Docs: Update for release FEX-2212

Merge pull request #2198 from lioncash/add
OpcodeDispatcher: Handle VADDPD/VADDPS/VPADDB/VPADDW/VPADDD/VPADDQ
2026-10-07 12:00:17 +02:00 · 2022-12-05 14:02:57 -08:00 · 2022-12-05 13:19:30 -08:00 · 2022-12-05 17:53:33 +00:00 · 2022-12-05 17:44:12 +00:00 · 2022-12-05 17:33:52 +00:00
509 changed files with 26064 additions and 11178 deletions
@@ -40,7 +40,6 @@ jobs:
      # 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
      run: $GITHUB_WORKSPACE/Scripts/CI_FetchRootFS.py

    - name : submodule checkout
@@ -65,7 +64,7 @@ jobs:
      # 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 -DENABLE_INTERPRETER=True -DBUILD_FEX_LINUX_TESTS=True
+      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE -G Ninja -DENABLE_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True -DENABLE_INTERPRETER=True -DBUILD_FEX_LINUX_TESTS=True -DBUILD_THUNKS=True -DCMAKE_INSTALL_PREFIX=${{runner.workspace}}/build/install

    - name: Build
      working-directory: ${{runner.workspace}}/build
@@ -178,6 +177,51 @@ jobs:
      working-directory: ${{runner.workspace}}/build
      run: mv ${{runner.workspace}}/build/Testing/Temporary/LastTest.log ${{runner.workspace}}/build/Testing/Temporary/LastTest_FEXLinuxTests.log || true

+    - 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
+
+    - name: Install
+      if: matrix.arch[1] == 'x64'
+      working-directory: ${{runner.workspace}}/build
+      shell: bash
+      run: cmake --build . --config $BUILD_TYPE --target install
+
+    - name: Test GL No-Thunks
+      if: matrix.arch[1] == 'x64'
+      working-directory: ${{runner.workspace}}/build
+      shell: bash
+      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
+
+    - name: Test GL Thunks
+      if: matrix.arch[1] == 'x64'
+      working-directory: ${{runner.workspace}}/build
+      shell: bash
+      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
+
    - name: Truncate test results
      if: ${{ always() }}
      shell: bash
@@ -197,4 +241,3 @@ jobs:
        name: Results-${{ env.runner_name }}
        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
        retention-days: 3
-
@@ -0,0 +1,118 @@
+name: Vixl Simulator run
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - main
+
+env:
+  # Customize the CMake build type here (Release, Debug, RelWithDebInfo, etc.)
+  BUILD_TYPE: Release
+  CC: clang
+  CXX: clang++
+
+jobs:
+  build:
+    runs-on: ${{ matrix.arch }}
+    strategy:
+      matrix:
+        # Only the x86-64 runner is fast enough to run this
+        arch: [[self-hosted, x64], [self-hosted, ARMv8.4]]
+      fail-fast: false
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Set runner label
+      run: echo "runner_label=${{ matrix.arch[1] }}" >> $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: 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_LTO=False -DENABLE_ASSERTIONS=True -DENABLE_X86_HOST_DEBUG=True
+
+    - 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: 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: IR Tests
+      working-directory: ${{runner.workspace}}/build
+      shell: bash
+      # Execute the unit tests
+      run: cmake --build . --config $BUILD_TYPE --target ir_tests
+
+    - name: IR 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_IR.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
+
+    - name: Upload results
+      if: ${{ always() }}
+      uses: 'actions/upload-artifact@v2'
+      with:
+        name: Results-${{ env.runner_name }}
+        path: ${{runner.workspace}}/build/Testing/Temporary/LastTest_*.log
+        retention-days: 3
+
@@ -10,3 +10,4 @@ out/
 .vscode/
 .vs/
 *.pyc
+.cache
@@ -49,3 +49,7 @@
 	shallow = true
 	path = External/robin-map
 	url = https://github.com/Tessil/robin-map.git
+[submodule "External/Vulkan-Headers"]
+	shallow = true
+	path = External/Vulkan-Headers
+	url = https://github.com/KhronosGroup/Vulkan-Headers.git
@@ -0,0 +1,5 @@
+{
+  "ThunksDB": {
+    "GL": 1
+  }
+}
@@ -0,0 +1,5 @@
+{
+  "ThunksDB": {
+    "Vulkan": 1
+  }
+}
@@ -0,0 +1,21 @@
+if(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
+  message(FATAL_ERROR "Cannot find install manifest: @CMAKE_BINARY_DIR@/install_manifest.txt")
+endif()
+
+file(READ "@CMAKE_BINARY_DIR@/install_manifest.txt" files)
+string(REGEX REPLACE "\n" ";" files "${files}")
+foreach(file ${files})
+  message(STATUS "Uninstalling $ENV{DESTDIR}${file}")
+  if(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
+    exec_program(
+      "@CMAKE_COMMAND@" ARGS "-E remove \"$ENV{DESTDIR}${file}\""
+      OUTPUT_VARIABLE rm_out
+      RETURN_VALUE rm_retval
+      )
+    if(NOT "${rm_retval}" STREQUAL 0)
+      message(FATAL_ERROR "Problem when removing $ENV{DESTDIR}${file}")
+    endif()
+  else(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
+    message(STATUS "File $ENV{DESTDIR}${file} does not exist.")
+  endif()
+endforeach()
@@ -7,6 +7,7 @@ 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_THUNKS "Build thunks" FALSE)
+option(ENABLE_CLANG_THUNKS "Build thunks with clang" FALSE)
 option(ENABLE_CLANG_FORMAT "Run clang format over the source" FALSE)
 option(ENABLE_IWYU "Enables include what you use program" FALSE)
 option(ENABLE_LTO "Enable LTO with compilation" TRUE)
@@ -20,7 +21,6 @@ option(ENABLE_GDB_SYMBOLS "Enables GDBSymbols integration support" ${HAVE_GDB_JI
 option(ENABLE_VISUAL_DEBUGGER "Enables the visual debugger for compiling" FALSE)
 option(ENABLE_STRICT_WERROR "Enables stricter -Werror for CI" FALSE)
 option(ENABLE_WERROR "Enables -Werror" FALSE)
-option(ENABLE_STATIC_PIE "Enables static-pie build" FALSE)
 option(ENABLE_JEMALLOC "Enables jemalloc allocator" TRUE)
 option(ENABLE_OFFLINE_TELEMETRY "Enables FEX offline telemetry" TRUE)
 option(ENABLE_COMPILE_TIME_TRACE "Enables time trace compile option" FALSE)
@@ -28,10 +28,36 @@ option(ENABLE_LIBCXX "Enables LLVM libc++" FALSE)
 option(ENABLE_INTERPRETER "Enables FEX's Interpreter" FALSE)
 option(ENABLE_CCACHE "Enables ccache for compile caching" TRUE)
 option(ENABLE_TERMUX_BUILD "Forces building for Termux on a non-Termux build machine" FALSE)
+option(ENABLE_VIXL_SIMULATOR "Forces the FEX JIT to use the VIXL simulator" 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 (X86_TOOLCHAIN_FILE "${CMAKE_CURRENT_SOURCE_DIR}/toolchain_x86.cmake" CACHE FILEPATH "Toolchain file for the x86 (cross-)compiler")
+set (X86_32_TOOLCHAIN_FILE "${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")

+if (ENABLE_FEXCORE_PROFILER)
+  add_definitions(-DENABLE_FEXCORE_PROFILER=1)
+  string(TOUPPER "${FEXCORE_PROFILER_BACKEND}" FEXCORE_PROFILER_BACKEND)
+
+  if (FEXCORE_PROFILER_BACKEND STREQUAL "GPUVIS")
+    add_definitions(-DFEXCORE_PROFILER_BACKEND=1)
+  else()
+    message(FATAL_ERROR "Unknown FEXCore profiler backend ${FEXCORE_PROFILER_BACKEND}")
+  endif()
+endif()
+
+# uninstall target
+if(NOT TARGET uninstall)
+  configure_file(
+    "${CMAKE_CURRENT_SOURCE_DIR}/CMakeFiles/cmake_uninstall.cmake.in"
+    "${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/cmake_uninstall.cmake"
+    IMMEDIATE @ONLY)
+
+  add_custom_target(uninstall
+    COMMAND ${CMAKE_COMMAND} -P ${CMAKE_CURRENT_BINARY_DIR}/CMakeFiles/cmake_uninstall.cmake)
+endif()
+
 # These options are meant for package management
 set (TUNE_CPU "native" CACHE STRING "Override the CPU the build is tuned for")
 set (TUNE_ARCH "generic" CACHE STRING "Override the Arch the build is tuned for")
@@ -85,10 +111,9 @@ if (CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64")
  set(_M_X86_64 1)
  add_definitions(-D_M_X86_64=1)
  set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcx16")
-  set (X86_TOOLCHAIN_FILE "")
 endif()

-if (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
+if (CMAKE_SYSTEM_PROCESSOR MATCHES "^aarch64|^arm64|^armv8\.*")
  set(_M_ARM_64 1)
  add_definitions(-D_M_ARM_64=1)
 endif()
@@ -140,130 +165,6 @@ if(DEFINED ENV{TERMUX_VERSION} OR ENABLE_TERMUX_BUILD)
  set(ENABLE_JEMALLOC FALSE)
 endif()

-if (ENABLE_STATIC_PIE)
-  if (_M_ARM_64 AND ENABLE_LLD)
-    message (FATAL_ERROR "Static linking does not currently work with AArch64+lld. Use GNU ld for now.")
-  endif()
-
-  file(WRITE ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/Determine_iplt.c
-    "int main(int argc, char* argv[])
-      {
-        return 0;
-      }")
-
-  # Compile the test application with our LD_OVERRIDE and static-pie options
-  try_compile(
-    COMPILE_RESULT
-    ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp
-    ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/Determine_iplt.c
-    COMPILE_DEFINITIONS "-fPIE ${LD_OVERRIDE}"
-    LINK_LIBRARIES "-static-pie ${LD_OVERRIDE}"
-    COPY_FILE ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/Determine_iplt
-  )
-
-  if (${COMPILE_RESULT})
-    # Read the symbols from the elf
-    execute_process(COMMAND
-      readelf -s ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/Determine_iplt
-      OUTPUT_FILE ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/plt_out.txt
-      OUTPUT_VARIABLE PLT_SYMBOLS)
-
-    # Pull out the __rela_iplt_{start,end} symbols if they exist
-    execute_process(COMMAND
-      "grep" "__rela_iplt" ${PROJECT_BINARY_DIR}/CMakeFiles/CMakeTmp/plt_out.txt
-      OUTPUT_VARIABLE PLT_SYMBOLS)
-
-    set (SYMBOLS_FINE TRUE)
-    set (HAS_IPLT -1)
-    # Check if we have any symbols in our grep output
-    # The symbols must either not exist at all OR the symbols are zero
-    if (PLT_SYMBOLS)
-      string(FIND ${PLT_SYMBOLS} "__rela_iplt_start" HAS_IPLT)
-    endif()
-
-    if (NOT HAS_IPLT EQUAL -1)
-      # We have some symbols from readelf. Let's parse the results to check if they are zero
-      # Format: '35: 0000000000000000     0 NOTYPE  LOCAL  HIDDEN   UND __rela_iplt_start'
-      string(REPLACE "\n" ";" SYMBOL_LIST ${PLT_SYMBOLS})
-      foreach (SYMBOL ${SYMBOL_LIST})
-        # strip any leading and trailing whitespace
-        string (STRIP ${SYMBOL} SYMBOL)
-        # Convert string to a list
-        string(REPLACE " " ";" SYMBOL_VALUES ${SYMBOL}})
-        # Pull out the address argument
-        list(GET SYMBOL_VALUES 1 OFFSET)
-
-        # Check against integer zero
-        if (NOT ${OFFSET} EQUAL 0)
-          # Symbol wasn't zero, this now fails
-          set (SYMBOLS_FINE FALSE)
-        endif()
-      endforeach()
-    endif()
-
-    if (SYMBOLS_FINE)
-      # We can now exnable static-pie
-      set (STATIC_PIE_OPTIONS "-static-pie")
-      # Pthreads has an issue with exposing symbols
-      # We need to make some concessions to the pthread gods
-      if (ENABLE_LLD)
-        set (PTHREAD_LIB
-          -Wl,--undefined-glob=pthread_*
-          -Wl,--undefined=__cxa_finalize
-          -Wl,--undefined=_pthread_cleanup_push_defer
-          -Wl,--undefined=_pthread_cleanup_pop_restore
-          -Wl,--undefined=__pthread_cleanup_upto
-          pthread)
-      else()
-        set (PTHREAD_LIB
-          -Wl,--undefined=pthread_join
-          -Wl,--undefined=pthread_attr_getdetachstate
-          -Wl,--undefined=pthread_sigmask
-          -Wl,--undefined=pthread_mutex_lock
-          -Wl,--undefined=pthread_cond_init
-          -Wl,--undefined=pthread_attr_init
-          -Wl,--undefined=pthread_mutex_unlock
-          -Wl,--undefined=pthread_mutexattr_destroy
-          -Wl,--undefined=pthread_detach
-          -Wl,--undefined=pthread_mutex_init
-          -Wl,--undefined=pthread_getattr_np
-          -Wl,--undefined=pthread_cond_timedwait
-          -Wl,--undefined=pthread_attr_destroy
-          -Wl,--undefined=pthread_mutexattr_settype
-          -Wl,--undefined=pthread_rwlock_unlock
-          -Wl,--undefined=pthread_rwlock_wrlock
-          -Wl,--undefined=pthread_setspecific
-          -Wl,--undefined=pthread_create
-          -Wl,--undefined=pthread_cond_clockwait
-          -Wl,--undefined=pthread_key_create
-          -Wl,--undefined=pthread_rwlock_rdlock
-          -Wl,--undefined=pthread_setname_np
-          -Wl,--undefined=pthread_cond_signal
-          -Wl,--undefined=pthread_mutexattr_init
-          -Wl,--undefined=pthread_attr_setstack
-          -Wl,--undefined=pthread_self
-          -Wl,--undefined=pthread_getaffinity_np
-          -Wl,--undefined=pthread_cond_wait
-          -Wl,--undefined=pthread_mutex_trylock
-          -Wl,--undefined=pthread_cond_broadcast
-          -Wl,--undefined=pthread_cond_destroy
-          -Wl,--undefined=pthread_getspecific
-          -Wl,--undefined=pthread_key_delete
-          -Wl,--undefined=pthread_once
-          -Wl,--undefined=__cxa_finalize
-          -Wl,--undefined=_pthread_cleanup_push_defer
-          -Wl,--undefined=_pthread_cleanup_pop_restore
-          -Wl,--undefined=__pthread_cleanup_upto
-          pthread)
-      endif()
-    else()
-      message (FATAL_ERROR "Application has __rela_iplt_{start,end} symbols. Which means static-pie can't be enabled")
-    endif()
-  else()
-    message (FATAL_ERROR "Couldn't compile static-pie test. Static-pie can't be enabled! Is your glibc compiled without static-pie?")
-  endif()
-endif()
-
 if (ENABLE_ASAN)
  add_definitions(-DENABLE_ASAN=1)
  add_compile_options(-fno-omit-frame-pointer -fsanitize=address -fsanitize-address-use-after-scope)
@@ -488,8 +389,6 @@ if (BUILD_TESTS)
 endif()

 add_subdirectory(FEXHeaderUtils/)
-include_directories(FEXHeaderUtils/)
-
 add_subdirectory(External/FEXCore)

 # Binfmt_misc files must be installed prior to Source/ installs
@@ -499,15 +398,20 @@ add_subdirectory(Source/)
 add_subdirectory(Data/AppConfig/)

 # Install the ThunksDB file
-install(
-  FILES ${CMAKE_CURRENT_SOURCE_DIR}/Data/ThunksDB.json
-  DESTINATION ${DATA_DIRECTORY}/)
+file(GLOB CONFIG_SOURCES CONFIGURE_DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/Data/*.json)
+
+# Any application configuration json file gets installed
+foreach(CONFIG_SRC ${CONFIG_SOURCES})
+  install(FILES ${CONFIG_SRC}
+    DESTINATION ${DATA_DIRECTORY}/)
+endforeach()

 if (BUILD_TESTS)
  add_subdirectory(unittests/)
 endif()

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

  # Thunk targets for both host libraries and IDE integration
@@ -523,10 +427,31 @@ if (BUILD_THUNKS)
    SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/ThunkLibs/GuestLibs"
    BINARY_DIR "Guest"
    CMAKE_ARGS
+      "-DBITNESS=64"
      "-DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE}"
-      "-DCMAKE_TOOLCHAIN_FILE:FILEPATH=${X86_TOOLCHAIN_FILE}"
+      "-DENABLE_CLANG_THUNKS=${ENABLE_CLANG_THUNKS}"
+      "-DCMAKE_TOOLCHAIN_FILE:FILEPATH=${X86_64_TOOLCHAIN_FILE}"
      "-DCMAKE_INSTALL_PREFIX=${CMAKE_INSTALL_PREFIX}"
      "-DSTRUCT_VERIFIER=${CMAKE_SOURCE_DIR}/Scripts/StructPackVerifier.py"
+      "-DFEX_PROJECT_SOURCE_DIR=${FEX_PROJECT_SOURCE_DIR}"
+      "-DGENERATOR_EXE=$<TARGET_FILE:thunkgen>"
+    INSTALL_COMMAND ""
+    BUILD_ALWAYS ON
+    DEPENDS thunkgen
+  )
+
+  ExternalProject_Add(guest-libs-32
+    PREFIX guest-libs-32
+    SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/ThunkLibs/GuestLibs"
+    BINARY_DIR "Guest_32"
+    CMAKE_ARGS
+      "-DBITNESS=32"
+      "-DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE}"
+      "-DENABLE_CLANG_THUNKS=${ENABLE_CLANG_THUNKS}"
+      "-DCMAKE_TOOLCHAIN_FILE:FILEPATH=${X86_32_TOOLCHAIN_FILE}"
+      "-DCMAKE_INSTALL_PREFIX=${CMAKE_INSTALL_PREFIX}"
+      "-DSTRUCT_VERIFIER=${CMAKE_SOURCE_DIR}/Scripts/StructPackVerifier.py"
+      "-DFEX_PROJECT_SOURCE_DIR=${FEX_PROJECT_SOURCE_DIR}"
      "-DGENERATOR_EXE=$<TARGET_FILE:thunkgen>"
    INSTALL_COMMAND ""
    BUILD_ALWAYS ON
@@ -541,6 +466,28 @@ if (BUILD_THUNKS)
    )"
    DEPENDS guest-libs
  )
+
+  install(
+    CODE "MESSAGE(\"-- Installing: guest-libs-32\")"
+    CODE "
+    EXECUTE_PROCESS(COMMAND ${CMAKE_COMMAND} --build . --target install
+    WORKING_DIRECTORY ${CMAKE_BINARY_DIR}/Guest_32
+    )"
+    DEPENDS guest-libs-32
+  )
+
+  add_custom_target(uninstall_guest-libs
+    COMMAND ${CMAKE_COMMAND} "--build" "." "--target" "uninstall"
+    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
+  )
+
+  add_dependencies(uninstall uninstall_guest-libs)
+  add_dependencies(uninstall uninstall_guest-libs-32)
 endif()

 set(FEX_VERSION_MAJOR "0")
@@ -593,15 +540,9 @@ endif()

 # Package creation
 set (CPACK_GENERATOR "DEB")
-if (ENABLE_STATIC_PIE)
-  set (CPACK_PACKAGE_NAME fex-emu-static)
-  set (CPACK_DEBIAN_PACKAGE_CONFLICTS "fex-emu")
-else()
-  set (CPACK_PACKAGE_NAME fex-emu)
-  set (CPACK_DEBIAN_PACKAGE_CONFLICTS "fex-emu-static")
-endif()
+set (CPACK_PACKAGE_NAME fex-emu)
 set (CPACK_PACKAGE_FILE_NAME "${CPACK_PACKAGE_NAME}-${GIT_DESCRIBE_STRING}_${CMAKE_SYSTEM_PROCESSOR}")
-set (CPACK_PACKAGE_CONTACT "FEX-Emu Maintainers <team@fex-emu.org>")
+set (CPACK_PACKAGE_CONTACT "FEX-Emu Maintainers <team@fex-emu.com>")
 set (CPACK_PACKAGE_VERSION_MAJOR "${FEX_VERSION_MAJOR}")
 set (CPACK_PACKAGE_VERSION_MINOR "${FEX_VERSION_MINOR}")
 set (CPACK_PACKAGE_VERSION_PATCH "${FEX_VERSION_PATCH}")
@@ -55,7 +55,7 @@ further defined and clarified by project maintainers.
 ## Enforcement

 Instances of abusive, harassing, or otherwise unacceptable behavior may be
-reported by contacting the project team at team@fex-emu.org. All
+reported by contacting the project team at team@fex-emu.com. All
 complaints will be reviewed and investigated and will result in a response that
 is deemed necessary and appropriate to the circumstances. The project team is
 obligated to maintain confidentiality with regard to the reporter of an incident.
@@ -11,15 +11,15 @@ endforeach()
 # First generate then install it
 foreach(GEN_CONFIG_SRC ${GEN_CONFIG_SOURCES})
  # Get the filename only component
-  get_filename_component(CONFIG_NAME ${GEN_CONFIG_SRC} NAME_WE)
+  get_filename_component(CONFIG_NAME ${GEN_CONFIG_SRC} NAME_WLE)

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

  # Then install the configured json
  install(
-    FILES ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME}.json
+    FILES ${CMAKE_BINARY_DIR}/Data/AppConfig/${CONFIG_NAME}
    DESTINATION ${DATA_DIRECTORY}/AppConfig/)
 endforeach()
@@ -0,0 +1,5 @@
+{
+  "Config": {
+    "x86dec_SynchronizeRIPOnAllBlocks": "1"
+  }
+}
@@ -0,0 +1,5 @@
+{
+  "Config": {
+    "x86dec_SynchronizeRIPOnAllBlocks": "1"
+  }
+}
@@ -0,0 +1,5 @@
+{
+  "Config": {
+    "x86dec_SynchronizeRIPOnAllBlocks": "1"
+  }
+}
@@ -0,0 +1,5 @@
+{
+  "Config": {
+    "x86dec_SynchronizeRIPOnAllBlocks": "1"
+  }
+}
@@ -165,6 +165,14 @@
        "@PREFIX_LIB@/x86_64-linux-gnu/libXfixes.so.3.1.0"
      ]
    },
+    "OpenCL": {
+      "Library" : "libOpenCL-guest.so",
+      "Overlay": [
+        "@PREFIX_LIB@/x86_64-linux-gnu/libOpenCL.so",
+        "@PREFIX_LIB@/x86_64-linux-gnu/libOpenCL.so.1",
+        "@PREFIX_LIB@/x86_64-linux-gnu/libOpenCL.so.1.0.0"
+      ]
+    },
    "":{}
  }
 }
@@ -9,12 +9,19 @@ if (CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64")
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcx16")
 endif()

-if (CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64")
+if (CMAKE_SYSTEM_PROCESSOR MATCHES "^aarch64|^arm64|^armv8\.*")
  set(_M_ARM_64 1)
 endif()

-set(ENABLE_JIT_X86_64 ${_M_X86_64} CACHE BOOL "Enable the x86_64 JIT")
-set(ENABLE_JIT_ARM64 ${_M_ARM_64} CACHE BOOL "Enable the ARM64 JIT")
+if (ENABLE_VIXL_SIMULATOR)
+  # If the vixl simulator is enabled then we are using the ARM64 JIT
+  option(ENABLE_JIT_X86_64 "Enable the x86_64 JIT" FALSE)
+  option(ENABLE_JIT_ARM64 "Enable the ARM64 JIT" TRUE)
+else()
+  option(ENABLE_JIT_X86_64 "Enable the x86_64 JIT" ${_M_X86_64})
+  option(ENABLE_JIT_ARM64 "Enable the ARM64 JIT" ${_M_ARM_64})
+endif()
+
 option(ENABLE_CLANG_FORMAT "Run clang format over the source" FALSE)

 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
@@ -27,7 +34,6 @@ set(CMAKE_INCLUDE_CURRENT_DIR ON)
 include(CheckCXXCompilerFlag)
 include(CheckIncludeFileCXX)

-
 if (EXISTS ${CMAKE_CURRENT_DIR}/External/vixl/)
    # Useful to have for freestanding libFEXCore
    add_subdirectory(External/vixl/)
@@ -46,14 +52,14 @@ if (OVERRIDE_VERSION STREQUAL "detect")

  if (GIT_FOUND)
    execute_process(
-      COMMAND ${GIT_EXECUTABLE} rev-parse --short HEAD
+      COMMAND ${GIT_EXECUTABLE} rev-parse --short=7 HEAD
      WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
      OUTPUT_VARIABLE GIT_SHORT_HASH
      ERROR_QUIET
      OUTPUT_STRIP_TRAILING_WHITESPACE
    )
    execute_process(
-      COMMAND ${GIT_EXECUTABLE} describe
+      COMMAND ${GIT_EXECUTABLE} describe --abbrev=7
      WORKING_DIRECTORY "${CMAKE_SOURCE_DIR}"
      OUTPUT_VARIABLE GIT_DESCRIBE_STRING
      ERROR_QUIET
@@ -321,8 +321,6 @@ def print_ir_structs(defines):


        if op.SSAArgNum > 0:
-            # Add helpers for accessing SSA arguments, given how frequently they're accessed
-
            output_file.write("\t// Get index of argument by name\n")
            SSAArg = 0
            for arg in op.Arguments:
@@ -330,14 +328,6 @@ def print_ir_structs(defines):
                    output_file.write("\tstatic constexpr size_t {}_Index = {};\n".format(arg.Name, SSAArg))
                    SSAArg = SSAArg + 1

-            output_file.write("\n")
-            output_file.write("\t[[nodiscard]] OrderedNodeWrapper& Args(size_t Index) {\n")
-            output_file.write("\t\treturn Header.Args[Index];\n")
-            output_file.write("\t}\n")
-            output_file.write("\t[[nodiscard]] const OrderedNodeWrapper& Args(size_t Index) const {\n")
-            output_file.write("\t\treturn Header.Args[Index];\n")
-            output_file.write("\t}\n")
-

        output_file.write("};\n")

@@ -135,7 +135,6 @@ set (SRCS
  Interface/IR/Passes/PhiValidation.cpp
  Interface/IR/Passes/RedundantFlagCalculationElimination.cpp
  Interface/IR/Passes/DeadStoreElimination.cpp
-  Interface/IR/Passes/StaticRegisterAllocationPass.cpp
  Interface/IR/Passes/RegisterAllocationPass.cpp
  Interface/IR/Passes/SyscallOptimization.cpp
  Utils/Allocator.cpp
@@ -143,6 +142,7 @@ set (SRCS
  Utils/NetStream.cpp
  Utils/Telemetry.cpp
  Utils/Threads.cpp
+  Utils/Profiler.cpp
  )

 if (ENABLE_INTERPRETER)
@@ -177,6 +177,11 @@ if (_M_ARM_64)
  list(APPEND DEFINES -D_M_ARM_64=1)
 endif()

+if (ENABLE_VIXL_SIMULATOR)
+  # We can run the simulator on both x86-64 or AArch64 hosts
+  list(APPEND DEFINES -DVIXL_SIMULATOR=1 -DVIXL_INCLUDE_SIMULATOR_AARCH64=1)
+endif()
+
 if (ENABLE_JIT_X86_64)
  list(APPEND SRCS
    Interface/Core/JIT/x86_64/JIT.cpp
@@ -213,7 +218,7 @@ if (ENABLE_JIT_ARM64)
    )
 endif()

-set (LIBS vixl dl xxhash tiny-json)
+set (LIBS fmt::fmt vixl dl xxhash tiny-json FEXHeaderUtils)
 if (ENABLE_JEMALLOC)
  list (APPEND LIBS FEX_jemalloc)
 endif()
@@ -359,14 +364,14 @@ endfunction()

 # Build FEXCore_Config static library
 add_library(FEXCore_Base STATIC ${FEXCORE_BASE_SRCS})
-target_link_libraries(FEXCore_Base fmt::fmt tiny-json)
+target_link_libraries(FEXCore_Base ${LIBS})
 AddDefaultOptionsToTarget(FEXCore_Base)

 function(AddObject Name Type)
  add_library(${Name} ${Type} ${SRCS})
  add_dependencies(${Name} IR_INC)

-  target_link_libraries(${Name} FEXCore_Base ${LIBS})
+  target_link_libraries(${Name} FEXCore_Base)
  AddDefaultOptionsToTarget(${Name})

  set_target_properties(${Name} PROPERTIES OUTPUT_NAME FEXCore)
@@ -374,7 +379,7 @@ endfunction()

 function(AddLibrary Name Type)
  add_library(${Name} ${Type} $<TARGET_OBJECTS:${PROJECT_NAME}_object>)
-  target_link_libraries(${Name} FEXCore_Base ${LIBS})
+  target_link_libraries(${Name} FEXCore_Base)
  set_target_properties(${Name} PROPERTIES OUTPUT_NAME FEXCore)

  AddDefaultOptionsToTarget(${Name})
@@ -20,12 +20,12 @@ struct BitSet final {
  ElementType *Memory;
  void Allocate(size_t Elements) {
    size_t AllocateSize = AlignUp(Elements, MinimumSizeBits) / MinimumSize;
-    LOGMAN_THROW_A_FMT((AllocateSize * MinimumSize) >= Elements, "Fail");
+    LOGMAN_THROW_AA_FMT((AllocateSize * MinimumSize) >= Elements, "Fail");
    Memory = static_cast<ElementType*>(FEXCore::Allocator::malloc(AllocateSize));
  }
  void Realloc(size_t Elements) {
    size_t AllocateSize = AlignUp(Elements, MinimumSizeBits) / MinimumSize;
-    LOGMAN_THROW_A_FMT((AllocateSize * MinimumSize) >= Elements, "Fail");
+    LOGMAN_THROW_AA_FMT((AllocateSize * MinimumSize) >= Elements, "Fail");
    Memory = static_cast<ElementType*>(FEXCore::Allocator::realloc(Memory, AllocateSize));
  }
  void Free() {
@@ -64,7 +64,7 @@ struct BitSetView final {
  ElementType *Memory;

  void GetView(BitSet<T> &Set, uint64_t ElementOffset) {
-    LOGMAN_THROW_A_FMT((ElementOffset % MinimumSize) == 0,
+    LOGMAN_THROW_AA_FMT((ElementOffset % MinimumSize) == 0,
                       "Bitset view offset needs to be aligned to size of backing element");
    Memory = &Set.Memory[ElementOffset / MinimumSizeBits];
  }
@@ -7,6 +7,11 @@

 namespace FEXCore {
  JITSymbols::JITSymbols() : fp{nullptr, std::fclose} {
+  }
+
+  JITSymbols::~JITSymbols() = default;
+
+  void JITSymbols::InitFile() {
    const auto PerfMap = fmt::format("/tmp/perf-{}.map", getpid());

    fp.reset(fopen(PerfMap.c_str(), "wb"));
@@ -16,8 +21,6 @@ namespace FEXCore {
    }
  }

-  JITSymbols::~JITSymbols() = default;
-
  void JITSymbols::Register(const void *HostAddr, uint64_t GuestAddr, uint32_t CodeSize) {
    if (!fp) return;

@@ -11,6 +11,7 @@ public:
  JITSymbols();
  ~JITSymbols();

+  void InitFile();
  void Register(const void *HostAddr, uint64_t GuestAddr, uint32_t CodeSize);
  void Register(const void *HostAddr, uint32_t CodeSize, std::string_view Name);
  void Register(const void *HostAddr, uint32_t CodeSize, std::string_view Name, uintptr_t Offset);
@@ -82,7 +82,7 @@ namespace JSON {
    json_t const* ConfigList = json_getProperty(json, "Config");

    if (!ConfigList) {
-      LogMan::Msg::EFmt("Couldn't get config list");
+      // This is a non-error if the configuration file exists but no Config section
      return;
    }

@@ -154,15 +154,20 @@ namespace JSON {
    return ConfigDir;
  }

-  std::string GetConfigFileLocation() {
+  std::string GetConfigFileLocation(bool Global) {
    std::string ConfigFile{};
-    const char *AppConfig = getenv("FEX_APP_CONFIG");
-    if (AppConfig) {
-      // App config environment variable overwrites only the config file
-      ConfigFile = AppConfig;
+    if (Global) {
+      ConfigFile = GetConfigDirectory(true) + "Config.json";
    }
    else {
-      ConfigFile = GetConfigDirectory(false) + "Config.json";
+      const char *AppConfig = getenv("FEX_APP_CONFIG");
+      if (AppConfig) {
+        // App config environment variable overwrites only the config file
+        ConfigFile = AppConfig;
+      }
+      else {
+        ConfigFile = GetConfigDirectory(false) + "Config.json";
+      }
    }
    return ConfigFile;
  }
@@ -206,9 +211,12 @@ namespace JSON {
  static std::map<FEXCore::Config::LayerType, std::unique_ptr<FEXCore::Config::Layer>> ConfigLayers;
  static FEXCore::Config::Layer *Meta{};

-  constexpr std::array<FEXCore::Config::LayerType, 6> LoadOrder = {
+  constexpr std::array<FEXCore::Config::LayerType, 9> LoadOrder = {
+    FEXCore::Config::LayerType::LAYER_GLOBAL_MAIN,
    FEXCore::Config::LayerType::LAYER_MAIN,
+    FEXCore::Config::LayerType::LAYER_GLOBAL_STEAM_APP,
    FEXCore::Config::LayerType::LAYER_GLOBAL_APP,
+    FEXCore::Config::LayerType::LAYER_LOCAL_STEAM_APP,
    FEXCore::Config::LayerType::LAYER_LOCAL_APP,
    FEXCore::Config::LayerType::LAYER_ARGUMENTS,
    FEXCore::Config::LayerType::LAYER_ENVIRONMENT,
@@ -613,7 +621,7 @@ namespace JSON {
  // Application loaders
  class MainLoader final : public FEXCore::Config::OptionMapper {
  public:
-    explicit MainLoader();
+    explicit MainLoader(FEXCore::Config::LayerType Type);
    explicit MainLoader(std::string ConfigFile);
    void Load() override;

@@ -623,7 +631,7 @@ namespace JSON {

  class AppLoader final : public FEXCore::Config::OptionMapper {
  public:
-    explicit AppLoader(const std::string& Filename, bool Global);
+    explicit AppLoader(const std::string& Filename, FEXCore::Config::LayerType Type);
    void Load();

  private:
@@ -659,9 +667,9 @@ namespace JSON {
    }
  }

-  MainLoader::MainLoader()
-    : FEXCore::Config::OptionMapper(FEXCore::Config::LayerType::LAYER_MAIN)
-    , Config{FEXCore::Config::GetConfigFileLocation()} {
+  MainLoader::MainLoader(FEXCore::Config::LayerType Type)
+    : FEXCore::Config::OptionMapper(Type)
+    , Config{FEXCore::Config::GetConfigFileLocation(Type == FEXCore::Config::LayerType::LAYER_GLOBAL_MAIN)} {
  }

  MainLoader::MainLoader(std::string ConfigFile)
@@ -675,8 +683,10 @@ namespace JSON {
    });
  }

-  AppLoader::AppLoader(const std::string& Filename, bool Global)
-    : FEXCore::Config::OptionMapper(Global ? FEXCore::Config::LayerType::LAYER_GLOBAL_APP : FEXCore::Config::LayerType::LAYER_LOCAL_APP) {
+  AppLoader::AppLoader(const std::string& Filename, FEXCore::Config::LayerType Type)
+    : FEXCore::Config::OptionMapper(Type) {
+    const bool Global = Type == FEXCore::Config::LayerType::LAYER_GLOBAL_STEAM_APP ||
+                        Type == FEXCore::Config::LayerType::LAYER_LOCAL_STEAM_APP;
    Config = FEXCore::Config::GetApplicationConfig(Filename, Global);

    // Immediately load so we can reload the meta layer
@@ -735,17 +745,21 @@ namespace JSON {
    }
  }

+  std::unique_ptr<FEXCore::Config::Layer> CreateGlobalMainLayer() {
+    return std::make_unique<FEXCore::Config::MainLoader>(FEXCore::Config::LayerType::LAYER_GLOBAL_MAIN);
+  }
+
  std::unique_ptr<FEXCore::Config::Layer> CreateMainLayer(std::string const *File) {
    if (File) {
      return std::make_unique<FEXCore::Config::MainLoader>(*File);
    }
    else {
-      return std::make_unique<FEXCore::Config::MainLoader>();
+      return std::make_unique<FEXCore::Config::MainLoader>(FEXCore::Config::LayerType::LAYER_MAIN);
    }
  }

-  std::unique_ptr<FEXCore::Config::Layer> CreateAppLayer(const std::string& Filename, bool Global) {
-    return std::make_unique<FEXCore::Config::AppLoader>(Filename, Global);
+  std::unique_ptr<FEXCore::Config::Layer> CreateAppLayer(const std::string& Filename, FEXCore::Config::LayerType Type) {
+    return std::make_unique<FEXCore::Config::AppLoader>(Filename, Type);
  }

  std::unique_ptr<FEXCore::Config::Layer> CreateEnvironmentLayer(char *const _envp[]) {
@@ -16,10 +16,11 @@
      },
      "Multiblock": {
        "Type": "bool",
-        "Default": "true",
+        "Default": "false",
        "ShortArg": "m",
        "Desc": [
-          "Controls multiblock code compilation"
+          "Controls multiblock code compilation",
+          "Can cause long JIT compilation times and stutter"
        ]
      },
      "MaxInst": {
@@ -49,6 +50,13 @@
          "Cache JIT object code to drive.",
          "Allows JIT code to be shared between applications"
        ]
+      },
+      "EnableAVX": {
+        "Type": "bool",
+        "Default": "true",
+        "Desc": [
+          "Determines whether or not we use the expanded register file for AVX or not"
+        ]
      }
    },
    "Emulation": {
@@ -83,6 +91,13 @@
          "Folder to find the guest-side thunking libraries."
        ]
      },
+      "ThunkGuestLibs32": {
+        "Type": "str",
+        "Default": "@CMAKE_INSTALL_PREFIX@/share/fex-emu/GuestThunks_32/",
+        "Desc": [
+          "Folder to find the 32-bit guest-side thunking libraries."
+        ]
+      },
      "ThunkConfig": {
        "Type": "str",
        "Default": "",
@@ -309,6 +324,16 @@
          "Forces a process to stall out on initialization",
          "Useful for a process that keeps restarting and doesn't work"
        ]
+      },
+      "x86dec_SynchronizeRIPOnAllBlocks": {
+        "Type": "bool",
+        "Default": "false",
+        "Desc": [
+          "An application that uses try-catch or longjump extensively needs the ability to do context aware state flushing",
+          "In the case of FEX's block-linking, it won't always ensure that RIP is synchronized.",
+          "If an exception occurs and RIP isn't synchronized, then FEX's exception stack restore may not long jump as expected",
+          "Can be useful for Wine applications that rely on stack unwinding"
+        ]
      }
    },
    "Misc": {
@@ -334,6 +359,13 @@
        "Desc": [
          "Loads an AOT IR cache for the loaded executable."
        ]
+      },
+      "ServerSocketPath": {
+        "Type": "str",
+        "Default": "",
+        "Desc": [
+          "Override for a FEXServer socket path. Only useful for chroots."
+        ]
      }
    }
  },
@@ -351,6 +383,17 @@
        "Type": "str",
        "Default": ""
      },
+      "APP_CONFIG_NAME": {
+        "Type": "str",
+        "Default": "",
+        "Desc": [
+          "This is the application config name that has been loaded.",
+          "This differs from APP_FILENAME in two ways",
+          "Where APP_FILENAME always points to the executable path that FEX-Emu is executing.",
+          "This matches what is used to load the AppLayer configuration name.",
+          "When running through a compatibility layer like wine, this will only be the exe name, instead of wine full path."
+        ]
+      },
      "IS64BIT_MODE": {
        "Type": "bool",
        "Default": "false"
@@ -110,6 +110,10 @@ namespace FEXCore::Context {
  void RegisterExternalSyscallVisitor(FEXCore::Context::Context *CTX, [[maybe_unused]] uint64_t Syscall, [[maybe_unused]] FEXCore::HLE::SyscallVisitor *Visitor) {
  }

+  HostFeatures GetHostFeatures(const FEXCore::Context::Context *CTX) {
+    return CTX->HostFeatures;
+  }
+
  void HandleCallback(FEXCore::Context::Context *CTX, FEXCore::Core::InternalThreadState *Thread, uint64_t RIP) {
    CTX->HandleCallback(Thread, RIP);
  }
@@ -198,6 +202,10 @@ namespace FEXCore::Context {
    return CTX->AddCustomIREntrypoint(Entrypoint, Handler, Creator, Data);
  }

+  void AppendThunkDefinitions(FEXCore::Context::Context *CTX, std::vector<FEXCore::IR::ThunkDefinition> const& Definitions) {
+    CTX->AppendThunkDefinitions(Definitions);
+  }
+
 namespace Debug {
  void CompileRIP(FEXCore::Context::Context *CTX, uint64_t RIP) {
    CTX->CompileRIP(CTX->ParentThread, RIP);
@@ -1,8 +1,8 @@
 #pragma once

 #include "Common/JitSymbols.h"
+#include "FEXHeaderUtils/ScopedSignalMask.h"
 #include "Interface/Core/CPUID.h"
-#include "Interface/Core/HostFeatures.h"
 #include "Interface/Core/X86HelperGen.h"
 #include "Interface/Core/ObjectCache/ObjectCacheService.h"
 #include "Interface/Core/Dispatcher/Dispatcher.h"
@@ -10,10 +10,12 @@
 #include <FEXCore/Config/Config.h>
 #include <FEXCore/Core/Context.h>
 #include <FEXCore/Core/CoreState.h>
+#include <FEXCore/Core/HostFeatures.h>
 #include <FEXCore/Core/SignalDelegator.h>
 #include <FEXCore/Debug/InternalThreadState.h>
 #include <FEXCore/Utils/CompilerDefs.h>
 #include <FEXCore/Utils/Event.h>
+#include <FEXHeaderUtils/Syscalls.h>
 #include <stdint.h>


@@ -113,6 +115,8 @@ namespace FEXCore::Context {
      FEX_CONFIG_OPT(ParanoidTSO, PARANOIDTSO);
      FEX_CONFIG_OPT(CacheObjectCodeCompilation, CACHEOBJECTCODECOMPILATION);
      FEX_CONFIG_OPT(x87ReducedPrecision, X87REDUCEDPRECISION);
+      FEX_CONFIG_OPT(x86dec_SynchronizeRIPOnAllBlocks, X86DEC_SYNCHRONIZERIPONALLBLOCKS);
+      FEX_CONFIG_OPT(EnableAVX, ENABLEAVX);
    } Config;

    FEXCore::HostFeatures HostFeatures;
@@ -121,6 +125,7 @@ namespace FEXCore::Context {
    FEXCore::Core::InternalThreadState* ParentThread;
    std::vector<FEXCore::Core::InternalThreadState*> Threads;
    std::atomic_bool CoreShuttingDown{false};
+    bool NeedToCheckXID{true};

    std::mutex IdleWaitMutex;
    std::condition_variable IdleWaitCV;
@@ -129,8 +134,8 @@ namespace FEXCore::Context {
    Event PauseWait;
    bool Running{};

-    std::shared_mutex CodeInvalidationMutex; 
-    
+    std::shared_mutex CodeInvalidationMutex;
+
    FEXCore::CPUIDEmu CPUID;
    FEXCore::HLE::SyscallHandler *SyscallHandler{};
    FEXCore::HLE::SourcecodeResolver *SourcecodeResolver{};
@@ -170,13 +175,26 @@ namespace FEXCore::Context {
    void RegisterHostSignalHandler(int Signal, HostSignalDelegatorFunction Func, bool Required);
    void RegisterFrontendHostSignalHandler(int Signal, HostSignalDelegatorFunction Func, bool Required);

-    // Must be called from owning thread
-    static void RemoveThreadCodeEntry(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP);
+    static void ThreadRemoveCodeEntry(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP);
+    static void ThreadAddBlockLink(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestDestination, uintptr_t HostLink, const std::function<void()> &delinker);

-    // Wrapper which takes CpuStateFrame instead of InternalThreadState
+    template<auto Fn>
+    static uint64_t ThreadExitFunctionLink(FEXCore::Core::CpuStateFrame *Frame, uint64_t *record) {
+      FHU::ScopedSignalMaskWithSharedLock lk(Frame->Thread->CTX->CodeInvalidationMutex);
+
+      return Fn(Frame, record);
+    }
+
+    // Wrapper which takes CpuStateFrame instead of InternalThreadState and unique_locks CodeInvalidationMutex
    // Must be called from owning thread
-    static void RemoveThreadCodeEntryFromJit(FEXCore::Core::CpuStateFrame *Frame, uint64_t GuestRIP) {
-      RemoveThreadCodeEntry(Frame->Thread, GuestRIP);
+    static void ThreadRemoveCodeEntryFromJit(FEXCore::Core::CpuStateFrame *Frame, uint64_t GuestRIP) {
+      auto Thread = Frame->Thread;
+      
+      LogMan::Throw::AFmt(Thread->ThreadManager.GetTID() == FHU::Syscalls::gettid(), "Must be called from owning thread {}, not {}", Thread->ThreadManager.GetTID(), FHU::Syscalls::gettid());
+
+      FHU::ScopedSignalMaskWithUniqueLock lk(Thread->CTX->CodeInvalidationMutex);
+
+      ThreadRemoveCodeEntry(Thread, GuestRIP);
    }

    // returns false if a handler was already registered
@@ -305,6 +323,8 @@ namespace FEXCore::Context {
      IRCaptureCache.SetAOTIRRenamer(CacheRenamer);
    }

+    void AppendThunkDefinitions(std::vector<FEXCore::IR::ThunkDefinition> const& Definitions);
+
    FEXCore::Utils::PooledAllocatorMMap OpDispatcherAllocator;
    FEXCore::Utils::PooledAllocatorMMap FrontendAllocator;

@@ -350,7 +370,7 @@ namespace FEXCore::Context {
    bool StartPaused = false;
    bool IsMemoryShared = false;
    FEX_CONFIG_OPT(AppFilename, APP_FILENAME);
-    
+
    std::shared_mutex CustomIRMutex;
    std::unordered_map<uint64_t, std::tuple<std::function<void(uintptr_t Entrypoint, FEXCore::IR::IREmitter *)>, void *, void *>> CustomIRHandlers;
    FEXCore::CPU::CPUBackendFeatures BackendFeatures;
@@ -3,6 +3,7 @@

 #include <aarch64/cpu-aarch64.h>

+#include <FEXCore/Utils/EnumUtils.h>
 #include <FEXCore/Utils/LogManager.h>
 #include <FEXCore/Utils/Telemetry.h>

@@ -1986,7 +1987,8 @@ uint64_t HandleAtomicLoadstoreExclusive(void *_ucontext, void *_info) {
        DesiredFunction = NEGDesired;
        break;
      default:
-        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}", AtomicOp);
+        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}",
+                          ToUnderlying(AtomicOp));
        return false;
    }

@@ -2039,7 +2041,8 @@ uint64_t HandleAtomicLoadstoreExclusive(void *_ucontext, void *_info) {
        DesiredFunction = NEGDesired;
        break;
      default:
-        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}", AtomicOp);
+        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}",
+                          ToUnderlying(AtomicOp));
        return false;
    }

@@ -2092,7 +2095,8 @@ uint64_t HandleAtomicLoadstoreExclusive(void *_ucontext, void *_info) {
        DesiredFunction = NEGDesired;
        break;
      default:
-        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}", AtomicOp);
+        LogMan::Msg::EFmt("Unhandled JIT SIGBUS Atomic mem op 0x{:02x}",
+                          ToUnderlying(AtomicOp));
        return false;
    }

@@ -17,23 +17,35 @@
 #include <utility>

 namespace FEXCore::CPU {
-#define STATE x28

 // We want vixl to not allocate a default buffer. Jit and dispatcher will manually create one.
 Arm64Emitter::Arm64Emitter(FEXCore::Context::Context *ctx, size_t size)
-  : vixl::aarch64::Assembler(size, vixl::aarch64::PositionDependentCode)
+  : vixl::aarch64::Assembler(size ? (byte*)FEXCore::Allocator::mmap(nullptr, size, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0) : reinterpret_cast<byte*>(~0ULL),
+                             size,
+                             vixl::aarch64::PositionDependentCode)
  , EmitterCTX {ctx} {
  CPU.SetUp();

+#ifdef VIXL_SIMULATOR
+  auto Features = vixl::CPUFeatures::All();
+#else
  auto Features = vixl::CPUFeatures::InferFromOS();
  if (ctx->HostFeatures.SupportsAtomics) {
    // Hypervisor can hide this on the c630?
    Features.Combine(vixl::CPUFeatures::Feature::kLORegions);
  }
+#endif

  SetCPUFeatures(Features);
 }

+Arm64Emitter::~Arm64Emitter() {
+  auto CodeBuffer = GetBuffer();
+  if (CodeBuffer->GetCapacity()) {
+    FEXCore::Allocator::munmap(CodeBuffer->GetStartAddress<void*>(), CodeBuffer->GetCapacity());
+  }
+}
+
 void Arm64Emitter::LoadConstant(vixl::aarch64::Register Reg, uint64_t Constant, bool NOPPad) {
  bool Is64Bit = Reg.IsX();
  int Segments = Is64Bit ? 4 : 2;
@@ -209,19 +221,30 @@ void Arm64Emitter::SpillStaticRegs(bool FPRs, uint32_t GPRSpillMask, uint32_t FP
    }

    if (FPRs) {
-      for (size_t i = 0; i < SRAFPR.size(); i+=2) {
-        auto Reg1 = SRAFPR[i];
-        auto Reg2 = SRAFPR[i+1];
+      if (EmitterCTX->HostFeatures.SupportsAVX) {
+        for (size_t i = 0; i < SRAFPR.size(); i++) {
+          const auto Reg = SRAFPR[i];

-        if (((1U << Reg1.GetCode()) & FPRSpillMask) &&
-            ((1U << Reg2.GetCode()) & FPRSpillMask)) {
-          stp(Reg1.Q(), Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i][0])));
+          if (((1U << Reg.GetCode()) & FPRSpillMask) != 0) {
+            mov(TMP4, offsetof(Core::CpuStateFrame, State.xmm.avx.data[i][0]));
+            st1b(Reg.Z().VnB(), PRED_TMP_32B, SVEMemOperand(STATE, TMP4));
+          }
        }
-        else if (((1U << Reg1.GetCode()) & FPRSpillMask)) {
-          str(Reg1.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i][0])));
-        }
-        else if (((1U << Reg2.GetCode()) & FPRSpillMask)) {
-          str(Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i+1][0])));
+      } else {
+        for (size_t i = 0; i < SRAFPR.size(); i += 2) {
+          const auto Reg1 = SRAFPR[i];
+          const auto Reg2 = SRAFPR[i + 1];
+
+          if (((1U << Reg1.GetCode()) & FPRSpillMask) &&
+              ((1U << Reg2.GetCode()) & FPRSpillMask)) {
+            stp(Reg1.Q(), Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i][0])));
+          }
+          else if (((1U << Reg1.GetCode()) & FPRSpillMask)) {
+            str(Reg1.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i][0])));
+          }
+          else if (((1U << Reg2.GetCode()) & FPRSpillMask)) {
+            str(Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i+1][0])));
+          }
        }
      }
    }
@@ -246,19 +269,37 @@ void Arm64Emitter::FillStaticRegs(bool FPRs, uint32_t GPRFillMask, uint32_t FPRF
    }

    if (FPRs) {
-      for (size_t i = 0; i < SRAFPR.size(); i+=2) {
-        auto Reg1 = SRAFPR[i];
-        auto Reg2 = SRAFPR[i+1];
+      if (EmitterCTX->HostFeatures.SupportsAVX) {
+        // Set up predicate registers.
+        // We don't bother spilling these in SpillStaticRegs,
+        // since all that matters is we restore them on a fill.
+        // It's not a concern if they get trounced by something else.
+        ptrue(PRED_TMP_16B.VnB(), SVE_VL16);
+        ptrue(PRED_TMP_32B.VnB(), SVE_VL32);

-        if (((1U << Reg1.GetCode()) & FPRFillMask) &&
-            ((1U << Reg2.GetCode()) & FPRFillMask)) {
-          ldp(Reg1.Q(), Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i][0])));
+        for (size_t i = 0; i < SRAFPR.size(); i++) {
+          const auto Reg = SRAFPR[i];
+
+          if (((1U << Reg.GetCode()) & FPRFillMask) != 0) {
+            mov(TMP4, offsetof(Core::CpuStateFrame, State.xmm.avx.data[i][0]));
+            ld1b(Reg.Z().VnB(), PRED_TMP_32B.Zeroing(), SVEMemOperand(STATE, TMP4));
+          }
        }
-        else if (((1U << Reg1.GetCode()) & FPRFillMask)) {
-          ldr(Reg1.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i][0])));
-        }
-        else if (((1U << Reg2.GetCode()) & FPRFillMask)) {
-          ldr(Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm[i+1][0])));
+      } else {
+        for (size_t i = 0; i < SRAFPR.size(); i += 2) {
+          const auto Reg1 = SRAFPR[i];
+          const auto Reg2 = SRAFPR[i + 1];
+
+          if (((1U << Reg1.GetCode()) & FPRFillMask) &&
+              ((1U << Reg2.GetCode()) & FPRFillMask)) {
+            ldp(Reg1.Q(), Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i][0])));
+          }
+          else if (((1U << Reg1.GetCode()) & FPRFillMask)) {
+            ldr(Reg1.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i][0])));
+          }
+          else if (((1U << Reg2.GetCode()) & FPRFillMask)) {
+            ldr(Reg2.Q(), MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, State.xmm.sse.data[i+1][0])));
+          }
        }
      }
    }
@@ -266,20 +307,31 @@ void Arm64Emitter::FillStaticRegs(bool FPRs, uint32_t GPRFillMask, uint32_t FPRF
 }

 void Arm64Emitter::PushDynamicRegsAndLR() {
-  uint64_t SPOffset = AlignUp((RA64.size() + 1) * 8 + RAFPR.size() * 16, 16);
+  const auto CanUseSVE = EmitterCTX->HostFeatures.SupportsAVX;
+  const auto GPRSize = (RA64.size() + 1) * Core::CPUState::GPR_REG_SIZE;
+  const auto FPRRegSize = CanUseSVE ? Core::CPUState::XMM_AVX_REG_SIZE
+                                    : Core::CPUState::XMM_SSE_REG_SIZE;
+  const auto FPRSize = RAFPR.size() * FPRRegSize;
+  const uint64_t SPOffset = AlignUp(GPRSize + FPRSize, 16);

  sub(sp, sp, SPOffset);
  int i = 0;

-  for (auto RA : RAFPR)
-  {
-    str(RA.Q(), MemOperand(sp, i * 8));
-    i+=2;
+  if (CanUseSVE) {
+    for (const auto& RA : RAFPR) {
+      mov(TMP4, i * 8);
+      st1b(RA.Z().VnB(), PRED_TMP_32B, SVEMemOperand(sp, TMP4));
+      i += 4;
+    }
+  } else {
+    for (const auto& RA : RAFPR) {
+      str(RA.Q(), MemOperand(sp, i * 8));
+      i += 2;
+    }
  }

 #if 0 // All GPRs should be caller saved
-  for (auto RA : RA64)
-  {
+  for (const auto& RA : RA64) {
    str(RA, MemOperand(sp, i * 8));
    i++;
  }
@@ -289,18 +341,29 @@ void Arm64Emitter::PushDynamicRegsAndLR() {
 }

 void Arm64Emitter::PopDynamicRegsAndLR() {
-  uint64_t SPOffset = AlignUp((RA64.size() + 1) * 8 + RAFPR.size() * 16, 16);
+  const auto CanUseSVE = EmitterCTX->HostFeatures.SupportsAVX;
+  const auto GPRSize = (RA64.size() + 1) * Core::CPUState::GPR_REG_SIZE;
+  const auto FPRRegSize = CanUseSVE ? Core::CPUState::XMM_AVX_REG_SIZE
+                                    : Core::CPUState::XMM_SSE_REG_SIZE;
+  const auto FPRSize = RAFPR.size() * FPRRegSize;
+  const uint64_t SPOffset = AlignUp(GPRSize + FPRSize, 16);
  int i = 0;

-  for (auto RA : RAFPR)
-  {
-    ldr(RA.Q(), MemOperand(sp, i * 8));
-    i+=2;
+  if (CanUseSVE) {
+    for (const auto& RA : RAFPR) {
+      mov(TMP4, i * 8);
+      ld1b(RA.Z().VnB(), PRED_TMP_32B.Zeroing(), SVEMemOperand(sp, TMP4));
+      i += 4;
+    }
+  } else {
+    for (const auto& RA : RAFPR) {
+      ldr(RA.Q(), MemOperand(sp, i * 8));
+      i += 2;
+    }
  }

 #if 0 // All GPRs should be caller saved
-  for (auto RA : RA64)
-  {
+  for (const auto& RA : RA64) {
    ldr(RA, MemOperand(sp, i * 8));
    i++;
  }
@@ -8,6 +8,10 @@
 #include <aarch64/cpu-aarch64.h>
 #include <aarch64/operands-aarch64.h>
 #include <platform-vixl.h>
+#ifdef VIXL_SIMULATOR
+#include <aarch64/simulator-aarch64.h>
+#include <aarch64/simulator-constants-aarch64.h>
+#endif

 #include <FEXCore/Config/Config.h>

@@ -58,15 +62,41 @@ const std::array<aarch64::VRegister, 12> RAFPR = {
  v8,  v9,  v10, v11, v12, v13, v14, v15
 };

+// Contains the address to the currently available CPU state
+#define STATE x28
+
+// GPR temporaries. Only x3 can be used across spill boundaries
+// so if these ever need to change, be very careful about that.
+#define TMP1 x0
+#define TMP2 x1
+#define TMP3 x2
+#define TMP4 x3
+
+// Vector temporaries
+#define VTMP1 v1
+#define VTMP2 v2
+#define VTMP3 v3
+
+// Predicate register temporaries (used when AVX support is enabled)
+// PRED_TMP_16B indicates a predicate register that indicates the first 16 bytes set to 1.
+// PRED_TMP_32B indicates a predicate register that indicates the first 32 bytes set to 1.
+#define PRED_TMP_16B p6
+#define PRED_TMP_32B p7
+
 // This class contains common emitter utility functions that can
 // be used by both Arm64 JIT and ARM64 Dispatcher
 class Arm64Emitter : public vixl::aarch64::Assembler {
 protected:
  Arm64Emitter(FEXCore::Context::Context *ctx, size_t size);
+  ~Arm64Emitter();

  FEXCore::Context::Context *EmitterCTX;
  vixl::aarch64::CPU CPU;
  void LoadConstant(vixl::aarch64::Register Reg, uint64_t Constant, bool NOPPad = false);
+
+  // NOTE: These functions WILL clobber the register TMP4 if AVX support is enabled
+  //       and FPRs are being spilled or filled. If only GPRs are spilled/filled, then
+  //       TMP4 is left alone.
  void SpillStaticRegs(bool FPRs = true, uint32_t GPRSpillMask = ~0U, uint32_t FPRSpillMask = ~0U);
  void FillStaticRegs(bool FPRs = true, uint32_t GPRFillMask = ~0U, uint32_t FPRFillMask = ~0U);

@@ -83,6 +113,71 @@ protected:
  void PopCalleeSavedRegisters();

  void Align16B();
+#ifdef VIXL_SIMULATOR
+  // Generates a vixl simulator runtime call.
+  //
+  // This matches behaviour of vixl's macro assembler, but we need to reimplement it since we aren't using the macro assembler.
+  // This isn't too complex with how vixl emits this.
+  //
+  // Emit:
+  // 1) hlt(kRuntimeCallOpcode)
+  // 2) Simulator wrapper handler
+  // 3) Function to call
+  // 4) Style of the function call (Call versus tail-call)
+  template<typename R, typename... P>
+  void GenerateRuntimeCall(R (*Function)(P...)) {
+    uintptr_t SimulatorWrapperAddress = reinterpret_cast<uintptr_t>(
+      &(Simulator::RuntimeCallStructHelper<R, P...>::Wrapper));
+
+    uintptr_t FunctionAddress = reinterpret_cast<uintptr_t>(Function);
+
+    hlt(kRuntimeCallOpcode);
+
+    // Simulator wrapper address pointer.
+    dc(SimulatorWrapperAddress);
+
+    // Runtime function address to call
+    dc(FunctionAddress);
+
+    // Call type
+    dc32(kCallRuntime);
+  }
+
+  template<typename R, typename... P>
+  void GenerateIndirectRuntimeCall(vixl::aarch64::Register Reg) {
+    uintptr_t SimulatorWrapperAddress = reinterpret_cast<uintptr_t>(
+      &(Simulator::RuntimeCallStructHelper<R, P...>::Wrapper));
+
+    hlt(kIndirectRuntimeCallOpcode);
+
+    // Simulator wrapper address pointer.
+    dc(SimulatorWrapperAddress);
+
+    // Register that contains the function to call
+    dc(Reg.GetCode());
+
+    // Call type
+    dc32(kCallRuntime);
+  }
+
+  template<>
+  void GenerateIndirectRuntimeCall<float, __uint128_t>(vixl::aarch64::Register Reg) {
+    uintptr_t SimulatorWrapperAddress = reinterpret_cast<uintptr_t>(
+      &(Simulator::RuntimeCallStructHelper<float, __uint128_t>::Wrapper));
+
+    hlt(kIndirectRuntimeCallOpcode);
+
+    // Simulator wrapper address pointer.
+    dc(SimulatorWrapperAddress);
+
+    // Register that contains the function to call
+    dc(Reg.GetCode());
+
+    // Call type
+    dc32(kCallRuntime);
+  }
+
+#endif

  FEX_CONFIG_OPT(StaticRegisterAllocation, SRA);
 };
@@ -124,7 +124,7 @@ static inline void SetArmReg(void* ucontext, uint32_t id, uint64_t val) {
 static inline __uint128_t GetArmFPR(void* ucontext, uint32_t id) {
  auto MContext = GetMContext(ucontext);
  HostFPRState *HostState = reinterpret_cast<HostFPRState*>(&MContext->__reserved[0]);
-  LOGMAN_THROW_A_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);
+  LOGMAN_THROW_AA_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);

  return HostState->FPRs[id];
 }
@@ -143,7 +143,7 @@ static inline void BackupContext(void* ucontext, T *Backup) {

    // Host FPR state starts at _mcontext->reserved[0];
    HostFPRState *HostState = reinterpret_cast<HostFPRState*>(&_mcontext->__reserved[0]);
-    LOGMAN_THROW_A_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);
+    LOGMAN_THROW_AA_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);
    Backup->FPSR = HostState->FPSR;
    Backup->FPCR = HostState->FPCR;
    memcpy(&Backup->FPRs[0], &HostState->FPRs[0], 32 * sizeof(__uint128_t));
@@ -163,7 +163,7 @@ static inline void RestoreContext(void* ucontext, T *Backup) {
   auto _mcontext = GetMContext(ucontext);

    HostFPRState *HostState = reinterpret_cast<HostFPRState*>(&_mcontext->__reserved[0]);
-    LOGMAN_THROW_A_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);
+    LOGMAN_THROW_AA_FMT(HostState->Head.Magic == FPR_MAGIC, "Wrong FPR Magic: 0x{:08x}", HostState->Head.Magic);
    memcpy(&HostState->FPRs[0], &Backup->FPRs[0], 32 * sizeof(__uint128_t));
    HostState->FPCR = Backup->FPCR;
    HostState->FPSR = Backup->FPSR;
@@ -58,8 +58,8 @@ auto CPUBackend::AllocateNewCodeBuffer(size_t Size) -> CodeBuffer {
  Buffer.Size = Size;
  Buffer.Ptr = static_cast<uint8_t *>(
      FEXCore::Allocator::mmap(nullptr, Buffer.Size, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0));
-  LOGMAN_THROW_A_FMT(!!Buffer.Ptr, "Couldn't allocate code buffer");
-  
+  LOGMAN_THROW_AA_FMT(!!Buffer.Ptr, "Couldn't allocate code buffer");
+
  if (ThreadState->CTX->Config.GlobalJITNaming()) {
    ThreadState->CTX->Symbols.RegisterJITSpace(Buffer.Ptr, Buffer.Size);
  }
@@ -84,4 +84,4 @@ bool CPUBackend::IsAddressInCodeBuffer(uintptr_t Address) const {
 }

 }
-}
+}
@@ -8,11 +8,11 @@ $end_info$
 #include "Common/StringConv.h"
 #include "Interface/Context/Context.h"
 #include "Interface/Core/CPUID.h"
-#include "Interface/Core/HostFeatures.h"
 #include "Utils/FileLoading.h"

 #include <FEXCore/Config/Config.h>
 #include <FEXCore/Core/CPUID.h>
+#include <FEXCore/Core/HostFeatures.h>
 #include <FEXHeaderUtils/Syscalls.h>

 #include "git_version.h"
@@ -39,6 +39,7 @@ namespace ProductNames {
  static const char ARM_A78C[] = "Cortex-A78C";
  static const char ARM_A710[] = "Cortex-A710";
  static const char ARM_X1[] = "Cortex-X1";
+  static const char ARM_X1C[] = "Cortex-X1C";
  static const char ARM_X2[] = "Cortex-X2";
  static const char ARM_N1[] = "Neoverse N1";
  static const char ARM_N2[] = "Neoverse N2";
@@ -83,7 +84,10 @@ static uint32_t CalculateNumberOfCPUs() {
  return CPUs;
 }

+// TODO: Replace usages with CTX->HostFeatures.EnableAVX
+//       when AVX implementations are further along.
 constexpr uint32_t SUPPORTS_AVX = 0;
+
 // #define CPUID_AMD
 #ifdef CPUID_AMD
 constexpr uint32_t FAMILY_IDENTIFIER =
@@ -151,7 +155,7 @@ void CPUIDEmu::SetupHostHybridFlag() {
  // CPU priority order
  // This is mostly arbitrary but will sort by some sort of CPU priority by performance
  // Relative list so things they will commonly end up in big.little configurations sort of relate
-  static constexpr std::array<CPUMIDR, 35> CPUMIDRs = {{
+  static constexpr std::array<CPUMIDR, 36> CPUMIDRs = {{
    // Typically big CPU cores
    {0x61, 0x023, 1, ProductNames::ARM_Firestorm}, // Apple M1 Firestorm

@@ -161,6 +165,7 @@ void CPUIDEmu::SetupHostHybridFlag() {
    {0x41, 0xd49, 1, ProductNames::ARM_N2}, // N2
    {0x41, 0xd48, 1, ProductNames::ARM_X2}, // X2
    {0x41, 0xd47, 1, ProductNames::ARM_A710}, // A710
+    {0x41, 0xd4C, 1, ProductNames::ARM_X1C}, // X1C
    {0x41, 0xd44, 1, ProductNames::ARM_X1}, // X1
    {0x41, 0xd42, 1, ProductNames::ARM_A78AE}, // A78AE
    {0x41, 0xd41, 1, ProductNames::ARM_A78}, // A78
@@ -416,7 +421,7 @@ FEXCore::CPUID::FunctionResults CPUIDEmu::Function_01h(uint32_t Leaf) {

  Res.ecx =
    (1 <<  0) | // SSE3
-    (1 <<  1) | // PCLMULQDQ
+    (CTX->HostFeatures.SupportsPMULL_128Bit <<  1) | // PCLMULQDQ
    (1 <<  2) | // DS area supports 64bit layout
    (1 <<  3) | // MWait
    (0 <<  4) | // DS-CPL
@@ -44,6 +44,7 @@ $end_info$
 #include <FEXCore/Utils/Event.h>
 #include <FEXCore/Utils/LogManager.h>
 #include <FEXCore/Utils/Threads.h>
+#include <FEXCore/Utils/Profiler.h>
 #include <FEXHeaderUtils/Syscalls.h>
 #include <FEXHeaderUtils/TodoDefines.h>

@@ -154,6 +155,16 @@ namespace FEXCore::Context {
    if (Config.CacheObjectCodeCompilation() != FEXCore::Config::ConfigObjectCodeHandler::CONFIG_NONE) {
      CodeObjectCacheService = std::make_unique<FEXCore::CodeSerialize::CodeObjectSerializeService>(this);
    }
+    if (!Config.EnableAVX) {
+      HostFeatures.SupportsAVX = false;
+    }
+
+    if (Config.BlockJITNaming() ||
+        Config.GlobalJITNaming() ||
+        Config.LibraryJITNaming()) {
+      // Only initialize symbols file if enabled. Ensures we don't pollute /tmp with empty files.
+      Symbols.InitFile();
+    }
  }

  Context::~Context() {
@@ -184,9 +195,11 @@ namespace FEXCore::Context {
      greg = 0;
    }

-    for (auto& xmm : NewThreadState.xmm) {
+    for (auto& xmm : NewThreadState.xmm.avx.data) {
      xmm[0] = 0xDEADBEEFULL;
      xmm[1] = 0xBAD0DAD1ULL;
+      xmm[2] = 0xDEADCAFEULL;
+      xmm[3] = 0xBAD2CAD3ULL;
    }
    memset(NewThreadState.flags, 0, Core::CPUState::NUM_EFLAG_BITS);
    NewThreadState.flags[1] = 1;
@@ -209,7 +222,7 @@ namespace FEXCore::Context {
 #if (_M_X86_64 && JIT_X86_64)
      FEXCore::CPU::InitializeX86JITSignalHandlers(this);
      BackendFeatures = FEXCore::CPU::GetX86JITBackendFeatures();
-#elif (_M_ARM_64 && JIT_ARM64)
+#elif (_M_ARM_64 && JIT_ARM64) || defined(VIXL_SIMULATOR)
      FEXCore::CPU::InitializeArm64JITSignalHandlers(this);
      BackendFeatures = FEXCore::CPU::GetArm64JITBackendFeatures();
 #else
@@ -226,16 +239,16 @@ namespace FEXCore::Context {

    DispatcherConfig.StaticRegisterAllocation = Config.StaticRegisterAllocation && BackendFeatures.SupportsStaticRegisterAllocation;

-#if (_M_X86_64)
-    Dispatcher = FEXCore::CPU::Dispatcher::CreateX86(this, DispatcherConfig);
-#elif (_M_ARM_64)
+#if JIT_ARM64
    Dispatcher = FEXCore::CPU::Dispatcher::CreateArm64(this, DispatcherConfig);
+#elif JIT_X86_64
+    Dispatcher = FEXCore::CPU::Dispatcher::CreateX86(this, DispatcherConfig);
 #else
    ERROR_AND_DIE_FMT("FEXCore has been compiled with an unknown target");
 #endif

    // Initialize common signal handlers
-    
+
    auto PauseHandler = [](FEXCore::Core::InternalThreadState *Thread, int Signal, void *info, void *ucontext) -> bool {
      return Thread->CTX->Dispatcher->HandleSignalPause(Thread, Signal, info, ucontext);
    };
@@ -495,10 +508,21 @@ namespace FEXCore::Context {
    ExecutionThreadHandler *Arg = reinterpret_cast<ExecutionThreadHandler*>(FEXCore::Allocator::malloc(sizeof(ExecutionThreadHandler)));
    Arg->This = this;
    Arg->Thread = Thread;
+    Thread->StartPaused = NeedToCheckXID;
    Thread->ExecutionThread = FEXCore::Threads::Thread::Create(ThreadHandler, Arg);

    // Wait for the thread to have started
    Thread->ThreadWaiting.Wait();
+
+    if (NeedToCheckXID) {
+      // The first time an application creates a thread, GLIBC installs their SETXID signal handler.
+      // FEX needs to capture all signals and defer them to the guest.
+      // Once FEX creates its first guest thread, overwrite the GLIBC SETXID handler *again* to ensure
+      // FEX maintains control of the signal handler on this signal.
+      NeedToCheckXID = false;
+      SignalDelegation->CheckXIDHandler();
+      Thread->StartRunning.NotifyAll();
+    }
  }

  void Context::InitializeThreadTLSData(FEXCore::Core::InternalThreadState *Thread) {
@@ -546,11 +570,11 @@ namespace FEXCore::Context {
      break;
 #endif
    case FEXCore::Config::CONFIG_IRJIT:
-      Thread->PassManager->InsertRegisterAllocationPass(DoSRA);
+      Thread->PassManager->InsertRegisterAllocationPass(DoSRA, HostFeatures.SupportsAVX);

 #if (_M_X86_64 && JIT_X86_64)
      Thread->CPUBackend = FEXCore::CPU::CreateX86JITCore(this, Thread);
-#elif (_M_ARM_64 && JIT_ARM64)
+#elif (_M_ARM_64 && JIT_ARM64) || defined(VIXL_SIMULATOR)
      Thread->CPUBackend = FEXCore::CPU::CreateArm64JITCore(this, Thread);
 #else
      ERROR_AND_DIE_FMT("FEXCore has been compiled without a viable JIT core");
@@ -651,6 +675,8 @@ namespace FEXCore::Context {
  }

  void Context::ClearCodeCache(FEXCore::Core::InternalThreadState *Thread) {
+    FEXCORE_PROFILE_INSTANT("ClearCodeCache");
+
    {
      // Ensure the Code Object Serialization service has fully serialized this thread's data before clearing the cache
      // Use the thread's object cache ref counter for this
@@ -717,7 +743,9 @@ namespace FEXCore::Context {
    }
  }

-  Context::GenerateIRResult Context::GenerateIR(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP, bool ExtendedDebugInfo) {    
+  Context::GenerateIRResult Context::GenerateIR(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP, bool ExtendedDebugInfo) {
+    FEXCORE_PROFILE_SCOPED("GenerateIR");
+
    Thread->OpDispatcher->ReownOrClaimBuffer();
    Thread->OpDispatcher->ResetWorkingList();

@@ -726,7 +754,7 @@ namespace FEXCore::Context {


    std::shared_lock lk(CustomIRMutex);
-    
+
    auto Handler = CustomIRHandlers.find(GuestRIP);
    if (Handler != CustomIRHandlers.end()) {
      TotalInstructions = 1;
@@ -762,6 +790,13 @@ namespace FEXCore::Context {
        // Reset any block-specific state
        Thread->OpDispatcher->StartNewBlock();

+        if (Config.x86dec_SynchronizeRIPOnAllBlocks) {
+          // Ensure the RIP is synchronized to the context on block entry.
+          // In the case of block linking, the RIP may not have synchronized.
+          auto NewRIP = Thread->OpDispatcher->_EntrypointOffset(Block.Entry - GuestRIP, GPRSize);
+          Thread->OpDispatcher->_StoreContext(GPRSize, IR::GPRClass, NewRIP, offsetof(FEXCore::Core::CPUState, rip));
+        }
+
        uint64_t InstsInBlock = Block.NumInstructions;

        for (size_t i = 0; i < InstsInBlock; ++i) {
@@ -775,7 +810,7 @@ namespace FEXCore::Context {
          if (ExtendedDebugInfo) {
            Thread->OpDispatcher->_GuestOpcode(Block.Entry + BlockInstructionsLength - GuestRIP);
          }
-          
+
          if (Config.SMCChecks == FEXCore::Config::CONFIG_SMC_FULL) {
            auto ExistingCodePtr = reinterpret_cast<uint64_t*>(Block.Entry + BlockInstructionsLength);

@@ -788,7 +823,7 @@ namespace FEXCore::Context {
            Thread->OpDispatcher->SetTrueJumpTarget(InvalidateCodeCond, CodeWasChangedBlock);

            Thread->OpDispatcher->SetCurrentCodeBlock(CodeWasChangedBlock);
-            Thread->OpDispatcher->_RemoveThreadCodeEntry();
+            Thread->OpDispatcher->_ThreadRemoveCodeEntry();
            Thread->OpDispatcher->_ExitFunction(Thread->OpDispatcher->_EntrypointOffset(Block.Entry + BlockInstructionsLength - GuestRIP, GPRSize));

            auto NextOpBlock = Thread->OpDispatcher->CreateNewCodeBlockAfter(CurrentBlock);
@@ -842,7 +877,7 @@ namespace FEXCore::Context {
          }
        }
      }
-      
+
      Thread->OpDispatcher->Finalize();

      Thread->FrontendDecoder->DelayedDisownBuffer();
@@ -981,6 +1016,7 @@ namespace FEXCore::Context {
  }

  uintptr_t Context::CompileBlock(FEXCore::Core::CpuStateFrame *Frame, uint64_t GuestRIP) {
+    FEXCORE_PROFILE_SCOPED("CompileBlock");
    auto Thread = Frame->Thread;

    // Invalidate might take a unique lock on this, to guarantee that during invalidation no code gets compiled
@@ -1090,7 +1126,7 @@ namespace FEXCore::Context {
    // Now notify the thread that we are initialized
    Thread->ThreadWaiting.NotifyAll();

-    if (Thread != Thread->CTX->ParentThread || StartPaused) {
+    if (Thread != Thread->CTX->ParentThread || StartPaused || Thread->StartPaused) {
      // Parent thread doesn't need to wait to run
      Thread->StartRunning.Wait();
    }
@@ -1144,24 +1180,30 @@ namespace FEXCore::Context {

    for (auto it = lower; it != upper; it++) {
      for (auto Address: it->second) {
-        Context::RemoveThreadCodeEntry(Thread, Address);
+        Context::ThreadRemoveCodeEntry(Thread, Address);
      }
      it->second.clear();
    }
  }

-  void InvalidateGuestCodeRange(FEXCore::Context::Context *CTX, uint64_t Start, uint64_t Length) {
+  static void InvalidateGuestCodeRangeInternal(FEXCore::Context::Context *CTX, uint64_t Start, uint64_t Length) {
    std::lock_guard lk(CTX->ThreadCreationMutex);
-    
+
    for (auto &Thread : CTX->Threads) {
      InvalidateGuestThreadCodeRange(Thread, Start, Length);
    }
  }

-  void InvalidateGuestCodeRange(FEXCore::Context::Context *CTX, uint64_t Start, uint64_t Length, std::function<void(uint64_t start, uint64_t Length)> CallAfter) {
-    std::unique_lock CodeInvalidationLock(CTX->CodeInvalidationMutex);
+  void InvalidateGuestCodeRange(FEXCore::Context::Context *CTX, uint64_t Start, uint64_t Length) {
+    FHU::ScopedSignalMaskWithUniqueLock CodeInvalidationLock(CTX->CodeInvalidationMutex);

-    InvalidateGuestCodeRange(CTX, Start, Length);
+    InvalidateGuestCodeRangeInternal(CTX, Start, Length);
+  }
+
+  void InvalidateGuestCodeRange(FEXCore::Context::Context *CTX, uint64_t Start, uint64_t Length, std::function<void(uint64_t start, uint64_t Length)> CallAfter) {
+    FHU::ScopedSignalMaskWithUniqueLock CodeInvalidationLock(CTX->CodeInvalidationMutex);
+
+    InvalidateGuestCodeRangeInternal(CTX, Start, Length);
    CallAfter(Start, Length);
  }

@@ -1170,8 +1212,6 @@ namespace FEXCore::Context {
      IsMemoryShared = true;

      if (Config.TSOAutoMigration) {
-        LogMan::Msg::IFmt("Migrating to shared memory mode");
-
        std::lock_guard<std::mutex> lkThreads(ThreadCreationMutex);
        LogMan::Throw::AFmt(Threads.size() == 1, "First MarkMemoryShared called must be before creating any threads");

@@ -1191,7 +1231,15 @@ namespace FEXCore::Context {
    CTX->MarkMemoryShared();
  }

-  void Context::RemoveThreadCodeEntry(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP) {
+  void Context::ThreadAddBlockLink(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestDestination, uintptr_t HostLink, const std::function<void()> &delinker) {
+    std::shared_lock lk(Thread->CTX->CodeInvalidationMutex);
+
+    Thread->LookupCache->AddBlockLink(GuestDestination, HostLink, delinker);
+  }
+
+  void Context::ThreadRemoveCodeEntry(FEXCore::Core::InternalThreadState *Thread, uint64_t GuestRIP) {
+    LogMan::Throw::AFmt(Thread->CTX->CodeInvalidationMutex.try_lock() == false, "CodeInvalidationMutex needs to be unique_locked here");
+
    std::lock_guard<std::recursive_mutex> lk(Thread->LookupCache->WriteLock);

    Thread->DebugStore.erase(GuestRIP);
@@ -1230,7 +1278,7 @@ namespace FEXCore::Context {
    Thread->CurrentFrame->State.rip = RIP;

    // Erase the RIP from all the storage backings if it exists
-    RemoveThreadCodeEntry(Thread, RIP);
+    ThreadRemoveCodeEntry(Thread, RIP);

    // We don't care if compilation passes or not
    CompileBlock(Thread->CurrentFrame, RIP);
@@ -1288,8 +1336,13 @@ namespace FEXCore::Context {
    }
  }

+
+  void Context::AppendThunkDefinitions(std::vector<FEXCore::IR::ThunkDefinition> const& Definitions) {
+    ThunkHandler->AppendThunkDefinitions(Definitions);
+  }
+
  void ConfigureAOTGen(FEXCore::Core::InternalThreadState *Thread, std::set<uint64_t> *ExternalBranches, uint64_t SectionMaxAddress) {
    Thread->FrontendDecoder->SetExternalBranches(ExternalBranches);
    Thread->FrontendDecoder->SetSectionMaxAddress(SectionMaxAddress);
  }
-}
+}
@@ -38,12 +38,20 @@ namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;

-static constexpr size_t MAX_DISPATCHER_CODE_SIZE = 4096;
-#define STATE x28
+constexpr size_t MAX_DISPATCHER_CODE_SIZE = 8192;

 Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const DispatcherConfig &config)
-  : FEXCore::CPU::Dispatcher(ctx), Arm64Emitter(ctx, MAX_DISPATCHER_CODE_SIZE)
-  , config(config) {
+  : FEXCore::CPU::Dispatcher(ctx, config), Arm64Emitter(ctx, MAX_DISPATCHER_CODE_SIZE)
+#ifdef VIXL_SIMULATOR
+  , Simulator {&Decoder}
+#endif
+{
+
+#ifdef VIXL_SIMULATOR
+  // Hardcode a 256-bit vector width if we are running in the simulator.
+  Simulator.SetVectorLengthInBits(256);
+#endif
+
  SetAllowAssembler(true);

  DispatchPtr = GetCursorAddress<AsmDispatch>();
@@ -179,7 +187,12 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    ret();
  }

+#ifdef VIXL_SIMULATOR
+  // VIXL simulator can't run syscalls.
+  constexpr bool SignalSafeCompile = false;
+#else
  constexpr bool SignalSafeCompile = true;
+#endif
  {
    ExitFunctionLinkerAddress = GetCursorAddress<uint64_t>();
    if (config.StaticRegisterAllocation)
@@ -207,8 +220,12 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    mov(x0, STATE);
    mov(x1, lr);

-    ldr(x3, STATE_PTR(CpuStateFrame, Pointers.Common.ExitFunctionLink));
-    blr(x3);
+    ldr(x2, STATE_PTR(CpuStateFrame, Pointers.Common.ExitFunctionLink));
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<uintptr_t, void *, void *>(x2);
+#else
+    blr(x2);
+#endif

    if (SignalSafeCompile) {
      // Now restore the signal mask
@@ -267,8 +284,11 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    ldr(x3, &l_CompileBlock);

    // X2 contains our guest RIP
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<void, void *, uint64_t, void *>(x3);
+#else
    blr(x3); // { CTX, Frame, RIP}
-
+#endif

    if (SignalSafeCompile) {
      // Now restore the signal mask
@@ -301,7 +321,7 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
  {
    // Guest SIGILL handler
    // Needs to be distinct from the SignalHandlerReturnAddress
-    UnimplementedInstructionAddress = GetCursorAddress<uint64_t>();
+    GuestSignal_SIGILL  = GetCursorAddress<uint64_t>();

    if (config.StaticRegisterAllocation)
      SpillStaticRegs();
@@ -310,26 +330,29 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
  }

  {
-    // Guest Overflow handler
+    // Guest SIGTRAP handler
    // Needs to be distinct from the SignalHandlerReturnAddress
-    OverflowExceptionInstructionAddress = GetCursorAddress<uint64_t>();
+    GuestSignal_SIGTRAP  = GetCursorAddress<uint64_t>();

    if (config.StaticRegisterAllocation)
      SpillStaticRegs();

-    LoadConstant(w1, 1);
-    strb(w1, STATE_PTR(CpuStateFrame, SynchronousFaultData.FaultToTopAndGeneratedException));
-    LoadConstant(w1, X86State::X86_TRAPNO_OF);
-    str(w1, STATE_PTR(CpuStateFrame, SynchronousFaultData.TrapNo));
-    LoadConstant(w1, 0x80);
-    str(w1, STATE_PTR(CpuStateFrame, SynchronousFaultData.si_code));
-    LoadConstant(x1, 0);
-    str(w1, STATE_PTR(CpuStateFrame, SynchronousFaultData.err_code));
+    brk(0);
+  }
+
+  {
+    // Guest Overflow handler
+    // Needs to be distinct from the SignalHandlerReturnAddress
+    GuestSignal_SIGSEGV = GetCursorAddress<uint64_t>();
+
+    if (config.StaticRegisterAllocation)
+      SpillStaticRegs();

    // hlt/udf = SIGILL
    // brk = SIGTRAP
    // ??? = SIGSEGV
    // Force a SIGSEGV by loading zero
+    LoadConstant(x1, 0);
    ldr(x1, MemOperand(x1));
  }

@@ -347,7 +370,11 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    ldr(x0, &l_CTX);
    mov(x1, STATE);
    ldr(x2, &l_Sleep);
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<void, void *, void *>(x2);
+#else
    blr(x2);
+#endif

    PauseReturnInstruction = GetCursorAddress<uint64_t>();
    // Fault to start running again
@@ -410,11 +437,14 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    LUDIVHandlerAddress = GetCursorAddress<uint64_t>();

    PushDynamicRegsAndLR();
+    SpillStaticRegs();

    ldr(x3, STATE_PTR(CpuStateFrame, Pointers.AArch64.LUDIV));
-
-    SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t, uint64_t>(x3);
+#else
    blr(x3);
+#endif
    FillStaticRegs();

    // Result is now in x0
@@ -429,11 +459,14 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    LDIVHandlerAddress = GetCursorAddress<uint64_t>();

    PushDynamicRegsAndLR();
+    SpillStaticRegs();

    ldr(x3, STATE_PTR(CpuStateFrame, Pointers.AArch64.LDIV));
-
-    SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t, uint64_t>(x3);
+#else
    blr(x3);
+#endif
    FillStaticRegs();

    // Result is now in x0
@@ -448,11 +481,14 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    LUREMHandlerAddress = GetCursorAddress<uint64_t>();

    PushDynamicRegsAndLR();
+    SpillStaticRegs();

    ldr(x3, STATE_PTR(CpuStateFrame, Pointers.AArch64.LUREM));
-
-    SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t, uint64_t>(x3);
+#else
    blr(x3);
+#endif
    FillStaticRegs();

    // Result is now in x0
@@ -467,11 +503,14 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
    LREMHandlerAddress = GetCursorAddress<uint64_t>();

    PushDynamicRegsAndLR();
+    SpillStaticRegs();

    ldr(x3, STATE_PTR(CpuStateFrame, Pointers.AArch64.LREM));
-
-    SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+    GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t, uint64_t>(x3);
+#else
    blr(x3);
+#endif
    FillStaticRegs();

    // Result is now in x0
@@ -502,13 +541,27 @@ Arm64Dispatcher::Arm64Dispatcher(FEXCore::Context::Context *ctx, const Dispatche
  }
 }

+#ifdef VIXL_SIMULATOR
+void Arm64Dispatcher::ExecuteDispatch(FEXCore::Core::CpuStateFrame *Frame) {
+  Simulator.WriteXRegister(0, reinterpret_cast<int64_t>(Frame));
+  Simulator.RunFrom(reinterpret_cast<Instruction const*>(DispatchPtr));
+}
+
+void Arm64Dispatcher::ExecuteJITCallback(FEXCore::Core::CpuStateFrame *Frame, uint64_t RIP) {
+  Simulator.WriteXRegister(0, reinterpret_cast<int64_t>(Frame));
+  Simulator.WriteXRegister(1, RIP);
+  Simulator.RunFrom(reinterpret_cast<Instruction const*>(CallbackPtr));
+}
+
+#endif
+
 // Used by GenerateGDBPauseCheck, GenerateInterpreterTrampoline, destination buffer is set before use
 static thread_local vixl::aarch64::Assembler emit((uint8_t*)&emit, 1);

 size_t Arm64Dispatcher::GenerateGDBPauseCheck(uint8_t *CodeBuffer, uint64_t GuestRIP) {

  *emit.GetBuffer() = vixl::CodeBuffer(CodeBuffer, MaxGDBPauseCheckSize);
-  
+
  vixl::CodeBufferCheckScope scope(&emit, MaxGDBPauseCheckSize, vixl::CodeBufferCheckScope::kDontReserveBufferSpace, vixl::CodeBufferCheckScope::kNoAssert);

  aarch64::Label RunBlock;
@@ -543,8 +596,8 @@ size_t Arm64Dispatcher::GenerateGDBPauseCheck(uint8_t *CodeBuffer, uint64_t Gues
 }

 size_t Arm64Dispatcher::GenerateInterpreterTrampoline(uint8_t *CodeBuffer) {
-  LOGMAN_THROW_A_FMT(!config.StaticRegisterAllocation, "GenerateInterpreterTrampoline dispatcher does not support SRA");
-  
+  LOGMAN_THROW_AA_FMT(!config.StaticRegisterAllocation, "GenerateInterpreterTrampoline dispatcher does not support SRA");
+
  *emit.GetBuffer() = vixl::CodeBuffer(CodeBuffer, MaxInterpreterTrampolineSize);

  vixl::CodeBufferCheckScope scope(&emit, MaxInterpreterTrampolineSize, vixl::CodeBufferCheckScope::kDontReserveBufferSpace, vixl::CodeBufferCheckScope::kNoAssert);
@@ -570,7 +623,7 @@ size_t Arm64Dispatcher::GenerateInterpreterTrampoline(uint8_t *CodeBuffer) {
 }

 void Arm64Dispatcher::SpillSRA(FEXCore::Core::InternalThreadState *Thread, void *ucontext, uint32_t IgnoreMask) {
-  for(int i = 0; i < SRA64.size(); i++) {
+  for (size_t i = 0; i < SRA64.size(); i++) {
    if (IgnoreMask & (1U << SRA64[i].GetCode())) {
      // Skip this one, it's already spilled
      continue;
@@ -578,14 +631,21 @@ void Arm64Dispatcher::SpillSRA(FEXCore::Core::InternalThreadState *Thread, void
    Thread->CurrentFrame->State.gregs[i] = ArchHelpers::Context::GetArmReg(ucontext, SRA64[i].GetCode());
  }

-  for(int i = 0; i < SRAFPR.size(); i++) {
-    auto FPR = ArchHelpers::Context::GetArmFPR(ucontext, SRAFPR[i].GetCode());
-    memcpy(&Thread->CurrentFrame->State.xmm[i][0], &FPR, sizeof(__uint128_t));
+  if (EmitterCTX->HostFeatures.SupportsAVX) {
+    for (size_t i = 0; i < SRAFPR.size(); i++) {
+      auto FPR = ArchHelpers::Context::GetArmFPR(ucontext, SRAFPR[i].GetCode());
+      memcpy(&Thread->CurrentFrame->State.xmm.avx.data[i][0], &FPR, sizeof(__uint128_t));
+    }
+  } else {
+    for (size_t i = 0; i < SRAFPR.size(); i++) {
+      auto FPR = ArchHelpers::Context::GetArmFPR(ucontext, SRAFPR[i].GetCode());
+      memcpy(&Thread->CurrentFrame->State.xmm.sse.data[i][0], &FPR, sizeof(__uint128_t));
+    }
  }
 }

 void Arm64Dispatcher::InitThreadPointers(FEXCore::Core::InternalThreadState *Thread) {
-// Setup dispatcher specific pointers that need to be accessed from JIT code
+  // Setup dispatcher specific pointers that need to be accessed from JIT code
  {
    auto &Common = Thread->CurrentFrame->Pointers.Common;

@@ -594,8 +654,9 @@ void Arm64Dispatcher::InitThreadPointers(FEXCore::Core::InternalThreadState *Thr
    Common.ExitFunctionLinker = ExitFunctionLinkerAddress;
    Common.ThreadStopHandlerSpillSRA = ThreadStopHandlerAddressSpillSRA;
    Common.ThreadPauseHandlerSpillSRA = ThreadPauseHandlerAddressSpillSRA;
-    Common.UnimplementedInstructionHandler = UnimplementedInstructionAddress;
-    Common.OverflowExceptionHandler = OverflowExceptionInstructionAddress;
+    Common.GuestSignal_SIGILL = GuestSignal_SIGILL;
+    Common.GuestSignal_SIGTRAP = GuestSignal_SIGTRAP;
+    Common.GuestSignal_SIGSEGV = GuestSignal_SIGSEGV;
    Common.SignalReturnHandler = SignalHandlerReturnAddress;

    auto &AArch64 = Thread->CurrentFrame->Pointers.AArch64;
@@ -610,4 +671,4 @@ std::unique_ptr<Dispatcher> Dispatcher::CreateArm64(FEXCore::Context::Context *C
  return std::make_unique<Arm64Dispatcher>(CTX, Config);
 }

-}
+}
@@ -3,6 +3,10 @@
 #include "Interface/Core/ArchHelpers/Arm64Emitter.h"
 #include "Interface/Core/Dispatcher/Dispatcher.h"

+#ifdef VIXL_SIMULATOR
+#include <aarch64/simulator-aarch64.h>
+#endif
+
 namespace FEXCore::Context {
 struct Context;
 }
@@ -20,6 +24,11 @@ class Arm64Dispatcher final : public Dispatcher, public Arm64Emitter {
    size_t GenerateGDBPauseCheck(uint8_t *CodeBuffer, uint64_t GuestRIP) override;
    size_t GenerateInterpreterTrampoline(uint8_t *CodeBuffer) override;

+#ifdef VIXL_SIMULATOR
+  void ExecuteDispatch(FEXCore::Core::CpuStateFrame *Frame) override;
+  void ExecuteJITCallback(FEXCore::Core::CpuStateFrame *Frame, uint64_t RIP) override;
+#endif
+
  protected:
    void SpillSRA(FEXCore::Core::InternalThreadState *Thread, void *ucontext, uint32_t IgnoreMask) override;

@@ -29,7 +38,11 @@ class Arm64Dispatcher final : public Dispatcher, public Arm64Emitter {
    uint64_t LDIVHandlerAddress{};
    uint64_t LUREMHandlerAddress{};
    uint64_t LREMHandlerAddress{};
-    DispatcherConfig config;
+
+#ifdef VIXL_SIMULATOR
+    vixl::aarch64::Decoder Decoder;
+    vixl::aarch64::Simulator Simulator;
+#endif
 };

 }
@@ -99,6 +99,7 @@ void Dispatcher::RestoreThreadState(FEXCore::Core::InternalThreadState *Thread,
    SignalFrames.pop();
  }

+  const bool IsAVXEnabled = CTX->Config.EnableAVX;
  uintptr_t NewSP = OldSP;
  auto Context = reinterpret_cast<ArchHelpers::Context::ContextBackup*>(NewSP);

@@ -159,10 +160,22 @@ void Dispatcher::RestoreThreadState(FEXCore::Core::InternalThreadState *Thread,
            COPY_REG(RCX);
            COPY_REG(RSP);
 #undef COPY_REG
-        FEXCore::x86_64::_libc_fpstate *fpstate = reinterpret_cast<FEXCore::x86_64::_libc_fpstate*>(guest_uctx->uc_mcontext.fpregs);
+        auto *xstate = reinterpret_cast<x86_64::xstate*>(guest_uctx->uc_mcontext.fpregs);
+        auto *fpstate = &xstate->fpstate;
+
        // Copy float registers
        memcpy(Frame->State.mm, fpstate->_st, sizeof(Frame->State.mm));
-        memcpy(Frame->State.xmm, fpstate->_xmm, sizeof(Frame->State.xmm));
+
+        if (IsAVXEnabled) {
+          for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+            memcpy(&Frame->State.xmm.avx.data[i][0], &fpstate->_xmm[i], sizeof(__uint128_t));
+          }
+          for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+            memcpy(&Frame->State.xmm.avx.data[i][2], &xstate->ymmh.ymmh_space[i], sizeof(__uint128_t));
+          }
+        } else {
+          memcpy(Frame->State.xmm.sse.data, fpstate->_xmm, sizeof(Frame->State.xmm.sse.data));
+        }

        // FCW store default
        Frame->State.FCW = fpstate->fcw;
@@ -199,12 +212,20 @@ void Dispatcher::RestoreThreadState(FEXCore::Core::InternalThreadState *Thread,
        Frame->State.flags[9] = 1;

        Frame->State.rip = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_EIP];
-        Frame->State.cs = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_CS];
-        Frame->State.ds = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_DS];
-        Frame->State.es = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ES];
-        Frame->State.fs = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_FS];
-        Frame->State.gs = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_GS];
-        Frame->State.ss = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_SS];
+        Frame->State.cs_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_CS];
+        Frame->State.ds_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_DS];
+        Frame->State.es_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ES];
+        Frame->State.fs_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_FS];
+        Frame->State.gs_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_GS];
+        Frame->State.ss_idx = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_SS];
+
+        Frame->State.cs_cached = Frame->State.gdt[Frame->State.cs_idx >> 3].base;
+        Frame->State.ds_cached = Frame->State.gdt[Frame->State.ds_idx >> 3].base;
+        Frame->State.es_cached = Frame->State.gdt[Frame->State.es_idx >> 3].base;
+        Frame->State.fs_cached = Frame->State.gdt[Frame->State.fs_idx >> 3].base;
+        Frame->State.gs_cached = Frame->State.gdt[Frame->State.gs_idx >> 3].base;
+        Frame->State.ss_cached = Frame->State.gdt[Frame->State.ss_idx >> 3].base;
+
 #define COPY_REG(x) \
      Frame->State.gregs[X86State::REG_##x] = guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_##x];
        COPY_REG(RDI);
@@ -216,7 +237,8 @@ void Dispatcher::RestoreThreadState(FEXCore::Core::InternalThreadState *Thread,
        COPY_REG(RCX);
        COPY_REG(RSP);
 #undef COPY_REG
-        FEXCore::x86::_libc_fpstate *fpstate = reinterpret_cast<FEXCore::x86::_libc_fpstate*>(guest_uctx->uc_mcontext.fpregs);
+        auto *xstate = reinterpret_cast<x86::xstate*>(guest_uctx->uc_mcontext.fpregs);
+        auto *fpstate = &xstate->fpstate;

        // Copy float registers
        for (size_t i = 0; i < Core::CPUState::NUM_MMS; ++i) {
@@ -225,7 +247,16 @@ void Dispatcher::RestoreThreadState(FEXCore::Core::InternalThreadState *Thread,
        }

        // Extended XMM state
-        memcpy(fpstate->_xmm, Frame->State.xmm, sizeof(Frame->State.xmm));
+        if (IsAVXEnabled) {
+          for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+            memcpy(&fpstate->_xmm[i], &Frame->State.xmm.avx.data[i][0], sizeof(__uint128_t));
+          }
+          for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+            memcpy(&xstate->ymmh.ymmh_space[i], &Frame->State.xmm.avx.data[i][2], sizeof(__uint128_t));
+          }
+        } else {
+          memcpy(Frame->State.xmm.sse.data, fpstate->_xmm, sizeof(Frame->State.xmm.sse.data));
+        }

        // FCW store default
        Frame->State.FCW = fpstate->fcw;
@@ -274,6 +305,26 @@ static uint32_t ConvertSignalToError(int Signal, siginfo_t *HostSigInfo) {
  return 0;
 }

+template <typename T>
+static void SetXStateInfo(T* xstate, bool is_avx_enabled) {
+  auto* fpstate = &xstate->fpstate;
+
+  fpstate->sw_reserved.magic1 = x86_64::fpx_sw_bytes::FP_XSTATE_MAGIC;
+  fpstate->sw_reserved.extended_size = is_avx_enabled ? sizeof(T) : 0;
+
+  fpstate->sw_reserved.xfeatures |= x86_64::fpx_sw_bytes::FEATURE_FP |
+                                    x86_64::fpx_sw_bytes::FEATURE_SSE;
+  if (is_avx_enabled) {
+    fpstate->sw_reserved.xfeatures |= x86_64::fpx_sw_bytes::FEATURE_YMM;
+  }
+
+  fpstate->sw_reserved.xstate_size = fpstate->sw_reserved.extended_size;
+
+  if (is_avx_enabled) {
+    xstate->xstate_hdr.xfeatures = 0;
+  }
+}
+
 bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, int Signal, void *info, void *ucontext, GuestSigAction *GuestAction, stack_t *GuestStack) {
  auto ContextBackup = StoreThreadState(Thread, Signal, ucontext);

@@ -293,13 +344,14 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
  uint64_t NewGuestSP = OldGuestSP;

  // Pulling from context here
-  bool Is64BitMode = CTX->Config.Is64BitMode;
-  uint64_t SignalReturn = CTX->X86CodeGen.SignalReturn;
+  const bool Is64BitMode = CTX->Config.Is64BitMode;
+  const bool IsAVXEnabled = CTX->Config.EnableAVX;
+  const uint64_t SignalReturn = CTX->X86CodeGen.SignalReturn;

  // Spill the SRA regardless of signal handler type
  // We are going to be returning to the top of the dispatcher which will fill again
  // Otherwise we might load garbage
-  if (SRAEnabled) {
+  if (config.StaticRegisterAllocation) {
    if (Thread->CPUBackend->IsAddressInCodeBuffer(OldPC)) {
      uint32_t IgnoreMask{};
 #ifdef _M_ARM_64
@@ -374,8 +426,13 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
  if (GuestAction->sa_flags & SA_SIGINFO) {
    // Setup ucontext a bit
    if (Is64BitMode) {
-      NewGuestSP -= sizeof(FEXCore::x86_64::_libc_fpstate);
-      NewGuestSP = AlignDown(NewGuestSP, alignof(FEXCore::x86_64::_libc_fpstate));
+      if (IsAVXEnabled) {
+        NewGuestSP -= sizeof(x86_64::xstate);
+        NewGuestSP = AlignDown(NewGuestSP, alignof(x86_64::xstate));
+      } else {
+        NewGuestSP -= sizeof(x86_64::_libc_fpstate);
+        NewGuestSP = AlignDown(NewGuestSP, alignof(x86_64::_libc_fpstate));
+      }
      uint64_t FPStateLocation = NewGuestSP;

      NewGuestSP -= sizeof(FEXCore::x86_64::ucontext_t);
@@ -397,8 +454,9 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
      guest_uctx->uc_flags = FEXCore::x86_64::UC_FP_XSTATE;

      // Pointer to where the fpreg memory is
-      guest_uctx->uc_mcontext.fpregs = reinterpret_cast<FEXCore::x86_64::_libc_fpstate*>(FPStateLocation);
-      FEXCore::x86_64::_libc_fpstate *fpstate = reinterpret_cast<FEXCore::x86_64::_libc_fpstate*>(FPStateLocation);
+      guest_uctx->uc_mcontext.fpregs = reinterpret_cast<x86_64::_libc_fpstate*>(FPStateLocation);
+      auto *xstate = reinterpret_cast<x86_64::xstate*>(FPStateLocation);
+      SetXStateInfo(xstate, IsAVXEnabled);

      guest_uctx->uc_mcontext.gregs[FEXCore::x86_64::FEX_REG_RIP] = Frame->State.rip;
      guest_uctx->uc_mcontext.gregs[FEXCore::x86_64::FEX_REG_EFL] = 0;
@@ -415,6 +473,7 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i

        // Overwrite si_code
        guest_siginfo->si_code = Thread->CurrentFrame->SynchronousFaultData.si_code;
+        Signal = Frame->SynchronousFaultData.Signal;
      }
      else {
        guest_uctx->uc_mcontext.gregs[FEXCore::x86_64::FEX_REG_TRAPNO] = ConvertSignalToTrapNo(Signal, HostSigInfo);
@@ -443,9 +502,21 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
      COPY_REG(RSP);
 #undef COPY_REG

+      auto* fpstate = &xstate->fpstate;
+
      // Copy float registers
      memcpy(fpstate->_st, Frame->State.mm, sizeof(Frame->State.mm));
-      memcpy(fpstate->_xmm, Frame->State.xmm, sizeof(Frame->State.xmm));
+
+      if (IsAVXEnabled) {
+        for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+          memcpy(&fpstate->_xmm[i], &Frame->State.xmm.avx.data[i][0], sizeof(__uint128_t));
+        }
+        for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+          memcpy(&xstate->ymmh.ymmh_space[i], &Frame->State.xmm.avx.data[i][2], sizeof(__uint128_t));
+        }
+      } else {
+        memcpy(fpstate->_xmm, Frame->State.xmm.sse.data, sizeof(Frame->State.xmm.sse.data));
+      }

      // FCW store default
      fpstate->fcw = Frame->State.FCW;
@@ -470,8 +541,13 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
    else {
      ContextBackup->Flags |= ArchHelpers::Context::ContextFlags::CONTEXT_FLAG_32BIT;

-      NewGuestSP -= sizeof(FEXCore::x86::_libc_fpstate);
-      NewGuestSP = AlignDown(NewGuestSP, alignof(FEXCore::x86::_libc_fpstate));
+      if (IsAVXEnabled) {
+        NewGuestSP -= sizeof(x86::xstate);
+        NewGuestSP = AlignDown(NewGuestSP, alignof(x86::xstate));
+      } else {
+        NewGuestSP -= sizeof(x86::_libc_fpstate);
+        NewGuestSP = AlignDown(NewGuestSP, alignof(x86::_libc_fpstate));
+      }
      uint64_t FPStateLocation = NewGuestSP;

      NewGuestSP -= sizeof(FEXCore::x86::ucontext_t);
@@ -494,27 +570,30 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i

      // Pointer to where the fpreg memory is
      guest_uctx->uc_mcontext.fpregs = static_cast<uint32_t>(FPStateLocation);
-      FEXCore::x86::_libc_fpstate *fpstate = reinterpret_cast<FEXCore::x86::_libc_fpstate*>(FPStateLocation);
+      auto *xstate = reinterpret_cast<x86::xstate*>(FPStateLocation);
+      SetXStateInfo(xstate, IsAVXEnabled);
+
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_CS] = Frame->State.cs_idx;
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_DS] = Frame->State.ds_idx;
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ES] = Frame->State.es_idx;
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_FS] = Frame->State.fs_idx;
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_GS] = Frame->State.gs_idx;
+      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_SS] = Frame->State.ss_idx;

-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_GS] = Frame->State.gs;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_FS] = Frame->State.fs;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ES] = Frame->State.es;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_DS] = Frame->State.ds;
      if (ContextBackup->FaultToTopAndGeneratedException) {
        guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_TRAPNO] = Frame->SynchronousFaultData.TrapNo;
        guest_siginfo->si_code = Frame->SynchronousFaultData.si_code;
        guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ERR] = Frame->SynchronousFaultData.err_code;
+        Signal = Frame->SynchronousFaultData.Signal;
      }
      else {
        guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_TRAPNO] = ConvertSignalToTrapNo(Signal, HostSigInfo);
        guest_siginfo->si_code = HostSigInfo->si_code;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ERR] = ConvertSignalToError(Signal, HostSigInfo);
+        guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_ERR] = ConvertSignalToError(Signal, HostSigInfo);
      }
      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_EIP] = Frame->State.rip;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_CS] = Frame->State.cs;
      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_EFL] = 0;
      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_UESP] = 0;
-      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_SS] = Frame->State.ss;

 #define COPY_REG(x) \
      guest_uctx->uc_mcontext.gregs[FEXCore::x86::FEX_REG_##x] = Frame->State.gregs[X86State::REG_##x];
@@ -528,6 +607,8 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
      COPY_REG(RSP);
 #undef COPY_REG

+      auto *fpstate = &xstate->fpstate;
+
      // Copy float registers
      for (size_t i = 0; i < Core::CPUState::NUM_MMS; ++i) {
        // 32-bit st register size is only 10 bytes. Not padded to 16byte like x86-64
@@ -536,7 +617,16 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i

      // Extended XMM state
      fpstate->status = FEXCore::x86::fpstate_magic::MAGIC_XFPSTATE;
-      memcpy(fpstate->_xmm, Frame->State.xmm, sizeof(Frame->State.xmm));
+      if (IsAVXEnabled) {
+        for (size_t i = 0; i < std::size(Frame->State.xmm.avx.data); i++) {
+          memcpy(&fpstate->_xmm[i], &Frame->State.xmm.avx.data[i][0], sizeof(__uint128_t));
+        }
+        for (size_t i = 0; i < std::size(Frame->State.xmm.avx.data); i++) {
+          memcpy(&xstate->ymmh.ymmh_space[i], &Frame->State.xmm.avx.data[i][2], sizeof(__uint128_t));
+        }
+      } else {
+        memcpy(fpstate->_xmm, Frame->State.xmm.sse.data, sizeof(Frame->State.xmm.sse.data));
+      }

      // FCW store default
      fpstate->fcw = Frame->State.FCW;
@@ -619,14 +709,14 @@ bool Dispatcher::HandleGuestSignal(FEXCore::Core::InternalThreadState *Thread, i
  else {
    NewGuestSP -= 4;
    *(uint32_t*)NewGuestSP = SignalReturn;
-    LOGMAN_THROW_A_FMT(SignalReturn < 0x1'0000'0000ULL, "This needs to be below 4GB");
+    LOGMAN_THROW_AA_FMT(SignalReturn < 0x1'0000'0000ULL, "This needs to be below 4GB");
    Frame->State.gregs[FEXCore::X86State::REG_RSP] = NewGuestSP;
  }

  // The guest starts its signal frame with a zero initialized FPU
  // Set that up now. Little bit costly but it's a requirement
  // This state will be restored on rt_sigreturn
-  memset(Frame->State.xmm, 0, sizeof(Frame->State.xmm));
+  memset(Frame->State.xmm.avx.data, 0, sizeof(Frame->State.xmm));
  memset(Frame->State.mm, 0, sizeof(Frame->State.mm));
  Frame->State.FCW = 0x37F;
  Frame->State.FTW = 0xFFFF;
@@ -665,11 +755,11 @@ bool Dispatcher::HandleSignalPause(FEXCore::Core::InternalThreadState *Thread, i
    // Store our thread state so we can come back to this
    StoreThreadState(Thread, Signal, ucontext);

-    if (SRAEnabled && Thread->CPUBackend->IsAddressInCodeBuffer(ArchHelpers::Context::GetPc(ucontext))) {
+    if (config.StaticRegisterAllocation && Thread->CPUBackend->IsAddressInCodeBuffer(ArchHelpers::Context::GetPc(ucontext))) {
      // We are in jit, SRA must be spilled
      ArchHelpers::Context::SetPc(ucontext, ThreadPauseHandlerAddressSpillSRA);
    } else {
-      if (SRAEnabled) {
+      if (config.StaticRegisterAllocation) {
        // We are in non-jit, SRA is already spilled
        LOGMAN_THROW_A_FMT(!IsAddressInDispatcher(ArchHelpers::Context::GetPc(ucontext)),
                           "Signals in dispatcher have unsynchronized context");
@@ -698,11 +788,11 @@ bool Dispatcher::HandleSignalPause(FEXCore::Core::InternalThreadState *Thread, i
    Thread->CurrentFrame->SignalHandlerRefCounter = 0;

    // Set the new PC
-    if (SRAEnabled && Thread->CPUBackend->IsAddressInCodeBuffer(ArchHelpers::Context::GetPc(ucontext))) {
+    if (config.StaticRegisterAllocation && Thread->CPUBackend->IsAddressInCodeBuffer(ArchHelpers::Context::GetPc(ucontext))) {
      // We are in jit, SRA must be spilled
      ArchHelpers::Context::SetPc(ucontext, ThreadStopHandlerAddressSpillSRA);
    } else {
-      if (SRAEnabled) {
+      if (config.StaticRegisterAllocation) {
        // We are in non-jit, SRA is already spilled
        LOGMAN_THROW_A_FMT(!IsAddressInDispatcher(ArchHelpers::Context::GetPc(ucontext)),
                           "Signals in dispatcher have unsynchronized context");
@@ -32,7 +32,7 @@ struct DispatcherConfig {
 class Dispatcher {
 public:
  virtual ~Dispatcher() = default;
-  
+
  /**
   * @name Dispatch Helper functions
   * @{ */
@@ -44,8 +44,9 @@ public:
  uint64_t ThreadPauseHandlerAddressSpillSRA{};
  uint64_t ExitFunctionLinkerAddress{};
  uint64_t SignalHandlerReturnAddress{};
-  uint64_t UnimplementedInstructionAddress{};
-  uint64_t OverflowExceptionInstructionAddress{};
+  uint64_t GuestSignal_SIGILL{};
+  uint64_t GuestSignal_SIGTRAP{};
+  uint64_t GuestSignal_SIGSEGV{};
  uint64_t IntCallbackReturnAddress{};

  uint64_t PauseReturnInstruction{};
@@ -74,28 +75,29 @@ public:

  static std::unique_ptr<Dispatcher> CreateX86(FEXCore::Context::Context *CTX, const DispatcherConfig &Config);
  static std::unique_ptr<Dispatcher> CreateArm64(FEXCore::Context::Context *CTX, const DispatcherConfig &Config);
-  
-  void ExecuteDispatch(FEXCore::Core::CpuStateFrame *Frame) {
+
+  virtual void ExecuteDispatch(FEXCore::Core::CpuStateFrame *Frame) {
    DispatchPtr(Frame);
  }

-  void ExecuteJITCallback(FEXCore::Core::CpuStateFrame *Frame, uint64_t RIP) {
+  virtual void ExecuteJITCallback(FEXCore::Core::CpuStateFrame *Frame, uint64_t RIP) {
    CallbackPtr(Frame, RIP);
  }

 protected:
-  Dispatcher(FEXCore::Context::Context *ctx)
+  Dispatcher(FEXCore::Context::Context *ctx, const DispatcherConfig &Config)
    : CTX {ctx}
+    , config {Config}
    {}

  ArchHelpers::Context::ContextBackup* StoreThreadState(FEXCore::Core::InternalThreadState *Thread, int Signal, void *ucontext);
  void RestoreThreadState(FEXCore::Core::InternalThreadState *Thread, void *ucontext);
  std::stack<uint64_t, std::vector<uint64_t>> SignalFrames;

-  bool SRAEnabled = false;
  virtual void SpillSRA(FEXCore::Core::InternalThreadState *Thread, void *ucontext, uint32_t IgnoreMask) {}

  FEXCore::Context::Context *CTX;
+  DispatcherConfig config;

  static void SleepThread(FEXCore::Context::Context *ctx, FEXCore::Core::CpuStateFrame *Frame);

@@ -28,12 +28,12 @@ static constexpr size_t MAX_DISPATCHER_CODE_SIZE = 4096;
 #define STATE r14

 X86Dispatcher::X86Dispatcher(FEXCore::Context::Context *ctx, const DispatcherConfig &config)
-  : Dispatcher(ctx)
+  : Dispatcher(ctx, config)
  , Xbyak::CodeGenerator(MAX_DISPATCHER_CODE_SIZE,
      FEXCore::Allocator::mmap(nullptr, MAX_DISPATCHER_CODE_SIZE, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0),
      nullptr) {

-  LOGMAN_THROW_A_FMT(!config.StaticRegisterAllocation, "X86 dispatcher does not support SRA");
+  LOGMAN_THROW_AA_FMT(!config.StaticRegisterAllocation, "X86 dispatcher does not support SRA");

  using namespace Xbyak;
  using namespace Xbyak::util;
@@ -347,24 +347,29 @@ X86Dispatcher::X86Dispatcher(FEXCore::Context::Context *ctx, const DispatcherCon
  {
    // Guest SIGILL handler
    // Needs to be distinct from the SignalHandlerReturnAddress
-    UnimplementedInstructionAddress = getCurr<uint64_t>();
+    GuestSignal_SIGILL = getCurr<uint64_t>();
    ud2();
  }

  {
-    // Guest Overflow handler
+    // Guest SIGTRAP handler
    // Needs to be distinct from the SignalHandlerReturnAddress
-    OverflowExceptionInstructionAddress = getCurr<uint64_t>();
+    GuestSignal_SIGTRAP = getCurr<uint64_t>();

    // ud2 = SIGILL
    // int3 = SIGTRAP
    // hlt = SIGSEGV
-  
-    add(byte STATE_PTR(CpuStateFrame, SynchronousFaultData.FaultToTopAndGeneratedException), 1);
-    mov(dword STATE_PTR(CpuStateFrame, SynchronousFaultData.TrapNo), X86State::X86_TRAPNO_OF);
-    mov(dword STATE_PTR(CpuStateFrame, SynchronousFaultData.err_code), 0);
-    mov(dword STATE_PTR(CpuStateFrame, SynchronousFaultData.si_code), 0x80);
+    int3();
+  }

+  {
+    // Guest SIGSEGV handler
+    // Needs to be distinct from the SignalHandlerReturnAddress
+    GuestSignal_SIGSEGV = getCurr<uint64_t>();
+
+    // ud2 = SIGILL
+    // int3 = SIGTRAP
+    // hlt = SIGSEGV
    hlt();
  }

@@ -477,8 +482,9 @@ void X86Dispatcher::InitThreadPointers(FEXCore::Core::InternalThreadState *Threa
    Common.ExitFunctionLinker = ExitFunctionLinkerAddress;
    Common.ThreadStopHandlerSpillSRA = ThreadStopHandlerAddress;
    Common.ThreadPauseHandlerSpillSRA = ThreadPauseHandlerAddress;
-    Common.UnimplementedInstructionHandler = UnimplementedInstructionAddress;
-    Common.OverflowExceptionHandler = OverflowExceptionInstructionAddress;
+    Common.GuestSignal_SIGILL = GuestSignal_SIGILL;
+    Common.GuestSignal_SIGTRAP = GuestSignal_SIGTRAP;
+    Common.GuestSignal_SIGSEGV = GuestSignal_SIGSEGV;
    Common.SignalReturnHandler = SignalHandlerReturnAddress;

    auto &Interpreter = Thread->CurrentFrame->Pointers.Interpreter;
@@ -18,6 +18,7 @@ $end_info$
 #include <FEXCore/HLE/SyscallHandler.h>
 #include <FEXCore/Utils/Allocator.h>
 #include <FEXCore/Utils/LogManager.h>
+#include <FEXCore/Utils/Profiler.h>
 #include <FEXCore/Utils/Telemetry.h>
 #include <FEXHeaderUtils/TypeDefines.h>
 #include <set>
@@ -188,7 +189,7 @@ Decoder::~Decoder() {

 uint8_t Decoder::ReadByte() {
  uint8_t Byte = InstStream[InstructionSize];
-  LOGMAN_THROW_A_FMT(InstructionSize < MAX_INST_SIZE, "Max instruction size exceeded!");
+  LOGMAN_THROW_AA_FMT(InstructionSize < MAX_INST_SIZE, "Max instruction size exceeded!");
  Instruction[InstructionSize] = Byte;
  InstructionSize++;
  return Byte;
@@ -200,14 +201,7 @@ uint8_t Decoder::PeekByte(uint8_t Offset) const {
 }

 uint64_t Decoder::ReadData(uint8_t Size) {
-  if (Size == 0) {
-    return 0;
-  }
-
-  if (Size > sizeof(uint64_t)) {
-    LOGMAN_MSG_A_FMT("Unknown data size to read");
-    return 0;
-  }
+  LOGMAN_THROW_AA_FMT(Size != 0 && Size <= sizeof(uint64_t), "Unknown data size to read");

  uint64_t Res = 0;
  std::memcpy(&Res, &InstStream[InstructionSize], Size);
@@ -347,13 +341,15 @@ void Decoder::DecodeModRM_64(X86Tables::DecodedOperand *Operand, X86Tables::ModR
    Operand->Data.SIB.Index = MapModRMToReg(DecodeInst->Flags & DecodeFlags::FLAG_REX_XGPR_X ? 1 : 0, SIB.index, false, false, false, false, 0b100);
    Operand->Data.SIB.Base  = MapModRMToReg(DecodeInst->Flags & DecodeFlags::FLAG_REX_XGPR_B ? 1 : 0, SIB.base, false, false, false, false, ModRM.mod == 0 ? 0b101 : 16);

-    LOGMAN_THROW_A_FMT(Displacement <= 4, "Number of bytes should be <= 4 for literal src");
+    LOGMAN_THROW_AA_FMT(Displacement <= 4, "Number of bytes should be <= 4 for literal src");

-    uint64_t Literal = ReadData(Displacement);
-    if (Displacement == 1) {
-      Literal = static_cast<int8_t>(Literal);
+    if (Displacement) {
+      uint64_t Literal = ReadData(Displacement);
+      if (Displacement == 1) {
+        Literal = static_cast<int8_t>(Literal);
+      }
+      Operand->Data.SIB.Offset = Literal;
    }
-    Operand->Data.SIB.Offset = Literal;
  }
  else if (ModRM.mod == 0) {
    // Explained in Table 1-14. "Operand Addressing Using ModRM and SIB Bytes"
@@ -404,7 +400,7 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
    return false;
  }

-  LOGMAN_THROW_A_FMT(!(Info->Type >= FEXCore::X86Tables::TYPE_GROUP_1 && Info->Type <= FEXCore::X86Tables::TYPE_GROUP_P),
+  LOGMAN_THROW_AA_FMT(!(Info->Type >= FEXCore::X86Tables::TYPE_GROUP_1 && Info->Type <= FEXCore::X86Tables::TYPE_GROUP_P),
                     "Group Ops should have been decoded before this!");

  uint8_t DestSize{};
@@ -455,8 +451,13 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
      DestSize = 2;
    }
    else if (DstSizeFlag == FEXCore::X86Tables::InstFlags::SIZE_128BIT) {
-      DecodeInst->Flags |= DecodeFlags::GenSizeDstSize(DecodeFlags::SIZE_128BIT);
-      DestSize = 16;
+      if (Options.L) {
+        DecodeInst->Flags |= DecodeFlags::GenSizeDstSize(DecodeFlags::SIZE_256BIT);
+        DestSize = 32;
+      } else {
+        DecodeInst->Flags |= DecodeFlags::GenSizeDstSize(DecodeFlags::SIZE_128BIT);
+        DestSize = 16;
+      }
    }
    else if (HasNarrowingDisplacement &&
      (DstSizeFlag == FEXCore::X86Tables::InstFlags::SIZE_DEF ||
@@ -487,7 +488,14 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
      DecodeInst->Flags |= DecodeFlags::GenSizeSrcSize(DecodeFlags::SIZE_16BIT);
    }
    else if (SrcSizeFlag == FEXCore::X86Tables::InstFlags::SIZE_128BIT) {
-      DecodeInst->Flags |= DecodeFlags::GenSizeSrcSize(DecodeFlags::SIZE_128BIT);
+      if (Options.L) {
+        DecodeInst->Flags |= DecodeFlags::GenSizeSrcSize(DecodeFlags::SIZE_256BIT);
+      } else {
+        DecodeInst->Flags |= DecodeFlags::GenSizeSrcSize(DecodeFlags::SIZE_128BIT);
+      }
+    }
+    else if (SrcSizeFlag == FEXCore::X86Tables::InstFlags::SIZE_256BIT) {
+      DecodeInst->Flags |= DecodeFlags::GenSizeSrcSize(DecodeFlags::SIZE_256BIT);
    }
    else if (HasNarrowingDisplacement &&
      (SrcSizeFlag == FEXCore::X86Tables::InstFlags::SIZE_DEF ||
@@ -528,7 +536,7 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
  }

  if (HAS_NON_XMM_SUBFLAG(Info->Flags, FEXCore::X86Tables::InstFlags::FLAGS_SF_REX_IN_BYTE)) {
-    LOGMAN_THROW_A_FMT(!HasMODRM, "This instruction shouldn't have ModRM!");
+    LOGMAN_THROW_AA_FMT(!HasMODRM, "This instruction shouldn't have ModRM!");

    // If the REX is in the byte that means the lower nibble of the OP contains the destination GPR
    // This also means that the destination is always a GPR on these ones
@@ -637,7 +645,7 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
  }

  if (Bytes != 0) {
-    LOGMAN_THROW_A_FMT(Bytes <= 8, "Number of bytes should be <= 8 for literal src");
+    LOGMAN_THROW_AA_FMT(Bytes <= 8, "Number of bytes should be <= 8 for literal src");

    DecodeInst->Src[CurrentSrc].Data.Literal.Size = Bytes;

@@ -662,7 +670,7 @@ bool Decoder::NormalOp(FEXCore::X86Tables::X86InstInfo const *Info, uint16_t Op,
    DecodeInst->Src[CurrentSrc].Data.Literal.Value = Literal;
  }

-  LOGMAN_THROW_A_FMT(Bytes == 0, "Inst at 0x{:x}: 0x{:04x} '{}' Had an instruction of size {} with {} remaining",
+  LOGMAN_THROW_AA_FMT(Bytes == 0, "Inst at 0x{:x}: 0x{:04x} '{}' Had an instruction of size {} with {} remaining",
                     DecodeInst->PC, DecodeInst->OP, DecodeInst->TableInfo->Name ?: "UND", InstructionSize, Bytes);
  DecodeInst->InstSize = InstructionSize;
  return true;
@@ -688,7 +696,7 @@ bool Decoder::NormalOpHeader(FEXCore::X86Tables::X86InstInfo const *Info, uint16
    return false;
  }

-  LOGMAN_THROW_A_FMT(Info->Type != FEXCore::X86Tables::TYPE_REX_PREFIX,
+  LOGMAN_THROW_AA_FMT(Info->Type != FEXCore::X86Tables::TYPE_REX_PREFIX,
                     "REX PREFIX should have been decoded before this!");

  if (Info->Type >= FEXCore::X86Tables::TYPE_GROUP_1 &&
@@ -745,7 +753,7 @@ bool Decoder::NormalOpHeader(FEXCore::X86Tables::X86InstInfo const *Info, uint16
        3,
      };
      uint8_t Field = RegToField[ModRM.reg];
-      LOGMAN_THROW_A_FMT(Field != 255, "Invalid field selected!");
+      LOGMAN_THROW_AA_FMT(Field != 255, "Invalid field selected!");

      LocalOp = (Field << 3) | ModRM.rm;
      return NormalOp(&SecondModRMTableOps[LocalOp], LocalOp);
@@ -781,6 +789,7 @@ bool Decoder::NormalOpHeader(FEXCore::X86Tables::X86InstInfo const *Info, uint16
    if (Op == 0xC5) { // Two byte VEX
      pp = Byte1 & 0b11;
      options.vvvv = 15 - ((Byte1 & 0b01111000) >> 3);
+      options.L = (Byte1 & 0b100) != 0;
    }
    else { // 0xC4 = Three byte VEX
      const uint8_t Byte2 = ReadByte();
@@ -788,6 +797,7 @@ bool Decoder::NormalOpHeader(FEXCore::X86Tables::X86InstInfo const *Info, uint16
      map_select = Byte1 & 0b11111;
      options.vvvv = 15 - ((Byte2 & 0b01111000) >> 3);
      options.w = (Byte2 & 0b10000000) != 0;
+      options.L = (Byte2 & 0b100) != 0;
      if ((Byte1 & 0b01000000) == 0) {
        LOGMAN_THROW_A_FMT(CTX->Config.Is64BitMode, "VEX.X shouldn't be 0 in 32-bit mode!");
        DecodeInst->Flags |= DecodeFlags::FLAG_REX_XGPR_X;
@@ -1137,6 +1147,7 @@ const uint8_t *Decoder::AdjustAddrForSpecialRegion(uint8_t const* _InstStream, u
 }

 void Decoder::DecodeInstructionsAtEntry(uint8_t const* _InstStream, uint64_t PC, std::function<void(uint64_t BlockEntry, uint64_t Start, uint64_t Length)> AddContainedCodePage) {
+  FEXCORE_PROFILE_SCOPED("DecodeInstructions");
  Blocks.clear();
  BlocksToDecode.clear();
  HasBlocks.clear();
@@ -1171,7 +1182,6 @@ void Decoder::DecodeInstructionsAtEntry(uint8_t const* _InstStream, uint64_t PC,
  std::set<uint64_t> CodePages = { CurrentCodePage };

  AddContainedCodePage(PC, CurrentCodePage, FHU::FEX_PAGE_SIZE);
-  

  while (!BlocksToDecode.empty()) {
    auto BlockDecodeIt = BlocksToDecode.begin();
@@ -49,6 +49,7 @@ private:
  struct DecodedHeader {
    uint8_t vvvv; // Encoded operand in a VEX prefix.
    bool w;       // VEX.W bit.
+    bool L;       // VEX.L bit (if set then 256 bit operation, if unset then scalar or 128-bit operation)
  };

  FEXCore::Context::Context *CTX;
@@ -259,7 +259,7 @@ struct FEX_PACKED GDBContextDefinition {
  uint32_t fctrl;
  uint32_t fstat;
  uint32_t dummies[6];
-  uint64_t xmm[Core::CPUState::NUM_XMMS][2];
+  uint64_t xmm[Core::CPUState::NUM_XMMS][4];
  uint32_t mxcsr;
 };

@@ -306,7 +306,7 @@ std::string GdbServer::readRegs() {
  GDB.fstat |= static_cast<uint32_t>(state.flags[FEXCore::X86State::X87FLAG_C2_LOC]) << 10;
  GDB.fstat |= static_cast<uint32_t>(state.flags[FEXCore::X86State::X87FLAG_C3_LOC]) << 14;

-  memcpy(&GDB.xmm[0], &state.xmm[0], sizeof(GDB.xmm));
+  memcpy(&GDB.xmm[0], &state.xmm.avx.data[0], sizeof(GDB.xmm));

  return encodeHex((unsigned char *)&GDB, sizeof(GDBContextDefinition));
 }
@@ -382,9 +382,9 @@ GdbServer::HandledPacketType GdbServer::readReg(const std::string& packet) {
  }
  else if (addr >= offsetof(GDBContextDefinition, xmm[0][0]) &&
           addr < offsetof(GDBContextDefinition, xmm[16][0])) {
-    const auto XmmIndex = (addr - offsetof(GDBContextDefinition, xmm[0][0])) / Core::CPUState::XMM_REG_SIZE;
-    const auto *Data = (unsigned char *)&state.xmm[XmmIndex][0];
-    return {encodeHex(Data, Core::CPUState::XMM_REG_SIZE), HandledPacketType::TYPE_ACK};
+    const auto XmmIndex = (addr - offsetof(GDBContextDefinition, xmm[0][0])) / Core::CPUState::XMM_AVX_REG_SIZE;
+    const auto *Data = (unsigned char *)&state.xmm.avx.data[XmmIndex][0];
+    return {encodeHex(Data, Core::CPUState::XMM_AVX_REG_SIZE), HandledPacketType::TYPE_ACK};
  }
  else if (addr == offsetof(GDBContextDefinition, mxcsr)) {
    uint32_t Empty{};
@@ -490,6 +490,30 @@ std::string buildTargetXML() {
      reg("mxcsr", "int", 32);

    xml << "</feature>\n";
+
+    xml << "<feature name='org.gnu.gdb.i386.avx'>";
+    xml <<
+        R"(<vector id="v4f" type="ieee_single" count="4"/>
+        <vector id="v2d" type="ieee_double" count="2"/>
+        <vector id="v16i8" type="int8" count="16"/>
+        <vector id="v8i16" type="int16" count="8"/>
+        <vector id="v4i32" type="int32" count="4"/>
+        <vector id="v2i64" type="int64" count="2"/>
+        <union id="vec128">
+          <field name="v4_float" type="v4f"/>
+          <field name="v2_double" type="v2d"/>
+          <field name="v16_int8" type="v16i8"/>
+          <field name="v8_int16" type="v8i16"/>
+          <field name="v4_int32" type="v4i32"/>
+          <field name="v2_int64" type="v2i64"/>
+          <field name="uint128" type="uint128"/>
+        </union>
+        )";
+    for (size_t i = 0; i < Core::CPUState::NUM_XMMS; i++) {
+        reg(fmt::format("ymm{}h", i), "vec128", 128);
+    }
+    xml << "</feature>\n";
+
  xml << "</target>";
  xml << std::flush;

@@ -1,7 +1,7 @@
 #include "Interface/Core/CPUID.h"
-#include "Interface/Core/HostFeatures.h"
+#include <FEXCore/Core/HostFeatures.h>

-#ifdef _M_ARM_64
+#if defined(_M_ARM_64) || defined(VIXL_SIMULATOR)
 #include "aarch64/assembler-aarch64.h"
 #include "aarch64/cpu-aarch64.h"
 #include "aarch64/disasm-aarch64.h"
@@ -50,8 +50,12 @@ static uint32_t GetDCZID() {


 HostFeatures::HostFeatures() {
-#ifdef _M_ARM_64
+#if defined(_M_ARM_64) || defined(VIXL_SIMULATOR)
+#ifdef VIXL_SIMULATOR
+  auto Features = vixl::CPUFeatures::All();
+#else
  auto Features = vixl::CPUFeatures::InferFromOS();
+#endif
  SupportsAES = Features.Has(vixl::CPUFeatures::Feature::kAES);
  SupportsCRC = Features.Has(vixl::CPUFeatures::Feature::kCRC32);
  SupportsAtomics = Features.Has(vixl::CPUFeatures::Feature::kAtomics);
@@ -61,7 +65,26 @@ HostFeatures::HostFeatures() {
  SupportsFlushInputsToZero = Features.Has(vixl::CPUFeatures::Feature::kAFP);
  SupportsRCPC = Features.Has(vixl::CPUFeatures::Feature::kRCpc);
  SupportsTSOImm9 = Features.Has(vixl::CPUFeatures::Feature::kRCpcImm);
+  SupportsPMULL_128Bit = Features.Has(vixl::CPUFeatures::Feature::kPmull1Q);

+  Supports3DNow = true;
+  SupportsSSE4A = true;
+#ifdef VIXL_SIMULATOR
+  // Hardcode enable SVE with 256-bit wide registers.
+  SupportsAVX = true;
+#else
+  SupportsAVX = Features.Has(vixl::CPUFeatures::Feature::kSVE2) &&
+                vixl::aarch64::CPU::ReadSVEVectorLengthInBits() >= 256;
+#endif
+  SupportsSHA = true;
+  SupportsBMI1 = true;
+  SupportsBMI2 = true;
+
+  if (!SupportsAtomics) {
+    WARN_ONCE_FMT("Host CPU doesn't support atomics. Expect bad performance");
+  }
+
+#ifdef _M_ARM_64
  // We need to get the CPU's cache line size
  // We expect sane targets that have correct cacheline sizes across clusters
  uint64_t CTR;
@@ -71,31 +94,6 @@ HostFeatures::HostFeatures() {
  DCacheLineSize = 4 << ((CTR >> 16) & 0xF);
  ICacheLineSize = 4 << (CTR & 0xF);

-  if (!SupportsAtomics) {
-    WARN_ONCE_FMT("Host CPU doesn't support atomics. Expect bad performance");
-  }
-#endif
-#ifdef _M_X86_64
-  Xbyak::util::Cpu Features{};
-  SupportsAES = Features.has(Xbyak::util::Cpu::tAESNI);
-  SupportsCRC = Features.has(Xbyak::util::Cpu::tSSE42);
-  SupportsRAND = Features.has(Xbyak::util::Cpu::tRDRAND) && Features.has(Xbyak::util::Cpu::tRDSEED);
-  SupportsRCPC = true;
-  SupportsTSOImm9 = true;
-
-  // xbyak doesn't know how to check for CLZero
-  uint32_t eax, ebx, ecx, edx;
-  // First ensure we support a new enough extended CPUID function range
-  __cpuid(0x8000'0000, eax, ebx, ecx, edx);
-  if (eax >= 0x8000'0008U) {
-    // CLZero defined in 8000_00008_EBX[bit 0]
-    __cpuid(0x8000'0008, eax, ebx, ecx, edx);
-    SupportsCLZERO = ebx & 1;
-  }
-
-  SupportsFlushInputsToZero = true;
-  SupportsFloatExceptions = true;
-#else
  // Test if this CPU supports float exception trapping by attempting to enable
  // On unsupported these bits are architecturally defined as RAZ/WI
  constexpr uint32_t ExceptionEnableTraps =
@@ -116,6 +114,40 @@ HostFeatures::HostFeatures() {
  SetFPCR(OriginalFPCR);
 #endif

+#endif
+#if defined(_M_X86_64) && !defined(VIXL_SIMULATOR)
+  Xbyak::util::Cpu Features{};
+  SupportsAES = Features.has(Xbyak::util::Cpu::tAESNI);
+  SupportsCRC = Features.has(Xbyak::util::Cpu::tSSE42);
+  SupportsRAND = Features.has(Xbyak::util::Cpu::tRDRAND) && Features.has(Xbyak::util::Cpu::tRDSEED);
+  SupportsRCPC = true;
+  SupportsTSOImm9 = true;
+  Supports3DNow = Features.has(Xbyak::util::Cpu::t3DN) && Features.has(Xbyak::util::Cpu::tE3DN);
+  SupportsSSE4A = Features.has(Xbyak::util::Cpu::tSSE4a);
+  SupportsAVX = true;
+  SupportsSHA = Features.has(Xbyak::util::Cpu::tSHA);
+  SupportsBMI1 = Features.has(Xbyak::util::Cpu::tBMI1);
+  SupportsBMI2 = Features.has(Xbyak::util::Cpu::tBMI2);
+  SupportsPMULL_128Bit = Features.has(Xbyak::util::Cpu::tPCLMULQDQ);
+
+  // xbyak doesn't know how to check for CLZero
+  uint32_t eax, ebx, ecx, edx;
+  // First ensure we support a new enough extended CPUID function range
+  __cpuid(0x8000'0000, eax, ebx, ecx, edx);
+  if (eax >= 0x8000'0008U) {
+    // CLZero defined in 8000_00008_EBX[bit 0]
+    __cpuid(0x8000'0008, eax, ebx, ecx, edx);
+    SupportsCLZERO = ebx & 1;
+  }
+
+  SupportsFlushInputsToZero = true;
+  SupportsFloatExceptions = true;
+#endif
+
+#ifdef VIXL_SIMULATOR
+  // simulator doesn't support dc(ZVA)
+  SupportsCLZERO = false;
+#else
  // Check if we can support cacheline clears
  uint32_t DCZID = GetDCZID();
  if ((DCZID & DCZID_DZP_MASK) == 0) {
@@ -125,5 +157,6 @@ HostFeatures::HostFeatures() {
    // This means we can use the instruction
    SupportsCLZERO = DCZID_Bytes == CPUIDEmu::CACHELINE_SIZE;
  }
+#endif
 }
 }
@@ -20,7 +20,7 @@ DEF_OP(TruncElementPair) {

  switch (IROp->Size) {
    case 4: {
-      uint64_t *Src = GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+      uint64_t *Src = GetSrc<uint64_t*>(Data->SSAData, Op->Pair);
      uint64_t Result{};
      Result = Src[0] & ~0U;
      Result |= Src[1] << 32;
@@ -69,11 +69,11 @@ DEF_OP(CycleCounter) {

 DEF_OP(Add) {
  auto Op = IROp->C<IR::IROp_Add>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
-  auto Func = [](auto a, auto b) { return a + b; };
+  auto *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  auto *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
+  const auto Func = [](auto a, auto b) { return a + b; };

  switch (OpSize) {
    DO_OP(4, uint32_t, Func)
@@ -84,11 +84,11 @@ DEF_OP(Add) {

 DEF_OP(Sub) {
  auto Op = IROp->C<IR::IROp_Sub>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
-  auto Func = [](auto a, auto b) { return a - b; };
+  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
+  const auto Func = [](auto a, auto b) { return a - b; };

  switch (OpSize) {
    DO_OP(4, uint32_t, Func)
@@ -99,9 +99,9 @@ DEF_OP(Sub) {

 DEF_OP(Neg) {
  auto Op = IROp->C<IR::IROp_Neg>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src = *GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint64_t Src = *GetSrc<int64_t*>(Data->SSAData, Op->Src);
  switch (OpSize) {
    case 4:
      GD = -static_cast<int32_t>(Src);
@@ -115,10 +115,10 @@ DEF_OP(Neg) {

 DEF_OP(Mul) {
  auto Op = IROp->C<IR::IROp_Mul>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 4:
@@ -138,10 +138,10 @@ DEF_OP(Mul) {

 DEF_OP(UMul) {
  auto Op = IROp->C<IR::IROp_UMul>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 4:
@@ -161,9 +161,9 @@ DEF_OP(UMul) {

 DEF_OP(Div) {
  auto Op = IROp->C<IR::IROp_Div>();
-  uint8_t OpSize = IROp->Size;
-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint8_t OpSize = IROp->Size;
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 1:
@@ -179,7 +179,7 @@ DEF_OP(Div) {
      GD = static_cast<int64_t>(Src1) / static_cast<int64_t>(Src2);
      break;
    case 16: {
-      __int128_t Tmp = *GetSrc<__int128_t*>(Data->SSAData, Op->Header.Args[0]) / *GetSrc<__int128_t*>(Data->SSAData, Op->Header.Args[1]);
+      __int128_t Tmp = *GetSrc<__int128_t*>(Data->SSAData, Op->Src1) / *GetSrc<__int128_t*>(Data->SSAData, Op->Src2);
      memcpy(GDP, &Tmp, 16);
      break;
    }
@@ -189,10 +189,10 @@ DEF_OP(Div) {

 DEF_OP(UDiv) {
  auto Op = IROp->C<IR::IROp_UDiv>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 1:
@@ -208,7 +208,7 @@ DEF_OP(UDiv) {
      GD = static_cast<uint64_t>(Src1) / static_cast<uint64_t>(Src2);
      break;
    case 16: {
-      __uint128_t Tmp = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]) / *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+      __uint128_t Tmp = *GetSrc<__uint128_t*>(Data->SSAData, Op->Src1) / *GetSrc<__uint128_t*>(Data->SSAData, Op->Src2);
      memcpy(GDP, &Tmp, 16);
      break;
    }
@@ -218,10 +218,10 @@ DEF_OP(UDiv) {

 DEF_OP(Rem) {
  auto Op = IROp->C<IR::IROp_Rem>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 1:
@@ -237,7 +237,7 @@ DEF_OP(Rem) {
      GD = static_cast<int64_t>(Src1) % static_cast<int64_t>(Src2);
      break;
    case 16: {
-      __int128_t Tmp = *GetSrc<__int128_t*>(Data->SSAData, Op->Header.Args[0]) % *GetSrc<__int128_t*>(Data->SSAData, Op->Header.Args[1]);
+      __int128_t Tmp = *GetSrc<__int128_t*>(Data->SSAData, Op->Src1) % *GetSrc<__int128_t*>(Data->SSAData, Op->Src2);
      memcpy(GDP, &Tmp, 16);
      break;
    }
@@ -247,10 +247,10 @@ DEF_OP(Rem) {

 DEF_OP(URem) {
  auto Op = IROp->C<IR::IROp_URem>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 1:
@@ -266,7 +266,7 @@ DEF_OP(URem) {
      GD = static_cast<uint64_t>(Src1) % static_cast<uint64_t>(Src2);
      break;
    case 16: {
-      __uint128_t Tmp = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]) % *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+      __uint128_t Tmp = *GetSrc<__uint128_t*>(Data->SSAData, Op->Src1) % *GetSrc<__uint128_t*>(Data->SSAData, Op->Src2);
      memcpy(GDP, &Tmp, 16);
      break;
    }
@@ -276,10 +276,10 @@ DEF_OP(URem) {

 DEF_OP(MulH) {
  auto Op = IROp->C<IR::IROp_MulH>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);

  switch (OpSize) {
    case 4: {
@@ -298,10 +298,10 @@ DEF_OP(MulH) {

 DEF_OP(UMulH) {
  auto Op = IROp->C<IR::IROp_UMulH>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);
  switch (OpSize) {
    case 4:
      GD = static_cast<uint64_t>(Src1) * static_cast<uint64_t>(Src2);
@@ -324,11 +324,11 @@ DEF_OP(UMulH) {

 DEF_OP(Or) {
  auto Op = IROp->C<IR::IROp_Or>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
-  auto Func = [](auto a, auto b) { return a | b; };
+  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
+  const auto Func = [](auto a, auto b) { return a | b; };

  switch (OpSize) {
    DO_OP(1, uint8_t,  Func)
@@ -342,11 +342,11 @@ DEF_OP(Or) {

 DEF_OP(And) {
  auto Op = IROp->C<IR::IROp_And>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
-  auto Func = [](auto a, auto b) { return a & b; };
+  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
+  const auto Func = [](auto a, auto b) { return a & b; };

  switch (OpSize) {
    DO_OP(1, uint8_t,  Func)
@@ -361,8 +361,8 @@ DEF_OP(Andn) {
  auto Op = IROp->C<IR::IROp_Andn>();
  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
+  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
  constexpr auto Func = [](auto a, auto b) {
    using Type = decltype(a);
    return static_cast<Type>(a & static_cast<Type>(~b));
@@ -379,11 +379,11 @@ DEF_OP(Andn) {

 DEF_OP(Xor) {
  auto Op = IROp->C<IR::IROp_Xor>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
-  auto Func = [](auto a, auto b) { return a ^ b; };
+  void *Src1 = GetSrc<void*>(Data->SSAData, Op->Src1);
+  void *Src2 = GetSrc<void*>(Data->SSAData, Op->Src2);
+  const auto Func = [](auto a, auto b) { return a ^ b; };

  switch (OpSize) {
    DO_OP(1, uint8_t,  Func)
@@ -396,11 +396,11 @@ DEF_OP(Xor) {

 DEF_OP(Lshl) {
  auto Op = IROp->C<IR::IROp_Lshl>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  uint8_t Mask = OpSize * 8 - 1;
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);
+  const uint8_t Mask = OpSize * 8 - 1;
  switch (OpSize) {
    case 4:
      GD = static_cast<uint32_t>(Src1) << (Src2 & Mask);
@@ -414,11 +414,11 @@ DEF_OP(Lshl) {

 DEF_OP(Lshr) {
  auto Op = IROp->C<IR::IROp_Lshr>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  uint8_t Mask = OpSize * 8 - 1;
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);
+  const uint8_t Mask = OpSize * 8 - 1;
  switch (OpSize) {
    case 4:
      GD = static_cast<uint32_t>(Src1) >> (Src2 & Mask);
@@ -432,11 +432,11 @@ DEF_OP(Lshr) {

 DEF_OP(Ashr) {
  auto Op = IROp->C<IR::IROp_Ashr>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  uint8_t Mask = OpSize * 8 - 1;
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);
+  const uint8_t Mask = OpSize * 8 - 1;
  switch (OpSize) {
    case 4:
      GD = (uint32_t)(static_cast<int32_t>(Src1) >> (Src2 & Mask));
@@ -450,12 +450,12 @@ DEF_OP(Ashr) {

 DEF_OP(Ror) {
  auto Op = IROp->C<IR::IROp_Ror>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  auto Ror = [] (auto In, auto R) {
-    auto RotateMask = sizeof(In) * 8 - 1;
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src2);
+  const auto Ror = [] (auto In, auto R) {
+    const auto RotateMask = sizeof(In) * 8 - 1;
    R &= RotateMask;
    return (In >> R) | (In << (sizeof(In) * 8 - R));
  };
@@ -474,11 +474,11 @@ DEF_OP(Ror) {

 DEF_OP(Extr) {
  auto Op = IROp->C<IR::IROp_Extr>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  auto Extr = [] (auto Src1, auto Src2, uint8_t lsb) -> decltype(Src1) {
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Upper);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Lower);
+  const auto Extr = [] (auto Src1, auto Src2, uint8_t lsb) -> decltype(Src1) {
    __uint128_t Result{};
    Result = Src1;
    Result <<= sizeof(Src1) * 8;
@@ -500,7 +500,7 @@ DEF_OP(Extr) {
 }

 DEF_OP(PDep) {
-  const auto Op = IROp->C<IR::IROp_PExt>();
+  const auto Op = IROp->C<IR::IROp_PDep>();
  const auto OpSize = IROp->Size;

  if (OpSize != 4 && OpSize != 8) {
@@ -508,10 +508,10 @@ DEF_OP(PDep) {
    return;
  }

-  const uint64_t Input = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Args(0))
-                                     : *GetSrc<uint64_t*>(Data->SSAData, Op->Args(0));
-  uint64_t Mask = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Args(1))
-                              : *GetSrc<uint64_t*>(Data->SSAData, Op->Args(1));
+  const uint64_t Input = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Input)
+                                     : *GetSrc<uint64_t*>(Data->SSAData, Op->Input);
+  uint64_t Mask = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Mask)
+                              : *GetSrc<uint64_t*>(Data->SSAData, Op->Mask);

  uint64_t Result = 0;
  for (uint64_t Index = 0; Mask > 0; Index++) {
@@ -532,10 +532,10 @@ DEF_OP(PExt) {
    return;
  }

-  const uint64_t Input = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Args(0))
-                                     : *GetSrc<uint64_t*>(Data->SSAData, Op->Args(0));
-  uint64_t Mask = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Args(1))
-                              : *GetSrc<uint64_t*>(Data->SSAData, Op->Args(1));
+  const uint64_t Input = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Input)
+                                     : *GetSrc<uint64_t*>(Data->SSAData, Op->Input);
+  uint64_t Mask = OpSize == 4 ? *GetSrc<uint32_t*>(Data->SSAData, Op->Mask)
+                              : *GetSrc<uint64_t*>(Data->SSAData, Op->Mask);

  uint64_t Result = 0;
  for (uint64_t Offset = 0; Mask > 0; Offset++) {
@@ -549,39 +549,39 @@ DEF_OP(PExt) {

 DEF_OP(LDiv) {
  auto Op = IROp->C<IR::IROp_LDiv>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  // Each source is OpSize in size
  // So you can have up to a 128bit divide from x86-64
  switch (OpSize) {
    case 2: {
-      uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[1]);
-      int16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[2]);
-      int32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
-      int32_t Res = Source / Divisor;
+      const uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Lower);
+      const uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Upper);
+      const int16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Divisor);
+      const int32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
+      const int32_t Res = Source / Divisor;

      // We only store the lower bits of the result
      GD = static_cast<int16_t>(Res);
      break;
    }
    case 4: {
-      uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[1]);
-      int32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[2]);
-      int64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
-      int64_t Res = Source / Divisor;
+      const uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Lower);
+      const uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Upper);
+      const int32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Divisor);
+      const int64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
+      const int64_t Res = Source / Divisor;

      // We only store the lower bits of the result
      GD = static_cast<int32_t>(Res);
      break;
    }
    case 8: {
-      uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-      int64_t Divisor = *GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[2]);
-      __int128_t Source = (static_cast<__int128_t>(SrcHigh) << 64) | SrcLow;
-      __int128_t Res = Source / Divisor;
+      const uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Lower);
+      const uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Upper);
+      const int64_t Divisor = *GetSrc<int64_t*>(Data->SSAData, Op->Divisor);
+      const __int128_t Source = (static_cast<__int128_t>(SrcHigh) << 64) | SrcLow;
+      const __int128_t Res = Source / Divisor;

      // We only store the lower bits of the result
      memcpy(GDP, &Res, OpSize);
@@ -593,39 +593,39 @@ DEF_OP(LDiv) {

 DEF_OP(LUDiv) {
  auto Op = IROp->C<IR::IROp_LUDiv>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  // Each source is OpSize in size
  // So you can have up to a 128bit divide from x86-64
  switch (OpSize) {
    case 2: {
-      uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[2]);
-      uint32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
-      uint32_t Res = Source / Divisor;
+      const uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Lower);
+      const uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Upper);
+      const uint16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Divisor);
+      const uint32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
+      const uint32_t Res = Source / Divisor;

      // We only store the lower bits of the result
      GD = static_cast<uint16_t>(Res);
      break;
    }
    case 4: {
-      uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[2]);
-      uint64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
-      uint64_t Res = Source / Divisor;
+      const uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Lower);
+      const uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Upper);
+      const uint32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Divisor);
+      const uint64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
+      const uint64_t Res = Source / Divisor;

      // We only store the lower bits of the result
      GD = static_cast<uint32_t>(Res);
      break;
    }
    case 8: {
-      uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint64_t Divisor = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[2]);
-      __uint128_t Source = (static_cast<__uint128_t>(SrcHigh) << 64) | SrcLow;
-      __uint128_t Res = Source / Divisor;
+      const uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Lower);
+      const uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Upper);
+      const uint64_t Divisor = *GetSrc<uint64_t*>(Data->SSAData, Op->Divisor);
+      const __uint128_t Source = (static_cast<__uint128_t>(SrcHigh) << 64) | SrcLow;
+      const __uint128_t Res = Source / Divisor;

      // We only store the lower bits of the result
      memcpy(GDP, &Res, OpSize);
@@ -637,39 +637,39 @@ DEF_OP(LUDiv) {

 DEF_OP(LRem) {
  auto Op = IROp->C<IR::IROp_LRem>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  // Each source is OpSize in size
  // So you can have up to a 128bit Remainder from x86-64
  switch (OpSize) {
    case 2: {
-      uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[1]);
-      int16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[2]);
-      int32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
-      int32_t Res = Source % Divisor;
+      const uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Lower);
+      const uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Upper);
+      const int16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Divisor);
+      const int32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
+      const int32_t Res = Source % Divisor;

      // We only store the lower bits of the result
      GD = static_cast<int16_t>(Res);
      break;
    }
    case 4: {
-      uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[1]);
-      int32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[2]);
-      int64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
-      int64_t Res = Source % Divisor;
+      const uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Lower);
+      const uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Upper);
+      const int32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Divisor);
+      const int64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
+      const int64_t Res = Source % Divisor;

      // We only store the lower bits of the result
      GD = static_cast<int32_t>(Res);
      break;
    }
    case 8: {
-      uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-      int64_t Divisor = *GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[2]);
-      __int128_t Source = (static_cast<__int128_t>(SrcHigh) << 64) | SrcLow;
-      __int128_t Res = Source % Divisor;
+      const uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Lower);
+      const uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Upper);
+      const int64_t Divisor = *GetSrc<int64_t*>(Data->SSAData, Op->Divisor);
+      const __int128_t Source = (static_cast<__int128_t>(SrcHigh) << 64) | SrcLow;
+      const __int128_t Res = Source % Divisor;
      // We only store the lower bits of the result
      memcpy(GDP, &Res, OpSize);
      break;
@@ -680,39 +680,39 @@ DEF_OP(LRem) {

 DEF_OP(LURem) {
  auto Op = IROp->C<IR::IROp_LURem>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  // Each source is OpSize in size
  // So you can have up to a 128bit Remainder from x86-64
  switch (OpSize) {
    case 2: {
-      uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[2]);
-      uint32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
-      uint32_t Res = Source % Divisor;
+      const uint16_t SrcLow = *GetSrc<uint16_t*>(Data->SSAData, Op->Lower);
+      const uint16_t SrcHigh = *GetSrc<uint16_t*>(Data->SSAData, Op->Upper);
+      const uint16_t Divisor = *GetSrc<uint16_t*>(Data->SSAData, Op->Divisor);
+      const uint32_t Source = (static_cast<uint32_t>(SrcHigh) << 16) | SrcLow;
+      const uint32_t Res = Source % Divisor;

      // We only store the lower bits of the result
      GD = static_cast<uint16_t>(Res);
      break;
    }
    case 4: {
-      uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[2]);
-      uint64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
-      uint64_t Res = Source % Divisor;
+      const uint32_t SrcLow = *GetSrc<uint32_t*>(Data->SSAData, Op->Lower);
+      const uint32_t SrcHigh = *GetSrc<uint32_t*>(Data->SSAData, Op->Upper);
+      const uint32_t Divisor = *GetSrc<uint32_t*>(Data->SSAData, Op->Divisor);
+      const uint64_t Source = (static_cast<uint64_t>(SrcHigh) << 32) | SrcLow;
+      const uint64_t Res = Source % Divisor;

      // We only store the lower bits of the result
      GD = static_cast<uint32_t>(Res);
      break;
    }
    case 8: {
-      uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-      uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-      uint64_t Divisor = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[2]);
-      __uint128_t Source = (static_cast<__uint128_t>(SrcHigh) << 64) | SrcLow;
-      __uint128_t Res = Source % Divisor;
+      const uint64_t SrcLow = *GetSrc<uint64_t*>(Data->SSAData, Op->Lower);
+      const uint64_t SrcHigh = *GetSrc<uint64_t*>(Data->SSAData, Op->Upper);
+      const uint64_t Divisor = *GetSrc<uint64_t*>(Data->SSAData, Op->Divisor);
+      const __uint128_t Source = (static_cast<__uint128_t>(SrcHigh) << 64) | SrcLow;
+      const __uint128_t Res = Source % Divisor;
      // We only store the lower bits of the result
      memcpy(GDP, &Res, OpSize);
      break;
@@ -723,62 +723,62 @@ DEF_OP(LURem) {

 DEF_OP(Not) {
  auto Op = IROp->C<IR::IROp_Not>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
  const uint64_t mask[9]= { 0, 0xFF, 0xFFFF, 0, 0xFFFFFFFF, 0, 0, 0, 0xFFFFFFFFFFFFFFFFULL };
-  uint64_t Mask = mask[OpSize];
+  const uint64_t Mask = mask[OpSize];
  GD = (~Src) & Mask;
 }

 DEF_OP(Popcount) {
  auto Op = IROp->C<IR::IROp_Popcount>();
-  uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
  GD = std::popcount(Src);
 }

 DEF_OP(FindLSB) {
  auto Op = IROp->C<IR::IROp_FindLSB>();
-  uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Result = FindFirstSetBit(Src);
+  const uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
+  const uint64_t Result = FindFirstSetBit(Src);
  GD = Result - 1;
 }

 DEF_OP(FindMSB) {
  auto Op = IROp->C<IR::IROp_FindMSB>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  switch (OpSize) {
-    case 1: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]))) - 1; break;
-    case 2: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]))) - 1; break;
-    case 4: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]))) - 1; break;
-    case 8: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]))) - 1; break;
+    case 1: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint8_t*>(Data->SSAData, Op->Src))) - 1; break;
+    case 2: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint16_t*>(Data->SSAData, Op->Src))) - 1; break;
+    case 4: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint32_t*>(Data->SSAData, Op->Src))) - 1; break;
+    case 8: GD = (OpSize * 8 - std::countl_zero(*GetSrc<uint64_t*>(Data->SSAData, Op->Src))) - 1; break;
    default: LOGMAN_MSG_A_FMT("Unknown FindMSB size: {}", OpSize); break;
  }
 }

 DEF_OP(FindTrailingZeros) {
  auto Op = IROp->C<IR::IROp_FindTrailingZeros>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  switch (OpSize) {
    case 1: {
-      auto Src = *GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint8_t*>(Data->SSAData, Op->Src);
      GD = std::countr_zero(Src);
      break;
    }
    case 2: {
-      auto Src = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint16_t*>(Data->SSAData, Op->Src);
      GD = std::countr_zero(Src);
      break;
    }
    case 4: {
-      auto Src = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint32_t*>(Data->SSAData, Op->Src);
      GD = std::countr_zero(Src);
      break;
    }
    case 8: {
-      auto Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
      GD = std::countr_zero(Src);
      break;
    }
@@ -788,26 +788,26 @@ DEF_OP(FindTrailingZeros) {

 DEF_OP(CountLeadingZeroes) {
  auto Op = IROp->C<IR::IROp_CountLeadingZeroes>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  switch (OpSize) {
    case 1: {
-      auto Src = *GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint8_t*>(Data->SSAData, Op->Src);
      GD = std::countl_zero(Src);
      break;
    }
    case 2: {
-      auto Src = *GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint16_t*>(Data->SSAData, Op->Src);
      GD = std::countl_zero(Src);
      break;
    }
    case 4: {
-      auto Src = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint32_t*>(Data->SSAData, Op->Src);
      GD = std::countl_zero(Src);
      break;
    }
    case 8: {
-      auto Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+      const auto Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
      GD = std::countl_zero(Src);
      break;
    }
@@ -817,12 +817,12 @@ DEF_OP(CountLeadingZeroes) {

 DEF_OP(Rev) {
  auto Op = IROp->C<IR::IROp_Rev>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  switch (OpSize) {
-    case 2: GD = BSwap16(*GetSrc<uint16_t*>(Data->SSAData, Op->Header.Args[0])); break;
-    case 4: GD = BSwap32(*GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[0])); break;
-    case 8: GD = BSwap64(*GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0])); break;
+    case 2: GD = BSwap16(*GetSrc<uint16_t*>(Data->SSAData, Op->Src)); break;
+    case 4: GD = BSwap32(*GetSrc<uint32_t*>(Data->SSAData, Op->Src)); break;
+    case 8: GD = BSwap64(*GetSrc<uint64_t*>(Data->SSAData, Op->Src)); break;
    default: LOGMAN_MSG_A_FMT("Unknown REV size: {}", OpSize); break;
  }
 }
@@ -830,34 +830,36 @@ DEF_OP(Rev) {
 DEF_OP(Bfi) {
  auto Op = IROp->C<IR::IROp_Bfi>();
  uint64_t SourceMask = (1ULL << Op->Width) - 1;
-  if (Op->Width == 64)
+  if (Op->Width == 64) {
    SourceMask = ~0ULL;
-  uint64_t DestMask = ~(SourceMask << Op->lsb);
-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
-  uint64_t Res = (Src1 & DestMask) | ((Src2 & SourceMask) << Op->lsb);
+  }
+  const uint64_t DestMask = ~(SourceMask << Op->lsb);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Dest);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
+  const uint64_t Res = (Src1 & DestMask) | ((Src2 & SourceMask) << Op->lsb);
  GD = Res;
 }

 DEF_OP(Bfe) {
  auto Op = IROp->C<IR::IROp_Bfe>();

-  LOGMAN_THROW_A_FMT(IROp->Size <= 8, "OpSize is too large for BFE: {}", IROp->Size);
+  LOGMAN_THROW_AA_FMT(IROp->Size <= 8, "OpSize is too large for BFE: {}", IROp->Size);
  uint64_t SourceMask = (1ULL << Op->Width) - 1;
-  if (Op->Width == 64)
+  if (Op->Width == 64) {
    SourceMask = ~0ULL;
+  }
  SourceMask <<= Op->lsb;
-  uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Src);
  GD = (Src & SourceMask) >> Op->lsb;
 }

 DEF_OP(Sbfe) {
  auto Op = IROp->C<IR::IROp_Sbfe>();

-  LOGMAN_THROW_A_FMT(IROp->Size <= 8, "OpSize is too large for SBFE: {}", IROp->Size);
-  int64_t Src = *GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t ShiftLeftAmount = (64 - (Op->Width + Op->lsb));
-  uint64_t ShiftRightAmount = ShiftLeftAmount + Op->lsb;
+  LOGMAN_THROW_AA_FMT(IROp->Size <= 8, "OpSize is too large for SBFE: {}", IROp->Size);
+  int64_t Src = *GetSrc<int64_t*>(Data->SSAData, Op->Src);
+  const uint64_t ShiftLeftAmount = (64 - (Op->Width + Op->lsb));
+  const uint64_t ShiftRightAmount = ShiftLeftAmount + Op->lsb;
  Src <<= ShiftLeftAmount;
  Src >>= ShiftRightAmount;
  GD = Src;
@@ -865,20 +867,20 @@ DEF_OP(Sbfe) {

 DEF_OP(Select) {
  auto Op = IROp->C<IR::IROp_Select>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp2);

  uint64_t ArgTrue;
  uint64_t ArgFalse;

  if (OpSize == 4) {
-    ArgTrue = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[2]);
-    ArgFalse = *GetSrc<uint32_t*>(Data->SSAData, Op->Header.Args[3]);
+    ArgTrue = *GetSrc<uint32_t*>(Data->SSAData, Op->TrueVal);
+    ArgFalse = *GetSrc<uint32_t*>(Data->SSAData, Op->FalseVal);
  } else {
-    ArgTrue = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[2]);
-    ArgFalse = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[3]);
+    ArgTrue = *GetSrc<uint64_t*>(Data->SSAData, Op->TrueVal);
+    ArgFalse = *GetSrc<uint64_t*>(Data->SSAData, Op->FalseVal);
  }

  bool CompResult;
@@ -892,57 +894,75 @@ DEF_OP(Select) {
 }

 DEF_OP(VExtractToGPR) {
-  auto Op = IROp->C<IR::IROp_VExtractToGPR>();
+  const auto Op = IROp->C<IR::IROp_VExtractToGPR>();
+  const auto OpSize = IROp->Size;

-  uint32_t SourceSize = GetOpSize(Data->CurrentIR, Op->Header.Args[0]);
+  constexpr auto AVXRegSize = Core::CPUState::XMM_AVX_REG_SIZE;
+  constexpr auto SSERegSize = Core::CPUState::XMM_SSE_REG_SIZE;
+  constexpr auto SSEBitSize = SSERegSize * 8;

-  LOGMAN_THROW_A_FMT(IROp->Size <= 16, "OpSize is too large for VExtractToGPR: {}", IROp->Size);
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto ElementSizeBits = ElementSize * 8;
+  const auto Shift = ElementSizeBits * Op->Index;

-  if (SourceSize == 16) {
-    __uint128_t SourceMask = (1ULL << (Op->Header.ElementSize * 8)) - 1;
-    uint64_t Shift = Op->Header.ElementSize * Op->Index * 8;
-    if (Op->Header.ElementSize == 8)
+  const uint32_t SourceSize = GetOpSize(Data->CurrentIR, Op->Vector);
+
+  LOGMAN_THROW_AA_FMT(OpSize <= AVXRegSize,
+                      "OpSize is too large for VExtractToGPR: {}", OpSize);
+
+  if (SourceSize >= SSERegSize) {
+    __uint128_t SourceMask = (1ULL << ElementSizeBits) - 1;
+    if (ElementSize == 8) {
      SourceMask = ~0ULL;
+    }

-    __uint128_t Src = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-    Src >>= Shift;
-    Src &= SourceMask;
-    memcpy(GDP, &Src, Op->Header.ElementSize);
+    const auto Src = *GetSrc<InterpVector256*>(Data->SSAData, Op->Vector);
+
+    const auto GetResult = [&] {
+      if (Shift >= SSEBitSize) {
+        const auto NormalizedShift = Shift - SSEBitSize;
+        return (Src.Upper >> NormalizedShift) & SourceMask;
+      } else {
+        return (Src.Lower >> Shift) & SourceMask;
+      }
+    };
+
+    const auto Result = GetResult();
+    memcpy(GDP, &Result, ElementSize);
  }
  else {
-    uint64_t SourceMask = (1ULL << (Op->Header.ElementSize * 8)) - 1;
-    uint64_t Shift = Op->Header.ElementSize * Op->Index * 8;
-    if (Op->Header.ElementSize == 8)
+    uint64_t SourceMask = (1ULL << ElementSizeBits) - 1;
+    if (ElementSize == 8) {
      SourceMask = ~0ULL;
+    }

-    uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-    Src >>= Shift;
-    Src &= SourceMask;
-    GD = Src;
+    const uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Vector);
+    const uint64_t Result = (Src >> Shift) & SourceMask;
+    GD = Result;
  }
 }

 DEF_OP(Float_ToGPR_ZS) {
  auto Op = IROp->C<IR::IROp_Float_ToGPR_ZS>();
-  uint16_t Conv = (IROp->Size << 8) | Op->SrcElementSize;
+  const uint16_t Conv = (IROp->Size << 8) | Op->SrcElementSize;
  switch (Conv) {
    case 0x0804: { // int64_t <- float
-      int64_t Dst = (int64_t)std::trunc(*GetSrc<float*>(Data->SSAData, Op->Header.Args[0]));
+      const int64_t Dst = (int64_t)std::trunc(*GetSrc<float*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0808: { // int64_t <- double
-      int64_t Dst = (int64_t)std::trunc(*GetSrc<double*>(Data->SSAData, Op->Header.Args[0]));
+      const int64_t Dst = (int64_t)std::trunc(*GetSrc<double*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0404: { // int32_t <- float
-      int32_t Dst = (int32_t)std::trunc(*GetSrc<float*>(Data->SSAData, Op->Header.Args[0]));
+      const int32_t Dst = (int32_t)std::trunc(*GetSrc<float*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0408: { // int32_t <- double
-      int32_t Dst = (int32_t)std::trunc(*GetSrc<double*>(Data->SSAData, Op->Header.Args[0]));
+      const int32_t Dst = (int32_t)std::trunc(*GetSrc<double*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
@@ -951,25 +971,25 @@ DEF_OP(Float_ToGPR_ZS) {

 DEF_OP(Float_ToGPR_S) {
  auto Op = IROp->C<IR::IROp_Float_ToGPR_S>();
-  uint16_t Conv = (IROp->Size << 8) | Op->SrcElementSize;
+  const uint16_t Conv = (IROp->Size << 8) | Op->SrcElementSize;
  switch (Conv) {
    case 0x0804: { // int64_t <- float
-      int64_t Dst = (int64_t)std::nearbyint(*GetSrc<float*>(Data->SSAData, Op->Header.Args[0]));
+      const int64_t Dst = (int64_t)std::nearbyint(*GetSrc<float*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0808: { // int64_t <- double
-      int64_t Dst = (int64_t)std::nearbyint(*GetSrc<double*>(Data->SSAData, Op->Header.Args[0]));
+      const int64_t Dst = (int64_t)std::nearbyint(*GetSrc<double*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0404: { // int32_t <- float
-      int32_t Dst = (int32_t)std::nearbyint(*GetSrc<float*>(Data->SSAData, Op->Header.Args[0]));
+      const int32_t Dst = (int32_t)std::nearbyint(*GetSrc<float*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
    case 0x0408: { // int32_t <- double
-      int32_t Dst = (int32_t)std::nearbyint(*GetSrc<double*>(Data->SSAData, Op->Header.Args[0]));
+      const int32_t Dst = (int32_t)std::nearbyint(*GetSrc<double*>(Data->SSAData, Op->Scalar));
      memcpy(GDP, &Dst, IROp->Size);
      break;
    }
@@ -980,9 +1000,9 @@ DEF_OP(FCmp) {
  auto Op = IROp->C<IR::IROp_FCmp>();
  uint32_t ResultFlags{};
  if (Op->ElementSize == 4) {
-    float Src1 = *GetSrc<float*>(Data->SSAData, Op->Header.Args[0]);
-    float Src2 = *GetSrc<float*>(Data->SSAData, Op->Header.Args[1]);
-    bool Unordered = std::isnan(Src1) || std::isnan(Src2);
+    const float Src1 = *GetSrc<float*>(Data->SSAData, Op->Scalar1);
+    const float Src2 = *GetSrc<float*>(Data->SSAData, Op->Scalar2);
+    const bool Unordered = std::isnan(Src1) || std::isnan(Src2);
    if (Op->Flags & (1 << IR::FCMP_FLAG_LT)) {
      if (Unordered || (Src1 < Src2)) {
        ResultFlags |= (1 << IR::FCMP_FLAG_LT);
@@ -1000,9 +1020,9 @@ DEF_OP(FCmp) {
    }
  }
  else {
-    double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-    double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-    bool Unordered = std::isnan(Src1) || std::isnan(Src2);
+    const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Scalar1);
+    const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Scalar2);
+    const bool Unordered = std::isnan(Src1) || std::isnan(Src2);
    if (Op->Flags & (1 << IR::FCMP_FLAG_LT)) {
      if (Unordered || (Src1 < Src2)) {
        ResultFlags |= (1 << IR::FCMP_FLAG_LT);
@@ -42,7 +42,7 @@ DEF_OP(ExitFunction) {
  uintptr_t* ContextPtr = reinterpret_cast<uintptr_t*>(Data->State->CurrentFrame);

  void *ContextData = reinterpret_cast<void*>(ContextPtr);
-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
+  void *Src = GetSrc<void*>(Data->SSAData, Op->NewRIP);

  memcpy(ContextData, Src, OpSize);

@@ -51,22 +51,22 @@ DEF_OP(ExitFunction) {

 DEF_OP(Jump) {
  auto Op = IROp->C<IR::IROp_Jump>();
-  uintptr_t ListBegin = Data->CurrentIR->GetListData();
-  uintptr_t DataBegin = Data->CurrentIR->GetData();
+  const uintptr_t ListBegin = Data->CurrentIR->GetListData();
+  const uintptr_t DataBegin = Data->CurrentIR->GetData();

-  Data->BlockIterator = IR::NodeIterator(ListBegin, DataBegin, Op->Header.Args[0]);
+  Data->BlockIterator = IR::NodeIterator(ListBegin, DataBegin, Op->TargetBlock);
  Data->BlockResults.Redo = true;
 }

 DEF_OP(CondJump) {
  auto Op = IROp->C<IR::IROp_CondJump>();
-  uintptr_t ListBegin = Data->CurrentIR->GetListData();
-  uintptr_t DataBegin = Data->CurrentIR->GetData();
+  const uintptr_t ListBegin = Data->CurrentIR->GetListData();
+  const uintptr_t DataBegin = Data->CurrentIR->GetData();

  bool CompResult;

-  uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp1);
-  uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp2);
+  const uint64_t Src1 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp1);
+  const uint64_t Src2 = *GetSrc<uint64_t*>(Data->SSAData, Op->Cmp2);

  if (Op->CompareSize == 4)
    CompResult = IsConditionTrue<uint32_t, int32_t, float>(Op->Cond.Val, Src1, Src2);
@@ -127,7 +127,7 @@ DEF_OP(Thunk) {
  auto Op = IROp->C<IR::IROp_Thunk>();

  auto thunkFn = Data->State->CTX->ThunkHandler->LookupThunk(Op->ThunkNameHash);
-  thunkFn(*GetSrc<void**>(Data->SSAData, Op->Header.Args[0]));
+  thunkFn(*GetSrc<void**>(Data->SSAData, Op->ArgPtr));
 }

 DEF_OP(ValidateCode) {
@@ -141,15 +141,15 @@ DEF_OP(ValidateCode) {
  }
 }

-DEF_OP(RemoveThreadCodeEntry) {
-  Data->State->CTX->RemoveThreadCodeEntry(Data->State, Data->CurrentEntry);
+DEF_OP(ThreadRemoveCodeEntry) {
+  Data->State->CTX->ThreadRemoveCodeEntryFromJit(Data->State->CurrentFrame, Data->CurrentEntry);
 }

 DEF_OP(CPUID) {
  auto Op = IROp->C<IR::IROp_CPUID>();
  uint64_t *DstPtr = GetDest<uint64_t*>(Data->SSAData, Node);
-  uint64_t Arg = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-  uint64_t Leaf = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const uint64_t Arg = *GetSrc<uint64_t*>(Data->SSAData, Op->Function);
+  const uint64_t Leaf = *GetSrc<uint64_t*>(Data->SSAData, Op->Leaf);

  auto Results = Data->State->CTX->CPUID.RunFunction(Arg, Leaf);
  memcpy(DstPtr, &Results, sizeof(uint32_t) * 4);
@@ -13,53 +13,77 @@ $end_info$
 namespace FEXCore::CPU {
 #define DEF_OP(x) void InterpreterOps::Op_##x(IR::IROp_Header *IROp, IROpData *Data, IR::NodeID Node)
 DEF_OP(VInsGPR) {
-  auto Op = IROp->C<IR::IROp_VInsGPR>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_VInsGPR>();
+  const auto OpSize = IROp->Size;

-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-  __uint128_t Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto ElementSizeBits = ElementSize * 8;
+  constexpr auto SSEBitSize = Core::CPUState::XMM_SSE_REG_SIZE * 8;

-  uint64_t Offset = Op->DestIdx * Op->Header.ElementSize * 8;
-  __uint128_t Mask = (1ULL << (Op->Header.ElementSize * 8)) - 1;
-  if (Op->Header.ElementSize == 8) {
+  const uint64_t Offset = Op->DestIdx * ElementSizeBits;
+  const auto InUpperLane = Offset >= SSEBitSize;
+
+  __uint128_t Mask = (1ULL << ElementSizeBits) - 1;
+  if (ElementSize == 8) {
    Mask = ~0ULL;
  }
-  Src2 = Src2 & Mask;
-  Mask <<= Offset;
+
+  const auto Src1 = *GetSrc<InterpVector256*>(Data->SSAData, Op->DestVector);
+  const auto Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Src);
+
+  const auto Scalar = Src2 & Mask;
+  const auto ScaledOffset = InUpperLane ? Offset - SSEBitSize
+                                        : Offset;
+
+  // Now shift into place and set all bits but
+  // the ones where we're going to insert our value.
+  Mask <<= ScaledOffset;
  Mask = ~Mask;
-  __uint128_t Dst = Src1 & Mask;
-  Dst |= Src2 << Offset;
+
+  const auto Dst = [&] {
+    if (InUpperLane) {
+      return InterpVector256{
+        .Lower = Src1.Lower,
+        .Upper = (Src1.Upper & Mask) | (Scalar << ScaledOffset),
+      };
+    } else {
+      return InterpVector256{
+        .Lower = (Src1.Lower & Mask) | (Scalar << ScaledOffset),
+        .Upper = Src1.Upper,
+      };
+    }
+  }();

  memcpy(GDP, &Dst, OpSize);
 }

 DEF_OP(VCastFromGPR) {
  auto Op = IROp->C<IR::IROp_VCastFromGPR>();
-  memcpy(GDP, GetSrc<void*>(Data->SSAData, Op->Header.Args[0]), Op->Header.ElementSize);
+  memcpy(GDP, GetSrc<void*>(Data->SSAData, Op->Src), Op->Header.ElementSize);
 }

 DEF_OP(Float_FromGPR_S) {
  auto Op = IROp->C<IR::IROp_Float_FromGPR_S>();

-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
  switch (Conv) {
    case 0x0404: { // Float <- int32_t
-      float Dst = (float)*GetSrc<int32_t*>(Data->SSAData, Op->Header.Args[0]);
+      const float Dst = (float)*GetSrc<int32_t*>(Data->SSAData, Op->Src);
      memcpy(GDP, &Dst, Op->Header.ElementSize);
      break;
    }
    case 0x0408: { // Float <- int64_t
-      float Dst = (float)*GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[0]);
+      const float Dst = (float)*GetSrc<int64_t*>(Data->SSAData, Op->Src);
      memcpy(GDP, &Dst, Op->Header.ElementSize);
      break;
    }
    case 0x0804: { // Double <- int32_t
-      double Dst = (double)*GetSrc<int32_t*>(Data->SSAData, Op->Header.Args[0]);
+      const double Dst = (double)*GetSrc<int32_t*>(Data->SSAData, Op->Src);
      memcpy(GDP, &Dst, Op->Header.ElementSize);
      break;
    }
    case 0x0808: { // Double <- int64_t
-      double Dst = (double)*GetSrc<int64_t*>(Data->SSAData, Op->Header.Args[0]);
+      const double Dst = (double)*GetSrc<int64_t*>(Data->SSAData, Op->Src);
      memcpy(GDP, &Dst, Op->Header.ElementSize);
      break;
    }
@@ -68,15 +92,15 @@ DEF_OP(Float_FromGPR_S) {

 DEF_OP(Float_FToF) {
  auto Op = IROp->C<IR::IROp_Float_FToF>();
-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
  switch (Conv) {
    case 0x0804: { // Double <- Float
-      double Dst = (double)*GetSrc<float*>(Data->SSAData, Op->Header.Args[0]);
+      const double Dst = (double)*GetSrc<float*>(Data->SSAData, Op->Scalar);
      memcpy(GDP, &Dst, 8);
      break;
    }
    case 0x0408: { // Float <- Double
-      float Dst = (float)*GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
+      const float Dst = (float)*GetSrc<double*>(Data->SSAData, Op->Scalar);
      memcpy(GDP, &Dst, 4);
      break;
    }
@@ -86,68 +110,78 @@ DEF_OP(Float_FToF) {

 DEF_OP(Vector_SToF) {
  auto Op = IROp->C<IR::IROp_Vector_SToF>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  uint8_t Tmp[16]{};
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Vector);
+  uint8_t Tmp[Core::CPUState::XMM_AVX_REG_SIZE]{};

-  uint8_t Elements = OpSize / Op->Header.ElementSize;
+  const uint8_t ElementSize = Op->Header.ElementSize;
+  const uint8_t Elements = OpSize / ElementSize;

-  auto Func = [](auto a, auto min, auto max) { return a; };
-  switch (Op->Header.ElementSize) {
+  const auto Func = [](auto a, auto min, auto max) { return a; };
+  switch (ElementSize) {
    DO_VECTOR_1SRC_2TYPE_OP(4, float, int32_t, Func, 0, 0)
    DO_VECTOR_1SRC_2TYPE_OP(8, double, int64_t, Func, 0, 0)
-    default: LOGMAN_MSG_A_FMT("Unknown Element Size: {}", Op->Header.ElementSize); break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+      break;
  }
  memcpy(GDP, Tmp, OpSize);
 }

 DEF_OP(Vector_FToZS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToZS>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_Vector_FToZS>();
+  const uint8_t OpSize = IROp->Size;

-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  uint8_t Tmp[16]{};
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Vector);
+  uint8_t Tmp[Core::CPUState::XMM_AVX_REG_SIZE]{};

-  uint8_t Elements = OpSize / Op->Header.ElementSize;
+  const uint8_t ElementSize = Op->Header.ElementSize;
+  const uint8_t Elements = OpSize / ElementSize;

-  auto Func = [](auto a, auto min, auto max) { return std::trunc(a); };
-  switch (Op->Header.ElementSize) {
+  const auto Func = [](auto a, auto min, auto max) { return std::trunc(a); };
+  switch (ElementSize) {
    DO_VECTOR_1SRC_2TYPE_OP(4, int32_t, float, Func, 0, 0)
    DO_VECTOR_1SRC_2TYPE_OP(8, int64_t, double, Func, 0, 0)
-    default: LOGMAN_MSG_A_FMT("Unknown Element Size: {}", Op->Header.ElementSize); break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+      break;
  }
  memcpy(GDP, Tmp, OpSize);
 }

 DEF_OP(Vector_FToS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToS>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_Vector_FToS>();
+  const uint8_t OpSize = IROp->Size;

-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  uint8_t Tmp[16]{};
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Vector);
+  uint8_t Tmp[Core::CPUState::XMM_AVX_REG_SIZE]{};

-  uint8_t Elements = OpSize / Op->Header.ElementSize;
+  const uint8_t ElementSize = Op->Header.ElementSize;
+  const uint8_t Elements = OpSize / ElementSize;

-  auto Func = [](auto a, auto min, auto max) { return std::nearbyint(a); };
-  switch (Op->Header.ElementSize) {
+  const auto Func = [](auto a, auto min, auto max) { return std::nearbyint(a); };
+  switch (ElementSize) {
    DO_VECTOR_1SRC_2TYPE_OP(4, int32_t, float, Func, 0, 0)
    DO_VECTOR_1SRC_2TYPE_OP(8, int64_t, double, Func, 0, 0)
-    default: LOGMAN_MSG_A_FMT("Unknown Element Size: {}", Op->Header.ElementSize); break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+      break;
  }
  memcpy(GDP, Tmp, OpSize);
 }

 DEF_OP(Vector_FToF) {
-  auto Op = IROp->C<IR::IROp_Vector_FToF>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_Vector_FToF>();
+  const uint8_t OpSize = IROp->Size;

-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  uint8_t Tmp[16]{};
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Vector);
+  uint8_t Tmp[Core::CPUState::XMM_AVX_REG_SIZE]{};

-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const uint16_t ElementSize = Op->Header.ElementSize;
+  const uint16_t Conv = (ElementSize << 8) | Op->SrcElementSize;

-  auto Func = [](auto a, auto min, auto max) { return a; };
+  const auto Func = [](auto a, auto min, auto max) { return a; };
  switch (Conv) {
    case 0x0804: { // Double <- float
      // Only the lower elements from the source
@@ -165,52 +199,55 @@ DEF_OP(Vector_FToF) {
      DO_VECTOR_1SRC_2TYPE_OP_NOSIZE(float, double, Func, 0, 0)
      break;
    }
-    default: LOGMAN_MSG_A_FMT("Unknown Conversion Type : 0x{:04x}", Conv); break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Conversion Type : 0x{:04x}", Conv);
+      break;
  }
  memcpy(GDP, Tmp, OpSize);
 }

 DEF_OP(Vector_FToI) {
-  auto Op = IROp->C<IR::IROp_Vector_FToI>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_Vector_FToI>();
+  const uint8_t OpSize = IROp->Size;

-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  uint8_t Tmp[16]{};
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Vector);
+  uint8_t Tmp[Core::CPUState::XMM_AVX_REG_SIZE]{};

-  uint8_t Elements = OpSize / Op->Header.ElementSize;
-  auto Func_Nearest = [](auto a) { return std::rint(a); };
-  auto Func_Neg = [](auto a) { return std::floor(a); };
-  auto Func_Pos = [](auto a) { return std::ceil(a); };
-  auto Func_Trunc = [](auto a) { return std::trunc(a); };
-  auto Func_Host = [](auto a) { return std::rint(a); };
+  const uint8_t ElementSize = Op->Header.ElementSize;
+  const uint8_t Elements = OpSize / ElementSize;
+  const auto Func_Nearest = [](auto a) { return std::rint(a); };
+  const auto Func_Neg = [](auto a) { return std::floor(a); };
+  const auto Func_Pos = [](auto a) { return std::ceil(a); };
+  const auto Func_Trunc = [](auto a) { return std::trunc(a); };
+  const auto Func_Host = [](auto a) { return std::rint(a); };

  switch (Op->Round) {
    case FEXCore::IR::Round_Nearest.Val:
-      switch (Op->Header.ElementSize) {
+      switch (ElementSize) {
        DO_VECTOR_1SRC_OP(4, float, Func_Nearest)
        DO_VECTOR_1SRC_OP(8, double, Func_Nearest)
      }
    break;
    case FEXCore::IR::Round_Negative_Infinity.Val:
-      switch (Op->Header.ElementSize) {
+      switch (ElementSize) {
        DO_VECTOR_1SRC_OP(4, float, Func_Neg)
        DO_VECTOR_1SRC_OP(8, double, Func_Neg)
      }
    break;
    case FEXCore::IR::Round_Positive_Infinity.Val:
-      switch (Op->Header.ElementSize) {
+      switch (ElementSize) {
        DO_VECTOR_1SRC_OP(4, float, Func_Pos)
        DO_VECTOR_1SRC_OP(8, double, Func_Pos)
      }
    break;
    case FEXCore::IR::Round_Towards_Zero.Val:
-      switch (Op->Header.ElementSize) {
+      switch (ElementSize) {
        DO_VECTOR_1SRC_OP(4, float, Func_Trunc)
        DO_VECTOR_1SRC_OP(8, double, Func_Trunc)
      }
    break;
    case FEXCore::IR::Round_Host.Val:
-      switch (Op->Header.ElementSize) {
+      switch (ElementSize) {
        DO_VECTOR_1SRC_OP(4, float, Func_Host)
        DO_VECTOR_1SRC_OP(8, double, Func_Host)
      }
@@ -360,7 +360,7 @@ namespace FEXCore::CPU {

 DEF_OP(AESImc) {
  auto Op = IROp->C<IR::IROp_VAESImc>();
-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
+  auto Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Vector);

  // Pseudo-code
  // Dst = InvMixColumns(STATE)
@@ -371,8 +371,8 @@ DEF_OP(AESImc) {

 DEF_OP(AESEnc) {
  auto Op = IROp->C<IR::IROp_VAESEnc>();
-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-  __uint128_t Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+  auto Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->State);
+  auto Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Key);

  // Pseudo-code
  // STATE = Src1
@@ -391,8 +391,8 @@ DEF_OP(AESEnc) {

 DEF_OP(AESEncLast) {
  auto Op = IROp->C<IR::IROp_VAESEncLast>();
-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-  __uint128_t Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+  auto Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->State);
+  auto Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Key);

  // Pseudo-code
  // STATE = Src1
@@ -409,8 +409,8 @@ DEF_OP(AESEncLast) {

 DEF_OP(AESDec) {
  auto Op = IROp->C<IR::IROp_VAESDec>();
-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-  __uint128_t Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+  auto Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->State);
+  auto Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Key);

  // Pseudo-code
  // STATE = Src1
@@ -429,8 +429,8 @@ DEF_OP(AESDec) {

 DEF_OP(AESDecLast) {
  auto Op = IROp->C<IR::IROp_VAESDecLast>();
-  __uint128_t Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-  __uint128_t Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[1]);
+  auto Src1 = *GetSrc<__uint128_t*>(Data->SSAData, Op->State);
+  auto Src2 = *GetSrc<__uint128_t*>(Data->SSAData, Op->Key);

  // Pseudo-code
  // STATE = Src1
@@ -447,7 +447,7 @@ DEF_OP(AESDecLast) {

 DEF_OP(AESKeyGenAssist) {
  auto Op = IROp->C<IR::IROp_VAESKeyGenAssist>();
-  uint8_t *Src1 = GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint8_t *Src1 = GetSrc<uint8_t*>(Data->SSAData, Op->Src);

  // Pseudo-code
  // X3 = Src1[127:96]
@@ -20,99 +20,90 @@ DEF_OP(F80LOADFCW) {

 DEF_OP(F80ADD) {
  auto Op = IROp->C<IR::IROp_F80Add>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FADD(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FADD(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80SUB) {
  auto Op = IROp->C<IR::IROp_F80Sub>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FSUB(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FSUB(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80MUL) {
  auto Op = IROp->C<IR::IROp_F80Mul>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FMUL(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FMUL(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80DIV) {
  auto Op = IROp->C<IR::IROp_F80Div>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FDIV(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FDIV(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80FYL2X) {
  auto Op = IROp->C<IR::IROp_F80FYL2X>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FYL2X(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FYL2X(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80ATAN) {
  auto Op = IROp->C<IR::IROp_F80ATAN>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FATAN(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FATAN(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80FPREM1) {
  auto Op = IROp->C<IR::IROp_F80FPREM1>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FREM1(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FREM1(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80FPREM) {
  auto Op = IROp->C<IR::IROp_F80FPREM>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FREM(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FREM(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80SCALE) {
  auto Op = IROp->C<IR::IROp_F80SCALE>();
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FSCALE(Src1, Src2);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
+  const auto Tmp = X80SoftFloat::FSCALE(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80CVT) {
  auto Op = IROp->C<IR::IROp_F80CVT>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);

  switch (OpSize) {
    case 4: {
@@ -131,9 +122,9 @@ DEF_OP(F80CVT) {

 DEF_OP(F80CVTINT) {
  auto Op = IROp->C<IR::IROp_F80CVTInt>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);

  switch (OpSize) {
    case 2: {
@@ -160,13 +151,13 @@ DEF_OP(F80CVTTO) {

  switch (Op->SrcSize) {
    case 4: {
-      float Src = *GetSrc<float *>(Data->SSAData, Op->Header.Args[0]);
+      float Src = *GetSrc<float *>(Data->SSAData, Op->X80Src);
      X80SoftFloat Tmp = Src;
      memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
      break;
    }
    case 8: {
-      double Src = *GetSrc<double *>(Data->SSAData, Op->Header.Args[0]);
+      double Src = *GetSrc<double *>(Data->SSAData, Op->X80Src);
      X80SoftFloat Tmp = Src;
      memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
      break;
@@ -180,13 +171,13 @@ DEF_OP(F80CVTTOINT) {

  switch (Op->SrcSize) {
    case 2: {
-      int16_t Src = *GetSrc<int16_t*>(Data->SSAData, Op->Header.Args[0]);
+      int16_t Src = *GetSrc<int16_t*>(Data->SSAData, Op->Src);
      X80SoftFloat Tmp = Src;
      memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
      break;
    }
    case 4: {
-      int32_t Src = *GetSrc<int32_t*>(Data->SSAData, Op->Header.Args[0]);
+      int32_t Src = *GetSrc<int32_t*>(Data->SSAData, Op->Src);
      X80SoftFloat Tmp = Src;
      memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
      break;
@@ -197,77 +188,73 @@ DEF_OP(F80CVTTOINT) {

 DEF_OP(F80ROUND) {
  auto Op = IROp->C<IR::IROp_F80Round>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FRNDINT(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FRNDINT(Src);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80F2XM1) {
  auto Op = IROp->C<IR::IROp_F80F2XM1>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::F2XM1(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::F2XM1(Src);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80TAN) {
  auto Op = IROp->C<IR::IROp_F80TAN>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FTAN(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FTAN(Src);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80SQRT) {
  auto Op = IROp->C<IR::IROp_F80SQRT>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FSQRT(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FSQRT(Src);

  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80SIN) {
  auto Op = IROp->C<IR::IROp_F80SIN>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FSIN(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FSIN(Src);
+
  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80COS) {
  auto Op = IROp->C<IR::IROp_F80COS>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FCOS(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FCOS(Src);
+
  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80XTRACT_EXP) {
  auto Op = IROp->C<IR::IROp_F80XTRACT_EXP>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FXTRACT_EXP(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FXTRACT_EXP(Src);
+
  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80XTRACT_SIG) {
  auto Op = IROp->C<IR::IROp_F80XTRACT_SIG>();
-  X80SoftFloat Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Tmp;
-  Tmp = X80SoftFloat::FXTRACT_SIG(Src);
+  const auto Src = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src);
+  const auto Tmp = X80SoftFloat::FXTRACT_SIG(Src);
+
  memcpy(GDP, &Tmp, sizeof(X80SoftFloat));
 }

 DEF_OP(F80CMP) {
  auto Op = IROp->C<IR::IROp_F80Cmp>();
  uint32_t ResultFlags{};
-  X80SoftFloat Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]);
-  X80SoftFloat Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[1]);
+  const auto Src1 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src1);
+  const auto Src2 = *GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src2);
  bool eq, lt, nan;
  X80SoftFloat::FCMP(Src1, Src2, &eq, &lt, &nan);
  if (Op->Flags & (1 << IR::FCMP_FLAG_LT) &&
@@ -288,7 +275,7 @@ DEF_OP(F80CMP) {

 DEF_OP(F80BCDLOAD) {
  auto Op = IROp->C<IR::IROp_F80BCDLoad>();
-  uint8_t *Src1 = GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+  const uint8_t *Src1 = GetSrc<uint8_t*>(Data->SSAData, Op->X80Src);
  uint64_t BCD{};
  // We walk through each uint8_t and pull out the BCD encoding
  // Each 4bit split is a digit
@@ -323,7 +310,7 @@ DEF_OP(F80BCDLOAD) {

 DEF_OP(F80BCDSTORE) {
  auto Op = IROp->C<IR::IROp_F80BCDStore>();
-  X80SoftFloat Src1 = X80SoftFloat::FRNDINT(*GetSrc<X80SoftFloat*>(Data->SSAData, Op->Header.Args[0]));
+  X80SoftFloat Src1 = X80SoftFloat::FRNDINT(*GetSrc<X80SoftFloat*>(Data->SSAData, Op->X80Src));
  bool Negative = Src1.Sign;

  // Clear the Sign bit
@@ -358,74 +345,74 @@ DEF_OP(F80BCDSTORE) {

 DEF_OP(F64SIN) {
  auto Op = IROp->C<IR::IROp_F64SIN>();
-  double Src = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Tmp = sin(Src);
+  const double Src = *GetSrc<double*>(Data->SSAData, Op->Src);
+  const double Tmp = sin(Src);
  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64COS) {
  auto Op = IROp->C<IR::IROp_F64COS>();
-  double Src = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Tmp = cos(Src);
+  const double Src = *GetSrc<double*>(Data->SSAData, Op->Src);
+  const double Tmp = cos(Src);
  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64TAN) {
  auto Op = IROp->C<IR::IROp_F64TAN>();
-  double Src = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Tmp = tan(Src);
+  const double Src = *GetSrc<double*>(Data->SSAData, Op->Src);
+  const double Tmp = tan(Src);
  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64F2XM1) {
  auto Op = IROp->C<IR::IROp_F64F2XM1>();
-  double Src = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Tmp = exp2(Src) - 1.0;
+  const double Src = *GetSrc<double*>(Data->SSAData, Op->Src);
+  const double Tmp = exp2(Src) - 1.0;
  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64ATAN) {
  auto Op = IROp->C<IR::IROp_F64ATAN>();
-  double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-  double Tmp = atan2(Src1, Src2);
+  const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Src1);
+  const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Src2);
+  const double Tmp = atan2(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64FPREM) {
  auto Op = IROp->C<IR::IROp_F64FPREM>();
-  double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-  double Tmp = fmod(Src1, Src2);
+  const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Src1);
+  const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Src2);
+  const double Tmp = fmod(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64FPREM1) {
  auto Op = IROp->C<IR::IROp_F64FPREM1>();
-  double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-  double Tmp = remainder(Src1, Src2);
+  const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Src1);
+  const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Src2);
+  const double Tmp = remainder(Src1, Src2);

  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64FYL2X) {
  auto Op = IROp->C<IR::IROp_F64FYL2X>();
-  double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-  double Tmp = Src2 * log2(Src1);
+  const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Src);
+  const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Src2);
+  const double Tmp = Src2 * log2(Src1);

  memcpy(GDP, &Tmp, sizeof(double));
 }

 DEF_OP(F64SCALE) {
  auto Op = IROp->C<IR::IROp_F64SCALE>();
-  double Src1 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[0]);
-  double Src2 = *GetSrc<double*>(Data->SSAData, Op->Header.Args[1]);
-  double trunc = (double)(int64_t)(Src2); //truncate
-  double Tmp = Src1 * exp2(trunc);
+  const double Src1 = *GetSrc<double*>(Data->SSAData, Op->Src1);
+  const double Src2 = *GetSrc<double*>(Data->SSAData, Op->Src2);
+  const double trunc = (double)(int64_t)(Src2); //truncate
+  const double Tmp = Src1 * exp2(trunc);

  memcpy(GDP, &Tmp, sizeof(double));
 }
@@ -14,7 +14,7 @@ namespace FEXCore::CPU {
 #define DEF_OP(x) void InterpreterOps::Op_##x(IR::IROp_Header *IROp, IROpData *Data, IR::NodeID Node)
 DEF_OP(GetHostFlag) {
  auto Op = IROp->C<IR::IROp_GetHostFlag>();
-  GD = (*GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]) >> Op->Flag) & 1;
+  GD = (*GetSrc<uint64_t*>(Data->SSAData, Op->Value) >> Op->Flag) & 1;
 }
 #undef DEF_OP

@@ -159,21 +159,26 @@
  break;                                            \
  }

+struct InterpVector256 {
+  __uint128_t Lower;
+  __uint128_t Upper;
+};
+
 template<typename Res>
 Res GetDest(void* SSAData, FEXCore::IR::OrderedNodeWrapper Op) {
-  auto DstPtr = &reinterpret_cast<__uint128_t*>(SSAData)[Op.ID().Value];
+  auto DstPtr = &reinterpret_cast<InterpVector256*>(SSAData)[Op.ID().Value];
  return reinterpret_cast<Res>(DstPtr);
 }

 template<typename Res>
 Res GetDest(void* SSAData, FEXCore::IR::NodeID Op) {
-  auto DstPtr = &reinterpret_cast<__uint128_t*>(SSAData)[Op.Value];
+  auto DstPtr = &reinterpret_cast<InterpVector256*>(SSAData)[Op.Value];
  return reinterpret_cast<Res>(DstPtr);
 }


 template<typename Res>
 Res GetSrc(void* SSAData, FEXCore::IR::OrderedNodeWrapper Src) {
-  auto DstPtr = &reinterpret_cast<__uint128_t*>(SSAData)[Src.ID().Value];
+  auto DstPtr = &reinterpret_cast<InterpVector256*>(SSAData)[Src.ID().Value];
  return reinterpret_cast<Res>(DstPtr);
 }
@@ -146,7 +146,7 @@ void InterpreterOps::FillFallbackIndexPointers(uint64_t *Info) {

 }

-bool InterpreterOps::GetFallbackHandler(IR::IROp_Header *IROp, FallbackInfo *Info) {
+bool InterpreterOps::GetFallbackHandler(IR::IROp_Header const *IROp, FallbackInfo *Info) {
  uint8_t OpSize = IROp->Size;
  switch(IROp->Op) {
    case IR::OP_F80LOADFCW: {
@@ -1,5 +1,6 @@
 #include "Interface/Context/Context.h"
 #include "Interface/Core/CPUID.h"
+#include "InterpreterDefines.h"
 #include "InterpreterOps.h"
 #include "F80Ops.h"

@@ -121,7 +122,7 @@ constexpr OpHandlerArray InterpreterOpHandlers = [] {
  REGISTER_OP(INLINESYSCALL,          InlineSyscall);
  REGISTER_OP(THUNK,                  Thunk);
  REGISTER_OP(VALIDATECODE,           ValidateCode);
-  REGISTER_OP(REMOVETHREADCODEENTRY,        RemoveThreadCodeEntry);
+  REGISTER_OP(THREADREMOVECODEENTRY,        ThreadRemoveCodeEntry);
  REGISTER_OP(CPUID,                  CPUID);

  // Conversion ops
@@ -153,8 +154,6 @@ constexpr OpHandlerArray InterpreterOpHandlers = [] {
  REGISTER_OP(STOREMEM,               StoreMem);
  REGISTER_OP(LOADMEMTSO,             LoadMem);
  REGISTER_OP(STOREMEMTSO,            StoreMem);
-  REGISTER_OP(VLOADMEMELEMENT,        VLoadMemElement);
-  REGISTER_OP(VSTOREMEMELEMENT,       VStoreMemElement);
  REGISTER_OP(CACHELINECLEAR,         CacheLineClear);
  REGISTER_OP(CACHELINEZERO,          CacheLineZero);

@@ -180,13 +179,10 @@ constexpr OpHandlerArray InterpreterOpHandlers = [] {
  // Move ops
  REGISTER_OP(EXTRACTELEMENTPAIR,     ExtractElementPair);
  REGISTER_OP(CREATEELEMENTPAIR,      CreateElementPair);
-  REGISTER_OP(MOV,                    Mov);

  // Vector ops
  REGISTER_OP(VECTORZERO,             VectorZero);
  REGISTER_OP(VECTORIMM,              VectorImm);
-  REGISTER_OP(SPLATVECTOR2,           SplatVector);
-  REGISTER_OP(SPLATVECTOR4,           SplatVector);
  REGISTER_OP(VMOV,                   VMov);
  REGISTER_OP(VAND,                   VAnd);
  REGISTER_OP(VBIC,                   VBic);
@@ -245,18 +241,13 @@ constexpr OpHandlerArray InterpreterOpHandlers = [] {
  REGISTER_OP(VUSHRS,                 VUShrS);
  REGISTER_OP(VSSHRS,                 VSShrS);
  REGISTER_OP(VINSELEMENT,            VInsElement);
-  REGISTER_OP(VINSSCALARELEMENT,      VInsScalarElement);
-  REGISTER_OP(VEXTRACTELEMENT,        VExtractElement);
  REGISTER_OP(VDUPELEMENT,            VDupElement);
  REGISTER_OP(VEXTR,                  VExtr);
-  REGISTER_OP(VSLI,                   VSLI);
-  REGISTER_OP(VSRI,                   VSRI);
  REGISTER_OP(VUSHRI,                 VUShrI);
  REGISTER_OP(VSSHRI,                 VSShrI);
  REGISTER_OP(VSHLI,                  VShlI);
  REGISTER_OP(VUSHRNI,                VUShrNI);
  REGISTER_OP(VUSHRNI2,               VUShrNI2);
-  REGISTER_OP(VBITCAST,               VBitcast);
  REGISTER_OP(VSXTL,                  VSXTL);
  REGISTER_OP(VSXTL2,                 VSXTL2);
  REGISTER_OP(VUXTL,                  VUXTL);
@@ -336,29 +327,34 @@ void InterpreterOps::Op_NoOp(FEXCore::IR::IROp_Header *IROp, IROpData *Data, IR:
 void InterpreterOps::InterpretIR(FEXCore::Core::CpuStateFrame *Frame, FEXCore::IR::IRListView const *CurrentIR) {
  volatile void *StackEntry = alloca(0);

-  uintptr_t ListSize = CurrentIR->GetSSACount();
+  const uintptr_t ListSize = CurrentIR->GetSSACount();

  static_assert(sizeof(FEXCore::IR::IROp_Header) == 4);
  static_assert(sizeof(FEXCore::IR::OrderedNode) == 16);

  auto BlockEnd = CurrentIR->GetBlocks().end();

-  InterpreterOps::IROpData OpData{};
-  OpData.State = Frame->Thread;
-  OpData.SSAData = alloca(ListSize * 16);
-  OpData.CurrentEntry = Frame->State.rip;
-  OpData.CurrentIR = CurrentIR;
-  OpData.StackEntry = StackEntry;
-  OpData.BlockIterator = CurrentIR->GetBlocks().begin();
+  constexpr size_t ListEntrySizeInBytes = sizeof(InterpVector256);
+  const size_t SSADataSize = ListSize * ListEntrySizeInBytes;

-  // Clear them all to zero. Required for Zero-extend semantics
-  memset(OpData.SSAData, 0, ListSize * 16);
+  InterpreterOps::IROpData OpData{
+    .State = Frame->Thread,
+    .CurrentEntry = Frame->State.rip,
+    .CurrentIR = CurrentIR,
+    .StackEntry = StackEntry,
+    .SSAData = alloca(SSADataSize),
+    .BlockResults = {},
+    .BlockIterator = CurrentIR->GetBlocks().begin(),
+  };
+
+  // Clear all SSAData entries to zero. Required for Zero-extend semantics
+  memset(OpData.SSAData, 0, SSADataSize);

  while (1) {
    using namespace FEXCore::IR;
    auto [BlockNode, BlockHeader] = OpData.BlockIterator();
    auto BlockIROp = BlockHeader->CW<IROp_CodeBlock>();
-    LOGMAN_THROW_A_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");
+    LOGMAN_THROW_AA_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");

    // Reset the block results per block
    memset(&OpData.BlockResults, 0, sizeof(OpData.BlockResults));
@@ -49,7 +49,7 @@ namespace FEXCore::CPU {
    public:
      static void InterpretIR(FEXCore::Core::CpuStateFrame *Frame, FEXCore::IR::IRListView const *IR);
      static void FillFallbackIndexPointers(uint64_t *Info);
-      static bool GetFallbackHandler(IR::IROp_Header *IROp, FallbackInfo *Info);
+      static bool GetFallbackHandler(IR::IROp_Header const *IROp, FallbackInfo *Info);

      struct IROpData {
        FEXCore::Core::InternalThreadState *State{};
@@ -151,7 +151,7 @@ namespace FEXCore::CPU {
  DEF_OP(InlineSyscall);
  DEF_OP(Thunk);
  DEF_OP(ValidateCode);
-  DEF_OP(RemoveThreadCodeEntry);
+  DEF_OP(ThreadRemoveCodeEntry);
  DEF_OP(CPUID);

  ///< Conversion ops
@@ -181,8 +181,6 @@ namespace FEXCore::CPU {
  DEF_OP(StoreFlag);
  DEF_OP(LoadMem);
  DEF_OP(StoreMem);
-  DEF_OP(VLoadMemElement);
-  DEF_OP(VStoreMemElement);
  DEF_OP(CacheLineClear);
  DEF_OP(CacheLineZero);

@@ -207,7 +205,6 @@ namespace FEXCore::CPU {
  ///< Vector ops
  DEF_OP(VectorZero);
  DEF_OP(VectorImm);
-  DEF_OP(SplatVector);
  DEF_OP(VMov);
  DEF_OP(VAnd);
  DEF_OP(VBic);
@@ -264,18 +261,13 @@ namespace FEXCore::CPU {
  DEF_OP(VUShrS);
  DEF_OP(VSShrS);
  DEF_OP(VInsElement);
-  DEF_OP(VInsScalarElement);
-  DEF_OP(VExtractElement);
  DEF_OP(VDupElement);
  DEF_OP(VExtr);
-  DEF_OP(VSLI);
-  DEF_OP(VSRI);
  DEF_OP(VUShrI);
  DEF_OP(VSShrI);
  DEF_OP(VShlI);
  DEF_OP(VUShrNI);
  DEF_OP(VUShrNI2);
-  DEF_OP(VBitcast);
  DEF_OP(VSXTL);
  DEF_OP(VSXTL2);
  DEF_OP(VUXTL);
@@ -25,93 +25,139 @@ static inline void CacheLineFlush(char *Addr) {

 #define DEF_OP(x) void InterpreterOps::Op_##x(IR::IROp_Header *IROp, IROpData *Data, IR::NodeID Node)
 DEF_OP(LoadContext) {
-  auto Op = IROp->C<IR::IROp_LoadContext>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_LoadContext>();
+  const auto OpSize = IROp->Size;
+
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Src = ContextPtr + Op->Offset;

-  uintptr_t ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
-  ContextPtr += Op->Offset;
  #define LOAD_CTX(x, y) \
    case x: { \
-      y const *MemData = reinterpret_cast<y const*>(ContextPtr); \
+      y const *MemData = reinterpret_cast<y const*>(Src); \
      GD = *MemData; \
      break; \
    }
+
  switch (OpSize) {
    LOAD_CTX(1, uint8_t)
    LOAD_CTX(2, uint16_t)
    LOAD_CTX(4, uint32_t)
    LOAD_CTX(8, uint64_t)
-    case 16: {
-      void const *MemData = reinterpret_cast<void const*>(ContextPtr);
+    case 16:
+    case 32: {
+      void const *MemData = reinterpret_cast<void const*>(Src);
      memcpy(GDP, MemData, OpSize);
      break;
    }
-    default:  LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+      break;
  }
  #undef LOAD_CTX
 }

 DEF_OP(StoreContext) {
-  auto Op = IROp->C<IR::IROp_StoreContext>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_StoreContext>();
+  const auto OpSize = IROp->Size;

-  uintptr_t ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
-  ContextPtr += Op->Offset;
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Dst = ContextPtr + Op->Offset;

-  void *MemData = reinterpret_cast<void*>(ContextPtr);
-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
+  void *MemData = reinterpret_cast<void*>(Dst);
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Value);
  memcpy(MemData, Src, OpSize);
 }

 DEF_OP(LoadRegister) {
-  LOGMAN_MSG_A_FMT("Unimplemented");
-}
+  const auto Op = IROp->C<IR::IROp_LoadRegister>();
+  const auto OpSize = IROp->Size;

-DEF_OP(StoreRegister) {
-  LOGMAN_MSG_A_FMT("Unimplemented");
-}
-
-DEF_OP(LoadContextIndexed) {
-  auto Op = IROp->C<IR::IROp_LoadContextIndexed>();
-  uint64_t Index = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
-
-  uintptr_t ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
-
-  ContextPtr += Op->BaseOffset;
-  ContextPtr += Index * Op->Stride;
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Src = ContextPtr + Op->Offset;

  #define LOAD_CTX(x, y) \
    case x: { \
-      y const *MemData = reinterpret_cast<y const*>(ContextPtr); \
+      y const *MemData = reinterpret_cast<y const*>(Src); \
      GD = *MemData; \
      break; \
    }
-  switch (IROp->Size) {
+
+  switch (OpSize) {
    LOAD_CTX(1, uint8_t)
    LOAD_CTX(2, uint16_t)
    LOAD_CTX(4, uint32_t)
    LOAD_CTX(8, uint64_t)
-    case 16: {
-      void const *MemData = reinterpret_cast<void const*>(ContextPtr);
-      memcpy(GDP, MemData, IROp->Size);
+    case 16:
+    case 32: {
+      void const *MemData = reinterpret_cast<void const*>(Src);
+      memcpy(GDP, MemData, OpSize);
      break;
    }
-    default:  LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", IROp->Size);
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+      break;
+  }
+  #undef LOAD_CTX
+}
+
+DEF_OP(StoreRegister) {
+  const auto Op = IROp->C<IR::IROp_StoreRegister>();
+  const auto OpSize = IROp->Size;
+
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Dst = ContextPtr + Op->Offset;
+
+  void *MemData = reinterpret_cast<void*>(Dst);
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Value);
+  memcpy(MemData, Src, OpSize);
+}
+
+DEF_OP(LoadContextIndexed) {
+  const auto Op = IROp->C<IR::IROp_LoadContextIndexed>();
+  const auto OpSize = IROp->Size;
+
+  const auto Index = *GetSrc<uint64_t*>(Data->SSAData, Op->Index);
+
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Src = ContextPtr + Op->BaseOffset + (Index * Op->Stride);
+
+  #define LOAD_CTX(x, y) \
+    case x: { \
+      y const *MemData = reinterpret_cast<y const*>(Src); \
+      GD = *MemData; \
+      break; \
+    }
+
+  switch (OpSize) {
+    LOAD_CTX(1, uint8_t)
+    LOAD_CTX(2, uint16_t)
+    LOAD_CTX(4, uint32_t)
+    LOAD_CTX(8, uint64_t)
+    case 16:
+    case 32: {
+      void const *MemData = reinterpret_cast<void const*>(Src);
+      memcpy(GDP, MemData, OpSize);
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", OpSize);
+      break;
  }
  #undef LOAD_CTX
 }

 DEF_OP(StoreContextIndexed) {
-  auto Op = IROp->C<IR::IROp_StoreContextIndexed>();
-  uint64_t Index = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[1]);
+  const auto Op = IROp->C<IR::IROp_StoreContextIndexed>();
+  const auto OpSize = IROp->Size;

-  uintptr_t ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
-  ContextPtr += Op->BaseOffset;
-  ContextPtr += Index * Op->Stride;
+  const auto Index = *GetSrc<uint64_t*>(Data->SSAData, Op->Index);

-  void *MemData = reinterpret_cast<void*>(ContextPtr);
-  void *Src = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  memcpy(MemData, Src, IROp->Size);
+  const auto ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
+  const auto Dst = ContextPtr + Op->BaseOffset + (Index * Op->Stride);
+
+  void *MemData = reinterpret_cast<void*>(Dst);
+  void *Src = GetSrc<void*>(Data->SSAData, Op->Value);
+  memcpy(MemData, Src, OpSize);
 }

 DEF_OP(SpillRegister) {
@@ -134,7 +180,7 @@ DEF_OP(LoadFlag) {

 DEF_OP(StoreFlag) {
  auto Op = IROp->C<IR::IROp_StoreFlag>();
-  uint8_t Arg = *GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+  uint8_t Arg = *GetSrc<uint8_t*>(Data->SSAData, Op->Value);

  uintptr_t ContextPtr = reinterpret_cast<uintptr_t>(Data->State->CurrentFrame);
  ContextPtr += offsetof(FEXCore::Core::CPUState, flags[0]);
@@ -144,8 +190,8 @@ DEF_OP(StoreFlag) {
 }

 DEF_OP(LoadMem) {
-  auto Op = IROp->C<IR::IROp_LoadMem>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_LoadMem>();
+  const auto OpSize = IROp->Size;

  uint8_t const *MemData = *GetSrc<uint8_t const**>(Data->SSAData, Op->Addr);

@@ -158,7 +204,8 @@ DEF_OP(LoadMem) {
      case IR::MEM_OFFSET_SXTW.Val: MemData += (int32_t)Offset; break;
    }
  }
-  memset(GDP, 0, 16);
+
+  memset(GDP, 0, Core::CPUState::XMM_AVX_REG_SIZE);
  switch (OpSize) {
    case 1: {
      auto D = reinterpret_cast<const std::atomic<uint8_t>*>(MemData);
@@ -180,16 +227,15 @@ DEF_OP(LoadMem) {
      GD = D->load();
      break;
    }
-
    default:
-      memcpy(GDP, MemData, IROp->Size);
+      memcpy(GDP, MemData, OpSize);
      break;
  }
 }

 DEF_OP(StoreMem) {
-  auto Op = IROp->C<IR::IROp_StoreMem>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_StoreMem>();
+  const auto OpSize = IROp->Size;

  uint8_t *MemData = *GetSrc<uint8_t **>(Data->SSAData, Op->Addr);

@@ -221,41 +267,11 @@ DEF_OP(StoreMem) {
    }

    default:
-      memcpy(MemData, GetSrc<void*>(Data->SSAData, Op->Value), IROp->Size);
+      memcpy(MemData, GetSrc<void*>(Data->SSAData, Op->Value), OpSize);
      break;
  }
 }

-DEF_OP(VLoadMemElement) {
-  auto Op = IROp->C<IR::IROp_VLoadMemElement>();
-  void const *MemData = *GetSrc<void const**>(Data->SSAData, Op->Header.Args[0]);
-
-  memcpy(GDP, GetSrc<void*>(Data->SSAData, Op->Header.Args[1]), 16);
-  memcpy(reinterpret_cast<void*>(reinterpret_cast<uintptr_t>(GDP) + (Op->Header.ElementSize * Op->Index)),
-    MemData, Op->Header.ElementSize);
-}
-
-DEF_OP(VStoreMemElement) {
-  #define STORE_DATA(x, y) \
-    case x: { \
-      y *MemData = *GetSrc<y**>(Data->SSAData, Op->Header.Args[0]); \
-      memcpy(MemData, &GetSrc<y*>(Data->SSAData, Op->Header.Args[1])[Op->Index], sizeof(y)); \
-      break; \
-    }
-
-  auto Op = IROp->C<IR::IROp_VStoreMemElement>();
-  uint8_t OpSize = IROp->Size;
-
-  switch (OpSize) {
-    STORE_DATA(1, uint8_t)
-    STORE_DATA(2, uint16_t)
-    STORE_DATA(4, uint32_t)
-    STORE_DATA(8, uint64_t)
-    default: LOGMAN_MSG_A_FMT("Unhandled StoreMem size"); break;
-  }
-  #undef STORE_DATA
-}
-
 DEF_OP(CacheLineClear) {
  auto Op = IROp->C<IR::IROp_CacheLineClear>();

@@ -19,13 +19,6 @@ $end_info$
 #include <sys/random.h>

 namespace FEXCore::CPU {
-[[noreturn]]
-static void StopThread(FEXCore::Core::InternalThreadState *Thread) {
-  Thread->CTX->StopThread(Thread);
-
-  LOGMAN_MSG_A_FMT("unreachable");
-  FEX_UNREACHABLE;
-}

 #define DEF_OP(x) void InterpreterOps::Op_##x(IR::IROp_Header *IROp, IROpData *Data, IR::NodeID Node)
 DEF_OP(Fence) {
@@ -46,14 +39,26 @@ DEF_OP(Fence) {

 DEF_OP(Break) {
  auto Op = IROp->C<IR::IROp_Break>();
-  switch (Op->Reason) {
-    case FEXCore::IR::Break_Halt: // HLT
-      StopThread(Data->State);
+
+  Data->State->CurrentFrame->SynchronousFaultData.FaultToTopAndGeneratedException = 1;
+  Data->State->CurrentFrame->SynchronousFaultData.Signal = Op->Reason.Signal;
+  Data->State->CurrentFrame->SynchronousFaultData.TrapNo = Op->Reason.TrapNumber;
+  Data->State->CurrentFrame->SynchronousFaultData.err_code = Op->Reason.ErrorRegister;
+  Data->State->CurrentFrame->SynchronousFaultData.si_code = Op->Reason.si_code;
+
+  switch (Op->Reason.Signal) {
+  case SIGILL:
+    FHU::Syscalls::tgkill(Data->State->ThreadManager.PID, Data->State->ThreadManager.TID, SIGILL);
    break;
-    case FEXCore::IR::Break_InvalidInstruction:
-      FHU::Syscalls::tgkill(Data->State->ThreadManager.PID, Data->State->ThreadManager.TID, SIGILL);
+  case SIGTRAP:
+    FHU::Syscalls::tgkill(Data->State->ThreadManager.PID, Data->State->ThreadManager.TID, SIGTRAP);
+    break;
+  case SIGSEGV:
+    FHU::Syscalls::tgkill(Data->State->ThreadManager.PID, Data->State->ThreadManager.TID, SIGSEGV);
+    break;
+  default:
+    FHU::Syscalls::tgkill(Data->State->ThreadManager.PID, Data->State->ThreadManager.TID, SIGTRAP);
    break;
-  default: LOGMAN_MSG_A_FMT("Unknown Break Reason: {}", Op->Reason); break;
  }
 }

@@ -88,7 +93,7 @@ DEF_OP(GetRoundingMode) {

 DEF_OP(SetRoundingMode) {
  auto Op = IROp->C<IR::IROp_SetRoundingMode>();
-  uint8_t GuestRounding = *GetSrc<uint8_t*>(Data->SSAData, Op->Header.Args[0]);
+  const auto GuestRounding = *GetSrc<uint8_t*>(Data->SSAData, Op->RoundMode);
 #ifdef _M_ARM_64
  uint64_t HostRounding{};
  __asm volatile(R"(
@@ -128,16 +133,16 @@ DEF_OP(SetRoundingMode) {

 DEF_OP(Print) {
  auto Op = IROp->C<IR::IROp_Print>();
-  uint8_t OpSize = IROp->Size;
+  const uint8_t OpSize = IROp->Size;

  if (OpSize <= 8) {
-    uint64_t Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Header.Args[0]);
+    const auto Src = *GetSrc<uint64_t*>(Data->SSAData, Op->Value);
    LogMan::Msg::IFmt(">>>> Value in Arg: 0x{:x}, {}", Src, Src);
  }
  else if (OpSize == 16) {
-    __uint128_t Src = *GetSrc<__uint128_t*>(Data->SSAData, Op->Header.Args[0]);
-    uint64_t Src0 = Src;
-    uint64_t Src1 = Src >> 64;
+    const auto Src = *GetSrc<__uint128_t*>(Data->SSAData, Op->Value);
+    const uint64_t Src0 = Src;
+    const uint64_t Src1 = Src >> 64;
    LogMan::Msg::IFmt(">>>> Value[0] in Arg: 0x{:x}, {}", Src0, Src0);
    LogMan::Msg::IFmt("     Value[1] in Arg: 0x{:x}, {}", Src1, Src1);
  }
@@ -14,15 +14,15 @@ namespace FEXCore::CPU {
 #define DEF_OP(x) void InterpreterOps::Op_##x(IR::IROp_Header *IROp, IROpData *Data, IR::NodeID Node)
 DEF_OP(ExtractElementPair) {
  auto Op = IROp->C<IR::IROp_ExtractElementPair>();
-  uintptr_t Src = GetSrc<uintptr_t>(Data->SSAData, Op->Header.Args[0]);
+  const auto Src = GetSrc<uintptr_t>(Data->SSAData, Op->Pair);
  memcpy(GDP,
    reinterpret_cast<void*>(Src + Op->Header.Size * Op->Element), Op->Header.Size);
 }

 DEF_OP(CreateElementPair) {
  auto Op = IROp->C<IR::IROp_CreateElementPair>();
-  void *Src_Lower = GetSrc<void*>(Data->SSAData, Op->Header.Args[0]);
-  void *Src_Upper = GetSrc<void*>(Data->SSAData, Op->Header.Args[1]);
+  const void *Src_Lower = GetSrc<void*>(Data->SSAData, Op->Lower);
+  const void *Src_Upper = GetSrc<void*>(Data->SSAData, Op->Upper);

  uint8_t *Dst = GetDest<uint8_t*>(Data->SSAData, Node);

@@ -30,13 +30,6 @@ DEF_OP(CreateElementPair) {
  memcpy(Dst + IROp->ElementSize, Src_Upper, IROp->ElementSize);
 }

-DEF_OP(Mov) {
-  auto Op = IROp->C<IR::IROp_Mov>();
-  uint8_t OpSize = IROp->Size;
-
-  memcpy(GDP, GetSrc<void*>(Data->SSAData, Op->Header.Args[0]), OpSize);
-}
-
 #undef DEF_OP

 } // namespace FEXCore::CPU
@@ -14,7 +14,7 @@ namespace FEXCore::CPU {

 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)
 DEF_OP(TruncElementPair) {
  auto Op = IROp->C<IR::IROp_TruncElementPair>();

@@ -1084,8 +1084,8 @@ DEF_OP(Bfi) {

 DEF_OP(Bfe) {
  auto Op = IROp->C<IR::IROp_Bfe>();
-  LOGMAN_THROW_A_FMT(IROp->Size <= 8, "OpSize is too large for BFE: {}", IROp->Size);
-  LOGMAN_THROW_A_FMT(Op->Width != 0, "Invalid BFE width of 0");
+  LOGMAN_THROW_AA_FMT(IROp->Size <= 8, "OpSize is too large for BFE: {}", IROp->Size);
+  LOGMAN_THROW_AA_FMT(Op->Width != 0, "Invalid BFE width of 0");

  auto Dst = GetReg<RA_64>(Node);
  ubfx(Dst, GetReg<RA_64>(Op->Src.ID()), Op->lsb, Op->Width);
@@ -1168,25 +1168,79 @@ DEF_OP(Select) {
 }

 DEF_OP(VExtractToGPR) {
-  auto Op = IROp->C<IR::IROp_VExtractToGPR>();
-  const uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_VExtractToGPR>();
+  const auto OpSize = IROp->Size;

-  switch (OpSize) {
-    case 1:
-      umov(GetReg<RA_32>(Node), GetSrc(Op->Vector.ID()).V16B(), Op->Index);
-    break;
-    case 2:
-      umov(GetReg<RA_32>(Node), GetSrc(Op->Vector.ID()).V8H(), Op->Index);
-    break;
-    case 4:
-      umov(GetReg<RA_32>(Node), GetSrc(Op->Vector.ID()).V4S(), Op->Index);
-    break;
-    case 8:
-      umov(GetReg<RA_64>(Node), GetSrc(Op->Vector.ID()).V2D(), Op->Index);
-    break;
-    default:
-      LOGMAN_MSG_A_FMT("Unhandled ExtractElementSize: {}", OpSize);
-    break;
+  constexpr auto AVXRegBitSize = Core::CPUState::XMM_AVX_REG_SIZE * 8;
+  constexpr auto SSERegBitSize = Core::CPUState::XMM_SSE_REG_SIZE * 8;
+  const auto ElementSizeBits = Op->Header.ElementSize * 8;
+
+  const auto Offset = ElementSizeBits * Op->Index;
+  const auto Is256Bit = Offset >= SSERegBitSize;
+
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  const auto PerformMove = [&](const aarch64::VRegister& reg, int index) {
+    switch (OpSize) {
+      case 1:
+        umov(GetReg<RA_32>(Node), reg.V16B(), index);
+        break;
+      case 2:
+        umov(GetReg<RA_32>(Node), reg.V8H(), index);
+        break;
+      case 4:
+        umov(GetReg<RA_32>(Node), reg.V4S(), index);
+        break;
+      case 8:
+        umov(GetReg<RA_64>(Node), reg.V2D(), index);
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled ExtractElementSize: {}", OpSize);
+        break;
+    }
+  };
+
+  if (Offset < SSERegBitSize) {
+    // Desired data lies within the lower 128-bit lane, so we
+    // can treat the operation as a 128-bit operation, even
+    // when acting on larger register sizes.
+    PerformMove(Vector, Op->Index);
+  } else {
+    LOGMAN_THROW_AA_FMT(HostSupportsSVE,
+                        "Host doesn't support SVE. Cannot perform 256-bit operation.");
+    LOGMAN_THROW_AA_FMT(Is256Bit,
+                        "Can't perform 256-bit extraction with op side: {}", OpSize);
+    LOGMAN_THROW_AA_FMT(Offset < AVXRegBitSize,
+                        "Trying to extract element outside bounds of register. Offset={}, Index={}",
+                        Offset, Op->Index);
+
+    // We need to use the upper 128-bit lane, so lets move it down.
+    // Inverting our dedicated predicate for 128-bit operations selects
+    // all of the top lanes. We can then compact those into a temporary.
+    const auto CompactPred = p0;
+    not_(CompactPred.VnB(), PRED_TMP_32B.Zeroing(), PRED_TMP_16B.VnB());
+    compact(VTMP1.Z().VnD(), CompactPred, Vector.Z().VnD());
+
+    // Sanitize the zero-based index to work on the now-moved
+    // upper half of the vector.
+    const auto SanitizedIndex = [OpSize, Op] {
+      switch (OpSize) {
+        case 1:
+          return Op->Index - 16;
+        case 2:
+          return Op->Index - 8;
+        case 4:
+          return Op->Index - 4;
+        case 8:
+          return Op->Index - 2;
+        default:
+          LOGMAN_MSG_A_FMT("Unhandled OpSize: {}", OpSize);
+          return 0;
+      }
+    }();
+
+    // Move the value from the now-low-lane data.
+    PerformMove(VTMP1, SanitizedIndex);
  }
 }

@@ -1249,7 +1303,7 @@ DEF_OP(FCmp) {
  bool set = false;

  if (Op->Flags & (1 << IR::FCMP_FLAG_EQ)) {
-    LOGMAN_THROW_A_FMT(IR::FCMP_FLAG_EQ == 0, "IR::FCMP_FLAG_EQ must equal 0");
+    LOGMAN_THROW_AA_FMT(IR::FCMP_FLAG_EQ == 0, "IR::FCMP_FLAG_EQ must equal 0");
    // EQ or unordered
    cset(Dst, Condition::eq); // Z = 1
    csinc(Dst, Dst, xzr, Condition::vc); // IF !V ? Z : 1
@@ -81,7 +81,7 @@ bool Arm64JITCore::ApplyRelocations(uint64_t GuestEntry, uint64_t CodeEntry, uin
  size_t DataIndex{};
  for (size_t j = 0; j < NumRelocations; ++j) {
    const FEXCore::CPU::Relocation *Reloc = reinterpret_cast<const FEXCore::CPU::Relocation *>(&EntryRelocations[DataIndex]);
-    LOGMAN_THROW_A_FMT((DataIndex % alignof(Relocation)) == 0, "Alignment of relocation wasn't adhered to");
+    LOGMAN_THROW_AA_FMT((DataIndex % alignof(Relocation)) == 0, "Alignment of relocation wasn't adhered to");

    switch (Reloc->Header.Type) {
      case FEXCore::CPU::RelocationTypes::RELOC_NAMED_SYMBOL_LITERAL: {
@@ -10,7 +10,7 @@ $end_info$
 namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)
 DEF_OP(CASPair) {
  auto Op = IROp->C<IR::IROp_CASPair>();
  // Size is the size of each pair element
@@ -19,7 +19,7 @@ $end_info$
 namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)

 DEF_OP(SignalReturn) {
  // First we must reset the stack
@@ -40,7 +40,7 @@ DEF_OP(CallbackReturn) {
  ResetStack();

  // We can now lower the ref counter again
-  
+
  ldr(w2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, SignalHandlerRefCounter)));
  sub(w2, w2, 1);
  str(w2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, SignalHandlerRefCounter)));
@@ -197,7 +197,11 @@ DEF_OP(Syscall) {
  ldr(x3, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.SyscallHandlerFunc)));
  mov(x1, STATE);
  mov(x2, sp);
+#ifdef VIXL_SIMULATOR
+  GenerateIndirectRuntimeCall<uint64_t, void*, void*, void*>(x3);
+#else
  blr(x3);
+#endif

  add(sp, sp, SPOffset);

@@ -239,7 +243,6 @@ DEF_OP(InlineSyscall) {
  bool Intersects{};
  // We always need to spill x8 since we can't know if it is live at this SSA location
  uint32_t SpillMask = 1U << 8;
-  std::vector<vixl::aarch64::Register> IntersectRegs(FEXCore::HLE::SyscallArguments::MAX_ARGS);
  for (uint32_t i = 0; i < FEXCore::HLE::SyscallArguments::MAX_ARGS-1; ++i) {
    if (Op->Header.Args[i].IsInvalid()) break;

@@ -381,7 +384,11 @@ DEF_OP(Thunk) {

  auto thunkFn = ThreadState->CTX->ThunkHandler->LookupThunk(Op->ThunkNameHash);
  LoadConstant(x2, (uintptr_t)thunkFn);
+#ifdef VIXL_SIMULATOR
+  GenerateIndirectRuntimeCall<void, void*, void*>(x2);
+#else
  blr(x2);
+#endif

  PopDynamicRegsAndLR();

@@ -436,7 +443,7 @@ DEF_OP(ValidateCode) {
  }
 }

-DEF_OP(RemoveThreadCodeEntry) {
+DEF_OP(ThreadRemoveCodeEntry) {
  // Arguments are passed as follows:
  // X0: Thread
  // X1: RIP
@@ -446,9 +453,13 @@ DEF_OP(RemoveThreadCodeEntry) {
  mov(x0, STATE);
  LoadConstant(x1, Entry);

-  ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.RemoveThreadCodeEntryFromJIT)));
+  ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadRemoveCodeEntryFromJIT)));
  SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+  GenerateIndirectRuntimeCall<void, void*, void*>(x2);
+#else
  blr(x2);
+#endif
  FillStaticRegs();

  // Fix the stack and any values that were stepped on
@@ -459,6 +470,7 @@ DEF_OP(CPUID) {
  auto Op = IROp->C<IR::IROp_CPUID>();

  PushDynamicRegsAndLR();
+  SpillStaticRegs();

  // x0 = CPUID Handler
  // x1 = CPUID Function
@@ -467,10 +479,13 @@ DEF_OP(CPUID) {
  ldr(x3, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.CPUIDFunction)));
  mov(x1, GetReg<RA_64>(Op->Function.ID()));
  mov(x2, GetReg<RA_64>(Op->Leaf.ID()));
-  SpillStaticRegs();
+#ifdef VIXL_SIMULATOR
+  GenerateIndirectRuntimeCall<__uint128_t, void*, uint64_t, uint64_t>(x3);
+#else
  blr(x3);
-  FillStaticRegs();
+#endif

+  FillStaticRegs();
  PopDynamicRegsAndLR();

  // Results are in x0, x1
@@ -492,7 +507,7 @@ void Arm64JITCore::RegisterBranchHandlers() {
  REGISTER_OP(INLINESYSCALL,     InlineSyscall);
  REGISTER_OP(THUNK,             Thunk);
  REGISTER_OP(VALIDATECODE,      ValidateCode);
-  REGISTER_OP(REMOVETHREADCODEENTRY,   RemoveThreadCodeEntry);
+  REGISTER_OP(THREADREMOVECODEENTRY,   ThreadRemoveCodeEntry);
  REGISTER_OP(CPUID,             CPUID);
 #undef REGISTER_OP
 }
@@ -10,28 +10,117 @@ namespace FEXCore::CPU {

 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)
 DEF_OP(VInsGPR) {
-  auto Op = IROp->C<IR::IROp_VInsGPR>();
-  mov(GetDst(Node), GetSrc(Op->DestVector.ID()));
-  switch (Op->Header.ElementSize) {
-    case 1: {
-      ins(GetDst(Node).V16B(), Op->DestIdx, GetReg<RA_32>(Op->Src.ID()));
-    break;
+  const auto Op = IROp->C<IR::IROp_VInsGPR>();
+  const auto OpSize = IROp->Size;
+
+  const auto DestIdx = Op->DestIdx;
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto DestVector = GetSrc(Op->DestVector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    const auto ElementSizeBits = ElementSize * 8;
+    const auto Offset = ElementSizeBits * DestIdx;
+
+    const auto SSEBitSize = Core::CPUState::XMM_SSE_REG_SIZE * 8;
+    const auto InUpperLane = Offset >= SSEBitSize;
+
+    // This is going to be a little gross. Pls forgive me.
+    // Since SVE has the whole vector length agnostic programming
+    // thing going on, we can't exactly freely insert entries into
+    // arbitrary locations in the vector.
+    //
+    // SVE *does* have INSR, however this only shifts the entire
+    // vector to the left by an element size and inserts a value
+    // at the beginning of the vector. Not *quite* what we need.
+    // (though INSR *is* very useful for other things).
+    //
+    // The idea is (in the case of the upper lane), move the upper
+    // lane down, insert into it and recombine with the lower lane.
+    //
+    // In the case of the lower lane, insert and then recombine with
+    // the upper lane.
+
+    if (InUpperLane) {
+      // Move the upper lane down for the insertion.
+      const auto CompactPred = p0;
+      not_(CompactPred.VnB(), PRED_TMP_32B.Zeroing(), PRED_TMP_16B.VnB());
+      compact(VTMP1.Z().VnD(), CompactPred, DestVector.Z().VnD());
    }
-    case 2: {
-      ins(GetDst(Node).V8H(), Op->DestIdx, GetReg<RA_32>(Op->Src.ID()));
-    break;
+
+    // Put data in place for destructive SPLICE below.
+    mov(Dst.Z().VnD(), DestVector.Z().VnD());
+
+    // Inserts the GPR value into the given V register.
+    // Also automatically adjusts the index in the case of using the
+    // moved upper lane.
+    const auto Insert = [&](const aarch64::VRegister& reg, int index) {
+      switch (ElementSize) {
+        case 1:
+          if (InUpperLane) {
+            index -= 16;
+          }
+          ins(reg.V16B(), index, GetReg<RA_32>(Op->Src.ID()));
+          break;
+        case 2:
+          if (InUpperLane) {
+            index -= 8;
+          }
+          ins(reg.V8H(), index, GetReg<RA_32>(Op->Src.ID()));
+          break;
+        case 4:
+          if (InUpperLane) {
+            index -= 4;
+          }
+          ins(reg.V4S(), index, GetReg<RA_32>(Op->Src.ID()));
+          break;
+        case 8:
+          if (InUpperLane) {
+            index -= 2;
+          }
+          ins(reg.V2D(), index, GetReg<RA_64>(Op->Src.ID()));
+          break;
+        default:
+          LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+          break;
+      }
+    };
+
+    if (InUpperLane) {
+      Insert(VTMP1, DestIdx);
+      splice(Dst.Z().VnD(), PRED_TMP_16B, Dst.Z().VnD(), VTMP1.Z().VnD());
+    } else {
+      Insert(Dst, DestIdx);
+      splice(Dst.Z().VnD(), PRED_TMP_16B, Dst.Z().VnD(), DestVector.Z().VnD());
    }
-    case 4: {
-      ins(GetDst(Node).V4S(), Op->DestIdx, GetReg<RA_32>(Op->Src.ID()));
-    break;
+  } else {
+    mov(Dst, DestVector);
+
+    switch (ElementSize) {
+      case 1: {
+        ins(Dst.V16B(), DestIdx, GetReg<RA_32>(Op->Src.ID()));
+        break;
+      }
+      case 2: {
+        ins(Dst.V8H(), DestIdx, GetReg<RA_32>(Op->Src.ID()));
+        break;
+      }
+      case 4: {
+        ins(Dst.V4S(), DestIdx, GetReg<RA_32>(Op->Src.ID()));
+        break;
+      }
+      case 8: {
+        ins(Dst.V2D(), DestIdx, GetReg<RA_64>(Op->Src.ID()));
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+        break;
    }
-    case 8: {
-      ins(GetDst(Node).V2D(), Op->DestIdx, GetReg<RA_64>(Op->Src.ID()));
-    break;
-    }
-    default: LOGMAN_MSG_A_FMT("Unknown Element Size: {}", Op->Header.ElementSize); break;
  }
 }

@@ -57,8 +146,11 @@ DEF_OP(VCastFromGPR) {
 }

 DEF_OP(Float_FromGPR_S) {
-  auto Op = IROp->C<IR::IROp_Float_FromGPR_S>();
-  const uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const auto Op = IROp->C<IR::IROp_Float_FromGPR_S>();
+
+  const uint16_t ElementSize = Op->Header.ElementSize;
+  const uint16_t Conv = (ElementSize << 8) | Op->SrcElementSize;
+
  switch (Conv) {
    case 0x0404: { // Float <- int32_t
      scvtf(GetDst(Node).S(), GetReg<RA_32>(Op->Src.ID()));
@@ -76,6 +168,10 @@ DEF_OP(Float_FromGPR_S) {
      scvtf(GetDst(Node).D(), GetReg<RA_64>(Op->Src.ID()));
      break;
    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled conversion mask: Mask=0x{:04x}, ElementSize={}, SrcElementSize={}",
+                       Conv, ElementSize, Op->SrcElementSize);
+      break;
  }
 }

@@ -96,116 +192,379 @@ DEF_OP(Float_FToF) {
 }

 DEF_OP(Vector_SToF) {
-  auto Op = IROp->C<IR::IROp_Vector_SToF>();
-  switch (Op->Header.ElementSize) {
-    case 4:
-      scvtf(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
-    break;
-    case 8:
-      scvtf(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_SToF element size: {}", Op->Header.ElementSize);
+  const auto Op = IROp->C<IR::IROp_Vector_SToF>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    const auto Mask = PRED_TMP_32B.Merging();
+
+    switch (ElementSize) {
+      case 2:
+        scvtf(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+        break;
+      case 4:
+        scvtf(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+        break;
+      case 8:
+        scvtf(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_SToF element size: {}", ElementSize);
+        break;
+    }
+  } else {
+    switch (ElementSize) {
+      case 2:
+        scvtf(Dst.V8H(), Vector.V8H());
+        break;
+      case 4:
+        scvtf(Dst.V4S(), Vector.V4S());
+        break;
+      case 8:
+        scvtf(Dst.V2D(), Vector.V2D());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_SToF element size: {}", ElementSize);
+        break;
+    }
  }
 }

 DEF_OP(Vector_FToZS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToZS>();
-  switch (Op->Header.ElementSize) {
-    case 4:
-      fcvtzs(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
-    break;
-    case 8:
-      fcvtzs(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToZS element size: {}", Op->Header.ElementSize);
+  const auto Op = IROp->C<IR::IROp_Vector_FToZS>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    const auto Mask = PRED_TMP_32B.Merging();
+
+    switch (ElementSize) {
+      case 2:
+        fcvtzs(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+        break;
+      case 4:
+        fcvtzs(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+        break;
+      case 8:
+        fcvtzs(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToZS element size: {}", ElementSize);
+        break;
+    }
+  } else {
+    switch (ElementSize) {
+      case 2:
+        fcvtzs(Dst.V8H(), Vector.V8H());
+        break;
+      case 4:
+        fcvtzs(Dst.V4S(), Vector.V4S());
+        break;
+      case 8:
+        fcvtzs(Dst.V2D(), Vector.V2D());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToZS element size: {}", ElementSize);
+        break;
+    }
  }
 }

 DEF_OP(Vector_FToS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToS>();
-  switch (Op->Header.ElementSize) {
-    case 4:
-      frinti(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
-      fcvtzs(GetDst(Node).V4S(), GetDst(Node).V4S());
-    break;
-    case 8:
-      frinti(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
-      fcvtzs(GetDst(Node).V2D(), GetDst(Node).V2D());
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToS element size: {}", Op->Header.ElementSize);
+  const auto Op = IROp->C<IR::IROp_Vector_FToS>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    const auto Mask = PRED_TMP_32B.Merging();
+
+    switch (ElementSize) {
+      case 2:
+        frinti(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+        fcvtzs(Dst.Z().VnH(), Mask, Dst.Z().VnH());
+        break;
+      case 4:
+        frinti(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+        fcvtzs(Dst.Z().VnS(), Mask, Dst.Z().VnS());
+        break;
+      case 8:
+        frinti(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+        fcvtzs(Dst.Z().VnD(), Mask, Dst.Z().VnD());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToS element size: {}", ElementSize);
+        break;
+    }
+  } else {
+    switch (ElementSize) {
+      case 2:
+        frinti(Dst.V8H(), Vector.V8H());
+        fcvtzs(Dst.V8H(), Dst.V8H());
+        break;
+      case 4:
+        frinti(Dst.V4S(), Vector.V4S());
+        fcvtzs(Dst.V4S(), Dst.V4S());
+        break;
+      case 8:
+        frinti(Dst.V2D(), Vector.V2D());
+        fcvtzs(Dst.V2D(), Dst.V2D());
+        break;
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToS element size: {}", ElementSize);
+        break;
+    }
  }
 }

 DEF_OP(Vector_FToF) {
-  auto Op = IROp->C<IR::IROp_Vector_FToF>();
-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const auto Op = IROp->C<IR::IROp_Vector_FToF>();
+  const auto OpSize = IROp->Size;

-  switch (Conv) {
-    case 0x0804: { // Double <- Float
-      fcvtl(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2S());
-      break;
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto Conv = (ElementSize << 8) | Op->SrcElementSize;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    // Curiously, FCVTLT and FCVTNT have no bottom variants,
+    // and also interesting is that FCVTLT will iterate the
+    // source vector by accessing each odd element and storing
+    // them consecutively in the destination.
+    //
+    // FCVTNT is somewhat like the opposite. It will read each
+    // consecutive element, but store each result into every odd
+    // element in the destination vector.
+    //
+    // We need to undo the behavior of FCVTNT with UZP2. In the case
+    // of FCVTLT, we instead need to set the vector up with ZIP1, so
+    // that the elements will be processed correctly.
+
+    const auto Mask = PRED_TMP_32B.Merging();
+
+    switch (Conv) {
+      case 0x0402: { // Float <- Half
+        zip1(Dst.Z().VnH(), Vector.Z().VnH(), Vector.Z().VnH());
+        fcvtlt(Dst.Z().VnS(), Mask, Dst.Z().VnH());
+        break;
+      }
+      case 0x0804: { // Double <- Float
+        zip1(Dst.Z().VnS(), Vector.Z().VnS(), Vector.Z().VnS());
+        fcvtlt(Dst.Z().VnD(), Mask, Dst.Z().VnS());
+        break;
+      }
+      case 0x0204: { // Half <- Float
+        fcvtnt(Dst.Z().VnH(), Mask, Vector.Z().VnS());
+        uzp2(Dst.Z().VnH(), Dst.Z().VnH(), Dst.Z().VnH());
+        break;
+      }
+      case 0x0408: { // Float <- Double
+        fcvtnt(Dst.Z().VnS(), Mask, Vector.Z().VnD());
+        uzp2(Dst.Z().VnS(), Dst.Z().VnS(), Dst.Z().VnS());
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToF Type : 0x{:04x}", Conv);
+        break;
    }
-    case 0x0408: { // Float <- Double
-      fcvtn(GetDst(Node).V2S(), GetSrc(Op->Vector.ID()).V2D());
-      break;
+  } else {
+    switch (Conv) {
+      case 0x0402: { // Float <- Half
+        fcvtl(Dst.V4S(), Vector.V4H());
+        break;
+      }
+      case 0x0804: { // Double <- Float
+        fcvtl(Dst.V2D(), Vector.V2S());
+        break;
+      }
+      case 0x0204: { // Half <- Float
+        fcvtn(Dst.V4H(), Vector.V4S());
+        break;
+      }
+      case 0x0408: { // Float <- Double
+        fcvtn(Dst.V2S(), Vector.V2D());
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Vector_FToF Type : 0x{:04x}", Conv);
+        break;
    }
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToF Type : 0x{:04x}", Conv); break;
  }
 }

 DEF_OP(Vector_FToI) {
-  auto Op = IROp->C<IR::IROp_Vector_FToI>();
-  switch (Op->Round) {
-    case FEXCore::IR::Round_Nearest.Val:
-      switch (Op->Header.ElementSize) {
-        case 4:
-          frintn(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
+  const auto Op = IROp->C<IR::IROp_Vector_FToI>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  if (HostSupportsSVE && Is256Bit) {
+    const auto Mask = PRED_TMP_32B.Merging();
+    
+    switch (Op->Round) {
+      case FEXCore::IR::Round_Nearest.Val:
+        switch (ElementSize) {
+          case 2:
+            frintn(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+            break;
+          case 4:
+            frintn(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+            break;
+          case 8:
+            frintn(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+            break;
+        }
        break;
-        case 8:
-          frintn(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
+
+      case FEXCore::IR::Round_Negative_Infinity.Val:
+        switch (ElementSize) {
+          case 2:
+            frintm(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+            break;
+          case 4:
+            frintm(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+            break;
+          case 8:
+            frintm(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+            break;
+        }
        break;
-      }
-    break;
-    case FEXCore::IR::Round_Negative_Infinity.Val:
-      switch (Op->Header.ElementSize) {
-        case 4:
-          frintm(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
+
+      case FEXCore::IR::Round_Positive_Infinity.Val:
+        switch (ElementSize) {
+          case 2:
+            frintp(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+            break;
+          case 4:
+            frintp(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+            break;
+          case 8:
+            frintp(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+            break;
+        }
        break;
-        case 8:
-          frintm(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
+
+      case FEXCore::IR::Round_Towards_Zero.Val:
+        switch (ElementSize) {
+          case 2:
+            frintz(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+            break;
+          case 4:
+            frintz(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+            break;
+          case 8:
+            frintz(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+            break;
+        }
        break;
-      }
-    break;
-    case FEXCore::IR::Round_Positive_Infinity.Val:
-      switch (Op->Header.ElementSize) {
-        case 4:
-          frintp(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
+
+      case FEXCore::IR::Round_Host.Val:
+        switch (ElementSize) {
+          case 2:
+            frinti(Dst.Z().VnH(), Mask, Vector.Z().VnH());
+            break;
+          case 4:
+            frinti(Dst.Z().VnS(), Mask, Vector.Z().VnS());
+            break;
+          case 8:
+            frinti(Dst.Z().VnD(), Mask, Vector.Z().VnD());
+            break;
+        }
        break;
-        case 8:
-          frintp(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
+    }
+  } else {
+    switch (Op->Round) {
+      case FEXCore::IR::Round_Nearest.Val:
+        switch (ElementSize) {
+          case 2:
+            frintn(Dst.V8H(), Vector.V8H());
+            break;
+          case 4:
+            frintn(Dst.V4S(), Vector.V4S());
+            break;
+          case 8:
+            frintn(Dst.V2D(), Vector.V2D());
+            break;
+        }
        break;
-      }
-    break;
-    case FEXCore::IR::Round_Towards_Zero.Val:
-      switch (Op->Header.ElementSize) {
-        case 4:
-          frintz(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
+
+      case FEXCore::IR::Round_Negative_Infinity.Val:
+        switch (ElementSize) {
+          case 2:
+            frintm(Dst.V8H(), Vector.V8H());
+            break;
+          case 4:
+            frintm(Dst.V4S(), Vector.V4S());
+            break;
+          case 8:
+            frintm(Dst.V2D(), Vector.V2D());
+            break;
+        }
        break;
-        case 8:
-          frintz(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
+
+      case FEXCore::IR::Round_Positive_Infinity.Val:
+        switch (ElementSize) {
+          case 2:
+            frintp(Dst.V8H(), Vector.V8H());
+            break;
+          case 4:
+            frintp(Dst.V4S(), Vector.V4S());
+            break;
+          case 8:
+            frintp(Dst.V2D(), Vector.V2D());
+            break;
+        }
        break;
-      }
-    break;
-    case FEXCore::IR::Round_Host.Val:
-      switch (Op->Header.ElementSize) {
-        case 4:
-          frinti(GetDst(Node).V4S(), GetSrc(Op->Vector.ID()).V4S());
+
+      case FEXCore::IR::Round_Towards_Zero.Val:
+        switch (ElementSize) {
+          case 2:
+            frintz(Dst.V8H(), Vector.V8H());
+            break;
+          case 4:
+            frintz(Dst.V4S(), Vector.V4S());
+            break;
+          case 8:
+            frintz(Dst.V2D(), Vector.V2D());
+            break;
+        }
        break;
-        case 8:
-          frinti(GetDst(Node).V2D(), GetSrc(Op->Vector.ID()).V2D());
+
+      case FEXCore::IR::Round_Host.Val:
+        switch (ElementSize) {
+          case 2:
+            frinti(Dst.V8H(), Vector.V8H());
+            break;
+          case 4:
+            frinti(Dst.V4S(), Vector.V4S());
+            break;
+          case 8:
+            frinti(Dst.V2D(), Vector.V2D());
+            break;
+        }
        break;
-      }
-    break;
+    }
  }
 }

@@ -10,7 +10,7 @@ $end_info$
 namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)

 DEF_OP(AESImc) {
  auto Op = IROp->C<IR::IROp_VAESImc>();
@@ -18,7 +18,7 @@ DEF_OP(AESImc) {
 }

 DEF_OP(AESEnc) {
-	auto Op = IROp->C<IR::IROp_VAESEnc>();
+  auto Op = IROp->C<IR::IROp_VAESEnc>();
  eor(VTMP2.V16B(), VTMP2.V16B(), VTMP2.V16B());
  mov(VTMP1.V16B(), GetSrc(Op->State.ID()).V16B());
  aese(VTMP1.V16B(), VTMP2.V16B());
@@ -27,7 +27,7 @@ DEF_OP(AESEnc) {
 }

 DEF_OP(AESEncLast) {
-	auto Op = IROp->C<IR::IROp_VAESEncLast>();
+  auto Op = IROp->C<IR::IROp_VAESEncLast>();
  eor(VTMP2.V16B(), VTMP2.V16B(), VTMP2.V16B());
  mov(VTMP1.V16B(), GetSrc(Op->State.ID()).V16B());
  aese(VTMP1.V16B(), VTMP2.V16B());
@@ -35,7 +35,7 @@ DEF_OP(AESEncLast) {
 }

 DEF_OP(AESDec) {
-	auto Op = IROp->C<IR::IROp_VAESDec>();
+  auto Op = IROp->C<IR::IROp_VAESDec>();
  eor(VTMP2.V16B(), VTMP2.V16B(), VTMP2.V16B());
  mov(VTMP1.V16B(), GetSrc(Op->State.ID()).V16B());
  aesd(VTMP1.V16B(), VTMP2.V16B());
@@ -44,7 +44,7 @@ DEF_OP(AESDec) {
 }

 DEF_OP(AESDecLast) {
-	auto Op = IROp->C<IR::IROp_VAESDecLast>();
+  auto Op = IROp->C<IR::IROp_VAESDecLast>();
  eor(VTMP2.V16B(), VTMP2.V16B(), VTMP2.V16B());
  mov(VTMP1.V16B(), GetSrc(Op->State.ID()).V16B());
  aesd(VTMP1.V16B(), VTMP2.V16B());
@@ -52,7 +52,7 @@ DEF_OP(AESDecLast) {
 }

 DEF_OP(AESKeyGenAssist) {
-	auto Op = IROp->C<IR::IROp_VAESKeyGenAssist>();
+  auto Op = IROp->C<IR::IROp_VAESKeyGenAssist>();

  aarch64::Literal ConstantLiteral (0x0C030609'0306090CULL, 0x040B0E01'0B0E0104ULL);
  aarch64::Label PastConstant;
@@ -69,9 +69,8 @@ DEF_OP(AESKeyGenAssist) {
  if (Op->RCON) {
    tbl(VTMP1.V16B(), VTMP1.V16B(), VTMP3.V16B());

-    LoadConstant(TMP1.W(), Op->RCON);
-    ins(VTMP2.V4S(), 1, TMP1.W());
-    ins(VTMP2.V4S(), 3, TMP1.W());
+    LoadConstant(TMP1, static_cast<uint64_t>(Op->RCON) << 32);
+    dup(VTMP2.V2D(), TMP1);
    eor(GetDst(Node).V16B(), VTMP1.V16B(), VTMP2.V16B());
  }
  else {
@@ -10,7 +10,7 @@ namespace FEXCore::CPU {

 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)
 DEF_OP(GetHostFlag) {
  auto Op = IROp->C<IR::IROp_GetHostFlag>();
  ubfx(GetReg<RA_64>(Node), GetReg<RA_64>(Op->Value.ID()), Op->Flag, 1);
@@ -27,6 +27,8 @@ $end_info$
 #include <FEXCore/Core/UContext.h>
 #include <FEXCore/Utils/Allocator.h>
 #include <FEXCore/Utils/CompilerDefs.h>
+#include <FEXCore/Utils/EnumUtils.h>
+#include <FEXCore/Utils/Profiler.h>

 #include "Interface/Core/Interpreter/InterpreterOps.h"

@@ -78,7 +80,7 @@ namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;

-void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
+void Arm64JITCore::Op_Unhandled(IR::IROp_Header const *IROp, IR::NodeID Node) {
  FallbackInfo Info;
  if (!InterpreterOps::GetFallbackHandler(IROp, &Info)) {
 #if defined(ASSERTIONS_ENABLED) && ASSERTIONS_ENABLED
@@ -93,7 +95,12 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {

        uxth(w0, GetReg<RA_32>(IROp->Args[0].ID()));
        ldr(x1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<void, uint16_t>(x1);
+#else
        blr(x1);
+#endif

        PopDynamicRegsAndLR();

@@ -108,7 +115,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {

        fmov(v0.S(), GetSrc(IROp->Args[0].ID()).S()) ;
        ldr(x0, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<__uint128_t, float>(x0);
+#else
        blr(x0);
+#endif

        PopDynamicRegsAndLR();

@@ -127,7 +138,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {

        mov(v0.D(), GetSrc(IROp->Args[0].ID()).D());
        ldr(x0, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<__uint128_t, double>(x0);
+#else
        blr(x0);
+#endif

        PopDynamicRegsAndLR();

@@ -152,7 +167,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
          mov(w0, GetReg<RA_32>(IROp->Args[0].ID()));
        }
        ldr(x1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<__uint128_t, uint32_t>(x1);
+#else
        blr(x1);
+#endif

        PopDynamicRegsAndLR();

@@ -173,7 +192,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<float, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -192,7 +215,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<double, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -209,7 +236,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {

        mov(v0.D(), GetSrc(IROp->Args[0].ID()).D());
        ldr(x0, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<double, double>(x0);
+#else
        blr(x0);
+#endif

        PopDynamicRegsAndLR();

@@ -228,7 +259,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        mov(v0.D(), GetSrc(IROp->Args[0].ID()).D());
        mov(v1.D(), GetSrc(IROp->Args[1].ID()).D());
        ldr(x0, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<double, double, double>(x0);
+#else
        blr(x0);
+#endif

        PopDynamicRegsAndLR();

@@ -248,7 +283,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<uint32_t, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -266,7 +305,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<uint32_t, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -284,7 +327,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -305,8 +352,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w3, GetSrc(IROp->Args[1].ID()).V8H(), 4);

        ldr(x4, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<uint64_t, uint64_t, uint64_t, uint64_t, uint64_t>(x4);
+#else
        blr(x4);
-
+#endif
        PopDynamicRegsAndLR();

        FillStaticRegs();
@@ -323,7 +373,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w1, GetSrc(IROp->Args[0].ID()).V8H(), 4);

        ldr(x2, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<__uint128_t, uint64_t, uint64_t>(x2);
+#else
        blr(x2);
+#endif

        PopDynamicRegsAndLR();

@@ -346,7 +400,11 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
        umov(w3, GetSrc(IROp->Args[1].ID()).V8H(), 4);

        ldr(x4, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.FallbackHandlerPointers[Info.HandlerIndex])));
+#ifdef VIXL_SIMULATOR
+        GenerateIndirectRuntimeCall<__uint128_t, uint64_t, uint64_t, uint64_t, uint64_t>(x4);
+#else
        blr(x4);
+#endif

        PopDynamicRegsAndLR();

@@ -361,7 +419,8 @@ void Arm64JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
      case FABI_UNKNOWN:
      default:
 #if defined(ASSERTIONS_ENABLED) && ASSERTIONS_ENABLED
-        LOGMAN_MSG_A_FMT("Unhandled IR Fallback ABI: {} {}", FEXCore::IR::GetName(IROp->Op), Info.ABI);
+        LOGMAN_MSG_A_FMT("Unhandled IR Fallback ABI: {} {}",
+                         FEXCore::IR::GetName(IROp->Op), ToUnderlying(Info.ABI));
 #endif
      break;
    }
@@ -395,7 +454,7 @@ static uint64_t Arm64JITCore_ExitFunctionLink(FEXCore::Core::CpuStateFrame *Fram
    vixl::aarch64::CPU::EnsureIAndDCacheCoherency((void*)branch, 24);

    // Add de-linking handler
-    Thread->LookupCache->AddBlockLink(GuestRip, (uintptr_t)record, [branch, LinkerAddress]{
+    Context::Context::ThreadAddBlockLink(Thread, GuestRip, (uintptr_t)record, [branch, LinkerAddress]{
      vixl::aarch64::Assembler emit((uint8_t*)(branch), 24);
      vixl::CodeBufferCheckScope scope(&emit, 24, vixl::CodeBufferCheckScope::kDontReserveBufferSpace, vixl::CodeBufferCheckScope::kNoAssert);
      Literal l_BranchHost{LinkerAddress};
@@ -410,7 +469,7 @@ static uint64_t Arm64JITCore_ExitFunctionLink(FEXCore::Core::CpuStateFrame *Fram
    record[0] = HostCode;

    // Add de-linking handler
-    Thread->LookupCache->AddBlockLink(GuestRip, (uintptr_t)record, [record, LinkerAddress]{
+    Context::Context::ThreadAddBlockLink(Thread, GuestRip, (uintptr_t)record, [record, LinkerAddress]{
      record[0] = LinkerAddress;
    });
  }
@@ -418,12 +477,13 @@ static uint64_t Arm64JITCore_ExitFunctionLink(FEXCore::Core::CpuStateFrame *Fram
  return HostCode;
 }

-void Arm64JITCore::Op_NoOp(IR::IROp_Header *IROp, IR::NodeID Node) {
+void Arm64JITCore::Op_NoOp(IR::IROp_Header const *IROp, IR::NodeID Node) {
 }

 Arm64JITCore::Arm64JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::InternalThreadState *Thread)
  : CPUBackend(Thread, INITIAL_CODE_SIZE, MAX_CODE_SIZE)
  , Arm64Emitter(ctx, 0)
+  , HostSupportsSVE{ctx->HostFeatures.SupportsAVX}
  , CTX {ctx} {

  RAPass = Thread->PassManager->GetPass<IR::RegisterAllocationPass>("RA");
@@ -470,10 +530,10 @@ Arm64JITCore::Arm64JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::Intern

    // Common
    auto &Common = ThreadState->CurrentFrame->Pointers.Common;
-    
+
    Common.PrintValue = reinterpret_cast<uint64_t>(PrintValue);
    Common.PrintVectorValue = reinterpret_cast<uint64_t>(PrintVectorValue);
-    Common.RemoveThreadCodeEntryFromJIT = reinterpret_cast<uintptr_t>(&Context::Context::RemoveThreadCodeEntryFromJit);
+    Common.ThreadRemoveCodeEntryFromJIT = reinterpret_cast<uintptr_t>(&Context::Context::ThreadRemoveCodeEntryFromJit);
    Common.CPUIDObj = reinterpret_cast<uint64_t>(&CTX->CPUID);

    {
@@ -483,7 +543,7 @@ Arm64JITCore::Arm64JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::Intern

    Common.SyscallHandlerObj = reinterpret_cast<uint64_t>(CTX->SyscallHandler);
    Common.SyscallHandlerFunc = reinterpret_cast<uint64_t>(FEXCore::Context::HandleSyscall);
-    Common.ExitFunctionLink = reinterpret_cast<uintptr_t>(&Arm64JITCore_ExitFunctionLink);
+    Common.ExitFunctionLink = reinterpret_cast<uintptr_t>(&Context::Context::ThreadExitFunctionLink<Arm64JITCore_ExitFunctionLink>);


    // Fill in the fallback handlers
@@ -491,7 +551,7 @@ Arm64JITCore::Arm64JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::Intern

    // Platform Specific
    auto &AArch64 = ThreadState->CurrentFrame->Pointers.AArch64;
-    
+
    AArch64.LUDIV = reinterpret_cast<uint64_t>(LUDIV);
    AArch64.LDIV = reinterpret_cast<uint64_t>(LDIV);
    AArch64.LUREM = reinterpret_cast<uint64_t>(LUREM);
@@ -508,6 +568,7 @@ void Arm64JITCore::InitializeSignalHandlers(FEXCore::Context::Context *CTX) {
    return Thread->CTX->Dispatcher->HandleSIGILL(Thread, Signal, info, ucontext);
  }, true);

+#ifdef _M_ARM_64
  CTX->SignalDelegation->RegisterHostSignalHandler(SIGBUS, [](FEXCore::Core::InternalThreadState *Thread, int Signal, void *info, void *ucontext) -> bool {
    if (!Thread->CPUBackend->IsAddressInCodeBuffer(ArchHelpers::Context::GetPc(ucontext))) {
      // Wasn't a sigbus in JIT code
@@ -516,6 +577,7 @@ void Arm64JITCore::InitializeSignalHandlers(FEXCore::Context::Context *CTX) {

    return FEXCore::ArchHelpers::Arm64::HandleSIGBUS(Thread->CTX->Config.ParanoidTSO(), Signal, info, ucontext);
  }, true);
+#endif
 }

 void Arm64JITCore::EmitDetectionString() {
@@ -527,7 +589,7 @@ void Arm64JITCore::EmitDetectionString() {

 void Arm64JITCore::ClearCache() {
  // Get the backing code buffer
-  
+
  auto CodeBuffer = GetEmptyCodeBuffer();
  *GetBuffer() = vixl::CodeBuffer(CodeBuffer->Ptr, CodeBuffer->Size);
  EmitDetectionString();
@@ -549,12 +611,12 @@ template<>
 aarch64::Register Arm64JITCore::GetReg<Arm64JITCore::RA_32>(IR::NodeID Node) const {
  auto Reg = GetPhys(Node);

+  LOGMAN_THROW_AA_FMT(Reg.Class == IR::GPRFixedClass.Val || Reg.Class == IR::GPRClass.Val, "Unexpected Class: {}", Reg.Class);
+
  if (Reg.Class == IR::GPRFixedClass.Val) {
    return SRA64[Reg.Reg].W();
  } else if (Reg.Class == IR::GPRClass.Val) {
    return RA64[Reg.Reg].W();
-  } else {
-    LOGMAN_THROW_A_FMT(false, "Unexpected Class: {}", Reg.Class);
  }

  FEX_UNREACHABLE;
@@ -564,12 +626,12 @@ template<>
 aarch64::Register Arm64JITCore::GetReg<Arm64JITCore::RA_64>(IR::NodeID Node) const {
  auto Reg = GetPhys(Node);

+  LOGMAN_THROW_AA_FMT(Reg.Class == IR::GPRFixedClass.Val || Reg.Class == IR::GPRClass.Val, "Unexpected Class: {}", Reg.Class);
+
  if (Reg.Class == IR::GPRFixedClass.Val) {
    return SRA64[Reg.Reg];
  } else if (Reg.Class == IR::GPRClass.Val) {
    return RA64[Reg.Reg];
-  } else {
-    LOGMAN_THROW_A_FMT(false, "Unexpected Class: {}", Reg.Class);
  }

  FEX_UNREACHABLE;
@@ -590,12 +652,12 @@ std::pair<aarch64::Register, aarch64::Register> Arm64JITCore::GetSrcPair<Arm64JI
 aarch64::VRegister Arm64JITCore::GetSrc(IR::NodeID Node) const {
  auto Reg = GetPhys(Node);

+  LOGMAN_THROW_AA_FMT(Reg.Class == IR::FPRFixedClass.Val || Reg.Class == IR::FPRClass.Val, "Unexpected Class: {}", Reg.Class);
+
  if (Reg.Class == IR::FPRFixedClass.Val) {
    return SRAFPR[Reg.Reg];
  } else if (Reg.Class == IR::FPRClass.Val) {
    return RAFPR[Reg.Reg];
-  } else {
-    LOGMAN_THROW_A_FMT(false, "Unexpected Class: {}", Reg.Class);
  }

  FEX_UNREACHABLE;
@@ -604,12 +666,12 @@ aarch64::VRegister Arm64JITCore::GetSrc(IR::NodeID Node) const {
 aarch64::VRegister Arm64JITCore::GetDst(IR::NodeID Node) const {
  auto Reg = GetPhys(Node);

+  LOGMAN_THROW_AA_FMT(Reg.Class == IR::FPRFixedClass.Val || Reg.Class == IR::FPRClass.Val, "Unexpected Class: {}", Reg.Class);
+
  if (Reg.Class == IR::FPRFixedClass.Val) {
    return SRAFPR[Reg.Reg];
  } else if (Reg.Class == IR::FPRClass.Val) {
    return RAFPR[Reg.Reg];
-  } else {
-    LOGMAN_THROW_A_FMT(false, "Unexpected Class: {}", Reg.Class);
  }

  FEX_UNREACHABLE;
@@ -665,7 +727,13 @@ bool Arm64JITCore::IsGPR(IR::NodeID Node) const {
  return Class == IR::GPRClass || Class == IR::GPRFixedClass;
 }

-void *Arm64JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IRListView const *IR, [[maybe_unused]] FEXCore::Core::DebugData *DebugData, FEXCore::IR::RegisterAllocationData *RAData, bool GDBEnabled) {
+void *Arm64JITCore::CompileCode(uint64_t Entry,
+                                FEXCore::IR::IRListView const *IR,
+                                FEXCore::Core::DebugData *DebugData,
+                                FEXCore::IR::RegisterAllocationData *RAData,
+                                bool GDBEnabled) {
+  FEXCORE_PROFILE_SCOPED("Arm64::CompileCode");
+
  using namespace aarch64;
  JumpTargets.clear();
  uint32_t SSACount = IR->GetSSACount();
@@ -718,10 +786,12 @@ void *Arm64JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IR
  SpillSlots = RAData->SpillSlots();

  if (SpillSlots) {
-    if (IsImmAddSub(SpillSlots * 16)) {
-      sub(sp, sp, SpillSlots * 16);
+    const auto TotalSpillSlotsSize = SpillSlots * MaxSpillSlotSize;
+
+    if (IsImmAddSub(TotalSpillSlotsSize)) {
+      sub(sp, sp, TotalSpillSlotsSize);
    } else {
-      LoadConstant(x0, SpillSlots * 16);
+      LoadConstant(x0, TotalSpillSlotsSize);
      sub(sp, sp, x0);
    }
  }
@@ -732,7 +802,7 @@ void *Arm64JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IR
    using namespace FEXCore::IR;
 #if defined(ASSERTIONS_ENABLED) && ASSERTIONS_ENABLED
    auto BlockIROp = BlockHeader->CW<FEXCore::IR::IROp_CodeBlock>();
-    LOGMAN_THROW_A_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");
+    LOGMAN_THROW_AA_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");
 #endif

    auto BlockStartHostCode = GetCursorAddress<uint8_t *>();
@@ -789,14 +859,17 @@ void *Arm64JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IR
 }

 void Arm64JITCore::ResetStack() {
-  if (SpillSlots == 0)
+  if (SpillSlots == 0) {
    return;
+  }

-  if (IsImmAddSub(SpillSlots * 16)) {
-    add(sp, sp, SpillSlots * 16);
+  const auto TotalSpillSlotsSize = SpillSlots * MaxSpillSlotSize;
+
+  if (IsImmAddSub(TotalSpillSlotsSize)) {
+    add(sp, sp, TotalSpillSlotsSize);
  } else {
   // Too big to fit in a 12bit immediate
-   LoadConstant(x0, SpillSlots * 16);
+   LoadConstant(x0, TotalSpillSlotsSize);
   add(sp, sp, x0);
  }
 }
@@ -23,16 +23,6 @@ $end_info$
 #include <utility>
 #include <vector>

-#define STATE x28
-#define TMP1 x0
-#define TMP2 x1
-#define TMP3 x2
-#define TMP4 x3
-
-#define VTMP1 v1
-#define VTMP2 v2
-#define VTMP3 v3
-
 namespace FEXCore::Core {
  struct InternalThreadState;
 }
@@ -66,6 +56,7 @@ public:

 private:
  FEX_CONFIG_OPT(ParanoidTSO, PARANOIDTSO);
+  const bool HostSupportsSVE{};

  Label *PendingTargetLabel;
  FEXCore::Context::Context *CTX;
@@ -127,6 +118,17 @@ private:
                                              IR::MemOffsetType OffsetType,
                                              uint8_t OffsetScale);

+  // NOTE: Will use TMP1 as a way to encode immediates that happen to fall outside
+  //       the limits of the scalar plus immediate variant of SVE load/stores.
+  //
+  //       TMP1 is safe to use again once this memory operand is used with its
+  //       equivalent loads or stores that this was called for.
+  [[nodiscard]] SVEMemOperand GenerateSVEMemOperand(uint8_t AccessSize,
+                                                    aarch64::Register Base,
+                                                    IR::OrderedNodeWrapper Offset,
+                                                    IR::MemOffsetType OffsetType,
+                                                    uint8_t OffsetScale);
+
  [[nodiscard]] bool IsInlineConstant(const IR::OrderedNodeWrapper& Node, uint64_t* Value = nullptr) const;
  [[nodiscard]] bool IsInlineEntrypointOffset(const IR::OrderedNodeWrapper& WNode, uint64_t* Value) const;

@@ -211,7 +213,7 @@ private:
  */
  uint8_t *GuestEntry{};

-  using OpHandler = void (Arm64JITCore::*)(IR::IROp_Header *IROp, IR::NodeID Node);
+  using OpHandler = void (Arm64JITCore::*)(IR::IROp_Header const *IROp, IR::NodeID Node);
  std::array<OpHandler, IR::IROps::OP_LAST + 1> OpHandlers {};
  void RegisterALUHandlers();
  void RegisterAtomicHandlers();
@@ -223,7 +225,7 @@ private:
  void RegisterMoveHandlers();
  void RegisterVectorHandlers();
  void RegisterEncryptionHandlers();
-#define DEF_OP(x) void Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)

  ///< Unhandled handler
  DEF_OP(Unhandled);
@@ -309,7 +311,7 @@ private:
  DEF_OP(InlineSyscall);
  DEF_OP(Thunk);
  DEF_OP(ValidateCode);
-  DEF_OP(RemoveThreadCodeEntry);
+  DEF_OP(ThreadRemoveCodeEntry);
  DEF_OP(CPUID);

  ///< Conversion ops
@@ -343,8 +345,6 @@ private:
  DEF_OP(StoreMemTSO);
  DEF_OP(ParanoidLoadMemTSO);
  DEF_OP(ParanoidStoreMemTSO);
-  DEF_OP(VLoadMemElement);
-  DEF_OP(VStoreMemElement);
  DEF_OP(CacheLineClear);
  DEF_OP(CacheLineZero);

@@ -364,13 +364,10 @@ private:
  ///< Move ops
  DEF_OP(ExtractElementPair);
  DEF_OP(CreateElementPair);
-  DEF_OP(Mov);

  ///< Vector ops
  DEF_OP(VectorZero);
  DEF_OP(VectorImm);
-  DEF_OP(SplatVector2);
-  DEF_OP(SplatVector4);
  DEF_OP(VMov);
  DEF_OP(VAnd);
  DEF_OP(VBic);
@@ -429,18 +426,13 @@ private:
  DEF_OP(VUShrS);
  DEF_OP(VSShrS);
  DEF_OP(VInsElement);
-  DEF_OP(VInsScalarElement);
-  DEF_OP(VExtractElement);
  DEF_OP(VDupElement);
  DEF_OP(VExtr);
-  DEF_OP(VSLI);
-  DEF_OP(VSRI);
  DEF_OP(VUShrI);
  DEF_OP(VSShrI);
  DEF_OP(VShlI);
  DEF_OP(VUShrNI);
  DEF_OP(VUShrNI2);
-  DEF_OP(VBitcast);
  DEF_OP(VSXTL);
  DEF_OP(VSXTL2);
  DEF_OP(VUXTL);
@@ -11,7 +11,7 @@ $end_info$
 namespace FEXCore::CPU {
 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)

 DEF_OP(GuestOpcode) {
  auto Op = IROp->C<IR::IROp_GuestOpcode>();
@@ -37,43 +37,41 @@ DEF_OP(Fence) {

 DEF_OP(Break) {
  auto Op = IROp->C<IR::IROp_Break>();
-  switch (Op->Reason) {
-    case FEXCore::IR::Break_Unimplemented: // Hard fault
-    case FEXCore::IR::Break_Interrupt: // Guest ud2
-      hlt(4);
-      break;
-    case FEXCore::IR::Break_Overflow: // overflow
-      ResetStack();
-      ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.OverflowExceptionHandler)));
-      br(TMP1);
-      break;
-    case FEXCore::IR::Break_Halt: { // HLT
-      // Time to quit
-      // Set our stack to the starting stack location
-      ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, ReturningStackLocation)));
-      add(sp, TMP1, 0);

-      // Now we need to jump to the thread stop handler
-      ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadStopHandlerSpillSRA)));
-      br(TMP1);
-      break;
-    }
-    case FEXCore::IR::Break_Interrupt3: { // INT3
-      ResetStack();
-      ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadPauseHandlerSpillSRA)));
-      br(TMP1);
-      break;
-    }
-    case FEXCore::IR::Break_InvalidInstruction:
-    {
-      ResetStack();
+  // First we must reset the stack
+  ResetStack();

-      ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.UnimplementedInstructionHandler)));
-      br(TMP1);
+  Core::CpuStateFrame::SynchronousFaultDataStruct State = {
+    .FaultToTopAndGeneratedException = 1,
+    .Signal = Op->Reason.Signal,
+    .TrapNo = Op->Reason.TrapNumber,
+    .si_code = Op->Reason.si_code,
+    .err_code = Op->Reason.ErrorRegister,
+  };

-      break;
-    }
-    default: LOGMAN_MSG_A_FMT("Unknown Break reason: {}", Op->Reason);
+  uint64_t Constant{};
+  memcpy(&Constant, &State, sizeof(State));
+
+  LoadConstant(x1, Constant);
+  str(x1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, SynchronousFaultData)));
+
+  switch (Op->Reason.Signal) {
+  case SIGILL:
+    ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGILL)));
+    br(TMP1);
+    break;
+  case SIGTRAP:
+    ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGTRAP)));
+    br(TMP1);
+    break;
+  case SIGSEGV:
+    ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGSEGV)));
+    br(TMP1);
+    break;
+  default:
+    ldr(TMP1, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGTRAP)));
+    br(TMP1);
+    break;
  }
 }

@@ -124,8 +122,11 @@ DEF_OP(SetRoundingMode) {

  mrs(TMP1, FPCR);

+  // vixl simulator doesn't support anything beyond ties-to-even rounding
+#ifndef VIXL_SIMULATOR
  // Insert the rounding flags
  bfi(TMP1, TMP2, 22, 2);
+#endif

  // Insert the FTZ flag
  lsr(TMP2, Src, 2);
@@ -139,6 +140,7 @@ DEF_OP(Print) {
  auto Op = IROp->C<IR::IROp_Print>();

  PushDynamicRegsAndLR();
+  SpillStaticRegs();

  if (IsGPR(Op->Value.ID())) {
    mov(x0, GetReg<RA_64>(Op->Value.ID()));
@@ -150,10 +152,10 @@ DEF_OP(Print) {
    fmov(x1, GetSrc(Op->Value.ID()).V1D(), 1);
    ldr(x3, MemOperand(STATE, offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.PrintVectorValue)));
  }
-  SpillStaticRegs();
-  blr(x3);
-  FillStaticRegs();

+  blr(x3);
+
+  FillStaticRegs();
  PopDynamicRegsAndLR();
 }

@@ -10,7 +10,7 @@ namespace FEXCore::CPU {

 using namespace vixl;
 using namespace vixl::aarch64;
-#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
+#define DEF_OP(x) void Arm64JITCore::Op_##x(IR::IROp_Header const *IROp, IR::NodeID Node)
 DEF_OP(ExtractElementPair) {
  auto Op = IROp->C<IR::IROp_ExtractElementPair>();
  switch (Op->Header.Size) {
@@ -68,17 +68,11 @@ DEF_OP(CreateElementPair) {
  }
 }

-DEF_OP(Mov) {
-  auto Op = IROp->C<IR::IROp_Mov>();
-  mov(GetReg<RA_64>(Node), GetReg<RA_64>(Op->Value.ID()));
-}
-
 #undef DEF_OP
 void Arm64JITCore::RegisterMoveHandlers() {
 #define REGISTER_OP(op, x) OpHandlers[FEXCore::IR::IROps::OP_##op] = &Arm64JITCore::Op_##x
  REGISTER_OP(EXTRACTELEMENTPAIR, ExtractElementPair);
  REGISTER_OP(CREATEELEMENTPAIR,  CreateElementPair);
-  REGISTER_OP(MOV,                Mov);
 #undef REGISTER_OP
 }
 }
@@ -33,7 +33,7 @@ namespace FEXCore::CPU {
 DEF_OP(SignalReturn) {
  // Adjust the stack first for a regular return
  if (SpillSlots) {
-    add(rsp, SpillSlots * 16); // + 8 to consume return address
+    add(rsp, SpillSlots * MaxSpillSlotSize); // + 8 to consume return address
  }

  jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.SignalReturnHandler)]);
@@ -42,7 +42,7 @@ DEF_OP(SignalReturn) {
 DEF_OP(CallbackReturn) {
  // Adjust the stack first for a regular return
  if (SpillSlots) {
-    add(rsp, SpillSlots * 16); // + 8 to consume return address
+    add(rsp, SpillSlots * MaxSpillSlotSize); // + 8 to consume return address
  }

  // Make sure to adjust the refcounter so we don't clear the cache now
@@ -71,7 +71,7 @@ DEF_OP(ExitFunction) {


  if (SpillSlots) {
-    add(rsp, SpillSlots * 16);
+    add(rsp, SpillSlots * MaxSpillSlotSize);
  }

  uint64_t NewRIP;
@@ -117,9 +117,9 @@ DEF_OP(ExitFunction) {

 DEF_OP(Jump) {
  const auto Op = IROp->C<IR::IROp_Jump>();
-  const auto ArgID = Op->Args(0).ID();
+  const auto Target = Op->TargetBlock.ID();

-  PendingTargetLabel = &JumpTargets.try_emplace(ArgID).first->second;
+  PendingTargetLabel = &JumpTargets.try_emplace(Target).first->second;
 }

 #define GRCMP(Node) (Op->CompareSize == 4 ? GetSrc<RA_32>(Node) : GetSrc<RA_64>(Node))
@@ -209,7 +209,7 @@ DEF_OP(Thunk) {
  if (NumPush & 1)
    sub(rsp, 8); // Align

-  mov(rdi, GetSrc<RA_64>(Op->Header.Args[0].ID()));
+  mov(rdi, GetSrc<RA_64>(Op->ArgPtr.ID()));

  auto thunkFn = ThreadState->CTX->ThunkHandler->LookupThunk(Op->ThunkNameHash);

@@ -253,7 +253,7 @@ DEF_OP(ValidateCode) {
  }
 }

-DEF_OP(RemoveThreadCodeEntry) {
+DEF_OP(ThreadRemoveCodeEntry) {
  auto NumPush = RA64.size();

  for (auto &Reg : RA64)
@@ -266,7 +266,7 @@ DEF_OP(RemoveThreadCodeEntry) {
  mov(rax, Entry); // imm64 move
  mov(rsi, rax);

-  call(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.RemoveThreadCodeEntryFromJIT)]);
+  call(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadRemoveCodeEntryFromJIT)]);

  if (NumPush & 1)
    add(rsp, 8); // Align
@@ -288,8 +288,8 @@ DEF_OP(CPUID) {
  // Result: RAX, RDX. 4xi32

  // rsi can be in the source registers, so copy argument to edx first
-  mov (edx, GetSrc<RA_32>(Op->Header.Args[1].ID()));
-  mov (esi, GetSrc<RA_32>(Op->Header.Args[0].ID()));
+  mov (edx, GetSrc<RA_32>(Op->Leaf.ID()));
+  mov (esi, GetSrc<RA_32>(Op->Function.ID()));
  mov (rdi, qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.CPUIDObj)]);

  auto NumPush = RA64.size();
@@ -322,7 +322,7 @@ void X86JITCore::RegisterBranchHandlers() {
  REGISTER_OP(SYSCALL,           Syscall);
  REGISTER_OP(THUNK,             Thunk);
  REGISTER_OP(VALIDATECODE,      ValidateCode);
-  REGISTER_OP(REMOVETHREADCODEENTRY,   RemoveThreadCodeEntry);
+  REGISTER_OP(THREADREMOVECODEENTRY,   ThreadRemoveCodeEntry);
  REGISTER_OP(CPUID,             CPUID);
 #undef REGISTER_OP
 }
@@ -17,27 +17,76 @@ namespace FEXCore::CPU {

 #define DEF_OP(x) void X86JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)
 DEF_OP(VInsGPR) {
-  auto Op = IROp->C<IR::IROp_VInsGPR>();
-  movapd(GetDst(Node), GetSrc(Op->DestVector.ID()));
+  const auto Op = IROp->C<IR::IROp_VInsGPR>();
+  const auto OpSize = IROp->Size;

-  switch (Op->Header.ElementSize) {
-    case 1: {
-      pinsrb(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()), Op->DestIdx);
-      break;
+  const auto Dst = GetDst(Node);
+  const auto DestVector = GetSrc(Op->DestVector.ID());
+
+  const auto DestIdx = Op->DestIdx;
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto ElementSizeBits = ElementSize * 8;
+  const auto Offset = ElementSizeBits * DestIdx;
+
+  constexpr auto SSEBitSize = Core::CPUState::XMM_SSE_REG_SIZE * 8;
+
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto InUpperLane = Offset >= SSEBitSize;
+
+  if (InUpperLane && !Is256Bit) {
+    LOGMAN_MSG_A_FMT("Attempt to access upper 128-bit lane in 128-bit operation! Offset={}",
+                     Offset);
+    return;
+  }
+
+  if (Is256Bit) {
+    vmovapd(ToYMM(Dst), ToYMM(DestVector));
+  } else {
+    vmovapd(Dst, DestVector);
+  }
+
+  const auto Insert = [&](const Xbyak::Xmm& reg, int index) {
+    switch (ElementSize) {
+      case 1: {
+        if (InUpperLane) {
+          index -= 16;
+        }
+        pinsrb(reg, GetSrc<RA_32>(Op->Src.ID()), index);
+        break;
+      }
+      case 2: {
+        if (InUpperLane) {
+          index -= 8;
+        }
+        pinsrw(reg, GetSrc<RA_32>(Op->Src.ID()), index);
+        break;
+      }
+      case 4: {
+        if (InUpperLane) {
+          index -= 4;
+        }
+        pinsrd(reg, GetSrc<RA_32>(Op->Src.ID()), index);
+        break;
+      }
+      case 8: {
+        if (InUpperLane) {
+          index -= 2;
+        }
+        pinsrq(reg, GetSrc<RA_64>(Op->Src.ID()), index);
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unknown Element Size: {}", ElementSize);
+        break;
    }
-    case 2: {
-      pinsrw(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()), Op->DestIdx);
-      break;
-    }
-    case 4: {
-      pinsrd(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()), Op->DestIdx);
-      break;
-    }
-    case 8: {
-      pinsrq(GetDst(Node), GetSrc<RA_64>(Op->Src.ID()), Op->DestIdx);
-      break;
-    }
-    default: LOGMAN_MSG_A_FMT("Unknown Element Size: {}", Op->Header.ElementSize); break;
+  };
+
+  if (InUpperLane) {
+    vextracti128(xmm15, ToYMM(Dst), 1);
+    Insert(xmm15, DestIdx);
+    vinserti128(ToYMM(Dst), ToYMM(Dst), xmm15, 1);
+  } else {
+    Insert(Dst, DestIdx);
  }
 }

@@ -45,56 +94,64 @@ DEF_OP(VCastFromGPR) {
  auto Op = IROp->C<IR::IROp_VCastFromGPR>();
  switch (Op->Header.ElementSize) {
    case 1:
-      movzx(rax, GetSrc<RA_8>(Op->Header.Args[0].ID()));
+      movzx(rax, GetSrc<RA_8>(Op->Src.ID()));
      vmovq(GetDst(Node), rax);
      break;
    case 2:
-      movzx(rax, GetSrc<RA_16>(Op->Header.Args[0].ID()));
+      movzx(rax, GetSrc<RA_16>(Op->Src.ID()));
      vmovq(GetDst(Node), rax);
      break;
    case 4:
-      vmovd(GetDst(Node), GetSrc<RA_32>(Op->Header.Args[0].ID()).cvt32());
+      vmovd(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()).cvt32());
      break;
    case 8:
-      vmovq(GetDst(Node), GetSrc<RA_64>(Op->Header.Args[0].ID()).cvt64());
+      vmovq(GetDst(Node), GetSrc<RA_64>(Op->Src.ID()).cvt64());
      break;
    default: LOGMAN_MSG_A_FMT("Unknown VCastFromGPR element size: {}", Op->Header.ElementSize);
  }
 }

 DEF_OP(Float_FromGPR_S) {
-  auto Op = IROp->C<IR::IROp_Float_FromGPR_S>();
-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const auto Op = IROp->C<IR::IROp_Float_FromGPR_S>();
+
+  const uint16_t ElementSize = Op->Header.ElementSize;
+  const uint16_t Conv = (ElementSize << 8) | Op->SrcElementSize;
+
  switch (Conv) {
    case 0x0404: { // Float <- int32_t
-      cvtsi2ss(GetDst(Node), GetSrc<RA_32>(Op->Header.Args[0].ID()));
+      cvtsi2ss(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()));
      break;
    }
    case 0x0408: { // Float <- int64_t
-      cvtsi2ss(GetDst(Node), GetSrc<RA_64>(Op->Header.Args[0].ID()));
+      cvtsi2ss(GetDst(Node), GetSrc<RA_64>(Op->Src.ID()));
      break;
    }
    case 0x0804: { // Double <- int32_t
-      cvtsi2sd(GetDst(Node), GetSrc<RA_32>(Op->Header.Args[0].ID()));
+      cvtsi2sd(GetDst(Node), GetSrc<RA_32>(Op->Src.ID()));
      break;
    }
    case 0x0808: { // Double <- int64_t
-      cvtsi2sd(GetDst(Node), GetSrc<RA_64>(Op->Header.Args[0].ID()));
+      cvtsi2sd(GetDst(Node), GetSrc<RA_64>(Op->Src.ID()));
      break;
    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled conversion mask: Mask=0x{:04x}, ElementSize={}, SrcElementSize={}",
+                       Conv, ElementSize, Op->SrcElementSize);
+      break;
  }
 }

 DEF_OP(Float_FToF) {
  auto Op = IROp->C<IR::IROp_Float_FToF>();
-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+
  switch (Conv) {
    case 0x0804: { // Double <- Float
-      cvtss2sd(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
+      cvtss2sd(GetDst(Node), GetSrc(Op->Scalar.ID()));
      break;
    }
    case 0x0408: { // Float <- Double
-      cvtsd2ss(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
+      cvtsd2ss(GetDst(Node), GetSrc(Op->Scalar.ID()));
      break;
    }
    default: LOGMAN_MSG_A_FMT("Unknown Float_FToF sizes: 0x{:x}", Conv);
@@ -102,99 +159,194 @@ DEF_OP(Float_FToF) {
 }

 DEF_OP(Vector_SToF) {
-  auto Op = IROp->C<IR::IROp_Vector_SToF>();
-  switch (Op->Header.ElementSize) {
+  const auto Op = IROp->C<IR::IROp_Vector_SToF>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  switch (ElementSize) {
    case 4:
-      cvtdq2ps(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
-    break;
+      if (Is256Bit) {
+        vcvtdq2ps(ToYMM(Dst), ToYMM(Vector));
+      } else {
+        vcvtdq2ps(Dst, Vector);
+      }
+      break;
    case 8:
      // This operation is a bit disgusting in x86
      // There is no vector form of this instruction until AVX512VL + AVX512DQ (vcvtqq2pd)
      // 1) First extract the top 64bits
      // 2) Do a scalar conversion on each
      // 3) Make sure to merge them together at the end
-      pextrq(rax, GetSrc(Op->Header.Args[0].ID()), 1);
-      pextrq(rcx, GetSrc(Op->Header.Args[0].ID()), 0);
-      cvtsi2sd(GetDst(Node), rcx);
+      pextrq(rax, Vector, 1);
+      pextrq(rcx, Vector, 0);
+      cvtsi2sd(Dst, rcx);
      cvtsi2sd(xmm15, rax);
-      movlhps(GetDst(Node), xmm15);
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_SToF element size: {}", Op->Header.ElementSize);
+      vmovlhps(Dst, Dst, xmm15);
+      if (Is256Bit) {
+        vextracti128(xmm15, ToYMM(Vector), 1);
+
+        pextrq(rax, xmm15, 1);
+        pextrq(rcx, xmm15, 0);
+        cvtsi2sd(xmm15, rcx);
+        cvtsi2sd(xmm14, rax);
+        movlhps(xmm15, xmm14);
+
+        vinserti128(ToYMM(Dst), ToYMM(Dst), xmm15, 1);
+      }
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Vector_SToF element size: {}", ElementSize);
+      break;
  }
 }

 DEF_OP(Vector_FToZS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToZS>();
-  switch (Op->Header.ElementSize) {
+  const auto Op = IROp->C<IR::IROp_Vector_FToZS>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  switch (ElementSize) {
    case 4:
-      cvttps2dq(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
-    break;
+      if (Is256Bit) {
+        vcvttps2dq(ToYMM(Dst), ToYMM(Vector));
+      } else {
+        vcvttps2dq(Dst, Vector);
+      }
+      break;
    case 8:
-      cvttpd2dq(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToZS element size: {}", Op->Header.ElementSize);
+      if (Is256Bit) {
+        vcvttpd2dq(ToYMM(Dst), ToYMM(Vector));
+      } else {
+        vcvttpd2dq(Dst, Vector);
+      }
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Vector_FToZS element size: {}", ElementSize);
+      break;
  }
 }

 DEF_OP(Vector_FToS) {
-  auto Op = IROp->C<IR::IROp_Vector_FToS>();
-  switch (Op->Header.ElementSize) {
+  const auto Op = IROp->C<IR::IROp_Vector_FToS>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  switch (ElementSize) {
    case 4:
-      cvtps2dq(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
-    break;
+      if (Is256Bit) {
+        vcvtps2dq(ToYMM(Dst), ToYMM(Vector));
+      } else {
+        vcvtps2dq(Dst, Vector);
+      }
+      break;
    case 8:
-      cvtpd2dq(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
-    break;
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToS element size: {}", Op->Header.ElementSize);
+      if (Is256Bit) {
+        vcvtpd2dq(ToYMM(Dst), ToYMM(Vector));
+      } else {
+        vcvtpd2dq(Dst, Vector);
+      }
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Vector_FToS element size: {}", ElementSize);
+      break;
  }
 }

 DEF_OP(Vector_FToF) {
-  auto Op = IROp->C<IR::IROp_Vector_FToF>();
-  uint16_t Conv = (Op->Header.ElementSize << 8) | Op->SrcElementSize;
+  const auto Op = IROp->C<IR::IROp_Vector_FToF>();
+  const auto OpSize = IROp->Size;
+
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto Conv = (ElementSize << 8) | Op->SrcElementSize;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());

  switch (Conv) {
    case 0x0804: { // Double <- Float
-      cvtps2pd(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
+      if (Is256Bit) {
+        vcvtps2pd(ToYMM(Dst), Vector);
+      } else {
+        vcvtps2pd(Dst, Vector);
+      }
      break;
    }
    case 0x0408: { // Float <- Double
-      cvtpd2ps(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
+      if (Is256Bit) {
+        vcvtpd2ps(Dst, ToYMM(Vector));
+      } else {
+        vcvtpd2ps(Dst, Vector);
+      }
      break;
    }
-    default: LOGMAN_MSG_A_FMT("Unknown Vector_FToF conversion type : 0x{:04x}", Conv); break;
+    default:
+      LOGMAN_MSG_A_FMT("Unknown Vector_FToF conversion type : 0x{:04x}", Conv);
+      break;
  }
 }

 DEF_OP(Vector_FToI) {
-  auto Op = IROp->C<IR::IROp_Vector_FToI>();
-  uint8_t RoundMode{};
+  const auto Op = IROp->C<IR::IROp_Vector_FToI>();
+  const auto OpSize = IROp->Size;

-  switch (Op->Round) {
-    case FEXCore::IR::Round_Nearest.Val:
-      RoundMode = 0b0000'0'0'00;
-    break;
-    case FEXCore::IR::Round_Negative_Infinity.Val:
-      RoundMode = 0b0000'0'0'01;
-    break;
-    case FEXCore::IR::Round_Positive_Infinity.Val:
-      RoundMode = 0b0000'0'0'10;
-    break;
-    case FEXCore::IR::Round_Towards_Zero.Val:
-      RoundMode = 0b0000'0'0'11;
-    break;
-    case FEXCore::IR::Round_Host.Val:
-      RoundMode = 0b0000'0'1'00;
-    break;
-  }
+  const uint8_t RoundMode = [Op] {
+    switch (Op->Round) {
+      case FEXCore::IR::Round_Nearest.Val:
+        return 0b0000'0'0'00;
+      case FEXCore::IR::Round_Negative_Infinity.Val:
+        return 0b0000'0'0'01;
+      case FEXCore::IR::Round_Positive_Infinity.Val:
+        return 0b0000'0'0'10;
+      case FEXCore::IR::Round_Towards_Zero.Val:
+        return 0b0000'0'0'11;
+      case FEXCore::IR::Round_Host.Val:
+        return 0b0000'0'1'00;
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled rounding mode");
+        return 0;
+    }
+  }();

-  switch (Op->Header.ElementSize) {
+  const auto ElementSize = Op->Header.ElementSize;
+  const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  const auto Dst = GetDst(Node);
+  const auto Vector = GetSrc(Op->Vector.ID());
+
+  switch (ElementSize) {
    case 4:
-      roundps(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), RoundMode);
-    break;
+      if (Is256Bit) {
+        vroundps(ToYMM(Dst), ToYMM(Vector), RoundMode);
+      } else {
+        vroundps(Dst, Vector, RoundMode);
+      }
+      break;
    case 8:
-      roundpd(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), RoundMode);
-    break;
+      if (Is256Bit) {
+        vroundpd(ToYMM(Dst), ToYMM(Vector), RoundMode);
+      } else {
+        vroundpd(Dst, Vector, RoundMode);
+      }
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled element size: {}", ElementSize);
+      break;
  }
 }

@@ -17,44 +17,44 @@ namespace FEXCore::CPU {

 DEF_OP(AESImc) {
  auto Op = IROp->C<IR::IROp_VAESImc>();
-  vaesimc(GetDst(Node), GetSrc(Op->Header.Args[0].ID()));
+  vaesimc(GetDst(Node), GetSrc(Op->Vector.ID()));
 }

 DEF_OP(AESEnc) {
  auto Op = IROp->C<IR::IROp_VAESEnc>();
-  vaesenc(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), GetSrc(Op->Header.Args[1].ID()));
+  vaesenc(GetDst(Node), GetSrc(Op->State.ID()), GetSrc(Op->Key.ID()));
 }

 DEF_OP(AESEncLast) {
  auto Op = IROp->C<IR::IROp_VAESEncLast>();
-  vaesenclast(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), GetSrc(Op->Header.Args[1].ID()));
+  vaesenclast(GetDst(Node), GetSrc(Op->State.ID()), GetSrc(Op->Key.ID()));
 }

 DEF_OP(AESDec) {
  auto Op = IROp->C<IR::IROp_VAESDec>();
-  vaesdec(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), GetSrc(Op->Header.Args[1].ID()));
+  vaesdec(GetDst(Node), GetSrc(Op->State.ID()), GetSrc(Op->Key.ID()));
 }

 DEF_OP(AESDecLast) {
  auto Op = IROp->C<IR::IROp_VAESDecLast>();
-  vaesdeclast(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), GetSrc(Op->Header.Args[1].ID()));
+  vaesdeclast(GetDst(Node), GetSrc(Op->State.ID()), GetSrc(Op->Key.ID()));
 }

 DEF_OP(AESKeyGenAssist) {
  auto Op = IROp->C<IR::IROp_VAESKeyGenAssist>();
-  vaeskeygenassist(GetDst(Node), GetSrc(Op->Header.Args[0].ID()), Op->RCON);
+  vaeskeygenassist(GetDst(Node), GetSrc(Op->Src.ID()), Op->RCON);
 }

 DEF_OP(CRC32) {
  auto Op = IROp->C<IR::IROp_CRC32>();
  switch (IROp->Size) {
  case 4:
-    mov(TMP1, GetSrc<RA_32>(Op->Header.Args[1].ID()));
-    mov(GetDst<RA_32>(Node), GetSrc<RA_32>(Op->Header.Args[0].ID()));
+    mov(TMP1, GetSrc<RA_32>(Op->Src2.ID()));
+    mov(GetDst<RA_32>(Node), GetSrc<RA_32>(Op->Src1.ID()));
  break;
  case 8:
-    mov(TMP1, GetSrc<RA_64>(Op->Header.Args[1].ID()));
-    mov(GetDst<RA_64>(Node), GetSrc<RA_64>(Op->Header.Args[0].ID()));
+    mov(TMP1, GetSrc<RA_64>(Op->Src2.ID()));
+    mov(GetDst<RA_64>(Node), GetSrc<RA_64>(Op->Src1.ID()));
  break;
  default: LOGMAN_MSG_A_FMT("Unknown CRC32 size: {}", IROp->Size);
  }
@@ -18,7 +18,7 @@ namespace FEXCore::CPU {
 DEF_OP(GetHostFlag) {
  auto Op = IROp->C<IR::IROp_GetHostFlag>();

-  mov(rax, GetSrc<RA_64>(Op->Header.Args[0].ID()));
+  mov(rax, GetSrc<RA_64>(Op->Value.ID()));
  shr(rax, Op->Flag);
  and_(rax, 1);
  mov(GetDst<RA_64>(Node), rax);
@@ -25,7 +25,9 @@ $end_info$
 #include <FEXCore/IR/IntrusiveIRList.h>
 #include <FEXCore/IR/RegisterAllocationData.h>
 #include <FEXCore/Utils/Allocator.h>
+#include <FEXCore/Utils/EnumUtils.h>
 #include <FEXCore/Utils/LogManager.h>
+#include <FEXCore/Utils/Profiler.h>

 #include <algorithm>
 #include <array>
@@ -59,32 +61,42 @@ static void PrintVectorValue(uint64_t Value, uint64_t ValueUpper) {
 namespace FEXCore::CPU {

 void X86JITCore::PushRegs() {
-  sub(rsp, 16 * RAXMM_x.size());
+  const auto AVXRegSize = Core::CPUState::XMM_AVX_REG_SIZE;
+
+  sub(rsp, AVXRegSize * RAXMM_x.size());
  for (size_t i = 0; i < RAXMM_x.size(); ++i) {
-    movaps(ptr[rsp + i * 16], RAXMM_x[i]);
+    vmovups(ptr[rsp + i * AVXRegSize], ToYMM(RAXMM_x[i]));
  }

-  for (auto &Reg : RA64)
+  for (const auto &Reg : RA64) {
    push(Reg);
+  }

-  auto NumPush = RA64.size();
-  if (NumPush & 1)
-    sub(rsp, 8); // Align
+  const auto NumPush = RA64.size();
+  if ((NumPush & 1) != 0) {
+    // Align
+    sub(rsp, 8);
+  }
 }

 void X86JITCore::PopRegs() {
-  auto NumPush = RA64.size();
+  const auto AVXRegSize = Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto NumPush = RA64.size();

-  if (NumPush & 1)
-    add(rsp, 8); // Align
-  for (uint32_t i = RA64.size(); i > 0; --i)
-    pop(RA64[i - 1]);
-
-  for (size_t i = 0; i < RAXMM_x.size(); ++i) {
-    movaps(RAXMM_x[i], ptr[rsp + i * 16]);
+  if ((NumPush & 1) != 0) {
+    // Align
+    add(rsp, 8);
  }

-  add(rsp, 16 * RAXMM_x.size());
+  for (uint32_t i = RA64.size(); i > 0; --i) {
+    pop(RA64[i - 1]);
+  }
+
+  for (size_t i = 0; i < RAXMM_x.size(); ++i) {
+    vmovups(ToYMM(RAXMM_x[i]), ptr[rsp + i * AVXRegSize]);
+  }
+
+  add(rsp, AVXRegSize * RAXMM_x.size());
 }

 void X86JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
@@ -293,7 +305,8 @@ void X86JITCore::Op_Unhandled(IR::IROp_Header *IROp, IR::NodeID Node) {
      case FABI_UNKNOWN:
      default:
 #if defined(ASSERTIONS_ENABLED) && ASSERTIONS_ENABLED
-        LOGMAN_MSG_A_FMT("Unhandled IR Fallback ABI: {} {}", FEXCore::IR::GetName(IROp->Op), Info.ABI);
+        LOGMAN_MSG_A_FMT("Unhandled IR Fallback ABI: {} {}",
+                         IR::GetName(IROp->Op), ToUnderlying(Info.ABI));
 #endif
        break;
    }
@@ -312,7 +325,7 @@ static uint64_t X86JITCore_ExitFunctionLink(FEXCore::Core::CpuStateFrame *Frame,
  }

  auto LinkerAddress = Frame->Pointers.Common.ExitFunctionLinker;
-  Thread->LookupCache->AddBlockLink(GuestRip, (uintptr_t)record, [record, LinkerAddress]{
+  Context::Context::ThreadAddBlockLink(Thread, GuestRip, (uintptr_t)record, [record, LinkerAddress]{
    // undo the link
    record[0] = LinkerAddress;
  });
@@ -358,10 +371,10 @@ X86JITCore::X86JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::InternalTh

  {
    auto &Common = ThreadState->CurrentFrame->Pointers.Common;
-    
+
    Common.PrintValue = reinterpret_cast<uint64_t>(PrintValue);
    Common.PrintVectorValue = reinterpret_cast<uint64_t>(PrintVectorValue);
-    Common.RemoveThreadCodeEntryFromJIT = reinterpret_cast<uintptr_t>(&Context::Context::RemoveThreadCodeEntryFromJit);
+    Common.ThreadRemoveCodeEntryFromJIT = reinterpret_cast<uintptr_t>(&Context::Context::ThreadRemoveCodeEntryFromJit);
    Common.CPUIDObj = reinterpret_cast<uint64_t>(&CTX->CPUID);

    {
@@ -371,7 +384,7 @@ X86JITCore::X86JITCore(FEXCore::Context::Context *ctx, FEXCore::Core::InternalTh

    Common.SyscallHandlerObj = reinterpret_cast<uint64_t>(CTX->SyscallHandler);
    Common.SyscallHandlerFunc = reinterpret_cast<uint64_t>(FEXCore::Context::HandleSyscall);
-    Common.ExitFunctionLink = reinterpret_cast<uintptr_t>(&X86JITCore_ExitFunctionLink);
+    Common.ExitFunctionLink = reinterpret_cast<uintptr_t>(&Context::Context::ThreadExitFunctionLink<X86JITCore_ExitFunctionLink>);

    // Fill in the fallback handlers
    InterpreterOps::FillFallbackIndexPointers(Common.FallbackHandlerPointers);
@@ -407,7 +420,7 @@ void X86JITCore::ClearCache() {
 IR::PhysicalRegister X86JITCore::GetPhys(IR::NodeID Node) const {
  auto PhyReg = RAData->GetNodeRegister(Node);

-  LOGMAN_THROW_A_FMT(PhyReg.Raw != 255, "Couldn't Allocate register for node: ssa{}. Class: {}", Node, PhyReg.Class);
+  LOGMAN_THROW_AA_FMT(PhyReg.Raw != 255, "Couldn't Allocate register for node: ssa{}. Class: {}", Node, PhyReg.Class);

  return PhyReg;
 }
@@ -570,6 +583,8 @@ std::tuple<X86JITCore::SetCC, X86JITCore::CMovCC, X86JITCore::JCC> X86JITCore::G
 }

 void *X86JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IRListView const *IR, [[maybe_unused]] FEXCore::Core::DebugData *DebugData, FEXCore::IR::RegisterAllocationData *RAData, bool GDBEnabled) {
+
+  FEXCORE_PROFILE_SCOPED("x86::CompileCode");
  JumpTargets.clear();
  uint32_t SSACount = IR->GetSSACount();

@@ -592,12 +607,12 @@ void *X86JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IRLi
    setSize(getSize() + GDBSize);
  }

-  LOGMAN_THROW_A_FMT(RAData != nullptr, "Needs RA");
+  LOGMAN_THROW_AA_FMT(RAData != nullptr, "Needs RA");

  SpillSlots = RAData->SpillSlots();

  if (SpillSlots) {
-    sub(rsp, SpillSlots * 16);
+    sub(rsp, SpillSlots * MaxSpillSlotSize);
  }

 #ifdef BLOCKSTATS
@@ -649,7 +664,7 @@ void *X86JITCore::CompileCode(uint64_t Entry, [[maybe_unused]] FEXCore::IR::IRLi
    using namespace FEXCore::IR;
 #if defined(ASSERTIONS_ENABLED) && ASSERTIONS_ENABLED
    auto BlockIROp = BlockHeader->CW<IROp_CodeBlock>();
-    LOGMAN_THROW_A_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");
+    LOGMAN_THROW_AA_FMT(BlockIROp->Header.Op == IR::OP_CODEBLOCK, "IR type failed to be a code block");
 #endif

    auto BlockStartHostCode = getCurr<uint8_t *>();
@@ -181,6 +181,10 @@ private:
  [[nodiscard]] Xbyak::Xmm GetSrc(IR::NodeID Node) const;
  [[nodiscard]] Xbyak::Xmm GetDst(IR::NodeID Node) const;

+  [[nodiscard]] static Xbyak::Ymm ToYMM(const Xbyak::Xmm& xmm) {
+    return Xbyak::Ymm{xmm.getIdx()};
+  }
+
  [[nodiscard]] Xbyak::RegExp GenerateModRM(Xbyak::Reg Base, IR::OrderedNodeWrapper Offset,
                                            IR::MemOffsetType OffsetType, uint8_t OffsetScale) const;

@@ -205,7 +209,7 @@ private:
  * @brief Current guest RIP entrypoint
  */
  uint8_t *GuestEntry{};
-  
+
  using SetCC = void (X86JITCore::*)(const Operand& op);
  using CMovCC = void (X86JITCore::*)(const Reg& reg, const Operand& op);
  using JCC = void (X86JITCore::*)(const Label& label, LabelType type);
@@ -312,7 +316,7 @@ private:
  DEF_OP(Syscall);
  DEF_OP(Thunk);
  DEF_OP(ValidateCode);
-  DEF_OP(RemoveThreadCodeEntry);
+  DEF_OP(ThreadRemoveCodeEntry);
  DEF_OP(CPUID);

  ///< Conversion ops
@@ -333,6 +337,8 @@ private:
  ///< Memory ops
  DEF_OP(LoadContext);
  DEF_OP(StoreContext);
+  DEF_OP(LoadRegister);
+  DEF_OP(StoreRegister);
  DEF_OP(LoadContextIndexed);
  DEF_OP(StoreContextIndexed);
  DEF_OP(SpillRegister);
@@ -341,8 +347,6 @@ private:
  DEF_OP(StoreFlag);
  DEF_OP(LoadMem);
  DEF_OP(StoreMem);
-  DEF_OP(VLoadMemElement);
-  DEF_OP(VStoreMemElement);
  DEF_OP(CacheLineClear);
  DEF_OP(CacheLineZero);

@@ -362,12 +366,10 @@ private:
  ///< Move ops
  DEF_OP(ExtractElementPair);
  DEF_OP(CreateElementPair);
-  DEF_OP(Mov);

  ///< Vector ops
  DEF_OP(VectorZero);
  DEF_OP(VectorImm);
-  DEF_OP(SplatVector);
  DEF_OP(VMov);
  DEF_OP(VAnd);
  DEF_OP(VBic);
@@ -426,18 +428,13 @@ private:
  DEF_OP(VUShrS);
  DEF_OP(VSShrS);
  DEF_OP(VInsElement);
-  DEF_OP(VInsScalarElement);
-  DEF_OP(VExtractElement);
  DEF_OP(VDupElement);
  DEF_OP(VExtr);
-  DEF_OP(VSLI);
-  DEF_OP(VSRI);
  DEF_OP(VUShrI);
  DEF_OP(VSShrI);
  DEF_OP(VShlI);
  DEF_OP(VUShrNI);
  DEF_OP(VUShrNI2);
-  DEF_OP(VBitcast);
  DEF_OP(VSXTL);
  DEF_OP(VSXTL2);
  DEF_OP(VUXTL);
@@ -21,130 +21,266 @@ namespace FEXCore::CPU {
 #define DEF_OP(x) void X86JITCore::Op_##x(IR::IROp_Header *IROp, IR::NodeID Node)

 DEF_OP(LoadContext) {
-  auto Op = IROp->C<IR::IROp_LoadContext>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_LoadContext>();
+  const auto OpSize = IROp->Size;

  if (Op->Class == IR::GPRClass) {
    switch (OpSize) {
    case 1: {
      movzx(GetDst<RA_32>(Node), byte [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 2: {
      movzx(GetDst<RA_32>(Node), word [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 4: {
      mov(GetDst<RA_32>(Node), dword [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 8: {
      mov(GetDst<RA_64>(Node), qword [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 16: {
      LOGMAN_MSG_A_FMT("Invalid GPR load of size 16");
+      break;
    }
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+      break;
    }
  }
  else {
+    const auto Dst = GetDst(Node);
+
    switch (OpSize) {
    case 1: {
      movzx(rax, byte [STATE + Op->Offset]);
-      vmovq(GetDst(Node), rax);
+      vmovq(Dst, rax);
+      break;
    }
-    break;
    case 2: {
      movzx(rax, word [STATE + Op->Offset]);
-      vmovq(GetDst(Node), rax);
+      vmovq(Dst, rax);
+      break;
    }
-    break;
    case 4: {
-      vmovd(GetDst(Node), dword [STATE + Op->Offset]);
+      vmovd(Dst, dword [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 8: {
-      vmovq(GetDst(Node), qword [STATE + Op->Offset]);
+      vmovq(Dst, qword [STATE + Op->Offset]);
+      break;
    }
-    break;
    case 16: {
-      if (Op->Offset % 16 == 0)
-        movaps(GetDst(Node), xword [STATE + Op->Offset]);
-      else
-        movups(GetDst(Node), xword [STATE + Op->Offset]);
+      if (Op->Offset % 16 == 0) {
+        vmovaps(Dst, xword [STATE + Op->Offset]);
+      } else {
+        vmovups(Dst, xword [STATE + Op->Offset]);
+      }
+      break;
    }
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+    case 32: {
+      vmovups(ToYMM(Dst), yword [STATE + Op->Offset]);
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadContext size: {}", OpSize);
+      break;
    }
  }
 }

 DEF_OP(StoreContext) {
-  auto Op = IROp->C<IR::IROp_StoreContext>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_StoreContext>();
+  const auto OpSize = IROp->Size;

  if (Op->Class == IR::GPRClass) {
    switch (OpSize) {
    case 1: {
      mov(byte [STATE + Op->Offset], GetSrc<RA_8>(Op->Value.ID()));
+      break;
    }
-    break;
-
    case 2: {
      mov(word [STATE + Op->Offset], GetSrc<RA_16>(Op->Value.ID()));
+      break;
    }
-    break;
    case 4: {
      mov(dword [STATE + Op->Offset], GetSrc<RA_32>(Op->Value.ID()));
+      break;
    }
-    break;
    case 8: {
      mov(qword [STATE + Op->Offset], GetSrc<RA_64>(Op->Value.ID()));
+      break;
    }
-    break;
-    case 16:
-      LogMan::Msg::DFmt("Invalid store size of 16");
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled StoreContext size: {}", OpSize);
+    case 16: {
+      LOGMAN_MSG_A_FMT("Invalid store size of 16");
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled StoreContext size: {}", OpSize);
+      break;
    }
  }
  else {
+    const auto Value = GetSrc(Op->Value.ID());
+
    switch (OpSize) {
    case 1: {
-      pextrb(byte [STATE + Op->Offset], GetSrc(Op->Value.ID()), 0);
+      pextrb(byte [STATE + Op->Offset], Value, 0);
+      break;
    }
-    break;
-
    case 2: {
-      pextrw(word [STATE + Op->Offset], GetSrc(Op->Value.ID()), 0);
+      pextrw(word [STATE + Op->Offset], Value, 0);
+      break;
    }
-    break;
    case 4: {
-      vmovd(dword [STATE + Op->Offset], GetSrc(Op->Value.ID()));
+      vmovd(dword [STATE + Op->Offset], Value);
+      break;
    }
-    break;
    case 8: {
-      vmovq(qword [STATE + Op->Offset], GetSrc(Op->Value.ID()));
+      vmovq(qword [STATE + Op->Offset], Value);
+      break;
    }
-    break;
    case 16: {
-      if (Op->Offset % 16 == 0)
-        movaps(xword [STATE + Op->Offset], GetSrc(Op->Value.ID()));
-      else
-        movups(xword [STATE + Op->Offset], GetSrc(Op->Value.ID()));
+      if (Op->Offset % 16 == 0) {
+        vmovaps(xword [STATE + Op->Offset], Value);
+      } else {
+        vmovups(xword [STATE + Op->Offset], Value);
+      }
+      break;
    }
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled StoreContext size: {}", OpSize);
+    case 32: {
+      vmovups(yword [STATE + Op->Offset], ToYMM(Value));
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled StoreContext size: {}", OpSize);
+      break;
    }
  }
 }

+DEF_OP(LoadRegister) {
+  const auto Op = IROp->C<IR::IROp_LoadRegister>();
+  const auto OpSize = IROp->Size;
+
+  if (Op->Class == IR::GPRClass) {
+    switch (OpSize) {
+    case 1: {
+      movzx(GetSrc<RA_32>(Node), byte [STATE + Op->Offset]);
+      break;
+    }
+    case 2: {
+      movzx(GetSrc<RA_32>(Node), word [STATE + Op->Offset]);
+      break;
+    }
+    case 4: {
+      mov(GetSrc<RA_32>(Node), dword [STATE + Op->Offset]);
+      break;
+    }
+    case 8: {
+      mov(GetSrc<RA_64>(Node), qword [STATE + Op->Offset]);
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadRegister size: {}", OpSize);
+      break;
+    }
+  }
+  else {
+    const auto Dst = GetSrc(Node);
+
+    switch (OpSize) {
+    case 1: {
+      movzx(rax, byte [STATE + Op->Offset]);
+      vmovq(Dst, rax);
+      break;
+    }
+    case 2: {
+      movzx(rax, word [STATE + Op->Offset]);
+      vmovq(Dst, rax);
+      break;
+    }
+    case 4: {
+      vmovd(Dst, dword [STATE + Op->Offset]);
+      break;
+    }
+    case 8: {
+      vmovq(Dst, qword [STATE + Op->Offset]);
+      break;
+    }
+    case 16: {
+      if (Op->Offset % 16 == 0) {
+        vmovaps(Dst, xword [STATE + Op->Offset]);
+      } else {
+        vmovups(Dst, xword [STATE + Op->Offset]);
+      }
+      break;
+    }
+    case 32: {
+      vmovups(ToYMM(Dst), yword [STATE + Op->Offset]);
+      break;
+    }
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled LoadRegister size: {}", OpSize);
+      break;
+    }
+  }
+}
+
+DEF_OP(StoreRegister) {
+  const auto Op = IROp->C<IR::IROp_StoreRegister>();
+  const auto OpSize = IROp->Size;
+
+  if (Op->Class == IR::GPRClass) {
+    const auto regOffs = Op->Offset & 7;
+
+    switch (OpSize) {
+      case 4:
+        LOGMAN_THROW_AA_FMT(regOffs == 0, "unexpected regOffs");
+        mov(dword [STATE + Op->Offset], GetSrc<RA_32>(Op->Value.ID()));
+        break;
+
+      case 8:
+        LOGMAN_THROW_AA_FMT(regOffs == 0, "unexpected regOffs");
+        mov(qword [STATE + Op->Offset], GetSrc<RA_64>(Op->Value.ID()));
+        break;
+
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled StoreRegister GPR size: {}", OpSize);
+        break;
+    }
+  } else if (Op->Class == IR::FPRClass) {
+    const auto Value = GetSrc(Op->Value.ID());
+    switch (OpSize) {
+      case 16: {
+        if (Op->Offset % 16 == 0) {
+          vmovaps(xword [STATE + Op->Offset], Value);
+        } else {
+          vmovups(xword [STATE + Op->Offset], Value);
+        }
+        break;
+      }
+      case 32: {
+        vmovups(yword [STATE + Op->Offset], ToYMM(Value));
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled StoreContext size: {}", OpSize);
+        break;
+    }
+  } else {
+    LOGMAN_THROW_AA_FMT(false, "Unhandled Op->Class {}", Op->Class);
+  }
+}
+
 DEF_OP(LoadContextIndexed) {
-  auto Op = IROp->C<IR::IROp_LoadContextIndexed>();
-  size_t size = IROp->Size;
-  Reg index = GetSrc<RA_64>(Op->Index.ID());
+  const auto Op = IROp->C<IR::IROp_LoadContextIndexed>();
+  const auto OpSize = IROp->Size;
+
+  const Reg Index = GetSrc<RA_64>(Op->Index.ID());

  if (Op->Class == IR::GPRClass) {
    switch (Op->Stride) {
@@ -153,21 +289,21 @@ DEF_OP(LoadContextIndexed) {
    case 4:
    case 8: {
      lea(rax, dword [STATE + Op->BaseOffset]);
-      switch (size) {
+      switch (OpSize) {
      case 1:
-        movzx(GetDst<RA_32>(Node), byte [rax + index * Op->Stride]);
+        movzx(GetDst<RA_32>(Node), byte [rax + Index * Op->Stride]);
        break;
      case 2:
-        movzx(GetDst<RA_32>(Node), word [rax + index * Op->Stride]);
+        movzx(GetDst<RA_32>(Node), word [rax + Index * Op->Stride]);
        break;
      case 4:
-        mov(GetDst<RA_32>(Node),  dword [rax + index * Op->Stride]);
+        mov(GetDst<RA_32>(Node),  dword [rax + Index * Op->Stride]);
        break;
      case 8:
-        mov(GetDst<RA_64>(Node),  qword [rax + index * Op->Stride]);
+        mov(GetDst<RA_64>(Node),  qword [rax + Index * Op->Stride]);
        break;
      default:
-        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", IROp->Size);
+        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", OpSize);
        break;
      }
      break;
@@ -186,53 +322,63 @@ DEF_OP(LoadContextIndexed) {
    case 2:
    case 4:
    case 8: {
+      const auto Dst = GetDst(Node);
+
      lea(rax, dword [STATE + Op->BaseOffset]);
-      switch (size) {
+      switch (OpSize) {
      case 1:
-        movzx(eax, byte [rax + index * Op->Stride]);
-        vmovd(GetDst(Node), eax);
+        movzx(eax, byte [rax + Index * Op->Stride]);
+        vmovd(Dst, eax);
        break;
      case 2:
-        movzx(eax, word [rax + index * Op->Stride]);
-        vmovd(GetDst(Node), eax);
+        movzx(eax, word [rax + Index * Op->Stride]);
+        vmovd(Dst, eax);
        break;
      case 4:
-        vmovd(GetDst(Node),  dword [rax + index * Op->Stride]);
+        vmovd(Dst,  dword [rax + Index * Op->Stride]);
        break;
      case 8:
-        vmovq(GetDst(Node),  qword [rax + index * Op->Stride]);
+        vmovq(Dst,  qword [rax + Index * Op->Stride]);
        break;
      default:
-        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", IROp->Size);
+        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", OpSize);
        break;
      }
      break;
    }
-    case 16: {
-      mov(rax, index);
-      shl(rax, 4);
+    case 16:
+    case 32: {
+      const auto Dst = GetDst(Node);
+      const auto Shift = Op->Stride == 16 ? 4 : 5;
+
+      mov(rax, Index);
+      shl(rax, Shift);
      lea(rax, dword [rax + Op->BaseOffset]);
-      switch (size) {
+      switch (OpSize) {
      case 1:
-        pinsrb(GetDst(Node), byte [STATE + rax], 0);
+        pinsrb(Dst, byte [STATE + rax], 0);
        break;
      case 2:
-        pinsrw(GetDst(Node), word [STATE + rax], 0);
+        pinsrw(Dst, word [STATE + rax], 0);
        break;
      case 4:
-        vmovd(GetDst(Node), dword [STATE + rax]);
+        vmovd(Dst, dword [STATE + rax]);
        break;
      case 8:
-        vmovq(GetDst(Node), qword [STATE + rax]);
+        vmovq(Dst, qword [STATE + rax]);
        break;
      case 16:
-        if (Op->BaseOffset % 16 == 0)
-          movaps(GetDst(Node), xword [STATE + rax]);
-        else
-          movups(GetDst(Node), xword [STATE + rax]);
+        if (Op->BaseOffset % 16 == 0) {
+          vmovaps(Dst, xword [STATE + rax]);
+        } else {
+          vmovups(Dst, xword [STATE + rax]);
+        }
+        break;
+      case 32:
+        vmovups(ToYMM(Dst), yword [STATE + rax]);
        break;
      default:
-        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", IROp->Size);
+        LOGMAN_MSG_A_FMT("Unhandled LoadContextIndexed size: {}", OpSize);
        break;
      }
      break;
@@ -245,12 +391,13 @@ DEF_OP(LoadContextIndexed) {
 }

 DEF_OP(StoreContextIndexed) {
-  auto Op = IROp->C<IR::IROp_StoreContextIndexed>();
-  Reg index = GetSrc<RA_64>(Op->Index.ID());
-  size_t size = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_StoreContextIndexed>();
+  const auto OpSize = IROp->Size;
+
+  const Reg Index = GetSrc<RA_64>(Op->Index.ID());

  if (Op->Class == IR::GPRClass) {
-    auto value = GetSrc<RA_64>(Op->Value.ID());
+    const auto Value = GetSrc<RA_64>(Op->Value.ID());
    lea(rax, dword [STATE + Op->BaseOffset]);

    switch (Op->Stride) {
@@ -258,10 +405,10 @@ DEF_OP(StoreContextIndexed) {
    case 2:
    case 4:
    case 8: {
-      if (!(size == 1 || size == 2 || size == 4 || size == 8)) {
-        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", IROp->Size);
+      if (!(OpSize == 1 || OpSize == 2 || OpSize == 4 || OpSize == 8)) {
+        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", OpSize);
      }
-      mov(AddressFrame(IROp->Size * 8) [rax + index * Op->Stride], value);
+      mov(AddressFrame(OpSize * 8) [rax + Index * Op->Stride], Value);
      break;
    }
    default:
@@ -270,57 +417,64 @@ DEF_OP(StoreContextIndexed) {
    }
  }
  else {
-    auto value = GetSrc(Op->Value.ID());
+    const auto Value = GetSrc(Op->Value.ID());
    switch (Op->Stride) {
    case 1:
    case 2:
    case 4:
    case 8: {
      lea(rax, dword [STATE + Op->BaseOffset]);
-      switch (size) {
+      switch (OpSize) {
      case 1:
-        pextrb(AddressFrame(IROp->Size * 8) [rax + index * Op->Stride], value, 0);
+        pextrb(AddressFrame(OpSize * 8) [rax + Index * Op->Stride], Value, 0);
        break;
      case 2:
-        pextrw(AddressFrame(IROp->Size * 8) [rax + index * Op->Stride], value, 0);
+        pextrw(AddressFrame(OpSize * 8) [rax + Index * Op->Stride], Value, 0);
        break;
      case 4:
-        vmovd(AddressFrame(IROp->Size * 8) [rax + index * Op->Stride], value);
+        vmovd(AddressFrame(OpSize * 8) [rax + Index * Op->Stride], Value);
        break;
      case 8:
-        vmovq(AddressFrame(IROp->Size * 8) [rax + index * Op->Stride], value);
+        vmovq(AddressFrame(OpSize * 8) [rax + Index * Op->Stride], Value);
        break;
      default:
-        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", size);
+        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", OpSize);
        break;
      }
      break;
    }
-    case 16: {
-      mov(rax, index);
-      shl(rax, 4);
+    case 16:
+    case 32: {
+      const auto Shift = Op->Stride == 16 ? 4 : 5;
+
+      mov(rax, Index);
+      shl(rax, Shift);
      lea(rax, dword [rax + Op->BaseOffset]);
-      switch (size) {
+      switch (OpSize) {
      case 1:
-        pextrb(AddressFrame(IROp->Size * 8) [STATE + rax], value, 0);
+        pextrb(AddressFrame(OpSize * 8) [STATE + rax], Value, 0);
        break;
      case 2:
-        pextrw(AddressFrame(IROp->Size * 8) [STATE + rax], value, 0);
+        pextrw(AddressFrame(OpSize * 8) [STATE + rax], Value, 0);
        break;
      case 4:
-        vmovd(AddressFrame(IROp->Size * 8) [STATE + rax], value);
+        vmovd(AddressFrame(OpSize * 8) [STATE + rax], Value);
        break;
      case 8:
-        vmovq(AddressFrame(IROp->Size * 8) [STATE + rax], value);
+        vmovq(AddressFrame(OpSize * 8) [STATE + rax], Value);
        break;
      case 16:
-        if (Op->BaseOffset % 16 == 0)
-          movaps(xword [STATE + rax], value);
-        else
-          movups(xword [STATE + rax], value);
+        if (Op->BaseOffset % 16 == 0) {
+          vmovaps(xword [STATE + rax], Value);
+        } else {
+          vmovups(xword [STATE + rax], Value);
+        }
+        break;
+      case 32:
+        vmovups(yword [STATE + rax], ToYMM(Value));
        break;
      default:
-        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", size);
+        LOGMAN_MSG_A_FMT("Unhandled StoreContextIndexed size: {}", OpSize);
        break;
      }
      break;
@@ -333,10 +487,10 @@ DEF_OP(StoreContextIndexed) {
 }

 DEF_OP(SpillRegister) {
-  auto Op = IROp->C<IR::IROp_SpillRegister>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_SpillRegister>();
+  const uint8_t OpSize = IROp->Size;
+  const uint32_t SlotOffset = Op->Slot * MaxSpillSlotSize;

-  uint32_t SlotOffset = Op->Slot * 16;
  if (Op->Class == FEXCore::IR::GPRClass) {
    switch (OpSize) {
      case 1: {
@@ -355,36 +509,44 @@ DEF_OP(SpillRegister) {
        mov(qword [rsp + SlotOffset], GetSrc<RA_64>(Op->Value.ID()));
        break;
      }
-      default:  LOGMAN_MSG_A_FMT("Unhandled SpillRegister size: {}", OpSize);
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled SpillRegister size: {}", OpSize);
+        break;
    }
  } else if (Op->Class == FEXCore::IR::FPRClass) {
+    const auto Src = GetSrc(Op->Value.ID());
+
    switch (OpSize) {
      case 4: {
-        movss(dword [rsp + SlotOffset], GetSrc(Op->Value.ID()));
+        movss(dword [rsp + SlotOffset], Src);
        break;
      }
      case 8: {
-        movsd(qword [rsp + SlotOffset], GetSrc(Op->Value.ID()));
+        movsd(qword [rsp + SlotOffset], Src);
        break;
      }
      case 16: {
-        movaps(xword [rsp + SlotOffset], GetSrc(Op->Value.ID()));
+        movaps(xword [rsp + SlotOffset], Src);
        break;
      }
-      default:  LOGMAN_MSG_A_FMT("Unhandled SpillRegister size: {}", OpSize);
+      case 32: {
+        vmovups(yword [rsp + SlotOffset], ToYMM(Src));
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled SpillRegister size: {}", OpSize);
+        break;
    }
  } else {
    LOGMAN_MSG_A_FMT("Unhandled SpillRegister class: {}", Op->Class.Val);
  }
-
-
 }

 DEF_OP(FillRegister) {
-  auto Op = IROp->C<IR::IROp_FillRegister>();
-  uint8_t OpSize = IROp->Size;
+  const auto Op = IROp->C<IR::IROp_FillRegister>();
+  const uint8_t OpSize = IROp->Size;
+  const uint32_t SlotOffset = Op->Slot * MaxSpillSlotSize;

-  uint32_t SlotOffset = Op->Slot * 16;
  if (Op->Class == FEXCore::IR::GPRClass) {
    switch (OpSize) {
      case 1: {
@@ -403,23 +565,33 @@ DEF_OP(FillRegister) {
        mov(GetDst<RA_64>(Node), qword [rsp + SlotOffset]);
        break;
      }
-      default:  LOGMAN_MSG_A_FMT("Unhandled FillRegister size: {}", OpSize);
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled FillRegister size: {}", OpSize);
+        break;
    }
  } else if (Op->Class == FEXCore::IR::FPRClass) {
+    const auto Dst = GetDst(Node);
+
    switch (OpSize) {
      case 4: {
-        movss(GetDst(Node), dword [rsp + SlotOffset]);
+        vmovss(Dst, dword [rsp + SlotOffset]);
        break;
      }
      case 8: {
-        movsd(GetDst(Node), qword [rsp + SlotOffset]);
+        vmovsd(Dst, qword [rsp + SlotOffset]);
        break;
      }
      case 16: {
-        movaps(GetDst(Node), xword [rsp + SlotOffset]);
+        vmovaps(Dst, xword [rsp + SlotOffset]);
        break;
      }
-      default:  LOGMAN_MSG_A_FMT("Unhandled FillRegister size: {}", OpSize);
+      case 32: {
+        vmovups(ToYMM(Dst), yword [rsp + SlotOffset]);
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled FillRegister size: {}", OpSize);
+        break;
    }
  } else {
    LOGMAN_MSG_A_FMT("Unhandled FillRegister class: {}", Op->Class.Val);
@@ -464,130 +636,136 @@ Xbyak::RegExp X86JITCore::GenerateModRM(Xbyak::Reg Base, IR::OrderedNodeWrapper
 }

 DEF_OP(LoadMem) {
-  auto Op = IROp->C<IR::IROp_LoadMem>();
+  const auto Op = IROp->C<IR::IROp_LoadMem>();
+  const auto OpSize = IROp->Size;

-  Xbyak::Reg MemReg = GetSrc<RA_64>(Op->Addr.ID());
-
-  auto MemPtr = GenerateModRM(MemReg, Op->Offset, Op->OffsetType, Op->OffsetScale);
+  const Xbyak::Reg MemReg = GetSrc<RA_64>(Op->Addr.ID());
+  const auto MemPtr = GenerateModRM(MemReg, Op->Offset, Op->OffsetType, Op->OffsetScale);

  if (Op->Class == IR::GPRClass) {
-    auto Dst = GetDst<RA_64>(Node);
+    const auto Dst = GetDst<RA_64>(Node);

-    switch (IROp->Size) {
+    switch (OpSize) {
      case 1: {
-        movzx (Dst, byte [MemPtr]);
+        movzx(Dst, byte [MemPtr]);
+        break;
      }
-      break;
      case 2: {
-        movzx (Dst, word [MemPtr]);
+        movzx(Dst, word [MemPtr]);
+        break;
      }
-      break;
      case 4: {
        mov(Dst.cvt32(), dword [MemPtr]);
+        break;
      }
-      break;
      case 8: {
        mov(Dst, qword [MemPtr]);
+        break;
      }
-      break;
-      default:  LOGMAN_MSG_A_FMT("Unhandled LoadMem size: {}", IROp->Size);
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled LoadMem size: {}", OpSize);
+        break;
    }
  }
  else
  {
-    auto Dst = GetDst(Node);
+    const auto Dst = GetDst(Node);

-    switch (IROp->Size) {
+    switch (OpSize) {
      case 1: {
        movzx(eax, byte [MemPtr]);
        vmovd(Dst, eax);
+        break;
      }
-      break;
      case 2: {
        movzx(eax, word [MemPtr]);
        vmovd(Dst, eax);
+        break;
      }
-      break;
      case 4: {
        vmovd(Dst, dword [MemPtr]);
+        break;
      }
-      break;
      case 8: {
        vmovq(Dst, qword [MemPtr]);
+        break;
      }
-      break;
      case 16: {
-         if (IROp->Size == Op->Align)
-           movups(GetDst(Node), xword [MemPtr]);
-         else
-           movups(GetDst(Node), xword [MemPtr]);
-         if (MemoryDebug) {
-           movq(rcx, GetDst(Node));
-         }
-       }
-       break;
-      default:  LOGMAN_MSG_A_FMT("Unhandled LoadMem size: {}", IROp->Size);
+        vmovups(Dst, xword [MemPtr]);
+        if (MemoryDebug) {
+          movq(rcx, Dst);
+        }
+        break;
+      }
+      case 32: {
+        vmovups(ToYMM(Dst), yword [MemPtr]);
+        if (MemoryDebug) {
+          movq(rcx, Dst);
+        }
+        break;
+      }
+      default:
+        LOGMAN_MSG_A_FMT("Unhandled LoadMem size: {}", OpSize);
+        break;
    }
  }
 }

 DEF_OP(StoreMem) {
-  auto Op = IROp->C<IR::IROp_StoreMem>();
+  const auto Op = IROp->C<IR::IROp_StoreMem>();
+  const auto OpSize = IROp->Size;

-  Xbyak::Reg MemReg = GetSrc<RA_64>(Op->Addr.ID());
-
-  auto MemPtr = GenerateModRM(MemReg, Op->Offset, Op->OffsetType, Op->OffsetScale);
+  const Xbyak::Reg MemReg = GetSrc<RA_64>(Op->Addr.ID());
+  const auto MemPtr = GenerateModRM(MemReg, Op->Offset, Op->OffsetType, Op->OffsetScale);

  if (Op->Class == IR::GPRClass) {
-    switch (IROp->Size) {
+    switch (OpSize) {
    case 1:
      mov(byte [MemPtr], GetSrc<RA_8>(Op->Value.ID()));
-    break;
+      break;
    case 2:
      mov(word [MemPtr], GetSrc<RA_16>(Op->Value.ID()));
-    break;
+      break;
    case 4:
      mov(dword [MemPtr], GetSrc<RA_32>(Op->Value.ID()));
-    break;
+      break;
    case 8:
      mov(qword [MemPtr], GetSrc<RA_64>(Op->Value.ID()));
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled StoreMem size: {}", IROp->Size);
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled StoreMem size: {}", OpSize);
+      break;
    }
  }
  else {
-    switch (IROp->Size) {
+    const auto Value = GetSrc(Op->Value.ID());
+
+    switch (OpSize) {
    case 1:
-      pextrb(byte [MemPtr], GetSrc(Op->Value.ID()), 0);
-    break;
+      pextrb(byte [MemPtr], Value, 0);
+      break;
    case 2:
-      pextrw(word [MemPtr], GetSrc(Op->Value.ID()), 0);
-    break;
+      pextrw(word [MemPtr], Value, 0);
+      break;
    case 4:
-      vmovd(dword [MemPtr], GetSrc(Op->Value.ID()));
-    break;
+      vmovd(dword [MemPtr], Value);
+      break;
    case 8:
-      vmovq(qword [MemPtr], GetSrc(Op->Value.ID()));
-    break;
+      vmovq(qword [MemPtr], Value);
+      break;
    case 16:
-      if (IROp->Size == Op->Align)
-        movups(xword [MemPtr], GetSrc(Op->Value.ID()));
-      else
-        movups(xword [MemPtr], GetSrc(Op->Value.ID()));
-    break;
-    default:  LOGMAN_MSG_A_FMT("Unhandled StoreMem size: {}", IROp->Size);
+      vmovups(xword [MemPtr], Value);
+      break;
+    case 32:
+      vmovups(yword [MemPtr], ToYMM(Value));
+      break;
+    default:
+      LOGMAN_MSG_A_FMT("Unhandled StoreMem size: {}", OpSize);
+      break;
    }
  }
 }

-DEF_OP(VLoadMemElement) {
-  LOGMAN_MSG_A_FMT("Unimplemented");
-}
-
-DEF_OP(VStoreMemElement) {
-  LOGMAN_MSG_A_FMT("Unimplemented");
-}
-
 DEF_OP(CacheLineClear) {
  auto Op = IROp->C<IR::IROp_CacheLineClear>();

@@ -618,8 +796,8 @@ void X86JITCore::RegisterMemoryHandlers() {
 #define REGISTER_OP(op, x) OpHandlers[FEXCore::IR::IROps::OP_##op] = &X86JITCore::Op_##x
  REGISTER_OP(LOADCONTEXT,         LoadContext);
  REGISTER_OP(STORECONTEXT,        StoreContext);
-  REGISTER_OP(LOADREGISTER,        Unhandled); // SRA specific, not supported on this backend
-  REGISTER_OP(STOREREGISTER,       Unhandled);
+  REGISTER_OP(LOADREGISTER,        LoadRegister);
+  REGISTER_OP(STOREREGISTER,       StoreRegister);
  REGISTER_OP(LOADCONTEXTINDEXED,  LoadContextIndexed);
  REGISTER_OP(STORECONTEXTINDEXED, StoreContextIndexed);
  REGISTER_OP(SPILLREGISTER,       SpillRegister);
@@ -630,8 +808,6 @@ void X86JITCore::RegisterMemoryHandlers() {
  REGISTER_OP(STOREMEM,            StoreMem);
  REGISTER_OP(LOADMEMTSO,          LoadMem);
  REGISTER_OP(STOREMEMTSO,         StoreMem);
-  REGISTER_OP(VLOADMEMELEMENT,     VLoadMemElement);
-  REGISTER_OP(VSTOREMEMELEMENT,    VStoreMemElement);
  REGISTER_OP(CACHELINECLEAR,      CacheLineClear);
  REGISTER_OP(CACHELINEZERO,       CacheLineZero);
 #undef REGISTER_OP
@@ -45,56 +45,38 @@ DEF_OP(Fence) {

 DEF_OP(Break) {
  auto Op = IROp->C<IR::IROp_Break>();
-  switch (Op->Reason) {
-    case FEXCore::IR::Break_Unimplemented: // Hard fault
-    case FEXCore::IR::Break_Interrupt: // Guest ud2
-      ud2();
-      break;
-    case FEXCore::IR::Break_Overflow: // overflow
-      // Need to be outside of JIT cache space to ensure cache clearing correctness
-      jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.OverflowExceptionHandler)]);
-      break;
-    case FEXCore::IR::Break_Halt: { // HLT
-      // Time to quit
-      // Set our stack to the starting stack location
-      mov(rsp, qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, ReturningStackLocation)]);

-      // Now we need to jump to the thread stop handler
-      jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadStopHandlerSpillSRA)]);
-      break;
-    }
-    case FEXCore::IR::Break_Interrupt3: // INT3
-    {
-      if (CTX->GetGdbServerStatus()) {
-        // Adjust the stack first for a regular return
-        if (SpillSlots) {
-          add(rsp, SpillSlots * 16);
-        }
+  if (SpillSlots) {
+    add(rsp, SpillSlots * MaxSpillSlotSize);
+  }

-        // This jump target needs to be a constant offset here
-        jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadPauseHandlerSpillSRA)]);
-      }
-      else {
-        // If we don't have a gdb server attached then....crash?
-        // Treat this case like HLT
-        mov(rsp, qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, ReturningStackLocation)]);
+  Core::CpuStateFrame::SynchronousFaultDataStruct State = {
+    .FaultToTopAndGeneratedException = 1,
+    .Signal = Op->Reason.Signal,
+    .TrapNo = Op->Reason.TrapNumber,
+    .si_code = Op->Reason.si_code,
+    .err_code = Op->Reason.ErrorRegister,
+  };

-        // Now we need to jump to the thread stop handler
-        jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.ThreadStopHandlerSpillSRA)]);
-      }
+  uint64_t Constant{};
+  memcpy(&Constant, &State, sizeof(State));
+
+  mov(TMP1, Constant);
+  mov(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, SynchronousFaultData)], TMP1);
+
+  switch (Op->Reason.Signal) {
+  case SIGILL:
+    jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGILL)]);
+    break;
+  case SIGTRAP:
+    jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGTRAP)]);
+    break;
+  case SIGSEGV:
+    jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGSEGV)]);
+    break;
+  default:
+    jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.GuestSignal_SIGTRAP)]);
    break;
-    }
-    case FEXCore::IR::Break_InvalidInstruction:
-    {
-      if (SpillSlots) {
-        add(rsp, SpillSlots * 16);
-      }
-
-      // Need to be outside of JIT cache space to ensure cache clearing correctness
-      jmp(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.UnimplementedInstructionHandler)]);
-      break;
-    }
-    default: LOGMAN_MSG_A_FMT("Unknown Break reason: {}", Op->Reason);
  }
 }

@@ -110,7 +92,7 @@ DEF_OP(GetRoundingMode) {

 DEF_OP(SetRoundingMode) {
  auto Op = IROp->C<IR::IROp_SetRoundingMode>();
-  auto Src = GetSrc<RA_32>(Op->Header.Args[0].ID());
+  auto Src = GetSrc<RA_32>(Op->RoundMode.ID());

  // Load old mxcsr
  // Only stores to memory
@@ -135,13 +117,13 @@ DEF_OP(Print) {
  auto Op = IROp->C<IR::IROp_Print>();

  PushRegs();
-  if (IsGPR(Op->Header.Args[0].ID())) {
-    mov (rdi, GetSrc<RA_64>(Op->Header.Args[0].ID()));
+  if (IsGPR(Op->Value.ID())) {
+    mov (rdi, GetSrc<RA_64>(Op->Value.ID()));
    call(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.PrintValue)]);
  }
  else {
-    pextrq(rdi, GetSrc(Op->Header.Args[0].ID()), 0);
-    pextrq(rsi, GetSrc(Op->Header.Args[0].ID()), 1);
+    pextrq(rdi, GetSrc(Op->Value.ID()), 0);
+    pextrq(rsi, GetSrc(Op->Value.ID()), 1);

    call(qword [STATE + offsetof(FEXCore::Core::CpuStateFrame, Pointers.Common.PrintVectorValue)]);
  }
@@ -20,13 +20,13 @@ DEF_OP(ExtractElementPair) {
  auto Op = IROp->C<IR::IROp_ExtractElementPair>();
  switch (Op->Header.Size) {
    case 4: {
-      auto Src = GetSrcPair<RA_32>(Op->Header.Args[0].ID());
+      auto Src = GetSrcPair<RA_32>(Op->Pair.ID());
      std::array<Xbyak::Reg, 2> Regs = {Src.first, Src.second};
      mov (GetDst<RA_32>(Node), Regs[Op->Element]);
      break;
    }
    case 8: {
-      auto Src = GetSrcPair<RA_64>(Op->Header.Args[0].ID());
+      auto Src = GetSrcPair<RA_64>(Op->Pair.ID());
      std::array<Xbyak::Reg, 2> Regs = {Src.first, Src.second};
      mov (GetDst<RA_64>(Node), Regs[Op->Element]);
      break;
@@ -45,15 +45,15 @@ DEF_OP(CreateElementPair) {
  switch (IROp->ElementSize) {
    case 4: {
      Dst = GetSrcPair<RA_32>(Node);
-      RegFirst = GetSrc<RA_32>(Op->Header.Args[0].ID());
-      RegSecond = GetSrc<RA_32>(Op->Header.Args[1].ID());
+      RegFirst = GetSrc<RA_32>(Op->Lower.ID());
+      RegSecond = GetSrc<RA_32>(Op->Upper.ID());
      RegTmp = eax;
      break;
    }
    case 8: {
      Dst = GetSrcPair<RA_64>(Node);
-      RegFirst = GetSrc<RA_64>(Op->Header.Args[0].ID());
-      RegSecond = GetSrc<RA_64>(Op->Header.Args[1].ID());
+      RegFirst = GetSrc<RA_64>(Op->Lower.ID());
+      RegSecond = GetSrc<RA_64>(Op->Upper.ID());
      RegTmp = rax;
      break;
    }
@@ -73,17 +73,11 @@ DEF_OP(CreateElementPair) {
  }
 }

-DEF_OP(Mov) {
-  auto Op = IROp->C<IR::IROp_Mov>();
-  mov (GetDst<RA_64>(Node), GetSrc<RA_64>(Op->Header.Args[0].ID()));
-}
-
 #undef DEF_OP
 void X86JITCore::RegisterMoveHandlers() {
 #define REGISTER_OP(op, x) OpHandlers[FEXCore::IR::IROps::OP_##op] = &X86JITCore::Op_##x
  REGISTER_OP(EXTRACTELEMENTPAIR, ExtractElementPair);
  REGISTER_OP(CREATEELEMENTPAIR,  CreateElementPair);
-  REGISTER_OP(MOV,                Mov);
 #undef REGISTER_OP
 }
 }
@@ -91,7 +91,7 @@ bool X86JITCore::ApplyRelocations(uint64_t GuestEntry, uint64_t CodeEntry, uint6
  size_t DataIndex{};
  for (size_t j = 0; j < NumRelocations; ++j) {
    const FEXCore::CPU::Relocation *Reloc = reinterpret_cast<const FEXCore::CPU::Relocation *>(&EntryRelocations[DataIndex]);
-    LOGMAN_THROW_A_FMT((DataIndex % alignof(Relocation)) == 0, "Alignment of relocation wasn't adhered to");
+    LOGMAN_THROW_AA_FMT((DataIndex % alignof(Relocation)) == 0, "Alignment of relocation wasn't adhered to");

    switch (Reloc->Header.Type) {
      case FEXCore::CPU::RelocationTypes::RELOC_NAMED_SYMBOL_LITERAL: {
@@ -37,11 +37,11 @@ LookupCache::LookupCache(FEXCore::Context::Context *CTX)
  // We currently limit to 128MB of real memory for caching for the total cache size.
  // Can end up being inefficient if we compile a small number of blocks per page
  PageMemory = reinterpret_cast<uintptr_t>(FEXCore::Allocator::mmap(nullptr, CODE_SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0));
-  LOGMAN_THROW_A_FMT(PageMemory != -1ULL, "Failed to allocate page memory");
+  LOGMAN_THROW_AA_FMT(PageMemory != -1ULL, "Failed to allocate page memory");

  // L1 Cache
  L1Pointer = reinterpret_cast<uintptr_t>(FEXCore::Allocator::mmap(nullptr, L1_SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0));
-  LOGMAN_THROW_A_FMT(L1Pointer != -1ULL, "Failed to allocate L1Pointer");
+  LOGMAN_THROW_AA_FMT(L1Pointer != -1ULL, "Failed to allocate L1Pointer");

  VirtualMemSize = ctx->Config.VirtualMemSize;
 }
@@ -8,6 +8,7 @@
 #include <utility>
 #include <vector>
 #include <mutex>
+#include <tsl/robin_map.h>

 namespace FEXCore {
 namespace Context {
@@ -17,7 +18,7 @@ namespace Context {
 class LookupCache {
 public:

-  struct LookupCacheEntry { 
+  struct LookupCacheEntry {
    uintptr_t HostCode;
    uintptr_t GuestCode;
  };
@@ -54,7 +55,7 @@ public:
        return L1Entry.HostCode;
      }
    }
-    
+
    // Try L3
    auto HostCode = BlockList.find(Address);

@@ -62,7 +63,7 @@ public:
      CacheBlockMapping(Address, HostCode->second);
      return HostCode->second;
    }
-      
+
    // Failed to find
    return 0;
  }
@@ -73,7 +74,7 @@ public:
  // Returns true if new pages are marked as containing code
  bool AddBlockExecutableRange(uint64_t Address, uint64_t Start, uint64_t Length) {
    std::lock_guard<std::recursive_mutex> lk(WriteLock);
-    
+
    bool rv = false;

    for (auto CurrentPage = Start >> 12, EndPage = (Start + Length -1) >> 12; CurrentPage <= EndPage; CurrentPage++) {
@@ -88,9 +89,9 @@ public:
  // Adds to Guest -> Host code mapping
  void AddBlockMapping(uint64_t Address, void *HostCode) {
    std::lock_guard<std::recursive_mutex> lk(WriteLock);
-    
+
    [[maybe_unused]] auto Inserted = BlockList.emplace(Address, (uintptr_t)HostCode).second;
-    LOGMAN_THROW_A_FMT(Inserted, "Duplicate block mapping added");
+    LOGMAN_THROW_AA_FMT(Inserted, "Duplicate block mapping added");

    // There is no need to update L1 or L2, they will get updated on first lookup
    // However, adding to L1 here increases performance
@@ -158,7 +159,7 @@ public:
  constexpr static size_t L1_ENTRIES_MASK = L1_ENTRIES - 1;

  // This needs to be taken before reads or writes to L2, L3, CodePages, Thread::DebugStore,
-  // and before writes to L1. Concurrent access from a thread that this LookupCache doesn't belong to 
+  // and before writes to L1. Concurrent access from a thread that this LookupCache doesn't belong to
  // may only happen during cross thread invalidation (::Erase).
  // All other operations must be done from the owning thread.
  // Some care is taken so that L1 lookups can be done without locks, and even tearing is unlikely to lead to a crash.
@@ -167,7 +168,7 @@ public:
  std::recursive_mutex WriteLock;

 private:
-  void CacheBlockMapping(uint64_t Address, uintptr_t HostCode) { 
+  void CacheBlockMapping(uint64_t Address, uintptr_t HostCode) {
    std::lock_guard<std::recursive_mutex> lk(WriteLock);

    // Do L1
@@ -239,7 +240,7 @@ private:


  std::map<BlockLinkTag, std::function<void()>> BlockLinks;
-  std::map<uint64_t, uint64_t> BlockList;
+  tsl::robin_map<uint64_t, uint64_t> BlockList;

  constexpr static size_t CODE_SIZE = 128 * 1024 * 1024;
  constexpr static size_t SIZE_PER_PAGE = 4096 * sizeof(LookupCacheEntry);
@@ -76,10 +76,10 @@ public:
  OrderedNode* flagsOpSrcSigned{};

  FEXCore::Context::Context *CTX{};
-  
+
  // Used during new op bringup
  bool ShouldDump {false};
-  
+
  struct JumpTargetInfo {
    OrderedNode* BlockEntry;
    bool HaveEmitted;
@@ -278,6 +278,12 @@ public:
  void NOTOp(OpcodeArgs);
  void XADDOp(OpcodeArgs);
  void PopcountOp(OpcodeArgs);
+  void DAAOp(OpcodeArgs);
+  void DASOp(OpcodeArgs);
+  void AAAOp(OpcodeArgs);
+  void AASOp(OpcodeArgs);
+  void AAMOp(OpcodeArgs);
+  void AADOp(OpcodeArgs);
  void XLATOp(OpcodeArgs);
  template<bool Reseed>
  void RDRANDOp(OpcodeArgs);
@@ -292,6 +298,8 @@ public:
  void WriteSegmentReg(OpcodeArgs);
  void EnterOp(OpcodeArgs);

+  void SGDTOp(OpcodeArgs);
+
  // SSE
  void MOVAPSOp(OpcodeArgs);
  void MOVUPSOp(OpcodeArgs);
@@ -398,6 +406,26 @@ public:
  // ADX Ops
  void ADXOp(OpcodeArgs);

+  // AVX Ops
+  template <IROps IROp, size_t ElementSize>
+  void AVXVectorALUOp(OpcodeArgs);
+
+  void VANDNOp(OpcodeArgs);
+
+  void VMOVAPS_VMOVAPD_Op(OpcodeArgs);
+  void VMOVUPS_VMOVUPD_Op(OpcodeArgs);
+
+  void VMOVHPOp(OpcodeArgs);
+  void VMOVLPOp(OpcodeArgs);
+
+  void VMOVDDUPOp(OpcodeArgs);
+  void VMOVSHDUPOp(OpcodeArgs);
+  void VMOVSLDUPOp(OpcodeArgs);
+
+  void VMOVVectorNTOp(OpcodeArgs);
+
+  void VZEROOp(OpcodeArgs);
+
  // X87 Ops
  template<size_t width>
  void FLD(OpcodeArgs);
@@ -515,7 +543,7 @@ public:
  void X87FRSTORF64(OpcodeArgs);
  void X87FXAMF64(OpcodeArgs);
  void X87LDENVF64(OpcodeArgs);
-  
+
  template<size_t width, bool Integer, FCOMIFlags whichflags, bool poptwice>
  void FCOMIF64(OpcodeArgs);

@@ -646,6 +674,7 @@ private:
  OrderedNode *Current_HeaderNode{};

  OrderedNode *AppendSegmentOffset(OrderedNode *Value, uint32_t Flags, uint32_t DefaultPrefix = 0, bool Override = false);
+  void UpdatePrefixFromSegment(OrderedNode *Segment, uint32_t SegmentReg);

  enum class MemoryAccessType {
    // Choose TSO or Non-TSO depending on access type
@@ -657,6 +686,11 @@ private:
    // Non-temporal streaming
    ACCESS_STREAM,
  };
+  OrderedNode *LoadGPRRegister(uint32_t GPR, int8_t Size = -1, uint8_t Offset = 0);
+  OrderedNode *LoadXMMRegister(uint32_t XMM);
+  void StoreGPRRegister(uint32_t GPR, OrderedNode *const Src, int8_t Size = -1, uint8_t Offset = 0);
+  void StoreXMMRegister(uint32_t XMM, OrderedNode *const Src);
+
  OrderedNode *GetRelocatedPC(FEXCore::X86Tables::DecodedOp const& Op, int64_t Offset = 0);
  OrderedNode *LoadSource(FEXCore::IR::RegisterClassType Class, FEXCore::X86Tables::DecodedOp const& Op, FEXCore::X86Tables::DecodedOperand const& Operand, uint32_t Flags, int8_t Align, bool LoadData = true, bool ForceLoad = false, MemoryAccessType AccessType = MemoryAccessType::ACCESS_DEFAULT);
  OrderedNode *LoadSource_WithOpSize(FEXCore::IR::RegisterClassType Class, FEXCore::X86Tables::DecodedOp const& Op, FEXCore::X86Tables::DecodedOperand const& Operand, uint8_t OpSize, uint32_t Flags, int8_t Align, bool LoadData = true, bool ForceLoad = false, MemoryAccessType AccessType = MemoryAccessType::ACCESS_DEFAULT);
@@ -665,7 +699,7 @@ private:
  void StoreResult(FEXCore::IR::RegisterClassType Class, FEXCore::X86Tables::DecodedOp Op, OrderedNode *const Src, int8_t Align, MemoryAccessType AccessType = MemoryAccessType::ACCESS_DEFAULT);

  [[nodiscard]] static uint32_t GPROffset(X86State::X86Reg reg) {
-    LOGMAN_THROW_A_FMT(reg <= X86State::X86Reg::REG_R15, "Invalid reg used");
+    LOGMAN_THROW_AA_FMT(reg <= X86State::X86Reg::REG_R15, "Invalid reg used");
    return static_cast<uint32_t>(offsetof(Core::CPUState, gregs[static_cast<size_t>(reg)]));
  }

@@ -221,7 +221,8 @@ void OpDispatchBuilder::SHA256RNDS2Op(OpcodeArgs) {

  OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
-  OrderedNode *XMM0 = _LoadContext(16, FPRClass, offsetof(FEXCore::Core::CPUState, xmm[0]));
+  // Hardcoded to XMM0
+  auto XMM0 = LoadXMMRegister(0);

  auto A0 = _VExtractToGPR(16, 4, Src, 3);
  auto B0 = _VExtractToGPR(16, 4, Src, 2);
@@ -769,7 +769,11 @@ void OpDispatchBuilder::CalculcateFlags_ShiftLeftImmediate(uint8_t SrcSize, Orde
  // CF
  {
    // Extract the last bit shifted in to CF
-    SetRFLAG<FEXCore::X86State::RFLAG_CF_LOC>(_Bfe(1, SrcSize * 8 - Shift, Src1));
+    auto OpSize = SrcSize * 8;
+    if (OpSize < Shift) {
+      Shift &= (OpSize - 1);
+    }
+    SetRFLAG<FEXCore::X86State::RFLAG_CF_LOC>(_Bfe(1, OpSize - Shift, Src1));
  }

  // PF
@@ -934,6 +938,7 @@ void OpDispatchBuilder::CalculcateFlags_RotateRight(uint8_t SrcSize, OrderedNode
    auto OldOF = GetRFLAG(FEXCore::X86State::RFLAG_OF_LOC);

    // OF is set to the XOR of the new CF bit and the most significant bit of the result
+    // OF is architecturally only defined for 1-bit rotate, which is why this only happens when the shift is one.
    auto NewOF = _Xor(_Bfe(1, OpSize - 2, Res), NewCF);

    // If shift == 0, don't update flags
@@ -963,7 +968,9 @@ void OpDispatchBuilder::CalculcateFlags_RotateLeft(uint8_t SrcSize, OrderedNode
  // OF
  {
    auto OldOF = GetRFLAG(FEXCore::X86State::RFLAG_OF_LOC);
-    // OF is set to the XOR of the new CF bit and the most significant bit of the result
+    // OF is the LSB and MSB XOR'd together.
+    // OF is set to the XOR of the new CF bit and the most significant bit of the result.
+    // OF is architecturally only defined for 1-bit rotate, which is why this only happens when the shift is one.
    auto NewOF = _Xor(_Bfe(1, OpSize - 1, Res), NewCF);

    auto OF = _Select(FEXCore::IR::COND_EQ, Src2, _Constant(0), OldOF, NewOF);
@@ -977,8 +984,7 @@ void OpDispatchBuilder::CalculcateFlags_RotateRightImmediate(uint8_t SrcSize, Or
  if (Shift == 0) return;

  auto OpSize = SrcSize * 8;
-
-  auto NewCF = _Bfe(1, OpSize - Shift, Src1);
+  auto NewCF = _Bfe(1, OpSize - 1, Res);

  // CF
  {
@@ -989,8 +995,10 @@ void OpDispatchBuilder::CalculcateFlags_RotateRightImmediate(uint8_t SrcSize, Or
  // OF
  {
    if (Shift == 1) {
-      // OF is set to the XOR of the new CF bit and the most significant bit of the result
-      SetRFLAG<FEXCore::X86State::RFLAG_OF_LOC>(_Xor(_Bfe(1, OpSize - 1, Res), NewCF));
+      // OF is the top two MSBs XOR'd together
+      // OF is architecturally only defined for 1-bit rotate, which is why this only happens when the shift is one.
+      auto NewOF = _Xor(_Bfe(1, OpSize - 2, Res), NewCF);
+      SetRFLAG<FEXCore::X86State::RFLAG_OF_LOC>(NewOF);
    }
  }
 }
@@ -1000,17 +1008,22 @@ void OpDispatchBuilder::CalculcateFlags_RotateLeftImmediate(uint8_t SrcSize, Ord

  auto OpSize = SrcSize * 8;

+  auto NewCF = _Bfe(1, 0, Res);
  // CF
  {
    // Extract the last bit shifted in to CF
-    SetRFLAG<FEXCore::X86State::RFLAG_CF_LOC>(_Bfe(1, Shift, Src1));
+    SetRFLAG<FEXCore::X86State::RFLAG_CF_LOC>(NewCF);
  }

  // OF
  {
    if (Shift == 1) {
-      // OF is the top two MSBs XOR'd together
-      SetRFLAG<FEXCore::X86State::RFLAG_OF_LOC>(_Xor(_Bfe(1, OpSize - 1, Src1), _Bfe(1, OpSize - 2, Src1)));
+      // OF is the LSB and MSB XOR'd together.
+      // OF is set to the XOR of the new CF bit and the most significant bit of the result.
+      // OF is architecturally only defined for 1-bit rotate, which is why this only happens when the shift is one.
+      auto NewOF = _Xor(_Bfe(1, OpSize - 1, Res), NewCF);
+
+      SetRFLAG<FEXCore::X86State::RFLAG_OF_LOC>(NewOF);
    }
  }
 }
@@ -33,11 +33,46 @@ void OpDispatchBuilder::MOVVectorNTOp(OpcodeArgs) {
  StoreResult(FPRClass, Op, Src, 1, MemoryAccessType::ACCESS_STREAM);
 }

+void OpDispatchBuilder::VMOVVectorNTOp(OpcodeArgs) {
+  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 1, true, false, MemoryAccessType::ACCESS_STREAM);
+
+  // TODO: When stores and loads gain the ability to explicitly express
+  //       whether a vector extension or an insert is desirable, ensure
+  //       the 128-bit case here is a zero extend on store if the destination
+  //       is a register.
+
+  StoreResult(FPRClass, Op, Src, 1, MemoryAccessType::ACCESS_STREAM);
+}
+
 void OpDispatchBuilder::MOVAPSOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
  StoreResult(FPRClass, Op, Src, -1);
 }

+void OpDispatchBuilder::VMOVAPS_VMOVAPD_Op(OpcodeArgs) {
+  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
+  const auto Is128BitDest = GetDstSize(Op) == Core::CPUState::XMM_SSE_REG_SIZE;
+
+  if (Op->Dest.IsGPR() && Is128BitDest) {
+    // Perform 32 byte store to clear the upper lane.
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Src, 32, -1);
+  } else {
+    StoreResult(FPRClass, Op, Src, -1);
+  }
+}
+
+void OpDispatchBuilder::VMOVUPS_VMOVUPD_Op(OpcodeArgs) {
+  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 1);
+  const auto Is128BitDest = GetDstSize(Op) == Core::CPUState::XMM_SSE_REG_SIZE;
+
+  if (Op->Dest.IsGPR() && Is128BitDest) {
+    // Perform 32 byte store to clear the upper lane.
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Src, 32, 1);
+  } else {
+    StoreResult(FPRClass, Op, Src, 1);
+  }
+}
+
 void OpDispatchBuilder::MOVUPSOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 1);
  StoreResult(FPRClass, Op, Src, 1);
@@ -68,20 +103,33 @@ void OpDispatchBuilder::MOVHPDOp(OpcodeArgs) {
  }
 }

+void OpDispatchBuilder::VMOVHPOp(OpcodeArgs) {
+  if (Op->Dest.IsGPR()) {
+    OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 16);
+    OrderedNode *Src2 = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags, 8);
+    OrderedNode *Result = _VInsElement(16, 8, 1, 0, Src1, Src2);
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 32, -1);
+  } else {
+    OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 16);
+    OrderedNode *Result = _VInsElement(16, 8, 0, 1, Src, Src);
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 8, 8);
+  }
+}
+
 void OpDispatchBuilder::MOVLPOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 8);
  if (Op->Dest.IsGPR()) {
    // xmm, xmm is movhlps special case
    if (Op->Src[0].IsGPR()) {
      OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, 8, 16);
-      Src = _VExtractElement(16, 8, Src, 1);
-      auto Result = _VInsScalarElement(16, 8, 0, Dest, Src);
+      auto Result = _VInsElement(16, 8, 0, 1, Dest, Src);
      StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 16, 16);
    }
    else {
-      OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, 8, 16);
-      auto Result = _VInsScalarElement(16, 8, 0, Dest, Src);
-      StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 8, 16);
+      auto DstSize = GetDstSize(Op);
+      OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, DstSize, Op->Flags, -1);
+      auto Result = _VInsElement(16, 8, 0, 0, Dest, Src);
+      StoreResult(FPRClass, Op, Result, -1);
    }
  }
  else {
@@ -89,30 +137,68 @@ void OpDispatchBuilder::MOVLPOp(OpcodeArgs) {
  }
 }

+void OpDispatchBuilder::VMOVLPOp(OpcodeArgs) {
+  if (Op->Dest.IsGPR()) {
+    OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 16);
+    OrderedNode *Src2 = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags, 8);
+    OrderedNode *Result = _VInsElement(16, 8, 0, 0, Src1, Src2);
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 32, -1);
+  } else {
+    OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 8);
+    StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Src, 8, 8);
+  }
+}
+
 void OpDispatchBuilder::MOVSHDUPOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 8);
-  OrderedNode *Result = _VInsElement(16, 4, 3, 3, Src, Src);
-  Result = _VInsElement(16, 4, 2, 3, Result, Src);
-  Result = _VInsElement(16, 4, 1, 1, Result, Src);
+  OrderedNode *Result = _VInsElement(16, 4, 2, 3, Src, Src);
  Result = _VInsElement(16, 4, 0, 1, Result, Src);
  StoreResult(FPRClass, Op, Result, -1);
 }

+void OpDispatchBuilder::VMOVSHDUPOp(OpcodeArgs) {
+  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
+  const auto SrcSize = GetSrcSize(Op);
+  const auto Is256Bit = SrcSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  OrderedNode *Result = _VInsElement(SrcSize, 4, 2, 3, Src, Src);
+  Result = _VInsElement(SrcSize, 4, 0, 1, Result, Src);
+  if (Is256Bit) {
+    Result = _VInsElement(SrcSize, 4, 4, 5, Result, Src);
+    Result = _VInsElement(SrcSize, 4, 6, 7, Result, Src);
+  }
+
+  StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 32, -1);
+}
+
 void OpDispatchBuilder::MOVSLDUPOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, 8);
  OrderedNode *Result = _VInsElement(16, 4, 3, 2, Src, Src);
-  Result = _VInsElement(16, 4, 2, 2, Result, Src);
  Result = _VInsElement(16, 4, 1, 0, Result, Src);
-  Result = _VInsElement(16, 4, 0, 0, Result, Src);
  StoreResult(FPRClass, Op, Result, -1);
 }

+void OpDispatchBuilder::VMOVSLDUPOp(OpcodeArgs) {
+  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
+  const auto SrcSize = GetSrcSize(Op);
+  const auto Is256Bit = SrcSize == Core::CPUState::XMM_AVX_REG_SIZE;
+
+  OrderedNode *Result = _VInsElement(SrcSize, 4, 3, 2, Src, Src);
+  Result = _VInsElement(SrcSize, 4, 1, 0, Result, Src);
+  if (Is256Bit) {
+    Result = _VInsElement(SrcSize, 4, 5, 4, Result, Src);
+    Result = _VInsElement(SrcSize, 4, 7, 6, Result, Src);
+  }
+
+  StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Result, 32, -1);
+}
+
 void OpDispatchBuilder::MOVSSOp(OpcodeArgs) {
  if (Op->Dest.IsGPR() && Op->Src[0].IsGPR()) {
    // MOVSS xmm1, xmm2
    OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, 16, Op->Flags, -1);
    OrderedNode *Src = LoadSource_WithOpSize(FPRClass, Op, Op->Src[0], 4, Op->Flags, -1);
-    auto Result = _VInsScalarElement(16, 4, 0, Dest, Src);
+    auto Result = _VInsElement(16, 4, 0, 0, Dest, Src);
    StoreResult(FPRClass, Op, Result, -1);
  }
  else if (Op->Dest.IsGPR()) {
@@ -133,7 +219,7 @@ void OpDispatchBuilder::MOVSDOp(OpcodeArgs) {
    // xmm1[63:0] <- xmm2[63:0]
    OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
    OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
-    auto Result = _VInsScalarElement(16, 8, 0, Dest, Src);
+    auto Result = _VInsElement(16, 8, 0, 0, Dest, Src);
    StoreResult(FPRClass, Op, Result, -1);
  }
  else if (Op->Dest.IsGPR()) {
@@ -302,6 +388,48 @@ void OpDispatchBuilder::VectorALUOp<IR::OP_VUQSUB, 1>(OpcodeArgs);
 template
 void OpDispatchBuilder::VectorALUOp<IR::OP_VUQSUB, 2>(OpcodeArgs);

+template <IROps IROp, size_t ElementSize>
+void OpDispatchBuilder::AVXVectorALUOp(OpcodeArgs) {
+  const auto Size = GetSrcSize(Op);
+  const auto Is128Bit = Size == Core::CPUState::XMM_SSE_REG_SIZE;
+
+  OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
+  OrderedNode *Src2 = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags, -1);
+
+  auto ALUOp = _VAdd(Size, ElementSize, Src1, Src2);
+  // Overwrite our IR's op type
+  ALUOp.first->Header.Op = IROp;
+
+  OrderedNode* Result = ALUOp;
+  if (Is128Bit) {
+    // 128-bit variants need to zero the upper lane.
+    Result = _VMov(Size, ALUOp);
+  }
+
+  StoreResult(FPRClass, Op, Result, -1);
+}
+
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VADD, 1>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VADD, 2>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VADD, 4>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VADD, 8>(OpcodeArgs);
+
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VFADD, 4>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VFADD, 8>(OpcodeArgs);
+
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VAND, 16>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VOR, 16>(OpcodeArgs);
+template
+void OpDispatchBuilder::AVXVectorALUOp<IR::OP_VXOR, 16>(OpcodeArgs);
+
 template<FEXCore::IR::IROps IROp, size_t ElementSize>
 void OpDispatchBuilder::VectorALUROp(OpcodeArgs) {
  auto Size = GetSrcSize(Op);
@@ -322,8 +450,9 @@ void OpDispatchBuilder::VectorALUROp<IR::OP_VFSUB, 8>(OpcodeArgs);

 template<FEXCore::IR::IROps IROp, size_t ElementSize>
 void OpDispatchBuilder::VectorScalarALUOp(OpcodeArgs) {
-  auto Size = GetSrcSize(Op);
-  OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
+  auto DstSize = GetDstSize(Op);
+
+  OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, DstSize, Op->Flags, -1);
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);

  // If OpSize == ElementSize then it only does the lower scalar op
@@ -333,9 +462,9 @@ void OpDispatchBuilder::VectorScalarALUOp(OpcodeArgs) {

  OrderedNode* Result = ALUOp;

-  if (Size != ElementSize) {
+  if (DstSize != ElementSize) {
    // Insert the lower bits
-    Result = _VInsScalarElement(Size, ElementSize, 0, Dest, Result);
+    Result = _VInsElement(DstSize, ElementSize, 0, 0, Dest, ALUOp);
  }

  StoreResult(FPRClass, Op, Result, -1);
@@ -369,11 +498,12 @@ void OpDispatchBuilder::VectorScalarALUOp<IR::OP_VFMAX, 8>(OpcodeArgs);
 template<FEXCore::IR::IROps IROp, size_t ElementSize, bool Scalar>
 void OpDispatchBuilder::VectorUnaryOp(OpcodeArgs) {
  auto Size = GetSrcSize(Op);
+  auto DstSize = GetDstSize(Op);
  if constexpr (Scalar) {
    Size = ElementSize;
  }
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
-  OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
+  OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, DstSize, Op->Flags, -1);

  auto ALUOp = _VFSqrt(Size, ElementSize, Src);
  // Overwrite our IR's op type
@@ -381,7 +511,7 @@ void OpDispatchBuilder::VectorUnaryOp(OpcodeArgs) {

  if constexpr (Scalar) {
    // Insert the lower bits
-    auto Result = _VInsScalarElement(GetSrcSize(Op), ElementSize, 0, Dest, ALUOp);
+    auto Result = _VInsElement(DstSize, ElementSize, 0, 0, Dest, ALUOp);
    StoreResult(FPRClass, Op, Result, -1);
  }
  else {
@@ -440,10 +570,10 @@ void OpDispatchBuilder::MOVQOp(OpcodeArgs) {
  // This instruction is a bit special that if the destination is a register then it'll ZEXT the 64bit source to 128bit
  if (Op->Dest.IsGPR()) {
    const auto gpr = Op->Dest.Data.GPR.GPR;
+    const auto gprIndex = gpr - X86State::REG_XMM_0;

-    _StoreContext(8, FPRClass, Src, offsetof(FEXCore::Core::CPUState, xmm[gpr - FEXCore::X86State::REG_XMM_0][0]));
-    auto Const = _Constant(0);
-    _StoreContext(8, GPRClass, Const, offsetof(FEXCore::Core::CPUState, xmm[gpr - FEXCore::X86State::REG_XMM_0][1]));
+    auto Reg = _VMov(16, Src);
+    StoreXMMRegister(gprIndex, Reg);
  }
  else {
    // This is simple, just store the result
@@ -562,7 +692,7 @@ void OpDispatchBuilder::PSHUFBOp(OpcodeArgs) {

 template<size_t ElementSize, bool HalfSize, bool Low>
 void OpDispatchBuilder::PSHUFDOp(OpcodeArgs) {
-  LOGMAN_THROW_A_FMT(ElementSize != 0, "What. No element size?");
+  LOGMAN_THROW_AA_FMT(ElementSize != 0, "What. No element size?");
  const auto Size = GetSrcSize(Op);
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
  uint8_t Shuffle = Op->Src[1].Data.Literal.Value;
@@ -599,7 +729,7 @@ void OpDispatchBuilder::PSHUFDOp<4, false, true>(OpcodeArgs);

 template<size_t ElementSize>
 void OpDispatchBuilder::SHUFOp(OpcodeArgs) {
-  LOGMAN_THROW_A_FMT(ElementSize != 0, "What. No element size?");
+  LOGMAN_THROW_AA_FMT(ElementSize != 0, "What. No element size?");
  const auto Size = GetSrcSize(Op);
  OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
  OrderedNode *Src2 = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
@@ -654,6 +784,22 @@ void OpDispatchBuilder::ANDNOp(OpcodeArgs) {
  StoreResult(FPRClass, Op, Dest, -1);
 }

+void OpDispatchBuilder::VANDNOp(OpcodeArgs) {
+  const auto Size = GetSrcSize(Op);
+  const auto Is128Bit = Size == Core::CPUState::XMM_SSE_REG_SIZE;
+
+  OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
+  OrderedNode *Src2 = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags, -1);
+
+  Src1 = _VNot(Size, Size, Src1);
+  OrderedNode *Dest = _VAnd(Size, Size, Src1, Src2);
+  if (Is128Bit) {
+    Dest = _VMov(16, Dest);
+  }
+
+  StoreResult(FPRClass, Op, Dest, -1);
+}
+
 template<size_t ElementSize>
 void OpDispatchBuilder::PINSROp(OpcodeArgs) {
  auto Size = GetDstSize(Op);
@@ -921,7 +1067,11 @@ void OpDispatchBuilder::PSRLDQ(OpcodeArgs) {

  auto Size = GetDstSize(Op);

-  auto Result = _VSRI(Size, 16, Dest, Shift);
+  OrderedNode *Result = _VectorZero(Size);
+  if (Shift < Size) {
+    Result = _VExtr(Size, 1, Result, Dest, Shift);
+  }
+
  StoreResult(FPRClass, Op, Result, -1);
 }

@@ -933,7 +1083,10 @@ void OpDispatchBuilder::PSLLDQ(OpcodeArgs) {

  auto Size = GetDstSize(Op);

-  auto Result = _VSLI(Size, 16, Dest, Shift);
+  OrderedNode *Result = _VectorZero(Size);
+  if (Shift < Size) {
+    Result = _VExtr(Size, 1, Dest, Result, Size - Shift);
+  }
  StoreResult(FPRClass, Op, Result, -1);
 }

@@ -972,10 +1125,28 @@ void OpDispatchBuilder::PAVGOp<2>(OpcodeArgs);

 void OpDispatchBuilder::MOVDDUPOp(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
-  OrderedNode *Res =  _SplatVector2(Src);
+  OrderedNode *Res = _VDupElement(16, GetSrcSize(Op), Src, 0);
+
  StoreResult(FPRClass, Op, Res, -1);
 }

+void OpDispatchBuilder::VMOVDDUPOp(OpcodeArgs) {
+  const auto SrcSize = GetSrcSize(Op);
+  const auto IsSrcGPR = Op->Src[0].IsGPR();
+  const auto Is256Bit = SrcSize == Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto MemSize = Is256Bit ? 32 : 8;
+
+  OrderedNode *Src = IsSrcGPR ? LoadSource_WithOpSize(FPRClass, Op, Op->Src[0], SrcSize, Op->Flags, -1)
+                              : LoadSource_WithOpSize(FPRClass, Op, Op->Src[0], MemSize, Op->Flags, -1);
+
+  OrderedNode *Res = _VInsElement(SrcSize, 8, 1, 0, Src, Src);
+  if (Is256Bit) {
+    Res = _VInsElement(SrcSize, 8, 3, 2, Res, Src);
+  }
+
+  StoreResult_WithOpSize(FPRClass, Op, Op->Dest, Res, 32, -1);
+}
+
 template<size_t DstElementSize>
 void OpDispatchBuilder::CVTGPR_To_FPR(OpcodeArgs) {
  OrderedNode *Src = LoadSource(GPRClass, Op, Op->Src[0], Op->Flags, -1);
@@ -986,7 +1157,7 @@ void OpDispatchBuilder::CVTGPR_To_FPR(OpcodeArgs) {

  OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, 16, Op->Flags, -1);

-  Src = _VInsScalarElement(16, DstElementSize, 0, Dest, Src);
+  Src = _VInsElement(16, DstElementSize, 0, 0, Dest, Src);

  StoreResult(FPRClass, Op, Src, -1);
 }
@@ -1081,13 +1252,16 @@ void OpDispatchBuilder::Vector_CVT_Float_To_Int<8, true, false>(OpcodeArgs);

 template<size_t DstElementSize, size_t SrcElementSize>
 void OpDispatchBuilder::Scalar_CVT_Float_To_Float(OpcodeArgs) {
-  OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
+  const auto DstSize = GetDstSize(Op);
+
+  OrderedNode *Dest = LoadSource_WithOpSize(FPRClass, Op, Op->Dest, DstSize, Op->Flags, -1);
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);

  Src = _Float_FToF(DstElementSize, SrcElementSize, Src);
-  Src = _VInsScalarElement(16, DstElementSize, 0, Dest, Src);
+  Src = _VInsElement(16, DstElementSize, 0, 0, Dest, Src);

-  StoreResult(FPRClass, Op, Src, -1);
+  auto Result = _VInsElement(DstSize, DstElementSize, 0, 0, Dest, Src);
+  StoreResult(FPRClass, Op, Result, -1);
 }

 template
@@ -1097,8 +1271,9 @@ void OpDispatchBuilder::Scalar_CVT_Float_To_Float<8, 4>(OpcodeArgs);

 template<size_t DstElementSize, size_t SrcElementSize>
 void OpDispatchBuilder::Vector_CVT_Float_To_Float(OpcodeArgs) {
+  const auto Size = GetDstSize(Op);
+
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);
-  size_t Size = GetDstSize(Op);

  if constexpr (DstElementSize > SrcElementSize) {
    Src = _Vector_FToF(Size, SrcElementSize << 1, Src, SrcElementSize);
@@ -1175,13 +1350,12 @@ void OpDispatchBuilder::XMM_To_MMX_Vector_CVT_Float_To_Int<8, true>(OpcodeArgs);

 void OpDispatchBuilder::MASKMOVOp(OpcodeArgs) {
  // Until we get correct PHI nodes this is required to be a loop unroll
-  const auto GPRSize = CTX->GetGPRSize();
  const auto Size = uint32_t{GetSrcSize(Op)} * 8;

  OrderedNode *Src = LoadSource(GPRClass, Op, Op->Src[0], Op->Flags, -1);
  OrderedNode *Dest = LoadSource(GPRClass, Op, Op->Dest, Op->Flags, -1);

-  OrderedNode *MemDest = _LoadContext(GPRSize, GPRClass, GPROffset(X86State::REG_RDI));
+  auto MemDest = LoadGPRRegister(X86State::REG_RDI);

  const size_t NumElements = Size / 64;
  for (size_t Element = 0; Element < NumElements; ++Element) {
@@ -1275,7 +1449,7 @@ void OpDispatchBuilder::VFCMPOp(OpcodeArgs) {

  if constexpr (Scalar) {
    // Insert the lower bits
-    Result = _VInsScalarElement(GetDstSize(Op), ElementSize, 0, Dest, Result);
+    Result = _VInsElement(GetDstSize(Op), ElementSize, 0, 0, Dest, Result);
  }

  StoreResult(FPRClass, Op, Result, -1);
@@ -1390,10 +1564,11 @@ void OpDispatchBuilder::FXSaveOp(OpcodeArgs) {

    _StoreMem(FPRClass, 16, MemLocation, MMReg, 16);
  }
-  unsigned NumRegs = CTX->Config.Is64BitMode ? 16 : 8;
+
+  const auto NumRegs = CTX->Config.Is64BitMode ? 16U : 8U;

  for (unsigned i = 0; i < NumRegs; ++i) {
-    OrderedNode *XMMReg = _LoadContext(16, FPRClass, offsetof(FEXCore::Core::CPUState, xmm[i]));
+    OrderedNode *XMMReg = LoadXMMRegister(i);
    OrderedNode *MemLocation = _Add(Mem, _Constant(i * 16 + 160));

    _StoreMem(FPRClass, 16, MemLocation, XMMReg, 16);
@@ -1439,12 +1614,13 @@ void OpDispatchBuilder::FXRStoreOp(OpcodeArgs) {
    auto MMReg = _LoadMem(FPRClass, 16, MemLocation, 16);
    _StoreContext(16, FPRClass, MMReg, offsetof(FEXCore::Core::CPUState, mm[i]));
  }
-  unsigned NumRegs = CTX->Config.Is64BitMode ? 16 : 8;
+
+  const auto NumRegs = CTX->Config.Is64BitMode ? 16U : 8U;

  for (unsigned i = 0; i < NumRegs; ++i) {
    OrderedNode *MemLocation = _Add(Mem, _Constant(i * 16 + 160));
    auto XMMReg = _LoadMem(FPRClass, 16, MemLocation, 16);
-    _StoreContext(16, FPRClass, XMMReg, offsetof(FEXCore::Core::CPUState, xmm[i]));
+    StoreXMMRegister(i, XMMReg);
  }
 }

@@ -1541,6 +1717,9 @@ void OpDispatchBuilder::PACKSSOp<4>(OpcodeArgs);

 template<size_t ElementSize, bool Signed>
 void OpDispatchBuilder::PMULLOp(OpcodeArgs) {
+  static_assert(ElementSize == sizeof(uint32_t),
+                "Currently only handles 32-bit -> 64-bit");
+
  auto Size = GetSrcSize(Op);

  OrderedNode *Src1 = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
@@ -1557,17 +1736,8 @@ void OpDispatchBuilder::PMULLOp(OpcodeArgs) {
    }
  }
  else {
-    OrderedNode* Srcs1[2]{};
-    OrderedNode* Srcs2[2]{};
-
-    Srcs1[0] = _VExtr(Size, ElementSize, Src1, Src1, 0);
-    Srcs1[1] = _VExtr(Size, ElementSize, Src1, Src1, 2);
-
-    Srcs2[0] = _VExtr(Size, ElementSize, Src2, Src2, 0);
-    Srcs2[1] = _VExtr(Size, ElementSize, Src2, Src2, 2);
-
-    Src1 = _VInsElement(Size, ElementSize, 1, 0, Srcs1[0], Srcs1[1]);
-    Src2 = _VInsElement(Size, ElementSize, 1, 0, Srcs2[0], Srcs2[1]);
+    Src1 = _VInsElement(Size, ElementSize, 1, 2, Src1, Src1);
+    Src2 = _VInsElement(Size, ElementSize, 1, 2, Src2, Src2);

    if constexpr (Signed) {
      Res = _VSMull(Size, ElementSize, Src1, Src2);
@@ -1590,8 +1760,10 @@ void OpDispatchBuilder::MOVQ2DQ(OpcodeArgs) {

  // This instruction is a bit special in that if the source is MMX then it zexts to 128bit
  if constexpr (ToXMM) {
+    const auto Index = Op->Dest.Data.GPR.GPR - FEXCore::X86State::REG_XMM_0;
+
    Src = _VMov(16, Src);
-    _StoreContext(16, FPRClass, Src, offsetof(FEXCore::Core::CPUState, xmm[Op->Dest.Data.GPR.GPR - FEXCore::X86State::REG_XMM_0][0]));
+    StoreXMMRegister(Index, Src);
  }
  else {
    // This is simple, just store the result
@@ -1688,8 +1860,8 @@ void OpDispatchBuilder::PFNACCOp(OpcodeArgs) {

  OrderedNode *ResSubSrc{};
  OrderedNode *ResSubDest{};
-  auto UpperSubDest = _VExtractElement(Size, 4, Dest, 1);
-  auto UpperSubSrc = _VExtractElement(Size, 4, Src, 1);
+  auto UpperSubDest = _VDupElement(Size, 4, Dest, 1);
+  auto UpperSubSrc = _VDupElement(Size, 4, Src, 1);

  ResSubDest = _VFSub(4, 4, Dest, UpperSubDest);
  ResSubSrc = _VFSub(4, 4, Src, UpperSubSrc);
@@ -1707,7 +1879,7 @@ void OpDispatchBuilder::PFPNACCOp(OpcodeArgs) {

  OrderedNode *ResAdd{};
  OrderedNode *ResSub{};
-  auto UpperSubDest = _VExtractElement(Size, 4, Dest, 1);
+  auto UpperSubDest = _VDupElement(Size, 4, Dest, 1);

  ResSub = _VFSub(4, 4, Dest, UpperSubDest);
  ResAdd = _VFAddP(Size, 4, Src, Src);
@@ -1840,8 +2012,6 @@ void OpDispatchBuilder::PMADDWD(OpcodeArgs) {

  if (Size == 8) {
    Size <<= 1;
-    Src1 = _VBitcast(Size, 2, Src1);
-    Src2 = _VBitcast(Size, 2, Src2);
  }

  auto Src1_L = _VSXTL(Size, 2, Src1);  // [15:0 ], [31:16], [32:47 ], [63:48  ]
@@ -1928,9 +2098,6 @@ void OpDispatchBuilder::PMULHW(OpcodeArgs) {

  OrderedNode *Res{};
  if (Size == 8) {
-    Dest = _VBitcast(Size * 2, 2, Dest);
-    Src = _VBitcast(Size * 2, 2, Src);
-
    // Implementation is more efficient for 8byte registers
    if (Signed)
      Res = _VSMull(Size * 2, 2, Dest, Src);
@@ -2307,7 +2474,7 @@ void OpDispatchBuilder::VectorRound(OpcodeArgs) {
  if constexpr (Scalar) {
    // Insert the lower bits
    OrderedNode *Dest = LoadSource(FPRClass, Op, Op->Dest, Op->Flags, -1);
-    auto Result = _VInsScalarElement(GetDstSize(Op), ElementSize, 0, Dest, Src);
+    auto Result = _VInsElement(GetDstSize(Op), ElementSize, 0, 0, Dest, Src);
    StoreResult(FPRClass, Op, Result, -1);
  }
  else {
@@ -2356,7 +2523,8 @@ void OpDispatchBuilder::VectorVariableBlend(OpcodeArgs) {
  OrderedNode *Src = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags, -1);

  // The mask is hardcoded to be xmm0 in this instruction
-  OrderedNode *Mask = _LoadContext(16, FPRClass, offsetof(FEXCore::Core::CPUState, xmm[0]));
+  auto Mask = LoadXMMRegister(0);
+
  // Each element is selected by the high bit of that element size
  // Dest[ElementIdx] = Xmm0[ElementIndex][HighBit] ? Src : Dest;
  //
@@ -2591,4 +2759,28 @@ void OpDispatchBuilder::MPSADBWOp(OpcodeArgs) {
  StoreResult(FPRClass, Op, Result, -1);
 }

+void OpDispatchBuilder::VZEROOp(OpcodeArgs) {
+  const auto DstSize = GetDstSize(Op);
+  const auto IsVZEROALL = DstSize == Core::CPUState::XMM_AVX_REG_SIZE;
+  const auto NumRegs = CTX->Config.Is64BitMode ? 16U : 8U;
+
+  if (IsVZEROALL) {
+    // NOTE: Despite the name being VZEROALL, this will still only ever
+    //       zero out up to the first 16 registers (even on AVX-512, where we have 32 registers)
+
+    OrderedNode* ZeroVector = _VectorZero(DstSize);
+    for (uint32_t i = 0; i < NumRegs; i++) {
+      StoreXMMRegister(i, ZeroVector);
+    }
+  } else {
+    // Likewise, VZEROUPPER will only ever zero only up to the first 16 registers
+
+    for (uint32_t i = 0; i < NumRegs; i++) {
+      OrderedNode* Reg = LoadXMMRegister(i);
+      OrderedNode* Dst = _VMov(16, Reg);
+      StoreXMMRegister(i, Dst);
+    }
+  }
+}
+
 }
@@ -782,6 +782,12 @@ void OpDispatchBuilder::X87UnaryOp(OpcodeArgs) {
  // Overwrite the op
  result.first->Header.Op = IROp;

+  if constexpr (IROp == IR::OP_F80SIN ||
+                IROp == IR::OP_F80COS) {
+    // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+    SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+  }
+
  // Write to ST[TOP]
  _StoreContextIndexed(result, top, 16, MMBaseOffset(), 16, FPRClass);
 }
@@ -809,7 +815,8 @@ void OpDispatchBuilder::X87BinaryOp(OpcodeArgs) {
  // Overwrite the op
  result.first->Header.Op = IROp;

-  if constexpr (IROp == IR::OP_F80FPREM) {
+  if constexpr (IROp == IR::OP_F80FPREM ||
+    IROp == IR::OP_F80FPREM1) {
    //TODO: Set C0 to Q2, C3 to Q1, C1 to Q0
    SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
  }
@@ -854,6 +861,9 @@ void OpDispatchBuilder::X87SinCos(OpcodeArgs) {
  auto sin = _F80SIN(a);
  auto cos = _F80COS(a);

+  // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+  SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+
  // Write to ST[TOP]
  _StoreContextIndexed(sin, orig_top, 16, MMBaseOffset(), 16, FPRClass);
  _StoreContextIndexed(cos, top, 16, MMBaseOffset(), 16, FPRClass);
@@ -900,6 +910,9 @@ void OpDispatchBuilder::X87TAN(OpcodeArgs) {
  OrderedNode *data = _VCastFromGPR(16, 8, low);
  data = _VInsGPR(16, 8, 1, data, high);

+  // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+  SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+
  // Write to ST[TOP]
  _StoreContextIndexed(result, orig_top, 16, MMBaseOffset(), 16, FPRClass);
  _StoreContextIndexed(data, top, 16, MMBaseOffset(), 16, FPRClass);
@@ -1185,7 +1198,7 @@ void OpDispatchBuilder::X87FNSAVE(OpcodeArgs) {
  // upper 16 bits [79:64]
  _StoreMem(FPRClass, 8, ST0Location, data, 1);
  ST0Location = _Add(ST0Location, _Constant(8));
-  auto topBytes = _VExtractElement(16, 2, data, 4);
+  auto topBytes = _VDupElement(16, 2, data, 4);
  _StoreMem(FPRClass, 2, ST0Location, topBytes, 1);

  // reset to default
@@ -778,6 +778,12 @@ void OpDispatchBuilder::X87UnaryOpF64(OpcodeArgs) {
  // Overwrite the op
  result.first->Header.Op = IROp;

+  if constexpr (IROp == IR::OP_F64SIN ||
+                IROp == IR::OP_F64COS) {
+    // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+    SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+  }
+
  // Write to ST[TOP]
  _StoreContextIndexed(result, top, 8, MMBaseOffset(), 16, FPRClass);
 }
@@ -804,7 +810,8 @@ void OpDispatchBuilder::X87BinaryOpF64(OpcodeArgs) {
  // Overwrite the op
  result.first->Header.Op = IROp;

-  if constexpr (IROp == IR::OP_F64FPREM) {
+  if constexpr (IROp == IR::OP_F80FPREM ||
+    IROp == IR::OP_F80FPREM1) {
    //TODO: Set C0 to Q2, C3 to Q1, C1 to Q0
    SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
  }
@@ -831,6 +838,9 @@ void OpDispatchBuilder::X87SinCosF64(OpcodeArgs) {
  auto sin = _F64SIN(a);
  auto cos = _F64COS(a);

+  // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+  SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+
  // Write to ST[TOP]
  _StoreContextIndexed(sin, orig_top, 8, MMBaseOffset(), 16, FPRClass);
  _StoreContextIndexed(cos, top, 8, MMBaseOffset(), 16, FPRClass);
@@ -871,6 +881,9 @@ void OpDispatchBuilder::X87TANF64(OpcodeArgs) {

  auto one = _VCastFromGPR(8, 8, _Constant(0x3FF0000000000000));

+  // TODO: ACCURACY: should check source is in range –2^63 to +2^63
+  SetRFLAG<FEXCore::X86State::X87FLAG_C2_LOC>(_Constant(0));
+
  // Write to ST[TOP]
  _StoreContextIndexed(result, orig_top, 8, MMBaseOffset(), 16, FPRClass);
  _StoreContextIndexed(one, top, 8, MMBaseOffset(), 16, FPRClass);
@@ -996,7 +1009,7 @@ void OpDispatchBuilder::X87FNSAVEF64(OpcodeArgs) {
  // upper 16 bits [79:64]
  _StoreMem(FPRClass, 8, ST0Location, data, 1);
  ST0Location = _Add(ST0Location, _Constant(8));
-  auto topBytes = _VExtractElement(16, 2, data, 4);
+  auto topBytes = _VDupElement(16, 2, data, 4);
  _StoreMem(FPRClass, 2, ST0Location, topBytes, 1);

  // reset to default
@@ -266,10 +266,10 @@ void InitializeBaseTables(Context::OperatingMode Mode) {
    {0x17, 1, X86InstInfo{"POP SS",   TYPE_INST, GenFlagsSizes(SIZE_16BIT, SIZE_DEF) | FLAGS_DEBUG_MEM_ACCESS,    0, nullptr}},
    {0x1E, 1, X86InstInfo{"PUSH DS",  TYPE_INST, GenFlagsSrcSize(SIZE_16BIT) | FLAGS_DEBUG_MEM_ACCESS,            0, nullptr}},
    {0x1F, 1, X86InstInfo{"POP DS",   TYPE_INST, GenFlagsSizes(SIZE_16BIT, SIZE_DEF) | FLAGS_DEBUG_MEM_ACCESS,    0, nullptr}},
-    {0x27, 1, X86InstInfo{"DAA",      TYPE_INST, FLAGS_NONE,                                                      0, nullptr}},
-    {0x2F, 1, X86InstInfo{"DAS",      TYPE_INST, FLAGS_NONE,                                                      0, nullptr}},
-    {0x37, 1, X86InstInfo{"AAA",      TYPE_INST, FLAGS_NONE,                                                      0, nullptr}},
-    {0x3F, 1, X86InstInfo{"AAS",      TYPE_INST, FLAGS_NONE,                                                      0, nullptr}},
+    {0x27, 1, X86InstInfo{"DAA",      TYPE_INST, GenFlagsDstSize(SIZE_8BIT) | FLAGS_SF_DST_RAX,                   0, nullptr}},
+    {0x2F, 1, X86InstInfo{"DAS",      TYPE_INST, GenFlagsDstSize(SIZE_8BIT) | FLAGS_SF_DST_RAX,                   0, nullptr}},
+    {0x37, 1, X86InstInfo{"AAA",      TYPE_INST, GenFlagsDstSize(SIZE_16BIT) | FLAGS_SF_DST_RAX,                  0, nullptr}},
+    {0x3F, 1, X86InstInfo{"AAS",      TYPE_INST, GenFlagsDstSize(SIZE_16BIT) | FLAGS_SF_DST_RAX,                  0, nullptr}},

    {0x40, 8, X86InstInfo{"INC",    TYPE_INST, FLAGS_SF_REX_IN_BYTE,                                              0, nullptr}},
    {0x48, 8, X86InstInfo{"DEC",    TYPE_INST, FLAGS_SF_REX_IN_BYTE,                                              0, nullptr}},
@@ -283,8 +283,8 @@ void InitializeBaseTables(Context::OperatingMode Mode) {
    {0xA1, 1, X86InstInfo{"MOV",    TYPE_INST, FLAGS_SF_DST_RAX | FLAGS_MEM_OFFSET,                               4, nullptr}},
    {0xA3, 1, X86InstInfo{"MOV",    TYPE_INST, FLAGS_SF_SRC_RAX | FLAGS_MEM_OFFSET,                               4, nullptr}},
    {0xCE, 1, X86InstInfo{"INTO",   TYPE_INST, FLAGS_NONE,                                                        0, nullptr}},
-    {0xD4, 1, X86InstInfo{"AAM",    TYPE_INST, FLAGS_NONE,                                                        1, nullptr}},
-    {0xD5, 1, X86InstInfo{"AAD",    TYPE_INST, FLAGS_NONE,                                                        1, nullptr}},
+    {0xD4, 1, X86InstInfo{"AAM",    TYPE_INST, GenFlagsSameSize(SIZE_8BIT) | FLAGS_SF_DST_RAX,                    1, nullptr}},
+    {0xD5, 1, X86InstInfo{"AAD",    TYPE_INST, GenFlagsSameSize(SIZE_8BIT) | FLAGS_SF_DST_RAX,                    1, nullptr}},
    {0xEA, 1, X86InstInfo{"JMPF",   TYPE_INST, FLAGS_NONE,                                                        0, nullptr}},
  };

@@ -67,7 +67,7 @@ void InitializeSecondaryGroupTables() {
    {OPD(TYPE_GROUP_6, PF_F2, 7), 1, X86InstInfo{"",        TYPE_INVALID, FLAGS_NONE,    0, nullptr}},

    // GROUP 7
-    {OPD(TYPE_GROUP_7, PF_NONE, 0), 1, X86InstInfo{"SGDT", TYPE_UNDEC, FLAGS_MODRM | FLAGS_SF_MOD_DST,         0, nullptr}},
+    {OPD(TYPE_GROUP_7, PF_NONE, 0), 1, X86InstInfo{"SGDT", TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST,         0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_NONE, 1), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE, 0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_NONE, 2), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE, 0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_NONE, 3), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE, 0, nullptr}},
@@ -76,7 +76,7 @@ void InitializeSecondaryGroupTables() {
    {OPD(TYPE_GROUP_7, PF_NONE, 6), 1, X86InstInfo{"LMSW", TYPE_PRIV, FLAGS_MODRM,          0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_NONE, 7), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE, 0, nullptr}},

-    {OPD(TYPE_GROUP_7, PF_F3, 0), 1, X86InstInfo{"SGDT", TYPE_UNDEC, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
+    {OPD(TYPE_GROUP_7, PF_F3, 0), 1, X86InstInfo{"SGDT", TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F3, 1), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F3, 2), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F3, 3), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
@@ -85,7 +85,7 @@ void InitializeSecondaryGroupTables() {
    {OPD(TYPE_GROUP_7, PF_F3, 6), 1, X86InstInfo{"LMSW", TYPE_PRIV, FLAGS_MODRM,            0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F3, 7), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},

-    {OPD(TYPE_GROUP_7, PF_66, 0), 1, X86InstInfo{"SGDT", TYPE_UNDEC, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
+    {OPD(TYPE_GROUP_7, PF_66, 0), 1, X86InstInfo{"SGDT", TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_66, 1), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_66, 2), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_66, 3), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
@@ -94,7 +94,7 @@ void InitializeSecondaryGroupTables() {
    {OPD(TYPE_GROUP_7, PF_66, 6), 1, X86InstInfo{"LMSW", TYPE_PRIV, FLAGS_MODRM,            0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_66, 7), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},

-    {OPD(TYPE_GROUP_7, PF_F2, 0), 1, X86InstInfo{"SGDT", TYPE_UNDEC, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
+    {OPD(TYPE_GROUP_7, PF_F2, 0), 1, X86InstInfo{"SGDT", TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST,           0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F2, 1), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F2, 2), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
    {OPD(TYPE_GROUP_7, PF_F2, 3), 1, X86InstInfo{"",     TYPE_SECOND_GROUP_MODRM, FLAGS_NONE,   0, nullptr}},
@@ -295,7 +295,7 @@ void InitializeSecondaryTables(Context::OperatingMode Mode) {

    {0x20, 4, X86InstInfo{"",          TYPE_COPY_OTHER, FLAGS_NONE,                                     0, nullptr}},
    {0x24, 6, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                        0, nullptr}},
-    {0x2A, 1, X86InstInfo{"CVTSI2SS",  TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_SRC_GPR, 0, nullptr}},
+    {0x2A, 1, X86InstInfo{"CVTSI2SS",  TYPE_INST, GenFlagsDstSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_SRC_GPR, 0, nullptr}},
    {0x2B, 1, X86InstInfo{"MOVNTSS",   TYPE_INST, GenFlagsSameSize(SIZE_32BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
    {0x2C, 1, X86InstInfo{"CVTTSS2SI", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_DST_GPR, 0, nullptr}},
    {0x2D, 1, X86InstInfo{"CVTSS2SI",  TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_DST_GPR, 0, nullptr}},
@@ -305,18 +305,18 @@ void InitializeSecondaryTables(Context::OperatingMode Mode) {
    {0x40, 16, X86InstInfo{"",         TYPE_COPY_OTHER, FLAGS_NONE,                                     0, nullptr}},

    {0x50, 1, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                        0, nullptr}},
-    {0x51, 1, X86InstInfo{"SQRTSS",    TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x52, 1, X86InstInfo{"RSQRTSS",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x53, 1, X86InstInfo{"RCPSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x51, 1, X86InstInfo{"SQRTSS",    TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x52, 1, X86InstInfo{"RSQRTSS",   TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x53, 1, X86InstInfo{"RCPSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
    {0x54, 4, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                        0, nullptr}},
-    {0x58, 1, X86InstInfo{"ADDSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x59, 1, X86InstInfo{"MULSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5A, 1, X86InstInfo{"CVTSS2SD",  TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,   0, nullptr}},
+    {0x58, 1, X86InstInfo{"ADDSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x59, 1, X86InstInfo{"MULSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5A, 1, X86InstInfo{"CVTSS2SD",  TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,   0, nullptr}},
    {0x5B, 1, X86InstInfo{"CVTTPS2DQ", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,  0, nullptr}},
-    {0x5C, 1, X86InstInfo{"SUBSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5D, 1, X86InstInfo{"MINSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5E, 1, X86InstInfo{"DIVSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5F, 1, X86InstInfo{"MAXSS",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5C, 1, X86InstInfo{"SUBSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5D, 1, X86InstInfo{"MINSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5E, 1, X86InstInfo{"DIVSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5F, 1, X86InstInfo{"MAXSS",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_32BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},

    {0x60, 8, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                        0, nullptr}},
    {0x68, 7, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                        0, nullptr}},
@@ -383,16 +383,16 @@ void InitializeSecondaryTables(Context::OperatingMode Mode) {
    {0x40, 16, X86InstInfo{"",         TYPE_COPY_OTHER, FLAGS_NONE,                                                      0, nullptr}},

    {0x50, 1, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                                         0, nullptr}},
-    {0x51, 1, X86InstInfo{"SQRTSD",    TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x51, 1, X86InstInfo{"SQRTSD",    TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {0x52, 6, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                                         0, nullptr}},
-    {0x58, 1, X86InstInfo{"ADDSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x59, 1, X86InstInfo{"MULSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x58, 1, X86InstInfo{"ADDSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x59, 1, X86InstInfo{"MULSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
    {0x5A, 1, X86InstInfo{"CVTSD2SS",  TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
    {0x5B, 1, X86InstInfo{"",          TYPE_INVALID, FLAGS_NONE,                                                         0, nullptr}},
-    {0x5C, 1, X86InstInfo{"SUBSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5D, 1, X86InstInfo{"MINSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5E, 1, X86InstInfo{"DIVSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
-    {0x5F, 1, X86InstInfo{"MAXSD",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5C, 1, X86InstInfo{"SUBSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5D, 1, X86InstInfo{"MINSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5E, 1, X86InstInfo{"DIVSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},
+    {0x5F, 1, X86InstInfo{"MAXSD",     TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS,                    0, nullptr}},

    {0x60, 16, X86InstInfo{"",         TYPE_INVALID, FLAGS_NONE,                                                         0, nullptr}},

@@ -17,23 +17,23 @@ void InitializeVEXTables() {
  static constexpr U16U8InfoStruct VEXTable[] = {
    // Map 0 (Reserved)
    // VEX Map 1
-    {OPD(1, 0b00, 0x10), 1, X86InstInfo{"VMOVUPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x10), 1, X86InstInfo{"VMODUPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x10), 1, X86InstInfo{"VMOVUPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x10), 1, X86InstInfo{"VMOVUPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b10, 0x10), 1, X86InstInfo{"VMOVSS",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x10), 1, X86InstInfo{"VMOVSD",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x11), 1, X86InstInfo{"VMOVUPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x11), 1, X86InstInfo{"VMODUPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x11), 1, X86InstInfo{"VMOVUPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x11), 1, X86InstInfo{"VMOVUPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b10, 0x11), 1, X86InstInfo{"VMOVSS",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x11), 1, X86InstInfo{"VMOVSD",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x12), 1, X86InstInfo{"VMOVLPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x12), 1, X86InstInfo{"VMOVLPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b10, 0x12), 1, X86InstInfo{"VMOVSLDUP", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b11, 0x12), 1, X86InstInfo{"VMOVDDUP",  TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x12), 1, X86InstInfo{"VMOVLPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_XMM_FLAGS | FLAGS_VEX_1ST_SRC, 0, nullptr}},
+    {OPD(1, 0b01, 0x12), 1, X86InstInfo{"VMOVLPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_XMM_FLAGS | FLAGS_VEX_1ST_SRC, 0, nullptr}},
+    {OPD(1, 0b10, 0x12), 1, X86InstInfo{"VMOVSLDUP", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b11, 0x12), 1, X86InstInfo{"VMOVDDUP",  TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x13), 1, X86InstInfo{"VMOVLPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x13), 1, X86InstInfo{"VMOVLPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x13), 1, X86InstInfo{"VMOVLPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x13), 1, X86InstInfo{"VMOVLPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b00, 0x14), 1, X86InstInfo{"VUNPCKLPS", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x14), 1, X86InstInfo{"VUNPCKLPD", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -41,12 +41,12 @@ void InitializeVEXTables() {
    {OPD(1, 0b00, 0x15), 1, X86InstInfo{"VUNPCKHPS", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x15), 1, X86InstInfo{"VUNPCKHPD", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x16), 1, X86InstInfo{"VMOVHPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x16), 1, X86InstInfo{"VMOVHPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b10, 0x16), 1, X86InstInfo{"VMOVSHDUP", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x16), 1, X86InstInfo{"VMOVHPS",   TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_XMM_FLAGS | FLAGS_VEX_1ST_SRC, 0, nullptr}},
+    {OPD(1, 0b01, 0x16), 1, X86InstInfo{"VMOVHPD",   TYPE_INST, GenFlagsSizes(SIZE_128BIT, SIZE_64BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_XMM_FLAGS | FLAGS_VEX_1ST_SRC, 0, nullptr}},
+    {OPD(1, 0b10, 0x16), 1, X86InstInfo{"VMOVSHDUP", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x17), 1, X86InstInfo{"VMOVHPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x17), 1, X86InstInfo{"VMOVHPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x17), 1, X86InstInfo{"VMOVHPS",   TYPE_INST, GenFlagsSizes(SIZE_64BIT, SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x17), 1, X86InstInfo{"VMOVHPD",   TYPE_INST, GenFlagsSizes(SIZE_64BIT, SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b00, 0x50), 1, X86InstInfo{"VMOVMSKPS", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x50), 1, X86InstInfo{"VMOVMSKPD", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -62,17 +62,17 @@ void InitializeVEXTables() {
    {OPD(1, 0b00, 0x53), 1, X86InstInfo{"VRCPPS",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b10, 0x53), 1, X86InstInfo{"VRCPSS",    TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x54), 1, X86InstInfo{"VANDPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x54), 1, X86InstInfo{"VANDPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x54), 1, X86InstInfo{"VANDPS",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x54), 1, X86InstInfo{"VANDPD",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x55), 1, X86InstInfo{"VANDNPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x55), 1, X86InstInfo{"VANDNPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x55), 1, X86InstInfo{"VANDNPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x55), 1, X86InstInfo{"VANDNPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x56), 1, X86InstInfo{"VORPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x56), 1, X86InstInfo{"VORPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x56), 1, X86InstInfo{"VORPS",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x56), 1, X86InstInfo{"VORPD",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x57), 1, X86InstInfo{"VXORPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x57), 1, X86InstInfo{"VDORPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x57), 1, X86InstInfo{"VXORPS",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x57), 1, X86InstInfo{"VXORPD",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b01, 0x60), 1, X86InstInfo{"VPUNPCKLBW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x61), 1, X86InstInfo{"VPUNPCKLWD", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -95,7 +95,7 @@ void InitializeVEXTables() {
    {OPD(1, 0b01, 0x75), 1, X86InstInfo{"VPCMPEQW",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0x76), 1, X86InstInfo{"VPCMPEQD",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x77), 1, X86InstInfo{"VZERO*",     TYPE_INST, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x77), 1, X86InstInfo{"VZERO*",     TYPE_INST, GenFlagsDstSize(SIZE_128BIT), 0, nullptr}},

    {OPD(1, 0b00, 0xC2), 1, X86InstInfo{"VCMPccPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xC2), 1, X86InstInfo{"VCMPccPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -112,17 +112,17 @@ void InitializeVEXTables() {
    // This table doesn't state which VEX.pp is for which instruction
    // XXX: Confirm all the above encoding opcodes

-    {OPD(1, 0b00, 0x28), 1, X86InstInfo{"VMOVAPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x28), 1, X86InstInfo{"VMOVAPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x28), 1, X86InstInfo{"VMOVAPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x28), 1, X86InstInfo{"VMOVAPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b00, 0x29), 1, X86InstInfo{"VMOVAPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x29), 1, X86InstInfo{"VMOVAPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x29), 1, X86InstInfo{"VMOVAPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x29), 1, X86InstInfo{"VMOVAPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b10, 0x2A), 1, X86InstInfo{"VCVTSI2SS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x2A), 1, X86InstInfo{"VCVTSI2SD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x2B), 1, X86InstInfo{"VMOVNTPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x2B), 1, X86InstInfo{"VMOVNTPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x2B), 1, X86InstInfo{"VMOVNTPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x2B), 1, X86InstInfo{"VMOVNTPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b10, 0x2C), 1, X86InstInfo{"VCVTTSS2SI",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x2C), 1, X86InstInfo{"VCVTTSD2SI",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -136,8 +136,8 @@ void InitializeVEXTables() {
    {OPD(1, 0b00, 0x2F), 1, X86InstInfo{"VUCOMISS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x2F), 1, X86InstInfo{"VUCOMISD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b00, 0x58), 1, X86InstInfo{"VADDPS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0x58), 1, X86InstInfo{"VADDPD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b00, 0x58), 1, X86InstInfo{"VADDPS",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x58), 1, X86InstInfo{"VADDPD",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b10, 0x58), 1, X86InstInfo{"VADDSS",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x58), 1, X86InstInfo{"VADDSD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

@@ -179,8 +179,8 @@ void InitializeVEXTables() {
    {OPD(1, 0b01, 0x6D), 1, X86InstInfo{"VPUNPCKHQDQ", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0x6E), 1, X86InstInfo{"VMOV*",       TYPE_INST, GenFlagsDstSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_SRC_GPR, 0, nullptr}},

-    {OPD(1, 0b01, 0x6F), 1, X86InstInfo{"VMOVDQA",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b10, 0x6F), 1, X86InstInfo{"VMOVDQU",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x6F), 1, X86InstInfo{"VMOVDQA",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b10, 0x6F), 1, X86InstInfo{"VMOVDQU",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b01, 0x7C), 1, X86InstInfo{"VHADDPD",     TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x7C), 1, X86InstInfo{"VHADDPS",     TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -188,11 +188,11 @@ void InitializeVEXTables() {
    {OPD(1, 0b01, 0x7D), 1, X86InstInfo{"VHSUBPD",     TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0x7D), 1, X86InstInfo{"VHSUBPS",     TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b01, 0x7E), 1, X86InstInfo{"VMOV*",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b10, 0x7E), 1, X86InstInfo{"VMOVQ",     TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x7E), 1, X86InstInfo{"VMOV*",     TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_SF_DST_GPR | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b10, 0x7E), 1, X86InstInfo{"VMOVQ",     TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},

-    {OPD(1, 0b01, 0x7F), 1, X86InstInfo{"VMOVDQA",     TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b10, 0x7F), 1, X86InstInfo{"VMOVDQU",     TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0x7F), 1, X86InstInfo{"VMOVDQA",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b10, 0x7F), 1, X86InstInfo{"VMOVDQU",     TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b00, 0xAE), 1, X86InstInfo{"",     TYPE_VEX_GROUP_15, FLAGS_NONE, 0, nullptr}}, // VEX Group 15
    {OPD(1, 0b01, 0xAE), 1, X86InstInfo{"",     TYPE_VEX_GROUP_15, FLAGS_NONE, 0, nullptr}}, // VEX Group 15
@@ -205,19 +205,19 @@ void InitializeVEXTables() {
    {OPD(1, 0b01, 0xD1), 1, X86InstInfo{"VPSRLW",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xD2), 1, X86InstInfo{"VPSRLD",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xD3), 1, X86InstInfo{"VPSRLQ",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xD4), 1, X86InstInfo{"VPADDQ",      TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xD4), 1, X86InstInfo{"VPADDQ",      TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xD5), 1, X86InstInfo{"VPMULLW",     TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xD6), 1, X86InstInfo{"VMOVQ",       TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xD6), 1, X86InstInfo{"VMOVQ",       TYPE_INST, GenFlagsSameSize(SIZE_64BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xD7), 1, X86InstInfo{"VPMOVMSKB",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS | FLAGS_SF_DST_GPR | FLAGS_SF_MOD_REG_ONLY, 0, nullptr}},

    {OPD(1, 0b01, 0xD8), 1, X86InstInfo{"VPSUBUSB", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xD9), 1, X86InstInfo{"VPSUBUSW", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xDA), 1, X86InstInfo{"VPMINUB",  TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b01, 0xDB), 1, X86InstInfo{"VPAND",    TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xDB), 1, X86InstInfo{"VPAND",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xDC), 1, X86InstInfo{"VPADDUSB", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xDD), 1, X86InstInfo{"VPADDUSW", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xDE), 1, X86InstInfo{"VPMAXUB",  TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b01, 0xDF), 1, X86InstInfo{"VPANDN",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xDF), 1, X86InstInfo{"VPANDN",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b01, 0xE0), 1, X86InstInfo{"VPAVGB",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xE1), 1, X86InstInfo{"VPSRAW",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -230,16 +230,16 @@ void InitializeVEXTables() {
    {OPD(1, 0b10, 0xE6), 1, X86InstInfo{"VCVTDQ2PD",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b11, 0xE6), 1, X86InstInfo{"VCVTPD2DQ",   TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

-    {OPD(1, 0b01, 0xE7), 1, X86InstInfo{"VMOVNTDQ",    TYPE_INST, FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xE7), 1, X86InstInfo{"VMOVNTDQ",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_DST | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b01, 0xE8), 1, X86InstInfo{"VPSUBSB", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xE9), 1, X86InstInfo{"VPSUBSW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xEA), 1, X86InstInfo{"VPMINSW",  TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xEB), 1, X86InstInfo{"VPOR",    TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xEA), 1, X86InstInfo{"VPMINSW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b01, 0xEB), 1, X86InstInfo{"VPOR",    TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xEC), 1, X86InstInfo{"VPADDSB", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(1, 0b01, 0xED), 1, X86InstInfo{"VPADDSW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xEE), 1, X86InstInfo{"VPMAXSW",  TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(1, 0b01, 0xEF), 1, X86InstInfo{"VPXOR",   TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xEE), 1, X86InstInfo{"VPMAXSW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(1, 0b01, 0xEF), 1, X86InstInfo{"VPXOR",   TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

    {OPD(1, 0b11, 0xF0), 1, X86InstInfo{"VLDDQU",      TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},

@@ -255,9 +255,9 @@ void InitializeVEXTables() {
    {OPD(1, 0b01, 0xF9), 1, X86InstInfo{"VPSUBW", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xFA), 1, X86InstInfo{"VPSUBD", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(1, 0b01, 0xFB), 1, X86InstInfo{"VPSUBQ", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b01, 0xFC), 1, X86InstInfo{"VPADDB", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b01, 0xFD), 1, X86InstInfo{"VPADDW", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
-    {OPD(1, 0b01, 0xFE), 1, X86InstInfo{"VPADDD", TYPE_INST, GenFlagsSameSize(SIZE_256BIT) | FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xFC), 1, X86InstInfo{"VPADDB", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xFD), 1, X86InstInfo{"VPADDW", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},
+    {OPD(1, 0b01, 0xFE), 1, X86InstInfo{"VPADDD", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_VEX_1ST_SRC | FLAGS_XMM_FLAGS, 0, nullptr}},

    // VEX Map 2
    {OPD(2, 0b01, 0x00), 1, X86InstInfo{"VPSHUFB", TYPE_INST, FLAGS_MODRM | FLAGS_XMM_FLAGS, 0, nullptr}},
@@ -298,7 +298,7 @@ void InitializeVEXTables() {

    {OPD(2, 0b01, 0x28), 1, X86InstInfo{"VPMULDQ", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(2, 0b01, 0x29), 1, X86InstInfo{"VPCMPEQQ", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
-    {OPD(2, 0b01, 0x2A), 1, X86InstInfo{"VMOVNTDQA", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
+    {OPD(2, 0b01, 0x2A), 1, X86InstInfo{"VMOVNTDQA", TYPE_INST, GenFlagsSameSize(SIZE_128BIT) | FLAGS_MODRM | FLAGS_SF_MOD_MEM_ONLY | FLAGS_XMM_FLAGS, 0, nullptr}},
    {OPD(2, 0b01, 0x2B), 1, X86InstInfo{"VPACKUSDW", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(2, 0b01, 0x2C), 1, X86InstInfo{"VMASKMOVPS", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
    {OPD(2, 0b01, 0x2D), 1, X86InstInfo{"VMASKMOVPD", TYPE_UNDEC, FLAGS_NONE, 0, nullptr}},
@@ -33,7 +33,7 @@ static inline void GenerateTable(X86InstInfo *FinalTable, X86TablesInfoStruct<Op
    auto OpNum = Op.first;
    X86InstInfo const &Info = Op.Info;
    for (uint32_t i = 0; i < Op.second; ++i) {
-      LOGMAN_THROW_A_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
+      LOGMAN_THROW_AA_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
      FinalTable[OpNum + i] = Info;
 #ifndef NDEBUG
      ++Total;
@@ -51,7 +51,7 @@ static inline void GenerateTableWithCopy(X86InstInfo *FinalTable, X86TablesInfoS
    auto OpNum = Op.first;
    X86InstInfo const &Info = Op.Info;
    for (uint32_t i = 0; i < Op.second; ++i) {
-      LOGMAN_THROW_A_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
+      LOGMAN_THROW_AA_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
      if (Info.Type == TYPE_COPY_OTHER) {
        FinalTable[OpNum + i] = OtherLocal[OpNum + i];
      }
@@ -74,7 +74,7 @@ static inline void GenerateX87Table(X86InstInfo *FinalTable, X86TablesInfoStruct
    auto OpNum = Op.first;
    X86InstInfo const &Info = Op.Info;
    for (uint32_t i = 0; i < Op.second; ++i) {
-      LOGMAN_THROW_A_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
+      LOGMAN_THROW_AA_FMT(FinalTable[OpNum + i].Type == TYPE_UNKNOWN, "Duplicate Entry {}->{}", FinalTable[OpNum + i].Name, Info.Name);
      if ((OpNum & 0b11'000'000) == 0b11'000'000) {
        // If the mod field is 0b11 then it is a regular op
        FinalTable[OpNum + i] = Info;
@@ -82,7 +82,7 @@ static inline void GenerateX87Table(X86InstInfo *FinalTable, X86TablesInfoStruct
      else {
        // If the mod field is !0b11 then this instruction is duplicated through the whole mod [0b00, 0b10] range
        // and the modrm.rm space because that is used part of the instruction encoding
-        LOGMAN_THROW_A_FMT((OpNum & 0b11'000'000) == 0, "Only support mod field of zero in this path");
+        LOGMAN_THROW_AA_FMT((OpNum & 0b11'000'000) == 0, "Only support mod field of zero in this path");
        for (uint16_t mod = 0b00'000'000; mod < 0b11'000'000; mod += 0b01'000'000) {
          for (uint16_t rm = 0b000; rm < 0b1'000; ++rm) {
            FinalTable[(OpNum | mod | rm) + i] = Info;
@@ -28,6 +28,10 @@ $end_info$
 #include <string>
 #include <utility>

+#ifdef ENABLE_JEMALLOC
+#include "jemalloc/jemalloc.h"
+#endif
+
 struct LoadlibArgs {
    const char *Name;
 };
@@ -66,12 +70,23 @@ namespace FEXCore {
    struct ExportEntry { uint8_t *sha256; ThunkedFunction* Fn; };

    struct TrampolineInstanceInfo {
-      uintptr_t HostPacker;
+      void* HostPacker;
      uintptr_t CallCallback;
      uintptr_t GuestUnpacker;
      uintptr_t GuestTarget;
    };

+    // Opaque type pointing to an instance of HostToGuestTrampolineTemplate and its
+    // embedded TrampolineInstanceInfo
+    struct HostToGuestTrampolinePtr;
+    const auto HostToGuestTrampolineSize = __stop_HostToGuestTrampolineTemplate - __start_HostToGuestTrampolineTemplate;
+
+    static TrampolineInstanceInfo& GetInstanceInfo(HostToGuestTrampolinePtr* Trampoline) {
+      const auto Length = __stop_HostToGuestTrampolineTemplate - __start_HostToGuestTrampolineTemplate;
+      const auto InstanceInfoOffset = Length - sizeof(TrampolineInstanceInfo);
+      return *reinterpret_cast<TrampolineInstanceInfo*>(reinterpret_cast<char*>(Trampoline) + InstanceInfoOffset);
+    }
+
    struct GuestcallInfo {
      uintptr_t GuestUnpacker;
      uintptr_t GuestTarget;
@@ -94,7 +109,9 @@ namespace FEXCore {
      }
    };

-    class ThunkHandler_impl final: public ThunkHandler {
+    HostToGuestTrampolinePtr* MakeHostTrampolineForGuestFunction(void* HostPacker, uintptr_t GuestTarget, uintptr_t GuestUnpacker);
+
+    struct ThunkHandler_impl final: public ThunkHandler {
        std::shared_mutex ThunksMutex;

        std::unordered_map<IR::SHA256Sum, ThunkedFunction*, TruncatingSHA256Hash> Thunks = {
@@ -108,23 +125,28 @@ namespace FEXCore {
                { 0xee, 0x57, 0xba, 0x0c, 0x5f, 0x6e, 0xef, 0x2a, 0x8c, 0xb5, 0x19, 0x81, 0xc9, 0x23, 0xe6, 0x51, 0xae, 0x65, 0x02, 0x8f, 0x2b, 0x5d, 0x59, 0x90, 0x6a, 0x7e, 0xe2, 0xe7, 0x1c, 0x33, 0x8a, 0xff },
                &IsLibLoaded
            },
+            {
+                // sha256(fex:is_host_heap_allocation)
+                { 0xf5, 0x77, 0x68, 0x43, 0xbb, 0x6b, 0x28, 0x18, 0x40, 0xb0, 0xdb, 0x8a, 0x66, 0xfb, 0x0e, 0x2d, 0x98, 0xc2, 0xad, 0xe2, 0x5a, 0x18, 0x5a, 0x37, 0x2e, 0x13, 0xc9, 0xe7, 0xb9, 0x8c, 0xa9, 0x3e },
+                &IsHostHeapAllocation
+            },
            {
                // sha256(fex:link_address_to_function)
                { 0xe6, 0xa8, 0xec, 0x1c, 0x7b, 0x74, 0x35, 0x27, 0xe9, 0x4f, 0x5b, 0x6e, 0x2d, 0xc9, 0xa0, 0x27, 0xd6, 0x1f, 0x2b, 0x87, 0x8f, 0x2d, 0x35, 0x50, 0xea, 0x16, 0xb8, 0xc4, 0x5e, 0x42, 0xfd, 0x77 },
                &LinkAddressToGuestFunction
            },
            {
-                // sha256(fex:make_host_trampoline_for_guest_function)
-                { 0x1e, 0x51, 0x6b, 0x07, 0x39, 0xeb, 0x50, 0x59, 0xb3, 0xf3, 0x4f, 0xca, 0xdd, 0x58, 0x37, 0xe9, 0xf0, 0x30, 0xe5, 0x89, 0x81, 0xc7, 0x14, 0xfb, 0x24, 0xf9, 0xba, 0xe7, 0x0e, 0x00, 0x1e, 0x86 },
-                &MakeHostTrampolineForGuestFunction
-            }
+                // sha256(fex:allocate_host_trampoline_for_guest_function)
+                { 0x9b, 0xb2, 0xf4, 0xb4, 0x83, 0x7d, 0x28, 0x93, 0x40, 0xcb, 0xf4, 0x7a, 0x0b, 0x47, 0x85, 0x87, 0xf9, 0xbc, 0xb5, 0x27, 0xca, 0xa6, 0x93, 0xa5, 0xc0, 0x73, 0x27, 0x24, 0xae, 0xc8, 0xb8, 0x5a },
+                &AllocateHostTrampolineForGuestFunction
+            },
        };

        // Can't be a string_view. We need to keep a copy of the library name in-case string_view pointer goes away.
        // Ideally we track when a library has been unloaded and remove it from this set before the memory backing goes away.
        std::set<std::string> Libs;

-        std::unordered_map<GuestcallInfo, uintptr_t, GuestcallInfoHash> GuestcallToHostTrampoline;
+        std::unordered_map<GuestcallInfo, HostToGuestTrampolinePtr*, GuestcallInfoHash> GuestcallToHostTrampoline;

        uint8_t *HostTrampolineInstanceDataPtr;
        size_t HostTrampolineInstanceDataAvailable = 0;
@@ -157,11 +179,11 @@ namespace FEXCore {
            auto args = reinterpret_cast<args_t*>(argsv);
            auto CTX = Thread->CTX;

-            LOGMAN_THROW_A_FMT(args->original_callee, "Tried to link null pointer address to guest function");
-            LOGMAN_THROW_A_FMT(args->target_addr, "Tried to link address to null pointer guest function");
+            LOGMAN_THROW_AA_FMT(args->original_callee, "Tried to link null pointer address to guest function");
+            LOGMAN_THROW_AA_FMT(args->target_addr, "Tried to link address to null pointer guest function");
            if (!CTX->Config.Is64BitMode) {
-                LOGMAN_THROW_A_FMT((args->original_callee >> 32) == 0, "Tried to link 64-bit address in 32-bit mode");
-                LOGMAN_THROW_A_FMT((args->target_addr >> 32) == 0, "Tried to link 64-bit address in 32-bit mode");
+                LOGMAN_THROW_AA_FMT((args->original_callee >> 32) == 0, "Tried to link 64-bit address in 32-bit mode");
+                LOGMAN_THROW_AA_FMT((args->target_addr >> 32) == 0, "Tried to link 64-bit address in 32-bit mode");
            }

            LogMan::Msg::DFmt("Thunks: Adding guest trampoline from address {:#x} to guest function {:#x}",
@@ -177,7 +199,7 @@ namespace FEXCore {

                        const uint8_t GPRSize = CTX->GetGPRSize();

-                        emit->_StoreContext(GPRSize, IR::GPRClass, emit->_Constant(Entrypoint), offsetof(Core::CPUState, gregs[X86State::REG_R11]));
+                        emit->_StoreRegister(emit->_Constant(Entrypoint), false, offsetof(Core::CPUState, gregs[X86State::REG_R11]), IR::GPRClass, IR::GPRFixedClass, GPRSize);
                        emit->_ExitFunction(emit->_Constant(GuestThunkEntrypoint));
                    }, CTX->ThunkHandler.get(), (void*)args->target_addr);

@@ -194,98 +216,44 @@ namespace FEXCore {
        }

        /**
-         * Generates a host-callable trampoline to call guest functions via the host ABI.
+         * Guest-side helper to initiate creation of a host trampoline for
+         * calling guest functions. This must be followed by a host-side call
+         * to FinalizeHostTrampolineForGuestFunction to make the trampoline
+         * usable.
         *
-         * This trampoline uses the same calling convention as the given HostPacker. Trampolines
-         * are cached, so it's safe to call this function repeatedly on the same arguments without
-         * leaking memory.
-         *
-         * Invoking the returned trampoline has the effect of:
-         * - packing the arguments (using the HostPacker identified by its SHA256)
-         * - performing a host->guest transition
-         * - unpacking the arguments via GuestUnpacker
-         * - calling the function at GuestTarget
-         *
-         * The primary use case of this is ensuring that guest function pointers ("callbacks")
-         * passed to thunked APIs can safely be called by the native host library.
+         * This two-step initialization is equivalent to a host-side call to
+         * MakeHostTrampolineForGuestFunction. The split is needed if the
+         * host doesn't have all information needed to create the trampoline
+         * on its own.
         */
-        static void MakeHostTrampolineForGuestFunction(void* ArgsRV) {
-          struct ArgsRV_t {
-              IR::SHA256Sum *HostPackerSha256;
-              uintptr_t GuestUnpacker;
-              uintptr_t GuestTarget;
-              uintptr_t rv; // Pointer to host trampoline + TrampolineInstanceInfo
-          } *args = reinterpret_cast<ArgsRV_t*>(ArgsRV);
+        static void AllocateHostTrampolineForGuestFunction(void* ArgsRV) {
+            struct ArgsRV_t {
+                uintptr_t GuestUnpacker;
+                uintptr_t GuestTarget;
+                uintptr_t rv; // Pointer to host trampoline + TrampolineInstanceInfo
+            } *args = reinterpret_cast<ArgsRV_t*>(ArgsRV);

-          LOGMAN_THROW_A_FMT(args->GuestTarget, "Tried to create host-trampoline to null pointer guest function");
+            args->rv = (uintptr_t)MakeHostTrampolineForGuestFunction(nullptr, args->GuestTarget, args->GuestUnpacker);
+        }

-          const auto CTX = Thread->CTX;
-          const auto ThunkHandler = reinterpret_cast<ThunkHandler_impl *>(CTX->ThunkHandler.get());
+        /**
+         * Checks if the given pointer is allocated on the host heap.
+         *
+         * This is useful for thunking APIs that need to work with both guest
+         * and host heap pointers.
+         */
+        static void IsHostHeapAllocation(void* ArgsRV) {
+#ifdef ENABLE_JEMALLOC
+            struct ArgsRV_t {
+                void* ptr;
+                bool rv;
+            } *args = reinterpret_cast<ArgsRV_t*>(ArgsRV);

-          const GuestcallInfo gci = { args->GuestUnpacker, args->GuestTarget };
-
-          // Try first with shared_lock
-          {
-            std::shared_lock lk(ThunkHandler->ThunksMutex);
-
-            auto found = ThunkHandler->GuestcallToHostTrampoline.find(gci);
-            if (found != ThunkHandler->GuestcallToHostTrampoline.end()) {
-              args->rv = found->second;
-              return;
-            }
-          }
-
-          std::lock_guard lk(ThunkHandler->ThunksMutex);
-
-          // Retry lookup with full lock before making a new trampoline to avoid double trampolines
-          {
-            auto found = ThunkHandler->GuestcallToHostTrampoline.find(gci);
-            if (found != ThunkHandler->GuestcallToHostTrampoline.end()) {
-              args->rv = found->second;
-              return;
-            }
-          }
-
-          // No entry found => create new trampoline
-          auto HostPackerEntry = ThunkHandler->Thunks.find(*args->HostPackerSha256);
-          if (HostPackerEntry == ThunkHandler->Thunks.end()) {
-            ERROR_AND_DIE_FMT("Unknown host packing function for callback");
-          }
-
-          LogMan::Msg::DFmt("Thunks: Adding host trampoline for guest function {:#x}",
-                            args->GuestTarget);
-
-          const auto Length = __stop_HostToGuestTrampolineTemplate - __start_HostToGuestTrampolineTemplate;
-          const auto InstanceInfoOffset = Length - sizeof(TrampolineInstanceInfo);
-
-          if (ThunkHandler->HostTrampolineInstanceDataAvailable < Length) {
-            const auto allocation_step = 16 * 1024;
-            ThunkHandler->HostTrampolineInstanceDataAvailable = allocation_step;
-            ThunkHandler->HostTrampolineInstanceDataPtr = (uint8_t *)mmap(
-                0, ThunkHandler->HostTrampolineInstanceDataAvailable,
-                PROT_READ | PROT_WRITE | PROT_EXEC,
-                MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
-
-            LOGMAN_THROW_A_FMT(ThunkHandler->HostTrampolineInstanceDataPtr != MAP_FAILED, "Failed to mmap HostTrampolineInstanceDataPtr");
-          }
-
-          const TrampolineInstanceInfo NewTrampolineInfo {
-            .HostPacker = reinterpret_cast<uintptr_t>(HostPackerEntry->second),
-            .CallCallback = (uintptr_t)&CallCallback,
-            .GuestUnpacker = args->GuestUnpacker,
-            .GuestTarget = args->GuestTarget
-          };
-
-          uint8_t* const HostTrampoline = ThunkHandler->HostTrampolineInstanceDataPtr;
-          ThunkHandler->HostTrampolineInstanceDataAvailable -= Length;
-          ThunkHandler->HostTrampolineInstanceDataPtr += Length;
-
-          memcpy(HostTrampoline, (void*)&HostToGuestTrampolineTemplate, Length);
-          memcpy(HostTrampoline + InstanceInfoOffset, &NewTrampolineInfo, sizeof(NewTrampolineInfo));
-
-          args->rv = reinterpret_cast<uintptr_t>(HostTrampoline);
-
-          ThunkHandler->GuestcallToHostTrampoline[gci] = args->rv;
+            args->rv = je_is_known_allocation(args->ptr);
+#else
+            // Thunks usage without jemalloc isn't supported
+            ERROR_AND_DIE_FMT("Unsupported: Thunks querying for host heap allocation information");
+#endif
        }

        static void LoadLib(void *ArgsV) {
@@ -352,9 +320,7 @@ namespace FEXCore {
            }
        }

-        public:
-
-        ThunkedFunction* LookupThunk(const IR::SHA256Sum &sha256) {
+        ThunkedFunction* LookupThunk(const IR::SHA256Sum &sha256) override {

            std::shared_lock lk(ThunksMutex);

@@ -367,12 +333,115 @@ namespace FEXCore {
            }
        }

-        void RegisterTLSState(FEXCore::Core::InternalThreadState *Thread) {
+        void RegisterTLSState(FEXCore::Core::InternalThreadState *Thread) override {
            ::Thread = Thread;
        }
+
+        void AppendThunkDefinitions(std::vector<FEXCore::IR::ThunkDefinition> const& Definitions) override {
+          for (auto & Definition : Definitions) {
+            Thunks.emplace(Definition.Sum, Definition.ThunkFunction);
+          }
+        }
    };

    ThunkHandler* ThunkHandler::Create() {
-        return new ThunkHandler_impl();
+      return new ThunkHandler_impl();
+    }
+
+    /**
+     * Generates a host-callable trampoline to call guest functions via the host ABI.
+     *
+     * This trampoline uses the same calling convention as the given HostPacker. Trampolines
+     * are cached, so it's safe to call this function repeatedly on the same arguments without
+     * leaking memory.
+     *
+     * Invoking the returned trampoline has the effect of:
+     * - packing the arguments (using the HostPacker identified by its SHA256)
+     * - performing a host->guest transition
+     * - unpacking the arguments via GuestUnpacker
+     * - calling the function at GuestTarget
+     *
+     * The primary use case of this is ensuring that guest function pointers ("callbacks")
+     * passed to thunked APIs can safely be called by the native host library.
+     *
+     * Returns a pointer to the generated host trampoline and its TrampolineInstanceInfo.
+     *
+     * If HostPacker is zero, the trampoline will be partially initialized and needs to be
+     * finalized with a call to FinalizeHostTrampolineForGuestFunction. A typical use case
+     * is to allocate the trampoline for a given GuestTarget/GuestUnpacker on the guest-side,
+     * and provide the HostPacker host-side.
+     */
+    FEX_DEFAULT_VISIBILITY
+    HostToGuestTrampolinePtr* MakeHostTrampolineForGuestFunction(void* HostPacker, uintptr_t GuestTarget, uintptr_t GuestUnpacker) {
+      LOGMAN_THROW_AA_FMT(GuestTarget, "Tried to create host-trampoline to null pointer guest function");
+
+      const auto CTX = Thread->CTX;
+      const auto ThunkHandler = reinterpret_cast<ThunkHandler_impl *>(CTX->ThunkHandler.get());
+
+      const GuestcallInfo gci = { GuestUnpacker, GuestTarget };
+
+      // Try first with shared_lock
+      {
+        std::shared_lock lk(ThunkHandler->ThunksMutex);
+
+        auto found = ThunkHandler->GuestcallToHostTrampoline.find(gci);
+        if (found != ThunkHandler->GuestcallToHostTrampoline.end()) {
+          return found->second;
+        }
+      }
+
+      std::lock_guard lk(ThunkHandler->ThunksMutex);
+
+      // Retry lookup with full lock before making a new trampoline to avoid double trampolines
+      {
+        auto found = ThunkHandler->GuestcallToHostTrampoline.find(gci);
+        if (found != ThunkHandler->GuestcallToHostTrampoline.end()) {
+          return found->second;
+        }
+      }
+
+      LogMan::Msg::DFmt("Thunks: Adding host trampoline for guest function {:#x} via unpacker {:#x}",
+                        GuestTarget, GuestUnpacker);
+
+      if (ThunkHandler->HostTrampolineInstanceDataAvailable < HostToGuestTrampolineSize) {
+        const auto allocation_step = 16 * 1024;
+        ThunkHandler->HostTrampolineInstanceDataAvailable = allocation_step;
+        ThunkHandler->HostTrampolineInstanceDataPtr = (uint8_t *)mmap(
+            0, ThunkHandler->HostTrampolineInstanceDataAvailable,
+            PROT_READ | PROT_WRITE | PROT_EXEC,
+            MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+
+        LOGMAN_THROW_AA_FMT(ThunkHandler->HostTrampolineInstanceDataPtr != MAP_FAILED, "Failed to mmap HostTrampolineInstanceDataPtr");
+      }
+
+      auto HostTrampoline = reinterpret_cast<HostToGuestTrampolinePtr* const>(ThunkHandler->HostTrampolineInstanceDataPtr);
+      ThunkHandler->HostTrampolineInstanceDataAvailable -= HostToGuestTrampolineSize;
+      ThunkHandler->HostTrampolineInstanceDataPtr += HostToGuestTrampolineSize;
+      memcpy(HostTrampoline, (void*)&HostToGuestTrampolineTemplate, HostToGuestTrampolineSize);
+      GetInstanceInfo(HostTrampoline) = TrampolineInstanceInfo {
+          .HostPacker = HostPacker,
+          .CallCallback = (uintptr_t)&ThunkHandler_impl::CallCallback,
+          .GuestUnpacker = GuestUnpacker,
+          .GuestTarget = GuestTarget
+      };
+
+      ThunkHandler->GuestcallToHostTrampoline[gci] = HostTrampoline;
+      return HostTrampoline;
+    }
+
+    FEX_DEFAULT_VISIBILITY
+    void FinalizeHostTrampolineForGuestFunction(HostToGuestTrampolinePtr* TrampolineAddress, void* HostPacker) {
+
+      if (TrampolineAddress == nullptr) return;
+
+      auto& Trampoline = GetInstanceInfo(TrampolineAddress);
+
+      LOGMAN_THROW_A_FMT(Trampoline.CallCallback == (uintptr_t)&ThunkHandler_impl::CallCallback,
+                        "Invalid trampoline at {} passed to {}", fmt::ptr(TrampolineAddress), __FUNCTION__);
+
+      if (!Trampoline.HostPacker) {
+        LogMan::Msg::DFmt("Thunks: Finalizing trampoline at {} with host packer {}", fmt::ptr(TrampolineAddress), fmt::ptr(HostPacker));
+        Trampoline.HostPacker = HostPacker;
+      }
    }
 }
@@ -6,6 +6,10 @@ $end_info$

 #pragma once

+#include <FEXCore/IR/IR.h>
+
+#include <vector>
+
 namespace FEXCore::Context {
  struct Context;
 }
@@ -28,5 +32,7 @@ namespace FEXCore {
        virtual ~ThunkHandler() { }

        static ThunkHandler* Create();
+
+        virtual void AppendThunkDefinitions(std::vector<FEXCore::IR::ThunkDefinition> const& Definitions) = 0;
    };
 };
@@ -127,7 +127,7 @@ namespace FEXCore::IR {

    auto Array = (AOTIRInlineIndex *)((char*)FilePtr + IndexOffset);

-    LOGMAN_THROW_A_FMT(Entry->Array == nullptr && Entry->FilePtr == nullptr, "Entry must not be initialized here");
+    LOGMAN_THROW_AA_FMT(Entry->Array == nullptr && Entry->FilePtr == nullptr, "Entry must not be initialized here");
    Entry->Array = Array;
    Entry->FilePtr = FilePtr;
    Entry->Size = Size;
@@ -387,7 +387,7 @@ namespace FEXCore::IR {
      auto Inserted = AOTIRCache.insert({fileid, AOTIRCacheEntry { .FileId = fileid, .Filename = filename }});
      auto Entry = &(Inserted.first->second);

-      LOGMAN_THROW_A_FMT(Entry->Array == nullptr, "Duplicate LoadAOTIRCacheEntry");
+      LOGMAN_THROW_AA_FMT(Entry->Array == nullptr, "Duplicate LoadAOTIRCacheEntry");

      if (CTX->Config.AOTIRLoad && AOTIRLoader) {
        auto streamfd = AOTIRLoader(fileid);
@@ -403,7 +403,7 @@ namespace FEXCore::IR {
  }

  void AOTIRCaptureCache::UnloadAOTIRCacheEntry(AOTIRCacheEntry *Entry) {
-    LOGMAN_THROW_A_FMT(Entry != nullptr, "Removing not existing entry");
+    LOGMAN_THROW_AA_FMT(Entry != nullptr, "Removing not existing entry");

    if (Entry->Array) {
      FEXCore::Allocator::munmap(Entry->FilePtr, Entry->Size);
@@ -123,12 +123,12 @@
    "constexpr FEXCore::IR::MemOffsetType MEM_OFFSET_UXTW {1}",
    "constexpr FEXCore::IR::MemOffsetType MEM_OFFSET_SXTW {2}",

-    "constexpr FEXCore::IR::BreakReason Break_Unimplemented      {0}",
-    "constexpr FEXCore::IR::BreakReason Break_Interrupt          {1}",
-    "constexpr FEXCore::IR::BreakReason Break_Interrupt3         {2}",
-    "constexpr FEXCore::IR::BreakReason Break_Halt               {3}",
-    "constexpr FEXCore::IR::BreakReason Break_Overflow           {4}",
-    "constexpr FEXCore::IR::BreakReason Break_InvalidInstruction {5}"
+    "struct BreakDefinition {",
+    "  uint16_t ErrorRegister;",
+    "  uint8_t Signal;",
+    "  uint8_t TrapNumber;",
+    "  uint8_t si_code;",
+    "};"
  ],
  "IRTypes" : {
    "i1":  "bool",
@@ -150,7 +150,7 @@
    "SyscallFlags": "FEXCore::IR::SyscallFlags",
    "SHA256Sum": "SHA256Sum",
    "MemOffsetType": "MemOffsetType",
-    "BreakReason": "BreakReason",
+    "BreakDefinition": "BreakDefinition",
    "RoundType": "RoundType"
  },
  "Ops": {
@@ -192,7 +192,7 @@
        "DestSize": "8"
      },

-      "RemoveThreadCodeEntry": {
+      "ThreadRemoveCodeEntry": {
        "HasSideEffects": true
      },

@@ -261,7 +261,7 @@
        "HasSideEffects": true,
        "DestSize": "GetOpSize(_NewRIP)"
      },
-      "Break BreakReason:$Reason, u8:$Literal": {
+      "Break BreakDefinition:$Reason": {
        "HasSideEffects": true
      },
      "SignalReturn": {
@@ -302,10 +302,6 @@
      }
    },
    "Moves": {
-      "GPR = Mov GPR:$Value": {
-        "DestSize": "GetOpSize(_Value)"
-      },
-
      "GPR = ExtractElementPair GPRPair:$Pair, u8:$Element": {
        "Desc": ["Extracts a register for the register pair"],
        "DestSize": "GetOpSize(_Pair) >> 1"
@@ -359,22 +355,26 @@
        "DestSize": "ByteSize",
        "EmitValidation": [
          "($Class == GPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8)) || $Class == FPRClass",
-          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16)) || $Class == GPRClass"
+          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16 || #ByteSize == 32)) || $Class == GPRClass",
+          "!($Offset >= offsetof(Core::CPUState, gregs[0]) && $Offset < offsetof(Core::CPUState, gregs[16])) && \"Can't LoadContext to GPR\"",
+          "!($Offset >= offsetof(Core::CPUState, xmm.avx.data[0]) && $Offset < offsetof(Core::CPUState, xmm.avx.data[16])) && \"Can't LoadContext to XMM\""
        ]
      },

      "StoreContext u8:#ByteSize, RegisterClass:$Class, SSA:$Value, u32:$Offset": {
-				"Desc": ["Stores a value to the context with offset",
-								 "Ctx[Offset] = Value",
-								 "Zero Extends if value's type is too small",
-								 "Truncates if value's type is too large"
-								],
+        "Desc": ["Stores a value to the context with offset",
+                 "Ctx[Offset] = Value",
+                 "Zero Extends if value's type is too small",
+                 "Truncates if value's type is too large"
+                ],
        "HasSideEffects": true,
        "DestSize": "ByteSize",
        "EmitValidation": [
          "WalkFindRegClass($Value) == $Class",
          "($Class == GPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8)) || $Class == FPRClass",
-          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16)) || $Class == GPRClass"
+          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16 || #ByteSize == 32)) || $Class == GPRClass",
+          "!($Offset >= offsetof(Core::CPUState, gregs[0]) && $Offset < offsetof(Core::CPUState, gregs[16])) && \"Can't StoreContext to GPR\"",
+          "!($Offset >= offsetof(Core::CPUState, xmm.avx.data[0]) && $Offset < offsetof(Core::CPUState, xmm.avx.data[16])) && \"Can't StoreContext to XMM\""
        ]
      },

@@ -385,7 +385,9 @@
        "DestSize": "ByteSize",
        "EmitValidation": [
          "($Class == GPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8)) || $Class == FPRClass",
-          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16)) || $Class == GPRClass"
+          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16 || #ByteSize == 32)) || $Class == GPRClass",
+          "!($BaseOffset >= offsetof(Core::CPUState, gregs[0]) && $BaseOffset < offsetof(Core::CPUState, gregs[16])) && \"Can't LoadContextIndexed to GPR\"",
+          "!($BaseOffset >= offsetof(Core::CPUState, xmm.avx.data[0]) && $BaseOffset < offsetof(Core::CPUState, xmm.avx.data[16])) && \"Can't LoadContextIndexed to XMM\""
        ]
      },
      "StoreContextIndexed SSA:$Value, GPR:$Index, u8:#ByteSize, u32:$BaseOffset, u32:$Stride, RegisterClass:$Class": {
@@ -397,7 +399,9 @@
        "EmitValidation": [
          "WalkFindRegClass($Value) == $Class",
          "($Class == GPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8)) || $Class == FPRClass",
-          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16)) || $Class == GPRClass"
+          "($Class == FPRClass && (#ByteSize == 1 || #ByteSize == 2 || #ByteSize == 4 || #ByteSize == 8 || #ByteSize == 16 || #ByteSize == 32)) || $Class == GPRClass",
+          "!($BaseOffset >= offsetof(Core::CPUState, gregs[0]) && $BaseOffset < offsetof(Core::CPUState, gregs[16])) && \"Can't StoreContextIndexed to GPR\"",
+          "!($BaseOffset >= offsetof(Core::CPUState, xmm.avx.data[0]) && $BaseOffset < offsetof(Core::CPUState, xmm.avx.data[16])) && \"Can't StoreContextIndexed to XMM\""
        ]
      },

@@ -475,22 +479,6 @@
        ]
      },

-      "FPR = VLoadMemElement u8:#RegisterSize, u8:#ElementSize, FPR:$Value, GPR:$Addr, u8:$Index, u8:$Align{1}": {
-        "Desc": ["Loads an element of size #ElementSize in to $Value from $Addr at $Index"
-                ],
-        "OpClass": "Memory",
-        "DestSize": "RegisterSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
-
-      "VStoreMemElement u8:#RegisterSize, u8:#ElementSize, FPR:$Value, GPR:$Addr, u8:$Index, u8:$Align": {
-        "Desc": ["Stores an element of size #ElementSize from $Value[$Index] to $Addr"
-                ],
-        "HasSideEffects": true,
-        "DestSize": "ElementSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
-
      "CacheLineClear GPR:$Addr": {
        "Desc": ["Does a 64 byte cacheline clear at the address specified",
                 "Only clears the data cachelines. Doesn't do any zeroing"
@@ -845,27 +833,25 @@
        "DestSize": "std::max<uint8_t>(4, std::max<uint8_t>(GetOpSize(_TrueVal), GetOpSize(_FalseVal)))"
      },
      "GPR = Extr GPR:$Upper, GPR:$Lower, u8:$LSB": {
-				"Desc": ["Concats the two GPRs to create a value that is the size of the full two GPRs",
-								 "It then extracts a bitfield width that size of a GPR from the LSB",
-								 "Valid LSB range is 0-31 for 32bit and 0-63 for 64bit",
-								 "<Size * 2> ConcatValue = $Upper:$Lower",
-								 "Result = ConcatValue<LSB+Size - 1: LSB>"
-								]
+        "Desc": ["Concats the two GPRs to create a value that is the size of the full two GPRs",
+                 "It then extracts a bitfield width that size of a GPR from the LSB",
+                 "Valid LSB range is 0-31 for 32bit and 0-63 for 64bit",
+                 "<Size * 2> ConcatValue = $Upper:$Lower",
+                 "Result = ConcatValue<LSB+Size - 1: LSB>"
+                ]
      },
      "GPR = PDep GPR:$Input, GPR:$Mask": {
-				"Desc": [
-					"Performs a parallel bit deposit.",
-					"Takes the contiguous low-order bits and deposits them into",
-					"the destination at the locations specified by the Mask."
-				]
+        "Desc": ["Performs a parallel bit deposit.",
+                 "Takes the contiguous low-order bits and deposits them into",
+                 "the destination at the locations specified by the Mask."
+                ]
      },

      "GPR = PExt GPR:$Input, GPR:$Mask": {
-				"Desc": [
-					"Performs a parallel bit extract.",
-					"Each bit set in the mask will select the corresponding bit in the Input",
-					"and transfers them to the lower contiguous bits in the destination."
-				]
+        "Desc": ["Performs a parallel bit extract.",
+                 "Each bit set in the mask will select the corresponding bit in the Input",
+                 "and transfers them to the lower contiguous bits in the destination."
+                ]
      },

      "GPR = LDiv GPR:$Lower, GPR:$Upper, GPR:$Divisor": {
@@ -924,15 +910,6 @@
      }
    },
    "Vector": {
-      "FPR = SplatVector2 FPR:$Scalar": {
-        "NumElements": "2",
-        "DestSize": "GetOpSize(_Scalar) * 2"
-      },
-      "FPR = SplatVector4 FPR:$Scalar": {
-        "NumElements": "4",
-        "DestSize": "GetOpSize(_Scalar) * 4"
-      },
-
      "FPR = VMov u8:#RegisterSize, FPR:$Source": {
        "Desc" : ["Copy vector register",
                  "When Register size is smaller than Source register size,",
@@ -941,12 +918,6 @@
        "DestSize": "RegisterSize"
      },

-      "FPR = VBitcast u8:#RegisterSize, u8:#ElementSize, FPR:$Source": {
-        "Desc": ["Workaround for issue with LLVM breaking when loading scalar elements to vectors"],
-        "DestSize": "RegisterSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
-
      "FPR = VectorZero u8:#RegisterSize": {
        "Desc": ["Generates a vector zero",
                 "Useful to generate a zero vector without any previous dependencies"
@@ -1038,24 +1009,11 @@
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
      },
-      "FPR = VExtractElement u8:#RegisterSize, u8:#ElementSize, FPR:$Vector, u8:$Index": {
-        "DestSize": "ElementSize"
-      },
      "FPR = VDupElement u8:#RegisterSize, u8:#ElementSize, FPR:$Vector, u8:$Index": {
        "Desc": ["Duplicates one element from the source register across the whole register"],
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
      },
-      "FPR = VSLI u8:#RegisterSize, u8:#ElementSize, FPR:$Vector, u8:$ByteShift": {
-        "DestSize": "RegisterSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
-
-      "FPR = VSRI u8:#RegisterSize, u8:#ElementSize, FPR:$Vector, u8:$ByteShift": {
-        "DestSize": "RegisterSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
-
      "FPR = VShlI u8:#RegisterSize, u8:#ElementSize, FPR:$Vector, u8:$BitShift": {
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
@@ -1089,7 +1047,7 @@
      },
      "FPR = VSXTL2 u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": {
        "Desc": ["Sign extends elements from the source element size to the next size up",
-                 "Source elements come from the upper 64bits of the register"
+                 "Source elements come from the upper half of the register"
                ],
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / (ElementSize << 1)"
@@ -1101,7 +1059,7 @@
      },
      "FPR = VUXTL2 u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": {
        "Desc": ["Zero extends elements from the source element size to the next size up",
-                 "Source elements come from the upper 64bits of the register"
+                 "Source elements come from the upper half of the register"
                ],
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / (ElementSize << 1)"
@@ -1124,9 +1082,9 @@
      },

      "FPR = VRev64 u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": {
-				"Desc" : ["Reverses elements in 64-bit halfwords",
-									"Available element size: 1byte, 2 byte, 4 byte"
-								 ],
+        "Desc" : ["Reverses elements in 64-bit halfwords",
+                  "Available element size: 1byte, 2 byte, 4 byte"
+                 ],
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
      },
@@ -1320,10 +1278,6 @@
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
      },
-      "FPR = VInsScalarElement u8:#RegisterSize, u8:#ElementSize, u8:$DestIdx, FPR:$DestVector, FPR:$SrcScalar": {
-        "DestSize": "RegisterSize",
-        "NumElements": "RegisterSize / ElementSize"
-      },
      "FPR = VInsGPR u8:#RegisterSize, u8:#ElementSize, u8:$DestIdx, FPR:$DestVector, GPR:$Src": {
        "DestSize": "RegisterSize",
        "NumElements": "RegisterSize / ElementSize"
@@ -1587,9 +1541,9 @@
        "DestSize": "16"
      },
      "GPR = F80Cmp FPR:$X80Src1, FPR:$X80Src2, u32:$Flags": {
-				"Desc": ["Does a scalar unordered compare and stores the asked for flags in to a GPR",
-								 "Ordering flag result is true if either float input is NaN"
-								],
+        "Desc": ["Does a scalar unordered compare and stores the asked for flags in to a GPR",
+                 "Ordering flag result is true if either float input is NaN"
+                ],
        "DestSize": "4"
      },
      "FPR = F80BCDLoad FPR:$X80Src": {
@@ -182,7 +182,12 @@ static void PrintArg(std::stringstream *out, [[maybe_unused]] IRListView const*
  }
 }

-
+static void PrintArg(std::stringstream *out, [[maybe_unused]] IRListView const* IR, FEXCore::IR::BreakDefinition Arg) {
+  *out << "{" << Arg.ErrorRegister << ".";
+  *out << static_cast<uint32_t>(Arg.Signal) << ".";
+  *out << static_cast<uint32_t>(Arg.TrapNumber) << ".";
+  *out << static_cast<uint32_t>(Arg.si_code) << "}";
+}

 void Dump(std::stringstream *out, IRListView const* IR, IR::RegisterAllocationData *RAData) {
  auto HeaderOp = IR->GetHeader();
@@ -8,6 +8,7 @@ $end_info$
 #include <FEXCore/IR/IR.h>
 #include <FEXCore/IR/IREmitter.h>
 #include <FEXCore/IR/IntrusiveIRList.h>
+#include <FEXCore/Utils/EnumUtils.h>
 #include <FEXCore/Utils/LogManager.h>

 #include <array>
@@ -87,7 +88,8 @@ FEXCore::IR::RegisterClassType IREmitter::WalkFindRegClass(OrderedNode *Node) {
      break;
    }
    default:
-      LOGMAN_MSG_A_FMT("Unhandled op type: {} {} in argument class validation", IROp->Op, GetOpName(Node));
+      LOGMAN_MSG_A_FMT("Unhandled op type: {} {} in argument class validation",
+                       ToUnderlying(IROp->Op), GetOpName(Node));
      break;
  }
  return InvalidClass;
@@ -167,7 +169,7 @@ IREmitter::IRPair<IROp_CodeBlock> IREmitter::CreateNewCodeBlockAfter(OrderedNode
  if (insertAfter) {
    LinkCodeBlocks(insertAfter, CodeNode);
  } else {
-    LOGMAN_THROW_A_FMT(CurrentCodeBlock != nullptr, "CurrentCodeBlock must not be null here");
+    LOGMAN_THROW_AA_FMT(CurrentCodeBlock != nullptr, "CurrentCodeBlock must not be null here");

    // Find last block
    auto LastBlock = CurrentCodeBlock;