);
-    }
+  // System with result
+  // TODO: SYSL
+  // System Instruction
+  // TODO: AT
+  // TODO: CFP
+  // TODO: CPP
+  void dc(ARMEmitter::DataCacheOperation DCOp, ARMEmitter::Register rt) {
+    constexpr uint32_t Op = 0b1101'0101'0000'1000'0111 << 12;
+    SystemInstruction(Op, 0, FEXCore::ToUnderlying(DCOp), rt);
+  }
+  // TODO: DVP
+  // TODO: IC
+  // TODO: TLBI

-    // Hints
-    void nop() {
-      Hint(ARMEmitter::HintRegister::NOP);
-    }
-    void yield() {
-      Hint(ARMEmitter::HintRegister::YIELD);
-    }
-    void wfe() {
-      Hint(ARMEmitter::HintRegister::WFE);
-    }
-    void wfi() {
-      Hint(ARMEmitter::HintRegister::WFI);
-    }
-    void sev() {
-      Hint(ARMEmitter::HintRegister::SEV);
-    }
-    void sevl() {
-      Hint(ARMEmitter::HintRegister::SEVL);
-    }
-    void dgh() {
-      Hint(ARMEmitter::HintRegister::DGH);
-    }
-    void csdb() {
-      Hint(ARMEmitter::HintRegister::CSDB);
-    }
+  // Exception generation
+  void svc(uint32_t Imm) {
+    ExceptionGeneration(0b000, 0b000, 0b01, Imm);
+  }
+  void hvc(uint32_t Imm) {
+    ExceptionGeneration(0b000, 0b000, 0b10, Imm);
+  }
+  void smc(uint32_t Imm) {
+    ExceptionGeneration(0b000, 0b000, 0b11, Imm);
+  }
+  void brk(uint32_t Imm) {
+    ExceptionGeneration(0b001, 0b000, 0b00, Imm);
+  }
+  void hlt(uint32_t Imm) {
+    ExceptionGeneration(0b010, 0b000, 0b00, Imm);
+  }
+  void tcancel(uint32_t Imm) {
+    ExceptionGeneration(0b011, 0b000, 0b00, Imm);
+  }
+  void dcps1(uint32_t Imm) {
+    ExceptionGeneration(0b101, 0b000, 0b01, Imm);
+  }
+  void dcps2(uint32_t Imm) {
+    ExceptionGeneration(0b101, 0b000, 0b10, Imm);
+  }
+  void dcps3(uint32_t Imm) {
+    ExceptionGeneration(0b101, 0b000, 0b11, Imm);
+  }
+  // System instructions with register argument
+  void wfet(ARMEmitter::Register rt) {
+    SystemInstructionWithReg(0b0000, 0b000, rt);
+  }
+  void wfit(ARMEmitter::Register rt) {
+    SystemInstructionWithReg(0b0000, 0b001, rt);
+  }

-    // Barriers
-    void clrex(uint32_t imm = 15) {
-      LOGMAN_THROW_AA_FMT(imm < 16, "Immediate out of range");
-      Barrier(ARMEmitter::BarrierRegister::CLREX, imm);
-    }
-    void dsb(ARMEmitter::BarrierScope Scope) {
-      Barrier(ARMEmitter::BarrierRegister::DSB, FEXCore::ToUnderlying(Scope));
-    }
-    void dmb(ARMEmitter::BarrierScope Scope) {
-      Barrier(ARMEmitter::BarrierRegister::DMB, FEXCore::ToUnderlying(Scope));
-    }
-    void isb() {
-      Barrier(ARMEmitter::BarrierRegister::ISB, FEXCore::ToUnderlying(ARMEmitter::BarrierScope::SY));
-    }
-    void sb() {
-      Barrier(ARMEmitter::BarrierRegister::SB, 0);
-    }
-    void tcommit() {
-      Barrier(ARMEmitter::BarrierRegister::TCOMMIT, 0);
-    }
+  // Hints
+  void nop() {
+    Hint(ARMEmitter::HintRegister::NOP);
+  }
+  void yield() {
+    Hint(ARMEmitter::HintRegister::YIELD);
+  }
+  void wfe() {
+    Hint(ARMEmitter::HintRegister::WFE);
+  }
+  void wfi() {
+    Hint(ARMEmitter::HintRegister::WFI);
+  }
+  void sev() {
+    Hint(ARMEmitter::HintRegister::SEV);
+  }
+  void sevl() {
+    Hint(ARMEmitter::HintRegister::SEVL);
+  }
+  void dgh() {
+    Hint(ARMEmitter::HintRegister::DGH);
+  }
+  void csdb() {
+    Hint(ARMEmitter::HintRegister::CSDB);
+  }

-    // System register move
-    void msr(ARMEmitter::SystemRegister reg, ARMEmitter::Register rt) {
-      constexpr uint32_t Op = 0b1101'0101'0001 << 20;
-      SystemRegisterMove(Op, rt, reg);
-    }
+  // Barriers
+  void clrex(uint32_t imm = 15) {
+    LOGMAN_THROW_A_FMT(imm < 16, "Immediate out of range");
+    Barrier(ARMEmitter::BarrierRegister::CLREX, imm);
+  }
+  void dsb(ARMEmitter::BarrierScope Scope) {
+    Barrier(ARMEmitter::BarrierRegister::DSB, FEXCore::ToUnderlying(Scope));
+  }
+  void dmb(ARMEmitter::BarrierScope Scope) {
+    Barrier(ARMEmitter::BarrierRegister::DMB, FEXCore::ToUnderlying(Scope));
+  }
+  void isb() {
+    Barrier(ARMEmitter::BarrierRegister::ISB, FEXCore::ToUnderlying(ARMEmitter::BarrierScope::SY));
+  }
+  void sb() {
+    Barrier(ARMEmitter::BarrierRegister::SB, 0);
+  }
+  void tcommit() {
+    Barrier(ARMEmitter::BarrierRegister::TCOMMIT, 0);
+  }

-    void mrs(ARMEmitter::Register rd, ARMEmitter::SystemRegister reg) {
-      constexpr uint32_t Op = 0b1101'0101'0011 << 20;
-      SystemRegisterMove(Op, rd, reg);
-    }
+  // System register move
+  void msr(ARMEmitter::SystemRegister reg, ARMEmitter::Register rt) {
+    constexpr uint32_t Op = 0b1101'0101'0001 << 20;
+    SystemRegisterMove(Op, rt, reg);
+  }
+
+  void mrs(ARMEmitter::Register rd, ARMEmitter::SystemRegister reg) {
+    constexpr uint32_t Op = 0b1101'0101'0011 << 20;
+    SystemRegisterMove(Op, rd, reg);
+  }

 private:

-    // Exception Generation
-    void ExceptionGeneration(uint32_t opc, uint32_t op2, uint32_t LL, uint32_t Imm) {
-      LOGMAN_THROW_AA_FMT((Imm & 0xFFFF'0000) == 0, "Imm amount too large");
+  // Exception Generation
+  void ExceptionGeneration(uint32_t opc, uint32_t op2, uint32_t LL, uint32_t Imm) {
+    LOGMAN_THROW_A_FMT((Imm & 0xFFFF'0000) == 0, "Imm amount too large");

-      uint32_t Instr = 0b1101'0100 << 24;
+    uint32_t Instr = 0b1101'0100 << 24;

-      Instr |= opc << 21;
-      Instr |= Imm << 5;
-      Instr |= op2 << 2;
-      Instr |= LL;
+    Instr |= opc << 21;
+    Instr |= Imm << 5;
+    Instr |= op2 << 2;
+    Instr |= LL;

-      dc32(Instr);
-    }
+    dc32(Instr);
+  }

-    // System instructions with register argument
-    void SystemInstructionWithReg(uint32_t CRm, uint32_t op2, ARMEmitter::Register rt) {
-      uint32_t Instr = 0b1101'0101'0000'0011'0001 << 12;
+  // System instructions with register argument
+  void SystemInstructionWithReg(uint32_t CRm, uint32_t op2, ARMEmitter::Register rt) {
+    uint32_t Instr = 0b1101'0101'0000'0011'0001 << 12;

-      Instr |= CRm << 8;
-      Instr |= op2 << 5;
-      Instr |= Encode_rt(rt);
-      dc32(Instr);
-    }
+    Instr |= CRm << 8;
+    Instr |= op2 << 5;
+    Instr |= Encode_rt(rt);
+    dc32(Instr);
+  }

-    // Hints
-    void Hint(ARMEmitter::HintRegister Reg) {
-      uint32_t Instr = 0b1101'0101'0000'0011'0010'0000'0001'1111U;
-      Instr |= FEXCore::ToUnderlying(Reg);
-      dc32(Instr);
-    }
-    // Barriers
-    void Barrier(ARMEmitter::BarrierRegister Reg, uint32_t CRm) {
-      uint32_t Instr = 0b1101'0101'0000'0011'0011'0000'0001'1111U;
-      Instr |= CRm << 8;
-      Instr |= FEXCore::ToUnderlying(Reg);
-      dc32(Instr);
-    }
+  // Hints
+  void Hint(ARMEmitter::HintRegister Reg) {
+    uint32_t Instr = 0b1101'0101'0000'0011'0010'0000'0001'1111U;
+    Instr |= FEXCore::ToUnderlying(Reg);
+    dc32(Instr);
+  }
+  // Barriers
+  void Barrier(ARMEmitter::BarrierRegister Reg, uint32_t CRm) {
+    uint32_t Instr = 0b1101'0101'0000'0011'0011'0000'0001'1111U;
+    Instr |= CRm << 8;
+    Instr |= FEXCore::ToUnderlying(Reg);
+    dc32(Instr);
+  }

-    // System Instruction
-    void SystemInstruction(uint32_t Op, uint32_t L, uint32_t SubOp, ARMEmitter::Register rt) {
-      uint32_t Instr = Op;
+  // System Instruction
+  void SystemInstruction(uint32_t Op, uint32_t L, uint32_t SubOp, ARMEmitter::Register rt) {
+    uint32_t Instr = Op;

-      Instr |= L << 21;
-      Instr |= SubOp;
-      Instr |= Encode_rt(rt);
+    Instr |= L << 21;
+    Instr |= SubOp;
+    Instr |= Encode_rt(rt);

-      dc32(Instr);
-    }
+    dc32(Instr);
+  }

-    // System register move
-    void SystemRegisterMove(uint32_t Op, ARMEmitter::Register rt, ARMEmitter::SystemRegister reg) {
-      uint32_t Instr = Op;
+  // System register move
+  void SystemRegisterMove(uint32_t Op, ARMEmitter::Register rt, ARMEmitter::SystemRegister reg) {
+    uint32_t Instr = Op;

-      Instr |= FEXCore::ToUnderlying(reg);
-      Instr |= Encode_rt(rt);
+    Instr |= FEXCore::ToUnderlying(reg);
+    Instr |= Encode_rt(rt);

-      dc32(Instr);
-    }
+    dc32(Instr);
+  }

+#ifndef INCLUDED_BY_EMITTER
+}; // struct LoadstoreEmitterOps
+} // namespace ARMEmitter
+#endif
@@ -34,21 +34,15 @@
 // by the corresponding fields in the logical instruction.
 // If it can not be encoded, the function returns false, and the values pointed
 // to by n, imm_s and imm_r are undefined.
-static bool IsImmLogical(uint64_t value,
-                             unsigned width,
-                             unsigned* n = nullptr,
-                             unsigned* imm_s = nullptr,
-                             unsigned* imm_r = nullptr) {
+static bool IsImmLogical(uint64_t value, unsigned width, unsigned* n = nullptr, unsigned* imm_s = nullptr, unsigned* imm_r = nullptr) {
  [[maybe_unused]] constexpr auto kBRegSize = 8;
  [[maybe_unused]] constexpr auto kHRegSize = 16;
  [[maybe_unused]] constexpr auto kSRegSize = 32;
  [[maybe_unused]] constexpr auto kDRegSize = 64;

  constexpr auto kWRegSize = 32;
-  constexpr auto kXRegSize = 64;

-  LOGMAN_THROW_A_FMT((width == kBRegSize) || (width == kHRegSize) ||
-              (width == kSRegSize) || (width == kDRegSize), "Unexpected imm size");
+  LOGMAN_THROW_A_FMT((width == kBRegSize) || (width == kHRegSize) || (width == kSRegSize) || (width == kDRegSize), "Unexpected imm size");

  bool negate = false;

@@ -134,8 +128,8 @@ static bool IsImmLogical(uint64_t value,
    // Compute the repeat distance d, and set up a bitmask covering the basic
    // unit of repetition (i.e. a word with the bottom d bits set). Also, in all
    // of these cases the N bit of the output will be zero.
-    clz_a = CountLeadingZeros(a, kXRegSize);
-    int clz_c = CountLeadingZeros(c, kXRegSize);
+    clz_a = std::countl_zero(a);
+    int clz_c = std::countl_zero(c);
    d = clz_a - clz_c;
    mask = ((UINT64_C(1) << d) - 1);
    out_n = 0;
@@ -156,7 +150,7 @@ static bool IsImmLogical(uint64_t value,
      // of set bits in our word, meaning that we have the trivial case of
      // d == 64 and only one 'repetition'. Set up all the same variables as in
      // the general case above, and set the N bit in the output.
-      clz_a = CountLeadingZeros(a, kXRegSize);
+      clz_a = std::countl_zero(a);
      d = 64;
      mask = ~UINT64_C(0);
      out_n = 1;
@@ -164,7 +158,7 @@ static bool IsImmLogical(uint64_t value,
  }

  // If the repeat period d is not a power of two, it can't be encoded.
-  if (!IsPowerOf2(d)) {
+  if (!std::has_single_bit(uint32_t(d))) {
    return false;
  }

@@ -182,14 +176,9 @@ static bool IsImmLogical(uint64_t value,
  // (1 + 2^d + 2^(2d) + ...), i.e. 0x0001000100010001 or similar. These can
  // be derived using a table lookup on CLZ(d).
  static const uint64_t multipliers[] = {
-      0x0000000000000001UL,
-      0x0000000100000001UL,
-      0x0001000100010001UL,
-      0x0101010101010101UL,
-      0x1111111111111111UL,
-      0x5555555555555555UL,
+    0x0000000000000001UL, 0x0000000100000001UL, 0x0001000100010001UL, 0x0101010101010101UL, 0x1111111111111111UL, 0x5555555555555555UL,
  };
-  uint64_t multiplier = multipliers[CountLeadingZeros(d, kXRegSize) - 57];
+  uint64_t multiplier = multipliers[std::countl_zero(uint64_t(d)) - 57];
  uint64_t candidate = (b - a) * multiplier;

  if (value != candidate) {
@@ -204,7 +193,7 @@ static bool IsImmLogical(uint64_t value,
  // Count the set bits in our basic stretch. The special case of clz(0) == -1
  // makes the answer come out right for stretches that reach the very top of
  // the word (e.g. numbers like 0xffffc00000000000).
-  int clz_b = (b == 0) ? -1 : CountLeadingZeros(b, kXRegSize);
+  int clz_b = (b == 0) ? -1 : std::countl_zero(b);
  int s = clz_a - clz_b;

  // Decide how many bits to rotate right by, to put the low bit of that basic
@@ -234,9 +223,13 @@ static bool IsImmLogical(uint64_t value,
  //    11110s     2    UInt(s)
  //
  // So we 'or' (2 * -d) with our computed s to form imms.
-  if ((n != NULL) || (imm_s != NULL) || (imm_r != NULL)) {
+  if (n != nullptr) {
    *n = out_n;
+  }
+  if (imm_s != nullptr) {
    *imm_s = ((2 * -d) | (s - 1)) & 0x3f;
+  }
+  if (imm_r != nullptr) {
    *imm_r = r;
  }

@@ -244,7 +237,9 @@ static bool IsImmLogical(uint64_t value,
 }

 static inline bool IsIntN(unsigned n, int64_t x) {
-  if (n == 64) return true;
+  if (n == 64) {
+    return true;
+  }
  int64_t limit = INT64_C(1) << (n - 1);
  return (-limit <= x) && (x < limit);
 }
@@ -271,11 +266,15 @@ V(57) V(58) V(59) V(60) V(61) V(62) V(63)

 // clang-format on

-#define DECLARE_IS_INT_N(N)                                       \
-  static inline bool IsInt##N(int64_t x) { return IsIntN(N, x); }
+#define DECLARE_IS_INT_N(N)                \
+  static inline bool IsInt##N(int64_t x) { \
+    return IsIntN(N, x);                   \
+  }

-#define DECLARE_IS_UINT_N(N)                                        \
-  static inline bool IsUint##N(int64_t x) { return IsUintN(N, x); }
+#define DECLARE_IS_UINT_N(N)                \
+  static inline bool IsUint##N(int64_t x) { \
+    return IsUintN(N, x);                   \
+  }

 INT_1_TO_63_LIST(DECLARE_IS_INT_N)
 INT_1_TO_63_LIST(DECLARE_IS_UINT_N)
@@ -285,14 +284,9 @@ INT_1_TO_63_LIST(DECLARE_IS_UINT_N)

 private:

-template <typename V>
-static inline bool IsPowerOf2(V value) {
-  return (value != 0) && ((value & (value - 1)) == 0);
-}
-
 // Some compilers dislike negating unsigned integers,
 // so we provide an equivalent.
-template <typename T>
+template<typename T>
 static inline T UnsignedNegate(T value) {
  static_assert(std::is_unsigned<T>::value);
  return ~value + 1;
@@ -302,50 +296,4 @@ static inline uint64_t LowestSetBit(uint64_t value) {
  return value & UnsignedNegate(value);
 }

-template <typename V>
-static inline int CountLeadingZeros(V value, int width = (sizeof(V) * 8)) {
-#if COMPILER_HAS_BUILTIN_CLZ
-  if (width == 32) {
-    return (value == 0) ? 32 : __builtin_clz(static_cast<unsigned>(value));
-  } else if (width == 64) {
-    return (value == 0) ? 64 : __builtin_clzll(value);
-  }
-#endif
-  return CountLeadingZerosFallBack(value, width);
-}
-
-static inline int CountLeadingZerosFallBack(uint64_t value, int width) {
-  LOGMAN_THROW_A_FMT(IsPowerOf2(width) && (width <= 64), "Invalid width");
-  if (value == 0) {
-    return width;
-  }
-  int count = 0;
-  value = value << (64 - width);
-  if ((value & UINT64_C(0xffffffff00000000)) == 0) {
-    count += 32;
-    value = value << 32;
-  }
-  if ((value & UINT64_C(0xffff000000000000)) == 0) {
-    count += 16;
-    value = value << 16;
-  }
-  if ((value & UINT64_C(0xff00000000000000)) == 0) {
-    count += 8;
-    value = value << 8;
-  }
-  if ((value & UINT64_C(0xf000000000000000)) == 0) {
-    count += 4;
-    value = value << 4;
-  }
-  if ((value & UINT64_C(0xc000000000000000)) == 0) {
-    count += 2;
-    value = value << 2;
-  }
-  if ((value & UINT64_C(0x8000000000000000)) == 0) {
-    count += 1;
-  }
-  count += (value == 0);
-  return count;
-}
-
 public:
@@ -4,7 +4,8 @@ file(GLOB GEN_CONFIG_SOURCES CONFIGURE_DEPENDS *.json.in)
 # Any application configuration json file gets installed
 foreach(CONFIG_SRC ${CONFIG_SOURCES})
  install(FILES ${CONFIG_SRC}
-    DESTINATION ${DATA_DIRECTORY}/AppConfig/)
+    DESTINATION ${DATA_DIRECTORY}/AppConfig/
+    COMPONENT Runtime)
 endforeach()

 # Any configuration file json file that needs to be generated
@@ -14,12 +15,10 @@ foreach(GEN_CONFIG_SRC ${GEN_CONFIG_SOURCES})
  get_filename_component(CONFIG_NAME ${GEN_CONFIG_SRC} NAME_WLE)

  # Configure it
-  configure_file(
-    ${GEN_CONFIG_SRC}
-    ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME})
+  configure_file(${GEN_CONFIG_SRC} ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME})

  # Then install the configured json
-  install(
-    FILES ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME}
-    DESTINATION ${DATA_DIRECTORY}/AppConfig/)
+  install(FILES ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME}
+    DESTINATION ${DATA_DIRECTORY}/AppConfig/
+    COMPONENT Runtime)
 endforeach()
@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: MIT
+
+if (CMAKE_CROSSCOMPILING)
+    return()
+endif()
+
+include(FindPackageHandleStandardArgs)
+
+find_package(Zycore QUIET CONFIG)
+
+if (Zycore_CONSIDERED_CONFIGS)
+    find_package_handle_standard_args(Zycore CONFIG_MODE)
+else()
+    find_package(PkgConfig QUIET)
+    pkg_search_module(Zycore QUIET IMPORTED_TARGET zycore)
+    find_package_handle_standard_args(Zycore
+        REQUIRED_VARS zycore_LINK_LIBRARIES
+        VERSION_VAR zycore_VERSION)
+
+    if (TARGET PkgConfig::zycore)
+      add_library(Zycore::Zycore ALIAS PkgConfig::zycore)
+    endif()
+endif()
@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: MIT
+
+if (CMAKE_CROSSCOMPILING)
+    return()
+endif()
+
+include(FindPackageHandleStandardArgs)
+
+find_package(Zydis QUIET CONFIG)
+
+if (Zydis_CONSIDERED_CONFIGS)
+    find_package_handle_standard_args(Zydis CONFIG_MODE)
+else()
+    find_package(PkgConfig QUIET)
+    pkg_search_module(Zydis QUIET IMPORTED_TARGET zydis)
+    find_package_handle_standard_args(Zydis
+        REQUIRED_VARS zydis_LINK_LIBRARIES
+        VERSION_VAR zydis_VERSION)
+
+    if (TARGET PkgConfig::zydis)
+      add_library(Zydis::Zydis ALIAS PkgConfig::zydis)
+    endif()
+endif()
@@ -0,0 +1,18 @@
+# SPDX-License-Identifier: MIT
+
+include(FindPackageHandleStandardArgs)
+
+find_package(PkgConfig QUIET)
+pkg_search_module(xxhash QUIET IMPORTED_TARGET xxhash libxxhash)
+find_package_handle_standard_args(xxhash
+    REQUIRED_VARS xxhash_LINK_LIBRARIES
+    VERSION_VAR xxhash_VERSION
+)
+
+if (xxhash_FOUND AND NOT TARGET xxHash::xxhash)
+    if (TARGET PkgConfig::xxhash)
+        add_library(xxHash::xxhash ALIAS PkgConfig::xxhash)
+    else()
+        add_library(xxHash::xxhash ALIAS xxhash)
+    endif()
+endif()
@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: MIT
+
+# This applies some common linker options that reduce code size and linking time in Release mode. Namely:
+# --gc-sections: Linktime garbage collection, discards unused sections from the final output
+# --strip-all  : Similar to running `strip`, discards the symbol table from the final output
+# --as-needed  : Only includes libraries that are actually needed in the final output.
+
+macro(LinkerGC target)
+  if (CMAKE_BUILD_TYPE MATCHES "RELEASE")
+    target_link_options(${target} PRIVATE
+      "LINKER:--gc-sections"
+      "LINKER:--strip-all"
+      "LINKER:--as-needed")
+  endif()
+endmacro()
@@ -1,6 +1,6 @@
 # This is a reference AArch64 cross compile script
 # Pass in to cmake when building:
-# eg: cmake -DCMAKE_TOOLCHAIN_FILE=../CMakeToolchains/AArch64.cmake ..
+# eg: cmake --toolchain ../Data/CMake/toolchain_aarch64.cmake ..
 if (NOT DEFINED ENV{SYSROOT})
  message(FATAL_ERROR "Need to have SYSROOT environment variable set")
 endif()
@@ -4,12 +4,13 @@ set(CMAKE_RC_COMPILER ${MINGW_TRIPLE}-windres)
 set(CMAKE_C_COMPILER ${MINGW_TRIPLE}-clang)
 set(CMAKE_CXX_COMPILER ${MINGW_TRIPLE}-clang++)
 set(CMAKE_DLLTOOL ${MINGW_TRIPLE}-dlltool)
+set(CMAKE_AR ${MINGW_TRIPLE}-ar)

 # Compile everything as static to avoid requiring the MinGW runtime libraries, force page aligned sections so that
 # debug symbols work correctly, and disable loop alignment to workaround an LLVM bug
 # (https://github.com/llvm/llvm-project/issues/47432)
-set(CMAKE_SHARED_LINKER_FLAGS_INIT "-static -static-libgcc -static-libstdc++ -Wl,--file-alignment=4096,/mllvm:-align-loops=1")
-set(CMAKE_EXE_LINKER_FLAGS_INIT "-static -static-libgcc -static-libstdc++ -Wl,--file-alignment=4096,/mllvm:-align-loops=1")
+set(CMAKE_SHARED_LINKER_FLAGS_INIT "-static -Wl,--file-alignment=4096,/mllvm:-align-loops=1")
+set(CMAKE_EXE_LINKER_FLAGS_INIT "-static -Wl,--file-alignment=4096,/mllvm:-align-loops=1")
 set(CMAKE_C_STANDARD_LIBRARIES "" CACHE STRING "" FORCE)
 set(CMAKE_CXX_STANDARD_LIBRARIES "" CACHE STRING "" FORCE)
 set(CMAKE_STANDARD_LIBRARIES "" CACHE STRING "" FORCE)
@@ -0,0 +1,31 @@
+# --- Stage 1: Builder ---
+FROM ubuntu:22.04 as builder
+
+RUN DEBIAN_FRONTEND="noninteractive" apt-get update
+RUN DEBIAN_FRONTEND="noninteractive" apt install -y cmake \
+clang-13 llvm-13 nasm ninja-build pkg-config \
+libcap-dev libglfw3-dev libepoxy-dev python3-dev libsdl2-dev \
+python3 linux-headers-generic  \
+git  qtbase5-dev qtdeclarative5-dev lld
+
+RUN git clone --recurse-submodules https://github.com/FEX-Emu/FEX.git
+WORKDIR /FEX
+RUN mkdir build
+
+ARG CC=clang-13
+ARG CXX=clang++-13
+RUN cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Release -DUSE_LINKER=lld -DENABLE_LTO=True -DBUILD_TESTING=False -DENABLE_ASSERTIONS=False -G Ninja .
+RUN ninja
+
+WORKDIR /FEX/build
+
+# --- Stage 2: Runner ---
+FROM builder as runner
+
+RUN DEBIAN_FRONTEND="noninteractive" apt-get update
+RUN DEBIAN_FRONTEND="noninteractive" apt install -y \
+libcap-dev libglfw3-dev libepoxy-dev
+
+COPY --from=builder /FEX/Bin/* /usr/bin/
+
+WORKDIR /
@@ -9,25 +9,6 @@
        "@PREFIX_LIB@/libGL.so.1.7.0"
      ]
    },
-    "GLESv2": {
-      "Library": "libGLESv2-guest.so",
-      "Depends": [
-        "X11"
-      ],
-      "Overlay": [
-        "@PREFIX_LIB@/libGLESv2.so",
-        "@PREFIX_LIB@/libGLESv2.so.2",
-        "@PREFIX_LIB@/libGLESv2.so.2.0.0"
-      ]
-    },
-    "X11": {
-      "Library": "libX11-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libX11.so",
-        "@PREFIX_LIB@/libX11.so.6",
-        "@PREFIX_LIB@/libX11.so.6.4.0"
-      ]
-    },
    "Vulkan": {
      "Library": "libvulkan-guest.so",
      "Overlay": [
@@ -36,89 +17,6 @@
        "@HOME@/.local/share/Steam/ubuntu12_32/steam-runtime/pinned_libs_64/libvulkan.so.1"
      ]
    },
-    "xcb": {
-      "Depends": [
-        "X11"
-      ],
-      "Library": "libxcb-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb.so",
-        "@PREFIX_LIB@/libxcb.so.1",
-        "@PREFIX_LIB@/libxcb.so.1.1.0"
-      ]
-    },
-    "xcb-dri2": {
-      "Library": "libxcb-dri2-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-dri2.so",
-        "@PREFIX_LIB@/libxcb-dri2.so.0",
-        "@PREFIX_LIB@/libxcb-dri2.so.0.0.0"
-      ]
-    },
-    "xcb-dri3": {
-      "Library": "libxcb-dri3-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-dri3.so",
-        "@PREFIX_LIB@/libxcb-dri3.so.0",
-        "@PREFIX_LIB@/libxcb-dri3.so.0.0.0"
-      ]
-    },
-    "xcb-xfixes": {
-      "Library": "libxcb-xfixes-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-xfixes.so",
-        "@PREFIX_LIB@/libxcb-xfixes.so.0",
-        "@PREFIX_LIB@/libxcb-xfixes.so.0.0.0"
-      ]
-    },
-    "xcb-shm": {
-      "Library": "libxcb-shm-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-shm.so",
-        "@PREFIX_LIB@/libxcb-shm.so.0",
-        "@PREFIX_LIB@/libxcb-shm.so.0.0.0"
-      ]
-    },
-    "xcb-sync": {
-      "Library": "libxcb-sync-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-sync.so",
-        "@PREFIX_LIB@/libxcb-sync.so.1",
-        "@PREFIX_LIB@/libxcb-sync.so.1.0.0"
-      ]
-    },
-    "xcb-randr": {
-      "Library": "libxcb-randr-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-randr.so",
-        "@PREFIX_LIB@/libxcb-randr.so.0",
-        "@PREFIX_LIB@/libxcb-randr.so.0.1.0"
-      ]
-    },
-    "xcb-present": {
-      "Library": "libxcb-present-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-present.so",
-        "@PREFIX_LIB@/libxcb-present.so.0",
-        "@PREFIX_LIB@/libxcb-present.so.0.0.0"
-      ]
-    },
-    "xcb-glx": {
-      "Library": "libxcb-glx-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxcb-glx.so",
-        "@PREFIX_LIB@/libxcb-glx.so.0",
-        "@PREFIX_LIB@/libxcb-glx.so.0.0.0"
-      ]
-    },
-    "xshmfence": {
-      "Library": "libxshmfence-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libxshmfence.so",
-        "@PREFIX_LIB@/libxshmfence.so.1",
-        "@PREFIX_LIB@/libxshmfence.so.1.0.0"
-      ]
-    },
    "drm": {
      "Library": "libdrm-guest.so",
      "Overlay": [
@@ -141,38 +39,6 @@
        "@PREFIX_LIB@/libfex_thunk_test.so"
      ]
    },
-    "Xrender": {
-      "Library": "libXrender-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libXrender.so",
-        "@PREFIX_LIB@/libXrender.so.1",
-        "@PREFIX_LIB@/libXrender.so.1.3.0"
-      ]
-    },
-    "Xext": {
-      "Library": "libXext-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libXext.so",
-        "@PREFIX_LIB@/libXext.so.6",
-        "@PREFIX_LIB@/libXext.so.6.4.0"
-      ]
-    },
-    "Xfixes": {
-      "Library": "libXfixes-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libXfixes.so",
-        "@PREFIX_LIB@/libXfixes.so.3",
-        "@PREFIX_LIB@/libXfixes.so.3.1.0"
-      ]
-    },
-    "OpenCL": {
-      "Library" : "libOpenCL-guest.so",
-      "Overlay": [
-        "@PREFIX_LIB@/libOpenCL.so",
-        "@PREFIX_LIB@/libOpenCL.so.1",
-        "@PREFIX_LIB@/libOpenCL.so.1.0.0"
-      ]
-    },
    "WaylandClient": {
      "Library" : "libwayland-client-guest.so",
      "Overlay": [
@@ -181,6 +47,12 @@
        "@PREFIX_LIB@/libwayland-client.so.0.20.0"
      ]
    },
-    "":{}
+    "cuda": {
+      "Library" : "libcuda-guest.so",
+      "Overlay": [
+        "@PREFIX_LIB@/libcuda.so",
+        "@PREFIX_LIB@/libcuda.so.1"
+      ]
+    }
  }
 }
@@ -3,23 +3,21 @@ function(GenBinFmt Name)
  get_filename_component(FMT_NAME ${Name} NAME_WE)

  # Configure it
-  configure_file(
-    ${Name}
-    ${CMAKE_BINARY_DIR}/Data/binfmts/${FMT_NAME})
+  configure_file(${Name} ${CMAKE_BINARY_DIR}/Data/binfmts/${FMT_NAME})

  # Then install the configured binfmt
-  install(
-    FILES ${CMAKE_BINARY_DIR}/Data/binfmts/${FMT_NAME}
-    DESTINATION ${CMAKE_INSTALL_PREFIX}/share/binfmts/)
+  install(FILES ${CMAKE_BINARY_DIR}/Data/binfmts/${FMT_NAME}
+    DESTINATION ${CMAKE_INSTALL_PREFIX}/share/binfmts/
+    COMPONENT Runtime)
 endfunction()

 if (NOT USE_LEGACY_BINFMTMISC)
  configure_file(FEX-x86.conf.in ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86.conf)
  configure_file(FEX-x86_64.conf.in ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86_64.conf)

-  install(
-    FILES ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86.conf ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86_64.conf
-    DESTINATION ${CMAKE_INSTALL_PREFIX}/lib/binfmt.d/)
+  install(FILES ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86.conf ${CMAKE_BINARY_DIR}/Data/binfmts/FEX-x86_64.conf
+    DESTINATION ${CMAKE_INSTALL_PREFIX}/lib/binfmt.d/
+    COMPONENT Runtime)
 else()
  GenBinFmt(FEX-x86.in)
  GenBinFmt(FEX-x86_64.in)
@@ -1 +1 @@
-:FEX-x86:M:0:\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00:\xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff:@CMAKE_INSTALL_PREFIX@/bin/FEXInterpreter:POCF
+:FEX-x86:M:0:\x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00:\xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff:@CMAKE_INSTALL_PREFIX@/bin/FEX:POCF
@@ -1,5 +1,5 @@
 package fex
-interpreter @CMAKE_INSTALL_PREFIX@/bin/FEXInterpreter
+interpreter @CMAKE_INSTALL_PREFIX@/bin/FEX
 magic \x7fELF\x01\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x03\x00
 offset 0
 mask \xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff
@@ -1 +1 @@
-:FEX-x86_64:M:0:\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x3e\x00:\xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff:@CMAKE_INSTALL_PREFIX@/bin/FEXInterpreter:POCF
+:FEX-x86_64:M:0:\x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x3e\x00:\xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff:@CMAKE_INSTALL_PREFIX@/bin/FEX:POCF
@@ -1,5 +1,5 @@
 package fex
-interpreter @CMAKE_INSTALL_PREFIX@/bin/FEXInterpreter
+interpreter @CMAKE_INSTALL_PREFIX@/bin/FEX
 magic \x7fELF\x02\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x02\x00\x3e\x00
 offset 0
 mask \xff\xff\xff\xff\xff\xfe\xfe\x00\x00\x00\x00\xff\xff\xff\xff\xff\xfe\xff\xff\xff
@@ -0,0 +1,35 @@
+{ pkgs ? import <nixpkgs> { } }:
+
+let
+  pkgsCross32 = pkgs.pkgsCross.gnu32;
+  pkgsCross64 = pkgs.pkgsCross.gnu64;
+
+  gcc32 = pkgs.writeText "toolchain_nix_gcc_x86_32.txt" ''
+    set(CMAKE_SYSTEM_PROCESSOR i686)
+    set(CMAKE_C_COMPILER ${pkgsCross32.buildPackages.gcc}/bin/i686-unknown-linux-gnu-gcc)
+    set(CMAKE_CXX_COMPILER ${pkgsCross32.buildPackages.gcc}/bin/i686-unknown-linux-gnu-g++)
+  '';
+
+  gcc64 = pkgs.writeText "toolchain_nix_gcc_x86_64.txt" ''
+    set(CMAKE_SYSTEM_PROCESSOR x86_64)
+    set(CMAKE_C_COMPILER ${pkgsCross64.buildPackages.gcc}/bin/x86_64-unknown-linux-gnu-gcc)
+    set(CMAKE_CXX_COMPILER ${pkgsCross64.buildPackages.gcc}/bin/x86_64-unknown-linux-gnu-g++)
+  '';
+in
+pkgs.mkShell {
+  buildInputs = [
+    pkgsCross64.buildPackages.clang
+    pkgsCross32.buildPackages.clang
+  ];
+
+  shellHook = ''
+    if [[ $- == *i* ]]; then
+      echo "toolchain32: ${gcc32}"
+      echo "toolchain64: ${gcc64}"
+      echo ""
+      echo "Use \$FEX_CMAKE_TOOLCHAINS to configure CMake."
+    fi
+  '';
+
+  FEX_CMAKE_TOOLCHAINS = "-DX86_32_TOOLCHAIN_FILE=${gcc32} -DX86_64_TOOLCHAIN_FILE=${gcc64}";
+}
@@ -0,0 +1,83 @@
+{ pkgs ? import <nixpkgs> { } }:
+
+let
+  pkgsCross32 = pkgs.pkgsCross.gnu32;
+  pkgsCross64 = pkgs.pkgsCross.gnu64;
+
+  devRootFS = pkgs.buildEnv {
+    name = "fex-dev-rootfs";
+    paths = [
+      pkgsCross64.stdenv.cc.libc_dev
+      pkgsCross32.stdenv.cc.libc_dev
+      pkgsCross64.stdenv.cc.cc
+      pkgsCross32.stdenv.cc.cc
+
+      pkgs.alsa-lib.dev
+      pkgs.libdrm.dev
+      pkgs.libGL.dev
+      pkgs.wayland.dev
+      pkgs.xorg.libX11.dev
+      pkgs.xorg.libxcb.dev
+      pkgs.xorg.libXrandr.dev
+      pkgs.xorg.libXrender.dev
+      pkgs.xorg.xorgproto
+    ];
+    ignoreCollisions = true;
+    pathsToLink = [
+      "/include"
+      "/lib"
+    ];
+
+    postBuild = ''
+      mkdir -p $out/usr
+      ln -s $out/include $out/usr/
+    '';
+  };
+
+  toolchain32 = pkgs.writeText "toolchain_nix_x86_32.txt" ''
+    set(CMAKE_EXE_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_MODULE_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_SHARED_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_SYSTEM_PROCESSOR i686)
+    set(CMAKE_C_COMPILER clang)
+    set(CMAKE_CXX_COMPILER clang++)
+    set(CMAKE_C_COMPILER ${pkgsCross32.buildPackages.clang}/bin/i686-unknown-linux-gnu-clang)
+    set(CMAKE_CXX_COMPILER ${pkgsCross32.buildPackages.clang}/bin/i686-unknown-linux-gnu-clang++)
+    set(CLANG_FLAGS "-nodefaultlibs -nostartfiles -lstdc++ -target i686-linux-gnu -msse2 -mfpmath=sse --sysroot=${devRootFS} -iwithsysroot/include")
+    set(CMAKE_C_FLAGS "''${CMAKE_C_FLAGS} ''${CLANG_FLAGS}")
+    set(CMAKE_CXX_FLAGS "''${CMAKE_CXX_FLAGS} ''${CLANG_FLAGS}")
+  '';
+
+  toolchain64 = pkgs.writeText "toolchain_nix_x86_64.txt" ''
+    set(CMAKE_EXE_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_MODULE_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_SHARED_LINKER_FLAGS_INIT "-fuse-ld=lld")
+    set(CMAKE_SYSTEM_PROCESSOR x86_64)
+    set(CMAKE_C_COMPILER clang)
+    set(CMAKE_CXX_COMPILER clang++)
+    set(CMAKE_C_COMPILER ${pkgsCross64.buildPackages.clang}/bin/x86_64-unknown-linux-gnu-clang)
+    set(CMAKE_CXX_COMPILER ${pkgsCross64.buildPackages.clang}/bin/x86_64-unknown-linux-gnu-clang++)
+    set(CLANG_FLAGS "-nodefaultlibs -nostartfiles -lstdc++ -target x86_64-linux-gnu --sysroot=${devRootFS} -iwithsysroot/usr/include")
+    set(CMAKE_C_FLAGS "''${CMAKE_C_FLAGS} ''${CLANG_FLAGS}")
+    set(CMAKE_CXX_FLAGS "''${CMAKE_CXX_FLAGS} ''${CLANG_FLAGS}")
+  '';
+in
+pkgs.mkShell {
+  buildInputs = [
+    pkgsCross64.buildPackages.clang
+    pkgsCross32.buildPackages.clang
+  ];
+
+  shellHook = ''
+    if [[ $- == *i* ]]; then
+      echo "Set up dev RootFS at ${devRootFS}"
+      echo "toolchain32: ${toolchain32}"
+      echo "toolchain64: ${toolchain64}"
+      echo ""
+      echo "Use \$FEX_CMAKE_TOOLCHAINS to configure CMake."
+    fi
+  '';
+
+  FEX_CMAKE_TOOLCHAINS = "-DX86_32_TOOLCHAIN_FILE=${toolchain32} -DX86_64_TOOLCHAIN_FILE=${toolchain64} -DX86_DEV_ROOTFS=${devRootFS}";
+  ROOTFS = "${devRootFS}";
+}
@@ -0,0 +1,52 @@
+{ pkgs ? import <nixpkgs> { } }:
+
+let
+  toolchain = pkgs.fetchzip {
+    url = "https://github.com/bylaws/llvm-mingw/releases/download/20250920/llvm-mingw-20250920-ucrt-ubuntu-22.04-aarch64.tar.xz";
+    sha256 = "sha256-LaojKjC8KzY+soW5u6eoDoXE3qtYk9Ejr7M3enTqRAE=";
+  };
+
+  cmakeToolchainFile = pkgs.substitute {
+    # Use absolute paths that are discoverable outside of the nix shell
+    src = ../../CMake/toolchain_mingw.cmake;
+    substitutions = ["--replace-fail" "\${MINGW_TRIPLE}-" "${toolchain}/bin/\${MINGW_TRIPLE}-"];
+  };
+
+  mesonCrossFile = pkgs.writeText "crossfile_llvm_mingw.txt" ''
+    [binaries]
+    ar = '${toolchain}/bin/arm64ec-w64-mingw32-ar'
+    c = '${toolchain}/bin/arm64ec-w64-mingw32-gcc'
+    cpp = '${toolchain}/bin/arm64ec-w64-mingw32-g++'
+    ld = '${toolchain}/bin/arm64ec-w64-mingw32-ld'
+    windres = '${toolchain}/bin/arm64ec-w64-mingw32-windres'
+    strip = '${toolchain}/bin/strip'
+    widl = '${toolchain}/bin/arm64ec-w64-mingw32-widl'
+    pkgconfig = 'aarch64-linux-gnu-pkg-config'
+    [host_machine]
+    system = 'windows'
+    cpu_family = 'aarch64'
+    cpu = 'aarch64'
+    endian = 'little'
+  '';
+in
+pkgs.mkShell {
+  buildInputs = [
+    toolchain
+  ];
+
+  shellHook = ''
+    if [[ $- == *i* ]]; then
+      echo "llvm-mingw set up at ${toolchain}."
+      echo ""
+      echo "To configure DXVK/vkd3d-proton: meson setup \$FEX_MESON_CROSSFILE"
+      echo ""
+      echo "To configure 32-bit FEX build: cmake \$FEX_CMAKE_TOOLCHAIN_WOW64"
+      echo "To configure 64-bit FEX build: cmake \$FEX_CMAKE_TOOLCHAIN_ARM64EC"
+    fi
+  '';
+
+  # E.g. cmake $FEX_CMAKE_TOOLCHAIN_ARM64EC -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr -DENABLE_LTO=False -DBUILD_TESTING=False
+  FEX_CMAKE_TOOLCHAIN_ARM64EC = "--toolchain ${cmakeToolchainFile} -DMINGW_TRIPLE=arm64ec-w64-mingw32 -DCMAKE_INSTALL_LIBDIR=/usr/lib/wine/aarch64-windows";
+  FEX_CMAKE_TOOLCHAIN_WOW64 = "--toolchain ${cmakeToolchainFile} -DMINGW_TRIPLE=aarch64-w64-mingw32 -DCMAKE_INSTALL_LIBDIR=/usr/lib/wine/aarch64-windows";
+  FEX_MESON_CROSSFILE = "--cross-file ${mesonCrossFile}";
+}
@@ -0,0 +1,21 @@
+#! /usr/bin/env nix-shell
+#! nix-shell -i bash WineOnArm/shell.nix
+
+# Helper script to configure CMake for building FEX as library for emulation
+# of 32-bit applications in Wine/Proton.
+# The required cross-toolchains will be set up and managed by nix.
+
+if [ $# -eq 0 ]
+then
+  echo "Expected CMake argument list"
+  exit 1
+fi
+
+if [ -f CMakeCache.txt ]
+then
+  echo "Expected empty build folder"
+  exit 1
+fi
+
+set -o xtrace
+cmake $FEX_CMAKE_TOOLCHAIN_WOW64 -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr -DENABLE_LTO=False -DBUILD_TESTING=False $@
@@ -0,0 +1,21 @@
+#! /usr/bin/env nix-shell
+#! nix-shell -i bash WineOnArm/shell.nix
+
+# Helper script to configure CMake for building FEX as library for emulation
+# of 64-bit applications in Wine/Proton
+# Nix is used to install and manage the required cross-toolchains.
+
+if [ $# -eq 0 ]
+then
+  echo "Expected CMake argument list"
+  exit 1
+fi
+
+if [ -f CMakeCache.txt ]
+then
+  echo "Expected empty build folder"
+  exit 1
+fi
+
+set -o xtrace
+cmake $FEX_CMAKE_TOOLCHAIN_ARM64EC -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/usr -DENABLE_LTO=False -DBUILD_TESTING=False $@
@@ -0,0 +1,17 @@
+#! /usr/bin/env nix-shell
+#! nix-shell -i bash FEXLinuxTests/shell.nix
+
+# Helper script to configure CMake for building FEXLinuxTests.
+# Nix is used to install and manage the required cross-toolchains.
+
+if [ ! -f CMakeCache.txt ]
+then
+  echo "Must be run from a pre-configured CMake build folder"
+  exit 1
+fi
+
+# Remove previous build to ensure the new toolchain is applied
+rm -rf unittests/FEXLinuxTests
+
+set -o xtrace
+cmake . $FEX_CMAKE_TOOLCHAINS -DBUILD_TESTING=ON -DBUILD_FEX_LINUX_TESTS=ON
@@ -0,0 +1,22 @@
+# Helper script to configure CMake for library forwarding in FEX.
+# Nix is used to install and manage the required cross-toolchains.
+
+if [ ! -f CMakeCache.txt ]
+then
+  echo "Must be run from a pre-configured CMake build folder"
+  exit 1
+fi
+
+# Remove previous build to ensure the new toolchain is applied
+rm -rf guest-libs guest-libs-32 Guest Guest_32
+
+# Set clang executable path manually since the one from the nix store
+# will be picked up otherwise
+CLANG_EXEC_PATH=""
+if ! grep -q CLANG_EXEC_PATH CMakeCache.txt
+then
+  CLANG_EXEC_PATH="-DCLANG_EXEC_PATH=`which clang`"
+fi
+
+nix-shell `dirname -- "$0"`/LibraryForwarding/shell.nix \
+  --run "set -o xtrace; cmake . \$FEX_CMAKE_TOOLCHAINS -DBUILD_THUNKS=ON $CLANG_EXEC_PATH; set +o xtrace"
@@ -1,31 +0,0 @@
-# --- Stage 1: Builder ---
-FROM ubuntu:20.04 as builder
-
-RUN DEBIAN_FRONTEND="noninteractive" apt-get update
-RUN DEBIAN_FRONTEND="noninteractive" apt install -y cmake \
-clang-10 llvm-10 nasm ninja-build pkg-config \
-libcap-dev libglfw3-dev libepoxy-dev python3-dev libsdl2-dev \
-python3 linux-headers-generic \
-git
-
-RUN git clone --recurse-submodules https://github.com/FEX-Emu/FEX.git
-
-CMD [ "mkdir /opt/FEX/build" ]
-
-WORKDIR /opt/FEX/build
-
-ARG CC=clang-10
-ARG CXX=clang++-10
-RUN cmake -G Ninja .. -DCMAKE_BUILD_TYPE=Release
-RUN ninja
-
-# --- Stage 2: Runner ---
-FROM ubuntu:20.04
-
-RUN DEBIAN_FRONTEND="noninteractive" apt-get update
-RUN DEBIAN_FRONTEND="noninteractive" apt install -y \
-libcap-dev libglfw3-dev libepoxy-dev
-
-COPY --from=builder /opt/FEX/build/Bin/* /usr/bin/
-
-WORKDIR /root
@@ -0,0 +1 @@
+DisableFormat: true
@@ -0,0 +1,97 @@
+
+add_library(softfloat_3e STATIC
+  # F80 support
+  src/extF80_add.c
+  src/extF80_div.c
+  src/extF80_sub.c
+  src/extF80_mul.c
+  src/extF80_rem.c
+  src/extF80_sqrt.c
+  src/extF80_le.c
+  src/extF80_to_i32.c
+  src/extF80_to_i64.c
+  src/extF80_to_ui64.c
+  src/extF80_to_f32.c
+  src/extF80_to_f64.c
+  src/i32_to_extF80.c
+  src/ui64_to_extF80.c
+  src/extF80_to_f128.c
+  src/f128_to_extF80.c
+
+  # F128 support
+  src/f128_add.c
+  src/f128_div.c
+  src/f128_eq.c
+  src/f128_eq_signaling.c
+  src/f128_isSignalingNaN.c
+  src/f128_le.c
+  src/f128_le_quiet.c
+  src/f128_lt.c
+  src/f128_lt_quiet.c
+  src/f128_mulAdd.c
+  src/f128_mul.c
+  src/f128_rem.c
+  src/f128_sqrt.c
+  src/f128_sub.c
+  src/f128_to_f16.c
+  src/f128_to_f32.c
+  src/f128_to_f64.c
+  src/f128_to_i32.c
+  src/f128_to_i64.c
+  src/f128_to_ui32.c
+  src/f128_to_ui64.c
+  src/s_addMagsF128.c
+  src/s_subMagsF128.c
+  src/s_normRoundPackToF128.c
+  src/s_roundPackToF128.c
+  src/s_propagateNaNF128UI.c
+
+  # Conversion
+  src/f32_to_f128.c
+  src/i32_to_f128.c
+
+  src/s_roundToUI64.c
+  src/s_f128UIToCommonNaN.c
+  src/s_commonNaNToF128UI.c
+  src/s_normSubnormalF128Sig.c
+  src/s_roundToI32.c
+  src/s_roundToI64.c
+  src/s_roundPackToF32.c
+  src/s_addMagsExtF80.c
+  src/s_extF80UIToCommonNaN.c
+  src/s_commonNaNToF32UI.c
+  src/s_commonNaNToF64UI.c
+  src/s_roundPackToF64.c
+  src/s_propagateNaNExtF80UI.c
+  src/s_roundPackToExtF80.c
+  src/s_normSubnormalExtF80Sig.c
+  src/s_subMagsExtF80.c
+  src/s_shiftRightJam128.c
+  src/s_shiftRightJam128Extra.c
+  src/s_normRoundPackToExtF80.c
+  src/s_approxRecip_1Ks.c
+  src/s_approxRecipSqrt32_1.c
+  src/s_approxRecipSqrt_1Ks.c
+  src/softfloat_raiseFlags.c
+  src/f64_to_extF80.c
+  src/s_commonNaNToExtF80UI.c
+  src/s_normSubnormalF64Sig.c
+  src/s_f64UIToCommonNaN.c
+  src/extF80_roundToInt.c
+  src/extF80_eq.c
+  src/extF80_lt.c
+  src/f32_to_extF80.c
+  src/s_normSubnormalF32Sig.c
+  src/s_f32UIToCommonNaN.c)
+
+if (ARCHITECTURE_arm64 AND HAS_CLANG_PRESERVE_ALL)
+  list(APPEND DEFINES "-DFEXCORE_PRESERVE_ALL_ATTR=__attribute__((preserve_all));-DFEXCORE_HAS_PRESERVE_ALL_ATTR=1")
+else()
+  list(APPEND DEFINES "-DFEXCORE_PRESERVE_ALL_ATTR=;-DFEXCORE_HAS_PRESERVE_ALL_ATTR=0")
+endif()
+
+list(APPEND DEFINES "-DSOFTFLOAT_BUILTIN_CLZ=1;-DINLINE=static inline;-DINLINE_LEVEL=4;-DSOFTFLOAT_FAST_INT64=1;-DSOFTFLOAT_FAST_DIV32TO16=1;-DSOFTFLOAT_FAST_DIV64TO32=1")
+
+target_include_directories(softfloat_3e PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include/)
+target_include_directories(softfloat_3e PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include/SoftFloat-3e/)
+target_compile_definitions(softfloat_3e PUBLIC ${DEFINES})
@@ -149,7 +149,7 @@ float64_t f32_to_f64( float32_t );
 #ifdef SOFTFLOAT_FAST_INT64
 FEXCORE_PRESERVE_ALL_ATTR
 extFloat80_t f32_to_extF80( struct softfloat_state *, float32_t );
-float128_t f32_to_f128( float32_t );
+float128_t f32_to_f128( struct softfloat_state *, float32_t );
 #endif
 void f32_to_extF80M( float32_t, extFloat80_t * );
 void f32_to_f128M( float32_t, float128_t * );
@@ -243,13 +243,17 @@ FEXCORE_PRESERVE_ALL_ATTR
 extFloat80_t extF80_sqrt( struct softfloat_state *, extFloat80_t );
 FEXCORE_PRESERVE_ALL_ATTR
 bool extF80_eq( struct softfloat_state *, extFloat80_t, extFloat80_t );
-bool extF80_le( extFloat80_t, extFloat80_t );
+bool extF80_le( struct softfloat_state *, extFloat80_t, extFloat80_t );
 FEXCORE_PRESERVE_ALL_ATTR
 bool extF80_lt( struct softfloat_state *, extFloat80_t, extFloat80_t );
 bool extF80_eq_signaling( extFloat80_t, extFloat80_t );
 bool extF80_le_quiet( extFloat80_t, extFloat80_t );
 bool extF80_lt_quiet( extFloat80_t, extFloat80_t );
 bool extF80_isSignalingNaN( extFloat80_t );
+static inline extFloat80_t extF80_complement_sign(extFloat80_t a) {
+  a.signExp ^= 1ULL << 15;
+  return a;
+}
 #endif
 uint_fast32_t extF80M_to_ui32( const extFloat80_t *, uint_fast8_t, bool );
 uint_fast64_t extF80M_to_ui64( const extFloat80_t *, uint_fast8_t, bool );
@@ -284,34 +288,38 @@ bool extF80M_isSignalingNaN( const extFloat80_t * );
 | 128-bit (quadruple-precision) floating-point operations.
 *----------------------------------------------------------------------------*/
 #ifdef SOFTFLOAT_FAST_INT64
-uint_fast32_t f128_to_ui32( float128_t, uint_fast8_t, bool );
-uint_fast64_t f128_to_ui64( float128_t, uint_fast8_t, bool );
-int_fast32_t f128_to_i32( float128_t, uint_fast8_t, bool );
-int_fast64_t f128_to_i64( float128_t, uint_fast8_t, bool );
+uint_fast32_t f128_to_ui32( struct softfloat_state *, float128_t, uint_fast8_t, bool );
+uint_fast64_t f128_to_ui64( struct softfloat_state *, float128_t, uint_fast8_t, bool );
+int_fast32_t f128_to_i32( struct softfloat_state *, float128_t, uint_fast8_t, bool );
+int_fast64_t f128_to_i64( struct softfloat_state *, float128_t, uint_fast8_t, bool );
 uint_fast32_t f128_to_ui32_r_minMag( float128_t, bool );
 uint_fast64_t f128_to_ui64_r_minMag( float128_t, bool );
 int_fast32_t f128_to_i32_r_minMag( float128_t, bool );
 int_fast64_t f128_to_i64_r_minMag( float128_t, bool );
-float16_t f128_to_f16( float128_t );
-float32_t f128_to_f32( float128_t );
-float64_t f128_to_f64( float128_t );
+float16_t f128_to_f16( struct softfloat_state *, float128_t );
+float32_t f128_to_f32( struct softfloat_state *, float128_t );
+float64_t f128_to_f64( struct softfloat_state *, float128_t );
 FEXCORE_PRESERVE_ALL_ATTR
 extFloat80_t f128_to_extF80( struct softfloat_state *, float128_t );
 float128_t f128_roundToInt( float128_t, uint_fast8_t, bool );
-float128_t f128_add( float128_t, float128_t );
-float128_t f128_sub( float128_t, float128_t );
-float128_t f128_mul( float128_t, float128_t );
-float128_t f128_mulAdd( float128_t, float128_t, float128_t );
-float128_t f128_div( float128_t, float128_t );
-float128_t f128_rem( float128_t, float128_t );
-float128_t f128_sqrt( float128_t );
-bool f128_eq( float128_t, float128_t );
-bool f128_le( float128_t, float128_t );
-bool f128_lt( float128_t, float128_t );
-bool f128_eq_signaling( float128_t, float128_t );
-bool f128_le_quiet( float128_t, float128_t );
-bool f128_lt_quiet( float128_t, float128_t );
+float128_t f128_add( struct softfloat_state *, float128_t, float128_t );
+float128_t f128_sub( struct softfloat_state *, float128_t, float128_t );
+float128_t f128_mul( struct softfloat_state *, float128_t, float128_t );
+float128_t f128_mulAdd( struct softfloat_state *, float128_t, float128_t, float128_t );
+float128_t f128_div( struct softfloat_state *, float128_t, float128_t );
+float128_t f128_rem( struct softfloat_state *, float128_t, float128_t );
+float128_t f128_sqrt( struct softfloat_state *, float128_t );
+bool f128_eq( struct softfloat_state *, float128_t, float128_t );
+bool f128_le( struct softfloat_state *, float128_t, float128_t );
+bool f128_lt( struct softfloat_state *, float128_t, float128_t );
+bool f128_eq_signaling( struct softfloat_state *, float128_t, float128_t );
+bool f128_le_quiet( struct softfloat_state *, float128_t, float128_t );
+bool f128_lt_quiet( struct softfloat_state *, float128_t, float128_t );
 bool f128_isSignalingNaN( float128_t );
+static inline float128_t f128_complement_sign(float128_t a) {
+  a.v[1] ^= 1ULL << 63;
+  return a;
+}
 #endif
 uint_fast32_t f128M_to_ui32( const float128_t *, uint_fast8_t, bool );
 uint_fast64_t f128M_to_ui64( const float128_t *, uint_fast8_t, bool );
@@ -0,0 +1,73 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool extF80_le( struct softfloat_state *state, extFloat80_t a, extFloat80_t b )
+{
+    union { struct extFloat80M s; extFloat80_t f; } uA;
+    uint_fast16_t uiA64;
+    uint_fast64_t uiA0;
+    union { struct extFloat80M s; extFloat80_t f; } uB;
+    uint_fast16_t uiB64;
+    uint_fast64_t uiB0;
+    bool signA, signB;
+
+    uA.f = a;
+    uiA64 = uA.s.signExp;
+    uiA0  = uA.s.signif;
+    uB.f = b;
+    uiB64 = uB.s.signExp;
+    uiB0  = uB.s.signif;
+    if ( isNaNExtF80UI( uiA64, uiA0 ) || isNaNExtF80UI( uiB64, uiB0 ) ) {
+        softfloat_raiseFlags( state, softfloat_flag_invalid );
+        return false;
+    }
+    signA = signExtF80UI64( uiA64 );
+    signB = signExtF80UI64( uiB64 );
+    return
+        (signA != signB)
+            ? signA || ! (((uiA64 | uiB64) & 0x7FFF) | uiA0 | uiB0)
+            : ((uiA64 == uiB64) && (uiA0 == uiB0))
+                  || (signA ^ softfloat_lt128( uiA64, uiA0, uiB64, uiB0 ));
+
+}
+
@@ -0,0 +1,78 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "softfloat.h"
+
+float128_t f128_add( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    bool signA;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signB;
+#if ! defined INLINE_LEVEL || (INLINE_LEVEL < 2)
+    float128_t
+        (*magsFuncPtr)(
+            uint_fast64_t, uint_fast64_t, uint_fast64_t, uint_fast64_t, bool );
+#endif
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    signA = signF128UI64( uiA64 );
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    signB = signF128UI64( uiB64 );
+#if defined INLINE_LEVEL && (2 <= INLINE_LEVEL)
+    if ( signA == signB ) {
+        return softfloat_addMagsF128( state, uiA64, uiA0, uiB64, uiB0, signA );
+    } else {
+        return softfloat_subMagsF128( state, uiA64, uiA0, uiB64, uiB0, signA );
+    }
+#else
+    magsFuncPtr =
+        (signA == signB) ? softfloat_addMagsF128 : softfloat_subMagsF128;
+    return (*magsFuncPtr)( uiA64, uiA0, uiB64, uiB0, signA );
+#endif
+
+}
+
@@ -0,0 +1,199 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+float128_t f128_div( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    bool signA;
+    int_fast32_t expA;
+    struct uint128 sigA;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signB;
+    int_fast32_t expB;
+    struct uint128 sigB;
+    bool signZ;
+    struct exp32_sig128 normExpSig;
+    int_fast32_t expZ;
+    struct uint128 rem;
+    uint_fast32_t recip32;
+    int ix;
+    uint_fast64_t q64;
+    uint_fast32_t q;
+    struct uint128 term;
+    uint_fast32_t qs[3];
+    uint_fast64_t sigZExtra;
+    struct uint128 sigZ, uiZ;
+    union ui128_f128 uZ;
+
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    signA = signF128UI64( uiA64 );
+    expA  = expF128UI64( uiA64 );
+    sigA.v64 = fracF128UI64( uiA64 );
+    sigA.v0  = uiA0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    signB = signF128UI64( uiB64 );
+    expB  = expF128UI64( uiB64 );
+    sigB.v64 = fracF128UI64( uiB64 );
+    sigB.v0  = uiB0;
+    signZ = signA ^ signB;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    if ( expA == 0x7FFF ) {
+        if ( sigA.v64 | sigA.v0 ) goto propagateNaN;
+        if ( expB == 0x7FFF ) {
+            if ( sigB.v64 | sigB.v0 ) goto propagateNaN;
+            goto invalid;
+        }
+        goto infinity;
+    }
+    if ( expB == 0x7FFF ) {
+        if ( sigB.v64 | sigB.v0 ) goto propagateNaN;
+        goto zero;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    if ( ! expB ) {
+        if ( ! (sigB.v64 | sigB.v0) ) {
+            if ( ! (expA | sigA.v64 | sigA.v0) ) goto invalid;
+            softfloat_raiseFlags( state, softfloat_flag_infinite );
+            goto infinity;
+        }
+        normExpSig = softfloat_normSubnormalF128Sig( sigB.v64, sigB.v0 );
+        expB = normExpSig.exp;
+        sigB = normExpSig.sig;
+    }
+    if ( ! expA ) {
+        if ( ! (sigA.v64 | sigA.v0) ) goto zero;
+        normExpSig = softfloat_normSubnormalF128Sig( sigA.v64, sigA.v0 );
+        expA = normExpSig.exp;
+        sigA = normExpSig.sig;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    expZ = expA - expB + 0x3FFE;
+    sigA.v64 |= UINT64_C( 0x0001000000000000 );
+    sigB.v64 |= UINT64_C( 0x0001000000000000 );
+    rem = sigA;
+    if ( softfloat_lt128( sigA.v64, sigA.v0, sigB.v64, sigB.v0 ) ) {
+        --expZ;
+        rem = softfloat_add128( sigA.v64, sigA.v0, sigA.v64, sigA.v0 );
+    }
+    recip32 = softfloat_approxRecip32_1( sigB.v64>>17 );
+    ix = 3;
+    for (;;) {
+        q64 = (uint_fast64_t) (uint32_t) (rem.v64>>19) * recip32;
+        q = (q64 + 0x80000000)>>32;
+        --ix;
+        if ( ix < 0 ) break;
+        rem = softfloat_shortShiftLeft128( rem.v64, rem.v0, 29 );
+        term = softfloat_mul128By32( sigB.v64, sigB.v0, q );
+        rem = softfloat_sub128( rem.v64, rem.v0, term.v64, term.v0 );
+        if ( rem.v64 & UINT64_C( 0x8000000000000000 ) ) {
+            --q;
+            rem = softfloat_add128( rem.v64, rem.v0, sigB.v64, sigB.v0 );
+        }
+        qs[ix] = q;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    if ( ((q + 1) & 7) < 2 ) {
+        rem = softfloat_shortShiftLeft128( rem.v64, rem.v0, 29 );
+        term = softfloat_mul128By32( sigB.v64, sigB.v0, q );
+        rem = softfloat_sub128( rem.v64, rem.v0, term.v64, term.v0 );
+        if ( rem.v64 & UINT64_C( 0x8000000000000000 ) ) {
+            --q;
+            rem = softfloat_add128( rem.v64, rem.v0, sigB.v64, sigB.v0 );
+        } else if ( softfloat_le128( sigB.v64, sigB.v0, rem.v64, rem.v0 ) ) {
+            ++q;
+            rem = softfloat_sub128( rem.v64, rem.v0, sigB.v64, sigB.v0 );
+        }
+        if ( rem.v64 | rem.v0 ) q |= 1;
+    }
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+    sigZExtra = (uint64_t) ((uint_fast64_t) q<<60);
+    term = softfloat_shortShiftLeft128( 0, qs[1], 54 );
+    sigZ =
+        softfloat_add128(
+            (uint_fast64_t) qs[2]<<19, ((uint_fast64_t) qs[0]<<25) + (q>>4),
+            term.v64, term.v0
+        );
+    return
+        softfloat_roundPackToF128( state, signZ, expZ, sigZ.v64, sigZ.v0, sigZExtra );
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+ propagateNaN:
+    uiZ = softfloat_propagateNaNF128UI( state, uiA64, uiA0, uiB64, uiB0 );
+    goto uiZ;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+ invalid:
+    softfloat_raiseFlags( state, softfloat_flag_invalid );
+    uiZ.v64 = defaultNaNF128UI64;
+    uiZ.v0  = defaultNaNF128UI0;
+    goto uiZ;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+ infinity:
+    uiZ.v64 = packToF128UI64( signZ, 0x7FFF, 0 );
+    goto uiZ0;
+    /*------------------------------------------------------------------------
+    *------------------------------------------------------------------------*/
+ zero:
+    uiZ.v64 = packToF128UI64( signZ, 0, 0 );
+ uiZ0:
+    uiZ.v0 = 0;
+ uiZ:
+    uZ.ui = uiZ;
+    return uZ.f;
+
+}
+
@@ -0,0 +1,73 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool f128_eq( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        if (
+               softfloat_isSigNaNF128UI( uiA64, uiA0 )
+            || softfloat_isSigNaNF128UI( uiB64, uiB0 )
+        ) {
+            softfloat_raiseFlags( state, softfloat_flag_invalid );
+        }
+        return false;
+    }
+    return
+           (uiA0 == uiB0)
+        && (   (uiA64 == uiB64)
+            || (! uiA0 && ! ((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF )))
+           );
+
+}
+
@@ -0,0 +1,67 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "softfloat.h"
+
+bool f128_eq_signaling( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        softfloat_raiseFlags( state, softfloat_flag_invalid );
+        return false;
+    }
+    return
+           (uiA0 == uiB0)
+        && (   (uiA64 == uiB64)
+            || (! uiA0 && ! ((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF )))
+           );
+
+}
+
@@ -0,0 +1,51 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool f128_isSignalingNaN( float128_t a )
+{
+    union ui128_f128 uA;
+
+    uA.f = a;
+    return softfloat_isSigNaNF128UI( uA.ui.v64, uA.ui.v0 );
+
+}
+
@@ -0,0 +1,72 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "softfloat.h"
+
+bool f128_le( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signA, signB;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        softfloat_raiseFlags( state, softfloat_flag_invalid );
+        return false;
+    }
+    signA = signF128UI64( uiA64 );
+    signB = signF128UI64( uiB64 );
+    return
+        (signA != signB)
+            ? signA
+                  || ! (((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF ))
+                            | uiA0 | uiB0)
+            : ((uiA64 == uiB64) && (uiA0 == uiB0))
+                  || (signA ^ softfloat_lt128( uiA64, uiA0, uiB64, uiB0 ));
+
+}
+
@@ -0,0 +1,78 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool f128_le_quiet( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signA, signB;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        if (
+               softfloat_isSigNaNF128UI( uiA64, uiA0 )
+            || softfloat_isSigNaNF128UI( uiB64, uiB0 )
+        ) {
+            softfloat_raiseFlags( state, softfloat_flag_invalid );
+        }
+        return false;
+    }
+    signA = signF128UI64( uiA64 );
+    signB = signF128UI64( uiB64 );
+    return
+        (signA != signB)
+            ? signA
+                  || ! (((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF ))
+                            | uiA0 | uiB0)
+            : ((uiA64 == uiB64) && (uiA0 == uiB0))
+                  || (signA ^ softfloat_lt128( uiA64, uiA0, uiB64, uiB0 ));
+
+}
+
@@ -0,0 +1,72 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "softfloat.h"
+
+bool f128_lt( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signA, signB;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        softfloat_raiseFlags( state, softfloat_flag_invalid );
+        return false;
+    }
+    signA = signF128UI64( uiA64 );
+    signB = signF128UI64( uiB64 );
+    return
+        (signA != signB)
+            ? signA
+                  && (((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF ))
+                          | uiA0 | uiB0)
+            : ((uiA64 != uiB64) || (uiA0 != uiB0))
+                  && (signA ^ softfloat_lt128( uiA64, uiA0, uiB64, uiB0 ));
+
+}
+
@@ -0,0 +1,78 @@
+
+/*============================================================================
+
+This C source file is part of the SoftFloat IEEE Floating-Point Arithmetic
+Package, Release 3e, by John R. Hauser.
+
+Copyright 2011, 2012, 2013, 2014 The Regents of the University of California.
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+ 1. Redistributions of source code must retain the above copyright notice,
+    this list of conditions, and the following disclaimer.
+
+ 2. Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions, and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+ 3. Neither the name of the University nor the names of its contributors may
+    be used to endorse or promote products derived from this software without
+    specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS "AS IS", AND ANY
+EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, ARE
+DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE FOR ANY
+DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
+ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+=============================================================================*/
+
+#include <stdbool.h>
+#include <stdint.h>
+#include "platform.h"
+#include "internals.h"
+#include "specialize.h"
+#include "softfloat.h"
+
+bool f128_lt_quiet( struct softfloat_state *state, float128_t a, float128_t b )
+{
+    union ui128_f128 uA;
+    uint_fast64_t uiA64, uiA0;
+    union ui128_f128 uB;
+    uint_fast64_t uiB64, uiB0;
+    bool signA, signB;
+
+    uA.f = a;
+    uiA64 = uA.ui.v64;
+    uiA0  = uA.ui.v0;
+    uB.f = b;
+    uiB64 = uB.ui.v64;
+    uiB0  = uB.ui.v0;
+    if ( isNaNF128UI( uiA64, uiA0 ) || isNaNF128UI( uiB64, uiB0 ) ) {
+        if (
+               softfloat_isSigNaNF128UI( uiA64, uiA0 )
+            || softfloat_isSigNaNF128UI( uiB64, uiB0 )
+        ) {
+            softfloat_raiseFlags( state, softfloat_flag_invalid );
+        }
+        return false;
+    }
+    signA = signF128UI64( uiA64 );
+    signB = signF128UI64( uiB64 );
+    return
+        (signA != signB)
+            ? signA
+                  && (((uiA64 | uiB64) & UINT64_C( 0x7FFFFFFFFFFFFFFF ))
+                          | uiA0 | uiB0)
+            : ((uiA64 != uiB64) || (uiA0 != uiB0))
+                  && (signA ^ softfloat_lt128( uiA64, uiA0, uiB64, uiB0 ));
+
+}
+
				`@@ -0,0 +1 @@`
				`AI must not be used to generate code for contributions to this project.`