Bumped to v0.2.0

Fix PKGBUILDS (#452 )
[DYNAREC] Fixed RCR and RCL opcodes
2026-11-08 17:39:19 -10:00 · 2022-11-17 14:21:42 +01:00 · 2022-11-16 15:51:18 +01:00 · 2022-11-16 15:11:19 +01:00 · 2022-11-15 16:43:34 +01:00 · 2022-11-15 16:31:26 +01:00
651 changed files with 68644 additions and 10162 deletions
@@ -0,0 +1,95 @@
+### Manual build Box64 with Github Action
+name: Custom build Box64
+
+on:
+  workflow_dispatch:
+    inputs:
+      platform:
+        description: 'Target platform name'
+        required: true
+        default: 'GENERIC_ARM'
+        type: choice
+        options:
+        - X64
+        - GENERIC_ARM
+        - PHYTIUM
+        - RK3288
+        - RK3326
+        - RK3399
+        - RPI4ARM64
+        - SD845
+        - TEGRAX1
+      build_type:
+        description: 'Build type'
+        required: true
+        default: 'RelWithDebInfo'
+        type: choice
+        options:
+        - RelWithDebInfo
+        - Release
+        - Debug
+        - MinSizeRel
+      dynarec:
+        description: 'Enable dynarec'
+        required: false
+        default: 'true'
+        type: boolean
+      trace:
+        description: 'Enable trace'
+        required: false
+        default: 'false'
+        type: boolean
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: "Checkout Box64 Repository"
+        uses: actions/checkout@v2
+
+      - name: "Environment preparation"
+        run: |
+          sudo apt-get update
+          if [[ ${{ github.event.inputs.platform }} != 'X64' && ${{ github.event.inputs.platform }} != 'GENERIC_ARM' ]]; then
+            echo "BOX64_PLATFORM_MARCRO=-D${{ github.event.inputs.platform }}=1" >> $GITHUB_ENV
+            echo "BOX64_COMPILER=aarch64-linux-gnu-gcc" >> $GITHUB_ENV
+            sudo apt-get -y install git gcc-aarch64-linux-gnu cmake make python3
+          else
+            if [[ ${{ github.event.inputs.platform }} == 'X64' ]]; then
+              echo "BOX64_PLATFORM_MARCRO=-DLD80BITS=1 -DNOALIGN=1" >> $GITHUB_ENV
+              echo "BOX64_COMPILER=gcc" >> $GITHUB_ENV
+              sudo apt-get -y install git cmake make python3
+            else
+              echo BOX64_PLATFORM_MARCRO="-DARM_DYNAREC=ON" >> $GITHUB_ENV
+              echo "BOX64_COMPILER=aarch64-linux-gnu-gcc" >> $GITHUB_ENV
+              sudo apt-get -y install git gcc-aarch64-linux-gnu cmake make python3
+            fi
+          fi
+
+      - name: "Display Build info"
+        run: |
+          echo "CMake Platform Macro: ${{ env.BOX64_PLATFORM_MARCRO }}"
+          echo "CMake C Compiler: ${{ env.BOX64_COMPILER }}"
+          echo "Build type: ${{ github.event.inputs.build_type }}"
+          echo "Dynarec Enabled: ${{ github.event.inputs.dynarec }}"
+          echo "Trace Enabled: ${{ github.event.inputs.trace }}"
+
+      - name: "Build Box64"
+        run: |
+          mkdir build
+          cd build
+          cmake .. -DCMAKE_C_COMPILER=${{ env.BOX64_COMPILER }} ${{ env.BOX64_PLATFORM_MARCRO }}\
+                   -DCMAKE_BUILD_TYPE=${{ github.event.inputs.build_type }}\
+                   -DARM_DYNAREC:BOOL=${{ github.event.inputs.dynarec }}\
+                   -DHAVE_TRACE:BOOL=${{ github.event.inputs.trace }}\
+                   -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON
+          make -j$(nproc) VERBOSE=1
+
+      - name: "Upload Artifact"
+        uses: actions/upload-artifact@v2
+        with:
+          name: box64-${{ github.event.inputs.platform }}-${{ github.event.inputs.build_type }}
+          path: build/box64
+
+
+
@@ -0,0 +1,104 @@
+### Auto build Box64 and release its binary with Github Action
+name: Build and Release Box64
+
+on:
+  workflow_dispatch:
+  release:
+  push:
+    paths:
+      - "**/*.c"
+      - "**/*.h"
+      - "**/*.S"
+      - "**/*.py"
+      - "CMakeLists.txt"
+      - "**/*.yml"
+  pull_request:
+    types: [assigned, opened, synchronize, reopened]
+    paths:
+      - "**/*.c"
+      - "**/*.h"
+      - "**/*.S"
+      - "**/*.py"
+      - "CMakeLists.txt"
+      - "**/*.yml"
+
+jobs:
+  build:
+    strategy:
+      fail-fast: false
+      matrix:
+        platform: [X64, OTHER_ARM, RISCV, RPI4ARM64, RK3326, RK3399, RK3288, PHYTIUM, SD845]
+        type: [Release, Trace]
+        os: [ubuntu-latest]
+        include:
+          - platform: TEGRAX1
+            type: Release
+            os: ubuntu-18.04
+          - platform: TEGRAX1
+            type: Trace
+            os: ubuntu-18.04
+
+    runs-on: ${{ matrix.os }}
+    steps:
+      - name: "Checkout Box64 Repository"
+        uses: actions/checkout@v2
+
+      - name: "Environment preparation"
+        run: |
+          sudo apt-get update
+          if [[ ${{ matrix.platform }} != 'X64' && ${{ matrix.platform }} != 'OTHER_ARM' && ${{ matrix.platform }} != 'RISCV' ]]; then
+            echo "BOX64_PLATFORM_MARCRO=-D${{ matrix.platform }}=1" >> $GITHUB_ENV
+            echo "BOX64_COMPILER=aarch64-linux-gnu-gcc" >> $GITHUB_ENV
+            sudo apt-get -y install git gcc-aarch64-linux-gnu cmake make python3
+          else
+            if [[ ${{ matrix.platform }} == 'X64' ]]; then
+              echo "BOX64_PLATFORM_MARCRO=-DLD80BITS=1 -DNOALIGN=1" >> $GITHUB_ENV
+              echo "BOX64_COMPILER=gcc" >> $GITHUB_ENV
+              sudo apt-get -y install git cmake make python3
+            elif [[ ${{ matrix.platform }} == 'RISCV' ]]; then
+              echo BOX64_PLATFORM_MARCRO="-DRV64=ON" >> $GITHUB_ENV
+              echo "BOX64_COMPILER=riscv64-linux-gnu-gcc" >> $GITHUB_ENV
+              sudo apt-get -y install git gcc-riscv64-linux-gnu cmake make python3
+            else
+              echo BOX64_PLATFORM_MARCRO="-DARM_DYNAREC=ON" >> $GITHUB_ENV
+              echo "BOX64_COMPILER=aarch64-linux-gnu-gcc" >> $GITHUB_ENV
+              sudo apt-get -y install git gcc-aarch64-linux-gnu cmake make python3
+            fi
+          fi
+          if [[ ${{ matrix.type }} == 'Release' ]]; then
+            echo BOX64_BUILD_TYPE=Release >> $GITHUB_ENV
+            echo BOX64_HAVE_TRACE=0 >> $GITHUB_ENV
+          else
+            echo BOX64_BUILD_TYPE=RelWithDebInfo >> $GITHUB_ENV
+            echo BOX64_HAVE_TRACE=1 >> $GITHUB_ENV
+          fi
+
+      - name: "Display Build info"
+        run: |
+          echo "CMake Platform Macro: ${{ env.BOX64_PLATFORM_MARCRO }}"
+          echo "CMake C Compiler: ${{ env.BOX64_COMPILER }}"
+          echo "Build type: ${{ env.BOX64_BUILD_TYPE }}"
+          echo "Trace Enabled: ${{ env.BOX64_HAVE_TRACE }}"
+
+      - name: "Build Box64"
+        run: |
+          mkdir build
+          cd build
+          cmake .. -DCMAKE_C_COMPILER=${{ env.BOX64_COMPILER }}\
+                   ${{ env.BOX64_PLATFORM_MARCRO }}\
+                   -DCMAKE_BUILD_TYPE=${{ env.BOX64_BUILD_TYPE }}\
+                   -DHAVE_TRACE=${{ env.BOX64_HAVE_TRACE }}\
+                   -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON
+          make -j$(nproc) VERBOSE=1
+
+      - name: "Test Box64"
+        if: ${{ matrix.platform == 'X64' }}
+        run: |
+          cd build
+          ctest -j$(nproc)
+
+      - name: "Upload Artifact"
+        uses: actions/upload-artifact@v2
+        with:
+          name: box64-${{ matrix.platform }}-${{ matrix.type }}
+          path: build/box64
@@ -55,6 +55,7 @@ dkms.conf
 build/
 build*/
 .vscode
+.cache
 .gdb_history
 src/git_head.h
 backup/
@@ -0,0 +1,24 @@
+language: c
+arch:
+  - amd64
+  - arm64-graviton2
+os: linux
+dist: focal
+
+compiler:
+    - gcc
+
+addons:
+  apt:
+    packages:
+      - python3
+
+      #Build steps
+before_script:
+    - mkdir build
+    - cd build
+    - if [ "$TRAVIS_CPU_ARCH" = "arm64" ]; then cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo ; else cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo ; fi
+
+script:
+    - make
+    - ctest --output-on-failure
@@ -2,22 +2,68 @@ cmake_minimum_required(VERSION 3.4)

 cmake_policy(SET CMP0065 NEW)

+SET(CMAKE_C_FLAGS_DEBUG  "-O0 -g")
+SET(CMAKE_CXX_FLAGS_DEBUG  "-O0 -g")
+
 option(RPI4ARM64 "Set to ON if targeting an RaspberryPI4 device with multiarch arm64 and armhf" ${RPI4ARM64})
 option(RK3326 "Set to ON if targeting an Rockchip RK3326 based device" ${RK3326})
 option(RK3399 "Set to ON if targeting an Rockchip RK3399 based device" ${RK3399})
+option(ODROIDN2 "Set to ON if targeting an Odroid-N2 device" ${ODROIDN2})
 option(TEGRAX1 "Set to ON if targeting an Tegra X1 based device" ${TEGRAX1})
+option(PHYTIUM "Set to ON if targeting an Phytium (D2000 or FT2000/4) based device" ${PHYTIUM})
+option(SD845 "Set to ON if targeting a Snapragon 845 based device" ${SD845})
+option(M1 "Set to ON if targeting a AppleM1 running on Asahi computer" ${M1})
+option(LARCH64 "Set to ON if targeting an Loongarch64 based device" ${LARCH64})
+option(RV64 "Set to ON if targeting an RISC-V RV64GC based device" ${RV64})
+option(PPC64LE "Set to ON if targeting an PowerPC 64 LE based device" ${PPC64LE})
+option(LX2160A "Set to ON if targeting an LX2160A based device" ${LX2160A})
 option(USE_CCACHE "Set to ON to use ccache if present in the system" ${USE_CCACHE})
 option(HAVE_TRACE "Set to ON to have Trace ability (needs ZydisInfo library)" ${HAVE_TRACE})
 option(NOLOADADDR "Set to ON to avoid fixing the load address of Box64" OFF)
 option(NOGIT "Set to ON if not building from a git clone repo (like when building from a zip download from github)" ${NOGIT})
-if(RK3399 OR RPI4ARM64 OR RK3326 OR TEGRAX1)
+option(SW64 "Set ON if targeting an SW64 based device" ${SW64})
+
+if(LARCH64)
+    set(LD80BITS OFF CACHE BOOL "")
+    set(NOALIGN OFF CACHE BOOL "")
+    set(ARM_DYNAREC OFF CACHE BOOL "")
+endif()
+if(RV64)
+    set(LD80BITS OFF CACHE BOOL "")
+    set(NOALIGN OFF CACHE BOOL "")
+    set(ARM_DYNAREC OFF CACHE BOOL "")  
+    set(PAGE16K OFF CACHE BOOL "")
+endif()
+if(PPC64LE)
+    set(LD80BITS OFF CACHE BOOL "")
+    set(NOALIGN OFF CACHE BOOL "")
+    set(ARM_DYNAREC OFF CACHE BOOL "")
+    set(PAGE16K OFF CACHE BOOL "")
+endif()
+if(RK3399 OR ODROIDN2 OR RPI4ARM64 OR RK3326 OR TEGRAX1 OR PHYTIUM OR SD845 OR LX2160A OR M1)
    set(LD80BITS OFF CACHE BOOL "")
    set(NOALIGN OFF CACHE BOOL "")
    set(ARM_DYNAREC ON CACHE BOOL "")
 endif()
+if(RK3399 OR ODROIDN2 OR RPI4ARM64 OR RK3326 OR TEGRAX1 OR PHYTIUM OR SD845 OR LX2160A)
+    set(PAGE16K OFF CACHE BOOL "")
+endif()
+if(M1 OR LARCH64)
+    set(PAGE16K ON CACHE BOOL "")
+endif()
+if(SW64)
+    set(LD80BITS OFF CACHE BOOL "")
+    set(NOALIGN OFF CACHE BOOL "")
+    set(PAGE8K ON CACHE BOOL "")
+endif()
+
 option(LD80BITS "Set to ON if host device have 80bits long double (i.e. i386)" ${LD80BITS})
 option(NOALIGN "Set to ON if host device doesn't need re-align (i.e. i386)" ${NOALIGN})
 option(ARM_DYNAREC "Set to ON to use ARM Dynamic Recompilation" ${ARM_DYNAREC})
+option(PAGE8K "Set to ON if host device have PageSize of 8K (instead of 4K)" ${PAGE8K})
+option(PAGE16K "Set to ON if host device have PageSize of 16K (instead of 4K)" ${PAGE16K})
+option(PAGE64K "Set to ON if host device have PageSize of 64K (instead of 4K)" ${PAGE64K})
+option(STATICBUILD "Set to ON to have a static build (Warning, not working)" ${STATICBUILD})

 if(${CMAKE_VERSION} VERSION_LESS "3.12.2")
    find_package(PythonInterp 3)
@@ -59,14 +105,58 @@ elseif(RK3399)
    add_definitions(-DRK3399)
    add_definitions(-pipe -march=armv8-a+crc+simd+crypto -mcpu=cortex-a72+crypto)
    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crc+simd+crypto -mcpu=cortex-a72+crypto")
+elseif(ODROIDN2)
+    add_definitions(-DODROIDN2)
+    add_definitions(-march=armv8-a+crc+simd+crypto -mcpu=cortex-a73.cortex-a53+crypto)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crc+simd+crypto -mcpu=cortex-a73.cortex-a53+crypto")
 elseif(TEGRAX1)
    add_definitions(-DTEGRAX1)
    add_definitions(-pipe -march=armv8-a+crc+simd+crypto -mcpu=cortex-a57+crypto)
    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crc+simd+crypto -mcpu=cortex-a57+crypto")
+elseif(PHYTIUM)
+    add_definitions(-DPHYSIUM)
+    add_definitions(-pipe -march=armv8.1-a+crc+simd+crypto)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8.1-a+crc+simd+crypto+aes")
+elseif(SD845)
+    add_definitions(-DSD845)
+    add_definitions(-pipe -march=armv8.2-a+simd+crypto -mtune=cortex-a75.cortex-a55)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8.2-a+simd+crypto -mtune=cortex-a75.cortex-a55")
+elseif(M1)
+    add_definitions(-DM1)
+    add_definitions(-pipe -march=armv8.5-a+simd+crypto)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8.5-a+simd+crypto")
+elseif(LARCH64)
+    add_definitions(-DLA464)
+    add_definitions(-pipe -march=loongarch64)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=loongarch64")
+elseif(RV64)
+    add_definitions(-DRV64)
+    add_definitions(-pipe -march=rv64gc)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=rv64gc")
+elseif(PPC64LE)
+    add_definitions(-DPPC64LE)
+elseif(LX2160A)
+    add_definitions(-DLX2160A)
+    add_definitions(-pipe -march=armv8-a+crypto+crc -mcpu=cortex-a72+crypto)
+    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crypto+crc -mcpu=cortex-a72+crypto")
+elseif(SW64)
+    add_definitions(-DSW64)
 elseif(ARM_DYNAREC)
    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crc+simd+crypto")
 endif()

+if(PAGE8K)
+    add_definitions(-DPAGE8K)
+endif()
+
+if(PAGE16K)
+    add_definitions(-DPAGE16K)
+endif()
+
+if(PAGE64K)
+    add_definitions(-DPAGE64K)
+endif()
+
 if(NOGIT)
    add_definitions(-DNOGIT)
 endif()
@@ -96,6 +186,10 @@ if(ARM_DYNAREC)
    add_definitions(-DDYNAREC)
    add_definitions(-DARM64)
    enable_language(ASM)
+    include_directories("${BOX64_ROOT}/src/dynarec/arm64")
+    set(DYNAREC ON)
+else()
+    set(DYNAREC OFF)
 endif()

 set(BOX64_ROOT ${CMAKE_SOURCE_DIR})
@@ -129,7 +223,10 @@ set(ELFLOADER_SRC
    "${BOX64_ROOT}/src/box64context.c"
    "${BOX64_ROOT}/src/build_info.c"
    "${BOX64_ROOT}/src/custommem.c"
+    "${BOX64_ROOT}/src/mallochook.c"
+    "${BOX64_ROOT}/src/steam.c"
    "${BOX64_ROOT}/src/dynarec/dynarec.c"
+    "${BOX64_ROOT}/src/elfs/elfdwarf_private.c"
    "${BOX64_ROOT}/src/elfs/elfloader.c"
    "${BOX64_ROOT}/src/elfs/elfparser.c"
    "${BOX64_ROOT}/src/elfs/elfload_dump.c"
@@ -171,6 +268,7 @@ set(ELFLOADER_SRC
    "${BOX64_ROOT}/src/librarian/globalsymbols.c"
    "${BOX64_ROOT}/src/librarian/symbols.c"
    "${BOX64_ROOT}/src/libtools/auxval.c"
+    "${BOX64_ROOT}/src/libtools/obstack.c"
    "${BOX64_ROOT}/src/libtools/myalign.c"
    "${BOX64_ROOT}/src/libtools/sdl1rwops.c"
    "${BOX64_ROOT}/src/libtools/sdl2rwops.c"
@@ -180,6 +278,7 @@ set(ELFLOADER_SRC
    "${BOX64_ROOT}/src/tools/bridge.c"
    "${BOX64_ROOT}/src/tools/callback.c"
    "${BOX64_ROOT}/src/tools/my_cpuid.c"
+    "${BOX64_ROOT}/src/tools/gtkclass.c"
    "${BOX64_ROOT}/src/tools/pathcoll.c"
    "${BOX64_ROOT}/src/tools/fileutils.c"
    "${BOX64_ROOT}/src/tools/wine_tools.c"
@@ -191,6 +290,7 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedalut.c"
    "${BOX64_ROOT}/src/wrapped/wrappedatk.c"
    "${BOX64_ROOT}/src/wrapped/wrappedatkbridge.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedatomic.c"
    "${BOX64_ROOT}/src/wrapped/wrappedatspi.c"
    "${BOX64_ROOT}/src/wrapped/wrappedbz2.c"
    "${BOX64_ROOT}/src/wrapped/wrappedcairo.c"
@@ -198,13 +298,37 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedcrypto.c"
    "${BOX64_ROOT}/src/wrapped/wrappedcurl.c"
    "${BOX64_ROOT}/src/wrapped/wrappeddbus.c"
+    "${BOX64_ROOT}/src/wrapped/wrappeddbusglib1.c"
    "${BOX64_ROOT}/src/wrapped/wrappedexpat.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedfaudio.c"
    "${BOX64_ROOT}/src/wrapped/wrappedflac.c"
    "${BOX64_ROOT}/src/wrapped/wrappedfontconfig.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedfreebl3.c"
    "${BOX64_ROOT}/src/wrapped/wrappedfreetype.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgbm.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgconf2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgcrypt.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgdk3.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgdkpixbuf2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgdkx112.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgio2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedglib2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgmodule2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgmp.c"
    "${BOX64_ROOT}/src/wrapped/wrappedgnutls.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgobject2.c"
    "${BOX64_ROOT}/src/wrapped/wrappedgssapi.c"
    "${BOX64_ROOT}/src/wrapped/wrappedgssapikrb5.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstapp.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstaudio.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstbase.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstgl.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstreamer.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgsttag.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgstvideo.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgthread2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgtk3.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedgtkx112.c"
    "${BOX64_ROOT}/src/wrapped/wrappedkrb5.c"
    "${BOX64_ROOT}/src/wrapped/wrappedldlinux.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibasound.c"
@@ -212,9 +336,11 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedlibcrypt.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibcups.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibdl.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibdrm.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibform.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibformw.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibfuse.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibegl.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibgl.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibglu.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibice.c"
@@ -222,22 +348,33 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedlibncurses.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibncurses6.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibncursesw.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibncursesw6.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibogg.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibpanel.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibpci.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibpcre.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibpthread.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibresolv.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibrt.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibsm.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibsndfile.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibssl.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibtinfo.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibtinfo6.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibusb1.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibuuid.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibva.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibvadrm.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibvawayland.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibvax11.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibvdpau.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibvorbis.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibx11.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibx11xcb.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibxau.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibxcb.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibxcbdri2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlibxcbdri3.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibxcbglx.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibxcbkeysyms.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlibxcbimage.c"
@@ -265,10 +402,17 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedlibz.c"
    "${BOX64_ROOT}/src/wrapped/wrappedlzma.c"
    "${BOX64_ROOT}/src/wrapped/wrappedmpg123.c"
+    "${BOX64_ROOT}/src/wrapped/wrappednotify.c"
+    "${BOX64_ROOT}/src/wrapped/wrappednsl.c"
+    "${BOX64_ROOT}/src/wrapped/wrappednspr4.c"
+    "${BOX64_ROOT}/src/wrapped/wrappednss3.c"
+    "${BOX64_ROOT}/src/wrapped/wrappednssutil3.c"
    "${BOX64_ROOT}/src/wrapped/wrappedopenal.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpango.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpangocairo.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpangoft2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedplc4.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedplds4.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpng16.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpulse.c"
    "${BOX64_ROOT}/src/wrapped/wrappedpulsesimple.c"
@@ -278,27 +422,42 @@ set(WRAPPEDS
    "${BOX64_ROOT}/src/wrapped/wrappedsdl1net.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl1sound.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl1ttf.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedsecret1.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedsoftokn3.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsmpeg.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl2.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl2image.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl2mixer.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl2net.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsdl2ttf.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedselinux.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedsmime3.c"
    "${BOX64_ROOT}/src/wrapped/wrappedsmpeg2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedssl3.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedtbbmallocproxy.c"
    "${BOX64_ROOT}/src/wrapped/wrappedtcmallocminimal.c"
    "${BOX64_ROOT}/src/wrapped/wrappedudev0.c"
    "${BOX64_ROOT}/src/wrapped/wrappedudev1.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedunwind.c"
    "${BOX64_ROOT}/src/wrapped/wrappedutil.c"
    "${BOX64_ROOT}/src/wrapped/wrappedvorbisfile.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedwaylandclient.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedwaylandcursor.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedwaylandegl.c"
    "${BOX64_ROOT}/src/wrapped/wrappedxinerama.c"
    "${BOX64_ROOT}/src/wrapped/wrappedxkbcommon.c"
    "${BOX64_ROOT}/src/wrapped/wrappedxkbcommonx11.c"
    "${BOX64_ROOT}/src/wrapped/wrappedxml2.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedxslt.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedldapr.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedlber.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedvulkan.c"
+    "${BOX64_ROOT}/src/wrapped/wrappedxshmfence.c"
 )

 #set(WRAPPEDS_HEAD "${BOX64_ROOT}/src/wrapped/wrappedd3dadapter9_gen.h")
 foreach(A ${WRAPPEDS})
-    string(REPLACE ".c" "_private.h" B ${A})
+    string(REGEX REPLACE ".c$" "_private.h" B ${A})
    set(WRAPPEDS_HEAD ${WRAPPEDS_HEAD} ${B})
    set_source_files_properties(${A} PROPERTIES OBJECT_DEPENDS ${B})
 endforeach()
@@ -327,77 +486,89 @@ add_custom_command(
 add_custom_target(WRAPPERS DEPENDS "${BOX64_ROOT}/src/wrapped/generated/functions_list.txt")
 #add_custom_target(PRINTER DEPENDS "${BOX64_ROOT}/src/dynarec/last_run.txt")

-if(ARM_DYNAREC)
+if(DYNAREC)
    set(DYNAREC_SRC
        "${BOX64_ROOT}/src/dynarec/dynablock.c"
+        "${BOX64_ROOT}/src/dynarec/dynarec_native.c"
+    )
+endif()

-        "${BOX64_ROOT}/src/dynarec/dynarec_arm64.c"
-        "${BOX64_ROOT}/src/dynarec/dynarec_arm64_functions.c"
-        "${BOX64_ROOT}/src/dynarec/arm64_printer.c"
+if(ARM_DYNAREC)
+    set(DYNAREC_SRC
+        ${DYNAREC_SRC}

-        "${BOX64_ROOT}/src/dynarec/arm64_prolog.S"
-        "${BOX64_ROOT}/src/dynarec/arm64_epilog.S"
-        "${BOX64_ROOT}/src/dynarec/arm64_next.S"
-        "${BOX64_ROOT}/src/dynarec/arm64_lock.S"
+        "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_functions.c"
+        "${BOX64_ROOT}/src/dynarec/arm64/arm64_printer.c"
+        "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_jmpnext.c"
+
+        "${BOX64_ROOT}/src/dynarec/arm64/arm64_prolog.S"
+        "${BOX64_ROOT}/src/dynarec/arm64/arm64_epilog.S"
+        "${BOX64_ROOT}/src/dynarec/arm64/arm64_next.S"
+        "${BOX64_ROOT}/src/dynarec/arm64/arm64_lock.S"
    )

    set(DYNAREC_PASS
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_helper.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_emit_tests.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_emit_math.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_emit_logic.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_emit_shift.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_pass.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_00.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_0f.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_64.c"
-    #"${BOX64_ROOT}/src/dynarec/dynarec_arm64_65.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_66.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_67.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_d8.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_d9.c"
-    #"${BOX64_ROOT}/src/dynarec/dynarec_arm64_da.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_db.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_dc.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_dd.c"
-    #"${BOX64_ROOT}/src/dynarec/dynarec_arm64_de.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_df.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_f0.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_660f.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_6664.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_f20f.c"
-    "${BOX64_ROOT}/src/dynarec/dynarec_arm64_f30f.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_helper.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_emit_tests.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_emit_math.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_emit_logic.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_emit_shift.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_00.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_0f.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_64.c"
+    #"${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_65.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_66.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_67.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_d8.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_d9.c"
+    #"${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_da.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_db.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_dc.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_dd.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_de.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_df.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_f0.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_660f.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_6664.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_66f0.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_f20f.c"
+    "${BOX64_ROOT}/src/dynarec/arm64/dynarec_arm64_f30f.c"
+    )
+endif()
+
+if(DYNAREC)
+    set(DYNAREC_PASS 
+        "${BOX64_ROOT}/src/wrapped/generated/wrapper.h" 
+        ${DYNAREC_PASS}
+        "${BOX64_ROOT}/src/dynarec/dynarec_native_pass.c"
    )

-    set(DYNAREC_PASS "${BOX64_ROOT}/src/wrapped/generated/wrapper.h" ${DYNAREC_PASS})
+    add_library(dynarec_native OBJECT ${DYNAREC_SRC})

-    add_library(dynarec_arm OBJECT ${DYNAREC_SRC})
-
-    add_library(arm64_pass0 OBJECT ${DYNAREC_PASS})
-    set_target_properties(arm64_pass0 PROPERTIES COMPILE_FLAGS "-DSTEP=0")
-    add_library(arm64_pass1 OBJECT ${DYNAREC_PASS})
-    set_target_properties(arm64_pass1 PROPERTIES COMPILE_FLAGS "-DSTEP=1")
-    add_library(arm64_pass2 OBJECT ${DYNAREC_PASS})
-    set_target_properties(arm64_pass2 PROPERTIES COMPILE_FLAGS "-DSTEP=2")
-    add_library(arm64_pass3 OBJECT ${DYNAREC_PASS})
-    set_target_properties(arm64_pass3 PROPERTIES COMPILE_FLAGS "-DSTEP=3")
-    add_dependencies(arm64_pass0 WRAPPERS)
-    add_dependencies(arm64_pass1 WRAPPERS)
-    add_dependencies(arm64_pass2 WRAPPERS)
-    add_dependencies(arm64_pass3 WRAPPERS)
+    add_library(native_pass0 OBJECT ${DYNAREC_PASS})
+    set_target_properties(native_pass0 PROPERTIES COMPILE_FLAGS "-DSTEP=0")
+    add_library(native_pass1 OBJECT ${DYNAREC_PASS})
+    set_target_properties(native_pass1 PROPERTIES COMPILE_FLAGS "-DSTEP=1")
+    add_library(native_pass2 OBJECT ${DYNAREC_PASS})
+    set_target_properties(native_pass2 PROPERTIES COMPILE_FLAGS "-DSTEP=2")
+    add_library(native_pass3 OBJECT ${DYNAREC_PASS})
+    set_target_properties(native_pass3 PROPERTIES COMPILE_FLAGS "-DSTEP=3")
+    add_dependencies(native_pass0 WRAPPERS)
+    add_dependencies(native_pass1 WRAPPERS)
+    add_dependencies(native_pass2 WRAPPERS)
+    add_dependencies(native_pass3 WRAPPERS)

    add_library(dynarec STATIC
-        $<TARGET_OBJECTS:dynarec_arm>
-        $<TARGET_OBJECTS:arm64_pass0>
-        $<TARGET_OBJECTS:arm64_pass1>
-        $<TARGET_OBJECTS:arm64_pass2>
-        $<TARGET_OBJECTS:arm64_pass3>
+        $<TARGET_OBJECTS:dynarec_native>
+        $<TARGET_OBJECTS:native_pass0>
+        $<TARGET_OBJECTS:native_pass1>
+        $<TARGET_OBJECTS:native_pass2>
+        $<TARGET_OBJECTS:native_pass3>
    )
-
 endif()

 # creates git_head.h
-if(ARM_DYNAREC)
+if(DYNAREC)
    add_custom_command(
        OUTPUT "${BOX64_ROOT}/src/git_head.h"
        COMMAND bash -c "echo \\\#define GITREV \\\"$(git rev-parse --short HEAD)\\\">\"${BOX64_ROOT}/src/git_head.h\""
@@ -415,8 +586,13 @@ add_executable(${BOX64} ${ELFLOADER_SRC} ${WRAPPEDS} "${BOX64_ROOT}/src/git_head
 set_target_properties(${BOX64} PROPERTIES ENABLE_EXPORTS ON)
 add_dependencies(${BOX64} WRAPPERS)
 #add_dependencies(${BOX64} PRINTER)
-target_link_libraries(${BOX64} m dl rt pthread)
-if(ARM_DYNAREC)
+#target_link_libraries(${BOX64} c m dl rt pthread resolv)
+if(STATICBUILD)
+    set_target_properties(${BOX64} PROPERTIES LINK_FLAGS "-static -Wl,--no-as-needed -Wl,--whole-archive -Wl,--allow-multiple-definition -lm -ldl -lrt -lpthread -lresolv -lc -Wl,-defsym,_DYNAMIC=0 -pthread")
+else()
+    set_target_properties(${BOX64} PROPERTIES LINK_FLAGS "-Wl,--no-as-needed -lc -lm -ldl -lrt -lpthread -lresolv -Wl,--as-needed -pthread")
+endif()
+if(DYNAREC)
    target_link_libraries(${BOX64} dynarec)
 endif()

@@ -452,8 +628,10 @@ if(NOT _x86 AND NOT _x86_64)
    RUNTIME DESTINATION bin)
  configure_file(system/box64.conf.cmake system/box64.conf)
  install(FILES ${CMAKE_BINARY_DIR}/system/box64.conf DESTINATION /etc/binfmt.d/)
+  install(FILES ${CMAKE_SOURCE_DIR}/x64lib/libstdc++.so.5 DESTINATION /usr/lib/x86_64-linux-gnu/)
  install(FILES ${CMAKE_SOURCE_DIR}/x64lib/libstdc++.so.6 DESTINATION /usr/lib/x86_64-linux-gnu/)
  install(FILES ${CMAKE_SOURCE_DIR}/x64lib/libgcc_s.so.1 DESTINATION /usr/lib/x86_64-linux-gnu/)
+  install(FILES ${CMAKE_SOURCE_DIR}/x64lib/libpng12.so.0 DESTINATION /usr/lib/x86_64-linux-gnu/)
 endif()

 if(NOT TARGET uninstall)
@@ -496,101 +674,123 @@ endif()
 set(CPACK_DEBIAN_FILE_NAME "${BOX64}-${BOX64_MAJOR}.${BOX64_MINOR}.${BOX64_REVISION}_Linux-${CPACK_DEBIAN_PACKAGE_ARCHITECTURE}.deb")
 INCLUDE(CPack)

-add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test01 -D TEST_OUTPUT=tmpfile.txt 
+add_test(bootSyscall ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test01 -D TEST_OUTPUT=tmpfile01.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref01.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test02 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test02 -D TEST_OUTPUT=tmpfile.txt 
+add_test(bootSyscallC ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test02 -D TEST_OUTPUT=tmpfile02.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref02.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test03 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test03 -D TEST_OUTPUT=tmpfile.txt 
+add_test(printf ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test03 -D TEST_OUTPUT=tmpfile03.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref03.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test04 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test04 -D TEST_ARGS2=yeah -D TEST_OUTPUT=tmpfile.txt 
+add_test(args ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test04 -D TEST_ARGS2=yeah -D TEST_OUTPUT=tmpfile04.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref04.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test05 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test05 -D TEST_ARGS2=7 -D TEST_OUTPUT=tmpfile.txt 
+add_test(maths1 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test05 -D TEST_ARGS2=7 -D TEST_OUTPUT=tmpfile05.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref05.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test06 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test06 -D TEST_OUTPUT=tmpfile.txt 
+add_test(threadsStart ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test06 -D TEST_OUTPUT=tmpfile06.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref06.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test07 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test07 -D TEST_OUTPUT=tmpfile.txt 
+add_test(trig ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test07 -D TEST_OUTPUT=tmpfile07.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref07.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test08 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test08 -D TEST_OUTPUT=tmpfile.txt 
+add_test(pi ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test08 -D TEST_OUTPUT=tmpfile08.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref08.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test09 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test09 -D TEST_OUTPUT=tmpfile.txt 
+add_test(fork ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test09 -D TEST_OUTPUT=tmpfile09.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref09.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test10 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test10 -D TEST_OUTPUT=tmpfile.txt 
+add_test(cppThreads ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test10 -D TEST_OUTPUT=tmpfile10.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref10.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test11 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test11 -D TEST_OUTPUT=tmpfile.txt 
+add_test(tlsData ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test11 -D TEST_OUTPUT=tmpfile11.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref11.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test12 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test12 -D TEST_OUTPUT=tmpfile.txt 
+add_test(fpu ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test12 -D TEST_OUTPUT=tmpfile12.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref12.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test13 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test13 -D TEST_OUTPUT=tmpfile.txt 
+add_test(contexts ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test13 -D TEST_OUTPUT=tmpfile13.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref13.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-#add_test(test14 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-#    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test14 -D TEST_OUTPUT=tmpfile.txt 
+#add_test(conditionalThreads ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+#    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test14 -D TEST_OUTPUT=tmpfile14.txt
 #    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref14.txt
 #    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test15 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test15 -D TEST_OUTPUT=tmpfile.txt 
+add_test(linkingIndirectNoversion ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test15 -D TEST_OUTPUT=tmpfile15.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref15.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test16 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test16 -D TEST_OUTPUT=tmpfile.txt 
+add_test(linkingIndirectVersion ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test16 -D TEST_OUTPUT=tmpfile16.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref16.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(test17 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test17 -D TEST_OUTPUT=tmpfile.txt 
+add_test(sse ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test17 -D TEST_OUTPUT=tmpfile17.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref17.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-add_test(aes ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64} 
-    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test18 -D TEST_OUTPUT=tmpfile.txt 
+set_tests_properties(sse PROPERTIES ENVIRONMENT "BOX64_DYNAREC_FASTNAN=0")
+    
+add_test(aes ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test18 -D TEST_OUTPUT=tmpfile18.txt
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref18.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")
+add_test(backtrace ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test19 -D TEST_OUTPUT=tmpfile19.txt
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref19.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(irelative_reloc ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test20 -D TEST_OUTPUT=tmpfile20.txt
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref20.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+    add_test(longjumpInSignals ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test21 -D TEST_OUTPUT=tmpfile21.txt
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref21.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+    add_test(x87 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test22 -D TEST_OUTPUT=tmpfile22.txt
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref22.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+    file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")
 foreach(file ${extension_tests})
    get_filename_component(testname "${file}" NAME_WE)
    add_test(NAME "${testname}" COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX64}
-        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/extensions/${testname} -D TEST_OUTPUT=tmpfile.txt
+        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/extensions/${testname} -D TEST_OUTPUT=tmpfile-${testname}.txt
        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/extensions/${testname}.txt
        -P ${CMAKE_SOURCE_DIR}/runTest.cmake)
 endforeach()
@@ -0,0 +1,29 @@
+box64 (0.1.8) unstable; urgency=low
+
+  * Added an option to get generated -NAN on SSE2 operation
+  * Many small fixes and improvement to get SteamPlay (i.e. proton) working (along with box86)
+  * Added some workaround for program calling uname to mimic x86_64 system
+  * Added some granularity to mmap, and generic memory protection tracking improvment, to mimic x86 system (for wine stagging and proton)
+  * Better clone syscall and libc implementation
+  * Improved wrapping of GTK libs
+  * More wrapped libs (more gtk libs, like gstreamer-1.0 among other non-gtk libs too)
+  * More wrapped functions(with some tricky one, like dladdr1)
+  * Some fixes and improvment to some wrapped functions
+  * Refactor Dynarec:
+  *   - better handling of SSE/x87 registers
+  *   - optimisation to limit float/double conversion
+  *   - better handling of "internal" jump inside a dynablock
+  *   - make adding new architecture easier, by sharing as much infrastructure as possible
+  * Improved elfloader
+  * Improved some tests
+  * Improved a bit the "Hotpage" handling and messages
+  * Added more dynarec opcodes
+  * A few fixes in some Dynarec opcodes
+  * A bit more optimisation in some Dynarec opcodes
+  * Don't quit on unsupported syscall, return -1 / ENOSYS error
+  * Added some more Syscall
+  * Wrapped Vulkan library. Vulkan and dxvk now supported!
+  * Support building on PPC64LE (interpretor only)
+  * Added support for Risc-V (interpretor only)
+
+ -- Sebastien Chevalier <[email protected]>  Tue, 18 Jan 2022 11:20:33 -0500
@@ -0,0 +1 @@
+11
@@ -0,0 +1,14 @@
+Source: box64
+Section: utils
+Priority: optional
+Maintainer: Ryan Fortner <[email protected]>
+Build-Depends: cmake, debhelper (>=11~), python3
+Standards-Version: 4.1.4
+Homepage: https://github.com/ptitSeb/box64
+
+Package: box64
+Architecture: any
+Multi-Arch: foreign
+Depends: ${misc:Depends}, ${shlibs:Depends}
+Description: Linux Userspace x86_64 Emulator with a twist, targeted at ARM Linux devices 
+ Box86 lets you run x86_64 Linux programs (such as games) on non-x86 Linux systems, like ARM (host system needs to be 64bit little-endian).
@@ -0,0 +1,26 @@
+Format-Specification: http://dep.debian.net/deps/dep5/
+
+Files: *
+Copyright: © 2022 Sebastien Chevalier
+License: MIT
+
+License: MIT
+ Copyright (c) 2018-2021 Sebastien Chevalier ("ptitSeb")
+ .
+ Permission is hereby granted, free of charge, to any person obtaining a copy
+ of this software and associated documentation files (the "Software"), to deal
+ in the Software without restriction, including without limitation the rights
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ copies of the Software, and to permit persons to whom the Software is
+ furnished to do so, subject to the following conditions:
+ .
+ The above copyright notice and this permission notice shall be included in
+ all copies or substantial portions of the Software.
+ .
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ THE SOFTWARE.
@@ -0,0 +1 @@
+box64: no-manual-page
@@ -0,0 +1,12 @@
+#!/usr/bin/make -f
+
+%:
+	dh $@  
+
+override_dh_auto_configure:
+	dh_auto_configure -- \
+	      -DCMAKE_BUILD_TYPE=RelWithDebInfo \
+	      -DARM_DYNAREC=1
+
+override_dh_shlibdeps:
+	dh_shlibdeps --exclude=libgcc_s.so.1 --exclude=libpng12.so.0 --exclude=libstdc++.so.6 
@@ -0,0 +1 @@
+3.0 (quilt)
@@ -0,0 +1,5 @@
+Bug-Database: https://github.com/ptitSeb/box64/issues
+Bug-Submit: https://github.com/ptitSeb/box64/issues/new
+Repository: https://github.com/ptitSeb/box64.git
+Repository-Browse: https://github.com/ptitSeb/box64
+Security-Contact: [email protected]
@@ -0,0 +1 @@
+version=3
@@ -1,5 +1,92 @@
-Current Version
+v0.2.0
 ======
+* Wrapped more vulkan function, dxvk 2.0 now works
+* Added support for overriden malloc/free functions
+* Refactor Dynarec memory management, reducing memory footprint (from 20% to 40% of Dynarec size)
+* Improved elf loader
+* Improved dlsym handling with RTLD_NEXT
+* Added BOX64_DYNAREC_SAFEFLAGS to allow finetuning Dynarec generated code handling of flags accross function calls and returns.
+* Added BOX64_BASH to setup x86_64 bash binary, so shell script can be run in x86_64 world 
+* Added BOX64_ROLLING_LOG to have details log only on crash
+* Work on alignment for some of pthread structures
+* More wrapped libraries in the gtk family
+* Some fixes on the Dynarec x87 code (fixing a bunch of games like ut2004 or Piczle Puzzle)
+* Improved TLS Size handling (Unity3D games now works)
+* More PageSize Compile options
+* Improved `execv` family of function to have better box64 integration
+* Added support for wrapped libfuse so AppImage works
+* Adjusted the binfmt integration so AppImage are automaticaly picked
+* Improved FAudio, SDL2, GL (and more) wrapping
+* More wrapped libs, like libEGL.so, to helps emulated Qt4/5 libs to load
+* Improved Signal Handling and memory protection traking
+* Some signature fixes to some wrapped functions (in libc)
+* Added more syscalls
+* Added a lot of new opcodes to the ARM64 Dynarec
+=> Lot's of new games stat working, to name a few: Surviving Mars, Piczle Cross Adventure (using Proton experimental), UT2004, Obduction, Machinarium, Silicon Zero, Hacknet, Absolute Drift, Art of Rally, Dota2
+=> Some apps also runs now, like TeamSpeak 3 Client and Server or Discord
+=> The bash integration is usefull for device with no binfmt integration or for script that check the machine it's running on
+
+v0.1.8
+======
+* Added an option to get generated -NAN on SSE2 operation
+* Many small fixes and improvement to get SteamPlay (i.e. proton) working (along with box86)
+* Added some workaround for program calling uname to mimic x86_64 system
+* Added some granularity to mmap, and generic memory protection tracking improvment, to mimic x86 system (for wine stagging and proton)
+* Better clone syscall and libc implementation
+* Improved wrapping of GTK libs
+* More wrapped libs (more gtk libs, like gstreamer-1.0 among other non-gtk libs too)
+* More wrapped functions(with some tricky one, like dladdr1)
+* Some fixes and improvment to some wrapped functions
+* Refactor Dynarec:
+    * better handling of SSE/x87 registers
+    * optimisation to limit float/double conversion
+    * better handling of "internal" jump inside a dynablock
+    * make adding new architecture easier, by sharing as much infrastructure as possible
+* Improved elfloader
+* Improved some tests
+* Improved a bit the "Hotpage" handling and messages
+* Added more dynarec opcodes
+* A few fixes in some Dynarec opcodes
+* A bit more optimisation in some Dynarec opcodes
+* Don't quit on unsupported syscall, return -1 / ENOSYS error
+* Added some more Syscall
+* Wrapped Vulkan library. Vulkan and dxvk now supported!
+* Support building on PPC64LE (interpretor only)
+* Added support for Risc-V (interpretor only)
+
+v0.1.6
+======
+* Introduce "HotPage", to temporarily disable Dynarec on a page were writing is also occuring (can help speed up C# code)
+* Some work on Dynarec to limit the number of mutex use, and also allow smaller block to be built (for JIT'd programs)
+* Introduce BOX64_DYNAREC_STRONGMEM for the (Arm) dynarec to emulate StrngMemory Model (needed by some games, like RimWorld)
+* More functions wrapped
+* Improve speed of the custom allocator used by dynarec and Hash (speedup loading)
+* Added a workaround for streamwebhelper to not load it
+* More opcodes added
+* More wrapped libraries, including gtk2
+* Added a fix for linking box64 on system using glibc 2.34+
+* Some residual fixes on thread attribute handling
+* Better handling of app that want 32bits jumps, like unity3d/mono (thanks mogery)
+* More ARM hardware option in CMake (Phytium, NX...)
+* Added loongarch64 support (interpretor only)
+* Some fixes to elfloader
+* Added a mecanism to cancel a Dynarec block construction if it triggers a segfault (instead of a crash)
+
+v0.1.4
+======
+* Reworked README a bit
+* Added some more wrapped libs
+* Added some SSE4x opcodes
+* Preliminary support for VSyscall (still missing VDSO)
+* Added support for AES-NI cpu extension, also in Dynarec (and using ARM extension if present)
+* Added some syscall
+* Added some more wrapped functions
+* Improvment to the handling of longjmp inside signal handlers
+* Improved wrapper helpers (from rajdakin)
+* Some fixes to opcodes (especially C2 flags for some x87 opcodes)
+* Handling of some specific opcode as privileged instruction (instead of NOP)
+* Fixed ELF signature detection
+* Some fixes to a few env. var. handling

 v0.1.2
 ======
@@ -1,8 +1,20 @@
-Compiling
+Compiling/Installing
 ----
+
+#### Debian-based Linux 
+You can use [@ryanfortner](https://github.com/ryanfortner)'s apt repository to install precompiled box64 debs, updated every 24 hours. 
+
+```
+sudo wget https://ryanfortner.github.io/box64-debs/box64.list -O /etc/apt/sources.list.d/box64.list
+wget -O- https://ryanfortner.github.io/box64-debs/KEY.gpg | gpg --dearmor | sudo tee /usr/share/keyrings/box64-debs-archive-keyring.gpg 
+sudo apt update && sudo apt install box64 -y
+```
+
+Alternatively, you can generate your own package using the [instructions below](https://github.com/ptitSeb/box64/blob/main/docs/COMPILE.md#debian-packaging). 
+
 #### for RK3399

-Using a 64bits OS:
+Using a 64bit OS:
 ```
 git clone https://github.com/ptitSeb/box64
 cd box64
@@ -17,7 +29,7 @@ sudo systemctl restart systemd-binfmt

 #### for PI4

-Warning, you need a 64bits OS:
+Warning, you need a 64bit OS:

 ```
 git clone https://github.com/ptitSeb/box64
@@ -33,7 +45,7 @@ sudo systemctl restart systemd-binfmt

 #### for TEGRA X1

-Using a 64bits OS:
+Using a 64bit OS:

 ```
 git clone https://github.com/ptitSeb/box64
@@ -47,15 +59,129 @@ If it's the first install, you also need:
 sudo systemctl restart systemd-binfmt
 ```

+#### for Snapdragon 845
+
+Using a 64bit OS:
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DSD845=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+
+----
+#### for Phytium
+
+Using a 64bit OS:
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DPHYTIUM=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+#### for M1
+
+Only test on Asahi for now, using the default "16K page" kernel
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DM1=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+
 #### for Other ARM64 Linux platforms

 `mkdir build; cd build; cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j$(nproc)`

+#### for LoongArch
+
+Using a 64bit OS:
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DLARCH64=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+#### for RISC-V
+
+Using a 64bit OS:
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DRV64=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j`nproc`
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+#### for PowerPC 64 LE
+
+Using a 64bit OS:
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DPPC64LE=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
+#### for LX2160A
+
+Using a 64bit OS:
+
+```
+git clone https://github.com/ptitSeb/box64
+cd box64
+mkdir build; cd build; cmake .. -DLX2160A=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j$(nproc)
+sudo make install
+```
+If it's the first install, you also need:
+```
+sudo systemctl restart systemd-binfmt
+```
+
 #### for x86_64 Linux

 `mkdir build; cd build; cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j$(nproc)`

-If you encounter some linking errors, try activating `NOLOADADDR` (`cmake -DNOLOADADDR=ON; make -j$(nproc)`).
+If you encounter some linking errors, try using `NOLOADADDR=ON` (`cmake -DNOLOADADDR=ON; make -j$(nproc)`).

 ### use ccmake

@@ -63,30 +189,39 @@ Alternatively, you can **use the curses-bases ccmake (or any other gui frontend

 ### Customize your build

-*Use ccache if you have it* 
+#### Use ccache if you have it

 Add `-DUSE_CCACHE=1` if you have ccache (it's better if you plan to touch the sources)

-*To have some debug info* 
+#### To have some debug info

 The `-DCMAKE_BUILD_TYPE=RelWithDebInfo` argument makes a build that is both optimized for speed, and has debug information embedded. That way, if you have a crash or try to analyse performance, you'll have some symbols.

-*To have a Trace Enabled build* 
+#### To have a Trace Enabled build 

-To have a trace enabled build ( ***the interpretor will be slightly slower***), add `-DHAVE_TRACE=1` but you will need, at runtime, to have the [Zydis library](https://github.com/zyantific/zydis) library in your `LD_LIBRARY_PATH` or in the system library folders.
+To have a trace enabled build (***the interpreter will be slightly slower***), add `-DHAVE_TRACE=1`. But you will need to have the [Zydis library](https://github.com/zyantific/zydis) in your `LD_LIBRARY_PATH` or in the system library folders at runtime.

-*To have ARM Dynarec*
+#### To have ARM Dynarec

-The Dynarec is only available on the ARM architecture(Right now, anyways.). Notes also that VFPv3 and NEON are required for the Dynarec. Activate it by using `-DARM_DYNAREC=1`.
+Dynarec is only available on ARM (for the meantime), Activate it by using `-DARM_DYNAREC=1`.
+###### *Note: VFPv3 and NEON are required for Dynarec.*

-*Not building from a git clone*
+#### Not building from a git clone

-If you are not building from a git clone (for example, downloading a release source zipped from github), you need to activate `-DNOGIT=1` from cmake to be able to build (normal process include git sha1 of HEAD in the version that box64 print).
+If you are not building from a git clone (for example, downloading a release source code zip from github), you need to use `-DNOGIT=1` from cmake to be able to build (box64 uses git SHA1 to show last commit in version number).

 ----

 Testing
 ----
-A few tests are included.
-They can be launched with `ctest`
-They are very basic and don't test much for now.
+A few tests are included with box64.
+
+They can be launched using the `ctest` command.
+
+The tests are very basic and only tests some functionality for now.
+
+----
+
+Debian Packaging
+----
+Box64 can also be packaged into a .deb file with `DEB_BUILD_OPTIONS=nostrip dpkg-buildpackage -us -uc -nc`.
@@ -1,17 +1,20 @@
-# box64
-
 ![Official logo](img/Box64Logo.png "Official Logo")
+
 Linux Userspace x86_64 Emulator with a twist

+[View changelog](https://github.com/ptitSeb/box64/blob/main/docs/CHANGELOG.md) | [中文](https://github.com/ptitSeb/box64/blob/main/docs/README_CN.md) | [Report an error](https://github.com/ptitSeb/box64/issues/new)
+
+![build](https://app.travis-ci.com/ptitSeb/box64.svg?branch=main) ![stars](https://img.shields.io/github/stars/ptitSeb/box64) ![forks](https://img.shields.io/github/forks/ptitSeb/box64) ![contributors](https://img.shields.io/github/contributors/ptitSeb/box64) ![prs](https://img.shields.io/github/issues-pr/ptitSeb/box64) ![issues](https://img.shields.io/github/issues/ptitSeb/box64)
+
 ----

-Box64 lets you run x86_64 Linux programs (such as games) on non-x86_64 Linux systems, like ARM (host system needs to be 64bit little-endian).
+Box64 lets you run x86_64 Linux programs (such as games) on non-x86_64 Linux systems, like ARM (host system needs to be 64-bit little-endian).

-You can find many more Box64 video on Youtube in the [MicroLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ), [Pi Labs](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g) or [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA) Channels
+You can find many Box64 videos on the [MicroLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ), [Pi Labs](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g) or [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA) YouTube channels.

-Because Box64 uses the native versions of some "system" libraries, like libc, libm, SDL, and OpenGL, it's easy to integrate and use with most applications, and performance can be surprisingly high in many cases. Take a look at thoses bench analysis for an example [here](https://box86.org/index.php/2021/06/game-performances/).
+Since Box64 uses the native versions of some "system" libraries, like libc, libm, SDL, and OpenGL, it's easy to integrate and use with most applications, and performance can be surprisingly high in many cases. Take a look at thoses bench analysis for an example [here](https://box86.org/index.php/2021/06/game-performances/).

-Box64 integrates a DynaRec (dynamic recompiler) for the ARM64 platform, providing a speed boost between 5 to 10 times faster than only using the interpreter. Some high level information on how the Dynarec work can be found [here](https://box86.org/2021/07/inner-workings-a-high%e2%80%91level-view-of-box86-and-a-low%e2%80%91level-view-of-the-dynarec/).
+Box64 integrates with DynaRec (dynamic recompiler) for the ARM64 platform, providing a speed boost between 5 to 10 times faster than using only the interpreter. Some high level information on how DynaRec works can be found [here](https://box86.org/2021/07/inner-workings-a-high%e2%80%91level-view-of-box86-and-a-low%e2%80%91level-view-of-the-dynarec/).

 Some x64 internal opcodes use parts of "Realmode X86 Emulator Library", see [x64primop.c](../src/emu/x64primop.c) for copyright details

@@ -26,7 +29,7 @@ Usage

 There are a few environment variables to control the behaviour of Box64.

-See [here](USAGE.md) for all environment variables and what they do.
+See [here](USAGE.md) for all the environment variables and what they do.

 Note: Box64's Dynarec uses a mechanism with Memory Protection and a SegFault signal handler to handle JIT code. In simpler terms, if you want to use GDB to debug a running program that use JIT'd code (like mono/Unity3D), you will still have many "normal" segfaults triggering. It is suggested to use something like `handle SIGSEGV nostop` in GDB to not stop at each segfault, and maybe put a breakpoint inside `my_memprotectionhandler` in `signals.c` if you want to trap SegFaults.

@@ -34,77 +37,77 @@ Note: Box64's Dynarec uses a mechanism with Memory Protection and a SegFault sig

 Compiling/Installation
 ----
-> Compilation instructions can be found [here](COMPILE.md)  
+> Compilation instructions can be found [here](COMPILE.md).

 ----

-Version history
+Version history/Change log
 ----

-The change log is available [here](CHANGELOG.md)
+The change log is available [here](CHANGELOG.md).

 ----

 Notes about 32-bit platforms
 ----

-Because Box64 works by directly translating function calls from x86_64 to host system, the host system (the one Box64 is running on) needs to have 64-bits libraries. Box64 doesn't include any 64-bits <-> 32-bits translation. 
+Because Box64 works by directly translating function calls from x86_64 to host system, the host system (the one Box64 is running on) needs to have 64-bit libraries. Box64 doesn't include any 64-bit <-> 32-bit translation. 

-So understand that box64 will only runs 64bits linux binaries. For 32bits binaries, you need box86 (with all the multiarch or proot trickery it imply on 64bits OS).
-Note that many installer (mojo setup based) will fall back to "x86" when detecting ARM64 OS, and so will try to use box86 for the setup, even if an x86_64 version exist. You can hack your way around with a fake "uname" that return "x86_64" when argument is "-m"
+So understand that box64 will only runs 64-bit linux binaries. For 32-bit binaries, you need box86 (with all the multiarch or proot trickery it imply on 64-bit OS).
+Note that many installer (mojo setup based) will fall back to "x86" when detecting ARM64 OS, and so will try to use box86 for the setup, even if an x86_64 version exist. You can hack your way around with a fake "uname" that return "x86_64" when the argument is "-m"

 ----

 Notes about Unity game emulation
 ----

-Running Unity games is a hit or miss for now. Unity uses Mono (which uses signals that are not always emulated in a 100% accurate way).
-You should also note that some Unity3D games require OpenGL 3+ which can be tricky to provide on ARM SBC (single-board computers) for now.
-So, not all Unity games work and can require a high OpenGL profile, but many actualy works now
-Hint: on Pi4, use `MESA_GL_VERSION_OVERRIDE=3.2` and with Panfrost use `PAN_MESA_DEBUG=gl3` to use higher profile if the game start then quit before showing anything.
+Running Unity games should just works, but you should also note that many Unity3D games require OpenGL 3+ which can be tricky to provide on ARM SBC (single-board computers). Also many newer Unity3D (like KSP) games use the BC7 compressed textures, wich is not supported on many ARM integrated GPU.
+Hint: on Pi4, use `MESA_GL_VERSION_OVERRIDE=3.2` and with Panfrost use `PAN_MESA_DEBUG=gl3` to use higher profile if the game starts then quits before showing anything.

 ----

 Notes about GTK programs
 ----

-GTK libraries are not wrapped yet on box64 (contrary to box86). Many games use gtk (setup too), like Unity game for the Screen Selection dialog. While Unity game can run without that dialog (using default setup), many just can't. For now, you'll need o put the set of gtk x86_64 libraries in the game folder (or some other folder known by box64) to get gtk support. GTK Wrapping is planned on box64.
+GTK libraries are now wrapped on box64, both gtk2 and gtk3.

 ----

 Notes about Steam
 ----

-Not that Steam is an hybrid 32buts / 64bits. You NEED box86 to run steam, as the client app is a 32bits binaries. It also use a 64bits local server binaries, but like most stuff using libsef/chromium. it's not working correctly on box64 for now.
-So: no Steam for now on box64.
+Note that Steam is a hybrid 32-bit / 64-bit. You NEED box86 to run Steam, as the client app is a 32-bit binary. It also uses a 64-bit local server binaries, but like most stuff using libcef/chromium, it's not working correctly on box64 for now.
+So, no Steam for now on box64.

 ----

 Notes about Wine
 ----

-Wine 64bits is supported on box64. Be aware that 64bits wine also include 32bits componant, to be able to run 32bits windws programs. The 32bits will need box86 and will not run without it. On a system where both box64 and box86 are present and working, a wine 64bits setup can run both 32bits and 64bits windows program (just use `wine` and `wine64` respectivelly)
+Wine64 is supported on box64. Proton should also work. Be aware that 64-bit Wine also includes 32-bit components, to be able to run 32-bit Windows programs. The 32-bit apps will need box86 and will not run without it. On a system where both box64 and box86 are present and working, a wine 64-bit setup can run both 32-bit and 64-bit Windows programs (just use `wine` and `wine64` respectively).
+Note that the new 32bits PE in 64bits process that the Wine time in currently implementing in Wine 7.+ is not yet supported. I tested Wine 7.5 64bits and it worked, but more recent version will probably just not work yet.

 ----

 Notes about Vulkan
 ----

-Box64 Does not wrap Vulkan libraries yet. It's planned, but for now, vulkan program will not run on box64.
+Box64 wraps Vulkan libraries, but note that it as only been tested with a RX550 card, so some extensions may be missing depending on your graphics card.

 ----

 Final word
 ----

-I want to thank everyone who has contributed to box64 developpement.
+I want to thank everyone who has contributed to box64 development.
 There are many ways to contribute: code contribution, financial, hardware and advertisement!
 So, in no particular order, I want to thank:
- * For their major code contribution: rajdakin
- * For their major financial contribution: FlyingFathead, stormchaser3000
- * For their hardware contribution: [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/)
+ * For their major code contribution: rajdakin, mogery
+ * For their major financial contribution: FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/)
+ * For hardware contribution and LoongArch migration: [xiaoji](https://www.linuxgame.cn/), Deepin Beijing Develop Team
+ * For their hardware contribution: [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [StarFive](https://rvspace.org/)
 * For their continous advertisement of box64 project: salva ([microLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)), [PILab](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)/[TwisterOS](https://twisteros.com/) team, [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA), [NicoD](https://www.youtube.com/channel/UCpv7NFr0-9AB5xoklh3Snhg), ekianjo ([Boilingsteam](https://boilingsteam.com/))

-And I also thank the many other people who participated even once in the project.
+And I also thank the many other people who participated even once in this project.

 (If you use Box64 in your project, please don't forget to mention it!)
@@ -0,0 +1,119 @@
+# box64
+
+![Official logo](img/Box64Logo.png "Official Logo")
+Linux上用户空间的x86_64应用模拟器 
+
+----
+![box64 build status](https://app.travis-ci.com/ptitSeb/box64.svg?branch=main)
+
+[English](README.md)
+
+在ARM或其他非X86_64架构的Linux操作系统上，使用Box64可以运行x86_64 Linux程序（例如游戏），请注意，主机系统需要**64位小端**。
+
+您可以在Youtube的[MicroLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)、[Pi Labs](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)或[The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA)频道找到更多Box64视频。
+
+Box64使用了一些“系统”库的原生版本，例如libc、libm、SDL和OpenGL，所以它很容易与大多数应用程序集成和使用，并且在很多情况下性能会出奇的高。点击[此处](https://box86.org/index.php/2021/06/game-performances/)查看示例的基准测试分析。
+
+Box64为ARM64平台集成了DynaRec（动态重编译器），比仅使用解释器提升5到10倍性能。关于Dynarec如何工作的更多信息，可以在[这里](https://box86.org/2021/07/inner-workings-a-high%e2%80%91level-view-of-box86-and-a-low%e2%80%91level-view-of-the-dynarec/)找到。
+
+注：DynaRec需要针对不同平台编写一部分汇编代码，目前仅支持ARM64。
+
+一些x64内部操作码使用“Realmode X86 Emulator Library”的部分内容，请参阅[x64primop.c](../src/emu/x64primop.c)了解版权详情 
+
+<img src="img/Box64Icon.png" width="96" height="96">
+
+Logo和图标由grayduck制作，感谢!
+
+----
+
+使用方法
+----
+
+若干环境变量可以设置Box64的行为。
+
+点击[这里](USAGE.md)了解所有的环境变量和它们的作用。
+
+请注意：Box64的Dynarec提供了一种机制来处理JIT代码，包含内存保护和段错误信号处理程序。简单地说，如果您想用GDB来调试一个使用JIT代码的运行程序（比如mono/Unity3D），您仍然会触发很多“正常”的段错误。建议在GDB中使用类似于`handle SIGSEGV nostop`的东西，不要在每个段错误时都停止，如果您想捕获段错误，可以在`signals.c`中的`my_memprotectionhandler`中放置一个断点。
+
+----
+
+编译/安装
+----
+> 编译说明可以在[这里](COMPILE.md)查看。  
+
+----
+
+版本历史
+----
+
+版本日志在[这里](CHANGELOG.md)。
+
+----
+
+32位平台的注意事项
+----
+
+Box64的工作原理是直接将函数调用从x86_64转换到主机系统，主机系统（Box64运行的系统）需要有64位的库。Box64并不包括任何64位<->32位的转换。
+
+所以请理解，Box64只能运行64位的Linux二进制文件。对于32位的二进制文件，需要Box86（在64位的操作系统上需要所有的multiarch库或一些proot技巧）。
+
+请注意，许多(基于mojo的)安装程序，即使存在x86_64版本，但在检测到架构为ARM64时会退回到"x86"，这时会尝试使用（如果有）box86进行安装。这时你可以使用一个假的"uname"，当它的参数为"-m"时返回"x86_64"。
+
+----
+
+关于Unity游戏模拟的注意事项
+----
+
+当前Unity游戏模拟是一件碰运气的事情。Unity使用Mono（它的信号表示，并不总能100%成功模拟）。
+
+您还应该注意，一些Unity3D游戏需要OpenGL 3+，目前在ARM SBC（单板计算机）上提供OpenGL 3+是很棘手的。
+
+所以，并不是所有的Unity游戏都能工作，而且可能需要很高的OpenGL配置，但现在很多游戏都能工作。
+
+提示：如果游戏开始后在显示任何东西之前退出，在Pi4上使用`MESA_GL_VERSION_OVERRIDE=3.2`，在Panfrost上使用`PAN_MESA_DEBUG=gl3`来使用更高的配置文件。
+
+----
+
+关于GTK程序的注意事项
+----
+
+GTK 庫現在包裝在 box64 上，但只有 gtk2，而不是 gtk3（與 box86 相反）。
+
+----
+
+关于Steam的注意事项
+----
+
+Steam并不是32位/64位混合的，而是32位的，因此您需要Box86才能运行Steam。Steam也使用64位的本地服务器程序，但像大多数使用libcef/chromium的程序一样，在box64上暂时无法正常工作。
+所以：目前在box64上无法运行Steam。
+----
+
+关于Wine的注意事项
+----
+
+box64支持64位Wine，但请注意，64位的Wine还包含32位的组件，以便能够运行32位的Windws程序。32位程序需要Box86，没有它就不能运行。在Box64和Box86共存的系统上，设置64位Wine可以同时运行32位和64位windows程序（只需分别使用 `wine` 和 `wine64`）。
+
+----
+
+关于Vulkan注意事项
+----
+
+Box64 包裝 Vulkan 庫。但請注意，它僅使用 RX550 卡進行了測試，根據您的圖形卡，可能會缺少某些擴展。
+
+----
+
+结语
+----
+
+我要感谢所有为 box64 开发做出贡献的人。
+贡献方式有很多种：代码贡献、财务帮助、捐赠捐赠和广告合作！
+以下感谢不分先后： 
+ * 代码贡献：rajdakin, mogery
+ * 财务帮助：FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/)
+ * 硬件捐赠：[Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [StarFive](https://rvspace.org/)
+ * 硬件捐赠和LoongArch迁移：[xiaoji](https://www.linuxgame.cn)
+ * 为本项目持续的宣传：salva ([microLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)), [PILab](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)/[TwisterOS](https://twisteros.com/) team, [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA), [NicoD](https://www.youtube.com/channel/UCpv7NFr0-9AB5xoklh3Snhg), ekianjo ([Boilingsteam](https://boilingsteam.com/))
+
+我还要感谢很多人，即使他们只参与了一次本项目。
+
+(如果你的项目中使用了Box64，请不要忘记提及！)
@@ -11,6 +11,12 @@ Controls the Verbosity level of the logs
 * 2: DEBUG : Details a lot of stuff (Example: relocations or functions called).
 * 3: DUMP : All DEBUG plus DUMP of all ELF Info.

+#### BOX64_ROLLING_LOG
+Show last few wrapped function call when a Signal is caught
+ * 0: No last function call printed (Default.)
+ * 1: Last 16 wrapped functions calls printed when a signal is printed. Incompatible with BOX64_LOG>1 (may need BOX64_SHOWSEGV=1 also)
+ * N: Last N wrapped functions calls printed when a signal is printed. Incompatible with BOX64_LOG>1 (may need BOX64_SHOWSEGV=1 also)
+
 #### BOX64_NOBANNER
 Disables Box64 printing its version and build
 * 0 : Enable printing its banner. (Default.)
@@ -31,6 +37,7 @@ Enables/Disables the logging of `dlsym` errors.
 #### BOX64_TRACE_FILE
 Send all log and trace to a file instead of `stdout`
 Also, if name contains `%pid` then this is replaced by the actual PID of box64 instance
+End the filename with `+` to have thetrace appended instead of overwriten
 Use `stderr` to use this instead of default `stdout` 

 #### BOX64_TRACE
@@ -78,6 +85,16 @@ Disable handling of SigILL (to ease debugging mainly).
 * 0 : Let x86 program set sighandler for Illegal Instruction
 * 1 : Disables the handling of SigILL 

+#### BOX64_SHOWSEGV
+Show Segfault signal even if a signal handler is present
+ * 0 : Don"t force show the SIGSEGV analysis (Default.)
+ * 1 : Show SIGSEGV detail, even if a signal handler is present
+
+#### BOX64_SHOWBT
+Show some Backtrace (Nativ e and Emulated) whgen a signal (SEGV, ILL or BUS) is caught
+ * 0 : Don"t show backtraces (Default.)
+ * 1 : Show Backtrace detail (for native, box64 is rename as the x86_64 binary run)
+
 #### BOX64_X11THREADS
 Call XInitThreads when loading X11. (This is mostly for old Loki games with the Loki_Compat library.)
 * 0 : Don't force call XInitThreads. (Default.)
@@ -115,6 +132,35 @@ Enables/Disables trace for generated code.
 Forbid dynablock creation in the interval specified (helpfull for debugging behaviour difference between Dynarec and Interpretor)
 * 0xXXXXXXXX-0xYYYYYYYY : define the interval where dynablock cannot start (inclusive-exclusive)

+#### BOX64_DYNAREC_BIGBLOCK
+Enables/Disables Box64's Dynarec building BigBlock.
+ * 0 : Don't try to build block as big as possible (can help program using lots of thread and a JIT, like C#/Unity) (Default when libmonobdwgc-2.0.so is loaded)
+ * 1 : Build Dynarec block as big as possible (Default.)
+ * 2 : Build Dynarec block bigger (don't stop when block overlaps, but only for blocks in elf memory)
+ * 3 : Build Dynarec block bigger (don't stop when block overlaps, for all type of memory)
+
+#### BOX64_DYNAREC_STRONGMEM
+Enable/Disable simulation of Strong Memory model
+* 0 : Don't try anything special (Default.)
+* 1 : Enable some Memory Barrier when reading from memory (on some MOV opcode) to simulate Strong Memory Model while trying to limit performance impact (Default when libmonobdwgc-2.0.so is loaded)
+* 2 : Enable some Memory Barrier when reading from memory (on some MOV opcode) to simulate Strong Memory Model (enabled with dota2 is detected)
+
+#### BOX64_DYNAREC_X87DOUBLE
+Force the use of Double for x87 emulation
+* 0 : Try to use float when possible for x87 emulation (default, faster)
+* 1 : Only use Double for x87 emulation (slower, may be needed for some specific games, like Crysis)
+
+#### BOX64_DYNAREC_FASTNAN
+Enable/Disable generation of -NAN
+* 0 : Generate -NAN like on x86
+* 1 : Don't do anything special with NAN, to go as fast as possible (default)
+
+#### BOX64_DYNAREC_SAFEFLAGS
+Handling of flags on CALL/RET opcodes
+* 0 : Treat CALL/RET as if it never needs any flags (faster but not advised)
+* 1 : most of RET will need flags, most of CALLS will not (Default)
+* 2 : All CALL/RET will need flags (slower, but might be needed. Automatically enabled for Vara.exe)
+
 #### BOX64_LIBGL
 * libXXXX set the name for libGL (defaults to libGL.so.1).
 * /PATH/TO/libGLXXX : Sets the name and path for libGL
@@ -134,13 +180,50 @@ Allow Box64 to continue even if a library is missing.
 * 0 : Box64 will stop if a library cannot be loaded. (Default.)
 * 1 : Continue even if a needed library cannot be loaded. Unadvised, this will, in most cases, crash later on.

+#### BOX64_PREFER_WRAPPED
+Box64 will use wrapped libs even if the lib is specified with absolute path
+ * 0 : Try to use emulated libs when they are defined with absolute path  (Default.)
+ * 1 : Use Wrapped native libs even if path is absolute
+
+#### BOX64_PREFER_EMULATED
+Box64 will prefer emulated libs first (execpt for glibc, alsa, pulse, GL, vulkan and X11
+ * 0 : Native libs are prefered (Default.)
+ * 1 : Emulated libs are prefered (Default for program running inside pressure-vessel)
+
+#### BOX64_CRASHHANDLER
+Box64 will use a dummy crashhandler.so library
+ * 0 : Use Emulated crashhandler.so library if needed
+ * 1 : Use an internal dummy (completly empty) crashhandler.so library (defaut for steam and dota2)
+
 #### BOX64_NOPULSE
 Disables the load of pulseaudio libraries.
 * 0 : Load pulseaudio libraries if found. (Default.)
 * 1 : Disables the load of pulse audio libraries (libpulse and libpulse-simple), both the native library and the x86 library

-#### BOX64_JITGDB
+#### BOX64_NOGTK
+Disables the loading of wrapped GTK libraries.
+ * 0 : Load wrapped GTK libraries if found. (Default.)
+ * 1 : Disables loading wrapped GTK libraries.

+#### BOX64_NOVULKAN
+Disables the load of vulkan libraries.
+ * 0 : Load vulkan libraries if found.
+ * 1 : Disables the load of vulkan libraries, both the native and the i386 version (can be useful on Pi4, where the vulkan driver is not quite there yet.)
+
+#### BOX64_BASH
+Define x86_64 bash to launch script
+ * yyyy
+ Will use yyyy as x86_64 bash to launch script. yyyy needs to be a full path to a valid x86_64 version of bash
+
+#### BOX64_ENV
+ * XXX=yyyy
+ will add XXX=yyyy env. var.
+
+#### BOX64_ENV1
+ * XXX=yyyy
+ will add XXX=yyyy env. var. and continue with BOX86_ENV2 ... until var doesn't exist
+
+#### BOX64_JITGDB
 * 0 : Just print the Segfault message on segfault (default)
 * 1 : Launch `gdb` when a segfault, bus error or illegal instruction signal is trapped, attached to the offending process and go in an endless loop, waiting.
 When in gdb, you need to find the correct thread yourself (the one with `my_box64signalhandler` in is stack)
@@ -0,0 +1,177 @@
+# Installing Wine64
+Running Wine64/Wine with Box64/Box86 allows x64/x86 Windows programs to run on 64-bit ARM Linux devices.
+ - Box64 needs `wine-amd64` to be installed **manually** on ARM64 (aarch64) devices.  This will run 64-bit Windows programs (aka x86_64).
+ - Box86 needs `wine-i386` to be installed **manually** on ARM32 (armhf) devices (or on ARM64 with multiarch or chroot).  This will run 32-bit Windows programs (aka x86).
+ - _Note that `wine:arm64` and `wine:armhf` will not work with Box64/Box86. Your device's package manager might try to install those if you ask it to install wine._
+
+See installation steps below (in the [Examples](#examples) section).
+
+## Overview
+The general procedure for installing Wine64 & Wine for Box64 & Box86 is to...
+ - Download all the install files for the version of Wine you wish to install
+ - Unzip or dpkg the install files into one folder
+ - Move that folder to the directory that you wish Wine to run from (often `~/wine/` by default)
+ - Go to `/usr/local/bin` and make symlinks or scripts that will point to your main wine binaries.
+ - Boot wine to create a new wineprefix.
+ - Download winetricks (which is just a complicated bash script), make it executable, then copy it to `/usr/local/bin`.
+
+## Examples
+### Installing Wine64 & Wine for Box64 & Box86 on Raspberry Pi OS from WineHQ .deb files
+_Links from the [WineHQ repo](https://dl.winehq.org/wine-builds/debian/dists/)_
+
+This install method allows you to install different versions of Wine64/Wine.  You may install any version/branch of Wine that you wish.
+```
+	# NOTE: Can only run on aarch64 (since box64 can only run on aarch64)
+	# box64 runs wine-amd64, box86 runs wine-i386.
+
+	### User-defined Wine version variables ################
+	# - Replace the variables below with your system's info.
+	# - Note that we need the amd64 version for Box64 even though we're installing it on our ARM processor.
+	# - Note that we need the i386 version for Box86 even though we're installing it on our ARM processor.
+	# - Wine download links from WineHQ: https://dl.winehq.org/wine-builds/
+  
+	local branch="devel" #example: devel, staging, or stable (wine-staging 4.5+ requires libfaudio0:i386)
+	local version="7.1" #example: "7.1"
+	local id="debian" #example: debian, ubuntu
+	local dist="bullseye" #example (for debian): bullseye, buster, jessie, wheezy, ${VERSION_CODENAME}, etc 
+	local tag="-1" #example: -1 (some wine .deb files have -1 tag on the end and some don't)
+
+  ########################################################
+
+	# Clean up any old wine instances
+	wineserver -k # stop any old wine installations from running
+	rm -rf ~/.cache/wine # remove any old wine-mono/wine-gecko install files
+	rm -rf ~/.local/share/applications/wine # remove any old program shortcuts
+
+	# Backup any old wine installs
+	rm -rf ~/wine-old 2>/dev/null; mv ~/wine ~/wine-old 2>/dev/null
+	rm -rf ~/.wine-old 2>/dev/null; mv ~/.wine ~/.wine-old 2>/dev/null
+	sudo mv /usr/local/bin/wine /usr/local/bin/wine-old 2>/dev/null
+	sudo mv /usr/local/bin/wine64 /usr/local/bin/wine-old 2>/dev/null
+	sudo mv /usr/local/bin/wineboot /usr/local/bin/wineboot-old 2>/dev/null
+	sudo mv /usr/local/bin/winecfg /usr/local/bin/winecfg-old 2>/dev/null
+	sudo mv /usr/local/bin/wineserver /usr/local/bin/wineserver-old 2>/dev/null
+
+	# Wine download links from WineHQ: https://dl.winehq.org/wine-builds/
+	LNKA="https://dl.winehq.org/wine-builds/${id}/dists/${dist}/main/binary-amd64/" #amd64-wine links
+	DEB_A1="wine-${branch}-amd64_${version}~${dist}${tag}_amd64.deb" #wine64 main bin
+	DEB_A2="wine-${branch}_${version}~${dist}${tag}_amd64.deb" #wine64 support files (required for wine64 / can work alongside wine_i386 main bin)
+		#DEB_A3="winehq-${branch}_${version}~${dist}${tag}_amd64.deb" #shortcuts & docs
+	LNKB="https://dl.winehq.org/wine-builds/${id}/dists/${dist}/main/binary-i386/" #i386-wine links
+	DEB_B1="wine-${branch}-i386_${version}~${dist}${tag}_i386.deb" #wine_i386 main bin
+	DEB_B2="wine-${branch}_${version}~${dist}${tag}_i386.deb" #wine_i386 support files (required for wine_i386 if no wine64 / CONFLICTS WITH wine64 support files)
+		#DEB_B3="winehq-${branch}_${version}~${dist}${tag}_i386.deb" #shortcuts & docs
+
+	# Install amd64-wine (64-bit) alongside i386-wine (32-bit)
+	echo -e "Downloading wine . . ."
+	wget -q ${LNKA}${DEB_A1} 
+	wget -q ${LNKA}${DEB_A2} 
+	wget -q ${LNKB}${DEB_B1} 
+	echo -e "Extracting wine . . ."
+	dpkg-deb -x ${DEB_A1} wine-installer
+	dpkg-deb -x ${DEB_A2} wine-installer
+	dpkg-deb -x ${DEB_B1} wine-installer
+	echo -e "Installing wine . . ."
+	mv wine-installer/opt/wine* ~/wine
+	
+	# Download wine dependencies
+	# - these packages are needed for running box86/wine-i386 on a 64-bit RPiOS via multiarch
+	sudo dpkg --add-architecture armhf && sudo apt-get update # enable multi-arch
+	sudo apt-get install -y libasound2:armhf libc6:armhf libglib2.0-0:armhf libgphoto2-6:armhf libgphoto2-port12:armhf \
+		libgstreamer-plugins-base1.0-0:armhf libgstreamer1.0-0:armhf libldap-2.4-2:armhf libopenal1:armhf libpcap0.8:armhf \
+		libpulse0:armhf libsane1:armhf libudev1:armhf libusb-1.0-0:armhf libvkd3d1:armhf libx11-6:armhf libxext6:armhf \
+		libasound2-plugins:armhf ocl-icd-libopencl1:armhf libncurses6:armhf libncurses5:armhf libcap2-bin:armhf libcups2:armhf \
+		libdbus-1-3:armhf libfontconfig1:armhf libfreetype6:armhf libglu1-mesa:armhf libglu1:armhf libgnutls30:armhf \
+		libgssapi-krb5-2:armhf libkrb5-3:armhf libodbc1:armhf libosmesa6:armhf libsdl2-2.0-0:armhf libv4l-0:armhf \
+		libxcomposite1:armhf libxcursor1:armhf libxfixes3:armhf libxi6:armhf libxinerama1:armhf libxrandr2:armhf \
+		libxrender1:armhf libxxf86vm1 libc6:armhf libcap2-bin:armhf # to run wine-i386 through box86:armhf on aarch64
+		# This list found by downloading...
+		#	wget https://dl.winehq.org/wine-builds/debian/dists/bullseye/main/binary-i386/wine-devel-i386_7.1~bullseye-1_i386.deb
+		#	wget https://dl.winehq.org/wine-builds/debian/dists/bullseye/main/binary-i386/winehq-devel_7.1~bullseye-1_i386.deb
+		#	wget https://dl.winehq.org/wine-builds/debian/dists/bullseye/main/binary-i386/wine-devel_7.1~bullseye-1_i386.deb
+		# then `dpkg-deb -I package.deb`. Read output, add `:armhf` to packages in dep list, then try installing them on Pi aarch64.
+		
+	# - these packages are needed for running box64/wine-amd64 on RPiOS (box64 only runs on 64-bit OS's)
+	sudo apt-get install -y libasound2:arm64 libc6:arm64 libglib2.0-0:arm64 libgphoto2-6:arm64 libgphoto2-port12:arm64 \
+		libgstreamer-plugins-base1.0-0:arm64 libgstreamer1.0-0:arm64 libldap-2.4-2:arm64 libopenal1:arm64 libpcap0.8:arm64 \
+		libpulse0:arm64 libsane1:arm64 libudev1:arm64 libunwind8:arm64 libusb-1.0-0:arm64 libvkd3d1:arm64 libx11-6:arm64 libxext6:arm64 \
+		ocl-icd-libopencl1:arm64 libasound2-plugins:arm64 libncurses6:arm64 libncurses5:arm64 libcups2:arm64 \
+		libdbus-1-3:arm64 libfontconfig1:arm64 libfreetype6:arm64 libglu1-mesa:arm64 libgnutls30:arm64 \
+		libgssapi-krb5-2:arm64 libjpeg62-turbo:arm64 libkrb5-3:arm64 libodbc1:arm64 libosmesa6:arm64 libsdl2-2.0-0:arm64 libv4l-0:arm64 \
+		libxcomposite1:arm64 libxcursor1:arm64 libxfixes3:arm64 libxi6:arm64 libxinerama1:arm64 libxrandr2:arm64 \
+		libxrender1:arm64 libxxf86vm1:arm64 libc6:arm64 libcap2-bin:arm64
+		# This list found by downloading...
+		#	wget https://dl.winehq.org/wine-builds/debian/dists/bullseye/main/binary-amd64/wine-devel_7.1~bullseye-1_amd64.deb
+		#	wget https://dl.winehq.org/wine-builds/debian/dists/bullseye/main/binary-amd64/wine-devel-amd64_7.1~bullseye-1_amd64.deb
+		# then `dpkg-deb -I package.deb`. Read output, add `:arm64` to packages in dep list, then try installing them on Pi aarch64.	
+
+	# These packages are needed for running wine-staging on RPiOS (Credits: chills340)
+	sudo apt install libstb0 -y
+	cd ~/Downloads
+	wget -r -l1 -np -nd -A "libfaudio0_*~bpo10+1_i386.deb" http://ftp.us.debian.org/debian/pool/main/f/faudio/ # Download libfaudio i386 no matter its version number
+	dpkg-deb -xv libfaudio0_*~bpo10+1_i386.deb libfaudio
+	sudo cp -TRv libfaudio/usr/ /usr/
+	rm libfaudio0_*~bpo10+1_i386.deb # clean up
+	rm -rf libfaudio # clean up
+
+	# Install symlinks
+	sudo ln -s ~/wine/bin/wine /usr/local/bin/wine
+	sudo ln -s ~/wine/bin/wine64 /usr/local/bin/wine64
+	sudo ln -s ~/wine/bin/wineboot /usr/local/bin/wineboot
+	sudo ln -s ~/wine/bin/winecfg /usr/local/bin/winecfg
+	sudo ln -s ~/wine/bin/wineserver /usr/local/bin/wineserver
+	sudo chmod +x /usr/local/bin/wine /usr/local/bin/wine64 /usr/local/bin/wineboot /usr/local/bin/winecfg /usr/local/bin/wineserver
+```
+
+## Installing winetricks
+Winetricks is a bash script which makes it easier to install & configure any desired Windows core system software packages which may be dependencies for certain Windows programs.  You can either install it with `apt`, or manually using the steps below.
+
+```
+sudo apt-get install cabextract -y                                                                   # winetricks needs this installed
+sudo mv /usr/local/bin/winetricks /usr/local/bin/winetricks-old                                      # Backup old winetricks
+cd ~/Downloads && wget https://raw.githubusercontent.com/Winetricks/winetricks/master/src/winetricks # Download
+sudo chmod +x winetricks && sudo mv winetricks /usr/local/bin/                                       # Install
+```
+
+Whenever we run winetricks, we must tell Box86 to supress its info banner, otherwise winetricks will crash.  Box86's info banner can be suppressed using the `BOX86_NOBANNER=1` environment variable before invoking winetricks (example: `BOX86_NOBANNER=1 winetricks`).
+
+If `apt` installed a desktop menu shortcut for winetricks (or if you make your own desktop shortcut for winetricks) then you may need to edit that shortcut to include Box86's BOX86_NOBANNER=1 env var.  With your favorite text editor, edit `/usr/share/applications/winetricks.desktop` and change `Exec=winetricks --gui` to `Exec=env BOX86_NOBANNER=1 winetricks --gui`.
+
+## Example commands
+Creating a 64-bit wineprefix:
+`wine64 wineboot` or `wine wineboot` (wineprefix is created in `~/.wine/`)
+`WINEPREFIX="$HOME/prefix64" wine wineboot` (wineprefix is created in `~/prefix64/`) _Note: You will need to invoke `WINEPREFIX="$HOME/prefix64"` before every command to use this wineprefix._
+
+Creating a 32-bit wineprefix:
+`WINEARCH=win32 wine wineboot` (wineprefix is created in `~/.wine/`)
+
+Fore-quitting wine:
+`wineserver -k`
+
+Running Wine Configuration:
+`winecfg`
+
+Using winetricks:
+`winetricks -q corefonts vcrun2010 dotnet20sp1`  
+_This command will silently install three packages in series: Windows core fonts, VC++ 2010 Runtimes, and .NET 2.0 SP1.  `-q` is the "install silent/quiet" command._
+
+Invoking Box86's logging features (with `BOX86_LOG=1` or similar) will cause winetricks to crash.
+
+For a list of all the different Windows packages & libraries that winetricks can help you install, run `winetricks list-all`  
+
+## Other notes
+
+### Wineprefixes (and Wine initialization)
+When you first run or boot Wine (`wine wineboot`), Wine will create a new user environment within which to install Windows software.  This user environment is called a "wineprefix" (or "wine bottle") and is located (by default) in `~/.wine` (note that Linux folders with a `.` in front of them are "hidden" folders).  Think of a wineprefix as Wine's virtual 'harddrive' where it installs software and saves settings.  Wineprefixes are portable and deletable.  For more Wine documentation, see [WineHQ](https://www.winehq.org/documentation).
+
+If you at any point corrupt something inside your default wineprefix, you can start "fresh" by deleting your `~/.wine` directory (with the `rm -rf ~/.wine` command) and boot wine again to create a new default wineprefix.
+
+### Transplanting wineprefixes
+If software isn't installing in Wine with Box86, but is installing for you in Wine on a regular x86 Linux computer, you can copy a wineprefix from your x86 Linux computer to the device you're running Box86 on. This is most easily done by tarring the `~/.wine` folder on your x86 Linux computer (`tar -cvf winebottle.tar ~/.wine`), transferring the tar file to your device, then un-tarring the tar file (`tar -xf winebottle.tar`) on your device running Box86 & Wine. Tarring the wineprefix preserves any symlinks that are inside it.
+
+### Swapping out different versions of Wine
+Some versions of Wine work better with certain software. It is best to install a version of Wine that is known to work with the software you would like to run. There are three main development branches of Wine you can pick from, referred to as wine-stable, wine-devel, and wine-staging. _Note that the wine-staging branch requires extra installation steps on Raspberry Pi._
+
+Your entire Wine installation can reside within a single folder on your Linux computer. TwisterOS assumes that your Wine installation is located inside the `~/wine/` directory. The actual directory where you put your `wine` folder doesn't matter as long as you have symlinks within the `/usr/local/bin/` directory which point to the `wine` folder so that Linux can find Wine when you type `wine` into the terminal).
+
+You can change which version of Wine you are running simply by renaming your old `wine` and `.wine` folders to something else, then putting a new `wine` folder (containing your new version of Wine) in its place. Running `wine wineboot` again will migrate a wineprefix from an older version of Wine to be usable by the newer version of Wine you just installed.  You can check which version of Wine you're running with the `wine --version` command.
@@ -1,7 +1,7 @@
-# Maintainer: Jai-JAP <parjailu@gmail.com>
+# Maintainer: Jai-JAP <jai.jap.318@gmail.com>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box64-arm64-git
-pkgver=1116.9243ea4
+pkgver=r1495.e6e9fae
 pkgrel=1
 pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
 arch=('aarch64')
@@ -23,7 +23,7 @@ prepare() {
 	cd "${srcdir}/${pkgname%-arm64-git}"
 	if [[ ! -d ./build ]]; then
 		mkdir build && cd build
-		cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo
+		cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
 	fi
 }

@@ -35,10 +35,4 @@ build() {
 package() {
 	cd "$srcdir/${pkgname%-arm64-git}/build"
 	make DESTDIR="${pkgdir}/" install
-	# /usr/local/bin isn't in PATH by the default,
-	# we should move it to /usr/bin
-	cd ${pkgdir}
-	mv usr/local/bin/ usr/bin/
-	# cleanup when dir is empty
-	rmdir usr/local || exit 0
 }
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-arm64-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-arm64-git}-git" "${pkgname%-arm64-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-arm64-git}-git" "${pkgname%-arm64-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-arm64-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-arm64-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-arm64-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-arm64-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-lx2160a-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-lx2160a-git}-git" "${pkgname%-lx2160a-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-lx2160a-git}-git" "${pkgname%-lx2160a-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-lx2160a-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-lx2160a-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DLX2160A=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-lx2160a-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-lx2160a-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-lx2160a-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-lx2160a-git}-git" "${pkgname%-lx2160a-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-lx2160a-git}-git" "${pkgname%-lx2160a-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-lx2160a-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-lx2160a-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DLX2160A=1 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-lx2160a-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-lx2160a-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
-# Maintainer: Jai-JAP <parjailu@gmail.com>
+# Maintainer: Jai-JAP <jai.jap.318@gmail.com>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-rk3399-git
-pkgver=1116.9243ea4
+pkgver=r1495.e6e9fae
 pkgrel=1
 pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
 arch=('aarch64')
@@ -23,7 +23,7 @@ prepare() {
 	cd "${srcdir}/${pkgname%-rk3399-git}"
 	if [[ ! -d ./build ]]; then
 		mkdir build && cd build
-                cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+                cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
 	fi
 }

@@ -35,10 +35,4 @@ build() {
 package() {
 	cd "$srcdir/${pkgname%-rk3399-git}/build"
 	make DESTDIR="${pkgdir}/" install
-	# /usr/local/bin isn't in PATH by the default,
-	# we should move it to /usr/bin
-	cd ${pkgdir}
-	mv usr/local/bin/ usr/bin/
-	# cleanup when dir is empty
-	rmdir usr/local || exit 0
-}
+}
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-rk3399-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-rk3399-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-rk3399-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+                cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-rk3399-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-rk3399-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
-# Maintainer: Jai-JAP <parjailu@gmail.com>
+# Maintainer: Jai-JAP <jai.jap.318@gmail.com>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box64-rpi-git
-pkgver=1116.9243ea4
+pkgver=r1495.e6e9fae
 pkgrel=1
 pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
 arch=('aarch64')
@@ -23,7 +23,7 @@ prepare() {
 	cd "${srcdir}/${pkgname%-rpi-git}"
 	if [[ ! -d ./build ]]; then
 		mkdir build && cd build
-		cmake .. -DRPI4ARM64=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+		cmake .. -DRPI4ARM64=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
 	fi
 }

@@ -35,10 +35,4 @@ build() {
 package() {
 	cd "$srcdir/${pkgname%-rpi-git}/build"
 	make DESTDIR="${pkgdir}/" install
-	# /usr/local/bin isn't in PATH by the default,
-	# we should move it to /usr/bin
-	cd ${pkgdir}
-	mv usr/local/bin/ usr/bin/
-	# cleanup when dir is empty
-	rmdir usr/local || exit 0
 }
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-rpi-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-rpi-git}-git" "${pkgname%-rpi-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-rpi-git}-git" "${pkgname%-rpi-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-rpi-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-rpi-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DRPI4ARM64=1 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-rpi-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-rpi-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
-# Maintainer: Jai-JAP <parjailu@gmail.com>
+# Maintainer: Jai-JAP <jai.jap.318@gmail.com>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box64-tegra-git
-pkgver=1116.9243ea4
+pkgver=r1495.e6e9fae
 pkgrel=1
 pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
 arch=('aarch64')
@@ -23,7 +23,7 @@ prepare() {
 	cd "${srcdir}/${pkgname%-tegra-git}"
 	if [[ ! -d ./build ]]; then
 		mkdir build && cd build
-		cmake .. -DTEGRAX1=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+		cmake .. -DTEGRAX1=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
 	fi
 }

@@ -35,10 +35,4 @@ build() {
 package() {
 	cd "$srcdir/${pkgname%-tegra-git}/build"
 	make DESTDIR="${pkgdir}/" install
-	# /usr/local/bin isn't in PATH by the default,
-	# we should move it to /usr/bin
-	cd ${pkgdir}
-	mv usr/local/bin/ usr/bin/
-	# cleanup when dir is empty
-	rmdir usr/local || exit 0
 }
@@ -0,0 +1,38 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-tegra-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-tegra-git}-git" "${pkgname%-tegra-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-tegra-git}-git" "${pkgname%-tegra-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-tegra-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-tegra-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DTEGRAX1=1 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-tegra-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-tegra-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
-# Maintainer: Jai-JAP <parjailu@gmail.com>
+# Maintainer: Jai-JAP <jai.jap.318@gmail.com>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box64-x64-git
-pkgver=1116.9243ea4
+pkgver=r1495.e6e9fae
 pkgrel=1
 pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
 arch=('aarch64')
@@ -22,7 +22,7 @@ prepare() {
 	cd "${srcdir}/${pkgname%-x64-git}"
 	if [[ ! -d ./build ]]; then
 		mkdir build && cd build
-		cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+		cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
 	fi
 }

@@ -34,10 +34,4 @@ build() {
 package() {
 	cd "$srcdir/${pkgname%-x64-git}/build"
 	make DESTDIR="${pkgdir}/" install
-	# /usr/local/bin isn't in PATH by the default,
-	# we should move it to /usr/bin
-	cd ${pkgdir}
-	mv usr/local/bin/ usr/bin/
-	# cleanup when dir is empty
-	rmdir usr/local || exit 0
 }
@@ -0,0 +1,37 @@
+# Maintainer: Jai-JAP <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box64-x64-git
+pkgver=r1495.e6e9fae
+pkgrel=1
+pkgdesc="Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box64"
+license=('MIT')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-x64-git}-git" "${pkgname%-x64-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-x64-git}-git" "${pkgname%-x64-git}")
+source=('git+https://github.com/ptitSeb/box64')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-x64-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-x64-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/usr
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-x64-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-x64-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -12,13 +12,13 @@ This will copy the PKGBUILD for your platform (where you type your platform name
 In the future, I might try to do an unified PKGBUILD for all platforms that will recognize your platform automatically.

 ### 2. What platforms are currently supported.
-All specified in [COMPILE.md](../COMPILE.md).
+All specified in [COMPILE.md](../docs/COMPILE.md).

-### 3. Any advantages with that sollution?
-This might give you some advantages like those:
+### 3. Any advantages with this solution?
+This might give you some advantages like:
 - automatic updates (with some AUR frontends like [`pamac-aur`](https://aur.archlinux.org/packages/pamac-aur/) or [`yay`](https://aur.archlinux.org/packages/yay));
 - easily uninstall with `pacman -R` when no needed (no need other scripts nor Makefile to do this);
 - your `box64` installation is recognized as a package (where it wasn't when installing with `make install`).


-### Note: You can use [makedeb](https://github.com/makedeb/makedeb) (or) [pkgbuild2spec](https://github.com/prozum/pkgbuild2spec) to compile box64 to .deb or .rpm installer files.
+### Note: You can use [makedeb](https://github.com/makedeb/makedeb) (or) [pkgbuild2spec](https://github.com/prozum/pkgbuild2spec) to compile box64 to .deb or .rpm installer files.
@@ -388,7 +388,10 @@ def readFiles(files: Iterable[Filename]) -> Tuple[JumbledGlobals, JumbledRedirec
 							 or match("sdl1mixer", "sdl2mixer") \
 							 or match("sdl1net", "sdl2net") \
 							 or match("sdl1ttf", "sdl2ttf") \
-							 or match("libc", "tcmallocminimal"):
+							 or match("libGL", "libEGL") \
+							 or match("libc", "tcmallocminimal") \
+							 or match("libc", "tbbmallocproxy") \
+							 or match("tcmallocminimal", "tbbmallocproxy"):
 								continue
 							
 							# Note: this test is very (too) simple. If it ever raises, comment
@@ -32,6 +32,9 @@ execute_process(

 # if the return value is !=0 bail out
 if( TEST_RESULT )
+  get_filename_component(TESTNAME "${TEST_ARGS}" NAME)
+  file(RENAME "${TEST_OUTPUT}" "${CMAKE_BINARY_DIR}/${TESTNAME}.out")
+  file(WRITE  "${CMAKE_BINARY_DIR}/${TESTNAME}.err" ${TEST_ERROR})
  message( FATAL_ERROR "Failed: Test program ${TEST_PROGRAM} exited != 0.\n${TEST_ERROR}" )
 endif( TEST_RESULT )

@@ -43,8 +46,16 @@ execute_process(

 # again, if return value is !=0 scream and shout
 if( TEST_RESULT )
+  get_filename_component(TESTNAME "${TEST_ARGS}" NAME)
+  file(RENAME "${TEST_OUTPUT}" "${CMAKE_BINARY_DIR}/${TESTNAME}.out")
+  file(WRITE  "${CMAKE_BINARY_DIR}/${TESTNAME}.err" ${TEST_ERROR})
  message( FATAL_ERROR "Failed: The output of ${TEST_PROGRAM} did not match ${TEST_REFERENCE}")
 endif( TEST_RESULT )

+# remove the temporary files if they exist
+if( EXISTS ${TEST_OUTPUT} )
+  file(REMOVE "${TEST_OUTPUT}")
+endif()
+
 # everything went fine...
 message( "Passed: The output of ${TEST_PROGRAM} matches ${TEST_REFERENCE}" )
@@ -5,6 +5,7 @@
 #include <dlfcn.h>
 #include <signal.h>
 #include <sys/mman.h>
+#include <pthread.h>

 #include "box64context.h"
 #include "debug.h"
@@ -62,8 +63,8 @@ void free_tlsdatasize(void* p)
    if(!p)
        return;
    tlsdatasize_t *data = (tlsdatasize_t*)p;
-    free(data->tlsdata);
-    free(p);
+    box_free(data->ptr);
+    box_free(p);
 }

 void x64Syscall(x64emu_t *emu);
@@ -78,8 +79,6 @@ int unlockMutex()
            ret|=(1<<B);                \
        }

-    GO(my_context->mutex_once, 5)
-    GO(my_context->mutex_once2, 6)
    GO(my_context->mutex_trace, 7)
    #ifdef DYNAREC
    GO(my_context->mutex_dyndump, 8)
@@ -99,10 +98,8 @@ void relockMutex(int locks)
    relockCustommemMutex(locks);
    #define GO(A, B)                    \
        if(locks&(1<<B))                \
-            pthread_mutex_lock(&A);     \
+            pthread_mutex_trylock(&A);  \

-    GO(my_context->mutex_once, 5)
-    GO(my_context->mutex_once2, 6)
    GO(my_context->mutex_trace, 7)
    #ifdef DYNAREC
    GO(my_context->mutex_dyndump, 8)
@@ -120,8 +117,6 @@ static void init_mutexes(box64context_t* context)
    pthread_mutexattr_t attr;
    pthread_mutexattr_init(&attr);
    pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_ERRORCHECK);
-    pthread_mutex_init(&context->mutex_once, &attr);
-    pthread_mutex_init(&context->mutex_once2, &attr);
    pthread_mutex_init(&context->mutex_trace, &attr);
 #ifndef DYNAREC
    pthread_mutex_init(&context->mutex_lock, &attr);
@@ -141,6 +136,31 @@ static void atfork_child_box64context(void)
    init_mutexes(my_context);
 }

+void freeCycleLog(box64context_t* ctx)
+{
+    if(cycle_log) {
+        for(int i=0; i<cycle_log; ++i) {
+            box_free(ctx->log_call[i]);
+            box_free(ctx->log_ret[i]);
+        }
+        box_free(ctx->log_call);
+        box_free(ctx->log_ret);
+        ctx->log_call = NULL;
+        ctx->log_ret = NULL;
+    }
+}
+void initCycleLog(box64context_t* context)
+{
+    if(cycle_log) {
+        context->log_call = (char**)box_calloc(cycle_log, sizeof(char*));
+        context->log_ret = (char**)box_calloc(cycle_log, sizeof(char*));
+        for(int i=0; i<cycle_log; ++i) {
+            context->log_call[i] = (char*)box_calloc(256, 1);
+            context->log_ret[i] = (char*)box_calloc(128, 1);
+        }
+    }
+}
+
 EXPORTDYN
 box64context_t *NewBox64Context(int argc)
 {
@@ -151,7 +171,9 @@ box64context_t *NewBox64Context(int argc)
    }
 #endif
    // init and put default values
-    box64context_t *context = my_context = (box64context_t*)calloc(1, sizeof(box64context_t));
+    box64context_t *context = my_context = (box64context_t*)box_calloc(1, sizeof(box64context_t));
+
+    initCycleLog(context);

    context->deferedInit = 1;
    context->sel_serial = 1;
@@ -162,6 +184,8 @@ box64context_t *NewBox64Context(int argc)
    context->local_maplib = NewLibrarian(context, 1);
    context->versym = NewDictionnary();
    context->system = NewBridge();
+    context->globaldefver = NewDefaultVersion();
+    context->weakdefver = NewDefaultVersion();
    // create vsyscall
    context->vsyscall = AddBridge(context->system, vFEv, x64Syscall, 0, NULL);
    // create the vsyscalls
@@ -169,15 +193,15 @@ box64context_t *NewBox64Context(int argc)
    context->vsyscalls[1] = AddVSyscall(context->system, 201);
    context->vsyscalls[2] = AddVSyscall(context->system, 309);
    // create the alternate to map at address
-    addAlternate((void*)context->vsyscalls[0], (void*)0xffffffffff600000);
-    addAlternate((void*)context->vsyscalls[1], (void*)0xffffffffff600400);
-    addAlternate((void*)context->vsyscalls[2], (void*)0xffffffffff600800);
+    addAlternate((void*)0xffffffffff600000, (void*)context->vsyscalls[0]);
+    addAlternate((void*)0xffffffffff600400, (void*)context->vsyscalls[1]);
+    addAlternate((void*)0xffffffffff600800, (void*)context->vsyscalls[2]);
    // get handle to box64 itself
    context->box64lib = dlopen(NULL, RTLD_NOW|RTLD_GLOBAL);
    context->dlprivate = NewDLPrivate();

    context->argc = argc;
-    context->argv = (char**)calloc(context->argc+1, sizeof(char*));
+    context->argv = (char**)box_calloc(context->argc+1, sizeof(char*));

    init_mutexes(context);
    pthread_atfork(NULL, NULL, atfork_child_box64context);
@@ -210,11 +234,13 @@ void FreeBox64Context(box64context_t** context)
    if(ctx->maplib)
        FreeLibrarian(&ctx->maplib, NULL);
    FreeDictionnary(&ctx->versym);
+    FreeDefaultVersion(&ctx->globaldefver);
+    FreeDefaultVersion(&ctx->weakdefver);

    for(int i=0; i<ctx->elfsize; ++i) {
        FreeElfHeader(&ctx->elfs[i]);
    }
-    free(ctx->elfs);
+    box_free(ctx->elfs);

    FreeCollection(&ctx->box64_path);
    FreeCollection(&ctx->box64_ld_lib);
@@ -226,16 +252,16 @@ void FreeBox64Context(box64context_t** context)
        DeleteX64Trace(ctx);

    if(ctx->deferedInitList)
-        free(ctx->deferedInitList);
+        box_free(ctx->deferedInitList);

-    free(ctx->argv);
+    /*box_free(ctx->argv);*/
    
-    for (int i=0; i<ctx->envc; ++i)
-        free(ctx->envv[i]);
-    free(ctx->envv);
+    /*for (int i=0; i<ctx->envc; ++i)
+        box_free(ctx->envv[i]);
+    box_free(ctx->envv);*/

    if(ctx->atfork_sz) {
-        free(ctx->atforks);
+        box_free(ctx->atforks);
        ctx->atforks = NULL;
        ctx->atfork_sz = ctx->atfork_cap = 0;
    }
@@ -251,14 +277,18 @@ void FreeBox64Context(box64context_t** context)

    FreeDLPrivate(&ctx->dlprivate);

-    free(ctx->fullpath);
-    free(ctx->box64path);
+    box_free(ctx->fullpath);
+    box_free(ctx->box64path);
+    box_free(ctx->bashpath);

    FreeBridge(&ctx->system);

    freeGLProcWrapper(ctx);
    freeALProcWrapper(ctx);

+    if(ctx->stack_clone)
+        box_free(ctx->stack_clone);
+

    void* ptr;
    if ((ptr = pthread_getspecific(ctx->tlskey)) != NULL) {
@@ -268,10 +298,15 @@ void FreeBox64Context(box64context_t** context)
    pthread_key_delete(ctx->tlskey);

    if(ctx->tlsdata)
-        free(ctx->tlsdata);
+        box_free(ctx->tlsdata);
+
+    free_neededlib(&ctx->neededlibs);
+
+    if(ctx->emu_sig)
+        FreeX64Emu(&ctx->emu_sig);
+
+    finiAllHelpers(ctx);

-    pthread_mutex_destroy(&ctx->mutex_once);
-    pthread_mutex_destroy(&ctx->mutex_once2);
    pthread_mutex_destroy(&ctx->mutex_trace);
 #ifndef DYNAREC
    pthread_mutex_destroy(&ctx->mutex_lock);
@@ -282,14 +317,9 @@ void FreeBox64Context(box64context_t** context)
    pthread_mutex_destroy(&ctx->mutex_thread);
    pthread_mutex_destroy(&ctx->mutex_bridge);

-    free_neededlib(&ctx->neededlibs);
+    freeCycleLog(ctx);

-    if(ctx->emu_sig)
-        FreeX64Emu(&ctx->emu_sig);
-
-    finiAllHelpers(ctx);
-
-    free(ctx);
+    box_free(ctx);
 }

 int AddElfHeader(box64context_t* ctx, elfheader_t* head) {
@@ -297,7 +327,7 @@ int AddElfHeader(box64context_t* ctx, elfheader_t* head) {
    if(idx==ctx->elfcap) {
        // resize...
        ctx->elfcap += 16;
-        ctx->elfs = (elfheader_t**)realloc(ctx->elfs, sizeof(elfheader_t*) * ctx->elfcap);
+        ctx->elfs = (elfheader_t**)box_realloc(ctx->elfs, sizeof(elfheader_t*) * ctx->elfcap);
    }
    ctx->elfs[idx] = head;
    ctx->elfsize++;
@@ -308,7 +338,7 @@ int AddElfHeader(box64context_t* ctx, elfheader_t* head) {
 int AddTLSPartition(box64context_t* context, int tlssize) {
    int oldsize = context->tlssize;
    context->tlssize += tlssize;
-    context->tlsdata = realloc(context->tlsdata, context->tlssize);
+    context->tlsdata = box_realloc(context->tlsdata, context->tlssize);
    memmove(context->tlsdata+tlssize, context->tlsdata, oldsize);   // move to the top, using memmove as regions will probably overlap
    memset(context->tlsdata, 0, tlssize);           // fill new space with 0 (not mandatory)
    // clean GS segment for current emu
@@ -321,52 +351,3 @@ int AddTLSPartition(box64context_t* context, int tlssize) {
    return -context->tlssize;   // negative offset
 }

-void add_neededlib(needed_libs_t* needed, library_t* lib)
-{
-    if(!needed)
-        return;
-    for(int i=0; i<needed->size; ++i)
-        if(needed->libs[i] == lib)
-            return;
-    if(needed->size == needed->cap) {
-        needed->cap += 8;
-        needed->libs = (library_t**)realloc(needed->libs, needed->cap*sizeof(library_t*));
-    }
-    needed->libs[needed->size++] = lib;
-}
-
-void free_neededlib(needed_libs_t* needed)
-{
-    if(!needed)
-        return;
-    needed->cap = 0;
-    needed->size = 0;
-    if(needed->libs)
-        free(needed->libs);
-    needed->libs = NULL;
-}
-
-void add_dependedlib(needed_libs_t* depended, library_t* lib)
-{
-    if(!depended)
-        return;
-    for(int i=0; i<depended->size; ++i)
-        if(depended->libs[i] == lib)
-            return;
-    if(depended->size == depended->cap) {
-        depended->cap += 8;
-        depended->libs = (library_t**)realloc(depended->libs, depended->cap*sizeof(library_t*));
-    }
-    depended->libs[depended->size++] = lib;
-}
-
-void free_dependedlib(needed_libs_t* depended)
-{
-    if(!depended)
-        return;
-    depended->cap = 0;
-    depended->size = 0;
-    if(depended->libs)
-        free(depended->libs);
-    depended->libs = NULL;
-}
@@ -2,7 +2,7 @@
 #define __BOX64_VERSION_H_

 #define BOX64_MAJOR 0
-#define BOX64_MINOR 1
-#define BOX64_REVISION 4
+#define BOX64_MINOR 2
+#define BOX64_REVISION 0

 #endif //__BOX64_VERSION_H_
@@ -194,6 +194,10 @@
 #define SBCSw_REG(Rd, Rn, Rm)       EMIT(ADDSUBC_gen(0, 1, 1, Rm, Rn, Rd))
 #define SBCSxw_REG(Rd, Rn, Rm)      EMIT(ADDSUBC_gen(rex.w, 1, 1, Rm, Rn, Rd))

+// ADR
+#define ADR_gen(immlo, immhi, Rd)   ((immlo)<<29 | 0b10000<<24 | (immhi)<<5 | (Rd))
+#define ADR_S20(Rd, imm)            EMIT(ADR_gen((imm)&3, ((imm)>>2)&0x7ffff, (Rd)))
+
 // LDR
 #define LDR_gen(size, op1, imm9, op2, Rn, Rt)    ((size)<<30 | 0b111<<27 | (op1)<<24 | 0b01<<22 | (imm9)<<12 | (op2)<<10 | (Rn)<<5 | (Rt))
 #define LDRx_S9_postindex(Rt, Rn, imm9)   EMIT(LDR_gen(0b11, 0b00, (imm9)&0x1ff, 0b01, Rn, Rt))
@@ -351,6 +355,13 @@

 #define PST_L1_STREAM_U12(Rn, imm12)    EMIT(PRFM_imm(((imm12)>>3)&0xfff, Rn, 0b01001))

+// Data Memory Barrier
+#define DMB_gen(CRm)                    (0b1101010100<<22 | 0b011<<16 | 0b0011<<12 | (CRm)<<8 | 1<<7 | 0b01<<5 | 0b11111)
+#define DMB_ISH()                       EMIT(DMB_gen(0b1011))
+
+// Break
+#define BRK_gen(imm16)                  (0b11010100<<24 | 0b001<<21 | (((imm16)&0xffff)<<5))
+#define BRK(imm16)                      EMIT(BRK_gen(imm16))

 // BR and Branches
 #define BR_gen(Z, op, A, M, Rn, Rm)       (0b1101011<<25 | (Z)<<24 | (op)<<21 | 0b11111<<16 | (A)<<11 | (M)<<10 | (Rn)<<5 | (Rm))
@@ -672,7 +683,9 @@
 
 // VLDR
 #define VMEM_gen(size, opc, imm12, Rn, Rt)  ((size)<<30 | 0b111<<27 | 1<<26 | 0b01<<24 | (opc)<<22 | (imm12)<<10 | (Rn)<<5 | (Rt))
-// imm14 must be 3-aligned
+// imm13 must be 1-aligned
+#define VLDR16_U12(Ht, Rn, imm13)           EMIT(VMEM_gen(0b01, 0b01, ((uint32_t)((imm13)>>1))&0xfff, Rn, Ht))
+// imm14 must be 2-aligned
 #define VLDR32_U12(Dt, Rn, imm14)           EMIT(VMEM_gen(0b10, 0b01, ((uint32_t)((imm14)>>2))&0xfff, Rn, Dt))
 // imm15 must be 3-aligned
 #define VLDR64_U12(Dt, Rn, imm15)           EMIT(VMEM_gen(0b11, 0b01, ((uint32_t)((imm15)>>3))&0xfff, Rn, Dt))
@@ -684,22 +697,22 @@
 #define VSTR64_U12(Dt, Rn, imm15)           EMIT(VMEM_gen(0b11, 0b00, ((uint32_t)(imm15>>3))&0xfff, Rn, Dt))
 // imm16 must be 4-aligned
 #define VSTR128_U12(Qt, Rn, imm16)          EMIT(VMEM_gen(0b00, 0b10, ((uint32_t)((imm16)>>4))&0xfff, Rn, Qt))
-// (imm14) must be 1-aligned
-#define VSTR16_U12(Ht, Rn, imm14)           EMIT(VMEM_gen(0b01, 0b00, ((uint32_t)(imm14>>1))&0xfff, Rn, Ht))
+// (imm13) must be 1-aligned
+#define VSTR16_U12(Ht, Rn, imm13)           EMIT(VMEM_gen(0b01, 0b00, ((uint32_t)(imm13>>1))&0xfff, Rn, Ht))

 #define VMEMUR_vector(size, opc, imm9, Rn, Rt)  ((size)<<30 | 0b111<<27 | 1<<26 | (opc)<<22 | (imm9)<<12 | (Rn)<<5 | (Rt))
 // signed offset, no alignement!
-#define VLDR8_I9(Vt, Rn, imm9)              EMIT(VMEMUR(0b00, 0b01, (imm9)&0b111111111, Rn, Vt))
-#define VLDR16_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b01, 0b01, (imm9)&0b111111111, Rn, Vt))
-#define VLDR32_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b10, 0b01, (imm9)&0b111111111, Rn, Vt))
-#define VLDR64_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b11, 0b01, (imm9)&0b111111111, Rn, Vt))
-#define VLDR128_I9(Vt, Rn, imm9)            EMIT(VMEMUR(0b00, 0b11, (imm9)&0b111111111, Rn, Vt))
+#define VLDR8_I9(Vt, Rn, imm9)              EMIT(VMEMUR_vector(0b00, 0b01, (imm9)&0b111111111, Rn, Vt))
+#define VLDR16_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b01, 0b01, (imm9)&0b111111111, Rn, Vt))
+#define VLDR32_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b10, 0b01, (imm9)&0b111111111, Rn, Vt))
+#define VLDR64_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b11, 0b01, (imm9)&0b111111111, Rn, Vt))
+#define VLDR128_I9(Vt, Rn, imm9)            EMIT(VMEMUR_vector(0b00, 0b11, (imm9)&0b111111111, Rn, Vt))
 // signed offset, no alignement!
-#define VSTR8_I9(Vt, Rn, imm9)              EMIT(VMEMUR(0b00, 0b00, (imm9)&0b111111111, Rn, Vt))
-#define VSTR16_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b01, 0b00, (imm9)&0b111111111, Rn, Vt))
-#define VSTR32_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b10, 0b00, (imm9)&0b111111111, Rn, Vt))
-#define VSTR64_I9(Vt, Rn, imm9)             EMIT(VMEMUR(0b11, 0b00, (imm9)&0b111111111, Rn, Vt))
-#define VSTR128_I9(Vt, Rn, imm9)            EMIT(VMEMUR(0b00, 0b10, (imm9)&0b111111111, Rn, Vt))
+#define VSTR8_I9(Vt, Rn, imm9)              EMIT(VMEMUR_vector(0b00, 0b00, (imm9)&0b111111111, Rn, Vt))
+#define VSTR16_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b01, 0b00, (imm9)&0b111111111, Rn, Vt))
+#define VSTR32_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b10, 0b00, (imm9)&0b111111111, Rn, Vt))
+#define VSTR64_I9(Vt, Rn, imm9)             EMIT(VMEMUR_vector(0b11, 0b00, (imm9)&0b111111111, Rn, Vt))
+#define VSTR128_I9(Vt, Rn, imm9)            EMIT(VMEMUR_vector(0b00, 0b10, (imm9)&0b111111111, Rn, Vt))

 #define VMEMW_gen(size, opc, imm9, op2, Rn, Rt)  ((size)<<30 | 0b111<<27 | 1<<26 | (opc)<<22 | (imm9)<<12 | (op2)<<10 | 0b01<<10 | (Rn)<<5 | (Rt))
 #define VLDR64_S9_postindex(Rt, Rn, imm9)   EMIT(VMEMW_gen(0b11, 0b01, (imm9)&0x1ff, 0b01, Rn, Rt))
@@ -714,18 +727,18 @@
 #define VMEM_REG_gen(size, opc, Rm, option, S, Rn, Rt)  ((size)<<30 | 0b111<<27 | 1<<26 | (opc)<<22 | 1<<21 | (Rm)<<16 | (option)<<13 | (S)<<12 | 0b10<<10 | (Rn)<<5 | (Rt))

 #define VLDR32_REG(Dt, Rn, Rm)              EMIT(VMEM_REG_gen(0b10, 0b01, Rm, 0b011, 0, Rn, Dt))
-#define VLDR32_REG_LSL3(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b10, 0b01, Rm, 0b011, 1, Rn, Dt))
+#define VLDR32_REG_LSL2(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b10, 0b01, Rm, 0b011, 1, Rn, Dt))
 #define VLDR64_REG(Dt, Rn, Rm)              EMIT(VMEM_REG_gen(0b11, 0b01, Rm, 0b011, 0, Rn, Dt))
 #define VLDR64_REG_LSL3(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b11, 0b01, Rm, 0b011, 1, Rn, Dt))
-#define VLDR128_REG(Qt, Rn, Rm)             EMIT(VMEM_REG_gen(0b00, 0b11, Rm, 0b011, 0, Rn, Dt))
-#define VLDR128_REG_LSL4(Qt, Rn, Rm)        EMIT(VMEM_REG_gen(0b00, 0b11, Rm, 0b011, 1, Rn, Dt))
+#define VLDR128_REG(Qt, Rn, Rm)             EMIT(VMEM_REG_gen(0b00, 0b11, Rm, 0b011, 0, Rn, Qt))
+#define VLDR128_REG_LSL4(Qt, Rn, Rm)        EMIT(VMEM_REG_gen(0b00, 0b11, Rm, 0b011, 1, Rn, Qt))

 #define VSTR32_REG(Dt, Rn, Rm)              EMIT(VMEM_REG_gen(0b10, 0b00, Rm, 0b011, 0, Rn, Dt))
-#define VSTR32_REG_LSL3(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b10, 0b00, Rm, 0b011, 1, Rn, Dt))
+#define VSTR32_REG_LSL2(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b10, 0b00, Rm, 0b011, 1, Rn, Dt))
 #define VSTR64_REG(Dt, Rn, Rm)              EMIT(VMEM_REG_gen(0b11, 0b00, Rm, 0b011, 0, Rn, Dt))
 #define VSTR64_REG_LSL3(Dt, Rn, Rm)         EMIT(VMEM_REG_gen(0b11, 0b00, Rm, 0b011, 1, Rn, Dt))
-#define VSTR128_REG(Qt, Rn, Rm)             EMIT(VMEM_REG_gen(0b00, 0b10, Rm, 0b011, 0, Rn, Dt))
-#define VSTR128_REG_LSL4(Qt, Rn, Rm)        EMIT(VMEM_REG_gen(0b00, 0b10, Rm, 0b011, 1, Rn, Dt))
+#define VSTR128_REG(Qt, Rn, Rm)             EMIT(VMEM_REG_gen(0b00, 0b10, Rm, 0b011, 0, Rn, Qt))
+#define VSTR128_REG_LSL4(Qt, Rn, Rm)        EMIT(VMEM_REG_gen(0b00, 0b10, Rm, 0b011, 1, Rn, Qt))

 #define VLDR_PC_gen(opc, imm19, Rt)         ((opc)<<30 | 0b011<<27 | 1<<26 | (imm19)<<5 | (Rt))
 #define VLDR32_literal(Vt, imm19)           EMIT(VLDR_PC_gen(0b00, ((imm19)>>2)&0x7FFFF, Vt))
@@ -1232,6 +1245,11 @@
 #define VFRINTISQ(Vd,Vn)            EMIT(FRINT_vector(1, 1, 1, 0, 1, Vn, Vd))
 #define VFRINTIDQ(Vd,Vn)            EMIT(FRINT_vector(1, 1, 1, 1, 1, Vn, Vd))

+#define FRINTI_scalar(type, Rn, Rd)  (0b11110<<24 | (type)<<22 | 1<<21 | 0b001<<18 | 0b111<<15 | 0b10000<<10 | (Rn)<<5 | (Rd))
+#define FRINTIS(Sd, Sn)             EMIT(FRINTI_scalar(0b00, Sn, Sd))
+#define FRINTID(Dd, Dn)             EMIT(FRINTI_scalar(0b01, Dn, Dd))
+
+
 #define FRINTxx_scalar(type, op, Rn, Rd)  (0b11110<<24 | (type)<<22 | 1<<21 | 0b0100<<17 | (op)<<15 | 0b10000<<10 | (Rn)<<5 | (Rd))
 #define FRINT32ZS(Sd, Sn)           EMIT(FRINTxx_scalar(0b00, 0b00, Sn, Sd))
 #define FRINT32ZD(Dd, Dn)           EMIT(FRINTxx_scalar(0b01, 0b00, Dn, Dd))
@@ -1312,6 +1330,15 @@
 #define VUZP1Q_64(Rt, Rn, Rm)       EMIT(UZP_gen(1, 0b11, Rm, 0, Rn, Rt))
 #define VUZP2Q_64(Rt, Rn, Rm)       EMIT(UZP_gen(1, 0b11, Rm, 1, Rn, Rt))

+#define DUP_gen(Q, imm5, Rn, Rd)    ((Q)<<30 | 0b01110000<<21 | (imm5)<<16 | 1<<10 | (Rn)<<5 | (Rd))
+#define VDUP_8(Vd, Vn, idx)         EMIT(DUP_gen(0, ((idx)<<1|1), Vn, Vd))
+#define VDUPQ_8(Vd, Vn, idx)        EMIT(DUP_gen(1, ((idx)<<1|1), Vn, Vd))
+#define VDUP_16(Vd, Vn, idx)        EMIT(DUP_gen(0, ((idx)<<2|0b10), Vn, Vd))
+#define VDUPQ_16(Vd, Vn, idx)       EMIT(DUP_gen(1, ((idx)<<2|0b10), Vn, Vd))
+#define VDUP_32(Vd, Vn, idx)        EMIT(DUP_gen(0, ((idx)<<3|0b100), Vn, Vd))
+#define VDUPQ_32(Vd, Vn, idx)       EMIT(DUP_gen(1, ((idx)<<3|0b100), Vn, Vd))
+#define VDUPQ_64(Vd, Vn, idx)       EMIT(DUP_gen(1, ((idx)<<4|0b1000), Vn, Vd))
+
 // TBL
 #define TBL_gen(Q, Rm, len, op, Rn, Rd) ((Q)<<30 | 0b001110<<24 | (Rm)<<16 | (len)<<13 | (op)<<12 | (Rn)<<5 | (Rd))
 //Use Rm[] to pick from Rn element and store in Rd. Out-of-range element gets 0
@@ -1451,18 +1478,35 @@
 // Vector Float CMP
 // EQual
 #define FCMP_vector(Q, U, E, sz, Rm, ac, Rn, Rd)    ((Q)<<30 | (U)<<29 | 0b01110<<24 | (E)<<23 | (sz)<<22 | 1<<21 | (Rm)<<16 | 0b1110<<12 | (ac)<<11 | 1<<10 | (Rn)<<5 | (Rd))
-#define FCMEQQD(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 0, 0, 1, Rm, 0, Rn, Rd))
-#define FCMEQQS(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 0, 0, 0, Rm, 0, Rn, Rd))
+#define VFCMEQQD(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 0, 0, 1, Rm, 0, Rn, Rd))
+#define VFCMEQQS(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 0, 0, 0, Rm, 0, Rn, Rd))
 // Greater or Equal
-#define FCMGEQD(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 1, 0, 1, Rm, 0, Rn, Rd))
-#define FCMGEQS(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 1, 0, 0, Rm, 0, Rn, Rd))
-#define FCMGEQD_ABS(Rd, Rn, Rm)      EMIT(FCMP_vector(1, 1, 0, 1, Rm, 1, Rn, Rd))
-#define FCMGEQS_ABS(Rd, Rn, Rm)      EMIT(FCMP_vector(1, 1, 0, 0, Rm, 1, Rn, Rd))
+#define VFCMGEQD(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 1, 0, 1, Rm, 0, Rn, Rd))
+#define VFCMGEQS(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 1, 0, 0, Rm, 0, Rn, Rd))
+#define VFCMGEQD_ABS(Rd, Rn, Rm)     EMIT(FCMP_vector(1, 1, 0, 1, Rm, 1, Rn, Rd))
+#define VFCMGEQS_ABS(Rd, Rn, Rm)     EMIT(FCMP_vector(1, 1, 0, 0, Rm, 1, Rn, Rd))
 // Greater Than
-#define FCMGTQD(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 1, 1, 1, Rm, 0, Rn, Rd))
-#define FCMGTQS(Rd, Rn, Rm)          EMIT(FCMP_vector(1, 1, 1, 0, Rm, 0, Rn, Rd))
-#define FCMGTQD_ABS(Rd, Rn, Rm)      EMIT(FCMP_vector(1, 1, 1, 1, Rm, 1, Rn, Rd))
-#define FCMGTQS_ABS(Rd, Rn, Rm)      EMIT(FCMP_vector(1, 1, 1, 0, Rm, 1, Rn, Rd))
+#define VFCMGTQD(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 1, 1, 1, Rm, 0, Rn, Rd))
+#define VFCMGTQS(Rd, Rn, Rm)         EMIT(FCMP_vector(1, 1, 1, 0, Rm, 0, Rn, Rd))
+#define VFCMGTQD_ABS(Rd, Rn, Rm)     EMIT(FCMP_vector(1, 1, 1, 1, Rm, 1, Rn, Rd))
+#define VFCMGTQS_ABS(Rd, Rn, Rm)     EMIT(FCMP_vector(1, 1, 1, 0, Rm, 1, Rn, Rd))
+
+// Scalar Float CMP to 0
+#define FCMP_0_scalar(U, sz, op, Rn, Rd) (0b01<<30 | (U)<<29| 0b11110<<24 | 1<<23 | (sz)<<22 | 0b10000<<17 | 0b011<<14 | (op)<<12 | 0b10<<10 | (Rn)<<5 | (Rd))
+// Less or equal to 0
+#define FCMLES_0(Rd, Rn)             EMIT(FCMP_0_scalar(1, 0, 0b01, (Rn), (Rd)))
+#define FCMLED_0(Rd, Rn)             EMIT(FCMP_0_scalar(1, 1, 0b01, (Rn), (Rd)))
+// Greater than 0
+#define FCMGTS_0(Rd, Rn)             EMIT(FCMP_0_scalar(0, 0, 0b00, (Rn), (Rd)))
+#define FCMGTD_0(Rd, Rn)             EMIT(FCMP_0_scalar(0, 1, 0b00, (Rn), (Rd)))
+// Less than 0
+#define FCMLTS_0(Rd, Rn)             EMIT(FCMP_0_scalar(0, 0, 0b10, (Rn), (Rd)))
+#define FCMLTD_0(Rd, Rn)             EMIT(FCMP_0_scalar(0, 1, 0b10, (Rn), (Rd)))
+
+// Scalar Float CMP
+#define FCMP_op_scalar(U, E, sz, Rm, ac, Rn, Rd)    (0b01<<30 | (U)<<29 | 0b11110<<24 | (E)<<23 | (sz)<<22 | 1<<21 | (Rm)<<16 | 0b1110<<12 | (ac)<<11 | 1<<10 | (Rn)<<5 | (Rd))
+#define FCMEQS(Rd, Rn, Rm)          EMIT(FCMP_op_scalar(0, 0, 0, (Rm), 0, (Rn), (Rd)))
+#define FCMEQD(Rd, Rn, Rm)          EMIT(FCMP_op_scalar(0, 0, 1, (Rm), 0, (Rn), (Rd)))

 // UMULL / SMULL
 #define MULL_vector(Q, U, size, Rm, Rn, Rd) ((Q)<<30 | (U)<<29 | 0b01110<<24 | (size)<<22 | 1<<21 | (Rm)<<16 | 0b1100<<12 |(Rn)<<5 |(Rd))
@@ -1727,7 +1771,7 @@
 #define SQDMULHQ_16(Vd, Vn, Vm)     EMIT(QDMULH_vector(1, 0, 0b01, Vm, Vn, Vd))
 #define SQDMULHQ_32(Vd, Vn, Vm)     EMIT(QDMULH_vector(1, 0, 0b10, Vm, Vn, Vd))

-// AES extensions
+// AES extension
 #define AES_gen(D, Rn, Rd)      (0b01001110<<24 | 0b00<<22 | 0b10100<<17 | 0b0010<<13 | (D)<<12 | 0b10<<10 | (Rn)<<5 | (Rd))
 #define AESD(Vd, Vn)    EMIT(AES_gen(1, Vn, Vd))
 #define AESE(Vd, Vn)    EMIT(AES_gen(0, Vn, Vd))
@@ -1736,4 +1780,11 @@
 #define AESIMC(Vd, Vn)  EMIT(AESMC_gen(1, Vn, Vd))
 #define AESMC(Vd, Vn)   EMIT(AESMC_gen(0, Vn, Vd))

+// PMULL extension is PMULL_128
+#define PMULL_gen(Q, size, Rm, Rn, Rd)  (0<<31 | (Q)<<30 | 0b001110<<24 | (size)<<22 | 1<<21 | (Rm)<<16 | 0b1110<<12 | (Rn)<<5 | (Rd))
+#define PMULL(Rd, Rn, Rm)   EMIT(PMULL_gen(0, 0b00, Rm, Rn, Rd))
+#define PMULL2(Rd, Rn, Rm)  EMIT(PMULL_gen(1, 0b00, Rm, Rn, Rd))
+#define PMULL_128(Rd, Rn, Rm)   EMIT(PMULL_gen(0, 0b11, Rm, Rn, Rd))
+#define PMULL2_128(Rd, Rn, Rm)  EMIT(PMULL_gen(1, 0b11, Rm, Rn, Rd))
+
 #endif  //__ARM64_EMITTER_H__
@@ -30,7 +30,7 @@ arm64_epilog:
    ldp     d12, d13, [sp, (8 *14)]
    ldp     d14, d15, [sp, (8 *16)]
    add     sp,  sp, (8 * 18)
-    ldp     lr, fp, [sp], 16  // saved lr
+    ldp     x30, x29, [sp], 16  // saved lr
    //end, return...
    ret

@@ -49,6 +49,6 @@ arm64_epilog_fast:
    ldp     d12, d13, [sp, (8 *14)]
    ldp     d14, d15, [sp, (8 *16)]
    add     sp,  sp, (8 * 18)
-    ldp     lr, fp, [sp], 16  // saved lr
+    ldp     x30, x29, [sp], 16  // saved lr
    //end, return...
    ret
@@ -16,11 +16,13 @@
 .global arm64_lock_read_dq
 .global arm64_lock_write_dq
 .global arm64_lock_xchg
+.global arm64_lock_xchg_d
 .global arm64_lock_storeifnull
 .global arm64_lock_storeifref


 arm64_lock_read_b:
+    dmb     ish
    // address is x0, return is x0
    ldaxrb  w0, [x0]
    ret
@@ -29,9 +31,11 @@ arm64_lock_write_b:
    // address is x0, value is x1, return is x0
    mov     x2, x0
    stlxrb  w0, w1, [x2]
+    dmb     ish
    ret

 arm64_lock_read_h:
+    dmb     ish
    // address is x0, return is x0
    ldaxrh  w0, [x0]
    ret
@@ -40,9 +44,11 @@ arm64_lock_write_h:
    // address is x0, value is x1, return is x0
    mov     x2, x0
    stlxrh  w0, w1, [x2]
+    dmb     ish
    ret

 arm64_lock_read_d:
+    dmb     ish
    // address is x0, return is x0
    ldaxr    w0, [x0]
    ret
@@ -51,9 +57,11 @@ arm64_lock_write_d:
    // address is x0, value is w1, return is x0
    mov     x2, x0
    stlxr   w0, w1, [x2]
+    dmb     ish
    ret

 arm64_lock_read_dd:
+    dmb     ish
    // address is x0, return is x0
    ldaxr   x0, [x0]
    ret
@@ -62,9 +70,11 @@ arm64_lock_write_dd:
    // address is x0, value is x1, return is x0
    mov     x2, x0
    stlxr   w0, x1, [x2]
+    dmb     ish
    ret

 arm64_lock_read_dq:
+    dmb     ish
    // address is r2, return is r0, r1
    ldaxp   x4, x3, [x2]
    str     x4, [x0]
@@ -76,37 +86,58 @@ arm64_lock_write_dq:
    // r0 needs to be aligned
    stlxp   w3, x0, x1, [x2]
    mov     w0, w3
+    dmb     ish
    ret


 arm64_lock_xchg:
+    dmb     ish
+arm64_lock_xchg_0:
    // address is x0, value is x1, return old value in x0
    ldaxr   x2, [x0]
    stlxr   w3, x1, [x0]
-    cbnz    w3, arm64_lock_xchg
+    cbnz    w3, arm64_lock_xchg_0
+    dmb     ish
    mov     x0, x2
    ret

+arm64_lock_xchg_d:
+    dmb     ish
+arm64_lock_xchg_d_0:
+    // address is x0, value is x1, return old value in x0
+    ldaxr   w2, [x0]
+    stlxr   w3, w1, [x0]
+    cbnz    w3, arm64_lock_xchg_d_0
+    dmb     ish
+    mov     w0, w2
+    ret
+
 arm64_lock_storeifnull:
+    dmb     ish
+arm64_lock_storeifnull_0:
    // address is x0, value is x1, x1 store to x0 only if [x0] is 0. return new [x0] value (so x1 or old value)
    ldaxr   x2, [x0]
    cbnz    x2, arm64_lock_storeifnull_exit
    mov     x2, x1
    stlxr   w3, x2, [x0]
-    cbnz    w3, arm64_lock_storeifnull
+    cbnz    w3, arm64_lock_storeifnull_0
 arm64_lock_storeifnull_exit:
+    dmb     ish
    mov     x0, x2
    ret

 arm64_lock_storeifref:
+    dmb     ish
+arm64_lock_storeifref_0:
    // address is x0, value is x1, x1 store to x0 only if [x0] is x3. return new [x0] value (so x1 or old value)
    ldaxr   x3, [x0]
    cmp     x2, x3
    bne     arm64_lock_storeifref_exit
    stlxr   w4, x1, [x0]
-    cbnz    w4, arm64_lock_storeifref
+    cbnz    w4, arm64_lock_storeifref_0
    mov     x0, x1
    ret
 arm64_lock_storeifref_exit:
+    dmb     ish
    mov     x0, x3
    ret
@@ -30,6 +30,9 @@ extern int arm64_lock_write_dq(uint64_t a, uint64_t b, void* addr);
 // Atomicaly exchange value at [p] with val, return old p
 extern uintptr_t arm64_lock_xchg(void* p, uintptr_t val);

+// Atomicaly exchange value at [p] with val, return old p
+extern uint32_t arm64_lock_xchg_d(void* p, uint32_t val);
+
 // Atomicaly store value to [p] only if [p] is NULL. Return new [p] value (so val or old)
 extern void* arm64_lock_storeifnull(void*p, void* val);

@@ -18,9 +18,10 @@ arm64_next:
    stp     x12, x13, [sp, (8 *  4)]
    stp     x14, x15, [sp, (8 *  6)]
    stp     x16, x17, [sp, (8 *  8)]
-    str     x18, [sp, (8 * 10)]
+    stp     x18, x27, [sp, (8 * 10)]    // also save x27(rip) to allow change in LinkNext

-    mov     x2, lr      // "from" is in lr, so put in x2
+    mov     x2, x30      // "from" is in lr, so put in x2
+    add     x3, sp, 8*11    // x3 is address to change rip
    // call the function
    bl      LinkNext
    // preserve return value
@@ -31,7 +32,7 @@ arm64_next:
    ldp     x12, x13, [sp, (8 *  4)]
    ldp     x14, x15, [sp, (8 *  6)]
    ldp     x16, x17, [sp, (8 *  8)]
-    ldr     x18, [sp, (8 * 10)]
+    ldp     x18, x27, [sp, (8 * 10)]
    add     sp,  sp, (8 * 12)
    // return offset is jump address
    br      x3
@@ -236,6 +236,15 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
            snprintf(buff, sizeof(buff), "STR %s, [%s, %s, %s %d]", (size==2)?Wt[Rt]:Xt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option], amount);
        return buff;
    }
+    if(isMask(opcode, "00111000001mmmmmoooS10nnnnnttttt", &a)) {
+        const char* extend[] = {"?0", "?1", "UXTW", "LSL", "?4", "?5", "SXTW", "SXTX"};
+        int amount = a.S;
+        if(option==3 && !amount)
+            snprintf(buff, sizeof(buff), "STRB %s, [%s, %s]", Wt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm]);
+        else
+            snprintf(buff, sizeof(buff), "STRB %s, [%s, %s, %s %d]", Wt[Rt], XtSp[Rn], ((option&1)==0)?Wt[Rm]:Xt[Rm], extend[option], amount);
+        return buff;
+    }
    if(isMask(opcode, "0x111000010iiiiiiiii01nnnnnttttt", &a)) {
        int size = a.x;
        int offset = signExtend(imm, 9);
@@ -767,7 +776,7 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
        return buff;
    }
    if(isMask(opcode, "10011011U10mmmmm011111nnnnnddddd", &a)) {
-        snprintf(buff, sizeof(buff), "%cMULH %s, %s, %s", a.U?'U':'S', Xt[Rd], Wt[Rn], Wt[Rm]);
+        snprintf(buff, sizeof(buff), "%cMULH %s, %s, %s", a.U?'U':'S', Xt[Rd], Xt[Rn], Xt[Rm]);
        return buff;
    }
    if(isMask(opcode, "f0011011000mmmmm0aaaaannnnnddddd", &a)) {
@@ -930,6 +939,12 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
        return buff;
    }

+    // ADR
+    if(isMask(opcode, "0ss10000iiiiiiiiiiiiiiiiiiiddddd", &a)) {
+      snprintf(buff, sizeof(buff), "ADR, %s, %ld", Xt[Rd], signExtend((imm)<<2|(imms), 20));
+      return buff;
+    }
+
    // LDR / STR
    if(isMask(opcode, "ss111101cciiiiiiiiiiiinnnnnttttt", &a)) {
        char s = '?';
@@ -1036,7 +1051,7 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
    //FMIN/FMAX
    if(isMask(opcode, "00011110ff1mmmmm01oo10nnnnnddddd", &a)) {
        char s = (sf==0)?'S':((sf==1)?'D':'?');
-        snprintf(buff, sizeof(buff), "F%s %c%d, %c%d, %c%d", (option==3)?"MINNM":((option==2)?"MINNMP":((!option)?"MAXNM":"MAXNMP")), s, Rd, s, Rn, s, Rm);
+        snprintf(buff, sizeof(buff), "F%s %c%d, %c%d, %c%d", (option==3)?"MINNM":((option==2)?"MAXNM":((!option)?"MAX":"MIN")), s, Rd, s, Rn, s, Rm);
        return buff;
    }
    if(isMask(opcode, "0Q001110of1mmmmm110001nnnnnddddd", &a)) {
@@ -1311,6 +1326,21 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
            snprintf(buff, sizeof(buff), "%cSHLL%s V%d.%s, V%d.%s, #%d", a.U?'U':'S', a.Q?"2":"", Rd, Va, Rn, Vd, sh);
        return buff;
    }
+
+    // DUP
+    if(isMask(opcode, "0Q001110000iiiii000001nnnnnddddd", &a)) {
+        const char* Y[] = {"8B", "16B", "4H", "8H", "2S", "4S", "2D", "??"};
+        const char* Z[] = {"B", "H", "S", "D"};
+        int sz = 3;
+        if((imm&0b0001)==0b0001) sz=0;
+        else if((imm&0b0011)==0b0010) sz=1;
+        else if((imm&0b0111)==0b0100) sz=2;
+        int sh=imm - (1<<sz);
+        const char* Vd = Y[(sz<<1)|a.Q];
+        const char* Vn = Z[sz];
+        snprintf(buff, sizeof(buff), "DUP V%d.%s, V%d.%s[%d]", Rd, Vd, Rn, Vn, sh);
+        return buff;
+    }
    
    // AES
    if(isMask(opcode, "0100111000101000010f10nnnnnddddd", &a)) {
@@ -1321,8 +1351,23 @@ const char* arm64_print(uint32_t opcode, uintptr_t addr)
        snprintf(buff, sizeof(buff), "AES%sMC V%d.16B, V%d.16B", sf?"I":"", Rd, Rn);
        return buff;
    }
+    // PMULL
+    if(isMask(opcode, "0Q001110ff1mmmmm111000nnnnnddddd", &a)) {
+        const char* Y[] = {"8B", "16B", "??", "??", "??", "??", "1D", "2D"};
+        const char* Z[] = {"8H", "??", "??", "1Q"};
+        int sz = sf;
+        const char* Vn = Y[(sz<<1)|a.Q];
+        const char* Vd = Z[sz];
+        snprintf(buff, sizeof(buff), "PMULL%s V%d.%s, V%d.%s, V%d.%s", a.Q?"2":"", Rd, Vd, Rn, Vn, Rm, Vn);   
+        return buff;
+    }
+
+    // DMB ISH
+    if(isMask(opcode, "11010101000000110011nnnn10111111", &a)) {
+        snprintf(buff, sizeof(buff), "DMB %s", (Rn==0b1011)?"ISH":"???");
+    }


    snprintf(buff, sizeof(buff), "%08X ???", __builtin_bswap32(opcode));
    return buff;
-}
+}
@@ -9,7 +9,7 @@
 .global arm64_prolog
 arm64_prolog:
    //save all 18 used register
-    stp     lr, fp, [sp, -16]!  // save lr
+    stp     x30, x29, [sp, -16]!  // save lr
    sub     sp,  sp, (8 * 18)
    stp     x19, x20, [sp, (8 * 0)]
    stp     x21, x22, [sp, (8 * 2)]
@@ -17,10 +17,10 @@
 #include "bridge.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

@@ -39,12 +39,16 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    uint64_t u64;
    uint8_t wback, wb1, wb2, wb;
    int64_t fixedaddress;
+    int lock;
+    int cacheupd;

    opcode = F8;
    MAYUSE(eb1);
    MAYUSE(eb2);
    MAYUSE(j64);
    MAYUSE(wb);
+    MAYUSE(lock);
+    MAYUSE(cacheupd);

    switch(opcode) {
        case 0x00:
@@ -372,6 +376,9 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            i64 = F32S;
            emit_sub32c(dyn, ninst, rex, xRAX, i64, x3, x4, x5);
            break;
+        case 0x2E:
+            INST_NAME("CS:");
+            break;

        case 0x30:
            INST_NAME("XOR Eb, Gb");
@@ -523,14 +530,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                if(MODREG) {   // reg <= reg
                    SXTWx(gd, xRAX+(nextop&7)+(rex.b<<3));
                } else {                    // mem <= reg
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    LDRSW_U12(gd, ed, fixedaddress);
                }
            } else {
                if(MODREG) {   // reg <= reg
                    MOVw_REG(gd, xRAX+(nextop&7)+(rex.b<<3));
                } else {                    // mem <= reg
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    LDRw_U12(gd, ed, fixedaddress);
                }
            }
@@ -633,23 +640,30 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            }
            break;

-        #define GO(GETFLAGS, NO, YES, F)    \
-            READFLAGS(F);                   \
-            i8 = F8S;   \
-            BARRIER(2); \
-            JUMP(addr+i8);\
-            GETFLAGS;   \
-            if(dyn->insts) {    \
-                if(dyn->insts[ninst].x64.jmp_insts==-1) {   \
-                    /* out of the block */                  \
-                    i32 = dyn->insts[ninst+1].address-(dyn->arm_size); \
-                    Bcond(NO, i32);     \
-                    jump_to_next(dyn, addr+i8, 0, ninst); \
-                } else {    \
-                    /* inside the block */  \
-                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->arm_size);    \
-                    Bcond(YES, i32);    \
-                }   \
+        #define GO(GETFLAGS, NO, YES, F)                                \
+            READFLAGS(F);                                               \
+            i8 = F8S;                                                   \
+            BARRIER(BARRIER_MAYBE);                                     \
+            JUMP(addr+i8, 1);                                           \
+            GETFLAGS;                                                   \
+            if(dyn->insts[ninst].x64.jmp_insts==-1 ||                   \
+                CHECK_CACHE()) {                                        \
+                /* out of the block */                                  \
+                i32 = dyn->insts[ninst].epilog-(dyn->native_size);      \
+                Bcond(NO, i32);                                         \
+                if(dyn->insts[ninst].x64.jmp_insts==-1) {               \
+                    if(!(dyn->insts[ninst].x64.barrier&BARRIER_FLOAT))  \
+                        fpu_purgecache(dyn, ninst, 1, x1, x2, x3);      \
+                    jump_to_next(dyn, addr+i8, 0, ninst);               \
+                } else {                                                \
+                    CacheTransform(dyn, ninst, cacheupd, x1, x2, x3);   \
+                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);\
+                    B(i32);                                             \
+                }                                                       \
+            } else {                                                    \
+                /* inside the block */                                  \
+                i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);    \
+                Bcond(YES, i32);                                        \
            }

        GOCOND(0x70, "J", "ib");
@@ -846,14 +860,16 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                BFIx(gb1, x1, gb2*8, 8);
                BFIx(eb1, x4, eb2*8, 8);
            } else {
+                DMB_ISH();
                GETGB(x4);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                MARKLOCK;
                // do the swap with exclusive locking
                LDAXRB(x1, ed);
                // do the swap 14 -> strb(ed), 1 -> gd
                STLXRB(x3, x4, ed);
                CBNZx_MARKLOCK(x3);
+                DMB_ISH();
                BFIx(gb1, x1, gb2*8, 8);
            }
            break;
@@ -868,7 +884,8 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                MOVxw_REG(ed, x1);
            } else {
                GETGD;
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                DMB_ISH();
                TSTx_mask(ed, 1, 0, 1+rex.w);    // mask=3 or 7
                B_MARK(cNE);
                MARKLOCK;
@@ -880,6 +897,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                LDRxw_U12(x1, ed, 0);
                STRxw_U12(gd, ed, 0);
                MARK2;
+                DMB_ISH();
                MOVxw_REG(gd, x1);
            }
            break;
@@ -911,8 +929,12 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                }
                BFIx(eb1, gd, eb2*8, 8);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff, 0, rex, &lock, 0, 0);
                STRB_U12(gd, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;
        case 0x89:
@@ -922,8 +944,12 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            if(MODREG) {   // reg <= reg
                MOVxw_REG(xRAX+(nextop&7)+(rex.b<<3), gd);
            } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 0);
                STRxw_U12(gd, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;
        case 0x8A:
@@ -953,7 +979,11 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    ed = wback;
                }
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, &lock, 0, 0);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || !ninst || box64_dynarec_strongmem>1 || (ninst && dyn->insts[ninst-1].x64.barrier)))) {
+                    DMB_ISH();
+                }
                LDRB_U12(x4, wback, fixedaddress);
                ed = x4;
            }
@@ -963,10 +993,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            INST_NAME("MOV Gd, Ed");
            nextop=F8;
            GETGD;
-            if(MODREG) {   // reg <= reg
+            if(MODREG) {
                MOVxw_REG(gd, xRAX+(nextop&7)+(rex.b<<3));
-            } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, 0);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 0);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || !ninst || box64_dynarec_strongmem>1 || (ninst && dyn->insts[ninst-1].x64.barrier)))) {
+                    DMB_ISH();
+                }
                LDRxw_U12(gd, ed, fixedaddress);
            }
            break;
@@ -976,7 +1010,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            if((nextop&0xC0)==0xC0) {   // reg <= seg
                LDRH_U12(xRAX+(nextop&7)+(rex.b<<3), xEmu, offsetof(x64emu_t, segs[(nextop&0x38)>>3]));
            } else {                    // mem <= seg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                LDRH_U12(x3, xEmu, offsetof(x64emu_t, segs[(nextop&0x38)>>3]));
                STRH_U12(x3, ed, fixedaddress);
            }
@@ -988,7 +1022,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            if(MODREG) {   // reg <= reg? that's an invalid operation
                DEFAULT;
            } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, gd, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, gd, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                if(gd!=ed) {    // it's sometimes used as a 3 bytes NOP
                    MOVxw_REG(gd, ed);
                }
@@ -997,7 +1031,19 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                }
            }
            break;
-
+        case 0x8E:
+            INST_NAME("MOV Seg,Ew");
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                ed = xRAX+(nextop&7)+(rex.b<<3);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 0, rex, NULL, 0, 0);
+                LDRH_U12(x1, ed, fixedaddress);
+                ed = x1;
+            }
+            STRw_U12(ed, xEmu, offsetof(x64emu_t, segs[(nextop&0x38)>>3]));
+            STRw_U12(wZR, xEmu, offsetof(x64emu_t, segs_serial[(nextop&0x38)>>3]));
+            break;
        case 0x8F:
            INST_NAME("POP Ed");
            nextop = F8;
@@ -1005,7 +1051,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                POP1(xRAX+(nextop&7)+(rex.b<<3));
            } else {
                POP1(x2); // so this can handle POP [ESP] and maybe some variant too
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                if(ed==xRSP) {
                    STRx_U12(x2, ed, fixedaddress);
                } else {
@@ -1162,7 +1208,8 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            case 1:
            case 2:
                if(rep==1) {INST_NAME("REPNZ CMPSB");} else {INST_NAME("REPZ CMPSB");}
-                SETFLAGS(X_ALL, SF_MAYSET);
+                MAYSETFLAGS();
+                SETFLAGS(X_ALL, SF_SET_PENDING);
                CBZx_NEXT(xRCX);
                TBNZ_MARK2(xFlags, F_DF);
                MARK;   // Part with DF==0
@@ -1256,12 +1303,33 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            }
            break;

+        case 0xAD:
+            if(rep) {
+                INST_NAME("REP LODSD");
+            } else {
+                INST_NAME("LODSD");
+            }
+            LDRxw_U12(xRAX, xRSI, 0);
+            BFIw(x2, xFlags, F_DF, 1);
+            if(rex.w) {
+                MOV32w(x1, 8);
+            } else {
+                MOV32w(x1, 4);
+            }
+            EORxw_REG_LSL(x1, x1, x2, rex.w?63:31);
+            if(rep) {
+                MULxw(x1, x1, xRCX);
+                EORw_REG(xRCX, xRCX, xRCX);
+            }
+            ADDx_REG(xRSI, xRSI, x1);
+            break;
        case 0xAE:
            switch(rep) {
            case 1:
            case 2:
                if(rep==1) {INST_NAME("REPNZ SCASB");} else {INST_NAME("REPZ SCASB");}
-                SETFLAGS(X_ALL, SF_MAYSET);
+                MAYSETFLAGS();
+                SETFLAGS(X_ALL, SF_SET_PENDING);
                CBZx_NEXT(xRCX);
                UBFXw(x1, xRAX, 0, 8);
                TBNZ_MARK2(xFlags, F_DF);
@@ -1451,7 +1519,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("ROL Ed, Ib");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETED(1);
                    u8 = (F8)&(rex.w?0x3f:0x1f);
                    emit_rol32c(dyn, ninst, rex, ed, u8, x3, x4);
@@ -1459,7 +1527,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 1:
                    INST_NAME("ROR Ed, Ib");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETED(1);
                    u8 = (F8)&(rex.w?0x3f:0x1f);
                    emit_ror32c(dyn, ninst, rex, ed, u8, x3, x4);
@@ -1521,8 +1589,10 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC2:
            INST_NAME("RETN");
            //SETFLAGS(X_ALL, SF_SET);    // Hack, set all flags (to an unknown state...)
-            READFLAGS(X_PEND);  // lets play safe here too
-            BARRIER(2);
+            if(box64_dynarec_safeflags) {
+                READFLAGS(X_PEND);  // lets play safe here too
+            }
+            BARRIER(BARRIER_FLOAT);
            i32 = F16;
            retn_to_epilog(dyn, ninst, i32);
            *need_epilog = 0;
@@ -1531,8 +1601,10 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC3:
            INST_NAME("RET");
            // SETFLAGS(X_ALL, SF_SET);    // Hack, set all flags (to an unknown state...)
-            READFLAGS(X_PEND);  // so instead, force the defered flags, so it's not too slow, and flags are not lost
-            BARRIER(2);
+            if(box64_dynarec_safeflags) {
+                READFLAGS(X_PEND);  // so instead, force the defered flags, so it's not too slow, and flags are not lost
+            }
+            BARRIER(BARRIER_FLOAT);
            ret_to_epilog(dyn, ninst);
            *need_epilog = 0;
            *ok = 0;
@@ -1554,10 +1626,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                MOV32w(x3, u8);
                BFIx(eb1, x3, eb2*8, 8);
            } else {                    // mem <= u8
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff, 0, rex, 0, 1);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff, 0, rex, &lock, 0, 1);
                u8 = F8;
                MOV32w(x3, u8);
                STRB_U12(x3, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;
        case 0xC7:
@@ -1568,10 +1644,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                ed = xRAX+(nextop&7)+(rex.b<<3);
                MOV64xw(ed, i64);
            } else {                    // mem <= i32
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, 4);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 4);
                i64 = F32S;
                MOV64xw(x3, i64);
                STRxw_U12(x3, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;

@@ -1585,7 +1665,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            SETFLAGS(X_ALL, SF_SET);    // Hack, set all flags (to an unknown state...)
            if(PK(0)=='S' && PK(1)=='C') {
                addr+=2;
-                BARRIER(2);
+                BARRIER(BARRIER_FLOAT);
                INST_NAME("Special Box64 instruction");
                if((PK64(0)==0))
                {
@@ -1602,13 +1682,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    x87_forget(dyn, ninst, x3, x4, 0);
                    sse_purge07cache(dyn, ninst, x3);
                    tmp = isSimpleWrapper(*(wrapper_t*)(addr));
-                    if(box64_log<2 && tmp) {
+                    if((box64_log<2 || !cycle_log) && tmp) {
                        //GETIP(ip+3+8+8); // read the 0xCC
                        call_n(dyn, ninst, *(void**)(addr+8), tmp);
                        addr+=8+8;
                    } else {
                        GETIP(ip+1); // read the 0xCC
                        STORE_XEMU_CALL(xRIP);
+                        ADDx_U12(x1, xEmu, (uint32_t)offsetof(x64emu_t, ip)); // setup addr as &emu->ip
                        CALL_S(x64Int3, -1);
                        LOAD_XEMU_CALL(xRIP);
                        addr+=8+8;
@@ -1643,7 +1724,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xCF:
            INST_NAME("IRET");
            SETFLAGS(X_ALL, SF_SET);    // Not a hack, EFLAGS are restored
-            BARRIER(2);
+            BARRIER(BARRIER_FLOAT);
            iret_to_epilog(dyn, ninst, rex.w);
            *need_epilog = 0;
            *ok = 0;
@@ -1756,14 +1837,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("ROL Ed, 1");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETED(0);
                    emit_rol32c(dyn, ninst, rex, ed, 1, x3, x4);
                    WBACK;
                    break;
                case 1:
                    INST_NAME("ROR Ed, 1");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETED(0);
                    emit_ror32c(dyn, ninst, rex, ed, 1, x3, x4);
                    WBACK;
@@ -1775,7 +1856,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_OF|X_CF, SF_SET);
                    MOV32w(x2, 1);
                    GETEDW(x4, x1, 0);
-                    CALL_(rcl32, ed, x4);
+                    CALL_(rex.w?((void*)rcl64):((void*)rcl32), ed, x4);
                    WBACK;
                    break;
                case 3:
@@ -1785,7 +1866,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_OF|X_CF, SF_SET);
                    MOV32w(x2, 1);
                    GETEDW(x4, x1, 0);
-                    CALL_(rcr32, ed, x4);
+                    CALL_(rex.w?((void*)rcr64):((void*)rcr32), ed, x4);
                    WBACK;
                    break;
                case 4:
@@ -1825,7 +1906,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    }
                    MOV64xw(x4, (rex.w?64:32));
                    SUBx_REG(x3, x4, x3);
-                    GETEDW(x4, x2, 0);
+                    GETED(0);
                    if(!rex.w && MODREG) {MOVw_REG(ed, ed);}
                    B_NEXT(cEQ);
                    RORxw_REG(ed, ed, x3);
@@ -1849,7 +1930,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    } else {
                        ANDSw_mask(x3, xRCX, 0, 0b00100);  //mask=0x00000001f
                    }
-                    GETEDW(x4, x2, 0);
+                    GETED(0);
                    if(!rex.w && MODREG) {MOVw_REG(ed, ed);}
                    B_NEXT(cEQ);
                    RORxw_REG(ed, ed, x3);
@@ -1963,24 +2044,33 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xDD:
            addr = dynarec64_DD(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);
            break;
-
+        case 0xDE:
+            addr = dynarec64_DE(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);
+            break;
        case 0xDF:
            addr = dynarec64_DF(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);
            break;
        #define GO(Z)                                                   \
-            BARRIER(2);                                                 \
-            JUMP(addr+i8);                                              \
-            if(dyn->insts) {                                            \
+            BARRIER(BARRIER_MAYBE);                                     \
+            JUMP(addr+i8, 1);                                           \
+            if(dyn->insts[ninst].x64.jmp_insts==-1 ||                   \
+                CHECK_CACHE()) {                                        \
+                /* out of the block */                                  \
+                i32 = dyn->insts[ninst].epilog-(dyn->native_size);      \
+                if(Z) {CBNZx(xRCX, i32);} else {CBZx(xRCX, i32);};      \
                if(dyn->insts[ninst].x64.jmp_insts==-1) {               \
-                    /* out of the block */                              \
-                    i32 = dyn->insts[ninst+1].address-(dyn->arm_size);  \
-                    if(Z) {CBNZx(xRCX, i32);} else {CBZx(xRCX, i32);};  \
+                    if(!(dyn->insts[ninst].x64.barrier&BARRIER_FLOAT))  \
+                        fpu_purgecache(dyn, ninst, 1, x1, x2, x3);      \
                    jump_to_next(dyn, addr+i8, 0, ninst);               \
                } else {                                                \
-                    /* inside the block */                              \
-                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->arm_size);    \
-                    if(Z) {CBZx(xRCX, i32);} else {CBNZx(xRCX, i32);};  \
-                }   \
+                    CacheTransform(dyn, ninst, cacheupd, x1, x2, x3);   \
+                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);    \
+                    Bcond(c__, i32);                                    \
+                }                                                       \
+            } else {                                                    \
+                /* inside the block */                                  \
+                i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);    \
+                if(Z) {CBZx(xRCX, i32);} else {CBNZx(xRCX, i32);};      \
            }
        case 0xE0:
            INST_NAME("LOOPNZ");
@@ -2019,12 +2109,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                printf_log(LOG_INFO, "Warning, CALL to 0x0 at %p (%p)\n", (void*)addr, (void*)(addr-1));
                #endif
            }
-            #if STEP == 0
-            if(isNativeCall(dyn, addr+i32, NULL, NULL))
-                tmp = 3;
-            else 
-                tmp = 0;
-            #elif STEP < 2
+            #if STEP < 2
            if(isNativeCall(dyn, addr+i32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
                tmp = dyn->insts[ninst].pass2choice = 3;
            else 
@@ -2035,20 +2120,26 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch(tmp) {
                case 3:
                    SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
-                    BARRIER(1);
-                    BARRIER_NEXT(1);
-                    TABLE64(x2, addr);
+                    BARRIER(BARRIER_FULL);
+                    //BARRIER_NEXT(BARRIER_FULL);
+                    if(dyn->last_ip && (addr-dyn->last_ip<0x1000)) {
+                        ADDx_U12(x2, xRIP, addr-dyn->last_ip);
+                    } else {
+                        TABLE64(x2, addr);
+                    }
                    PUSH1(x2);
                    MESSAGE(LOG_DUMP, "Native Call to %s (retn=%d)\n", GetNativeName(GetNativeFnc(dyn->insts[ninst].natcall-1)), dyn->insts[ninst].retn);
                    // calling a native function
                    sse_purge07cache(dyn, ninst, x3);
-                    if(box64_log<2 && dyn->insts && (tmp=isSimpleWrapper(*(wrapper_t*)(dyn->insts[ninst].natcall+2)))) {
+                    if((box64_log<2 || !cycle_log) && dyn->insts[ninst].natcall && (tmp=isSimpleWrapper(*(wrapper_t*)(dyn->insts[ninst].natcall+2)))) {
                        //GETIP(ip+3+8+8); // read the 0xCC
                        call_n(dyn, ninst, *(void**)(dyn->insts[ninst].natcall+2+8), tmp);
                        POP1(xRIP);   // pop the return address
+                        dyn->last_ip = addr;
                    } else {
                        GETIP_(dyn->insts[ninst].natcall); // read the 0xCC already
                        STORE_XEMU_CALL(xRIP);
+                        ADDx_U12(x1, xEmu, (uint32_t)offsetof(x64emu_t, ip)); // setup addr as &emu->ip
                        CALL_S(x64Int3, -1);
                        LOAD_XEMU_CALL(xRIP);
                        TABLE64(x3, dyn->insts[ninst].natcall);
@@ -2067,21 +2158,20 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MARK;
                        LOAD_XEMU_REM();    // load remaining register, has they have changed
                        jump_to_epilog(dyn, 0, xRIP, ninst);
+                        dyn->last_ip = addr;
                    }
                    break;
                default:
-                    if(ninst && dyn->insts && dyn->insts[ninst-1].x64.set_flags) {
+                    if((box64_dynarec_safeflags>1) || (ninst && dyn->insts[ninst-1].x64.set_flags)) {
                        READFLAGS(X_PEND);  // that's suspicious
                    } else {
                        SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    }
                    // regular call
-                    BARRIER(1);
-                    BARRIER_NEXT(1);
-                    if(!dyn->insts || ninst==dyn->size-1) {
-                        *need_epilog = 0;
-                        *ok = 0;
-                    }
+                    BARRIER(BARRIER_FULL);
+                    //BARRIER_NEXT(1);
+                    *need_epilog = 0;
+                    *ok = 0;
                    TABLE64(x2, addr);
                    PUSH1(x2);
                    if(addr+i32==0) {   // self modifying code maybe? so use indirect address fetching
@@ -2095,7 +2185,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            break;
        case 0xE9:
        case 0xEB:
-            BARRIER(1);
+            BARRIER(BARRIER_MAYBE);
            if(opcode==0xE9) {
                INST_NAME("JMP Id");
                i32 = F32S;
@@ -2103,19 +2193,19 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                INST_NAME("JMP Ib");
                i32 = F8S;
            }
-            JUMP(addr+i32);
-            if(dyn->insts) {
-                PASS2IF(dyn->insts[ninst].x64.jmp_insts==-1, 1) {
-                    // out of the block
-                    jump_to_next(dyn, addr+i32, 0, ninst);
+            JUMP(addr+i32, 0);
+            if(dyn->insts[ninst].x64.jmp_insts==-1) {
+                // out of the block
+                fpu_purgecache(dyn, ninst, 1, x1, x2, x3);
+                jump_to_next(dyn, addr+i32, 0, ninst);
+            } else {
+                // inside the block
+                CacheTransform(dyn, ninst, CHECK_CACHE(), x1, x2, x3);
+                tmp = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);
+                if(tmp==4) {
+                    NOP;
                } else {
-                    // inside the block
-                    tmp = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->arm_size);
-                    if(tmp==4) {
-                        NOP;
-                    } else {
-                        B(tmp);
-                    }
+                    B(tmp);
                }
            }
            *need_epilog = 0;
@@ -2271,7 +2361,7 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOVw_REG(xRAX, x2);
                        MOVw_REG(xRDX, x4);
                    } else {
-                        if(ninst && dyn->insts 
+                        if(ninst 
                           && dyn->insts[ninst-1].x64.addr 
                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr)==0x31 
                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr+1)==0xD2) {
@@ -2318,9 +2408,15 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                            MOVx_REG(xRAX, x2);
                        } else {
                            GETEDH(x1, 0);  // get edd changed addr, so cannot be called 2 times for same op...
-                            CBZxw_MARK(xRDX);
+                            //Need to see if RDX==0 and RAX not signed
+                            // or RDX==-1 and RAX signed
+                            CBNZx_MARK2(xRDX);
+                            TBZ_MARK(xRAX, 31);
+                            MARK2;
                            MVNx_REG(x2, xRDX);
-                            CBZxw_MARK(x2);
+                            CBNZx_MARK3(x2);
+                            TBNZ_MARK(xRAX, 31);
+                            MARK3;
                            if(ed!=x1) {MOVx_REG(x1, ed);}
                            CALL((void*)idiv64, -1);
                            B_NEXT_nocond;
@@ -2334,6 +2430,18 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
            }
            break;
+        case 0xF8:
+            INST_NAME("CLC");
+            SETFLAGS(X_CF, SF_SUBSET);
+            SET_DFNONE(x1);
+            BFCx(xFlags, F_CF, 1);
+            break;
+        case 0xF9:
+            INST_NAME("STC");
+            SETFLAGS(X_CF, SF_SUBSET);
+            SET_DFNONE(x1);
+            ORRx_mask(xFlags, xFlags, 1, 0, 0); // xFlags | 1
+            break;
        
        case 0xFC:
            INST_NAME("CLD");
@@ -2349,14 +2457,14 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("INC Eb");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEB(x1, 0);
                    emit_inc8(dyn, ninst, x1, x2, x4);
                    EBBACK;
                    break;
                case 1:
                    INST_NAME("DEC Eb");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEB(x1, 0);
                    emit_dec8(dyn, ninst, x1, x2, x4);
                    EBBACK;
@@ -2370,22 +2478,22 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0: // INC Ed
                    INST_NAME("INC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETED(0);
                    emit_inc32(dyn, ninst, rex, ed, x3, x4);
                    WBACK;
                    break;
                case 1: //DEC Ed
                    INST_NAME("DEC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETED(0);
                    emit_dec32(dyn, ninst, rex, ed, x3, x4);
                    WBACK;
                    break;
                case 2: // CALL Ed
                    INST_NAME("CALL Ed");
-                    PASS2IF(dyn->insts && 
-                           ((ninst && dyn->insts[ninst-1].x64.set_flags)
+                    PASS2IF((box64_dynarec_safeflags>1) || 
+                        ((ninst && dyn->insts[ninst-1].x64.set_flags)
                        || ((ninst>1) && dyn->insts[ninst-2].x64.set_flags)), 1)
                    {
                        READFLAGS(X_PEND);          // that's suspicious
@@ -2393,19 +2501,19 @@ uintptr_t dynarec64_00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
                    }
                    GETEDx(0);
-                    BARRIER(1);
-                    BARRIER_NEXT(1);
+                    BARRIER(BARRIER_FLOAT);
+                    //BARRIER_NEXT(BARRIER_FULL);
                    if(!dyn->insts || ninst==dyn->size-1) {
                        *need_epilog = 0;
                        *ok = 0;
                    }
-                    GETIP(addr);
+                    GETIP_(addr);
                    PUSH1(xRIP);
                    jump_to_next(dyn, 0, ed, ninst);
                    break;
                case 4: // JMP Ed
                    INST_NAME("JMP Ed");
-                    BARRIER(1);
+                    BARRIER(BARRIER_FULL);
                    GETEDx(0);
                    jump_to_next(dyn, 0, ed, ninst);
                    *need_epilog = 0;
@@ -15,10 +15,10 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -31,10 +31,10 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    uint8_t opcode = F8;
    uint8_t nextop;
    uint8_t u8;
-    uint8_t gd, ed, eb1, eb2;
+    uint8_t gd, ed, eb1, eb2, gb1, gb2;
    uint8_t wback, wb1, wb2, wb;
    int64_t i64, j64;
-    int v0;
+    int v0, v1;
    int q0;
    int d0;
    int64_t fixedaddress;
@@ -42,10 +42,13 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    MAYUSE(eb2);
    MAYUSE(wb1);
    MAYUSE(wb2);
+    MAYUSE(gb1);
+    MAYUSE(gb2);
    MAYUSE(j64);
    MAYUSE(d0);
    MAYUSE(q0);
    MAYUSE(v0);
+    MAYUSE(v1);

    while((opcode==0xF2) || (opcode==0xF3)) {
        rep = opcode-0xF1;
@@ -82,13 +85,13 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                            GETG;
                            if(MODREG) {
                                ed = (nextop&7)+ (rex.b<<3);
-                                v0 = sse_get_reg(dyn, ninst, x1, gd);
-                                d0 = sse_get_reg(dyn, ninst, x1, ed);
+                                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
+                                d0 = sse_get_reg(dyn, ninst, x1, ed, 0);
                                VMOVeD(v0, 0, d0, 0);
                            } else {
                                grab_segdata(dyn, addr, ninst, x4, seg);
                                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                                ADDx_REG(x4, x4, ed);
                                VLDR64_U12(v0, x4, fixedaddress); // upper part reseted
                            }
@@ -98,13 +101,13 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                            nextop = F8;
                            GETG;
                            if(MODREG) {
-                                v0 = sse_get_reg(dyn, ninst, x1, gd);
-                                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
+                                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
+                                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 0);
                                VMOVeS(v0, 0, q0, 0);
                            } else {
                                grab_segdata(dyn, addr, ninst, x4, seg);
                                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                                ADDx_REG(x4, x4, ed);
                                VLDR32_U12(v0, x4, fixedaddress);
                            }
@@ -115,18 +118,34 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 0x11:
                    switch(rep) {
+                        case 0:
+                            INST_NAME("MOVUPS Ex,Gx");
+                            nextop = F8;
+                            GETG;
+                            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
+                            if(MODREG) {
+                                ed = (nextop&7)+(rex.b<<3);
+                                v1 = sse_get_reg_empty(dyn, ninst, x1, ed);
+                                VMOVQ(v1, v0);
+                            } else {
+                                grab_segdata(dyn, addr, ninst, x4, seg);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
+                                ADDx_REG(x4, x4, ed);
+                                VSTR128_U12(v0, x4, fixedaddress);
+                            }
+                            break;
                        case 1:
                            INST_NAME("MOVSD Ex, Gx");
                            nextop = F8;
                            GETG;
-                            v0 = sse_get_reg(dyn, ninst, x1, gd);
+                            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
                            if(MODREG) {
                                ed = (nextop&7)+ (rex.b<<3);
-                                d0 = sse_get_reg(dyn, ninst, x1, ed);
+                                d0 = sse_get_reg(dyn, ninst, x1, ed, 1);
                                VMOVeD(d0, 0, v0, 0);
                            } else {
                                grab_segdata(dyn, addr, ninst, x4, seg);
-                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                                ADDx_REG(x4, x4, ed);
                                VSTR64_U12(v0, x4, fixedaddress);
                            }
@@ -135,13 +154,13 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                            INST_NAME("MOVSS Ex, Gx");
                            nextop = F8;
                            GETG;
-                            v0 = sse_get_reg(dyn, ninst, x1, gd);
+                            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
                            if(MODREG) {
-                                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
+                                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 1);
                                VMOVeS(q0, 0, v0, 0);
                            } else {
                                grab_segdata(dyn, addr, ninst, x4, seg);
-                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                                ADDx_REG(x4, x4, ed);
                                VSTR32_U12(v0, x4, fixedaddress);
                            }
@@ -151,6 +170,29 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    }
                    break;

+                case 0x6F:
+                    switch(rep) {
+                        case 2:
+                            INST_NAME("MOVDQU Gx,Ex");// no alignment constraint on NEON here, so same as MOVDQA
+                            nextop = F8;
+                            GETG;
+                            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                            if(MODREG) {
+                                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
+                                VMOVQ(v0, v1);
+                            } else {
+                                grab_segdata(dyn, addr, ninst, x4, seg);
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
+                                ADDx_REG(x4, x4, ed);
+                                VLDR128_U12(v0, ed, fixedaddress);
+                            }
+                            break;
+                        default:
+                            DEFAULT;
+                    }
+                    break;
+
+
                case 0xAF:
                    INST_NAME("IMUL Gd, Ed");
                    SETFLAGS(X_ALL, SF_PENDING);
@@ -184,11 +226,42 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    }
                    break;

+                case 0xB6:
+                    INST_NAME("MOVZX Gd, Eb");
+                    nextop = F8;
+                    grab_segdata(dyn, addr, ninst, x4, seg);
+                    GETGD;
+                    if(MODREG) {
+                        if(rex.rex) {
+                            eb1 = xRAX+(nextop&7)+(rex.b<<3);
+                            eb2 = 0;                \
+                        } else {
+                            ed = (nextop&7);
+                            eb1 = xRAX+(ed&3);  // Ax, Cx, Dx or Bx
+                            eb2 = (ed&4)>>2;    // L or H
+                        }
+                        UBFXxw(gd, eb1, eb2*8, 8);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                        LDRB_REG(gd, ed, x4);
+                    }
+                    break;
+
                default:
                    DEFAULT;
            }
            break;

+        case 0x2B:
+            INST_NAME("SUB Gd, Seg:Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            grab_segdata(dyn, addr, ninst, x4, seg);
+            nextop = F8;
+            GETGD;
+            GETEDO(x4, 0);
+            emit_sub32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+
        case 0x33:
            INST_NAME("XOR Gd, Seg:Ed");
            SETFLAGS(X_ALL, SF_SET_PENDING);
@@ -209,16 +282,49 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            emit_cmp32(dyn, ninst, rex, ed, gd, x3, x4, x5);
            break;

+        case 0x3B:
+            INST_NAME("CMP Gd, Seg:Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            grab_segdata(dyn, addr, ninst, x4, seg);
+            nextop = F8;
+            GETGD;
+            GETEDO(x4, 0);
+            emit_cmp32(dyn, ninst, rex, gd, ed, x3, x4, x5);
+            break;
+
+        case 0x63:
+            INST_NAME("MOVSXD Gd, Ed");
+            nextop = F8;
+            GETGD;
+            if(rex.w) {
+                if(MODREG) {   // reg <= reg
+                    SXTWx(gd, xRAX+(nextop&7)+(rex.b<<3));
+                } else {                    // mem <= reg
+                    grab_segdata(dyn, addr, ninst, x4, seg);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                    LDRSW_REG(gd, ed, x4);
+                }
+            } else {
+                if(MODREG) {   // reg <= reg
+                    MOVw_REG(gd, xRAX+(nextop&7)+(rex.b<<3));
+                } else {                    // mem <= reg
+                    grab_segdata(dyn, addr, ninst, x4, seg);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                    LDRw_REG(gd, ed, x4);
+                }
+            }
+            break;
+
        case 0x66:
            addr = dynarec64_6664(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);
            break;

        case 0x80:
            nextop = F8;
-            grab_segdata(dyn, addr, ninst, x1, seg);
            switch((nextop>>3)&7) {
                case 0: //ADD
                    INST_NAME("ADD Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -227,6 +333,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 1: //OR
                    INST_NAME("OR Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -235,6 +342,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 2: //ADC
                    INST_NAME("ADC Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    READFLAGS(X_CF);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
@@ -244,6 +352,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 3: //SBB
                    INST_NAME("SBB Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    READFLAGS(X_CF);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
@@ -253,6 +362,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 4: //AND
                    INST_NAME("AND Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -261,6 +371,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 5: //SUB
                    INST_NAME("SUB Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -269,6 +380,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 6: //XOR
                    INST_NAME("XOR Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -277,6 +389,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 7: //CMP
                    INST_NAME("CMP Eb, Ib");
+                    grab_segdata(dyn, addr, ninst, x1, seg);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEBO(x1, 1);
                    u8 = F8;
@@ -369,7 +482,40 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
            }
            break;
-            
+        case 0x8A:
+            INST_NAME("MOV Gb, Eb");
+            nextop = F8;
+            if(rex.rex) {
+                gb1 = gd = xRAX+((nextop&0x38)>>3)+(rex.r<<3);
+                gb2=0;
+            } else {
+                gd = (nextop&0x38)>>3;
+                gb1 = xRAX+(gd&3);
+                gb2 = ((gd&4)>>2);
+            }
+            if(MODREG) {
+                if(rex.rex) {
+                    wback = xRAX+(nextop&7)+(rex.b<<3);
+                    wb2 = 0;
+                } else {
+                    wback = (nextop&7);
+                    wb2 = (wback>>2);
+                    wback = xRAX+(wback&3);
+                }
+                if(wb2) {
+                    UBFXw(x4, wback, wb2*8, 8);
+                    ed = x4;
+                } else {
+                    ed = wback;
+                }
+            } else {
+                grab_segdata(dyn, addr, ninst, x4, seg);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                LDRB_REG(x4, wback, x4);
+                ed = x4;
+            }
+            BFIx(gb1, ed, gb2*8, 8);
+            break;
        case 0x89:
            INST_NAME("MOV Seg:Ed, Gd");
            grab_segdata(dyn, addr, ninst, x4, seg);
@@ -378,7 +524,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            if(MODREG) {   // reg <= reg
                MOVxw_REG(xRAX+(nextop&7)+(rex.b<<3), gd);
            } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                STRxw_REG(gd, ed, x4);
            }
            break;
@@ -391,7 +537,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            if(MODREG) {   // reg <= reg
                MOVxw_REG(gd, xRAX+(nextop&7)+(rex.b<<3));
            } else {                    // mem <= reg
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                LDRxw_REG(gd, ed, x4);
            }
            break;
@@ -413,7 +559,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                MOV32w(x3, u8);
                BFIx(eb1, x3, eb2*8, 8);
            } else {                    // mem <= u8
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 1);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 1);
                u8 = F8;
                MOV32w(x3, u8);
                STRB_REG(x3, ed, x4);
@@ -428,7 +574,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                ed = xRAX+(nextop&7)+(rex.b<<3);
                MOV64xw(ed, i64);
            } else {                    // mem <= i32
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 4);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 4);
                i64 = F32S;
                MOV64xw(x3, i64);
                STRxw_REG(x3, ed, x4);
@@ -441,14 +587,14 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("ROL Ed, 1");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETEDO(x6, 0);
                    emit_rol32c(dyn, ninst, rex, ed, 1, x3, x4);
                    WBACKO(x6);
                    break;
                case 1:
                    INST_NAME("ROR Ed, 1");
-                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETEDO(x6, 0);
                    emit_ror32c(dyn, ninst, rex, ed, 1, x3, x4);
                    WBACKO(x6);
@@ -717,7 +863,7 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOVw_REG(xRAX, x2);
                        MOVw_REG(xRDX, x4);
                    } else {
-                        if(ninst && dyn->insts 
+                        if(ninst
                           && dyn->insts[ninst-1].x64.addr 
                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr)==0x31 
                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr+1)==0xD2) {
@@ -787,22 +933,22 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0: // INC Ed
                    INST_NAME("INC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEDO(x6, 0);
                    emit_inc32(dyn, ninst, rex, ed, x3, x4);
                    WBACKO(x6);
                    break;
                case 1: //DEC Ed
                    INST_NAME("DEC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEDO(x6, 0);
                    emit_dec32(dyn, ninst, rex, ed, x3, x4);
                    WBACKO(x6);
                    break;
                case 2: // CALL Ed
                    INST_NAME("CALL Ed");
-                    PASS2IF(dyn->insts && 
-                           ((ninst && dyn->insts[ninst-1].x64.set_flags)
+                    PASS2IF((box64_dynarec_safeflags>1) ||
+                        ((ninst && dyn->insts[ninst-1].x64.set_flags)
                        || ((ninst>1) && dyn->insts[ninst-2].x64.set_flags)), 1)
                    {
                        READFLAGS(X_PEND);          // that's suspicious
@@ -810,13 +956,12 @@ uintptr_t dynarec64_64(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
                    }
                    GETEDOx(x6, 0);
-                    BARRIER(1);
-                    BARRIER_NEXT(1);
+                    BARRIER(BARRIER_FLOAT);
                    if(!dyn->insts || ninst==dyn->size-1) {
                        *need_epilog = 0;
                        *ok = 0;
                    }
-                    GETIP(addr);
+                    GETIP_(addr);
                    PUSH1(xRIP);
                    jump_to_next(dyn, 0, ed, ninst);
                    break;
@@ -15,10 +15,10 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -35,19 +35,21 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    uint8_t gd, ed;
    uint8_t wback, wb1;
    int64_t fixedaddress;
+    int lock;
    MAYUSE(u8);
    MAYUSE(u16);
    MAYUSE(u64);
    MAYUSE(j64);
+    MAYUSE(lock);

-    while((opcode==0x2E) || (opcode==0x66))   // ignoring CS: or multiple 0x66
+    while((opcode==0x2E) || (opcode==0x36) || (opcode==0x66))   // ignoring CS:, SS: or multiple 0x66
        opcode = F8;

    while((opcode==0xF2) || (opcode==0xF3)) {
        rep = opcode-0xF1;
        opcode = F8;
    }
-    // REX prefix before the F0 are ignored
+    // REX prefix before the 66 are ignored
    rex.rex = 0;
    while(opcode>=0x40 && opcode<=0x4f) {
        rex.rex = opcode;
@@ -172,9 +174,9 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            INST_NAME("SBB AX, Iw");
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
-            i16 = F16;
+            i16 = F16S;
            UXTHw(x1, xRAX);
-            MOV32w(x2, i16);
+            MOVZw(x2, i16);
            emit_sbb16(dyn, ninst, x1, x2, x3, x4);
            BFIx(xRAX, x1, 0, 16);
            break;
@@ -301,12 +303,17 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            break;

        case 0x69:
-            INST_NAME("IMUL Gw,Ew,Iw");
+        case 0x6B:
+            if(opcode==0x69) {
+                INST_NAME("IMUL Gw,Ew,Iw");
+            } else {
+                INST_NAME("IMUL Gw,Ew,Ib");
+            }
            SETFLAGS(X_ALL, SF_PENDING);
            nextop = F8;
            UFLAG_DF(x1, d_imul16);
-            GETSEW(x1, 2);
-            i32 = F16S;
+            GETSEW(x1, (opcode==0x69)?2:1);
+            if(opcode==0x69) i32 = F16S; else i32 = F8S;
            MOV32w(x2, i32);
            MULw(x2, x2, x1);
            UFLAG_RES(x2);
@@ -314,34 +321,35 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            GWBACK;
            break;

-        case 0x6B:
-            INST_NAME("IMUL Gw,Ew,Ib");
-            SETFLAGS(X_ALL, SF_PENDING);
-            nextop = F8;
-            UFLAG_DF(x1, d_imul16);
-            GETSEW(x1, 1);
-            i32 = F8S;
-            MOV32w(x2, i32);
-            MULw(x2, x2, x1);
-            UFLAG_RES(x2);
-            gd=x2;
-            GWBACK;
-            break;
+        case 0x70:
+        case 0x71:
+        case 0x72:
+        case 0x73:
+        case 0x74:
+        case 0x75:
+        case 0x76:
+        case 0x77:
+        case 0x78:
+        case 0x79:
+        case 0x7a:
+        case 0x7b:
+        case 0x7c:
+        case 0x7d:
+        case 0x7e:
+        case 0x7f:
+            // just use regular conditional jump
+            return dynarec64_00(dyn, addr-1, ip, ninst, rex, rep, ok, need_epilog);

        case 0x81:
        case 0x83:
            nextop = F8;
            switch((nextop>>3)&7) {
                case 0: //ADD
-                    if(opcode==0x81) {
-                        INST_NAME("ADD Ew, Iw");
-                    } else {
-                        INST_NAME("ADD Ew, Ib");
-                    }
+                    if(opcode==0x81) {INST_NAME("ADD Ew, Iw");} else {INST_NAME("ADD Ew, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_add16(dyn, ninst, ed, x5, x2, x4);
                    EWBACK;
                    break;
@@ -350,7 +358,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_or16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -360,7 +368,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_adc16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -370,7 +378,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_sbb16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -379,7 +387,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_and16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -388,7 +396,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_sub16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -397,7 +405,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
-                    MOV32w(x5, i16);
+                    MOVZw(x5, i16);
                    emit_xor16(dyn, ninst, x1, x5, x2, x4);
                    EWBACK;
                    break;
@@ -407,7 +415,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    GETEW(x1, (opcode==0x81)?2:1);
                    if(opcode==0x81) i16 = F16S; else i16 = F8S;
                    if(i16) {
-                        MOV32w(x2, i16);
+                        MOVZw(x2, i16);
                        emit_cmp16(dyn, ninst, x1, x2, x3, x4, x5);
                    } else
                        emit_cmp16_0(dyn, ninst, x1, x3, x4);
@@ -424,6 +432,35 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            emit_test16(dyn, ninst, x1, x2, x3, x4, x5);
            break;

+        case 0x87:
+            INST_NAME("(LOCK)XCHG Ew, Gw");
+            nextop = F8;
+            if(MODREG) {
+                GETGD;
+                GETED(0);
+                MOVxw_REG(x1, gd);
+                BFIx(gd, ed, 0, 16);
+                BFIx(ed, x1, 0, 16);
+            } else {
+                GETGD;
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                DMB_ISH();
+                TSTx_mask(ed, 1, 0, 0);    // mask=1
+                B_MARK(cNE);
+                MARKLOCK;
+                LDAXRH(x1, ed);
+                STLXRH(x3, gd, ed);
+                CBNZx_MARKLOCK(x3);
+                B_MARK2_nocond;
+                MARK;
+                LDRH_U12(x1, ed, 0);
+                STRH_U12(gd, ed, 0);
+                MARK2;
+                DMB_ISH();
+                BFIx(gd, x1, 0, 16);
+            }
+            break;
+            
        case 0x89:
            INST_NAME("MOV Ew, Gw");
            nextop = F8;
@@ -434,8 +471,12 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    BFIx(ed, gd, 0, 16);
                }
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, &lock, 0, 0);
                STRH_U12(gd, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;
        case 0x8B:
@@ -448,7 +489,11 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    BFIx(gd, ed, 0, 16);
                }
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, &lock, 0, 0);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
                LDRH_U12(x1, ed, fixedaddress);
                BFIx(gd, x1, 0, 16);
            }
@@ -620,7 +665,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_PENDING);
                    GETEW(x1, 1);
                    u8 = F8;
-                    MOV32w(x2, (u8&0x1f));
+                    UFLAG_IF {MOV32w(x2, (u8&0x1f));}
                    UFLAG_OP12(ed, x2)
                    LSLw_IMM(ed, ed, u8&0x1f);
                    EWBACK;
@@ -633,7 +678,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_PENDING);
                    GETEW(x1, 1);
                    u8 = F8;
-                    MOV32w(x2, (u8&0x1f));
+                    UFLAG_IF {MOV32w(x2, (u8&0x1f));}
                    UFLAG_OP12(ed, x2)
                    LSRw_IMM(ed, ed, u8&0x1f);
                    EWBACK;
@@ -644,11 +689,11 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    INST_NAME("SAR Ed, Ib");
                    SETFLAGS(X_ALL, SF_PENDING);
                    UFLAG_IF {MESSAGE(LOG_DUMP, "Need Optimization for flags\n");}
-                    GETSEW(x1, 0);
+                    GETSEW(x1, 1);
                    u8 = F8;
-                    MOV32w(x2, (u8&0x1f));
+                    UFLAG_IF {MOV32w(x2, (u8&0x1f));}
                    UFLAG_OP12(ed, x2)
-                    ASRw_REG(ed, ed, x2);
+                    ASRw(ed, ed, (u8&0x1f));
                    EWBACK;
                    UFLAG_RES(ed);
                    UFLAG_DF(x3, d_sar16);
@@ -665,10 +710,14 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                MOV32w(x1, u16);
                BFIx(ed, x1, 0, 16);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 2);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<1, 1, rex, &lock, 0, 2);
                u16 = F16;
                MOV32w(x1, u16);
                STRH_U12(x1, ed, fixedaddress);
+                if(lock || (box64_dynarec_strongmem && 
+                 (dyn->insts[ninst].x64.barrier || box64_dynarec_strongmem>1 || (dyn->insts[ninst+1].x64.barrier || dyn->insts[ninst+1].x64.jmp)))) {
+                    DMB_ISH();
+                }
            }
            break;

@@ -682,7 +731,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV32w(x2, 1);
                    } else {
                        INST_NAME("ROL Ew, CL");
-                        ANDSw_mask(x2, xRCX, 0, 0b00100);
+                        ANDw_mask(x2, xRCX, 0, 0b00100);
                    }
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    SETFLAGS(X_OF|X_CF, SF_SET);
@@ -696,7 +745,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV32w(x2, 1);
                    } else {
                        INST_NAME("ROR Ew, CL");
-                        ANDSw_mask(x2, xRCX, 0, 0b00100);
+                        ANDw_mask(x2, xRCX, 0, 0b00100);
                    }
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    SETFLAGS(X_OF|X_CF, SF_SET);
@@ -709,7 +758,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD1) {MOV32w(x2, 1);} else {ANDSw_mask(x2, xRCX, 0, 0b00100);}
+                    if(opcode==0xD1) {MOV32w(x2, 1);} else {ANDw_mask(x2, xRCX, 0, 0b00100);}
                    GETEW(x1, 0);
                    CALL_(rcl16, x1, x3);
                    EWBACK;
@@ -719,7 +768,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD1) {MOV32w(x2, 1);} else {ANDSw_mask(x2, xRCX, 0, 0b00100);}
+                    if(opcode==0xD1) {MOV32w(x2, 1);} else {ANDw_mask(x2, xRCX, 0, 0b00100);}
                    GETEW(x1, 0);
                    CALL_(rcr16, x1, x3);
                    EWBACK;
@@ -731,7 +780,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV32w(x4, 1);
                    } else {
                        INST_NAME("SHL Ew, CL");
-                        ANDSw_mask(x4, xRCX, 0, 0b00100);
+                        ANDw_mask(x4, xRCX, 0, 0b00100);
                    }
                    UFLAG_IF {MESSAGE(LOG_DUMP, "Need Optimization for flags\n");}
                    SETFLAGS(X_ALL, SF_PENDING);
@@ -748,7 +797,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV32w(x4, 1);
                    } else {
                        INST_NAME("SHR Ew, CL");
-                        ANDSw_mask(x4, xRCX, 0, 0b00100);
+                        ANDw_mask(x4, xRCX, 0, 0b00100);
                    }
                    UFLAG_IF {MESSAGE(LOG_DUMP, "Need Optimization for flags\n");}
                    SETFLAGS(X_ALL, SF_PENDING);
@@ -765,7 +814,7 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV32w(x4, 1);
                    } else {
                        INST_NAME("SAR Ew, CL");
-                        ANDSw_mask(x4, xRCX, 0, 0b00100);
+                        ANDw_mask(x4, xRCX, 0, 0b00100);
                    }
                    UFLAG_IF {MESSAGE(LOG_DUMP, "Need Optimization for flags\n");}
                    SETFLAGS(X_ALL, SF_PENDING);
@@ -779,6 +828,9 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            }
            break;
            
+        case 0xF0:
+            return dynarec64_66F0(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);
+
        case 0xF7:
            nextop = F8;
            switch((nextop>>3)&7) {
@@ -828,17 +880,25 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 6:
                    INST_NAME("DIV Ew");
-                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    SETFLAGS(X_ALL, SF_SET);
                    GETEW(x1, 0);
-                    CALL(div16, -1);
+                    UXTHw(x2, xRAX);
+                    BFIw(x2, xRDX, 16, 16);
+                    UDIVw(x3, x2, ed);
+                    MSUBw(x4, x3, ed, x2);  // x4 = x2 mod ed (i.e. x2 - x3*ed)
+                    BFIx(xRAX, x3, 0, 16);
+                    BFIx(xRDX, x4, 0, 16);
                    break;
                case 7:
                    INST_NAME("IDIV Ew");
-                    MESSAGE(LOG_DUMP, "Need Optimization\n");
                    SETFLAGS(X_ALL, SF_SET);
-                    GETEW(x1, 0);
-                    CALL(idiv16, -1);
+                    GETSEW(x1, 0);
+                    UXTHw(x2, xRAX);
+                    BFIw(x2, xRDX, 16, 16);
+                    SDIVw(x3, x2, ed);
+                    MSUBw(x4, x3, ed, x2);  // x4 = x2 mod ed (i.e. x2 - x3*ed)
+                    BFIx(xRAX, x3, 0, 16);
+                    BFIx(xRDX, x4, 0, 16);
                    break;
            }
            break;
@@ -848,14 +908,14 @@ uintptr_t dynarec64_66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("INC Ew");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEW(x1, 0);
                    emit_inc16(dyn, ninst, x1, x2, x4);
                    EWBACK;
                    break;
                case 1:
                    INST_NAME("DEC Ew");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
                    GETEW(x1, 0);
                    emit_dec16(dyn, ninst, x1, x2, x4);
                    EWBACK;
@@ -15,10 +15,10 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -31,8 +31,10 @@ uintptr_t dynarec64_6664(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
    uint8_t opcode = F8;
    uint8_t nextop;
    uint8_t gd, ed;
+    int64_t j64;
    int v0, v1;
    int64_t fixedaddress;
+    MAYUSE(j64);

    // REX prefix before the 66 are ignored
    rex.rex = 0;
@@ -51,18 +53,37 @@ uintptr_t dynarec64_6664(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x0F:
            opcode = F8;
            switch(opcode) {
-
+                /*  //Something is wrong, and this gives a SIGILL
+                case 0x2E:
+                    // no special check...
+                case 0x2F:
+                    if(opcode==0x2F) {INST_NAME("COMISD Gx, Ex");} else {INST_NAME("UCOMISD Gx, Ex");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    nextop = F8;
+                    GETG;
+                    if(MODREG) {
+                        v0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 0);
+                    } else {
+                        grab_segdata(dyn, addr, ninst, x4, _FS);
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                        v1 = fpu_get_scratch(dyn);                                                                       \
+                        VLDR128_REG(v1, ed, x4);
+                    }
+                    FCMPD(v0, v1);
+                    FCOMI(x1, x2);
+                    break;
+                */
            case 0xD6:
                INST_NAME("MOVQ Ex, Gx");
                nextop = F8;
                GETG;
-                v0 = sse_get_reg(dyn, ninst, x1, gd);
+                v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
                if(MODREG) {
                    v1 = sse_get_reg_empty(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
                    FMOVD(v1, v0);
                } else {
                    grab_segdata(dyn, addr, ninst, x4, _FS);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    VSTR64_REG(v0, ed, x4);
                }
                break;
@@ -87,7 +108,7 @@ uintptr_t dynarec64_6664(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                }
            } else {
                grab_segdata(dyn, addr, ninst, x4, _FS);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                if(rex.w) {
                    STRx_REG(gd, ed, x4);
                } else {
@@ -111,7 +132,7 @@ uintptr_t dynarec64_6664(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
                }
            } else {                    // mem <= reg
                grab_segdata(dyn, addr, ninst, x4, _FS);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                if(rex.w) {
                    LDRx_REG(gd, ed, x4);
                } else {
@@ -0,0 +1,412 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "emu/x64run_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "dynarec_native.h"
+
+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
+#include "dynarec_arm64_helper.h"
+#include "dynarec_arm64_functions.h"
+
+
+uintptr_t dynarec64_66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
+{
+    (void)ip; (void)rep; (void)need_epilog;
+
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    uint8_t gd, ed, u8;
+    uint8_t wback, wb1, wb2, gb1, gb2;
+    int32_t i32;
+    int64_t i64, j64;
+    int64_t fixedaddress;
+    MAYUSE(gb1);
+    MAYUSE(gb2);
+    MAYUSE(wb1);
+    MAYUSE(wb2);
+    MAYUSE(j64);
+
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+    // REX prefix before the F0/66 are ignored
+    rex.rex = 0;
+    while(opcode>=0x40 && opcode<=0x4f) {
+        rex.rex = opcode;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+        case 0x09:
+            INST_NAME("LOCK OR Ew, Gw");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGW(x5);
+            DMB_ISH();
+            if(MODREG) {
+                ed = xRAX+(nextop&7)+(rex.b<<3);
+                UXTHw(x6, ed);
+                emit_or16(dyn, ninst, x6, x5, x3, x4);
+                BFIx(ed, x6, 0, 16);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                MARKLOCK;
+                LDAXRH(x1, wback);
+                emit_or16(dyn, ninst, x1, x5, x3, x4);
+                STLXRH(x3, x1, wback);
+                CBNZx_MARKLOCK(x3);
+            }
+            DMB_ISH();
+            break;
+
+        case 0x0F:
+            opcode = F8;
+            switch(opcode) {
+
+                case 0xB1:
+                    INST_NAME("LOCK CMPXCHG Ew, Gw");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    nextop = F8;
+                    GETGD;
+                    DMB_ISH();
+                    UXTHw(x6, xRAX);
+                    if(MODREG) {
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        wback = 0;
+                        UXTHw(x1, ed);
+                        UFLAG_IF {emit_cmp16(dyn, ninst, x6, x1, x3, x4, x5);}
+                        CMPSxw_REG(x6, x1);
+                        B_MARK(cNE);
+                        BFIx(ed, gd, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                        TSTx_mask(wback, 1, 0, 0);    // mask=1
+                        B_MARK3(cNE);
+                        // Aligned version
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        CMPSw_REG(x6, x1);
+                        B_MARK(cNE);
+                        // EAX == Ed
+                        STLXRH(x4, gd, wback);
+                        CBNZx_MARKLOCK(x4);
+                        // done
+                        B_MARK_nocond;
+                        // Unaligned version
+                        MARK3;
+                        LDRH_U12(x1, wback, 0);
+                        LDAXRB(x3, wback); // dummy read, to arm the write...
+                        CMPSw_REG(x6, x1);
+                        B_MARK(cNE);
+                        // EAX == Ed
+                        STLXRB(x4, gd, wback);
+                        CBNZx_MARK3(x4);
+                        STRH_U12(gd, wback, 0);
+                    }
+                    MARK;
+                    // Common part (and fallback for EAX != Ed)
+                    UFLAG_IF {emit_cmp32(dyn, ninst, rex, x6, x1, x3, x4, x5);}
+                    BFIx(xRAX, x1, 0, 16);
+                    DMB_ISH();
+                    break;
+                    
+                case 0xC1:
+                    INST_NAME("LOCK XADD Gw, Ew");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    nextop = F8;
+                    gd = xRAX+((nextop&0x38)>>3)+(rex.r<<3);
+                    UXTHx(x5, gd);
+                    DMB_ISH();
+                    if(MODREG) {
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        BFIx(gd, ed, 0, 16);
+                        emit_add16(dyn, ninst, x5, gd, x3, x4);
+                        BFIx(ed, x5, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        ADDxw_REG(x4, x1, x5);
+                        STLXRH(x3, x4, wback);
+                        CBNZx_MARKLOCK(x3);
+                        IFX(X_ALL|X_PEND) {
+                            MOVxw_REG(x2, x1);
+                            emit_add16(dyn, ninst, x2, x5, x3, x4);
+                        }
+                        BFIx(gd, x1, 0, 16);
+                    }
+                    DMB_ISH();
+                    break;
+
+                default:
+                    DEFAULT;
+            }
+            break;
+
+        case 0x81:
+        case 0x83:
+            nextop = F8;
+            DMB_ISH();
+            switch((nextop>>3)&7) {
+                case 0: //ADD
+                    if(opcode==0x81) {
+                        INST_NAME("LOCK ADD Ew, Iw");
+                    } else {
+                        INST_NAME("LOCK ADD Ew, Ib");
+                    }
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_add16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        TSTx_mask(wback, 1, 0, 0);    // mask=1
+                        B_MARK(cNE);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_add16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                        B_NEXT_nocond;
+                        MARK;   // unaligned! also, not enough 
+                        LDRH_U12(x1, wback, 0);
+                        LDAXRB(x4, wback);
+                        BFIw(x1, x4, 0, 8); // re-inject
+                        emit_add16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRB(x3, x1, wback);
+                        CBNZx_MARK(x3);
+                        STRH_U12(x1, wback, 0);    // put the whole value
+                    }
+                    break;
+                case 1: //OR
+                    if(opcode==0x81) {INST_NAME("LOCK OR Ew, Iw");} else {INST_NAME("LOCK OR Ew, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_or16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_or16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                    }
+                    break;
+                case 2: //ADC
+                    if(opcode==0x81) {INST_NAME("LOCK ADC Ew, Iw");} else {INST_NAME("LOCK ADC Ew, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_adc16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_adc16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                    }
+                    break;
+                case 3: //SBB
+                    if(opcode==0x81) {INST_NAME("LOCK SBB Ew, Iw");} else {INST_NAME("LOCK SBB Ew, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_sbb16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_sbb16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                    }
+                    break;
+                case 4: //AND
+                    if(opcode==0x81) {INST_NAME("LOCK AND Ew, Iw");} else {INST_NAME("LOCK AND Ew, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_and16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_and16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                    }
+                    break;
+                case 5: //SUB
+                    if(opcode==0x81) {INST_NAME("LOCK SUB Ew, Iw");} else {INST_NAME("LOCK SUB Ew, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_sub16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        TSTx_mask(wback, 1, 0, 0);    // mask=1
+                        B_MARK(cNE);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_sub16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                        B_NEXT_nocond;
+                        MARK;   // unaligned! also, not enough 
+                        LDRH_U12(x1, wback, 0);
+                        LDAXRB(x4, wback);
+                        BFIw(x1, x4, 0, 8); // re-inject
+                        emit_sub16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRB(x3, x1, wback);
+                        CBNZx_MARK(x3);
+                        STRH_U12(x1, wback, 0);    // put the whole value
+                    }
+                    break;
+                case 6: //XOR
+                    if(opcode==0x81) {INST_NAME("LOCK XOR Ew, Iw");} else {INST_NAME("LOCK XOR Ew, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    if(MODREG) {
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        ed = xRAX+(nextop&7)+(rex.b<<3);
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_xor16(dyn, ninst, x6, x5, x3, x4);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?2:1);
+                        if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                        MOV32w(x5, i32);
+                        MARKLOCK;
+                        LDAXRH(x1, wback);
+                        emit_xor16(dyn, ninst, x1, x5, x3, x4);
+                        STLXRH(x3, x1, wback);
+                        CBNZx_MARKLOCK(x3);
+                    }
+                    break;
+                case 7: //CMP
+                    if(opcode==0x81) {INST_NAME("(LOCK) CMP Ew, Iw");} else {INST_NAME("(LOCK) CMP Ew, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETEW(x6, (opcode==0x81)?2:1);
+                    (void)wb1;
+                    // No need to LOCK, this is readonly
+                    if(opcode==0x81) i32 = F16S; else i32 = F8S;
+                    if(i32) {
+                        MOV32w(x5, i32);
+                        UXTHw(x6, ed);
+                        emit_cmp16(dyn, ninst, x6, x5, x3, x4, x6);
+                        BFIx(ed, x6, 0, 16);
+                    } else {
+                        emit_cmp16_0(dyn, ninst, ed, x3, x4);
+                    }
+                    break;
+            }
+            DMB_ISH();
+            break;
+
+            case 0xFF:
+                nextop = F8;
+                switch((nextop>>3)&7)
+                {
+                    case 0: // INC Ew
+                        INST_NAME("LOCK INC Ew");
+                        SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                        DMB_ISH();
+                        if(MODREG) {
+                            ed = xRAX+(nextop&7)+(rex.b<<3);
+                            UXTHw(x6, ed);
+                            emit_inc16(dyn, ninst, x6, x5, x3);
+                            BFIx(ed, x6, 0, 16);
+                        } else {
+                            addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                            MARKLOCK;
+                            LDAXRH(x1, wback);
+                            emit_inc16(dyn, ninst, x1, x3, x4);
+                            STLXRH(x3, x1, wback);
+                            CBNZx_MARKLOCK(x3);
+                        }
+                        DMB_ISH();
+                        break;
+                    case 1: //DEC Ew
+                        INST_NAME("LOCK DEC Ew");
+                        SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                        DMB_ISH();
+                        if(MODREG) {
+                            ed = xRAX+(nextop&7)+(rex.b<<3);
+                            UXTHw(x6, ed);
+                            emit_dec16(dyn, ninst, x6, x5, x3);
+                            BFIx(ed, x6, 0, 16);
+                        } else {
+                            addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                            MARKLOCK;
+                            LDAXRH(x1, wback);
+                            emit_dec16(dyn, ninst, x1, x3, x4);
+                            STLXRH(x3, x1, wback);
+                            CBNZx_MARKLOCK(x3);
+                        }
+                        DMB_ISH();
+                        break;
+                    default:
+                        DEFAULT;
+                }
+                break;
+
+        default:
+            DEFAULT;
+    }
+    return addr;
+}
@@ -0,0 +1,968 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "emu/x64run_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "dynarec_native.h"
+
+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
+#include "dynarec_arm64_helper.h"
+#include "dynarec_arm64_functions.h"
+
+#define GETGX(a, w)                     \
+    gd = ((nextop&0x38)>>3)+(rex.r<<3); \
+    a = sse_get_reg(dyn, ninst, x1, gd, w)
+
+#define GETGM(a)                        \
+    gd = ((nextop&0x38)>>3);            \
+    a = mmx_get_reg(dyn, ninst, x1, x2, x3, gd)
+
+#define GETGm   gd = ((nextop&0x38)>>3)
+
+uintptr_t dynarec64_67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
+{
+    (void)ip; (void)need_epilog;
+
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    uint8_t gd, ed, wback, wb, wb1, wb2, gb1, gb2;
+    int64_t fixedaddress;
+    int8_t  i8;
+    uint8_t u8;
+    int32_t i32;
+    int64_t j64, i64;
+    int cacheupd;
+    int lock;
+    int v0, v1, s0;
+    MAYUSE(i32);
+    MAYUSE(j64);
+    MAYUSE(v0);
+    MAYUSE(v1);
+    MAYUSE(s0);
+    MAYUSE(lock);
+    MAYUSE(cacheupd);
+
+    // REX prefix before the 67 are ignored
+    rex.rex = 0;
+    while(opcode>=0x40 && opcode<=0x4f) {
+        rex.rex = opcode;
+        opcode = F8;
+    }
+    rep = 0;
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+        case 0x00:
+            INST_NAME("ADD Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_add8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x01:
+            INST_NAME("ADD Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_add32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x02:
+            INST_NAME("ADD Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_add8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x03:
+            INST_NAME("ADD Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_add32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x04:
+            INST_NAME("ADD AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_add8c(dyn, ninst, x1, u8, x3, x4);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x05:
+            INST_NAME("ADD EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            emit_add32c(dyn, ninst, rex, xRAX, i64, x3, x4, x5);
+            break;
+
+        case 0x08:
+            INST_NAME("OR Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_or8(dyn, ninst, x1, x2, x4, x2);
+            EBBACK;
+            break;
+        case 0x09:
+            INST_NAME("OR Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_or32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x0A:
+            INST_NAME("OR Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_or8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x0B:
+            INST_NAME("OR Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_or32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x0C:
+            INST_NAME("OR AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_or8c(dyn, ninst, x1, u8, x3, x4);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x0D:
+            INST_NAME("OR EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            emit_or32c(dyn, ninst, rex, xRAX, i64, x3, x4);
+            break;
+
+        case 0x0F:
+            opcode=F8;
+            switch(opcode) {
+
+                case 0x2E:
+                    // no special check...
+                case 0x2F:
+                    if(rep) {
+                        DEFAULT;
+                    } else {
+                        if(opcode==0x2F) {INST_NAME("COMISS Gx, Ex");} else {INST_NAME("UCOMISS Gx, Ex");}
+                        SETFLAGS(X_ALL, SF_SET);
+                        nextop = F8;
+                        GETGX(v0, 0);
+                        if(MODREG) {
+                            s0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 0);
+                        } else {
+                            s0 = fpu_get_scratch(dyn);
+                            addr = geted32(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                            VLDR32_U12(s0, ed, fixedaddress);
+                        }
+                        FCMPS(v0, s0);
+                        FCOMI(x1, x2);
+                    }
+                    break;
+
+                case 0x6F:
+                    INST_NAME("MOVQ Gm, Em");
+                    nextop = F8;
+                    GETGm;
+                    if(MODREG) {
+                        v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7); // no rex.b on MMX
+                        v0 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, gd);
+                        VMOV(v0, v1);
+                    } else {
+                        v0 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, gd);
+                        addr = geted32(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
+                        VLDR64_U12(v0, ed, fixedaddress);
+                    }
+                    break;
+
+                case 0x7F:
+                    INST_NAME("MOVQ Em, Gm");
+                    nextop = F8;
+                    GETGM(v0);
+                    if(MODREG) {
+                        v1 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, nextop&7);
+                        VMOV(v1, v0);
+                    } else {
+                        addr = geted32(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
+                        VSTR64_U12(v0, ed, fixedaddress);
+                    }
+                    break;
+
+                default:
+                    DEFAULT;
+            }
+            break;
+
+        case 0x10:
+            INST_NAME("ADC Eb, Gb");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_adc8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x11:
+            INST_NAME("ADC Ed, Gd");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_adc32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x12:
+            INST_NAME("ADC Gb, Eb");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_adc8(dyn, ninst, x1, x2, x4, x3);
+            GBBACK;
+            break;
+        case 0x13:
+            INST_NAME("ADC Gd, Ed");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_adc32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x14:
+            INST_NAME("ADC AL, Ib");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_adc8c(dyn, ninst, x1, u8, x3, x4, x5);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x15:
+            INST_NAME("ADC EAX, Id");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            MOV64xw(x1, i64);
+            emit_adc32(dyn, ninst, rex, xRAX, x1, x3, x4);
+            break;
+
+        case 0x18:
+            INST_NAME("SBB Eb, Gb");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_sbb8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x19:
+            INST_NAME("SBB Ed, Gd");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_sbb32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x1A:
+            INST_NAME("SBB Gb, Eb");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_sbb8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x1B:
+            INST_NAME("SBB Gd, Ed");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_sbb32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x1C:
+            INST_NAME("SBB AL, Ib");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_sbb8c(dyn, ninst, x1, u8, x3, x4, x5);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x1D:
+            INST_NAME("SBB EAX, Id");
+            READFLAGS(X_CF);
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            MOV64xw(x2, i64);
+            emit_sbb32(dyn, ninst, rex, xRAX, x2, x3, x4);
+            break;
+
+        case 0x20:
+            INST_NAME("AND Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_and8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x21:
+            INST_NAME("AND Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_and32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x22:
+            INST_NAME("AND Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_and8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x23:
+            INST_NAME("AND Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_and32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x24:
+            INST_NAME("AND AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_and8c(dyn, ninst, x1, u8, x3, x4);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x25:
+            INST_NAME("AND EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            emit_and32c(dyn, ninst, rex, xRAX, i64, x3, x4);
+            break;
+
+        case 0x28:
+            INST_NAME("SUB Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_sub8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x29:
+            INST_NAME("SUB Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_sub32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x2A:
+            INST_NAME("SUB Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_sub8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x2B:
+            INST_NAME("SUB Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_sub32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x2C:
+            INST_NAME("SUB AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_sub8c(dyn, ninst, x1, u8, x3, x4, x5);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x2D:
+            INST_NAME("SUB EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            emit_sub32c(dyn, ninst, rex, xRAX, i64, x3, x4, x5);
+            break;
+
+        case 0x30:
+            INST_NAME("XOR Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_xor8(dyn, ninst, x1, x2, x4, x5);
+            EBBACK;
+            break;
+        case 0x31:
+            INST_NAME("XOR Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_xor32(dyn, ninst, rex, ed, gd, x3, x4);
+            WBACK;
+            break;
+        case 0x32:
+            INST_NAME("XOR Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_xor8(dyn, ninst, x1, x2, x3, x4);
+            GBBACK;
+            break;
+        case 0x33:
+            INST_NAME("XOR Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_xor32(dyn, ninst, rex, gd, ed, x3, x4);
+            break;
+        case 0x34:
+            INST_NAME("XOR AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            emit_xor8c(dyn, ninst, x1, u8, x3, x4);
+            BFIx(xRAX, x1, 0, 8);
+            break;
+        case 0x35:
+            INST_NAME("XOR EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            emit_xor32c(dyn, ninst, rex, xRAX, i64, x3, x4);
+            break;
+
+        case 0x38:
+            INST_NAME("CMP Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x1, 0);
+            GETGB(x2);
+            emit_cmp8(dyn, ninst, x1, x2, x3, x4, x5);
+            break;
+        case 0x39:
+            INST_NAME("CMP Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_cmp32(dyn, ninst, rex, ed, gd, x3, x4, x5);
+            break;
+        case 0x3A:
+            INST_NAME("CMP Gb, Eb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETEB32(x2, 0);
+            GETGB(x1);
+            emit_cmp8(dyn, ninst, x1, x2, x3, x4, x5);
+            break;
+        case 0x3B:
+            INST_NAME("CMP Gd, Ed");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            GETED32(0);
+            emit_cmp32(dyn, ninst, rex, gd, ed, x3, x4, x5);
+            break;
+        case 0x3C:
+            INST_NAME("CMP AL, Ib");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            u8 = F8;
+            UXTBw(x1, xRAX);
+            if(u8) {
+                MOV32w(x2, u8);
+                emit_cmp8(dyn, ninst, x1, x2, x3, x4, x5);
+            } else {
+                emit_cmp8_0(dyn, ninst, x1, x3, x4);
+            }
+            break;
+        case 0x3D:
+            INST_NAME("CMP EAX, Id");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            i64 = F32S;
+            if(i64) {
+                MOV64xw(x2, i64);
+                emit_cmp32(dyn, ninst, rex, xRAX, x2, x3, x4, x5);
+            } else
+                emit_cmp32_0(dyn, ninst, rex, xRAX, x3, x4);
+            break;
+
+        case 0x81:
+        case 0x83:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0: //ADD
+                    if(opcode==0x81) {INST_NAME("ADD Ed, Id");} else {INST_NAME("ADD Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    emit_add32c(dyn, ninst, rex, ed, i64, x3, x4, x5);
+                    WBACK;
+                    break;
+                case 1: //OR
+                    if(opcode==0x81) {INST_NAME("OR Ed, Id");} else {INST_NAME("OR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    emit_or32c(dyn, ninst, rex, ed, i64, x3, x4);
+                    WBACK;
+                    break;
+                case 2: //ADC
+                    if(opcode==0x81) {INST_NAME("ADC Ed, Id");} else {INST_NAME("ADC Ed, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    MOV64xw(x5, i64);
+                    emit_adc32(dyn, ninst, rex, ed, x5, x3, x4);
+                    WBACK;
+                    break;
+                case 3: //SBB
+                    if(opcode==0x81) {INST_NAME("SBB Ed, Id");} else {INST_NAME("SBB Ed, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    MOV64xw(x5, i64);
+                    emit_sbb32(dyn, ninst, rex, ed, x5, x3, x4);
+                    WBACK;
+                    break;
+                case 4: //AND
+                    if(opcode==0x81) {INST_NAME("AND Ed, Id");} else {INST_NAME("AND Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    emit_and32c(dyn, ninst, rex, ed, i64, x3, x4);
+                    WBACK;
+                    break;
+                case 5: //SUB
+                    if(opcode==0x81) {INST_NAME("SUB Ed, Id");} else {INST_NAME("SUB Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    emit_sub32c(dyn, ninst, rex, ed, i64, x3, x4, x5);
+                    WBACK;
+                    break;
+                case 6: //XOR
+                    if(opcode==0x81) {INST_NAME("XOR Ed, Id");} else {INST_NAME("XOR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    emit_xor32c(dyn, ninst, rex, ed, i64, x3, x4);
+                    WBACK;
+                    break;
+                case 7: //CMP
+                    if(opcode==0x81) {INST_NAME("CMP Ed, Id");} else {INST_NAME("CMP Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32((opcode==0x81)?4:1);
+                    if(opcode==0x81) i64 = F32S; else i64 = F8S;
+                    if(i64) {
+                        MOV64xw(x2, i64);
+                        emit_cmp32(dyn, ninst, rex, ed, x2, x3, x4, x5);
+                    } else
+                        emit_cmp32_0(dyn, ninst, rex, ed, x3, x4);
+                    break;
+            }
+            break;
+            
+        case 0x89:
+            INST_NAME("MOV Ed, Gd");
+            nextop=F8;
+            GETGD;
+            if(MODREG) {   // reg <= reg
+                MOVxw_REG(xRAX+(nextop&7)+(rex.b<<3), gd);
+            } else {                    // mem <= reg
+                addr = geted32(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 0);
+                STRxw_U12(gd, ed, fixedaddress);
+                if(lock) {DMB_ISH();}
+            }
+            break;
+
+        case 0x8B:
+            INST_NAME("MOV Gd, Ed");
+            nextop=F8;
+            GETGD;
+            if(MODREG) {
+                MOVxw_REG(gd, xRAX+(nextop&7)+(rex.b<<3));
+            } else {
+                addr = geted32(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 0);
+                if(lock) {DMB_ISH();}
+                LDRxw_U12(gd, ed, fixedaddress);
+            }
+            break;
+
+        case 0x8D:
+            INST_NAME("LEA Gd, Ed");
+            nextop=F8;
+            GETGD;
+            if(MODREG) {   // reg <= reg? that's an invalid operation
+                DEFAULT;
+            } else {                    // mem <= reg
+                // should a geted32 be created, to use 32bits regs instead of 64bits?
+                addr = geted32(dyn, addr, ninst, nextop, &ed, gd, &fixedaddress, 0, 0, rex, NULL, 0, 0);
+                if(ed!=gd) {
+                    MOVw_REG(gd, ed);
+                }
+            }
+            break;
+
+        case 0xC1:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("ROL Ed, Ib");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                    GETED32(1);
+                    u8 = (F8)&(rex.w?0x3f:0x1f);
+                    emit_rol32c(dyn, ninst, rex, ed, u8, x3, x4);
+                    if(u8) { WBACK; }
+                    break;
+                case 1:
+                    INST_NAME("ROR Ed, Ib");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                    GETED32(1);
+                    u8 = (F8)&(rex.w?0x3f:0x1f);
+                    emit_ror32c(dyn, ninst, rex, ed, u8, x3, x4);
+                    if(u8) { WBACK; }
+                    break;
+                case 2:
+                    INST_NAME("RCL Ed, Ib");
+                    MESSAGE(LOG_DUMP, "Need Optimization\n");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    GETED32W(x4, x1, 1);
+                    u8 = F8;
+                    MOV32w(x2, u8);
+                    CALL_(rex.w?((void*)rcl64):((void*)rcl32), ed, x4);
+                    WBACK;
+                    break;
+                case 3:
+                    INST_NAME("RCR Ed, Ib");
+                    MESSAGE(LOG_DUMP, "Need Optimization\n");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    GETED32W(x4, x1, 1);
+                    u8 = F8;
+                    MOV32w(x2, u8);
+                    CALL_(rex.w?((void*)rcr64):((void*)rcr32), ed, x4);
+                    WBACK;
+                    break;
+                case 4:
+                case 6:
+                    INST_NAME("SHL Ed, Ib");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    GETED32(1);
+                    u8 = (F8)&(rex.w?0x3f:0x1f);
+                    emit_shl32c(dyn, ninst, rex, ed, u8, x3, x4);
+                    WBACK;
+                    break;
+                case 5:
+                    INST_NAME("SHR Ed, Ib");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    GETED32(1);
+                    u8 = (F8)&(rex.w?0x3f:0x1f);
+                    emit_shr32c(dyn, ninst, rex, ed, u8, x3, x4);
+                    if(u8) {
+                        WBACK;
+                    }
+                    break;
+                case 7:
+                    INST_NAME("SAR Ed, Ib");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    GETED32(1);
+                    u8 = (F8)&(rex.w?0x3f:0x1f);
+                    emit_sar32c(dyn, ninst, rex, ed, u8, x3, x4);
+                    if(u8) {
+                        WBACK;
+                    }
+                    break;
+            }
+            break;
+
+        case 0xC7:
+            INST_NAME("MOV Ed, Id");
+            nextop=F8;
+            if(MODREG) {   // reg <= i32
+                i64 = F32S;
+                ed = xRAX+(nextop&7)+(rex.b<<3);
+                MOV64xw(ed, i64);
+            } else {                    // mem <= i32
+                addr = geted32(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, &lock, 0, 4);
+                i64 = F32S;
+                MOV64xw(x3, i64);
+                STRxw_U12(x3, ed, fixedaddress);
+                if(lock) {DMB_ISH();}
+            }
+            break;
+
+
+        #define GO(NO, YES)                                             \
+            BARRIER(BARRIER_MAYBE);                                     \
+            JUMP(addr+i8, 1);                                           \
+            if(dyn->insts[ninst].x64.jmp_insts==-1 ||                   \
+                CHECK_CACHE()) {                                        \
+                /* out of the block */                                  \
+                i32 = dyn->insts[ninst].epilog-(dyn->native_size);      \
+                Bcond(NO, i32);                                         \
+                if(dyn->insts[ninst].x64.jmp_insts==-1) {               \
+                    if(!(dyn->insts[ninst].x64.barrier&BARRIER_FLOAT))  \
+                        fpu_purgecache(dyn, ninst, 1, x1, x2, x3);      \
+                    jump_to_next(dyn, addr+i8, 0, ninst);               \
+                } else {                                                \
+                    CacheTransform(dyn, ninst, cacheupd, x1, x2, x3);   \
+                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);\
+                    B(i32);                                             \
+                }                                                       \
+            } else {    \
+                /* inside the block */  \
+                i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->native_size);    \
+                Bcond(YES, i32);    \
+            }
+        case 0xE0:
+            INST_NAME("LOOPNZ (32bits)");
+            READFLAGS(X_ZF);
+            i8 = F8S;
+            MOVw_REG(x1, xRCX);
+            SUBSw_U12(x1, x1, 1);
+            BFIx(xRCX, x1, 0, 32);
+            B_NEXT(cEQ);    // ECX is 0, no LOOP
+            TSTw_mask(xFlags, 0b011010, 0); //mask=0x40
+            GO(cNE, cEQ);
+            break;
+        case 0xE1:
+            INST_NAME("LOOPZ (32bits)");
+            READFLAGS(X_ZF);
+            i8 = F8S;
+            MOVw_REG(x1, xRCX);
+            SUBSw_U12(x1, x1, 1);
+            BFIx(xRCX, x1, 0, 32);
+            B_NEXT(cEQ);    // ECX is 0, no LOOP
+            TSTw_mask(xFlags, 0b011010, 0); //mask=0x40
+            GO(cEQ, cNE);
+            break;
+        case 0xE2:
+            INST_NAME("LOOP (32bits)");
+            i8 = F8S;
+            MOVw_REG(x1, xRCX);
+            SUBSw_U12(x1, x1, 1);
+            BFIx(xRCX, x1, 0, 32);
+            GO(cEQ, cNE);
+            break;
+        case 0xE3:
+            INST_NAME("JECXZ");
+            i8 = F8S;
+            MOVw_REG(x1, xRCX);
+            TSTw_REG(x1, x1);
+            GO(cNE, cEQ);
+            break;
+        #undef GO
+
+        case 0xE8:
+            return dynarec64_00(dyn, addr-1, ip, ninst, rex, rep, ok, need_epilog); // addr-1, to "put back" opcode)
+
+        case 0xF7:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0:
+                case 1:
+                    INST_NAME("TEST Ed, Id");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32H(x1, 4);
+                    i64 = F32S;
+                    MOV64xw(x2, i64);
+                    emit_test32(dyn, ninst, rex, ed, x2, x3, x4);
+                    break;
+                case 2:
+                    INST_NAME("NOT Ed");
+                    GETED32(0);
+                    MVNxw_REG(ed, ed);
+                    WBACK;
+                    break;
+                case 3:
+                    INST_NAME("NEG Ed");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    GETED32(0);
+                    emit_neg32(dyn, ninst, rex, ed, x3, x4);
+                    WBACK;
+                    break;
+                case 4:
+                    INST_NAME("MUL EAX, Ed");
+                    SETFLAGS(X_ALL, SF_PENDING);
+                    UFLAG_DF(x2, rex.w?d_mul64:d_mul32);
+                    GETED32(0);
+                    if(rex.w) {
+                        if(ed==xRDX) gd=x3; else gd=xRDX;
+                        UMULH(gd, xRAX, ed);
+                        MULx(xRAX, xRAX, ed);
+                        if(gd!=xRDX) {MOVx_REG(xRDX, gd);}
+                    } else {
+                        UMULL(xRDX, xRAX, ed);  //64 <- 32x32
+                        MOVw_REG(xRAX, xRDX);
+                        LSRx(xRDX, xRDX, 32);
+                    }
+                    UFLAG_RES(xRAX);
+                    UFLAG_OP1(xRDX);
+                    break;
+                case 5:
+                    INST_NAME("IMUL EAX, Ed");
+                    SETFLAGS(X_ALL, SF_PENDING);
+                    UFLAG_DF(x2, rex.w?d_imul64:d_imul32);
+                    GETED32(0);
+                    if(rex.w) {
+                        if(ed==xRDX) gd=x3; else gd=xRDX;
+                        SMULH(gd, xRAX, ed);
+                        MULx(xRAX, xRAX, ed);
+                        if(gd!=xRDX) {MOVx_REG(xRDX, gd);}
+                    } else {
+                        SMULL(xRDX, xRAX, ed);  //64 <- 32x32
+                        MOVw_REG(xRAX, xRDX);
+                        LSRx(xRDX, xRDX, 32);
+                    }
+                    UFLAG_RES(xRAX);
+                    UFLAG_OP1(xRDX);
+                    break;
+                case 6:
+                    INST_NAME("DIV Ed");
+                    SETFLAGS(X_ALL, SF_SET);
+                    if(!rex.w) {
+                        SET_DFNONE(x2);
+                        GETED32(0);
+                        MOVw_REG(x3, xRAX);
+                        ORRx_REG_LSL(x3, x3, xRDX, 32);
+                        if(MODREG) {
+                            MOVw_REG(x4, ed);
+                            ed = x4;
+                        }
+                        UDIVx(x2, x3, ed);
+                        MSUBx(x4, x2, ed, xRAX);
+                        MOVw_REG(xRAX, x2);
+                        MOVw_REG(xRDX, x4);
+                    } else {
+                        if(ninst && dyn->insts 
+                           && dyn->insts[ninst-1].x64.addr 
+                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr)==0x31 
+                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr+1)==0xD2) {
+                            SET_DFNONE(x2);
+                            GETED32(0);
+                            UDIVx(x2, xRAX, ed);
+                            MSUBx(xRDX, x2, ed, xRAX);
+                            MOVx_REG(xRAX, x2);
+                        } else {
+                            GETED32H(x1, 0);  // get edd changed addr, so cannot be called 2 times for same op...
+                            CBZxw_MARK(xRDX);
+                            if(ed!=x1) {MOVx_REG(x1, ed);}
+                            CALL(div64, -1);
+                            B_NEXT_nocond;
+                            MARK;
+                            UDIVx(x2, xRAX, ed);
+                            MSUBx(xRDX, x2, ed, xRAX);
+                            MOVx_REG(xRAX, x2);
+                            SET_DFNONE(x2);
+                        }
+                    }
+                    break;
+                case 7:
+                    INST_NAME("IDIV Ed");
+                    SETFLAGS(X_ALL, SF_SET);
+                    if(!rex.w) {
+                        SET_DFNONE(x2)
+                        GETSED32w(0);
+                        MOVw_REG(x3, xRAX);
+                        ORRx_REG_LSL(x3, x3, xRDX, 32);
+                        SDIVx(x2, x3, wb);
+                        MSUBx(x4, x2, wb, x3);
+                        MOVw_REG(xRAX, x2);
+                        MOVw_REG(xRDX, x4);
+                    } else {
+                        if(ninst && dyn->insts
+                           &&  dyn->insts[ninst-1].x64.addr 
+                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr)==0x48
+                           && *(uint8_t*)(dyn->insts[ninst-1].x64.addr+1)==0x99) {
+                            SET_DFNONE(x2)
+                            GETED32(0);
+                            SDIVx(x2, xRAX, ed);
+                            MSUBx(xRDX, x2, ed, xRAX);
+                            MOVx_REG(xRAX, x2);
+                        } else {
+                            GETED32H(x1, 0);  // get edd changed addr, so cannot be called 2 times for same op...
+                            CBZxw_MARK(xRDX);
+                            MVNx_REG(x2, xRDX);
+                            CBZxw_MARK(x2);
+                            if(ed!=x1) {MOVx_REG(x1, ed);}
+                            CALL((void*)idiv64, -1);
+                            B_NEXT_nocond;
+                            MARK;
+                            SDIVx(x2, xRAX, ed);
+                            MSUBx(xRDX, x2, ed, xRAX);
+                            MOVx_REG(xRAX, x2);
+                            SET_DFNONE(x2)
+                        }
+                    }
+                    break;
+            }
+            break;
+
+        default:
+            DEFAULT;
+    }
+    return addr;
+}
@@ -0,0 +1,296 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "emu/x64run_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "emu/x87emu_private.h"
+#include "dynarec_native.h"
+
+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
+#include "dynarec_arm64_helper.h"
+#include "dynarec_arm64_functions.h"
+
+
+uintptr_t dynarec64_D8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
+{
+    (void)ip; (void)rep; (void)need_epilog;
+
+    uint8_t nextop = F8;
+    uint8_t ed;
+    int64_t fixedaddress;
+    int v1, v2;
+    int s0;
+
+    MAYUSE(s0);
+    MAYUSE(v2);
+    MAYUSE(v1);
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FADD ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FADDS(v1, v1, v2);
+            } else {
+                FADDD(v1, v1, v2);
+            }
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FMUL ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FMULS(v1, v1, v2);
+            } else {
+                FMULD(v1, v1, v2);
+            }
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCOM ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
+            FCOM(x1, x2, x3);
+            break;
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FCOMP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
+            FCOM(x1, x2, x3);
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FSUB ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v1, v2);
+            } else {
+                FSUBD(v1, v1, v2);
+            }
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FSUBR ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v2, v1);
+            } else {
+                FSUBD(v1, v2, v1);
+            }
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FDIV ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v1, v2);
+            } else {
+                FDIVD(v1, v1, v2);
+            }
+            break;
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            INST_NAME("FDIVR ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v2, v1);
+            } else {
+                FDIVD(v1, v2, v1);
+            }
+            break;
+      
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FADD ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FADDS(v1, v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FADDD(v1, v1, s0);
+                    }
+                    break;
+                case 1:
+                    INST_NAME("FMUL ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FMULS(v1, v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FMULD(v1, v1, s0);
+                    }
+                    break;
+                case 2:
+                    INST_NAME("FCOM ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FCMPS(v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FCMPD(v1, s0);
+                    }
+                    FCOM(x1, x2, x3);
+                    break;
+                case 3:
+                    INST_NAME("FCOMP ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FCMPS(v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FCMPD(v1, s0);
+                    }
+                    FCOM(x1, x2, x3);
+                    x87_do_pop(dyn, ninst, x3);
+                    break;
+                case 4:
+                    INST_NAME("FSUB ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FSUBS(v1, v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FSUBD(v1, v1, s0);
+                    }
+                    break;
+                case 5:
+                    INST_NAME("FSUBR ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FSUBS(v1, s0, v1);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FSUBD(v1, s0, v1);
+                    }
+                    break;
+                case 6:
+                    INST_NAME("FDIV ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FDIVS(v1, v1, s0);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FDIVD(v1, v1, s0);
+                    }
+                    break;
+                case 7:
+                    INST_NAME("FDIVR ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+                    s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
+                    VLDR32_U12(s0, ed, fixedaddress);
+                    if(ST_IS_F(0)) {
+                        FDIVS(v1, s0, v1);
+                    } else {
+                        FCVT_D_S(s0, s0);
+                        FDIVD(v1, s0, v1);
+                    }
+                    break;
+                default:
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
 #include "emu/x87emu_private.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -31,6 +31,7 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    uint8_t nextop = F8;
    uint8_t ed;
    uint8_t wback, wb1;
+    uint8_t u8;
    int64_t fixedaddress;
    int v1, v2;
    int s0;
@@ -50,12 +51,18 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC6:
        case 0xC7:
            INST_NAME("FLD STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            v2 = x87_do_push(dyn, ninst);
-            FMOVD(v2, v1);
+            v2 = x87_do_push(dyn, ninst, x1, X87_ST(nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, (nextop&7)+1, X87_COMBINE(0, (nextop&7)+1));
+            if(ST_IS_F(0)) {
+                FMOVS(v2, v1);
+            } else {
+                FMOVD(v2, v1);
+            }
            break;

        case 0xC8:
+            INST_NAME("FXCH ST0");
+            break;
        case 0xC9:
        case 0xCA:
        case 0xCB:
@@ -65,32 +72,61 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xCF:
            INST_NAME("FXCH STx");
            // swap the cache value, not the double value itself :p
-            i1 = x87_get_cache(dyn, ninst, x1, x2, nextop&7);
-            i2 = x87_get_cache(dyn, ninst, x1, x2, 0);
-            i3 = dyn->x87cache[i1];
-            dyn->x87cache[i1] = dyn->x87cache[i2];
-            dyn->x87cache[i2] = i3;
+            x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_ST(nextop&7));
+            x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            x87_swapreg(dyn, ninst, x1, x2, 0, nextop&7);
+            // should set C1 to 0
            break;

        case 0xD0:
            INST_NAME("FNOP");
            break;

+        case 0xD8:
+            INST_NAME("FSTPNCE ST0, ST0");
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FSTPNCE ST0, STx");
+            // copy the cache value for st0 to stx
+            x87_get_st_empty(dyn, ninst, x1, x2, nextop&7, X87_ST(nextop&7));
+            x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            x87_swapreg(dyn, ninst, x1, x2, 0, nextop&7);
+            x87_do_pop(dyn, ninst, x3);
+            break;
        case 0xE0:
            INST_NAME("FCHS");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            FNEGD(v1, v1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            if(ST_IS_F(0)) {
+                FNEGS(v1, v1);
+            } else {
+                FNEGD(v1, v1);
+            }
            break;
        case 0xE1:
            INST_NAME("FABS");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            FABSD(v1, v1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            if(ST_IS_F(0)) {
+                FABSS(v1, v1);
+            } else {
+                FABSD(v1, v1);
+            }
            break;

        case 0xE4:
            INST_NAME("FTST");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            FCMPD_0(v1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            if(ST_IS_F(0)) {
+                FCMPS_0(v1);
+            } else {
+                FCMPD_0(v1);
+            }
            FCOM(x1, x2, x3);   // same flags...
            break;
        case 0xE5:
@@ -102,64 +138,44 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin

        case 0xE8:
            INST_NAME("FLD1");
-            v1 = x87_do_push(dyn, ninst);
-            FTABLE64(v1, 1.0);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            if(ST_IS_F(0)) {
+                FMOVS_8(v1, 0b01110000);
+            } else {
+                FMOVD_8(v1, 0b01110000);
+            }
            break;
        case 0xE9:
            INST_NAME("FLDL2T");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
            FTABLE64(v1, L2T);
            break;
        case 0xEA:     
            INST_NAME("FLDL2E");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
            FTABLE64(v1, L2E);
            break;
        case 0xEB:
            INST_NAME("FLDPI");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
            FTABLE64(v1, PI);
            break;
        case 0xEC:
            INST_NAME("FLDLG2");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
            FTABLE64(v1, LG2);
            break;
        case 0xED:
            INST_NAME("FLDLN2");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
            FTABLE64(v1, LN2);
            break;
        case 0xEE:
            INST_NAME("FLDZ");
-            v1 = x87_do_push(dyn, ninst);
-            FTABLE64(v1, 0.0);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            VEOR(v1, v1, v1);
            break;

-        case 0xFA:
-            INST_NAME("FSQRT");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            FSQRTD(v1, v1);
-            break;
-
-        case 0xFC:
-            INST_NAME("FRNDINT");
-            MESSAGE(LOG_DUMP, "Need Optimization\n");
-            // use C helper for now, nothing staightforward is available
-            x87_forget(dyn, ninst, x1, x2, 0);
-            CALL(arm_frndint, -1);
-            /*
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            VCMP_F64_0(v1);
-            VMRS_APSR();
-            B_NEXT(cVS);    // Unordered, skip
-            B_NEXT(cEQ);    // Zero, skip
-            u8 = x87_setround(dyn, ninst, x1, x2, x3);
-            VCVT_S32_F64(x1, v1);   // limit to 32bits....
-            VCVT_F64_S32(v1, x1);
-            x87_restoreround(dyn, ninst, u8);
-            */
-            break;
        case 0xF0:
            INST_NAME("F2XM1");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
@@ -172,15 +188,19 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fyl2x, -1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF2:
-            INST_NAME("FTAN");
+            INST_NAME("FPTAN");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
            x87_forget(dyn, ninst, x1, x2, 0);
            CALL(arm_ftan, -1);
-            v1 = x87_do_push(dyn, ninst);
-            FTABLE64(v1, 1.0);
+            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            if(ST_IS_F(0)) {
+                FMOVS_8(v1, 0b01110000);
+            } else {
+                FMOVD_8(v1, 0b01110000);
+            }
            break;
        case 0xF3:
            INST_NAME("FPATAN");
@@ -188,7 +208,7 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fpatan, -1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF4:
            INST_NAME("FXTRACT");
@@ -206,7 +226,7 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            break;
        case 0xF6:
            INST_NAME("FDECSTP");
-            fpu_purgecache(dyn, ninst, x1, x2, x3);
+            fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
            LDRw_U12(x2, xEmu, offsetof(x64emu_t, top));
            SUBw_U12(x2, x2, 1);
            ANDw_mask(x2, x2, 0, 2);    //mask=7
@@ -214,7 +234,7 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            break;
        case 0xF7:
            INST_NAME("FINCSTP");
-            fpu_purgecache(dyn, ninst, x1, x2, x3);
+            fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
            LDRw_U12(x2, xEmu, offsetof(x64emu_t, top));
            ADDw_U12(x2, x2, 1);
            ANDw_mask(x2, x2, 0, 2);    //mask=7
@@ -233,7 +253,16 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fyl2xp1, -1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xFA:
+            INST_NAME("FSQRT");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            if(ST_IS_F(0)) {
+                FSQRTS(v1, v1);
+            } else {
+                FSQRTD(v1, v1);
+            }
            break;
        case 0xFB:
            INST_NAME("FSINCOS");
@@ -242,6 +271,20 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fsincos, -1);
            break;
+        case 0xFC:
+            INST_NAME("FRNDINT");
+            #if 0
+            MESSAGE(LOG_DUMP, "Need Optimization\n");
+            // use C helper for now, nothing staightforward is available
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_frndint, -1);
+            #else
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            u8 = x87_setround(dyn, ninst, x1, x2, x3);
+            FRINTID(v1, v1);
+            x87_restoreround(dyn, ninst, u8);
+            #endif
+            break;
        case 0xFD:
            INST_NAME("FSCALE");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
@@ -268,14 +311,6 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xD5:
        case 0xD6:
        case 0xD7:
-        case 0xD8:
-        case 0xD9:
-        case 0xDA:
-        case 0xDB:
-        case 0xDC:
-        case 0xDD:
-        case 0xDE:
-        case 0xDF:
        case 0xE2:
        case 0xE3:
        case 0xE6:
@@ -288,34 +323,47 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FLD ST0, float[ED]");
-                    v1 = x87_do_push(dyn, ninst);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    v1 = x87_do_push(dyn, ninst, x1, box64_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
+                    if(ST_IS_F(0))
+                        s0 = v1;
+                    else
+                        s0 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(v1, s0);
+                    if(!ST_IS_F(0)) {
+                        FCVT_D_S(v1, s0);
+                    }
                    break;
                case 2:
                    INST_NAME("FST float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    FCVT_S_D(s0, v1);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    if(ST_IS_F(0))
+                        s0 = v1;
+                    else {
+                        s0 = fpu_get_scratch(dyn);
+                        FCVT_S_D(s0, v1);
+                    }
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    VSTR32_U12(s0, ed, fixedaddress);
                    break;
                case 3:
                    INST_NAME("FSTP float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    FCVT_S_D(s0, v1);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    if(ST_IS_F(0))
+                        s0 = v1;
+                    else {
+                        s0 = fpu_get_scratch(dyn);
+                        FCVT_S_D(s0, v1);
+                    }
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    VSTR32_U12(s0, ed, fixedaddress);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 4:
                    INST_NAME("FLDENV Ed");
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
-                    fpu_purgecache(dyn, ninst, x1, x2, x3); // maybe only x87, not SSE?
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    fpu_purgecache(dyn, ninst, 0, x1, x2, x3); // maybe only x87, not SSE?
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {
                        MOVx_REG(x1, ed);
                    }
@@ -326,14 +374,12 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    INST_NAME("FLDCW Ew");
                    GETEW(x1, 0);
                    STRH_U12(x1, xEmu, offsetof(x64emu_t, cw));    // hopefully cw is not too far for an imm8
-                    UBFXw(x1, x1, 10, 2);    // extract round
-                    STRw_U12(x1, xEmu, offsetof(x64emu_t, round));
                    break;
                case 6:
                    INST_NAME("FNSTENV Ed");
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
-                    fpu_purgecache(dyn, ninst, x1, x2, x3); // maybe only x87, not SSE?
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    fpu_purgecache(dyn, ninst, 0, x1, x2, x3); // maybe only x87, not SSE?
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {
                        MOVx_REG(x1, ed);
                    }
@@ -342,7 +388,7 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 7:
                    INST_NAME("FNSTCW Ew");
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
                    ed = x1;
                    wb1 = 1;
                    LDRH_U12(x1, xEmu, offsetof(x64emu_t, cw));
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
 #include "emu/x87emu_private.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -53,10 +53,14 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC7:
            INST_NAME("FCMOVNB ST0, STx");
            READFLAGS(X_CF);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
            TSTw_mask(xFlags, 0, 0);    //mask=1<<F_CF
-            FCSELD(v1, v2, v1, cEQ);    // F_CF==0
+            if(ST_IS_F(0)) {
+                FCSELS(v1, v2, v1, cEQ);
+            } else {
+                FCSELD(v1, v2, v1, cEQ);    // F_CF==0
+            }
            break;
        case 0xC8:
        case 0xC9:
@@ -68,10 +72,14 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xCF:
            INST_NAME("FCMOVNE ST0, STx");
            READFLAGS(X_ZF);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
            TSTw_mask(xFlags, 0b011010, 0); //mask=1<<F_ZF
-            FCSELD(v1, v2, v1, cEQ);        // F_ZF==0
+            if(ST_IS_F(0)) {
+                FCSELS(v1, v2, v1, cEQ);
+            } else {
+                FCSELD(v1, v2, v1, cEQ);        // F_ZF==0
+            }
            break;
        case 0xD0:
        case 0xD1:
@@ -83,11 +91,15 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xD7:
            INST_NAME("FCMOVNBE ST0, STx");
            READFLAGS(X_CF|X_ZF);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
            MOV32w(x1, (1<<F_CF)|(1<<F_ZF));
            TSTw_REG(xFlags, x1);
-            FCSELD(v1, v2, v1, cEQ);   // F_CF==0 & F_ZF==0
+            if(ST_IS_F(0)) {
+                FCSELS(v1, v2, v1, cEQ);
+            } else {
+                FCSELD(v1, v2, v1, cEQ);   // F_CF==0 & F_ZF==0
+            }
            break;
        case 0xD8:
        case 0xD9:
@@ -99,10 +111,14 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xDF:
            INST_NAME("FCMOVNU ST0, STx");
            READFLAGS(X_PF);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
            TSTw_mask(xFlags, 0b011110, 0); //mask=1<<F_PF
-            FCSELD(v1, v2, v1, cEQ);        // F_PF==0
+            if(ST_IS_F(0)) {
+                FCSELS(v1, v2, v1, cEQ);
+            } else {
+                FCSELD(v1, v2, v1, cEQ);        // F_PF==0
+            }
            break;
        case 0xE1:
            INST_NAME("FDISI8087_NOP"); // so.. NOP?
@@ -118,7 +134,7 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xE3:
            INST_NAME("FNINIT");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
-            x87_purgecache(dyn, ninst, x1, x2, x3);
+            x87_purgecache(dyn, ninst, 0, x1, x2, x3);
            CALL(reset_fpu, -1);
            break;
        case 0xE8:
@@ -131,9 +147,13 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xEF:
            INST_NAME("FUCOMI ST0, STx");
            SETFLAGS(X_ALL, SF_SET);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOMI(x1, x2);
            break;
        case 0xF0:  
@@ -146,9 +166,13 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xF7:
            INST_NAME("FCOMI ST0, STx");
            SETFLAGS(X_ALL, SF_SET);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOMI(x1, x2);
            break;

@@ -164,21 +188,21 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FILD ST0, Ed");
-                    v1 = x87_do_push(dyn, ninst);
+                    v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                    VLDR32_U12(s0, ed, fixedaddress);
                    SXTL_32(v1, s0);
                    SCVTFDD(v1, v1);
                    break;
                case 1:
                    INST_NAME("FISTTP Ed, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        wback = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                        ed = x1;
                    }
                    s0 = fpu_get_scratch(dyn);
@@ -200,17 +224,17 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    STRw_U12(x5, wback, fixedaddress);
                    MARK3;
                    #endif
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 2:
                    INST_NAME("FIST Ed, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    u8 = x87_setround(dyn, ninst, x1, x2, x4); // x1 have the modified RPSCR reg
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        wback = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                        ed = x1;
                    }
                    s0 = fpu_get_scratch(dyn);
@@ -236,13 +260,13 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 3:
                    INST_NAME("FISTP Ed, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    u8 = x87_setround(dyn, ninst, x1, x2, x4); // x1 have the modified RPSCR reg
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        wback = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                        ed = x1;
                    }
                    s0 = fpu_get_scratch(dyn);
@@ -265,11 +289,11 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MARK3;
                    #endif
                    x87_restoreround(dyn, ninst, u8);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 5:
                    INST_NAME("FLD tbyte");
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(PK(0)==0xDB && ((PK(1)>>3)&7)==7) {
                        // the FLD is immediatly followed by an FSTP
                        LDRx_U12(x5, ed, 0);
@@ -278,7 +302,7 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MESSAGE(LOG_DUMP, "\tHack: FSTP tbyte\n");
                        nextop = F8;    //0xDB
                        nextop = F8;    //modrm
-                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                        STRx_U12(x5, ed, 0);
                        STRH_U12(x6, ed, 8);
                    } else {
@@ -292,12 +316,12 @@ uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                case 7:
                    INST_NAME("FSTP tbyte");
                    x87_forget(dyn, ninst, x1, x3, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {
                        MOVx_REG(x1, ed);
                    }
                    CALL(arm_fstp, -1);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                default:
                    DEFAULT;
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
 #include "emu/x87emu_private.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -46,9 +46,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC6:
        case 0xC7:
            INST_NAME("FADD STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FADDD(v1, v1, v2);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FADDS(v1, v1, v2);
+            } else {
+                FADDD(v1, v1, v2);
+            }
            break;
        case 0xC8:
        case 0xC9:
@@ -59,9 +63,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xCE:
        case 0xCF:
            INST_NAME("FMUL STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FMULD(v1, v1, v2);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FMULS(v1, v1, v2);
+            } else {
+                FMULD(v1, v1, v2);
+            }
            break;
        case 0xD0:
        case 0xD1:
@@ -72,9 +80,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xD6:
        case 0xD7:
            INST_NAME("FCOM ST0, STx"); //yep
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOM(x1, x2, x3);
            break;
        case 0xD8:
@@ -86,11 +98,15 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xDE:
        case 0xDF:
            INST_NAME("FCOMP ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOM(x1, x2, x3);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xE0:
        case 0xE1:
@@ -101,9 +117,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xE6:
        case 0xE7:
            INST_NAME("FSUBR STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FSUBD(v1, v2, v1);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v2, v1);
+            } else {
+                FSUBD(v1, v2, v1);
+            }
            break;
        case 0xE8:
        case 0xE9:
@@ -114,9 +134,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xEE:
        case 0xEF:
            INST_NAME("FSUB STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FSUBD(v1, v1, v2);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v1, v2);
+            } else {
+                FSUBD(v1, v1, v2);
+            }
            break;
        case 0xF0:
        case 0xF1:
@@ -127,9 +151,13 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xF6:
        case 0xF7:
            INST_NAME("FDIVR STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FDIVD(v1, v2, v1);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v2, v1);
+            } else {
+                FDIVD(v1, v2, v1);
+            }
            break;       
        case 0xF8:
        case 0xF9:
@@ -140,76 +168,80 @@ uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xFE:
        case 0xFF:
            INST_NAME("FDIV STx, ST0");
-            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FDIVD(v1, v1, v2);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v1, v2);
+            } else {
+                FDIVD(v1, v1, v2);
+            }
            break;
        default:
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FADD ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FADDD(v1, v1, v2);
                    break;
                case 1:
                    INST_NAME("FMUL ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FMULD(v1, v1, v2);
                    break;
                case 2:
                    INST_NAME("FCOM ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FCMPD(v1, v2);
                    FCOM(x1, x2, x3);
                    break;
                case 3:
                    INST_NAME("FCOMP ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FCMPD(v1, v2);
                    FCOM(x1, x2, x3);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 4:
                    INST_NAME("FSUB ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FSUBD(v1, v1, v2);
                    break;
                case 5:
                    INST_NAME("FSUBR ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FSUBD(v1, v2, v1);
                    break;
                case 6:
                    INST_NAME("FDIV ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FDIVD(v1, v1, v2);
                    break;
                case 7:
                    INST_NAME("FDIVR ST0, double[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    v2 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 3, rex, NULL, 0, 0);
                    VLDR64_U12(v2, wback, fixedaddress);
                    FDIVD(v1, v2, v1);
                    break;
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
 #include "emu/x87emu_private.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -49,8 +49,8 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC7:
            INST_NAME("FFREE STx");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
-            x87_purgecache(dyn, ninst, x1, x2, x3);
-            MOV32w(x1, nextop-0xC0);
+            x87_purgecache(dyn, ninst, 0, x1, x2, x3);
+            MOV32w(x1, nextop&7);
            CALL(fpu_do_free, -1);
            break;
        case 0xD0:
@@ -62,13 +62,17 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xD6:
        case 0xD7:
            INST_NAME("FST ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FMOVD(v2, v1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FMOVS(v2, v1);
+            } else {
+                FMOVD(v2, v1);
+            }
            break;
        case 0xD8:
            INST_NAME("FSTP ST0, ST0");
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xD9:
        case 0xDA:
@@ -78,10 +82,11 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xDE:
        case 0xDF:
            INST_NAME("FSTP ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FMOVD(v2, v1);
-            x87_do_pop(dyn, ninst);
+            // copy the cache value for st0 to stx
+            x87_get_st_empty(dyn, ninst, x1, x2, nextop&7, X87_ST(nextop&7));
+            x87_get_st(dyn, ninst, x1, x2, 0, X87_ST0);
+            x87_swapreg(dyn, ninst, x1, x2, 0, nextop&7);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xE0:
@@ -93,9 +98,13 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xE6:
        case 0xE7:
            INST_NAME("FUCOM ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOM(x1, x2, x3);
            break;
        case 0xE8:
@@ -107,11 +116,15 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xEE:
        case 0xEF:
            INST_NAME("FUCOMP ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOM(x1, x2, x3);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xC8:
@@ -145,56 +158,70 @@ uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FLD double");
-                    v1 = x87_do_push(dyn, ninst);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    v1 = x87_do_push(dyn, ninst, x3, NEON_CACHE_ST_D);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    VLDR64_U12(v1, ed, fixedaddress);
                    break;
                case 1:
                    INST_NAME("FISTTP i64, ST0");
-                    v1 = x87_do_push(dyn, ninst);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    s0 = fpu_get_scratch(dyn);
+                    #if 0
+                    // those are ARM 8.5 opcode!
                    FRINT64ZD(s0, v1);
+                    #else
+                    FRINTRRD(s0, v1, 3);    // not ideal, might averflow and not set 0x8000000000000000 correctly
+                    #endif
                    FCVTZSxD(x2, s0);
                    STRx_U12(x2, ed, fixedaddress);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 2:
                    INST_NAME("FST double");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    VSTR64_U12(v1, ed, fixedaddress);
                    break;
                case 3:
                    INST_NAME("FSTP double");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    VSTR64_U12(v1, ed, fixedaddress);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 4: 
                    INST_NAME("FRSTOR m108byte");
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
-                    fpu_purgecache(dyn, ninst, x1, x2, x3);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {MOVx_REG(x1, ed);}
                    CALL(arm_frstor, -1);
                    break;
                case 6: 
                    INST_NAME("FSAVE m108byte");
                    MESSAGE(LOG_DUMP, "Need Optimization\n");
-                    fpu_purgecache(dyn, ninst, x1, x2, x3);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {MOVx_REG(x1, ed);}
                    CALL(arm_fsave, -1);
                    break;
                case 7:
                    INST_NAME("FNSTSW m2byte");
-                    fpu_purgecache(dyn, ninst, x1, x2, x3);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x4, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
-                    LDRw_U12(x1, xEmu, offsetof(x64emu_t, top));
+                    fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x4, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    LDRw_U12(x2, xEmu, offsetof(x64emu_t, top));
                    LDRH_U12(x3, xEmu, offsetof(x64emu_t, sw));
-                    BFIw(x3, x1, 11, 3); // inject TOP at bit 11 (3 bits)
+                    if(dyn->n.x87stack) {
+                        // update top
+                        if(dyn->n.x87stack>0) {
+                            SUBw_U12(x2, x2, dyn->n.x87stack);
+                        } else {
+                            ADDw_U12(x2, x2, -dyn->n.x87stack);
+                        }
+                        ANDw_mask(x2, x2, 0, 2);
+                    }
+                    BFIw(x3, x2, 11, 3); // inject TOP at bit 11 (3 bits)
                    STRH_U12(x3, ed, fixedaddress);   // store whole sw flags
                    break;
                default:
@@ -0,0 +1,284 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "emu/x64run_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "emu/x87emu_private.h"
+#include "dynarec_native.h"
+
+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
+#include "dynarec_arm64_helper.h"
+#include "dynarec_arm64_functions.h"
+
+
+uintptr_t dynarec64_DE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
+{
+    (void)ip; (void)rep; (void)need_epilog;
+
+    uint8_t nextop = F8;
+    uint8_t wback;
+    int64_t fixedaddress;
+    int v1, v2;
+
+    MAYUSE(v2);
+    MAYUSE(v1);
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FADDP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FADDS(v1, v1, v2);
+            } else {
+                FADDD(v1, v1, v2);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FMULP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FMULS(v1, v1, v2);
+            } else {
+                FMULD(v1, v1, v2);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCOMP ST0, STx"); //yep
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
+            FCOM(x1, x2, x3);
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xD9:
+            INST_NAME("FCOMPP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
+            FCOM(x1, x2, x3);
+            x87_do_pop(dyn, ninst, x3);
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FSUBRP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v2, v1);
+            } else {
+                FSUBD(v1, v2, v1);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FSUBP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FSUBS(v1, v1, v2);
+            } else {
+                FSUBD(v1, v1, v2);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FDIVRP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v2, v1);
+            } else {
+                FDIVD(v1, v2, v1);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;       
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            INST_NAME("FDIVP STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0)) {
+                FDIVS(v1, v1, v2);
+            } else {
+                FDIVD(v1, v1, v2);
+            }
+            x87_do_pop(dyn, ninst, x3);
+            break;
+        case 0xD8:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            return 0;
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FIADD ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FADDD(v1, v1, v2);
+                    break;
+                case 1:
+                    INST_NAME("FIMUL ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FMULD(v1, v1, v2);
+                    break;
+                case 2:
+                    INST_NAME("FICOM ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FCMPD(v1, v2);
+                    FCOM(x1, x2, x3);
+                    break;
+                case 3:
+                    INST_NAME("FICOMP ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FCMPD(v1, v2);
+                    FCOM(x1, x2, x3);
+                    x87_do_pop(dyn, ninst, x3);
+                    break;
+                case 4:
+                    INST_NAME("FISUB ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FSUBD(v1, v1, v2);
+                    break;
+                case 5:
+                    INST_NAME("FISUBR ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FSUBD(v1, v2, v1);
+                    break;
+                case 6:
+                    INST_NAME("FIDIV ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FDIVD(v1, v1, v2);
+                    break;
+                case 7:
+                    INST_NAME("FIDIVR ST0, word[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
+                    v2 = fpu_get_scratch(dyn);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
+                    VLDR16_U12(v2, wback, fixedaddress);
+                    SXTL_16(v2, v2);
+                    SXTL_32(v2, v2);
+                    SCVTFDD(v2, v2);
+                    FDIVD(v1, v2, v1);
+                    break;
+            }
+    }
+    return addr;
+}
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
 #include "emu/x87emu_private.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -51,7 +51,7 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xC7:
            INST_NAME("FFREEP STx");
            // not handling Tag...
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xE0:
@@ -71,11 +71,16 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xEF:
            INST_NAME("FUCOMIP ST0, STx");
            SETFLAGS(X_ALL, SF_SET);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0))
+            {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOMI(x1, x2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF0:
        case 0xF1:
@@ -87,11 +92,16 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
        case 0xF7:
            INST_NAME("FCOMIP ST0, STx");
            SETFLAGS(X_ALL, SF_SET);
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0, X87_COMBINE(0, nextop&7));
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7, X87_COMBINE(0, nextop&7));
+            if(ST_IS_F(0))
+            {
+                FCMPS(v1, v2);
+            } else {
+                FCMPD(v1, v2);
+            }
            FCOMI(x1, x2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xC8:
@@ -140,19 +150,23 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FILD ST0, Ew");
-                    v1 = x87_do_push(dyn, ninst);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                    v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
                    LDRSHw_U12(x1, wback, fixedaddress);
-                    SCVTFDw(v1, x1);
+                    if(ST_IS_F(0)) {
+                        SCVTFSw(v1, x1);
+                    } else {
+                        SCVTFDw(v1, x1);
+                    }
                    break;
                case 1:
                    INST_NAME("FISTTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
                    ed = x1;
                    s0 = fpu_get_scratch(dyn);
                    #if 0
-                    // this version needs ARM v8.5, //TODO: add detection of this extensio to use it
+                    // this version needs ARM v8.5, //TODO: add detection of this extension to use it
                    FRINT32ZD(s0, v1);
                    // no saturation instruction on Arm, so using NEON
                    VFCVTZSd(s0, s0);
@@ -163,8 +177,12 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MRS_fpsr(x5);
                    BFCw(x5, FPSR_IOC, 1);   // reset IOC bit
                    MSR_fpsr(x5);
-                    VFCVTZSd(s0, v1);
-                    SQXTN_S_D(s0, s0);
+                    if(ST_IS_F(0)) {
+                        VFCVTZSs(s0, v1);
+                    } else {
+                        VFCVTZSd(s0, v1);
+                        SQXTN_S_D(s0, s0);
+                    }
                    SQXTN_H_S(s0, s0);
                    VSTR16_U12(s0, wback, fixedaddress);
                    MRS_fpsr(x5);   // get back FPSR to check the IOC bit
@@ -173,13 +191,13 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    STRH_U12(x5, wback, fixedaddress);
                    MARK3;
                    #endif
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 2:
                    INST_NAME("FIST Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
                    u8 = x87_setround(dyn, ninst, x1, x2, x4);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
                    ed = x1;
                    s0 = fpu_get_scratch(dyn);
                    #if 0
@@ -193,9 +211,14 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MRS_fpsr(x5);
                    BFCw(x5, FPSR_IOC, 1);   // reset IOC bit
                    MSR_fpsr(x5);
-                    FRINTXD(s0, v1);
-                    VFCVTZSd(s0, s0);
-                    SQXTN_S_D(s0, s0);
+                    if(ST_IS_F(0)) {
+                        FRINTXS(s0, v1);
+                        VFCVTZSs(s0, s0);
+                    } else {
+                        FRINTXD(s0, v1);
+                        VFCVTZSd(s0, s0);
+                        SQXTN_S_D(s0, s0);
+                    }
                    SQXTN_H_S(s0, s0);
                    VSTR16_U12(s0, wback, fixedaddress);
                    MRS_fpsr(x5);   // get back FPSR to check the IOC bit
@@ -208,9 +231,9 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    break;
                case 3:
                    INST_NAME("FISTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
                    u8 = x87_setround(dyn, ninst, x1, x2, x4);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<1, 1, rex, NULL, 0, 0);
                    ed = x1;
                    s0 = fpu_get_scratch(dyn);
                    #if 0
@@ -224,9 +247,14 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MRS_fpsr(x5);
                    BFCw(x5, FPSR_IOC, 1);   // reset IOC bit
                    MSR_fpsr(x5);
-                    FRINTXD(s0, v1);
-                    VFCVTZSd(s0, s0);
-                    SQXTN_S_D(s0, s0);
+                    if(ST_IS_F(0)) {
+                        FRINTXS(s0, v1);
+                        VFCVTZSs(s0, s0);
+                    } else {
+                        FRINTXD(s0, v1);
+                        VFCVTZSd(s0, s0);
+                        SQXTN_S_D(s0, s0);
+                    }
                    SQXTN_H_S(s0, s0);
                    VSTR16_U12(s0, wback, fixedaddress);
                    MRS_fpsr(x5);   // get back FPSR to check the IOC bit
@@ -235,36 +263,36 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    STRH_U12(x5, wback, fixedaddress);
                    MARK3;
                    #endif
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    x87_restoreround(dyn, ninst, u8);
                    break;
                case 4:
                    INST_NAME("FBLD ST0, tbytes");
                    x87_do_push_empty(dyn, ninst, x1);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {MOVx_REG(x1, ed);}
                    CALL(fpu_fbld, -1);
                    break;
                case 5:
                    INST_NAME("FILD ST0, i64");
-                    v1 = x87_do_push(dyn, ninst);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_D);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    LDRx_U12(x1, wback, fixedaddress);
                    SCVTFDx(v1, x1);
                    break;
                case 6:
                    INST_NAME("FBSTP tbytes, ST0");
                    x87_forget(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, rex, NULL, 0, 0);
                    if(ed!=x1) {MOVx_REG(x1, ed);}
                    CALL(fpu_fbst, -1);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 7:
                    INST_NAME("FISTP i64, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_D);
                    u8 = x87_setround(dyn, ninst, x1, x2, x4);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                    ed = x1;
                    s0 = fpu_get_scratch(dyn);
                    #if 0
@@ -285,7 +313,7 @@ uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MARK3;
                    #endif
                    x87_restoreround(dyn, ninst, u8);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                default:
                    DEFAULT;
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"
 #include "../tools/bridge_private.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

@@ -28,8 +28,6 @@ void emit_or32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3,
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_or64:d_or32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -60,14 +58,11 @@ void emit_or32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3,
 void emit_or32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, int s3, int s4)
 {
    IFX(X_PEND) {
-        MOV64xw(s3, c);
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_or64:d_or32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
    }
-    IFX(X_PEND) {} else {MOV64xw(s3, c);}
+    MOV64xw(s3, c);
    ORRxw_REG(s1, s1, s3);
    IFX(X_PEND) {
        STRxw_U12(s1, xEmu, offsetof(x64emu_t, res));
@@ -95,8 +90,6 @@ void emit_xor32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_xor64:d_xor32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -127,14 +120,11 @@ void emit_xor32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 void emit_xor32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, int s3, int s4)
 {
    IFX(X_PEND) {
-        MOV64xw(s3, c);
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_xor64:d_xor32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
    }
-    IFX(X_PEND) {} else {MOV64xw(s3, c);}
+    MOV64xw(s3, c);
    EORxw_REG(s1, s1, s3);
    IFX(X_PEND) {
        STRxw_U12(s1, xEmu, offsetof(x64emu_t, res));
@@ -162,8 +152,6 @@ void emit_and32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_and64:d_and32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -197,14 +185,11 @@ void emit_and32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 void emit_and32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, int s3, int s4)
 {
    IFX(X_PEND) {
-        MOV64xw(s3, c);
-        STRxw_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRxw_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, rex.w?d_and64:d_and32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
    }
-    IFX(X_PEND) {} else {MOV64xw(s3, c);}
+    MOV64xw(s3, c);
    IFX(X_ALL) {
        ANDSxw_REG(s1, s1, s3);
    } else {
@@ -235,8 +220,6 @@ void emit_or8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_or8);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -268,8 +251,6 @@ void emit_or8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
 {
    MOV32w(s3, c&0xff);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, d_or8);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -301,8 +282,6 @@ void emit_xor8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_xor8);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -334,8 +313,6 @@ void emit_xor8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
 {
    MOV32w(s3, c&0xff);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, d_xor8);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -367,8 +344,6 @@ void emit_and8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_and8);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -403,8 +378,6 @@ void emit_and8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
 {
    MOV32w(s3, c&0xff);
    IFX(X_PEND) {
-        STRB_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRB_U12(s3, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s4, d_and8);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
@@ -440,8 +413,6 @@ void emit_or16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRH_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRH_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_or16);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -519,8 +490,6 @@ void emit_xor16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRH_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRH_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_xor16);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -598,8 +567,6 @@ void emit_and16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
 {
    MAYUSE(s2);
    IFX(X_PEND) {
-        STRH_U12(s1, xEmu, offsetof(x64emu_t, op1));
-        STRH_U12(s2, xEmu, offsetof(x64emu_t, op2));
        SET_DF(s3, d_and16);
    } else IFX(X_ALL) {
        SET_DFNONE(s3);
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"
 #include "../tools/bridge_private.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

@@ -77,7 +77,7 @@ void emit_add32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 void emit_add32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, int s3, int s4, int s5)
 {
    MAYUSE(s5);
-    if(s1==xRSP && (!dyn->insts || dyn->insts[ninst].x64.need_flags==X_PEND))
+    if(s1==xRSP && (!dyn->insts || dyn->insts[ninst].x64.gen_flags==X_PEND))
    {
        // special case when doing math on ESP and only PEND is needed: ignoring it!
        if(c>=0 && c<0x1000) {
@@ -201,7 +201,7 @@ void emit_sub32(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int s3
 void emit_sub32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int64_t c, int s3, int s4, int s5)
 {
    MAYUSE(s5);
-    if(s1==xRSP && (!dyn->insts || dyn->insts[ninst].x64.need_flags==X_PEND))
+    if(s1==xRSP && (!dyn->insts || dyn->insts[ninst].x64.gen_flags==X_PEND))
    {
        // special case when doing math on RSP and only PEND is needed: ignoring it!
        if(c>=0 && c<0x1000) {
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"
 #include "../tools/bridge_private.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

@@ -209,7 +209,7 @@ void emit_shr32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int32_t c, in
        if(c>1) {
            LSRxw(s3, s1, c-1);
        }
-        BFIw(xFlags, s3, 0, 1);
+        BFIw(xFlags, (c>1)?s3:s1, 0, 1);
    }
    LSRxw(s1, s1, c);
    IFX(X_PEND) {
@@ -257,7 +257,7 @@ void emit_sar32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int32_t c, in
        if(c>1) {
            ASRxw(s3, s1, c-1);
        }
-        BFIw(xFlags, s3, 0, 1);
+        BFIw(xFlags, (c>1)?s3:s1, 0, 1);
    }
    ASRxw(s1, s1, c);
    IFX(X_PEND) {
@@ -366,7 +366,7 @@ void emit_shrd32c(dynarec_arm_t* dyn, int ninst, rex_t rex, int s1, int s2, int3
        if(c>1) {
            LSRxw(s3, s1, c-1);
        }
-        BFIw(xFlags, s3, 0, 1);
+        BFIw(xFlags, (c>1)?s3:s1, 0, 1);
    }
    LSRxw(s3, s1, c);
    ORRxw_REG_LSL(s1, s3, s2, (rex.w?64:32)-c);
@@ -15,11 +15,11 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"
 #include "../tools/bridge_private.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

@@ -15,10 +15,10 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_helper.h"
 #include "dynarec_arm64_functions.h"

@@ -30,10 +30,12 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
    uint8_t opcode = F8;
    uint8_t nextop;
    uint8_t gd, ed, u8;
-    uint8_t wback, wb1, wb2, gb1, gb2;
+    uint8_t wback, wb1, wb2, eb1, eb2, gb1, gb2;
    int32_t i32;
    int64_t i64, j64;
    int64_t fixedaddress;
+    MAYUSE(eb1);
+    MAYUSE(eb2);
    MAYUSE(gb1);
    MAYUSE(gb2);
    MAYUSE(wb1);
@@ -56,6 +58,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            INST_NAME("LOCK ADD Eb, Gb");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
+            DMB_ISH();
            GETGB(x2);
            if((nextop&0xC0)==0xC0) {
                if(rex.rex) {
@@ -70,59 +73,137 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                emit_add8(dyn, ninst, x1, x2, x4, x3);
                BFIx(wback, x1, wb2*8, 8);
            } else {                   
-                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                MARKLOCK;
                LDAXRB(x1, wback);
                emit_add8(dyn, ninst, x1, x2, x4, x3);
                STLXRB(x4, x1, wback);
                CBNZx_MARKLOCK(x4);
            }
+            DMB_ISH();
            break;
        case 0x01:
            INST_NAME("LOCK ADD Ed, Gd");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
+            DMB_ISH();
            if((nextop&0xC0)==0xC0) {
                ed = xRAX+(nextop&7)+(rex.b<<3);
                emit_add32(dyn, ninst, rex, ed, gd, x3, x4);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                MARKLOCK;
                LDAXRxw(x1, wback);
                emit_add32(dyn, ninst, rex, x1, gd, x3, x4);
                STLXRxw(x3, x1, wback);
                CBNZx_MARKLOCK(x3);
            }
+            DMB_ISH();
            break;

+        case 0x08:
+            INST_NAME("LOCK OR Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            DMB_ISH();
+            GETGB(x2);
+            if((nextop&0xC0)==0xC0) {
+                if(rex.rex) {
+                    wback = xRAX + (nextop&7) + (rex.b<<3);
+                    wb2 = 0;
+                } else {
+                    wback = (nextop&7);    
+                    wb2 = (wback>>2);      
+                    wback = xRAX+(wback&3);
+                }
+                UBFXw(x1, wback, wb2*8, 8);   
+                emit_or8(dyn, ninst, x1, x2, x4, x3);
+                BFIx(wback, x1, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                MARKLOCK;
+                LDAXRB(x1, wback);
+                emit_or8(dyn, ninst, x1, x2, x4, x3);
+                STLXRB(x4, x1, wback);
+                CBNZx_MARKLOCK(x4);
+            }
+            DMB_ISH();
+            break;
        case 0x09:
            INST_NAME("LOCK OR Ed, Gd");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
+            DMB_ISH();
            if(MODREG) {
                ed = xRAX+(nextop&7)+(rex.b<<3);
                emit_or32(dyn, ninst, rex, ed, gd, x3, x4);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                MARKLOCK;
                LDAXRxw(x1, wback);
                emit_or32(dyn, ninst, rex, x1, gd, x3, x4);
                STLXRxw(x3, x1, wback);
                CBNZx_MARKLOCK(x3);
            }
+            DMB_ISH();
            break;

        case 0x0F:
            nextop = F8;
            switch(nextop) {

+                case 0xB0:
+                    INST_NAME("LOCK CMPXCHG Eb, Gb");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);
+                    nextop = F8;
+                    GETGB(x1);
+                    UBFXx(x6, xRAX, 0, 8);
+                    DMB_ISH();
+                    if(MODREG) {
+                        if(rex.rex) {
+                            wback = xRAX+(nextop&7)+(rex.b<<3);
+                            wb2 = 0;
+                        } else { 
+                            wback = (nextop&7);
+                            wb2 = (wback>>2)*8;
+                            wback = xRAX+(wback&3);
+                        } 
+                        UBFXx(x2, wback, wb2, 8);
+                        wb1 = 0;
+                        ed = x2;
+                        UFLAG_IF {emit_cmp8(dyn, ninst, x6, ed, x3, x4, x5);}
+                        CMPSxw_REG(x6, x2);
+                        B_MARK2(cNE);
+                        BFIx(wback, x2, wb2, 8);
+                        MOVxw_REG(ed, gd);
+                        MARK2;
+                        BFIx(xRAX, x2, 0, 8);
+                        B_NEXT_nocond;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                        // Aligned version
+                        MARKLOCK;
+                        LDAXRB(x2, wback);
+                        CMPSxw_REG(x6, x2);
+                        B_MARK(cNE);
+                        // EAX == Ed
+                        STLXRB(x4, gd, wback);
+                        CBNZx_MARKLOCK(x4);
+                        // done
+                        MARK;
+                        UFLAG_IF {emit_cmp32(dyn, ninst, rex, x6, x2, x3, x4, x5);}
+                        BFIx(xRAX, x2, 0, 8);    // upper par of RAX will be erase on 32bits, no mater what
+                    }
+                    DMB_ISH();
+                    break;
                case 0xB1:
                    INST_NAME("LOCK CMPXCHG Ed, Gd");
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    nextop = F8;
                    GETGD;
+                    DMB_ISH();
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        wback = 0;
@@ -135,7 +216,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOVxw_REG(xRAX, x1);
                        B_NEXT_nocond;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK3(cNE);
                        // Aligned version
@@ -163,6 +244,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        UFLAG_IF {emit_cmp32(dyn, ninst, rex, xRAX, x1, x3, x4, x5);}
                        MOVxw_REG(xRAX, x1);    // upper par of RAX will be erase on 32bits, no mater what
                    }
+                    DMB_ISH();
                    break;

                case 0xC1:
@@ -170,6 +252,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    nextop = F8;
                    GETGD;
+                    DMB_ISH();
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        MOVxw_REG(x1, ed);
@@ -177,7 +260,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOVxw_REG(gd, x1);
                        emit_add32(dyn, ninst, rex, ed, gd, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK(cNE);    // unaligned
                        MARKLOCK;
@@ -201,13 +284,15 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        }
                        MOVxw_REG(gd, x1);
                    }
+                    DMB_ISH();
                    break;

                case 0xC7:
                    INST_NAME("LOCK CMPXCHG8B Gq, Eq");
                    SETFLAGS(X_ZF, SF_SUBSET);
                    nextop = F8;
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 0, 0, rex, 0, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                    DMB_ISH();
                    MARKLOCK;
                    LDAXPxw(x2, x3, wback);
                    CMPSxw_REG(xRAX, x2);
@@ -223,6 +308,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    MOVxw_REG(xRDX, x3);
                    MOV32w(x1, 0);
                    MARK3;
+                    DMB_ISH();
                    BFIw(xFlags, x1, F_ZF, 1);
                    break;

@@ -231,231 +317,52 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            }
            break;
                    
+        case 0x21:
+            INST_NAME("LOCK AND Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            nextop = F8;
+            GETGD;
+            DMB_ISH();
+            if(MODREG) {
+                ed = xRAX+(nextop&7)+(rex.b<<3);
+                emit_and32(dyn, ninst, rex, ed, gd, x3, x4);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                MARKLOCK;
+                LDAXRxw(x1, wback);
+                emit_and32(dyn, ninst, rex, x1, gd, x3, x4);
+                STLXRxw(x3, x1, wback);
+                CBNZx_MARKLOCK(x3);
+            }
+            DMB_ISH();
+            break;
+            
        case 0x29:
            INST_NAME("LOCK SUB Ed, Gd");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
+            DMB_ISH();
            if(MODREG) {
                ed = xRAX+(nextop&7)+(rex.b<<3);
                emit_sub32(dyn, ninst, rex, ed, gd, x3, x4);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                MARKLOCK;
                LDAXRxw(x1, wback);
                emit_sub32(dyn, ninst, rex, x1, gd, x3, x4);
                STLXRxw(x3, x1, wback);
                CBNZx_MARKLOCK(x3);
            }
+            DMB_ISH();
            break;

        case 0x66:
-            opcode = F8;
-            switch(opcode) {
-                case 0x81:
-                case 0x83:
-                    nextop = F8;
-                    switch((nextop>>3)&7) {
-                        case 0: //ADD
-                            if(opcode==0x81) {
-                                INST_NAME("LOCK ADD Ew, Iw");
-                            } else {
-                                INST_NAME("LOCK ADD Ew, Iw");
-                            }
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_add16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F32S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                TSTx_mask(wback, 1, 0, 0);    // mask=1
-                                B_MARK(cNE);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_add16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                                B_NEXT_nocond;
-                                MARK;   // unaligned! also, not enough 
-                                LDRH_U12(x1, wback, 0);
-                                LDAXRB(x4, wback);
-                                BFIw(x1, x4, 0, 8); // re-inject
-                                emit_add16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRB(x3, x1, wback);
-                                CBNZx_MARK(x3);
-                                STRH_U12(x1, wback, 0);    // put the whole value
-                            }
-                            break;
-                        case 1: //OR
-                            if(opcode==0x81) {INST_NAME("LOCK OR Ew, Iw");} else {INST_NAME("LOCK OR Ew, Iw");}
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_or16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_or16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                            }
-                            break;
-                        case 2: //ADC
-                            if(opcode==0x81) {INST_NAME("LOCK ADC Ew, Iw");} else {INST_NAME("LOCK ADC Ew, Ib");}
-                            READFLAGS(X_CF);
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_adc16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_adc16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                            }
-                            break;
-                        case 3: //SBB
-                            if(opcode==0x81) {INST_NAME("LOCK SBB Ew, Iw");} else {INST_NAME("LOCK SBB Ew, Ib");}
-                            READFLAGS(X_CF);
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_sbb16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_sbb16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                            }
-                            break;
-                        case 4: //AND
-                            if(opcode==0x81) {INST_NAME("LOCK AND Ew, Iw");} else {INST_NAME("LOCK AND Ew, Ib");}
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_and16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_and16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                            }
-                            break;
-                        case 5: //SUB
-                            if(opcode==0x81) {INST_NAME("LOCK SUB Ew, Iw");} else {INST_NAME("LOCK SUB Ew, Ib");}
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_sub16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                TSTx_mask(wback, 1, 0, 0);    // mask=1
-                                B_MARK(cNE);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_sub16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                                B_NEXT_nocond;
-                                MARK;   // unaligned! also, not enough 
-                                LDRH_U12(x1, wback, 0);
-                                LDAXRB(x4, wback);
-                                BFIw(x1, x4, 0, 8); // re-inject
-                                emit_sub16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRB(x3, x1, wback);
-                                CBNZx_MARK(x3);
-                                STRH_U12(x1, wback, 0);    // put the whole value
-                            }
-                            break;
-                        case 6: //XOR
-                            if(opcode==0x81) {INST_NAME("LOCK XOR Ew, Iw");} else {INST_NAME("LOCK XOR Ew, Ib");}
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            if(MODREG) {
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                ed = xRAX+(nextop&7)+(rex.b<<3);
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_xor16(dyn, ninst, x6, x5, x3, x4);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?2:1);
-                                if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                                MOV32w(x5, i32);
-                                MARKLOCK;
-                                LDAXRH(x1, wback);
-                                emit_xor16(dyn, ninst, x1, x5, x3, x4);
-                                STLXRH(x3, x1, wback);
-                                CBNZx_MARKLOCK(x3);
-                            }
-                            break;
-                        case 7: //CMP
-                            if(opcode==0x81) {INST_NAME("(LOCK) CMP Ew, Iw");} else {INST_NAME("(LOCK) CMP Ew, Ib");}
-                            SETFLAGS(X_ALL, SF_SET_PENDING);
-                            GETEW(x6, (opcode==0x81)?2:1);
-                            (void)wb1;
-                            // No need to LOCK, this is readonly
-                            if(opcode==0x81) i32 = F16S; else i32 = F8S;
-                            if(i32) {
-                                MOV32w(x5, i32);
-                                UXTHw(x6, ed);
-                                emit_cmp16(dyn, ninst, x6, x5, x3, x4, x6);
-                                BFIx(ed, x6, 0, 16);
-                            } else {
-                                emit_cmp16_0(dyn, ninst, ed, x3, x4);
-                            }
-                            break;
-                    }
-                    break;
-                default:
-                    DEFAULT;
-            }
-            break;
+            return dynarec64_66F0(dyn, addr, ip, ninst, rex, rep, ok, need_epilog);

        case 0x80:
            nextop = F8;
+            DMB_ISH();
            switch((nextop>>3)&7) {
                case 0: //ADD
                    INST_NAME("ADD Eb, Ib");
@@ -466,7 +373,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_add8c(dyn, ninst, x1, u8, x2, x4);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -486,7 +393,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_or8c(dyn, ninst, x1, u8, x2, x4);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -507,7 +414,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_adc8c(dyn, ninst, x1, u8, x2, x4, x5);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -528,7 +435,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_sbb8c(dyn, ninst, x1, u8, x2, x4, x5);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -548,7 +455,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_and8c(dyn, ninst, x1, u8, x2, x4);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -568,7 +475,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_sub8c(dyn, ninst, x1, u8, x2, x4, x5);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -588,7 +495,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        emit_xor8c(dyn, ninst, x1, u8, x2, x4);
                        wb1 = 0;
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, 0, 1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x5, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 1);
                        u8 = F8;
                        wb1 = 1;
                        MARKLOCK;
@@ -614,10 +521,12 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                default:
                    DEFAULT;
            }
+            DMB_ISH();
            break;
        case 0x81:
        case 0x83:
            nextop = F8;
+            DMB_ISH();
            switch((nextop>>3)&7) {
                case 0: //ADD
                    if(opcode==0x81) {
@@ -632,22 +541,22 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_add32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
-                        MOV64xw(x5, i64);
-                        TSTx_mask(wback, 1, 0, 0);    // mask=1
+                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK(cNE);
                        MARKLOCK;
                        LDAXRxw(x1, wback);
-                        emit_add32(dyn, ninst, rex, x1, x5, x3, x4);
+                        emit_add32c(dyn, ninst, rex, x1, i64, x3, x4, x5);
                        STLXRxw(x3, x1, wback);
                        CBNZx_MARKLOCK(x3);
+                        DMB_ISH();
                        B_NEXT_nocond;
                        MARK;   // unaligned! also, not enough 
                        LDRxw_U12(x1, wback, 0);
                        LDAXRB(x4, wback);
                        BFIxw(x1, x4, 0, 8); // re-inject
-                        emit_add32(dyn, ninst, rex, x1, x5, x3, x4);
+                        emit_add32c(dyn, ninst, rex, x1, i64, x3, x4, x5);
                        STLXRB(x3, x1, wback);
                        CBNZx_MARK(x3);
                        STRxw_U12(x1, wback, 0);    // put the whole value
@@ -662,7 +571,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_or32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
                        MOV64xw(x5, i64);
                        MARKLOCK;
@@ -682,7 +591,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_adc32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
                        MOV64xw(x5, i64);
                        MARKLOCK;
@@ -702,7 +611,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_sbb32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
                        MOV64xw(x5, i64);
                        MARKLOCK;
@@ -721,7 +630,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_and32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
                        MOV64xw(x5, i64);
                        MARKLOCK;
@@ -740,22 +649,22 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_sub32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
-                        MOV64xw(x5, i64);
-                        TSTx_mask(wback, 1, 0, 0);    // mask=1
+                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK(cNE);
                        MARKLOCK;
                        LDAXRxw(x1, wback);
-                        emit_sub32(dyn, ninst, rex, x1, x5, x3, x4);
+                        emit_sub32c(dyn, ninst, rex, x1, i64, x3, x4, x5);
                        STLXRxw(x3, x1, wback);
                        CBNZx_MARKLOCK(x3);
+                        DMB_ISH();
                        B_NEXT_nocond;
                        MARK;   // unaligned! also, not enough 
                        LDRxw_U12(x1, wback, 0);
                        LDAXRB(x4, wback);
                        BFIxw(x1, x4, 0, 8); // re-inject
-                        emit_sub32(dyn, ninst, rex, x1, x5, x3, x4);
+                        emit_sub32c(dyn, ninst, rex, x1, i64, x3, x4, x5);
                        STLXRB(x3, x1, wback);
                        CBNZx_MARK(x3);
                        STRxw_U12(x1, wback, 0);    // put the whole value
@@ -770,7 +679,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        MOV64xw(x5, i64);
                        emit_xor32(dyn, ninst, rex, ed, x5, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, (opcode==0x81)?4:1);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, (opcode==0x81)?4:1);
                        if(opcode==0x81) i64 = F32S; else i64 = F8S;
                        MOV64xw(x5, i64);
                        MARKLOCK;
@@ -794,8 +703,42 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                    }
                    break;
            }
+            DMB_ISH();
            break;
            
+        case 0x86:
+            INST_NAME("LOCK XCHG Eb, Gb");
+            // Do the swap
+            nextop = F8;
+            if(MODREG) {
+                GETGB(x4);
+                if(rex.rex) {
+                    ed = xRAX+(nextop&7)+(rex.b<<3);
+                    eb1 = ed;
+                    eb2 = 0;
+                } else {
+                    ed = (nextop&7);
+                    eb1 = xRAX+(ed&3);
+                    eb2 = ((ed&4)>>2);
+                }
+                UBFXw(x1, eb1, eb2*8, 8);
+                // do the swap 14 -> ed, 1 -> gd
+                BFIx(gb1, x1, gb2*8, 8);
+                BFIx(eb1, x4, eb2*8, 8);
+            } else {
+                DMB_ISH();
+                GETGB(x4);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
+                MARKLOCK;
+                // do the swap with exclusive locking
+                LDAXRB(x1, ed);
+                // do the swap 14 -> strb(ed), 1 -> gd
+                STLXRB(x3, x4, ed);
+                CBNZx_MARKLOCK(x3);
+                DMB_ISH();
+                BFIx(gb1, x1, gb2*8, 8);
+            }
+            break;
        case 0x87:
            INST_NAME("LOCK XCHG Ed, Gd");
            nextop = F8;
@@ -807,7 +750,8 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                MOVxw_REG(ed, x1);
            } else {
                GETGD;
-                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                DMB_ISH();
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                TSTx_mask(ed, 1, 0, 1+rex.w);    // mask=3 or 7
                B_MARK(cNE);
                MARKLOCK;
@@ -819,6 +763,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                LDRxw_U12(x1, ed, 0);
                STRxw_U12(gd, ed, 0);
                MARK2;
+                DMB_ISH();
                MOVxw_REG(gd, x1);
            }
            break;
@@ -829,12 +774,13 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
            {
                case 0: // INC Ed
                    INST_NAME("LOCK INC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                    DMB_ISH();
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        emit_inc32(dyn, ninst, rex, ed, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK(cNE);    // unaligned
                        MARKLOCK;
@@ -852,15 +798,17 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        CBNZw_MARK(x3);
                        STRxw_U12(x1, wback, 0);
                    }
+                    DMB_ISH();
                    break;
                case 1: //DEC Ed
                    INST_NAME("LOCK DEC Ed");
-                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
+                    DMB_ISH();
                    if(MODREG) {
                        ed = xRAX+(nextop&7)+(rex.b<<3);
                        emit_dec32(dyn, ninst, rex, ed, x3, x4);
                    } else {
-                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, 0);
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, LOCK_LOCK, 0, 0);
                        TSTx_mask(wback, 1, 0, 1+rex.w);    // mask=3 or 7
                        B_MARK(cNE);    // unaligned
                        MARKLOCK;
@@ -878,6 +826,7 @@ uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
                        CBNZw_MARK(x3);
                        STRxw_U12(x1, wback, 0);
                    }
+                    DMB_ISH();
                    break;
                default:
                    DEFAULT;
@@ -15,31 +15,35 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

 // Get Ex as a double, not a quad (warning, x2 get used)
-#define GETEX(a, D)                                                                                 \
-    if(MODREG) {                                                                                    \
-        a = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));                                     \
-    } else {                                                                                        \
-        a = fpu_get_scratch(dyn);                                                                   \
-        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, D);     \
-        VLDR64_U12(a, ed, fixedaddress);                                                            \
+#define GETEX(a, w, D)                                                                                  \
+    if(MODREG) {                                                                                        \
+        a = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), w);                                      \
+    } else {                                                                                            \
+        a = fpu_get_scratch(dyn);                                                                       \
+        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, D);   \
+        VLDR64_U12(a, ed, fixedaddress);                                                                \
    }

 #define GETG        gd = ((nextop&0x38)>>3)+(rex.r<<3)

-#define GETGX(a)    gd = ((nextop&0x38)>>3)+(rex.r<<3); \
-                    a = sse_get_reg(dyn, ninst, x1, gd)
+#define GETGX(a, w) gd = ((nextop&0x38)>>3)+(rex.r<<3); \
+                    a = sse_get_reg(dyn, ninst, x1, gd, w)

 #define GETGX_empty(a)  gd = ((nextop&0x38)>>3)+(rex.r<<3); \
                        a = sse_get_reg_empty(dyn, ninst, x1, gd)

+#define GETGM(a)                        \
+    gd = ((nextop&0x38)>>3);            \
+    a = mmx_get_reg(dyn, ninst, x1, x2, x3, gd)
+
 uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int* ok, int* need_epilog)
 {
    (void)ip; (void)need_epilog;
@@ -49,7 +53,7 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
    uint8_t gd, ed;
    uint8_t wback;
    uint8_t u8;
-    uint64_t u64;
+    uint64_t u64, j64;
    int v0, v1;
    int q0;
    int d0, d1;
@@ -73,12 +77,12 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            GETG;
            if(MODREG) {
                ed = (nextop&7)+ (rex.b<<3);
-                v0 = sse_get_reg(dyn, ninst, x1, gd);
-                d0 = sse_get_reg(dyn, ninst, x1, ed);
+                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
+                d0 = sse_get_reg(dyn, ninst, x1, ed, 0);
                VMOVeD(v0, 0, d0, 0);
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                VLDR64_U12(v0, ed, fixedaddress); // upper part reseted
            }
            break;
@@ -86,13 +90,13 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("MOVSD Ex, Gx");
            nextop = F8;
            GETG;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
            if(MODREG) {
                ed = (nextop&7)+ (rex.b<<3);
-                d0 = sse_get_reg(dyn, ninst, x1, ed);
+                d0 = sse_get_reg(dyn, ninst, x1, ed, 1);
                VMOVeD(d0, 0, v0, 0);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                VSTR64_U12(v0, ed, fixedaddress);
            }
            break;
@@ -101,12 +105,12 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            nextop = F8;
            GETG;
            if(MODREG) {
-                d0 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                d0 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVeD(v0, 0, d0, 0);
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                VLDR64_U12(v0, ed, fixedaddress);
            }
            VMOVeD(v0, 1, v0, 0);
@@ -115,7 +119,7 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x2A:
            INST_NAME("CVTSI2SD Gx, Ed");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            GETED(0);
            d1 = fpu_get_scratch(dyn);
            if(rex.w) {
@@ -130,23 +134,23 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("CVTTSD2SI Gd, Ex");
            nextop = F8;
            GETGD;
-            GETEX(q0, 0);
+            GETEX(q0, 0, 0);
            FCVTZSxwD(gd, q0);
            break;
        case 0x2D:
            INST_NAME("CVTSD2SI Gd, Ex");
            nextop = F8;
            GETGD;
-            GETEX(q0, 0);
+            GETEX(q0, 0, 0);
            #ifdef PRECISE_CVT
            LDRH_U12(x1, xEmu, offsetof(x64emu_t, mxcsr));
            UBFXx(x1, x1, 13, 2);   // extract round requested
            LSLx_REG(x1, x1, 3);
-            ADDx_U12(x1, x1, 8);    // add the actual add+jump opcodes
            // Construct a "switch case", with each case 2 instructions, so 8 bytes
-            BL(+4); // Branch with Link to next, so LR gets next PC address
+            ADR(xLR, GETMARK);
            ADDx_REG(xLR, xLR, x1);
-            B(xLR); // could use RET, but it's not really one
+            B(xLR);
+            MARK;
            FCVTNSxwD(gd, q0);  // 0: Nearest (even)
            B_NEXT_nocond;
            FCVTMSxwD(gd, q0);  // 1: Toward -inf
@@ -163,36 +167,70 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x51:
            INST_NAME("SQRTSD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
+            if(!box64_dynarec_fastnan) {
+                v1 = fpu_get_scratch(dyn);
+                FCMLTD_0(v1, d0);
+                SHL_64(v1, v1, 63);
+            }
            FSQRTD(d1, d0);
+            if(!box64_dynarec_fastnan) {
+                VORR(d1, d1, v1);
+            }
            VMOVeD(v0, 0, d1, 0);
            break;

        case 0x58:
            INST_NAME("ADDSD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
-            FADDD(d1, v0, d0);  // the high part of the vector is erased...
-            VMOVeD(v0, 0, d1, 0);
+            GETGX(d1, 1);
+            v1 = fpu_get_scratch(dyn);
+            GETEX(d0, 0, 0);
+            if(!box64_dynarec_fastnan) {
+                v0 = fpu_get_scratch(dyn);
+                q0 = fpu_get_scratch(dyn);
+                // check if any input value was NAN
+                FMAXD(v0, d0, d1);    // propagate NAN
+                FCMEQD(v0, v0, v0);    // 0 if NAN, 1 if not NAN
+            }
+            FADDD(v1, d1, d0);  // the high part of the vector is erased...
+            if(!box64_dynarec_fastnan) {
+                FCMEQD(q0, v1, v1);    // 0 => out is NAN
+                VBIC(q0, v0, q0);      // forget it in any input was a NAN already
+                SHL_64(q0, q0, 63);     // only keep the sign bit
+                VORR(v1, v1, q0);      // NAN -> -NAN
+            }
+            VMOVeD(d1, 0, v1, 0);
            break;
        case 0x59:
            INST_NAME("MULSD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
-            FMULD(d1, v0, d0);
-            VMOVeD(v0, 0, d1, 0);
+            GETGX(d1, 1);
+            v1 = fpu_get_scratch(dyn);
+            GETEX(d0, 0, 0);
+            if(!box64_dynarec_fastnan) {
+                v0 = fpu_get_scratch(dyn);
+                q0 = fpu_get_scratch(dyn);
+                // check if any input value was NAN
+                FMAXD(v0, d0, d1);    // propagate NAN
+                FCMEQD(v0, v0, v0);    // 0 if NAN, 1 if not NAN
+            }
+            FMULD(v1, d1, d0);
+            if(!box64_dynarec_fastnan) {
+                FCMEQD(q0, v1, v1);    // 0 => out is NAN
+                VBIC(q0, v0, q0);      // forget it in any input was a NAN already
+                SHL_64(q0, q0, 63);     // only keep the sign bit
+                VORR(v1, v1, q0);      // NAN -> -NAN
+            }
+            VMOVeD(d1, 0, v1, 0);
            break;
        case 0x5A:
            INST_NAME("CVTSD2SS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(d0, 0);
+            GETGX(v0, 1);
+            GETEX(d0, 0, 0);
            d1 = fpu_get_scratch(dyn);
            FCVT_S_D(d1, d0);
            VMOVeS(v0, 0, d1, 0);
@@ -201,49 +239,85 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x5C:
            INST_NAME("SUBSD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
-            FSUBD(d1, v0, d0);
-            VMOVeD(v0, 0, d1, 0);
+            GETGX(d1, 1);
+            v1 = fpu_get_scratch(dyn);
+            GETEX(d0, 0, 0);
+            if(!box64_dynarec_fastnan) {
+                v0 = fpu_get_scratch(dyn);
+                q0 = fpu_get_scratch(dyn);
+                // check if any input value was NAN
+                FMAXD(v0, d0, d1);    // propagate NAN
+                FCMEQD(v0, v0, v0);    // 0 if NAN, 1 if not NAN
+            }
+            FSUBD(v1, d1, d0);
+            if(!box64_dynarec_fastnan) {
+                FCMEQD(q0, v1, v1);    // 0 => out is NAN
+                VBIC(q0, v0, q0);      // forget it in any input was a NAN already
+                SHL_64(q0, q0, 63);     // only keep the sign bit
+                VORR(v1, v1, q0);      // NAN -> -NAN
+            }
+            VMOVeD(d1, 0, v1, 0);
            break;
        case 0x5D:
            INST_NAME("MINSD Gx, Ex");
            nextop = F8;
-            GETG;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            // MINSD: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
+            #if 0
            d0 = fpu_get_scratch(dyn);
            FMINNMD(d0, v0, v1);    // NaN handling may be slightly different, is that a problem?
            VMOVeD(v0, 0, d0, 0);   // to not erase uper part
+            #else
+            FCMPD(v0, v1);
+            B_NEXT(cLS);    //Less than or equal
+            VMOVeD(v0, 0, v1, 0);   // to not erase uper part
+            #endif
            break;
        case 0x5E:
            INST_NAME("DIVSD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
-            FDIVD(d1, v0, d0);
+            GETEX(v1, 0, 0);
+            if(!box64_dynarec_fastnan) {
+                d0 = fpu_get_scratch(dyn);
+                q0 = fpu_get_scratch(dyn);
+                // check if any input value was NAN
+                FMAXD(d0, v0, v1);      // propagate NAN
+                FCMEQD(d0, d0, d0);     // 0 if NAN, 1 if not NAN
+            }
+            FDIVD(d1, v0, v1);
+            if(!box64_dynarec_fastnan) {
+                FCMEQD(q0, d1, d1);     // 0 => out is NAN
+                VBIC(q0, d0, q0);       // forget it in any input was a NAN already
+                SHL_64(q0, q0, 63);     // only keep the sign bit
+                VORR(d1, d1, q0);       // NAN -> -NAN
+            }
            VMOVeD(v0, 0, d1, 0);
            break;
        case 0x5F:
            INST_NAME("MAXSD Gx, Ex");
            nextop = F8;
-            GETG;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            // MAXSD: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
+            #if 0
            d0 = fpu_get_scratch(dyn);
            FMAXNMD(d0, v0, v1);    // NaN handling may be slightly different, is that a problem?
            VMOVeD(v0, 0, d0, 0);   // to not erase uper part
+            #else
+            FCMPD(v0, v1);
+            B_NEXT(cGE);    //Greater than or equal
+            VMOVeD(v0, 0, v1, 0);   // to not erase uper part
+            #endif
            break;

        case 0x70:
            INST_NAME("PSHUFLW Gx, Ex, Ib");
            nextop = F8;
-            GETEX(v1, 1);
-            GETGX(v0);
+            GETEX(v1, 0, 1);
+            GETGX(v0, 1);

            u8 = F8;
            // only low part need to be suffled. VTBL only handle 8bits value, so the 16bits suffles need to be changed in 8bits
@@ -265,11 +339,11 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x7C:
            INST_NAME("HADDPS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            if(MODREG) {
-                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
                v1 = fpu_get_scratch(dyn);
                VLDR128_U12(v1, ed, fixedaddress);
            }
@@ -279,22 +353,18 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0xC2:
            INST_NAME("CMPSD Gx, Ex, Ib");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 1);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 1);
            u8 = F8;
-            if((u8&7)==6){
-                FCMPD(v1, v0);
-            } else {
-                FCMPD(v0, v1);
-            }
+            FCMPD(v0, v1);
            switch(u8&7) {
-                case 0: CSETMx(x2, cEQ); CSELx(x2, xZR, x2, cVS); break;   // Equal
-                case 1: CSETMx(x2, cMI); CSELx(x2, xZR, x2, cVS); break;   // Less than
-                case 2: CSETMx(x2, cLE); CSELx(x2, xZR, x2, cVS); break;   // Less or equal
+                case 0: CSETMx(x2, cEQ); break;   // Equal
+                case 1: CSETMx(x2, cCC); break;   // Less than
+                case 2: CSETMx(x2, cLS); break;   // Less or equal
                case 3: CSETMx(x2, cVS); break;   // NaN
-                case 4: CSETMx(x2, cNE); break;   // Not Equal
+                case 4: CSETMx(x2, cNE); break;   // Not Equal or unordered
                case 5: CSETMx(x2, cCS); break;   // Greater or equal or unordered
-                case 6: CSETMx(x2, cLT); break;   // Greater or unordered, test inverted, N!=V so unordered or less than (inverted)
+                case 6: CSETMx(x2, cHI); break;   // Greater or unordered, test inverted, N!=V so unordered or less than (inverted)
                case 7: CSETMx(x2, cVC); break;   // not NaN
            }
            VMOVQDfrom(v0, 0, x2);
@@ -303,8 +373,8 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0xD0:
            INST_NAME("ADDSUBPS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            q0 = fpu_get_scratch(dyn);
            static float addsubps[4] = {-1.f, 1.f, -1.f, 1.f};
            MAYUSE(addsubps);
@@ -313,17 +383,37 @@ uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            VFMLAQS(v0, v1, q0);
            break;

+        case 0xD6:
+            INST_NAME("MOVDQ2Q Gm, Ex");
+            nextop = F8;
+            GETGM(v0);
+            GETEX(v1, 0, 0);
+            VMOV(v0, v1);
+            break;
+
+        case 0xE6:
+            INST_NAME("CVTPD2DQ Gx, Ex");
+            nextop = F8;
+            GETEX(v1, 0, 0);
+            GETGX_empty(v0);
+            u8 = sse_setround(dyn, ninst, x1, x2, x3);
+            VFRINTIDQ(v0, v1);
+            x87_restoreround(dyn, ninst, u8);
+            VFCVTNSQD(v0, v0);  // convert double -> int64
+            SQXTN_32(v0, v0);   // convert int64 -> int32 with saturation in lower part, RaZ high part
+            break;
+
        case 0xF0:
            INST_NAME("LDDQU Gx,Ex");
            nextop = F8;
            GETG;
            if(MODREG) {
-                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 7, rex, NULL, 0, 0);
                VLDR128_U12(v0, ed, fixedaddress);
            }
            break;
@@ -15,27 +15,37 @@
 #include "callback.h"
 #include "emu/x64run_private.h"
 #include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
+#include "dynarec_native.h"

+#include "arm64_printer.h"
+#include "dynarec_arm64_private.h"
 #include "dynarec_arm64_functions.h"
 #include "dynarec_arm64_helper.h"

 // Get Ex as a single, not a quad (warning, x2 get used)
-#define GETEX(a, D)                                                                                 \
-    if(MODREG) {                                                                                    \
-        a = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));                                     \
-    } else {                                                                                        \
-        a = fpu_get_scratch(dyn);                                                                   \
-        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, D);     \
-        VLDR32_U12(a, ed, fixedaddress);                                                            \
+#define GETEX(a, w, D)                                                                                  \
+    if(MODREG) {                                                                                        \
+        a = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), w);                                      \
+    } else {                                                                                            \
+        a = fpu_get_scratch(dyn);                                                                       \
+        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, D);   \
+        VLDR32_U12(a, ed, fixedaddress);                                                                \
+    }
+
+// Get EX as a quad
+#define GETEXQ(a, w, D)                                                                                 \
+    if(MODREG) {                                                                                        \
+        a = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), w);                                      \
+    } else {                                                                                            \
+        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, D);  \
+        a = fpu_get_scratch(dyn);                                                                       \
+        VLDR128_U12(a, ed, fixedaddress);                                                               \
    }

 #define GETG        gd = ((nextop&0x38)>>3)+(rex.r<<3)

-#define GETGX(a)    gd = ((nextop&0x38)>>3)+(rex.r<<3); \
-                    a = sse_get_reg(dyn, ninst, x1, gd)
+#define GETGX(a, w) gd = ((nextop&0x38)>>3)+(rex.r<<3); \
+                    a = sse_get_reg(dyn, ninst, x1, gd, w)

 #define GETGX_empty(a)  gd = ((nextop&0x38)>>3)+(rex.r<<3);         \
                        a = sse_get_reg_empty(dyn, ninst, x1, gd)
@@ -70,12 +80,12 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            nextop = F8;
            GETG;
            if(MODREG) {
-                v0 = sse_get_reg(dyn, ninst, x1, gd);
-                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
+                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
+                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 0);
                VMOVeS(v0, 0, q0, 0);
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                VLDR32_U12(v0, ed, fixedaddress);
            }
            break;
@@ -83,12 +93,12 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("MOVSS Ex, Gx");
            nextop = F8;
            GETG;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
            if(MODREG) {
-                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
+                q0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 1);
                VMOVeS(q0, 0, v0, 0);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, 0);
                VSTR32_U12(v0, ed, fixedaddress);
            }
            break;
@@ -96,9 +106,9 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("MOVSLDUP Gx, Ex");
            nextop = F8;
            if(MODREG) {
-                q1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                q1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
                q1 = fpu_get_scratch(dyn);
                VLDR128_U12(q1, ed, fixedaddress);
            }
@@ -110,9 +120,9 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("MOVSHDUP Gx, Ex");
            nextop = F8;
            if(MODREG) {
-                q1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                q1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
                q1 = fpu_get_scratch(dyn);
                VLDR128_U12(q1, ed, fixedaddress);
            }
@@ -129,7 +139,7 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x2A:
            INST_NAME("CVTSI2SS Gx, Ed");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            GETED(0);
            d1 = fpu_get_scratch(dyn);
            if(rex.w) {
@@ -144,23 +154,22 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            INST_NAME("CVTTSS2SI Gd, Ex");
            nextop = F8;
            GETGD;
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FCVTZSxwS(gd, d0);
            break;
        case 0x2D:
            INST_NAME("CVTSS2SI Gd, Ex");
            nextop = F8;
            GETGD;
-            GETEX(q0, 0);
+            GETEX(q0, 0, 0);
            #ifdef PRECISE_CVT
            LDRH_U12(x1, xEmu, offsetof(x64emu_t, mxcsr));
            UBFXx(x1, x1, 13, 2);   // extract round requested
            LSLx_REG(x1, x1, 3);
-            ADDx_U12(x1, x1, 8);    // add the actual add+jump opcodes
            // Construct a "switch case", with each case 2 instructions, so 8 bytes
-            BL(+4); // Branch with Link to next, so LR gets next PC address
+            ADR(xLR, GETMARK);
            ADDx_REG(xLR, xLR, x1);
-            B(xLR); // could use RET, but it's not really one
+            B(xLR);
            FCVTNSxwS(gd, q0);  // 0: Nearest (even)
            B_NEXT_nocond;
            FCVTMSxwS(gd, q0);  // 1: Toward -inf
@@ -175,17 +184,17 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x51:
            INST_NAME("SQRTSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FSQRTS(d1, d0);
            VMOVeS(v0, 0, d1, 0);
            break;
        case 0x52:
            INST_NAME("RSQRTSS Gx, Ex");
            nextop = F8;
-            GETEX(v1, 0);
-            GETGX_empty(v0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            d0 = fpu_get_scratch(dyn);
            d1 = fpu_get_scratch(dyn);
            // so here: F32: Imm8 = abcd efgh that gives => aBbbbbbc defgh000 00000000 00000000
@@ -201,8 +210,8 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x53:
            INST_NAME("RCPSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            d0 = fpu_get_scratch(dyn);
            FMOVS_8(d0, 0b01110000);    //1.0f
            FDIVS(d0, d0, v1);
@@ -212,34 +221,34 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x58:
            INST_NAME("ADDSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FADDS(d1, v0, d0);  // the high part of the vector is erased...
            VMOVeS(v0, 0, d1, 0);
            break;
        case 0x59:
            INST_NAME("MULSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FMULS(d1, v0, d0);
            VMOVeS(v0, 0, d1, 0);
            break;
        case 0x5A:
            INST_NAME("CVTSS2SD Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            d0 = fpu_get_scratch(dyn);
            FCVT_D_S(d0, v1);
            VMOVeD(v0, 0, d0, 0);
            break;
        case 0x5B:
-            INST_NAME("CVTPS2DQ Gx, Ex");
+            INST_NAME("CVTTPS2DQ Gx, Ex");
            nextop = F8;
-            GETEX(d0, 0);
+            GETEXQ(d0, 0, 0);
            GETGX_empty(v0);
            VFCVTZSQS(v0, d0);
            break;
@@ -247,61 +256,72 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x5C:
            INST_NAME("SUBSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FSUBS(d1, v0, d0);
            VMOVeS(v0, 0, d1, 0);
            break;
        case 0x5D:
            INST_NAME("MINSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            // MINSS: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
+            #if 0
            d0 = fpu_get_scratch(dyn);
            FMINNMS(d0, v0, v1);    // NaN handling may be slightly different, is that a problem?
            VMOVeS(v0, 0, d0, 0);   // to not erase uper part
+            #else
+            FCMPS(v0, v1);
+            B_NEXT(cLS);    //Less than or equal
+            VMOVeS(v0, 0, v1, 0);   // to not erase uper part
+            #endif
            break;
        case 0x5E:
            INST_NAME("DIVSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
+            GETGX(v0, 1);
            d1 = fpu_get_scratch(dyn);
-            GETEX(d0, 0);
+            GETEX(d0, 0, 0);
            FDIVS(d1, v0, d0);
            VMOVeS(v0, 0, d1, 0);
            break;
        case 0x5F:
            INST_NAME("MAXSS Gx, Ex");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 0);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 0);
            // MAXSS: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
+            #if 0
            d0 = fpu_get_scratch(dyn);
            FMAXNMS(d0, v0, v1);    // NaN handling may be slightly different, is that a problem?
            VMOVeS(v0, 0, d0, 0);   // to not erase uper part
+            #else
+            FCMPS(v0, v1);
+            B_NEXT(cGE);    //Greater than or equal
+            VMOVeS(v0, 0, v1, 0);   // to not erase uper part
+            #endif
            break;
            
        case 0x6F:
            INST_NAME("MOVDQU Gx,Ex");// no alignment constraint on NEON here, so same as MOVDQA
            nextop = F8;
-            GETG;
-            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
            if(MODREG) {
-                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3));
+                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
+                GETGX_empty(v0);
                VMOVQ(v0, v1);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, 0, 0);
+                GETGX_empty(v0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
                VLDR128_U12(v0, ed, fixedaddress);
            }
            break;
        case 0x70:
            INST_NAME("PSHUFHW Gx, Ex, Ib");
            nextop = F8;
-            GETEX(v1, 1);
-            GETGX(v0);
-
+            GETEXQ(v1, 0, 1);
+            GETGX(v0, 1);
            u8 = F8;
            // only high part need to be suffled. VTBL only handle 8bits value, so the 16bits suffles need to be changed in 8bits
            u64 = 0;
@@ -322,28 +342,25 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0x7E:
            INST_NAME("MOVQ Gx, Ex");
            nextop = F8;
-            GETG;
            if(MODREG) {
-                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
-                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3), 0);
+                GETGX_empty(v0);
                FMOVD(v0, v1);
            } else {
-                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                GETGX_empty(v0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                VLDR64_U12(v0, ed, fixedaddress);
            }
            break;
        case 0x7F:
            INST_NAME("MOVDQU Ex,Gx");
            nextop = F8;
-            GETG;
+            GETGX(v0, 0);
            if(MODREG) {
-                v0 = sse_get_reg(dyn, ninst, x1, gd);
                v1 = sse_get_reg_empty(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
                VMOVQ(v1, v0);
            } else {
-                v0 = sse_get_reg(dyn, ninst, x1, gd);
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<4, 15, rex, NULL, 0, 0);
                VSTR128_U12(v0, ed, fixedaddress);
            }
            break;
@@ -381,22 +398,18 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0xC2:
            INST_NAME("CMPSS Gx, Ex, Ib");
            nextop = F8;
-            GETGX(v0);
-            GETEX(v1, 1);
+            GETGX(v0, 1);
+            GETEX(v1, 0, 1);
            u8 = F8;
-            if((u8&7)==6){
-                FCMPS(v1, v0);
-            } else {
-                FCMPS(v0, v1);
-            }
+            FCMPS(v0, v1);
            switch(u8&7) {
-                case 0: CSETMw(x2, cEQ); CSELw(x2, xZR, x2, cVS); break;   // Equal
-                case 1: CSETMw(x2, cMI); CSELw(x2, xZR, x2, cVS); break;   // Less than
-                case 2: CSETMw(x2, cLE); CSELw(x2, xZR, x2, cVS); break;   // Less or equal
+                case 0: CSETMw(x2, cEQ); break;   // Equal
+                case 1: CSETMw(x2, cCC); break;   // Less than
+                case 2: CSETMw(x2, cLS); break;   // Less or equal
                case 3: CSETMw(x2, cVS); break;   // NaN
-                case 4: CSETMw(x2, cNE); break;   // Not Equal
+                case 4: CSETMw(x2, cNE); break;   // Not Equal or unordered
                case 5: CSETMw(x2, cCS); break;   // Greater or equal or unordered
-                case 6: CSETMw(x2, cLT); break;   // Greater or unordered, test inverted, N!=V so unordered or less than (inverted)
+                case 6: CSETMw(x2, cHI); break;   // Greater or unordered, test inverted, N!=V so unordered or less than (inverted)
                case 7: CSETMw(x2, cVC); break;   // not NaN
            }
            VMOVQSfrom(v0, 0, x2);
@@ -407,11 +420,11 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
            nextop = F8;
            GETGX_empty(v0);
            if(MODREG) {
-                v1 = mmx_get_reg(dyn, ninst, x1, (nextop&7));
+                v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, (nextop&7));
                VEORQ(v0, v0, v0);  // usefull?
                VMOV(v0, v1);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, 0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
                VLDR64_U12(v0, ed, fixedaddress);
            }
            break;
@@ -419,7 +432,14 @@ uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int n
        case 0xE6:
            INST_NAME("CVTDQ2PD Gx, Ex");
            nextop = F8;
-            GETEX(v1, 0);
+            // cannot use GETEX because we want 64bits not 32bits
+            if(MODREG) {
+                v1 = sse_get_reg(dyn, ninst, x1, (nextop&7)+(rex.b<<3), 0);
+            } else {
+                v1 = fpu_get_scratch(dyn);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, 0);
+                VLDR64_U12(v1, ed, fixedaddress);
+            }
            GETGX_empty(v0);
            d0 = fpu_get_scratch(dyn);
            SXTL_32(v0, v1);
@@ -0,0 +1,946 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+#include <string.h>
+#include <math.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "tools/bridge_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "emu/x87emu_private.h"
+#include "x64trace.h"
+#include "signals.h"
+#include "dynarec_native.h"
+#include "dynarec_arm64_private.h"
+#include "dynarec_arm64_functions.h"
+#include "custommem.h"
+#include "bridge.h"
+
+void arm_fstp(x64emu_t* emu, void* p)
+{
+    if(ST0.q!=STld(0).uref)
+        D2LD(&ST0.d, p);
+    else
+        memcpy(p, &STld(0).ld, 10);
+}
+
+void arm_print_armreg(x64emu_t* emu, uintptr_t reg, uintptr_t n)
+{
+    (void)emu;
+    dynarec_log(LOG_DEBUG, "R%lu=0x%lx (%lu)\n", n, reg, reg);
+}
+
+void arm_f2xm1(x64emu_t* emu)
+{
+    ST0.d = exp2(ST0.d) - 1.0;
+}
+void arm_fyl2x(x64emu_t* emu)
+{
+    ST(1).d = log2(ST0.d)*ST(1).d;
+}
+void arm_ftan(x64emu_t* emu)
+{
+    ST0.d = tan(ST0.d);
+    emu->sw.f.F87_C2 = 0;
+}
+void arm_fpatan(x64emu_t* emu)
+{
+    ST1.d = atan2(ST1.d, ST0.d);
+}
+void arm_fxtract(x64emu_t* emu)
+{
+    int32_t tmp32s = (ST1.q&0x7ff0000000000000LL)>>52;
+    tmp32s -= 1023;
+    ST1.d /= exp2(tmp32s);
+    ST0.d = tmp32s;
+}
+void arm_fprem(x64emu_t* emu)
+{
+    int32_t tmp32s = ST0.d / ST1.d;
+    ST0.d -= ST1.d * tmp32s;
+    emu->sw.f.F87_C2 = 0;
+    emu->sw.f.F87_C0 = (tmp32s&1);
+    emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
+    emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
+}
+void arm_fyl2xp1(x64emu_t* emu)
+{
+    ST(1).d = log2(ST0.d + 1.0)*ST(1).d;
+}
+void arm_fsincos(x64emu_t* emu)
+{
+    sincos(ST1.d, &ST1.d, &ST0.d);
+    emu->sw.f.F87_C2 = 0;
+}
+void arm_frndint(x64emu_t* emu)
+{
+    ST0.d = fpu_round(emu, ST0.d);
+}
+void arm_fscale(x64emu_t* emu)
+{
+    if(ST0.d!=0.0)
+        ST0.d *= exp2(trunc(ST1.d));
+}
+void arm_fsin(x64emu_t* emu)
+{
+    ST0.d = sin(ST0.d);
+    emu->sw.f.F87_C2 = 0;
+}
+void arm_fcos(x64emu_t* emu)
+{
+    ST0.d = cos(ST0.d);
+    emu->sw.f.F87_C2 = 0;
+}
+
+void arm_fbld(x64emu_t* emu, uint8_t* ed)
+{
+    fpu_fbld(emu, ed);
+}
+
+void arm_fild64(x64emu_t* emu, int64_t* ed)
+{
+    int64_t tmp;
+    memcpy(&tmp, ed, sizeof(tmp));
+    ST0.d = tmp;
+    STll(0).sq = tmp;
+    STll(0).sref = ST0.sq;
+}
+
+void arm_fbstp(x64emu_t* emu, uint8_t* ed)
+{
+    fpu_fbst(emu, ed);
+}
+
+void arm_fistp64(x64emu_t* emu, int64_t* ed)
+{
+    // used of memcpy to avoid aligments issues
+    if(STll(0).sref==ST(0).sq) {
+        memcpy(ed, &STll(0).sq, sizeof(int64_t));
+    } else {
+        int64_t tmp;
+        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL) || !isfinite(ST0.d))
+            tmp = 0x8000000000000000LL;
+        else
+            tmp = fpu_round(emu, ST0.d);
+        memcpy(ed, &tmp, sizeof(tmp));
+    }
+}
+
+void arm_fistt64(x64emu_t* emu, int64_t* ed)
+{
+    // used of memcpy to avoid aligments issues
+    int64_t tmp = ST0.d;
+    memcpy(ed, &tmp, sizeof(tmp));
+}
+
+void arm_fld(x64emu_t* emu, uint8_t* ed)
+{
+    memcpy(&STld(0).ld, ed, 10);
+    LD2D(&STld(0), &ST(0).d);
+    STld(0).uref = ST0.q;
+}
+
+void arm_ud(x64emu_t* emu)
+{
+    emit_signal(emu, SIGILL, (void*)R_RIP, 0);
+}
+
+void arm_priv(x64emu_t* emu)
+{
+    emit_signal(emu, SIGSEGV, (void*)R_RIP, 0);
+}
+
+void arm_fsave(x64emu_t* emu, uint8_t* ed)
+{
+    fpu_savenv(emu, (char*)ed, 0);
+
+    uint8_t* p = ed;
+    p += 28;
+    for (int i=0; i<8; ++i) {
+        LD2D(p, &ST(i).d);
+        p+=10;
+    }
+}
+void arm_frstor(x64emu_t* emu, uint8_t* ed)
+{
+    fpu_loadenv(emu, (char*)ed, 0);
+
+    uint8_t* p = ed;
+    p += 28;
+    for (int i=0; i<8; ++i) {
+        D2LD(&ST(i).d, p);
+        p+=10;
+    }
+
+}
+
+void arm_fprem1(x64emu_t* emu)
+{
+    // simplified version
+    int32_t tmp32s = round(ST0.d / ST1.d);
+    ST0.d -= ST1.d*tmp32s;
+    emu->sw.f.F87_C2 = 0;
+    emu->sw.f.F87_C0 = (tmp32s&1);
+    emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
+    emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
+}
+
+static uint8_t ff_mult(uint8_t a, uint8_t b)
+{
+	int retval = 0;
+
+	for(int i = 0; i < 8; i++) {
+		if((b & 1) == 1) 
+			retval ^= a;
+		
+		if((a & 0x80)) {
+			a <<= 1;
+			a  ^= 0x1b;
+		} else {
+			a <<= 1;
+		}
+		
+		b >>= 1;
+	}
+	
+	return retval;
+}
+
+void arm_aesimc(x64emu_t* emu, int xmm)
+{
+    sse_regs_t eax1 = emu->xmm[xmm];
+
+    for(int j=0; j<4; ++j) {
+        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x0E, eax1.ub[0+j*4]) ^ ff_mult(0x0B, eax1.ub[1+j*4]) ^ ff_mult(0x0D, eax1.ub[2+j*4]) ^ ff_mult(0x09, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[1+j*4] = ff_mult(0x09, eax1.ub[0+j*4]) ^ ff_mult(0x0E, eax1.ub[1+j*4]) ^ ff_mult(0x0B, eax1.ub[2+j*4]) ^ ff_mult(0x0D, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[2+j*4] = ff_mult(0x0D, eax1.ub[0+j*4]) ^ ff_mult(0x09, eax1.ub[1+j*4]) ^ ff_mult(0x0E, eax1.ub[2+j*4]) ^ ff_mult(0x0B, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x0B, eax1.ub[0+j*4]) ^ ff_mult(0x0D, eax1.ub[1+j*4]) ^ ff_mult(0x09, eax1.ub[2+j*4]) ^ ff_mult(0x0E, eax1.ub[3+j*4]);
+    }
+}
+void arm_aesmc(x64emu_t* emu, int xmm)
+{
+    sse_regs_t eax1 = emu->xmm[xmm];
+
+    for(int j=0; j<4; ++j) {
+        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x02, eax1.ub[0+j*4]) ^ ff_mult(0x03, eax1.ub[1+j*4]) ^               eax1.ub[2+j*4]  ^               eax1.ub[3+j*4] ;
+        emu->xmm[xmm].ub[1+j*4] =               eax1.ub[0+j*4]  ^ ff_mult(0x02, eax1.ub[1+j*4]) ^ ff_mult(0x03, eax1.ub[2+j*4]) ^               eax1.ub[3+j*4] ;
+        emu->xmm[xmm].ub[2+j*4] =               eax1.ub[0+j*4]  ^               eax1.ub[1+j*4]  ^ ff_mult(0x02, eax1.ub[2+j*4]) ^ ff_mult(0x03, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x03, eax1.ub[0+j*4]) ^               eax1.ub[1+j*4]  ^               eax1.ub[2+j*4]  ^ ff_mult(0x02, eax1.ub[3+j*4]);
+    }
+}
+void arm_aesdlast(x64emu_t* emu, int xmm)
+{
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  N  K  H  E  B  O  L  I  F  C  P  M  J  G  D
+    const uint8_t invshiftrows[] = {0,13,10, 7, 4, 1,14,11, 8, 5, 2,15,12, 9, 6, 3};
+    const uint8_t invsubbytes[256] = {
+        0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38, 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
+        0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87, 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
+        0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d, 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
+        0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2, 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
+        0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
+        0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda, 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
+        0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a, 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
+        0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02, 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
+        0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea, 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
+        0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85, 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
+        0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89, 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
+        0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20, 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
+        0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31, 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
+        0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d, 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
+        0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0, 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
+        0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26, 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
+    };
+
+    sse_regs_t eax1;
+    for(int i=0; i<16; ++i)
+        eax1.ub[i] = emu->xmm[xmm].ub[invshiftrows[i]];
+    //STATE ← InvSubBytes( STATE );
+    for(int i=0; i<16; ++i)
+        emu->xmm[xmm].ub[i] = invsubbytes[eax1.ub[i]];
+
+}
+static const uint8_t shiftrows[] = {0, 5,10,15, 4, 9,14, 3, 8,13, 2, 7,12, 1, 6,11};
+static const uint8_t subbytes[256] = {
+    0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
+    0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
+    0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
+    0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
+    0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
+    0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
+    0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
+    0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
+    0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
+    0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
+    0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
+    0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
+    0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
+    0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
+    0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
+    0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
+};
+void arm_aeselast(x64emu_t* emu, int xmm)
+{
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  F  K  P  E  J  O  D  I  N  C  H  M  B  G  L
+    sse_regs_t eax1;
+    for(int i=0; i<16; ++i)
+        eax1.ub[i] = emu->xmm[xmm].ub[shiftrows[i]];
+    //STATE ← SubBytes( STATE );
+    for(int i=0; i<16; ++i)
+        emu->xmm[xmm].ub[i] = subbytes[eax1.ub[i]];
+}
+void arm_aesd(x64emu_t* emu, int xmm)
+{
+    arm_aesdlast(emu, xmm);
+    arm_aesimc(emu, xmm);
+}
+void arm_aese(x64emu_t* emu, int xmm)
+{
+    arm_aeselast(emu, xmm);
+    arm_aesmc(emu, xmm);
+}
+void arm_aeskeygenassist(x64emu_t* emu, int gx, int ex, void* p, uint32_t u8)
+{
+    sse_regs_t *EX = p?((sse_regs_t*)p):&emu->xmm[ex];
+    sse_regs_t *GX = &emu->xmm[gx];
+    for (int i = 4; i < 8; ++i)
+        GX->ub[i] = subbytes[EX->ub[i]];
+    for (int i = 12; i < 16; ++i)
+        GX->ub[i] = subbytes[EX->ub[i]];
+    GX->ud[0] = GX->ud[1];
+    uint8_t tmp8u = GX->ub[4];
+    GX->ud[1] = GX->ud[1] >> 8;
+    GX->ub[7] = tmp8u;
+    GX->ud[1] ^= u8;
+    GX->ud[2] = GX->ud[3];
+    tmp8u = GX->ub[12];
+    GX->ud[3] = GX->ud[3] >> 8;
+    GX->ub[15] = tmp8u;
+    GX->ud[3] ^= u8;
+}
+
+void arm_pclmul(x64emu_t* emu, int gx, int ex, void* p, uint32_t u8)
+{
+    sse_regs_t *EX = p?((sse_regs_t*)p):&emu->xmm[ex];
+    sse_regs_t *GX = &emu->xmm[gx];
+    int g = (u8&1)?1:0;
+    int e = (u8&0b10000)?1:0;
+    __int128 result = 0;
+    __int128 op2 = EX->q[e];
+    for (int i=0; i<64; ++i)
+        if(GX->q[g]&(1LL<<i))
+            result ^= (op2<<i);
+
+    GX->q[0] = result&0xffffffffffffffffLL;
+    GX->q[1] = (result>>64)&0xffffffffffffffffLL;
+}
+
+void arm_clflush(x64emu_t* emu, void* p)
+{
+    cleanDBFromAddressRange((uintptr_t)p, 8, 0);
+}
+
+
+#define XMM0    0
+#define XMM8    16
+#define X870    8
+#define EMM0    8
+#define SCRATCH0    24
+
+// Get a FPU scratch reg
+int fpu_get_scratch(dynarec_arm_t* dyn)
+{
+    return SCRATCH0 + dyn->n.fpu_scratch++;  // return an Sx
+}
+// Reset scratch regs counter
+void fpu_reset_scratch(dynarec_arm_t* dyn)
+{
+    dyn->n.fpu_scratch = 0;
+}
+// Get a x87 double reg
+int fpu_get_reg_x87(dynarec_arm_t* dyn, int t, int n)
+{
+    int i=X870;
+    while (dyn->n.fpuused[i]) ++i;
+    dyn->n.fpuused[i] = 1;
+    dyn->n.neoncache[i].n = n;
+    dyn->n.neoncache[i].t = t;
+    dyn->n.news |= (1<<i);
+    return i; // return a Dx
+}
+// Free a FPU double reg
+void fpu_free_reg(dynarec_arm_t* dyn, int reg)
+{
+    // TODO: check upper limit?
+    dyn->n.fpuused[reg] = 0;
+    if(dyn->n.neoncache[reg].t!=NEON_CACHE_ST_F && dyn->n.neoncache[reg].t!=NEON_CACHE_ST_D)
+        dyn->n.neoncache[reg].v = 0;
+}
+// Get an MMX double reg
+int fpu_get_reg_emm(dynarec_arm_t* dyn, int emm)
+{
+    dyn->n.fpuused[EMM0 + emm] = 1;
+    dyn->n.neoncache[EMM0 + emm].t = NEON_CACHE_MM;
+    dyn->n.neoncache[EMM0 + emm].n = emm;
+    dyn->n.news |= (1<<(EMM0 + emm));
+    return EMM0 + emm;
+}
+// Get an XMM quad reg
+int fpu_get_reg_xmm(dynarec_arm_t* dyn, int t, int xmm)
+{
+    int i;
+    if(xmm>7) {
+        i = XMM8 + xmm - 8;
+    } else {
+        i = XMM0 + xmm;
+    }
+    dyn->n.fpuused[i] = 1;
+    dyn->n.neoncache[i].t = t;
+    dyn->n.neoncache[i].n = xmm;
+    dyn->n.news |= (1<<i);
+    return i;
+}
+// Reset fpu regs counter
+void fpu_reset_reg(dynarec_arm_t* dyn)
+{
+    dyn->n.fpu_reg = 0;
+    for (int i=0; i<24; ++i) {
+        dyn->n.fpuused[i]=0;
+        dyn->n.neoncache[i].v = 0;
+    }
+
+}
+
+int neoncache_get_st(dynarec_arm_t* dyn, int ninst, int a)
+{
+    if (dyn->insts[ninst].n.swapped) {
+        if(dyn->insts[ninst].n.combined1 == a)
+            a = dyn->insts[ninst].n.combined2;
+        else if(dyn->insts[ninst].n.combined2 == a)
+            a = dyn->insts[ninst].n.combined1;
+    }
+    for(int i=0; i<24; ++i)
+        if((dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_F
+         || dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_D)
+         && dyn->insts[ninst].n.neoncache[i].n==a)
+            return dyn->insts[ninst].n.neoncache[i].t;
+    // not in the cache yet, so will be fetched...
+    return NEON_CACHE_ST_D;
+}
+
+int neoncache_get_current_st(dynarec_arm_t* dyn, int ninst, int a)
+{
+    (void)ninst;
+    if(!dyn->insts)
+        return NEON_CACHE_ST_D;
+    for(int i=0; i<24; ++i)
+        if((dyn->n.neoncache[i].t==NEON_CACHE_ST_F
+         || dyn->n.neoncache[i].t==NEON_CACHE_ST_D)
+         && dyn->n.neoncache[i].n==a)
+            return dyn->n.neoncache[i].t;
+    // not in the cache yet, so will be fetched...
+    return NEON_CACHE_ST_D;
+}
+
+int neoncache_get_st_f(dynarec_arm_t* dyn, int ninst, int a)
+{
+    /*if(a+dyn->insts[ninst].n.stack_next-st<0)
+        // The STx has been pushed at the end of instructon, so stop going back
+        return -1;*/
+    for(int i=0; i<24; ++i)
+        if(dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_F
+         && dyn->insts[ninst].n.neoncache[i].n==a)
+            return i;
+    return -1;
+} 
+int neoncache_get_st_f_noback(dynarec_arm_t* dyn, int ninst, int a)
+{
+    for(int i=0; i<24; ++i)
+        if(dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_F
+         && dyn->insts[ninst].n.neoncache[i].n==a)
+            return i;
+    return -1;
+} 
+int neoncache_get_current_st_f(dynarec_arm_t* dyn, int a)
+{
+    for(int i=0; i<24; ++i)
+        if(dyn->n.neoncache[i].t==NEON_CACHE_ST_F
+         && dyn->n.neoncache[i].n==a)
+            return i;
+    return -1;
+} 
+static void neoncache_promote_double_forward(dynarec_arm_t* dyn, int ninst, int maxinst, int a);
+static void neoncache_promote_double_internal(dynarec_arm_t* dyn, int ninst, int maxinst, int a);
+static void neoncache_promote_double_combined(dynarec_arm_t* dyn, int ninst, int maxinst, int a)
+{
+    if(a == dyn->insts[ninst].n.combined1 || a == dyn->insts[ninst].n.combined2) {
+        if(a == dyn->insts[ninst].n.combined1) {
+            a = dyn->insts[ninst].n.combined2;
+        } else 
+            a = dyn->insts[ninst].n.combined1;
+        int i = neoncache_get_st_f_noback(dyn, ninst, a);
+        //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_combined, ninst=%d combined%c %d i=%d (stack:%d/%d)\n", ninst, (a == dyn->insts[ninst].n.combined2)?'2':'1', a ,i, dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
+        if(i>=0) {
+            dyn->insts[ninst].n.neoncache[i].t = NEON_CACHE_ST_D;
+            if(!dyn->insts[ninst].n.barrier)
+                neoncache_promote_double_internal(dyn, ninst-1, maxinst, a-dyn->insts[ninst].n.stack_push);
+            // go forward is combined is not pop'd
+            if(a-dyn->insts[ninst].n.stack_pop>=0)
+                if(!dyn->insts[ninst+1].n.barrier)
+                    neoncache_promote_double_forward(dyn, ninst+1, maxinst, a-dyn->insts[ninst].n.stack_pop);
+        }
+    }
+}
+static void neoncache_promote_double_internal(dynarec_arm_t* dyn, int ninst, int maxinst, int a)
+{
+    if(dyn->insts[ninst+1].n.barrier)
+        return;
+    while(ninst>=0) {
+        a+=dyn->insts[ninst].n.stack_pop;    // adjust Stack depth: add pop'd ST (going backward)
+        int i = neoncache_get_st_f(dyn, ninst, a);
+        //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_internal, ninst=%d, a=%d st=%d:%d, i=%d\n", ninst, a, dyn->insts[ninst].n.stack, dyn->insts[ninst].n.stack_next, i);
+        if(i<0) return;
+        dyn->insts[ninst].n.neoncache[i].t = NEON_CACHE_ST_D;
+        // check combined propagation too
+        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2) {
+            if(dyn->insts[ninst].n.swapped) {
+                //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_internal, ninst=%d swapped %d/%d vs %d with st %d\n", ninst, dyn->insts[ninst].n.combined1 ,dyn->insts[ninst].n.combined2, a, dyn->insts[ninst].n.stack);
+                if (a==dyn->insts[ninst].n.combined1)
+                    a = dyn->insts[ninst].n.combined2;
+                else if (a==dyn->insts[ninst].n.combined2)
+                    a = dyn->insts[ninst].n.combined1;
+            } else {
+                //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_internal, ninst=%d combined %d/%d vs %d with st %d\n", ninst, dyn->insts[ninst].n.combined1 ,dyn->insts[ninst].n.combined2, a, dyn->insts[ninst].n.stack);
+                neoncache_promote_double_combined(dyn, ninst, maxinst, a);
+            }
+        }
+        a-=dyn->insts[ninst].n.stack_push;  // // adjust Stack depth: remove push'd ST (going backward)
+        --ninst;
+        if(ninst<0 || a<0 || dyn->insts[ninst].n.barrier)
+            return;
+    }
+}
+
+static void neoncache_promote_double_forward(dynarec_arm_t* dyn, int ninst, int maxinst, int a)
+{
+    while((ninst!=-1) && (ninst<maxinst) && (a>=0)) {
+        a+=dyn->insts[ninst].n.stack_push;  // // adjust Stack depth: add push'd ST (going forward)
+        if((dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2) && dyn->insts[ninst].n.swapped) {
+            //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_forward, ninst=%d swapped %d/%d vs %d with st %d\n", ninst, dyn->insts[ninst].n.combined1 ,dyn->insts[ninst].n.combined2, a, dyn->insts[ninst].n.stack);
+            if (a==dyn->insts[ninst].n.combined1)
+                a = dyn->insts[ninst].n.combined2;
+            else if (a==dyn->insts[ninst].n.combined2)
+                a = dyn->insts[ninst].n.combined1;
+        }
+        int i = neoncache_get_st_f_noback(dyn, ninst, a);
+        //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_forward, ninst=%d, a=%d st=%d:%d(%d/%d), i=%d\n", ninst, a, dyn->insts[ninst].n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop, i);
+        if(i<0) return;
+        dyn->insts[ninst].n.neoncache[i].t = NEON_CACHE_ST_D;
+        // check combined propagation too
+        if((dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2) && !dyn->insts[ninst].n.swapped) {
+            //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double_forward, ninst=%d combined %d/%d vs %d with st %d\n", ninst, dyn->insts[ninst].n.combined1 ,dyn->insts[ninst].n.combined2, a, dyn->insts[ninst].n.stack);
+            neoncache_promote_double_combined(dyn, ninst, maxinst, a);
+        }
+        a-=dyn->insts[ninst].n.stack_pop;    // adjust Stack depth: remove pop'd ST (going forward)
+        if(dyn->insts[ninst].x64.has_next && !dyn->insts[ninst].n.barrier)
+            ++ninst;
+        else
+            ninst=-1;
+    }
+    if(ninst==maxinst)
+        neoncache_promote_double(dyn, ninst, a);
+}
+
+void neoncache_promote_double(dynarec_arm_t* dyn, int ninst, int a)
+{
+    int i = neoncache_get_current_st_f(dyn, a);
+    //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double, ninst=%d a=%d st=%d i=%d\n", ninst, a, dyn->n.stack, i);
+    if(i<0) return;
+    dyn->n.neoncache[i].t = NEON_CACHE_ST_D;
+    dyn->insts[ninst].n.neoncache[i].t = NEON_CACHE_ST_D;
+    // check combined propagation too
+    if(dyn->n.combined1 || dyn->n.combined2) {
+        if(dyn->n.swapped) {
+            //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double, ninst=%d swapped! %d/%d vs %d\n", ninst, dyn->n.combined1 ,dyn->n.combined2, a);
+            if(dyn->n.combined1 == a)
+                a = dyn->n.combined2;
+            else if(dyn->n.combined2 == a)
+                a = dyn->n.combined1;
+        } else {
+            //if(box64_dynarec_dump) dynarec_log(LOG_NONE, "neoncache_promote_double, ninst=%d combined! %d/%d vs %d\n", ninst, dyn->n.combined1 ,dyn->n.combined2, a);
+            if(dyn->n.combined1 == a)
+                neoncache_promote_double(dyn, ninst, dyn->n.combined2);
+            else if(dyn->n.combined2 == a)
+                neoncache_promote_double(dyn, ninst, dyn->n.combined1);
+        }
+    }
+    a-=dyn->insts[ninst].n.stack_push;  // // adjust Stack depth: remove push'd ST (going backward)
+    if(!ninst || a<0) return;
+    neoncache_promote_double_internal(dyn, ninst-1, ninst, a);
+}
+
+int neoncache_combine_st(dynarec_arm_t* dyn, int ninst, int a, int b)
+{
+    dyn->n.combined1=a;
+    dyn->n.combined2=b;
+    if( neoncache_get_current_st(dyn, ninst, a)==NEON_CACHE_ST_F
+     && neoncache_get_current_st(dyn, ninst, b)==NEON_CACHE_ST_F )
+        return NEON_CACHE_ST_F;
+    return NEON_CACHE_ST_D;
+}
+
+int isPred(dynarec_arm_t* dyn, int ninst, int pred) {
+    for(int i=0; i<dyn->insts[ninst].pred_sz; ++i)
+        if(dyn->insts[ninst].pred[i]==pred)
+            return pred;
+    return -1;
+}
+int getNominalPred(dynarec_arm_t* dyn, int ninst) {
+    if((ninst<=0) || !dyn->insts[ninst].pred_sz)
+        return -1;
+    if(isPred(dyn, ninst, ninst-1)!=-1)
+        return ninst-1;
+    return dyn->insts[ninst].pred[0];
+}
+
+int isCacheEmpty(dynarec_arm_t* dyn, int ninst) {
+    if(dyn->insts[ninst].n.stack_next) {
+        return 0;
+    }
+    for(int i=0; i<24; ++i)
+        if(dyn->insts[ninst].n.neoncache[i].v) {       // there is something at ninst for i
+            if(!(
+            (dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_F || dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_D)
+            && dyn->insts[ninst].n.neoncache[i].n<dyn->insts[ninst].n.stack_pop))
+                return 0;
+        }
+    return 1;
+
+}
+
+static int fpuCacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
+    int i2 = dyn->insts[ninst].x64.jmp_insts;
+    if(i2<0)
+        return 1;
+    if((dyn->insts[i2].x64.barrier&BARRIER_FLOAT))
+        // if the barrier as already been apply, no transform needed
+        return ((dyn->insts[ninst].x64.barrier&BARRIER_FLOAT))?0:(isCacheEmpty(dyn, ninst)?0:1);
+    int ret = 0;
+    if(!i2) { // just purge
+        if(dyn->insts[ninst].n.stack_next) {
+            return 1;
+        }
+        for(int i=0; i<24 && !ret; ++i)
+            if(dyn->insts[ninst].n.neoncache[i].v) {       // there is something at ninst for i
+                if(!(
+                (dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_F || dyn->insts[ninst].n.neoncache[i].t==NEON_CACHE_ST_D)
+                && dyn->insts[ninst].n.neoncache[i].n<dyn->insts[ninst].n.stack_pop))
+                    ret = 1;
+            }
+        return ret;
+    }
+    // Check if ninst can be compatible to i2
+    if(dyn->insts[ninst].n.stack_next != dyn->insts[i2].n.stack-dyn->insts[i2].n.stack_push) {
+        return 1;
+    }
+    neoncache_t cache_i2 = dyn->insts[i2].n;
+    neoncacheUnwind(&cache_i2);
+
+    for(int i=0; i<24; ++i) {
+        if(dyn->insts[ninst].n.neoncache[i].v) {       // there is something at ninst for i
+            if(!cache_i2.neoncache[i].v) {    // but there is nothing at i2 for i
+                ret = 1;
+            } else if(dyn->insts[ninst].n.neoncache[i].v!=cache_i2.neoncache[i].v) {  // there is something different
+                if(dyn->insts[ninst].n.neoncache[i].n!=cache_i2.neoncache[i].n) {   // not the same x64 reg
+                    ret = 1;
+                }
+                else if(dyn->insts[ninst].n.neoncache[i].t == NEON_CACHE_XMMR && cache_i2.neoncache[i].t == NEON_CACHE_XMMW)
+                    {/* nothing */ }
+                else
+                    ret = 1;
+            }
+        } else if(cache_i2.neoncache[i].v)
+            ret = 1;
+    }
+    return ret;
+}
+
+static int flagsCacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
+    int jmp = dyn->insts[ninst].x64.jmp_insts;
+    if(jmp<0)
+        return 0;
+    if(dyn->insts[ninst].f_exit.dfnone)  // flags are fully known, nothing we can do more
+        return 0;
+/*    if((dyn->f.pending!=SF_SET)
+    && (dyn->f.pending!=SF_SET_PENDING)) {
+        if(dyn->f.pending!=SF_PENDING) {*/
+    switch (dyn->insts[jmp].f_entry.pending) {
+        case SF_UNKNOWN: return 0;
+        case SF_SET: 
+            if(dyn->insts[ninst].f_exit.pending!=SF_SET && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING) 
+                return 1; 
+            else 
+                return 0;
+        case SF_SET_PENDING:
+            if(dyn->insts[ninst].f_exit.pending!=SF_SET 
+            && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING
+            && dyn->insts[ninst].f_exit.pending!=SF_PENDING) 
+                return 1; 
+            else 
+                return 0;
+        case SF_PENDING:
+            if(dyn->insts[ninst].f_exit.pending!=SF_SET 
+            && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING
+            && dyn->insts[ninst].f_exit.pending!=SF_PENDING)
+                return 1;
+            else
+                return (dyn->insts[jmp].f_entry.dfnone  == dyn->insts[ninst].f_exit.dfnone)?0:1;
+    }
+    if(dyn->insts[jmp].f_entry.dfnone && !dyn->insts[ninst].f_exit.dfnone)
+        return 1;
+    return 0;
+}
+int CacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
+    int ret = 0;
+    if (fpuCacheNeedsTransform(dyn, ninst)) ret|=1;
+    if (flagsCacheNeedsTransform(dyn, ninst)) ret|=2;
+    return ret;
+}
+
+void neoncacheUnwind(neoncache_t* cache)
+{
+    if(cache->swapped) {
+        // unswap
+        int a = -1; 
+        int b = -1;
+        for(int j=0; j<24 && ((a==-1) || (b==-1)); ++j)
+            if((cache->neoncache[j].t == NEON_CACHE_ST_D || cache->neoncache[j].t == NEON_CACHE_ST_F)) {
+                if(cache->neoncache[j].n == cache->combined1)
+                    a = j;
+                else if(cache->neoncache[j].n == cache->combined2)
+                    b = j;
+            }
+        if(a!=-1 && b!=-1) {
+            int tmp = cache->neoncache[a].n;
+            cache->neoncache[a].n = cache->neoncache[b].n;
+            cache->neoncache[b].n = tmp;
+        }
+        cache->swapped = 0;
+        cache->combined1 = cache->combined2 = 0;
+    }
+    if(cache->news) {
+        // reove the newly created neoncache
+        for(int i=0; i<24; ++i)
+            if(cache->news&(1<<i))
+                cache->neoncache[i].v = 0;
+        cache->news = 0;
+    }
+    if(cache->stack_push) {
+        // unpush
+        for(int j=0; j<24; ++j) {
+            if((cache->neoncache[j].t == NEON_CACHE_ST_D || cache->neoncache[j].t == NEON_CACHE_ST_F)) {
+                if(cache->neoncache[j].n<cache->stack_push)
+                    cache->neoncache[j].v = 0;
+                else
+                    cache->neoncache[j].n-=cache->stack_push;
+            }
+        }
+        cache->x87stack-=cache->stack_push;
+        cache->stack-=cache->stack_push;
+        cache->stack_push = 0;
+    }
+    cache->x87stack+=cache->stack_pop;
+    cache->stack_next = cache->stack;
+    cache->stack_pop = 0;
+    cache->barrier = 0;
+    // And now, rebuild the x87cache info with neoncache
+    cache->mmxcount = 0;
+    cache->fpu_scratch = 0;
+    cache->fpu_extra_qscratch = 0;
+    cache->fpu_reg = 0;
+    for(int i=0; i<8; ++i) {
+        cache->x87cache[i] = -1;
+        cache->mmxcache[i] = -1;
+        cache->x87reg[i] = 0;
+        cache->ssecache[i*2].v = -1;
+        cache->ssecache[i*2+1].v = -1;
+    }
+    int x87reg = 0;
+    for(int i=0; i<24; ++i) {
+        if(cache->neoncache[i].v) {
+            cache->fpuused[i] = 1;
+            switch (cache->neoncache[i].t) {
+                case NEON_CACHE_MM:
+                    cache->mmxcache[cache->neoncache[i].n] = i;
+                    ++cache->mmxcount;
+                    ++cache->fpu_reg;
+                    break;
+                case NEON_CACHE_XMMR:
+                case NEON_CACHE_XMMW:
+                    cache->ssecache[cache->neoncache[i].n].reg = i;
+                    cache->ssecache[cache->neoncache[i].n].write = (cache->neoncache[i].t==NEON_CACHE_XMMW)?1:0;
+                    ++cache->fpu_reg;
+                    break;
+                case NEON_CACHE_ST_F:
+                case NEON_CACHE_ST_D:
+                    cache->x87cache[x87reg] = cache->neoncache[i].n;
+                    cache->x87reg[x87reg] = i;
+                    ++x87reg;
+                    ++cache->fpu_reg;
+                    break;
+                case NEON_CACHE_SCR:
+                    cache->fpuused[i] = 0;
+                    cache->neoncache[i].v = 0;
+                    break;
+            }
+        } else {
+            cache->fpuused[i] = 0;
+        }
+    }
+}
+
+#define F8      *(uint8_t*)(addr++)
+#define F32     *(uint32_t*)(addr+=4, addr-4)
+#define F32S64  (uint64_t)(int64_t)*(int32_t*)(addr+=4, addr-4)
+// Get if ED will have the correct parity. Not emiting anything. Parity is 2 for DWORD or 3 for QWORD
+int getedparity(dynarec_arm_t* dyn, int ninst, uintptr_t addr, uint8_t nextop, int parity, int delta)
+{
+    (void)dyn; (void)ninst;
+
+    uint32_t tested = (1<<parity)-1;
+    if((nextop&0xC0)==0xC0)
+        return 0;   // direct register, no parity...
+    if(!(nextop&0xC0)) {
+        if((nextop&7)==4) {
+            uint8_t sib = F8;
+            int sib_reg = (sib>>3)&7;
+            if((sib&0x7)==5) {
+                uint64_t tmp = F32S64;
+                if (sib_reg!=4) {
+                    // if XXXXXX+reg<<N then check parity of XXXXX and N should be enough
+                    return ((tmp&tested)==0 && (sib>>6)>=parity)?1:0;
+                } else {
+                    // just a constant...
+                    return (tmp&tested)?0:1;
+                }
+            } else {
+                if(sib_reg==4 && parity<3)
+                    return 0;   // simple [reg]
+                // don't try [reg1 + reg2<<N], unless reg1 is ESP
+                return ((sib&0x7)==4 && (sib>>6)>=parity)?1:0;
+            }
+        } else if((nextop&7)==5) {
+            uint64_t tmp = F32S64;
+            tmp+=addr+delta;
+            return (tmp&tested)?0:1;
+        } else {
+            return 0;
+        }
+    } else {
+        return 0; //Form [reg1 + reg2<<N + XXXXXX]
+    }
+}
+
+// Do the GETED, but don't emit anything...
+uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop) 
+{
+    (void)dyn; (void)addr; (void)ninst;
+
+    if((nextop&0xC0)==0xC0)
+        return addr;
+    if(!(nextop&0xC0)) {
+        if((nextop&7)==4) {
+            uint8_t sib = F8;
+            if((sib&0x7)==5) {
+                addr+=4;
+            }
+        } else if((nextop&7)==5) {
+            addr+=4;
+        }
+    } else {
+        if((nextop&7)==4) {
+            ++addr;
+        }
+        if(nextop&0x80) {
+            addr+=4;
+        } else {
+            ++addr;
+        }
+    }
+    return addr;
+}
+#undef F8
+#undef F32
+
+int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn)
+{
+    (void)dyn;
+
+#define PK(a)       *(uint8_t*)(addr+a)
+#define PK32(a)     *(int32_t*)(addr+a)
+
+    if(!addr || !getProtection(addr))
+        return 0;
+    if(PK(0)==0xff && PK(1)==0x25) {            // "absolute" jump, maybe the GOT (well, RIP relative in fact)
+        uintptr_t a1 = addr + 6 + (PK32(2));    // need to add a check to see if the address is from the GOT !
+        addr = (uintptr_t)getAlternate(*(void**)a1);
+    }
+    if(!addr || !getProtection(addr))
+        return 0;
+    onebridge_t *b = (onebridge_t*)(addr);
+    if(b->CC==0xCC && b->S=='S' && b->C=='C' && b->w!=(wrapper_t)0 && b->f!=(uintptr_t)PltResolver) {
+        // found !
+        if(retn) *retn = (b->C3==0xC2)?b->N:0;
+        if(calladdress) *calladdress = addr+1;
+        return 1;
+    }
+    return 0;
+#undef PK32
+#undef PK
+}
+
+const char* getCacheName(int t, int n)
+{
+    static char buff[20];
+    switch(t) {
+        case NEON_CACHE_ST_D: sprintf(buff, "ST%d", n); break;
+        case NEON_CACHE_ST_F: sprintf(buff, "st%d", n); break;
+        case NEON_CACHE_MM: sprintf(buff, "MM%d", n); break;
+        case NEON_CACHE_XMMW: sprintf(buff, "XMM%d", n); break;
+        case NEON_CACHE_XMMR: sprintf(buff, "xmm%d", n); break;
+        case NEON_CACHE_SCR: sprintf(buff, "Scratch"); break;
+        case NEON_CACHE_NONE: buff[0]='\0'; break;
+    }
+    return buff;
+}
+
+// is inst clean for a son branch?
+int isInstClean(dynarec_arm_t* dyn, int ninst)
+{
+    // check flags cache
+    if(dyn->insts[ninst].f_entry.dfnone || dyn->insts[ninst].f_entry.pending)
+        return 0;
+    if(dyn->insts[ninst].x64.state_flags)
+        return 0;
+    // check neoncache
+    neoncache_t* n = &dyn->insts[ninst].n;
+    if(n->news || n->stack || n->stack_next)
+        return 0;
+    for(int i=0; i<24; ++i)
+        if(n->neoncache[i].v)
+            return 0;
+    return 1;
+}
@@ -34,25 +34,56 @@ void arm_aese(x64emu_t* emu, int xmm);
 void arm_aesdlast(x64emu_t* emu, int xmm);
 void arm_aeselast(x64emu_t* emu, int xmm);
 void arm_aesimc(x64emu_t* emu, int xmm);
+void arm_aeskeygenassist(x64emu_t* emu, int gx, int ex, void* p, uint32_t u8);
+void arm_pclmul(x64emu_t* emu, int gx, int ex, void* p, uint32_t u8);

+void arm_clflush(x64emu_t* emu, void* p);

 void arm_ud(x64emu_t* emu);
+void arm_priv(x64emu_t* emu);

 // Get an FPU scratch reg
 int fpu_get_scratch(dynarec_arm_t* dyn);
 // Reset scratch regs counter
 void fpu_reset_scratch(dynarec_arm_t* dyn);
 // Get an x87 double reg
-int fpu_get_reg_x87(dynarec_arm_t* dyn);
+int fpu_get_reg_x87(dynarec_arm_t* dyn, int t, int n);
 // Get an MMX double reg
 int fpu_get_reg_emm(dynarec_arm_t* dyn, int emm);
 // Get an XMM quad reg
-int fpu_get_reg_xmm(dynarec_arm_t* dyn, int xmm);
+int fpu_get_reg_xmm(dynarec_arm_t* dyn, int t, int xmm);
 // Free a FPU/MMX/XMM reg
 void fpu_free_reg(dynarec_arm_t* dyn, int reg);
 // Reset fpu regs counter
 void fpu_reset_reg(dynarec_arm_t* dyn);

+// ---- Neon cache functions
+// Get type for STx
+int neoncache_get_st(dynarec_arm_t* dyn, int ninst, int a);
+// Get if STx is FLOAT or DOUBLE
+int neoncache_get_st_f(dynarec_arm_t* dyn, int ninst, int a);
+// Get actual type for STx
+int neoncache_get_current_st(dynarec_arm_t* dyn, int ninst, int a);
+// Get actual STx is FLOAT or DOUBLE
+int neoncache_get_current_st_f(dynarec_arm_t* dyn, int a);
+// Back-propagate a change float->double
+void neoncache_promote_double(dynarec_arm_t* dyn, int ninst, int a);
+// Combine and propagate if needed (pass 1 only)
+int neoncache_combine_st(dynarec_arm_t* dyn, int ninst, int a, int b);  // with stack current dyn->n_stack*
+
+// FPU Cache transformation (for loops)
+int CacheNeedsTransform(dynarec_arm_t* dyn, int i1);
+
+// Undo the changes of a neoncache to get the status before the instruction
+void neoncacheUnwind(neoncache_t* cache);
+
+// is inst clean for a son branch?
+int isInstClean(dynarec_arm_t* dyn, int ninst);
+
+// predecessor access
+int isPred(dynarec_arm_t* dyn, int ninst, int pred);
+int getNominalPred(dynarec_arm_t* dyn, int ninst);
+
 // Get if ED will have the correct parity. Not emiting anything. Parity is 2 for DWORD or 3 for QWORD
 int getedparity(dynarec_arm_t* dyn, int ninst, uintptr_t addr, uint8_t nextop, int parity, int delta);
 // Do the GETED, but don't emit anything...
@@ -61,4 +92,6 @@ uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop);
 // Is what pointed at addr a native call? And if yes, to what function?
 int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn);

+const char* getCacheName(int t, int n);
+
 #endif //__DYNAREC_ARM_FUNCTIONS_H__
@@ -32,6 +32,9 @@
 #define PK64(a)   *(uint64_t*)(addr+a)
 #define PKip(a)   *(uint8_t*)(ip+a)

+//LOCK_* define
+#define LOCK_LOCK   (int*)1
+
 // GETGD    get x64 register in gd
 #define GETGD   gd = xRAX+((nextop&0x38)>>3)+(rex.r<<3)
 //GETED can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI
@@ -39,7 +42,7 @@
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
                    LDRxw_U12(x1, wback, fixedaddress); \
                    ed = x1;                            \
                }
@@ -47,7 +50,7 @@
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<3, 7, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<3, 7, rex, NULL, 0, D); \
                    LDRx_U12(x1, wback, fixedaddress);  \
                    ed = x1;                            \
                }
@@ -55,7 +58,7 @@
                    ed = xEAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL,0, D); \
                    LDRw_U12(x1, wback, fixedaddress);  \
                    ed = x1;                            \
                }
@@ -65,7 +68,25 @@
                    wb = x1;                            \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, D); \
+                    LDRSW_U12(x1, wback, fixedaddress); \
+                    wb = ed = x1;                       \
+                }
+#define GETED32(D)  if(MODREG) {                        \
+                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
+                    wback = 0;                          \
+                } else {                                \
+                    addr = geted32(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
+                    LDRxw_U12(x1, wback, fixedaddress); \
+                    ed = x1;                            \
+                }
+#define GETSED32w(D)  if((nextop&0xC0)==0xC0) {         \
+                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
+                    SXTWx(x1, ed);                      \
+                    wb = x1;                            \
+                    wback = 0;                          \
+                } else {                                \
+                    addr = geted32(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0xfff<<2, 3, rex, NULL, 0, D); \
                    LDRSW_U12(x1, wback, fixedaddress); \
                    wb = ed = x1;                       \
                }
@@ -74,17 +95,34 @@
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, (hint==x2)?x1:x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, (hint==x2)?x1:x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
+                    LDRxw_U12(hint, wback, fixedaddress); \
+                    ed = hint;                            \
+                }
+#define GETED32H(hint, D) if(MODREG) {                  \
+                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
+                    wback = 0;                          \
+                } else {                                \
+                    addr = geted32(dyn, addr, ninst, nextop, &wback, (hint==x2)?x1:x2, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
                    LDRxw_U12(hint, wback, fixedaddress); \
                    ed = hint;                            \
                }
 //GETEDW can use hint for wback and ret for ed. wback is 0 if ed is xEAX..xEDI
-#define GETEDW(hint, ret, D)   if(MODREG) {              \
+#define GETEDW(hint, ret, D)   if(MODREG) {             \
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    MOVxw_REG(ret, ed);                 \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, hint, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, hint, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
+                    ed = ret;                           \
+                    LDRxw_U12(ed, wback, fixedaddress); \
+                }
+#define GETED32W(hint, ret, D)   if(MODREG) {           \
+                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
+                    MOVxw_REG(ret, ed);                 \
+                    wback = 0;                          \
+                } else {                                \
+                    addr = geted32(dyn, addr, ninst, nextop, &wback, hint, &fixedaddress, 0xfff<<(2+rex.w), (1<<(2+rex.w))-1, rex, NULL, 0, D); \
                    ed = ret;                           \
                    LDRxw_U12(ed, wback, fixedaddress); \
                }
@@ -94,14 +132,12 @@
 #define WBACKx      if(wback) {STRx_U12(ed, wback, fixedaddress);}
 // Write back ed in wback (if wback not 0)
 #define WBACKw      if(wback) {STRw_U12(ed, wback, fixedaddress);}
-// Send back wb to either ed or wback
-#define SBACK(wb)   if(wback) {STRxw(wb, wback, fixedaddress);} else {MOVxw_REG(ed, wb);}
 //GETEDO can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI
 #define GETEDO(O, D)   if(MODREG) {                     \
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, NULL, 0, D); \
                    LDRxw_REG(x1, wback, O);            \
                    ed = x1;                            \
                }
@@ -111,7 +147,7 @@
                    ed = xRAX+(nextop&7)+(rex.b<<3);    \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, NULL, 0, D); \
                    LDRx_REG(x1, wback, O);             \
                    ed = x1;                            \
                }
@@ -121,7 +157,7 @@
                    wb = x1;                            \
                    wback = 0;                          \
                } else {                                \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, rex, NULL, 0, D); \
                    LDRSW_REG(x1, wback, O);            \
                    wb = ed = x1;                       \
                }
@@ -140,7 +176,7 @@
                    ed = i;                 \
                    wb1 = 0;                \
                } else {                    \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, w, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, w, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, NULL, 0, D); \
                    LDRH_U12(i, wback, fixedaddress); \
                    ed = i;                 \
                    wb1 = 1;                \
@@ -152,7 +188,7 @@
                    ed = i;                 \
                    wb1 = 0;                \
                } else {                    \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, NULL, 0, D); \
                    LDRH_U12(i, wback, fixedaddress); \
                    ed = i;                 \
                    wb1 = 1;                \
@@ -164,7 +200,7 @@
                    ed = i;                 \
                    wb1 = 0;                \
                } else {                    \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff<<1, (1<<1)-1, rex, NULL, 0, D); \
                    LDRSHx_U12(i, wback, fixedaddress);\
                    ed = i;                 \
                    wb1 = 1;                \
@@ -189,7 +225,7 @@
                    wb1 = 0;                    \
                    ed = i;                     \
                } else {                        \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, NULL, 0, D); \
                    LDRB_U12(i, wback, fixedaddress); \
                    wb1 = 1;                    \
                    ed = i;                     \
@@ -208,8 +244,9 @@
                    wb1 = 0;                    \
                    ed = i;                     \
                } else {                        \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, 0, D); \
-                    ADDx_REG(wback, wback, i);  \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, rex, NULL, 0, D); \
+                    ADDx_REG(x3, wback, i);     \
+                    if(wback!=x3) wback = x3;   \
                    LDRB_U12(i, wback, fixedaddress);      \
                    wb1 = 1;                    \
                    ed = i;                     \
@@ -228,11 +265,30 @@
                    wb1 = 0;                    \
                    ed = i;                     \
                } else {                        \
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, 0, D); \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, NULL, 0, D); \
                    LDRSBx_U12(i, wback, fixedaddress); \
                    wb1 = 1;                    \
                    ed = i;                     \
                }
+//GETEB will use i for ed, and can use r3 for wback.
+#define GETEB32(i, D) if(MODREG) {              \
+                    if(rex.rex) {               \
+                        wback = xRAX+(nextop&7)+(rex.b<<3);     \
+                        wb2 = 0;                \
+                    } else {                    \
+                        wback = (nextop&7);     \
+                        wb2 = (wback>>2)*8;     \
+                        wback = xRAX+(wback&3); \
+                    }                           \
+                    UBFXx(i, wback, wb2, 8);    \
+                    wb1 = 0;                    \
+                    ed = i;                     \
+                } else {                        \
+                    addr = geted32(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0xfff, 0, rex, NULL, 0, D); \
+                    LDRB_U12(i, wback, fixedaddress); \
+                    wb1 = 1;                    \
+                    ed = i;                     \
+                }
 // Write eb (ed) back to original register / memory
 #define EBBACK   if(wb1) {STRB_U12(ed, wback, fixedaddress);} else {BFIx(wback, ed, wb2, 8);}
 //GETGB will use i for gd
@@ -277,129 +333,153 @@
 // R0 will not be pushed/popd if ret is -2. Flags are not save/restored
 #define CALL_S(F, ret) call_c(dyn, ninst, F, x7, ret, 0, 0)

-#define MARK    if(dyn->insts) {dyn->insts[ninst].mark = dyn->arm_size;}
-#define GETMARK ((dyn->insts)?dyn->insts[ninst].mark:(dyn->arm_size+4))
-#define MARK2   if(dyn->insts) {dyn->insts[ninst].mark2 = dyn->arm_size;}
-#define GETMARK2 ((dyn->insts)?dyn->insts[ninst].mark2:(dyn->arm_size+4))
-#define MARK3   if(dyn->insts) {dyn->insts[ninst].mark3 = dyn->arm_size;}
-#define GETMARK3 ((dyn->insts)?dyn->insts[ninst].mark3:(dyn->arm_size+4))
-#define MARKF   if(dyn->insts) {dyn->insts[ninst].markf = dyn->arm_size;}
-#define GETMARKF ((dyn->insts)?dyn->insts[ninst].markf:(dyn->arm_size+4))
-#define MARKSEG if(dyn->insts) {dyn->insts[ninst].markseg = dyn->arm_size;}
-#define GETMARKSEG ((dyn->insts)?dyn->insts[ninst].markseg:(dyn->arm_size+4))
-#define MARKLOCK if(dyn->insts) {dyn->insts[ninst].marklock = dyn->arm_size;}
-#define GETMARKLOCK ((dyn->insts)?dyn->insts[ninst].marklock:(dyn->arm_size+4))
+#define MARK    dyn->insts[ninst].mark = dyn->native_size
+#define GETMARK dyn->insts[ninst].mark
+#define MARK2   dyn->insts[ninst].mark2 = dyn->native_size
+#define GETMARK2 dyn->insts[ninst].mark2
+#define MARK3   dyn->insts[ninst].mark3 = dyn->native_size
+#define GETMARK3 dyn->insts[ninst].mark3
+#define MARKF   dyn->insts[ninst].markf = dyn->native_size
+#define GETMARKF dyn->insts[ninst].markf
+#define MARKSEG dyn->insts[ninst].markseg = dyn->native_size
+#define GETMARKSEG dyn->insts[ninst].markseg
+#define MARKLOCK dyn->insts[ninst].marklock = dyn->native_size
+#define GETMARKLOCK dyn->insts[ninst].marklock

 // Branch to MARK if cond (use j64)
 #define B_MARK(cond)                \
-    j64 = GETMARK-(dyn->arm_size);  \
+    j64 = GETMARK-(dyn->native_size);  \
    Bcond(cond, j64)
 // Branch to MARK unconditionnal (use j64)
 #define B_MARK_nocond               \
-    j64 = GETMARK-(dyn->arm_size);  \
+    j64 = GETMARK-(dyn->native_size);  \
    B(j64)
 // Branch to MARK if reg is 0 (use j64)
 #define CBZxw_MARK(reg)             \
-    j64 = GETMARK-(dyn->arm_size);  \
+    j64 = GETMARK-(dyn->native_size);  \
    CBZxw(reg, j64)
 // Branch to MARK if reg is not 0 (use j64)
 #define CBNZx_MARK(reg)             \
-    j64 = GETMARK-(dyn->arm_size);  \
+    j64 = GETMARK-(dyn->native_size);  \
    CBNZx(reg, j64)
 // Branch to MARK if reg is not 0 (use j64)
 #define CBNZw_MARK(reg)             \
-    j64 = GETMARK-(dyn->arm_size);  \
+    j64 = GETMARK-(dyn->native_size);  \
    CBNZw(reg, j64)
+// Test bit N of A and branch to MARK if not set
+#define TBZ_MARK(A, N)              \
+    j64 = GETMARK-(dyn->native_size);  \
+    TBZ(A, N, j64)
+// Test bit N of A and branch to MARK if set
+#define TBNZ_MARK(A, N)             \
+    j64 = GETMARK-(dyn->native_size);  \
+    TBNZ(A, N, j64)
 // Branch to MARK2 if cond (use j64)
 #define B_MARK2(cond)               \
-    j64 = GETMARK2-(dyn->arm_size); \
+    j64 = GETMARK2-(dyn->native_size); \
    Bcond(cond, j64)
 // Branch to MARK2 unconditionnal (use j64)
 #define B_MARK2_nocond              \
-    j64 = GETMARK2-(dyn->arm_size); \
+    j64 = GETMARK2-(dyn->native_size); \
    B(j64)
 // Branch to MARK2 if reg is not 0 (use j64)
 #define CBNZx_MARK2(reg)            \
-    j64 = GETMARK2-(dyn->arm_size); \
+    j64 = GETMARK2-(dyn->native_size); \
    CBNZx(reg, j64)
 // Test bit N of A and branch to MARK2 if set
 #define TBNZ_MARK2(A, N)            \
-    j64 = GETMARK2-(dyn->arm_size); \
+    j64 = GETMARK2-(dyn->native_size); \
    TBNZ(A, N, j64)
 // Branch to MARK3 if cond (use j64)
 #define B_MARK3(cond)               \
-    j64 = GETMARK3-(dyn->arm_size); \
+    j64 = GETMARK3-(dyn->native_size); \
    Bcond(cond, j64)
+// Test bit N of A and branch to MARK3 if not set
+#define TBZ_MARK2(A, N)            \
+    j64 = GETMARK2-(dyn->native_size); \
+    TBZ(A, N, j64)
 // Branch to MARK3 unconditionnal (use j64)
 #define B_MARK3_nocond              \
-    j64 = GETMARK3-(dyn->arm_size); \
+    j64 = GETMARK3-(dyn->native_size); \
    B(j64)
 // Branch to MARK3 if reg is not 0 (use j64)
 #define CBNZx_MARK3(reg)            \
-    j64 = GETMARK3-(dyn->arm_size); \
+    j64 = GETMARK3-(dyn->native_size); \
    CBNZx(reg, j64)
-// Test bit N of A and branch to MARK3 if set
-#define TBNZ_MARK3(A, N)            \
-    j64 = GETMARK3-(dyn->arm_size); \
-    TBNZ(A, N, j64)
+// Branch to MARK3 if reg is not 0 (use j64)
+#define CBNZw_MARK3(reg)            \
+    j64 = GETMARK3-(dyn->native_size); \
+    CBNZw(reg, j64)
+// Branch to MARK3 if reg is 0 (use j64)
+#define CBZx_MARK3(reg)             \
+    j64 = GETMARK3-(dyn->native_size); \
+    CBZx(reg, j64)
+// Branch to MARK3 if reg is 0 (use j64)
+#define CBZw_MARK3(reg)             \
+    j64 = GETMARK3-(dyn->native_size); \
+    CBZw(reg, j64)
 // Test bit N of A and branch to MARK3 if not set
 #define TBZ_MARK3(A, N)             \
-    j64 = GETMARK3-(dyn->arm_size); \
+    j64 = GETMARK3-(dyn->native_size); \
    TBZ(A, N, j64)
+// Test bit N of A and branch to MARK3 if set
+#define TBNZ_MARK3(A, N)            \
+    j64 = GETMARK3-(dyn->native_size); \
+    TBNZ(A, N, j64)
 // Branch to next instruction if cond (use j64)
 #define B_NEXT(cond)     \
-    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    Bcond(cond, j64)
 // Branch to next instruction unconditionnal (use j64)
 #define B_NEXT_nocond                                               \
-    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0;\
+    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0;\
    B(j64)
 // Branch to next instruction if reg is 0 (use j64)
 #define CBZw_NEXT(reg)    \
-    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    CBZw(reg, j64)
 // Branch to next instruction if reg is 0 (use j64)
 #define CBZx_NEXT(reg)    \
-    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    CBZx(reg, j64)
 // Branch to next instruction if reg is not 0 (use j64)
 #define CBNZx_NEXT(reg)   \
-    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 =  (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    CBNZx(reg, j64)
 // Test bit N of A and branch to next instruction if not set
 #define TBZ_NEXT(A, N)              \
-    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    TBZ(A, N, j64)
 // Test bit N of A and branch to next instruction if set
 #define TBNZ_NEXT(A, N)             \
-    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size)):0; \
+    j64 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->native_size)):0; \
    TBNZ(A, N, j64)
 // Branch to MARKSEG if cond (use j64)
 #define B_MARKSEG(cond)    \
-    j64 = GETMARKSEG-(dyn->arm_size);   \
+    j64 = GETMARKSEG-(dyn->native_size);   \
    Bcond(cond, j64)
 // Branch to MARKSEG if reg is 0 (use j64)
 #define CBZw_MARKSEG(reg)    \
-    j64 = GETMARKSEG-(dyn->arm_size);   \
+    j64 = GETMARKSEG-(dyn->native_size);   \
    CBZw(reg, j64)
 // Branch to MARKSEG if reg is not 0 (use j64)
 #define CBNZw_MARKSEG(reg)              \
-    j64 = GETMARKSEG-(dyn->arm_size);   \
+    j64 = GETMARKSEG-(dyn->native_size);   \
    CBNZw(reg, j64)
 // Branch to MARKLOCK if cond (use j64)
 #define B_MARKLOCK(cond)    \
-    j64 = GETMARKLOCK-(dyn->arm_size);   \
+    j64 = GETMARKLOCK-(dyn->native_size);   \
    Bcond(cond, j64)
 // Branch to MARKLOCK if reg is not 0 (use j64)
 #define CBNZx_MARKLOCK(reg)             \
-    j64 = GETMARKLOCK-(dyn->arm_size);  \
+    j64 = GETMARKLOCK-(dyn->native_size);  \
    CBNZx(reg, j64)

-#define IFX(A)  if(dyn->insts && (dyn->insts[ninst].x64.need_flags&(A)))
-#define IFX_PENDOR0  if(dyn->insts && (dyn->insts[ninst].x64.need_flags&(X_PEND) || !dyn->insts[ninst].x64.need_flags))
-#define IFXX(A) if(dyn->insts && (dyn->insts[ninst].x64.need_flags==(A)))
-#define IFX2X(A, B) if(dyn->insts && (dyn->insts[ninst].x64.need_flags==(A) || dyn->insts[ninst].x64.need_flags==(B) || dyn->insts[ninst].x64.need_flags==((A)|(B))))
-#define IFXN(A, B)  if(dyn->insts && (dyn->insts[ninst].x64.need_flags&(A) && !(dyn->insts[ninst].x64.need_flags&(B))))
+#define IFX(A)  if((dyn->insts[ninst].x64.gen_flags&(A)))
+#define IFX_PENDOR0  if((dyn->insts[ninst].x64.gen_flags&(X_PEND) || !dyn->insts[ninst].x64.gen_flags))
+#define IFXX(A) if((dyn->insts[ninst].x64.gen_flags==(A)))
+#define IFX2X(A, B) if((dyn->insts[ninst].x64.gen_flags==(A) || dyn->insts[ninst].x64.gen_flags==(B) || dyn->insts[ninst].x64.gen_flags==((A)|(B))))
+#define IFXN(A, B)  if((dyn->insts[ninst].x64.gen_flags&(A) && !(dyn->insts[ninst].x64.gen_flags&(B))))

 // Generate FCOM with s1 and s2 scratch regs (the VCMP is already done)
 #define FCOM(s1, s2, s3)                                                    \
@@ -417,9 +497,19 @@

 // Generate FCOMI with s1 and s2 scratch regs (the VCMP is already done)
 #define FCOMI(s1, s2)    \
-    IFX(X_CF|X_PF|X_ZF|X_PEND) {                                            \
-        MOV32w(s2, 0b01000101);                                             \
+    IFX(X_OF|X_AF|X_SF|X_PEND) {                                            \
+        MOV32w(s2, 0b100011010101);                                         \
        BICw_REG(xFlags, xFlags, s2);                                       \
+        IFX(X_CF|X_PF|X_ZF|X_PEND) {                                        \
+            MOV32w(s2, 0b01000101);                                         \
+        }                                                                   \
+    } else {                                                                \
+        IFX(X_CF|X_PF|X_ZF|X_PEND) {                                        \
+            MOV32w(s2, 0b01000101);                                         \
+            BICw_REG(xFlags, xFlags, s2);                                   \
+        }                                                                   \
+    }                                                                       \
+    IFX(X_CF|X_PF|X_ZF|X_PEND) {                                            \
        CSETw(s1, cMI); /* 1 if less than, 0 else */                        \
        /*s2 already set */     /* unordered */                             \
        CSELw(s1, s2, s1, cVS);                                             \
@@ -429,60 +519,43 @@
        ORRw_REG(xFlags, xFlags, s1);                                       \
    }                                                                       \
    SET_DFNONE(s1);                                                         \
-    IFX(X_OF|X_PEND) {                                                      \
-        BFCw(xFlags, F_OF, 1);                                              \
-    }                                                                       \
-    IFX(X_AF|X_PEND) {                                                      \
-        BFCw(xFlags, F_AF, 1);                                              \
-    }                                                                       \
-    IFX(X_SF|X_PEND) {                                                      \
-        BFCw(xFlags, F_SF, 1);                                              \
-    }                                                                       \


 #define STORE_REG(A)    STRx_U12(x##A, xEmu, offsetof(x64emu_t, regs[_##A]))
 #define STP_REGS(A, B)  STPx_S7_offset(x##A, x##B, xEmu, offsetof(x64emu_t, regs[_##A]))
 #define LDP_REGS(A, B)  LDPx_S7_offset(x##A, x##B, xEmu, offsetof(x64emu_t, regs[_##A]))
 #define STORE_XEMU_REGS(A)  \
-    STORE_REG(RAX);         \
-    STORE_REG(RCX);         \
-    STORE_REG(RDX);         \
-    STORE_REG(RBX);         \
-    STORE_REG(RSP);         \
-    STORE_REG(RBP);         \
-    STORE_REG(RSI);         \
-    STORE_REG(RDI);         \
-    STORE_REG(R8);          \
-    STORE_REG(R9);          \
-    STORE_REG(R10);         \
-    STORE_REG(R11);         \
-    STORE_REG(R12);         \
-    STORE_REG(R13);         \
-    STORE_REG(R14);         \
-    STORE_REG(R15);         \
-    STRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
-    if(A) {STRx_U12(A, xEmu, offsetof(x64emu_t, ip));}
+    STP_REGS(RAX, RCX);     \
+    STP_REGS(RDX, RBX);     \
+    STP_REGS(RSP, RBP);     \
+    STP_REGS(RSI, RDI);     \
+    STP_REGS(R8, R9);       \
+    STP_REGS(R10, R11);     \
+    STP_REGS(R12, R13);     \
+    STP_REGS(R14, R15);     \
+    if(A==xRIP) {           \
+        STPx_S7_offset(xFlags, xRIP, xEmu, offsetof(x64emu_t, eflags)); \
+    } else {                \
+        STRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
+        if(A) {STRx_U12(A, xEmu, offsetof(x64emu_t, ip));}  \
+    }

 #define LOAD_REG(A)    LDRx_U12(x##A, xEmu, offsetof(x64emu_t, regs[_##A]))
-#define LOAD_XEMU_REGS(A)  \
-    LOAD_REG(RAX);         \
-    LOAD_REG(RCX);         \
-    LOAD_REG(RDX);         \
-    LOAD_REG(RBX);         \
-    LOAD_REG(RSP);         \
-    LOAD_REG(RBP);         \
-    LOAD_REG(RSI);         \
-    LOAD_REG(RDI);         \
-    LOAD_REG(R8);          \
-    LOAD_REG(R9);          \
-    LOAD_REG(R10);         \
-    LOAD_REG(R11);         \
-    LOAD_REG(R12);         \
-    LOAD_REG(R13);         \
-    LOAD_REG(R14);         \
-    LOAD_REG(R15);         \
-    LDRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
-    if(A) {LDRx_U12(A, xEmu, offsetof(x64emu_t, ip));}
+#define LOAD_XEMU_REGS(A)   \
+    LDP_REGS(RAX, RCX);     \
+    LDP_REGS(RDX, RBX);     \
+    LDP_REGS(RSP, RBP);     \
+    LDP_REGS(RSI, RDI);     \
+    LDP_REGS(R8, R9);       \
+    LDP_REGS(R10, R11);     \
+    LDP_REGS(R12, R13);     \
+    LDP_REGS(R14, R15);     \
+    if(A==xRIP) {           \
+        LDPx_S7_offset(xFlags, xRIP, xEmu, offsetof(x64emu_t, eflags)); \
+    } else {                \
+        LDRx_U12(xFlags, xEmu, offsetof(x64emu_t, eflags)); \
+        if(A) {LDRx_U12(A, xEmu, offsetof(x64emu_t, ip));}  \
+    }

 #define STORE_XEMU_MINIMUM(A)  \
    STORE_REG(RAX);         \
@@ -518,35 +591,49 @@
    LDP_REGS(R12, R13);     \
    LDP_REGS(R14, R15)

-#define SET_DFNONE(S)    if(!dyn->dfnone) {MOVZw(S, d_none); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->dfnone=1;}
-#define SET_DF(S, N)     if((N)!=d_none) {MOVZw(S, (N)); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->dfnone=0;} else SET_DFNONE(S)
-#define SET_NODF()          dyn->dfnone = 0
-#define SET_DFOK()          dyn->dfnone = 1
+#define SET_DFNONE(S)    if(!dyn->f.dfnone) {MOVZw(S, d_none); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->f.dfnone=1;}
+#define SET_DF(S, N)     if((N)!=d_none) {MOVZw(S, (N)); STRw_U12(S, xEmu, offsetof(x64emu_t, df)); dyn->f.dfnone=0;} else SET_DFNONE(S)
+#define SET_NODF()          dyn->f.dfnone = 0
+#define SET_DFOK()          dyn->f.dfnone = 1
+
+#ifndef MAYSETFLAGS
+#define MAYSETFLAGS()
+#endif

 #ifndef READFLAGS
 #define READFLAGS(A) \
-    if(((A)!=X_PEND) && dyn->state_flags!=SF_SET && dyn->state_flags!=SF_SET_PENDING) { \
-        if(dyn->state_flags!=SF_PENDING) {              \
+    if(((A)!=X_PEND && dyn->f.pending!=SF_SET)          \
+    && (dyn->f.pending!=SF_SET_PENDING)) {              \
+        if(dyn->f.pending!=SF_PENDING) {                \
            LDRw_U12(x3, xEmu, offsetof(x64emu_t, df)); \
-            j64 = (GETMARKF)-(dyn->arm_size);           \
+            j64 = (GETMARKF)-(dyn->native_size);        \
            CBZw(x3, j64);                              \
        }                                               \
        CALL_(UpdateFlags, -1, 0);                      \
        MARKF;                                          \
-        dyn->state_flags = SF_SET;                      \
+        dyn->f.pending = SF_SET;                        \
        SET_DFOK();                                     \
    }
 #endif
-#ifndef SETFLAGS
-#define SETFLAGS(A, B)  \
-    if(dyn->state_flags!=SF_SET && B==SF_SUBSET && (dyn->insts[ninst].x64.need_flags&(~((A)/*|X_PEND*/)))) \
-        READFLAGS(dyn->insts[ninst].x64.need_flags&(~(A)|X_PEND));  \
-    dyn->state_flags = (B==SF_SUBSET)?SF_SET:                       \
-        ((B==SF_SET_PENDING && !(dyn->insts[ninst].x64.need_flags&X_PEND)?SF_SET:B))

+#ifndef SETFLAGS
+#define SETFLAGS(A, B)                                                                          \
+    if(dyn->f.pending!=SF_SET                                                                   \
+    && ((B)&SF_SUB)                                                                             \
+    && (dyn->insts[ninst].x64.gen_flags&(~(A))))                                                \
+        READFLAGS(((dyn->insts[ninst].x64.gen_flags&X_PEND)?X_ALL:dyn->insts[ninst].x64.gen_flags)&(~(A)));\
+    if(dyn->insts[ninst].x64.gen_flags) switch(B) {                                             \
+        case SF_SUBSET:                                                                         \
+        case SF_SET: dyn->f.pending = SF_SET; break;                                            \
+        case SF_PENDING: dyn->f.pending = SF_PENDING; break;                                    \
+        case SF_SUBSET_PENDING:                                                                 \
+        case SF_SET_PENDING:                                                                    \
+            dyn->f.pending = (dyn->insts[ninst].x64.gen_flags&X_PEND)?SF_SET_PENDING:SF_SET;    \
+            break;                                                                              \
+    } else dyn->f.pending = SF_SET
 #endif
 #ifndef JUMP
-#define JUMP(A) 
+#define JUMP(A, C) 
 #endif
 #ifndef BARRIER
 #define BARRIER(A) 
@@ -554,12 +641,12 @@
 #ifndef BARRIER_NEXT
 #define BARRIER_NEXT(A)
 #endif
-#define UFLAG_OP1(A) if(dyn->insts && dyn->insts[ninst].x64.need_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, op1));}
-#define UFLAG_OP2(A) if(dyn->insts && dyn->insts[ninst].x64.need_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, op2));}
-#define UFLAG_OP12(A1, A2) if(dyn->insts && dyn->insts[ninst].x64.need_flags) {STRxw_U12(A1, xEmu, offsetof(x64emu_t, op1));STRxw_U12(A2, 0, offsetof(x64emu_t, op2));}
-#define UFLAG_RES(A) if(dyn->insts && dyn->insts[ninst].x64.need_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, res));}
-#define UFLAG_DF(r, A) if(dyn->insts && dyn->insts[ninst].x64.need_flags) {SET_DF(r, A)}
-#define UFLAG_IF if(dyn->insts && dyn->insts[ninst].x64.need_flags)
+#define UFLAG_OP1(A) if(dyn->insts[ninst].x64.gen_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, op1));}
+#define UFLAG_OP2(A) if(dyn->insts[ninst].x64.gen_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, op2));}
+#define UFLAG_OP12(A1, A2) if(dyn->insts[ninst].x64.gen_flags) {STRxw_U12(A1, xEmu, offsetof(x64emu_t, op1));STRxw_U12(A2, 0, offsetof(x64emu_t, op2));}
+#define UFLAG_RES(A) if(dyn->insts[ninst].x64.gen_flags) {STRxw_U12(A, xEmu, offsetof(x64emu_t, res));}
+#define UFLAG_DF(r, A) if(dyn->insts[ninst].x64.gen_flags) {SET_DF(r, A)}
+#define UFLAG_IF if(dyn->insts[ninst].x64.gen_flags)
 #ifndef DEFAULT
 #define DEFAULT      *ok = -1; BARRIER(2)
 #endif
@@ -603,6 +690,7 @@
            TABLE64(xRIP, (A));                         \
    }
 #endif
+#define CLEARIP()   dyn->last_ip=0

 #if STEP < 2
 #define PASS2IF(A, B) if(A)
@@ -623,7 +711,7 @@ void* arm64_next(x64emu_t* emu, uintptr_t addr);
 #define STEPNAME(N) STEPNAME2(N, STEP)
 #endif

-#define arm_pass        STEPNAME(arm_pass)
+#define native_pass        STEPNAME(native_pass)

 #define dynarec64_00       STEPNAME(dynarec64_00)
 #define dynarec64_0F       STEPNAME(dynarec64_0F)
@@ -642,6 +730,7 @@ void* arm64_next(x64emu_t* emu, uintptr_t addr);
 #define dynarec64_F0       STEPNAME(dynarec64_F0)
 #define dynarec64_660F     STEPNAME(dynarec64_660F)
 #define dynarec64_6664     STEPNAME(dynarec64_6664)
+#define dynarec64_66F0     STEPNAME(dynarec64_66F0)
 #define dynarec64_F20F     STEPNAME(dynarec64_F20F)
 #define dynarec64_F30F     STEPNAME(dynarec64_F30F)

@@ -731,12 +820,16 @@ void* arm64_next(x64emu_t* emu, uintptr_t addr);
 #define x87_do_push     STEPNAME(x87_do_push)
 #define x87_do_push_empty STEPNAME(x87_do_push_empty)
 #define x87_do_pop      STEPNAME(x87_do_pop)
+#define x87_get_current_cache   STEPNAME(x87_get_current_cache)
 #define x87_get_cache   STEPNAME(x87_get_cache)
+#define x87_get_neoncache STEPNAME(x87_get_neoncache)
 #define x87_get_st      STEPNAME(x87_get_st)
+#define x87_get_st_empty  STEPNAME(x87_get_st)
 #define x87_refresh     STEPNAME(x87_refresh)
 #define x87_forget      STEPNAME(x87_forget)
 #define x87_reget_st    STEPNAME(x87_reget_st)
 #define x87_stackcount  STEPNAME(x87_stackcount)
+#define x87_swapreg     STEPNAME(x87_swapreg)
 #define x87_setround    STEPNAME(x87_setround)
 #define x87_restoreround STEPNAME(x87_restoreround)
 #define sse_setround    STEPNAME(sse_setround)
@@ -757,11 +850,13 @@ void* arm64_next(x64emu_t* emu, uintptr_t addr);
 #define fpu_reflectcache STEPNAME(fpu_reflectcache)
 #endif

-/* setup r2 to address pointed by */
-uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int64_t* fixaddress, int absmax, uint32_t mask, rex_t rex, int s, int delta);
+#define CacheTransform       STEPNAME(CacheTransform)

 /* setup r2 to address pointed by */
-uintptr_t geted32(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int64_t* fixaddress, int absmax, uint32_t mask, rex_t rex, int s, int delta);
+uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int64_t* fixaddress, int absmax, uint32_t mask, rex_t rex, int* l, int s, int delta);
+
+/* setup r2 to address pointed by */
+uintptr_t geted32(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int64_t* fixaddress, int absmax, uint32_t mask, rex_t rex, int* l, int s, int delta);

 /* setup r2 to address pointed by */
 uintptr_t geted16(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int64_t* fixaddress, int absmax, uint32_t mask, int s);
@@ -852,21 +947,29 @@ void emit_pf(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
 // cache of the local stack counter, to avoid upadte at every call
 void x87_stackcount(dynarec_arm_t* dyn, int ninst, int scratch);
 // fpu push. Return the Dd value to be used
-int x87_do_push(dynarec_arm_t* dyn, int ninst);
+int x87_do_push(dynarec_arm_t* dyn, int ninst, int s1, int t);
 // fpu push. Do not allocate a cache register. Needs a scratch register to do x87stack synch (or 0 to not do it)
 void x87_do_push_empty(dynarec_arm_t* dyn, int ninst, int s1);
 // fpu pop. All previous returned Dd should be considered invalid
-void x87_do_pop(dynarec_arm_t* dyn, int ninst);
+void x87_do_pop(dynarec_arm_t* dyn, int ninst, int s1);
+// get cache index for a x87 reg, return -1 if cache doesn't exist
+int x87_get_current_cache(dynarec_arm_t* dyn, int ninst, int st, int t);
 // get cache index for a x87 reg, create the entry if needed
-int x87_get_cache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
+int x87_get_cache(dynarec_arm_t* dyn, int ninst, int populate, int s1, int s2, int a, int t);
+// get neoncache index for a x87 reg
+int x87_get_neoncache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
 // get vfpu register for a x87 reg, create the entry if needed
-int x87_get_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
+int x87_get_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a, int t);
+// get vfpu register for a x87 reg, create the entry if needed. Do not fetch the Stx if not already in cache
+int x87_get_st_empty(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a, int t);
 // refresh a value from the cache ->emu (nothing done if value is not cached)
 void x87_refresh(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
 // refresh a value from the cache ->emu and then forget the cache (nothing done if value is not cached)
 void x87_forget(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
 // refresh the cache value from emu
 void x87_reget_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
+// swap 2 x87 regs
+void x87_swapreg(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a, int b);
 // Set rounding according to cw flags, return reg to restore flags
 int x87_setround(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 // Restore round flag
@@ -874,15 +977,47 @@ void x87_restoreround(dynarec_arm_t* dyn, int ninst, int s1);
 // Set rounding according to mxcsr flags, return reg to restore flags
 int sse_setround(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);

+void CacheTransform(dynarec_arm_t* dyn, int ninst, int cacheupd, int s1, int s2, int s3);
+
+#if STEP < 2
+#define CHECK_CACHE()   0
+#else
+#define CHECK_CACHE()   (cacheupd = CacheNeedsTransform(dyn, ninst))
+#endif
+
+#define neoncache_st_coherency STEPNAME(neoncache_st_coherency)
+int neoncache_st_coherency(dynarec_arm_t* dyn, int ninst, int a, int b);
+
+#if STEP == 0
+#define ST_IS_F(A)          0
+#define X87_COMBINE(A, B)   NEON_CACHE_ST_D
+#define X87_ST0             NEON_CACHE_ST_D
+#define X87_ST(A)           NEON_CACHE_ST_D
+#elif STEP == 1
+#define ST_IS_F(A) (neoncache_get_current_st(dyn, ninst, A)==NEON_CACHE_ST_F)
+#define X87_COMBINE(A, B) neoncache_combine_st(dyn, ninst, A, B)
+#define X87_ST0     neoncache_get_current_st(dyn, ninst, 0)
+#define X87_ST(A)   neoncache_get_current_st(dyn, ninst, A)
+#else
+#define ST_IS_F(A) (neoncache_get_st(dyn, ninst, A)==NEON_CACHE_ST_F)
+#if STEP == 3
+#define X87_COMBINE(A, B) neoncache_st_coherency(dyn, ninst, A, B)
+#else
+#define X87_COMBINE(A, B) neoncache_get_st(dyn, ninst, A)
+#endif
+#define X87_ST0     neoncache_get_st(dyn, ninst, 0)
+#define X87_ST(A)   neoncache_get_st(dyn, ninst, A)
+#endif
+
 //MMX helpers
 // get neon register for a MMX reg, create the entry if needed
-int mmx_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a);
+int mmx_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int a);
 // get neon register for a MMX reg, but don't try to synch it if it needed to be created
-int mmx_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a);
+int mmx_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int a);

 //SSE/SSE2 helpers
 // get neon register for a SSE reg, create the entry if needed
-int sse_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a);
+int sse_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a, int forwrite);
 // get neon register for a SSE reg, but don't try to synch it if it needed to be created
 int sse_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a);
 // forget neon register for a SSE reg, create the entry if needed
@@ -892,13 +1027,13 @@ void sse_purge07cache(dynarec_arm_t* dyn, int ninst, int s1);

 // common coproc helpers
 // reset the cache
-void fpu_reset(dynarec_arm_t* dyn, int ninst);
+void fpu_reset(dynarec_arm_t* dyn);
 // purge the FPU cache (needs 3 scratch registers)
-void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 // purge MMX cache
-void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int s1);
+void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1);
 // purge x87 cache
-void x87_purgecache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+void x87_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 #ifdef HAVE_TRACE
 void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 #endif
@@ -917,11 +1052,12 @@ uintptr_t dynarec64_D9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nin
 uintptr_t dynarec64_DB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_DC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_DD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
-//uintptr_t dynarec64_DE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
+uintptr_t dynarec64_DE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_DF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_6664(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
+uintptr_t dynarec64_66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog);
 uintptr_t dynarec64_F20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int* ok, int* need_epilog);
 uintptr_t dynarec64_F30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int* ok, int* need_epilog);

@@ -0,0 +1,11 @@
+#include <stdint.h>
+
+#include "arm64_emitter.h"
+
+#define EMIT(A) *block = (A); ++block
+void CreateJmpNext(void* addr, void* next)
+{
+    uint32_t* block = (uint32_t*)addr;
+    LDRx_literal(x2, (intptr_t)next - (intptr_t)addr);
+    BR(x2);
+}
@@ -0,0 +1,53 @@
+
+#define INIT    uintptr_t sav_addr=addr
+#define FINI    \
+    dyn->isize = addr-sav_addr;         \
+    dyn->insts[ninst].x64.addr = addr;  \
+    if(ninst) dyn->insts[ninst-1].x64.size = dyn->insts[ninst].x64.addr - dyn->insts[ninst-1].x64.addr
+
+#define MESSAGE(A, ...)  
+#define MAYSETFLAGS()   dyn->insts[ninst].x64.may_set = 1
+#define READFLAGS(A)    \
+        dyn->insts[ninst].x64.use_flags = A; dyn->f.dfnone = 1;\
+        dyn->f.pending=SF_SET
+#define SETFLAGS(A,B)   \
+        dyn->insts[ninst].x64.set_flags = A;    \
+        dyn->insts[ninst].x64.state_flags = B;  \
+        dyn->f.pending=(B)&SF_SET_PENDING;      \
+        dyn->f.dfnone=((B)&SF_SET)?1:0;
+#define EMIT(A)     
+#define JUMP(A, C)         add_next(dyn, (uintptr_t)A); dyn->insts[ninst].x64.jmp = A; dyn->insts[ninst].x64.jmp_cond = C
+#define BARRIER(A)      if(A!=BARRIER_MAYBE) {fpu_purgecache(dyn, ninst, 0, x1, x2, x3); dyn->insts[ninst].x64.barrier = A;} else dyn->insts[ninst].barrier_maybe = 1
+#define BARRIER_NEXT(A) dyn->insts[ninst+1].x64.barrier = A
+#define NEW_INST \
+        ++dyn->size;                            \
+        if(dyn->size+3>=dyn->cap) {             \
+                dyn->insts = (instruction_native_t*)customRealloc(dyn->insts, sizeof(instruction_native_t)*dyn->cap*2);\
+                memset(&dyn->insts[dyn->cap], 0, sizeof(instruction_native_t)*dyn->cap);   \
+                dyn->cap *= 2;                  \
+        }                                       \
+        dyn->insts[ninst].x64.addr = ip;        \
+        dyn->n.combined1 = dyn->n.combined2 = 0;\
+        dyn->n.swapped = 0; dyn->n.barrier = 0; \
+        dyn->insts[ninst].f_entry = dyn->f;     \
+        if(ninst) {dyn->insts[ninst-1].x64.size = dyn->insts[ninst].x64.addr - dyn->insts[ninst-1].x64.addr;}
+
+#define INST_EPILOG                             \
+        dyn->insts[ninst].f_exit = dyn->f;      \
+        dyn->insts[ninst].n = dyn->n;           \
+        dyn->insts[ninst].x64.has_next = (ok>0)?1:0;
+#define INST_NAME(name) 
+#define DEFAULT                         \
+        --dyn->size;                    \
+        *ok = -1;                       \
+        if(box64_dynarec_log>=LOG_INFO) {\
+        dynarec_log(LOG_NONE, "%p: Dynarec stopped because of Opcode %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X", \
+        (void*)ip, PKip(0),             \
+        PKip(1), PKip(2), PKip(3),      \
+        PKip(4), PKip(5), PKip(6),      \
+        PKip(7), PKip(8), PKip(9),      \
+        PKip(10),PKip(11),PKip(12),     \
+        PKip(13),PKip(14));             \
+        printFunctionAddr(ip, " => ");  \
+        dynarec_log(LOG_NONE, "\n");    \
+        }
@@ -0,0 +1,14 @@
+#define INIT    
+#define FINI
+#define MESSAGE(A, ...)  
+#define EMIT(A)     
+#define NEW_INST                                \
+        dyn->insts[ninst].f_entry = dyn->f;     \
+        dyn->n.combined1 = dyn->n.combined2 = 0;\
+        dyn->n.swapped = 0; dyn->n.barrier = 0
+
+#define INST_EPILOG                             \
+        dyn->insts[ninst].n = dyn->n;           \
+        dyn->insts[ninst].f_exit = dyn->f
+
+#define INST_NAME(name)  
@@ -0,0 +1,16 @@
+#define INIT        dyn->native_size = 0
+#define FINI        if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
+
+#define MESSAGE(A, ...)  
+#define EMIT(A)     dyn->insts[ninst].size+=4; dyn->native_size+=4
+#define NEW_INST                                                                                        \
+        if(ninst) {                                                                                     \
+                dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);     \
+                if(ninst && isInstClean(dyn, ninst)) {                                                  \
+                        if(dyn->last_ip!=ip) dyn->last_ip = 0;                                          \
+                }                                                                                       \
+        }
+#define INST_EPILOG dyn->insts[ninst].epilog = dyn->native_size; 
+#define INST_NAME(name) 
+#define TABLE64(A, V)   {Table64(dyn, (V)); EMIT(0);}
+#define FTABLE64(A, V)  {mmx87_regs_t v = {.d = V}; Table64(dyn, v.q); EMIT(0);}
@@ -0,0 +1,64 @@
+#define INIT    
+#define FINI
+#define EMIT(A)     \
+    if(box64_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (uint32_t)(A), arm64_print(A, (uintptr_t)dyn->block));} \
+    *(uint32_t*)(dyn->block) = (uint32_t)(A);       \
+    dyn->block += 4; dyn->native_size += 4;         \
+    dyn->insts[ninst].size2 += 4
+
+#define MESSAGE(A, ...)  if(box64_dynarec_dump) dynarec_log(LOG_NONE, __VA_ARGS__)
+#define NEW_INST        \
+    if(ninst && isInstClean(dyn, ninst)) {                      \
+        if(dyn->last_ip!=ip) dyn->last_ip = 0;                  \
+    }
+#define INST_EPILOG     
+#define INST_NAME(name) \
+    if(box64_dynarec_dump) {\
+        printf_x64_instruction(my_context->dec, &dyn->insts[ninst].x64, name); \
+        dynarec_log(LOG_NONE, "%s%p: %d emited opcodes, inst=%d, barrier=%d state=%d/%d(%d), %s=%X/%X, use=%X, need=%X/%X", \
+            (box64_dynarec_dump>1)?"\e[32m":"", \
+            (void*)(dyn->native_start+dyn->insts[ninst].address),  \
+            dyn->insts[ninst].size/4,           \
+            ninst,                              \
+            dyn->insts[ninst].x64.barrier,      \
+            dyn->insts[ninst].x64.state_flags,  \
+            dyn->f.pending,                     \
+            dyn->f.dfnone,                      \
+            dyn->insts[ninst].x64.may_set?"may":"set",              \
+            dyn->insts[ninst].x64.set_flags,    \
+            dyn->insts[ninst].x64.gen_flags,    \
+            dyn->insts[ninst].x64.use_flags,    \
+            dyn->insts[ninst].x64.need_before,  \
+            dyn->insts[ninst].x64.need_after);  \
+        if(dyn->insts[ninst].pred_sz) {         \
+            dynarec_log(LOG_NONE, ", pred=");   \
+            for(int ii=0; ii<dyn->insts[ninst].pred_sz; ++ii)\
+                dynarec_log(LOG_NONE, "%s%d", ii?"/":"", dyn->insts[ninst].pred[ii]);\
+        }                                       \
+        if(dyn->insts[ninst].x64.jmp && dyn->insts[ninst].x64.jmp_insts>=0)\
+            dynarec_log(LOG_NONE, ", jmp=%d", dyn->insts[ninst].x64.jmp_insts);\
+        if(dyn->insts[ninst].x64.jmp && dyn->insts[ninst].x64.jmp_insts==-1)\
+            dynarec_log(LOG_NONE, ", jmp=out"); \
+        if(dyn->last_ip)                        \
+            dynarec_log(LOG_NONE, ", last_ip=%p", (void*)dyn->last_ip);\
+        for(int ii=0; ii<24; ++ii) {            \
+            switch(dyn->insts[ninst].n.neoncache[ii].t) {    \
+                case NEON_CACHE_ST_D: dynarec_log(LOG_NONE, " D%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;  \
+                case NEON_CACHE_ST_F: dynarec_log(LOG_NONE, " S%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;  \
+                case NEON_CACHE_MM: dynarec_log(LOG_NONE, " D%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;    \
+                case NEON_CACHE_XMMW: dynarec_log(LOG_NONE, " Q%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;  \
+                case NEON_CACHE_XMMR: dynarec_log(LOG_NONE, " Q%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;  \
+                case NEON_CACHE_SCR: dynarec_log(LOG_NONE, " D%d:%s", ii, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;   \
+                case NEON_CACHE_NONE:           \
+                default:    break;              \
+            }                                   \
+        }                                       \
+        if(dyn->n.stack || dyn->insts[ninst].n.stack_next || dyn->insts[ninst].n.x87stack)     \
+            dynarec_log(LOG_NONE, " X87:%d/%d(+%d/-%d)%d", dyn->n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, dyn->insts[ninst].n.stack_pop, dyn->insts[ninst].n.x87stack); \
+        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)                     \
+            dynarec_log(LOG_NONE, " %s:%d/%d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);   \
+        dynarec_log(LOG_NONE, "%s\n", (box64_dynarec_dump>1)?"\e[m":"");                       \
+    }
+
+#define TABLE64(A, V)   {int val64offset = Table64(dyn, (V)); MESSAGE(LOG_DUMP, "  Table64: 0x%lx\n", (V)); LDRx_literal(A, val64offset);}
+#define FTABLE64(A, V)  {mmx87_regs_t v = {.d = V}; int val64offset = Table64(dyn, v.q); MESSAGE(LOG_DUMP, "  FTable64: %g\n", v.d); VLDR64_literal(A, val64offset);}
@@ -0,0 +1,125 @@
+#ifndef __DYNAREC_ARM_PRIVATE_H_
+#define __DYNAREC_ARM_PRIVATE_H_
+
+#include "../dynarec_private.h"
+
+typedef struct x64emu_s x64emu_t;
+typedef struct dynablock_s dynablock_t;
+typedef struct instsize_s instsize_t;
+
+#define BARRIER_MAYBE   8
+
+#define NEON_CACHE_NONE 0
+#define NEON_CACHE_ST_D 1
+#define NEON_CACHE_ST_F 2
+#define NEON_CACHE_MM   3
+#define NEON_CACHE_XMMW 4
+#define NEON_CACHE_XMMR 5
+#define NEON_CACHE_SCR  6
+typedef union neon_cache_s {
+    int8_t           v;
+    struct {
+        uint8_t t:4;   // reg type
+        uint8_t n:4;   // reg number
+    };
+} neon_cache_t;
+typedef union sse_cache_s {
+    int8_t      v;
+    struct {
+        uint8_t reg:7;
+        uint8_t write:1;
+    };
+} sse_cache_t;
+typedef struct neoncache_s {
+    // Neon cache
+    neon_cache_t        neoncache[24];
+    int8_t              stack;
+    int8_t              stack_next;
+    int8_t              stack_pop;
+    int8_t              stack_push;
+    uint8_t             combined1;
+    uint8_t             combined2;
+    uint8_t             swapped;        // the combined reg were swapped
+    uint8_t             barrier;        // is there a barrier at instruction epilog?
+    uint32_t            news;           // bitmask, wich neoncache are new for this opcode
+    // fpu cache
+    int8_t              x87cache[8];    // cache status for the 8 x87 register behind the fpu stack
+    int8_t              x87reg[8];      // reg used for x87cache entry
+    int8_t              mmxcache[8];    // cache status for the 8 MMX registers
+    sse_cache_t         ssecache[16];   // cache status for the 16 SSE(2) registers
+    int8_t              fpuused[24];    // all 0..24 double reg from fpu, used by x87, sse and mmx
+    int8_t              x87stack;       // cache stack counter
+    int8_t              mmxcount;       // number of mmx register used (not both mmx and x87 at the same time)
+    int8_t              fpu_scratch;    // scratch counter
+    int8_t              fpu_extra_qscratch; // some opcode need an extra quad scratch register
+    int8_t              fpu_reg;        // x87/sse/mmx reg counter
+} neoncache_t;
+
+typedef struct flagcache_s {
+    int                 pending;    // is there a pending flags here, or to check?
+    int                 dfnone;     // if defered flags is already set to df_none
+} flagcache_t;
+
+typedef struct instruction_arm64_s {
+    instruction_x64_t   x64;
+    uintptr_t           address;    // (start) address of the arm emited instruction
+    uintptr_t           epilog;     // epilog of current instruction (can be start of next, or barrier stuff)
+    int                 size;       // size of the arm emited instruction
+    int                 size2;      // size of the arm emited instrucion after pass2
+    int                 pred_sz;    // size of predecessor list
+    int                 *pred;      // predecessor array
+    uintptr_t           mark, mark2, mark3;
+    uintptr_t           markf;
+    uintptr_t           markseg;
+    uintptr_t           marklock;
+    int                 pass2choice;// value for choices that are fixed on pass2 for pass3
+    uintptr_t           natcall;
+    int                 retn;
+    int                 barrier_maybe;
+    flagcache_t         f_exit;     // flags status at end of intruction
+    neoncache_t         n;          // neoncache at end of intruction (but before poping)
+    flagcache_t         f_entry;    // flags status before the instruction begin
+} instruction_arm64_t;
+
+typedef struct dynarec_arm_s {
+    instruction_arm64_t *insts;
+    int32_t             size;
+    int32_t             cap;
+    uintptr_t           start;      // start of the block
+    uint32_t            isize;      // size in byte of x64 instructions included
+    void*               block;      // memory pointer where next instruction is emited
+    uintptr_t           native_start;  // start of the arm code
+    size_t              native_size;   // size of emitted arm code
+    uintptr_t           last_ip;    // last set IP in RIP (or NULL if unclean state) TODO: move to a cache something
+    uint64_t            *table64;   // table of 64bits value
+    int                 table64size;// size of table (will be appended at end of executable code)
+    int                 table64cap;
+    uintptr_t           tablestart;
+    flagcache_t         f;
+    neoncache_t         n;          // cache for the 8..31 double reg from fpu, plus x87 stack delta
+    uintptr_t*          next;       // variable array of "next" jump address
+    int                 next_sz;
+    int                 next_cap;
+    int*                predecessor;// single array of all predecessor
+    dynablock_t*        dynablock;
+    instsize_t*         instsize;
+} dynarec_arm_t;
+
+void add_next(dynarec_arm_t *dyn, uintptr_t addr);
+uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr);
+int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n);
+int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n);
+
+int Table64(dynarec_arm_t *dyn, uint64_t val);  // add a value to etable64 (if needed) and gives back the imm19 to use in LDR_literal
+
+void CreateJmpNext(void* addr, void* next);
+
+#define GO_TRACE() \
+    GETIP_(ip);             \
+    MOVx_REG(x1, xRIP);     \
+    STORE_XEMU_CALL(xRIP);  \
+    MOV32w(x2, 1);          \
+    CALL(PrintTrace, -1);   \
+    LOAD_XEMU_CALL(xRIP)
+
+#endif //__DYNAREC_ARM_PRIVATE_H_
@@ -2,6 +2,8 @@
 #include <stdlib.h>
 #include <pthread.h>
 #include <errno.h>
+#include <setjmp.h>
+#include <sys/mman.h>

 #include "debug.h"
 #include "box64context.h"
@@ -18,12 +20,11 @@
 #include "dynablock_private.h"
 #include "dynarec_private.h"
 #include "elfloader.h"
-#ifdef ARM64
-#include "dynarec_arm64.h"
-#include "arm64_lock.h"
-#else
-#error Unsupported architecture!
-#endif
+#include "bridge.h"
+
+#include "dynarec_native.h"
+#include "native_lock.h"
+
 #include "custommem.h"
 #include "khash.h"

@@ -38,104 +39,37 @@ uint32_t X31_hash_code(void* addr, int len)
 	return (uint32_t)h;
 }

-dynablocklist_t* NewDynablockList(uintptr_t text, int textsz, int direct)
-{
-    if(!textsz) {
-        printf_log(LOG_NONE, "Error, creating a NULL sized Dynablock\n");
-        return NULL;
-    }
-    dynablocklist_t* ret = (dynablocklist_t*)calloc(1, sizeof(dynablocklist_t));
-    ret->text = text;
-    ret->textsz = textsz;
-    if(direct && textsz) {
-        ret->direct = (dynablock_t**)calloc(textsz, sizeof(dynablock_t*));
-        if(!ret->direct) {printf_log(LOG_NONE, "Warning, fail to create direct block for dynablock @%p\n", (void*)text);}
-    }
-    dynarec_log(LOG_DEBUG, "New Dynablocklist %p, from %p->%p\n", ret, (void*)text, (void*)(text+textsz));
-    return ret;
-}
-
-void FreeDynablock(dynablock_t* db)
+void FreeDynablock(dynablock_t* db, int need_lock)
 {
    if(db) {
-        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x64=%p:%p parent=%p, father=%p, with %d son(s) already gone=%d\n", db, db->block, db->x64_addr, db->x64_addr+db->x64_size, db->parent, db->father, db->sons_size, db->gone);
        if(db->gone)
            return; // already in the process of deletion!
-        db->done = 0;
-        db->gone = 1;
-        // remove from direct if there
-        uintptr_t startdb = db->parent->text;
-        uintptr_t enddb = db->parent->text + db->parent->textsz;
-        if(db->parent->direct) {
-            uintptr_t addr = (uintptr_t)db->x64_addr;
-            if(addr>=startdb && addr<enddb)
-                arm64_lock_xchg(&db->parent->direct[addr-startdb], 0);   // secured write
-        }
+        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x64=%p:%p already gone=%d\n", db, db->block, db->x64_addr, db->x64_addr+db->x64_size-1, db->gone);
+        if(need_lock)
+            pthread_mutex_lock(&my_context->mutex_dyndump);
        // remove jumptable
        setJumpTableDefault64(db->x64_addr);
-        // remove and free the sons
-        for (int i=0; i<db->sons_size; ++i) {
-            dynablock_t *son = (dynablock_t*)arm64_lock_xchg(&db->sons[i], 0);
-            FreeDynablock(son);
-        }
-        // only the father free the DynarecMap
-        if(!db->father) {
-            dynarec_log(LOG_DEBUG, " -- FreeDyrecMap(%p, %d)\n", db->block, db->size);
-            FreeDynarecMap(db, (uintptr_t)db->block, db->size);
-        }
-        free(db->sons);
-        free(db->instsize);
-        free(db);
+        dynarec_log(LOG_DEBUG, " -- FreeDyrecMap(%p, %d)\n", db->actual_block, db->size);
+        db->done = 0;
+        db->gone = 1;
+        FreeDynarecMap(db, (uintptr_t)db->actual_block, db->size);
+        customFree(db);
+        if(need_lock)
+            pthread_mutex_unlock(&my_context->mutex_dyndump);
    }
 }

-void FreeDynablockList(dynablocklist_t** dynablocks)
-{
-    if(!dynablocks)
-        return;
-    if(!*dynablocks)
-        return;
-    dynarec_log(LOG_DEBUG, "Free Dynablocklist %p, with Direct Blocks %p\n", *dynablocks, (*dynablocks)->direct);
-    if((*dynablocks)->direct) {
-        for (int i=0; i<(*dynablocks)->textsz; ++i) {
-            if((*dynablocks)->direct[i] && !(*dynablocks)->direct[i]->father) 
-                FreeDynablock((*dynablocks)->direct[i]);
-        }
-        free((*dynablocks)->direct);
-    }
-    (*dynablocks)->direct = NULL;
-
-    free(*dynablocks);
-    *dynablocks = NULL;
-}
-
 void MarkDynablock(dynablock_t* db)
 {
    if(db) {
-        if(db->father)
-            db = db->father;    // mark only father
        if(db->need_test)
            return; // already done
-        db->need_test = 1;  // test only blocks that can be marked (and so deleted)
-        setJumpTableDefault64(db->x64_addr);
-        for(int i=0; i<db->sons_size; ++i)
-            setJumpTableDefault64(db->sons[i]->x64_addr);
+        dynarec_log(LOG_DEBUG, "MarkDynablock %p %p-%p\n", db, db->x64_addr, db->x64_addr+db->x64_size-1);
+        db->need_test = 1;
+        setJumpTableIfRef64(db->x64_addr, db->jmpnext, db->block);
    }
 }

-uintptr_t StartDynablockList(dynablocklist_t* db)
-{
-    if(db)
-        return db->text;
-    return 0;
-}
-uintptr_t EndDynablockList(dynablocklist_t* db)
-{
-    if(db)
-        return db->text+db->textsz-1;
-    return 0;
-}
-
 int IntervalIntersects(uintptr_t start1, uintptr_t end1, uintptr_t start2, uintptr_t end2)
 {
    if(start1 > end2 || start2 > end1)
@@ -143,85 +77,38 @@ int IntervalIntersects(uintptr_t start1, uintptr_t end1, uintptr_t start2, uintp
    return 1;
 }

-void MarkDirectDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+static int MarkedDynablock(dynablock_t* db)
 {
-    if(!dynablocks)
-        return;
-    if(!dynablocks->direct)
-        return;
-    uintptr_t startdb = dynablocks->text;
-    uintptr_t enddb = startdb + dynablocks->textsz -1;
-    uintptr_t start = addr;
-    uintptr_t end = addr+size-1;
-    if(start<startdb)
-        start = startdb;
-    if(end>enddb)
-        end = enddb;
-    dynablock_t *db;
-    if(end>startdb && start<enddb)
-        for(uintptr_t i = start; i<end; ++i)
-            if((db=dynablocks->direct[i-startdb]))
-                if(IntervalIntersects((uintptr_t)db->x64_addr, (uintptr_t)db->x64_addr+db->x64_size-1, addr, addr+size+1))
-                    MarkDynablock(db);
+    if(db) {
+        if(db->need_test)
+            return 1; // already done
+    }
+    return 0;
 }

-int FreeRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+void MarkRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
 {
-    if(!dynablocks)
+    // Mark will try to find *any* blocks that intersect the range to mark
+    if(!db)
+        return;
+    dynarec_log(LOG_DEBUG, "MarkRangeDynablock %p-%p .. startdb=%p, sizedb=%p\n", (void*)addr, (void*)addr+size-1, (void*)db->x64_addr, (void*)db->x64_size);
+    if(!MarkedDynablock(db))
+        if(IntervalIntersects((uintptr_t)db->x64_addr, (uintptr_t)db->x64_addr+db->x64_size-1, addr, addr+size+1))
+            MarkDynablock(db);
+}
+
+int FreeRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
+{
+    if(!db)
        return 1;

-    if(dynablocks->direct) {
-        dynablock_t* db;
-        int ret;
-        khint_t k;
-        kh_dynablocks_t *blocks = kh_init(dynablocks);
-        // copy in a temporary list
-        uintptr_t startdb = dynablocks->text;
-        uintptr_t enddb = startdb + dynablocks->textsz;
-        uintptr_t start = addr;
-        uintptr_t end = addr+size;
-        if(start<startdb)
-            start = startdb;
-        if(end>enddb)
-            end = enddb;
-        if(end>startdb && start<enddb)
-            for(uintptr_t i = start; i<end; ++i) {
-                db = (dynablock_t*)arm64_lock_xchg(&dynablocks->direct[i-startdb], 0);
-                if(db) {
-                    if(db->father)
-                        db = db->father;
-                    if(db->parent==dynablocks) {
-                        k = kh_put(dynablocks, blocks, (uintptr_t)db, &ret);
-                        kh_value(blocks, k) = db;
-                    }
-                }
-            }
-        // purge the list
-        kh_foreach_value(blocks, db,
-            FreeDynablock(db);
-        );
-        kh_destroy(dynablocks, blocks);
-        // check emptyness
-        for(uintptr_t i=0; i<dynablocks->textsz; ++i)
-            if(dynablocks->direct[i])
-                return 0;
-        return 1;
+    int need_lock = my_context?1:0;
+    if(IntervalIntersects((uintptr_t)db->x64_addr, (uintptr_t)db->x64_addr+db->x64_size-1, addr, addr+size+1)) {
+        FreeDynablock(db, need_lock);
+        return 0;
    }
    return 1;
 }
-void MarkRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
-{
-    if(!dynablocks)
-        return;
-    if(dynablocks->direct) {
-        uintptr_t new_addr = addr - dynablocks->maxsz;
-        uintptr_t new_size = size + dynablocks->maxsz;
-        MarkDirectDynablock(dynablocks, new_addr, new_size);
-        // the blocks check before
-        for(unsigned idx=(new_addr)>>DYNAMAP_SHIFT; idx<(addr>>DYNAMAP_SHIFT); ++idx)
-            MarkDirectDynablock(getDB(idx), new_addr, new_size);
-    }
-}

 dynablock_t* FindDynablockDynablocklist(void* addr, kh_dynablocks_t* dynablocks)
 {
@@ -232,174 +119,130 @@ dynablock_t* FindDynablockDynablocklist(void* addr, kh_dynablocks_t* dynablocks)
        const uintptr_t s = (uintptr_t)db->block;
        const uintptr_t e = (uintptr_t)db->block+db->size;
        if((uintptr_t)addr>=s && (uintptr_t)addr<e)
-            return db->father?db->father:db;
+            return db;
    )
    return NULL;
 }

-static dynablocklist_t* getDBFromAddress(uintptr_t addr)
+dynablock_t *AddNewDynablock(uintptr_t addr)
 {
-    const uintptr_t idx = (addr>>DYNAMAP_SHIFT);
-    return getDB(idx);
-}
-
-dynablock_t *AddNewDynablock(dynablocklist_t* dynablocks, uintptr_t addr, int* created)
-{
-    if(!dynablocks) {
-        dynarec_log(LOG_INFO, "Warning: Ask to create a dynablock with a NULL dynablocklist (addr=%p)\n", (void*)addr);
-        *created = 0;
+    dynablock_t* block;
+    #if 0
+    // check if memory as the correct flags
+    int prot = getProtection(addr);
+    if(!(prot&(PROT_EXEC|PROT_DYNAREC|PROT_DYNAREC_R))) {
+        dynarec_log(LOG_VERBOSE, "Block asked on a memory with no execution flags 0x%02X\n", prot);
        return NULL;
    }
-    if((addr<dynablocks->text) || (addr>=(dynablocks->text+dynablocks->textsz))) {
-        return AddNewDynablock(getDBFromAddress(addr), addr, created);
-    }
-    dynablock_t* block = NULL;
-    // first, check if it exist in direct access mode
-    if(dynablocks->direct) {
-        block = dynablocks->direct[addr-dynablocks->text];
-        if(block) {
-            dynarec_log(LOG_VERBOSE, "Block already exist in Direct Map\n");
-            *created = 0;
-            return block;
-        }
-    }
    
-    if (!*created)
-        return block;
-    
-    if(!dynablocks->direct) {
-        dynablock_t** p = (dynablock_t**)calloc(dynablocks->textsz, sizeof(dynablock_t*));
-        if(arm64_lock_storeifnull(&dynablocks->direct, p)!=p)
-            free(p);    // someone already create the direct array, too late...
-    }
+    #endif
+    pthread_mutex_lock(&my_context->mutex_dyndump);

    // create and add new block
    dynarec_log(LOG_VERBOSE, "Ask for DynaRec Block creation @%p\n", (void*)addr);

-    block = (dynablock_t*)calloc(1, sizeof(dynablock_t));
-    block->parent = dynablocks; 
-    dynablock_t* tmp = (dynablock_t*)arm64_lock_storeifnull(&dynablocks->direct[addr-dynablocks->text], block);
-    if(tmp !=  block) {
-        // a block appeard!
-        free(block);
-        *created = 0;
-        return tmp;
-    }
+    block = (dynablock_t*)customCalloc(1, sizeof(dynablock_t));

-    *created = 1;
+    pthread_mutex_lock(&my_context->mutex_dyndump);
    return block;
 }

+//TODO: move this to dynrec_arm.c and track allocated structure to avoid memory leak
+static __thread struct __jmp_buf_tag dynarec_jmpbuf;
+
+void cancelFillBlock()
+{
+    longjmp(&dynarec_jmpbuf, 1);
+}
+
 /* 
    return NULL if block is not found / cannot be created. 
    Don't create if create==0
 */
-static dynablock_t* internalDBGetBlock(x64emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, dynablock_t* current)
+static dynablock_t* internalDBGetBlock(x64emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, int need_lock)
 {
-    // try the quickest way first: get parent of current and check if ok!
-    dynablocklist_t *dynablocks = NULL;
-    dynablock_t* block = NULL;
-    if(current) {
-        dynablocks = current->parent;
-        if(dynablocks && !(addr>=dynablocks->text && addr<(dynablocks->text+dynablocks->textsz)))
-            dynablocks = NULL;
-    }
-    // nope, lets do the long way
-    if(!dynablocks) {
-        dynablocks = getDBFromAddress(addr);
-        if(!dynablocks) {
-            dynablocks = GetDynablocksFromAddress(emu->context, addr);
-            if(!dynablocks)
-                return NULL;
-        }
-    }
-    // check direct first, without lock
-    if(dynablocks->direct/* && (addr>=dynablocks->text) && (addr<(dynablocks->text+dynablocks->textsz))*/)
-        if((block = dynablocks->direct[addr-dynablocks->text]))
-            return block;
+    if(hasAlternate((void*)addr))
+        return NULL;
+    dynablock_t* block = getDB(addr);
+    if(block || !create)
+        return block;

-    int created = create;
-    block = AddNewDynablock(dynablocks, addr, &created);
-    if(!created)
-        return block;   // existing block...
-
-    #if 0
-    if(box64_dynarec_dump)
+    if(need_lock)
        pthread_mutex_lock(&my_context->mutex_dyndump);
-    #endif
+    
+    block = getDB(addr);    // just in case
+    if(block) {
+        if(need_lock)
+            pthread_mutex_unlock(&my_context->mutex_dyndump);
+        return block;
+    }
+    
+    block = AddNewDynablock(addr);
+
    // fill the block
    block->x64_addr = (void*)addr;
-    pthread_mutex_lock(&my_context->mutex_dyndump);
+    if(sigsetjmp(&dynarec_jmpbuf, 1)) {
+        printf_log(LOG_INFO, "FillBlock at %p triggered a segfault, cancelling\n", (void*)addr);
+        if(need_lock)
+            pthread_mutex_unlock(&my_context->mutex_dyndump);
+        return NULL;
+    }
    void* ret = FillBlock64(block, filladdr);
-    pthread_mutex_unlock(&my_context->mutex_dyndump);
    if(!ret) {
        dynarec_log(LOG_DEBUG, "Fillblock of block %p for %p returned an error\n", block, (void*)addr);
-        void* old = (void*)arm64_lock_storeifref(&dynablocks->direct[addr-dynablocks->text], 0, block);
-        if(old!=block && old) {// put it back in place, strange things are happening here!
-            dynarec_log(LOG_INFO, "Warning, a wild block appeared at %p: %p\n", (void*)addr, old);
-            // doing nothing else, the block has not be writen
-        }
-        free(block);
+        customFree(block);
        block = NULL;
    }
-    #if 0
-    if(box64_dynarec_dump)
-        pthread_mutex_unlock(&my_context->mutex_dyndump);
-    #endif
    // check size
-    if(block && block->x64_size) {
+    if(block && (block->x64_size || (!block->x64_size && !block->done))) {
        int blocksz = block->x64_size;
-        if(dynablocks->maxsz<blocksz) {
-            dynablocks->maxsz = blocksz;
-            for(unsigned idx=(addr>>DYNAMAP_SHIFT)+1; idx<=((addr+blocksz)>>DYNAMAP_SHIFT); ++idx) {
-                dynablocklist_t* dblist;
-                if((dblist = getDB(idx)))
-                    if(dblist->maxsz<blocksz)
-                        dblist->maxsz = blocksz;
-            }
-        }
-        lockDB();
-        protectDBnolock((uintptr_t)block->x64_addr, block->x64_size);
+        if(blocksz>my_context->max_db_size)
+            my_context->max_db_size = blocksz;
        // fill-in jumptable
-        addJumpTableIfDefault64(block->x64_addr, block->block);
-        for(int i=0; i<block->sons_size; ++i) {
-            addJumpTableIfDefault64(block->sons[i]->x64_addr, block->sons[i]->block);
-            block->sons[i]->done = 1;
+        if(!addJumpTableIfDefault64(block->x64_addr, block->block)) {
+            FreeDynablock(block, 1);
+            block = getDB(addr);
+        } else {
+            if(block->x64_size)
+                block->done = 1;    // don't validate the block if the size is null, but keep the block
        }
-        block->done = 1;
-        unlockDB();
    }
+    if(need_lock)
+        pthread_mutex_unlock(&my_context->mutex_dyndump);

-    dynarec_log(LOG_DEBUG, " --- DynaRec Block %s @%p:%p (%p, 0x%x bytes, with %d son(s))\n", created?"created":"recycled", (void*)addr, (void*)(addr+((block)?block->x64_size:0)), (block)?block->block:0, (block)?block->size:0, (block)?block->sons_size:0);
+    dynarec_log(LOG_DEBUG, "%04d| --- DynaRec Block created @%p:%p (%p, 0x%x bytes)\n", GetTID(), (void*)addr, (void*)(addr+((block)?block->x64_size:1)-1), (block)?block->block:0, (block)?block->size:0);

    return block;
 }

-dynablock_t* DBGetBlock(x64emu_t* emu, uintptr_t addr, int create, dynablock_t** current)
+dynablock_t* DBGetBlock(x64emu_t* emu, uintptr_t addr, int create)
 {
-    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, *current);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = X31_hash_code(father->x64_addr, father->x64_size);
-        if(hash!=father->hash) {
-            dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) with %d son(s) for %p\n", father, father->x64_addr, father->x64_addr+father->x64_size, hash, father->hash, father->sons_size, (void*)addr);
-            // no more current if it gets invalidated too
-            if(*current && father->x64_addr>=(*current)->x64_addr && (father->x64_addr+father->x64_size)<(*current)->x64_addr)
-                *current = NULL;
-            // Free father, it's now invalid!
-            FreeDynablock(father);
-            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, addr, create, *current);
-        } else {
-            father->need_test = 0;
-            lockDB();
-            protectDBnolock((uintptr_t)father->x64_addr, father->x64_size);
-            // fill back jumptable
-            addJumpTableIfDefault64(father->x64_addr, father->block);
-            for(int i=0; i<father->sons_size; ++i)
-                addJumpTableIfDefault64(father->sons[i]->x64_addr, father->sons[i]->block);
-            unlockDB();
+    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, 1);
+    if(db && db->done && db->block && db->need_test) {
+        if(AreaInHotPage((uintptr_t)db->x64_addr, (uintptr_t)db->x64_addr + db->x64_size - 1)) {
+            dynarec_log(LOG_DEBUG, "Not running block %p from %p:%p with for %p because it's in a hotpage\n", db, db->x64_addr, db->x64_addr+db->x64_size-1, (void*)addr);
+            return NULL;
        }
+        uint32_t hash = X31_hash_code(db->x64_addr, db->x64_size);
+        if(pthread_mutex_trylock(&my_context->mutex_dyndump)) {
+            dynarec_log(LOG_DEBUG, "mutex_dyndump not available when trying to validate block %p from %p:%p (hash:%X) for %p\n", db, db->x64_addr, db->x64_addr+db->x64_size-1, db->hash, (void*)addr);
+            return NULL;
+        }
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x64_addr, db->x64_addr+db->x64_size-1, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            FreeDynablock(db, 0);
+            // start again... (will create a new block)
+            db = internalDBGetBlock(emu, addr, addr, create, 0);
+        } else {
+            db->need_test = 0;
+            dynarec_log(LOG_DEBUG, "Validating block %p from %p:%p (hash:%X) for %p\n", db, db->x64_addr, db->x64_addr+db->x64_size-1, db->hash, (void*)addr);
+            protectDB((uintptr_t)db->x64_addr, db->x64_size);
+            // fill back jumptable
+            setJumpTableIfRef64(db->x64_addr, db->block, db->jmpnext);
+        }
+        pthread_mutex_unlock(&my_context->mutex_dyndump);
    } 
    return db;
 }
@@ -408,26 +251,25 @@ dynablock_t* DBAlternateBlock(x64emu_t* emu, uintptr_t addr, uintptr_t filladdr)
 {
    dynarec_log(LOG_DEBUG, "Creating AlternateBlock at %p for %p\n", (void*)addr, (void*)filladdr);
    int create = 1;
-    dynablock_t *db = internalDBGetBlock(emu, addr, filladdr, create, NULL);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = X31_hash_code(father->x64_addr, father->x64_size);
-        if(hash!=father->hash) {
-            dynarec_log(LOG_DEBUG, "Invalidating alt block %p from %p:%p (hash:%X/%X) with %d son(s) for %p\n", father, father->x64_addr, father->x64_addr+father->x64_size, hash, father->hash, father->sons_size, (void*)addr);
-            // Free father, it's now invalid!
-            FreeDynablock(father);
+    dynablock_t *db = internalDBGetBlock(emu, addr, filladdr, create, 1);
+    if(db && db->done && db->block && db->need_test) {
+        if(pthread_mutex_trylock(&my_context->mutex_dyndump))
+            return NULL;
+        uint32_t hash = X31_hash_code(db->x64_addr, db->x64_size);
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating alt block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x64_addr, db->x64_addr+db->x64_size, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            FreeDynablock(db, 0);
            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, filladdr, create, NULL);
+            db = internalDBGetBlock(emu, addr, filladdr, create, 0);
        } else {
-            father->need_test = 0;
-            lockDB();
-            protectDBnolock((uintptr_t)father->x64_addr, father->x64_size);
+            db->need_test = 0;
+            protectDB((uintptr_t)db->x64_addr, db->x64_size);
            // fill back jumptable
-            addJumpTableIfDefault64(father->x64_addr, father->block);
-            for(int i=0; i<father->sons_size; ++i)
-                addJumpTableIfDefault64(father->sons[i]->x64_addr, father->sons[i]->block);
-            unlockDB();
+            addJumpTableIfDefault64(db->x64_addr, db->block);
        }
+        pthread_mutex_unlock(&my_context->mutex_dyndump);
    } 
    return db;
 }
@@ -1,16 +1,14 @@
 #ifndef __DYNABLOCK_PRIVATE_H_
 #define __DYNABLOCK_PRIVATE_H_

-typedef struct dynablocklist_s  dynablocklist_t;
-
 typedef struct instsize_s {
-    unsigned int x64:4;
-    unsigned int nat:4;
+    unsigned char x64:4;
+    unsigned char nat:4;
 } instsize_t;

 typedef struct dynablock_s {
-    dynablocklist_t* parent;
-    void*           block;
+    void*           block;  // block-sizeof(void*) == self
+    void*           actual_block;   // the actual start of the block (so block-sizeof(void*))
    int             size;
    void*           x64_addr;
    uintptr_t       x64_size;
@@ -20,17 +18,8 @@ typedef struct dynablock_s {
    uint8_t         gone;
    uint8_t         dummy;
    int             isize;
-    dynablock_t**   sons;   // sons (kind-of dummy dynablock...)
-    int             sons_size;
-    dynablock_t*    father; // set only in the case of a son
    instsize_t*     instsize;
+    void*           jmpnext;    // a branch jmpnext code when block is marked
 } dynablock_t;

-typedef struct dynablocklist_s {
-    uintptr_t           text;
-    int                 textsz;
-    int                 maxsz;     // maxblock size (for this block or previous block)
-    dynablock_t**       direct;    // direct mapping (waste of space, so the array is created at first write)
-} dynablocklist_t;
-
 #endif //__DYNABLOCK_PRIVATE_H_
@@ -19,45 +19,49 @@
 #ifdef DYNAREC
 #include "dynablock.h"
 #include "dynablock_private.h"
-#endif
-
-#ifdef DYNAREC
-#ifdef ARM64
-void arm64_prolog(x64emu_t* emu, void* addr) EXPORTDYN;
-void arm64_epilog() EXPORTDYN;
-void arm64_epilog_fast() EXPORTDYN;
-#endif
+#include "bridge.h"
+#include "dynarec_next.h"
 #endif

 #ifdef DYNAREC
 uintptr_t getX64Address(dynablock_t* db, uintptr_t arm_addr);

-void* LinkNext(x64emu_t* emu, uintptr_t addr, void* x2)
+void* LinkNext(x64emu_t* emu, uintptr_t addr, void* x2, uintptr_t* x3)
 {
    #ifdef HAVE_TRACE
    if(!addr) {
        dynablock_t* db = FindDynablockFromNativeAddress(x2-4);
-        printf_log(LOG_NONE, "Warning, jumping to NULL address from %p (db=%p, x64addr=%p)\n", x2-4, db, db?(void*)getX64Address(db, (uintptr_t)x2-4):NULL);
+        printf_log(LOG_NONE, "Warning, jumping to NULL address from %p (db=%p, x64addr=%p/%s)\n", x2-4, db, db?(void*)getX64Address(db, (uintptr_t)x2-4):NULL, db?getAddrFunctionName(getX64Address(db, (uintptr_t)x2-4)):"(nil)");
    }
    #endif
-    dynablock_t* current = NULL;
    void * jblock;
-    dynablock_t* block = DBGetBlock(emu, addr, 1, &current);
+    dynablock_t* block = DBGetBlock(emu, addr, 1);
    if(!block) {
        // no block, let link table as is...
-dynablock_t* db = FindDynablockFromNativeAddress(x2-4);
-printf_log(LOG_NONE, "Warning, jumping to a no-block address %p from %p (db=%p, x64addr=%p)\n", (void*)addr, x2-4, db, db?(void*)getX64Address(db, (uintptr_t)x2-4):NULL);
-        //tableupdate(arm64_epilog, addr, table);
-        return arm64_epilog;
+        if(hasAlternate((void*)addr)) {
+            printf_log(LOG_DEBUG, "Jmp address has alternate: %p", (void*)addr);
+            addr = (uintptr_t)getAlternate((void*)addr);    // set new address
+            R_RIP = addr;   // but also new RIP!
+            *x3 = addr; // and the RIP in x27 register
+            printf_log(LOG_DEBUG, " -> %p\n", (void*)addr);
+            block = DBGetBlock(emu, addr, 1);
+        }
+        if(!block) {
+            #ifdef HAVE_TRACE
+            dynablock_t* db = FindDynablockFromNativeAddress(x2-4);
+            dynarec_log(LOG_INFO, "Warning, jumping to a no-block address %p from %p (db=%p, x64addr=%p)\n", (void*)addr, x2-4, db, db?(void*)getX64Address(db, (uintptr_t)x2-4):NULL);
+            #endif
+            //tableupdate(native_epilog, addr, table);
+            return native_epilog;
+        }
    }
    if(!block->done) {
        // not finished yet... leave linker
-        //tableupdate(arm_linker, addr, table);
-        return arm64_epilog;
+        return native_epilog;
    }
    if(!(jblock=block->block)) {
        // null block, but done: go to epilog, no linker here
-        return arm64_epilog;
+        return native_epilog;
    }
    //dynablock_t *father = block->father?block->father:block;
    return jblock;
@@ -101,23 +105,18 @@ void DynaCall(x64emu_t* emu, uintptr_t addr)
        PushExit(emu);
        R_RIP = addr;
        emu->df = d_none;
-        dynablock_t* block = NULL;
-        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_RIP, 1, &current);
-            current = block;
+            dynablock_t* block = DBGetBlock(emu, R_RIP, 1);
            if(!block || !block->block || !block->done) {
                // no block, of block doesn't have DynaRec content (yet, temp is not null)
                // Use interpreter (should use single instruction step...)
                dynarec_log(LOG_DEBUG, "%04d|Calling Interpretor @%p, emu=%p\n", GetTID(), (void*)R_RIP, emu);
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "%04d|Calling DynaRec Block @%p (%p) of %d x64 instructions (father=%p) emu=%p\n", GetTID(), (void*)R_RIP, block->block, block->isize ,block->father, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Calling DynaRec Block @%p (%p) of %d x64 instructions emu=%p\n", GetTID(), (void*)R_RIP, block->block, block->isize ,emu);
                CHECK_FLAGS(emu);
                // block is here, let's run it!
-                #ifdef ARM64
-                arm64_prolog(emu, block->block);
-                #endif
+                native_prolog(emu, block->block);
            }
            if(emu->fork) {
                int forktype = emu->fork;
@@ -181,22 +180,17 @@ int DynaRun(x64emu_t* emu)
        return Run(emu, 0);
 #ifdef DYNAREC
    else {
-        dynablock_t* block = NULL;
-        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_RIP, 1, &current);
-            current = block;
+            dynablock_t* block = DBGetBlock(emu, R_RIP, 1);
            if(!block || !block->block || !block->done) {
                // no block, of block doesn't have DynaRec content (yet, temp is not null)
                // Use interpreter (should use single instruction step...)
                dynarec_log(LOG_DEBUG, "%04d|Running Interpretor @%p, emu=%p\n", GetTID(), (void*)R_RIP, emu);
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x64 insts (father=%p) emu=%p\n", GetTID(), (void*)R_RIP, block->block, block->isize, block->father, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x64 insts emu=%p\n", GetTID(), (void*)R_RIP, block->block, block->isize, emu);
                // block is here, let's run it!
-                #ifdef ARM64
-                arm64_prolog(emu, block->block);
-                #endif
+                native_prolog(emu, block->block);
            }
            if(emu->fork) {
                int forktype = emu->fork;
@@ -0,0 +1,25 @@
+#ifndef __DYNAREC_ARCH__H_
+#define __DYNAREC_ARCH__H_
+
+#ifdef ARM64
+#include "arm64/arm64_printer.h"
+#include "arm64/dynarec_arm64_private.h"
+#include "arm64/dynarec_arm64_functions.h"
+
+#define instruction_native_t        instruction_arm64_t
+#define dynarec_native_t            dynarec_arm_t
+
+#elif defined(LA464)
+#include "la464/la464_printer.h"
+#include "la464/dynarec_la464_private.h"
+#include "la464/dynarec_la464_functions.h"
+
+#define instruction_native_t        instruction_la464_t
+#define dynarec_native_t            dynarec_la464_t
+
+
+#else
+#error Unsupported platform
+#endif
+
+#endif //__DYNAREC_ARCH__H_
@@ -1,494 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <pthread.h>
-#include <errno.h>
-#include <string.h>
-
-#include "debug.h"
-#include "box64context.h"
-#include "custommem.h"
-#include "dynarec.h"
-#include "emu/x64emu_private.h"
-#include "tools/bridge_private.h"
-#include "x64run.h"
-#include "x64emu.h"
-#include "box64stack.h"
-#include "callback.h"
-#include "emu/x64run_private.h"
-#include "x64trace.h"
-#include "dynablock.h"
-#include "dynablock_private.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "dynarec_arm64_functions.h"
-#include "elfloader.h"
-
-void printf_x64_instruction(zydis_dec_t* dec, instruction_x64_t* inst, const char* name) {
-    uint8_t *ip = (uint8_t*)inst->addr;
-    if(ip[0]==0xcc && ip[1]=='S' && ip[2]=='C') {
-        uintptr_t a = *(uintptr_t*)(ip+3);
-        if(a==0) {
-            dynarec_log(LOG_NONE, "%s%p: Exit x64emu%s\n", (box64_dynarec_dump>1)?"\e[1m":"", (void*)ip, (box64_dynarec_dump>1)?"\e[m":"");
-        } else {
-            dynarec_log(LOG_NONE, "%s%p: Native call to %p%s\n", (box64_dynarec_dump>1)?"\e[1m":"", (void*)ip, (void*)a, (box64_dynarec_dump>1)?"\e[m":"");
-        }
-    } else {
-        if(dec) {
-            dynarec_log(LOG_NONE, "%s%p: %s", (box64_dynarec_dump>1)?"\e[1m":"", ip, DecodeX64Trace(dec, inst->addr));
-        } else {
-            dynarec_log(LOG_NONE, "%s%p: ", (box64_dynarec_dump>1)?"\e[1m":"", ip);
-            for(int i=0; i<inst->size; ++i) {
-                dynarec_log(LOG_NONE, "%02X ", ip[i]);
-            }
-            dynarec_log(LOG_NONE, " %s", name);
-        }
-        // print Call function name if possible
-        if(ip[0]==0xE8 || ip[0]==0xE9) { // Call / Jmp
-            uintptr_t nextaddr = (uintptr_t)ip + 5 + *((int32_t*)(ip+1));
-            printFunctionAddr(nextaddr, "=> ");
-        } else if(ip[0]==0xFF) {
-            if(ip[1]==0x25) {
-                uintptr_t nextaddr = (uintptr_t)ip + 6 + *((int32_t*)(ip+2));
-                printFunctionAddr(nextaddr, "=> ");
-            }
-        }
-        // end of line and colors
-        dynarec_log(LOG_NONE, "%s\n", (box64_dynarec_dump>1)?"\e[m":"");
-    }
-}
-
-void add_next(dynarec_arm_t *dyn, uintptr_t addr) {
-    if(dyn->next_sz == dyn->next_cap) {
-        dyn->next_cap += 16;
-        dyn->next = (uintptr_t*)realloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
-    }
-    for(int i=0; i<dyn->next_sz; ++i)
-        if(dyn->next[i]==addr)
-            return;
-    dyn->next[dyn->next_sz++] = addr;
-}
-uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr) {
-    // get closest, but no addresses befores
-    uintptr_t best = 0;
-    int i = 0;
-    while((i<dyn->next_sz) && (best!=addr)) {
-        if(dyn->next[i]<addr) { // remove the address, it's before current address
-            memmove(dyn->next+i, dyn->next+i+1, (dyn->next_sz-i-1)*sizeof(uintptr_t));
-            --dyn->next_sz;
-        } else {
-            if((dyn->next[i]<best) || !best)
-                best = dyn->next[i];
-            ++i;
-        }
-    }
-    return best;
-}
-#define PK(A) (*((uint8_t*)(addr+(A))))
-int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n)
-{
-    if(!n)
-        return 1;
-    if(PK(0)==0x90)
-        return is_nops(dyn, addr+1, n-1);
-    if(n>1 && PK(0)==0x66)  // if opcode start with 0x66, and there is more after, than is *can* be a NOP
-        return is_nops(dyn, addr+1, n-1);
-    if(n>2 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x00)
-        return is_nops(dyn, addr+3, n-3);
-    if(n>2 && PK(0)==0x8d && PK(1)==0x76 && PK(2)==0x00)    // lea esi, [esi]
-        return is_nops(dyn, addr+3, n-3);
-    if(n>3 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x40 && PK(3)==0x00)
-        return is_nops(dyn, addr+4, n-4);
-    if(n>3 && PK(0)==0x8d && PK(1)==0x74 && PK(2)==0x26 && PK(3)==0x00)
-        return is_nops(dyn, addr+4, n-4);
-    if(n>4 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x44 && PK(3)==0x00 && PK(4)==0x00)
-        return is_nops(dyn, addr+5, n-5);
-    if(n>5 && PK(0)==0x8d && PK(1)==0xb6 && PK(2)==0x00 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00)
-        return is_nops(dyn, addr+6, n-6);
-    if(n>6 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x80 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00)
-        return is_nops(dyn, addr+7, n-7);
-    if(n>6 && PK(0)==0x8d && PK(1)==0xb4 && PK(2)==0x26 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00) // lea esi, [esi+0]
-        return is_nops(dyn, addr+7, n-7);
-    if(n>7 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x84 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00 && PK(7)==0x00)
-        return is_nops(dyn, addr+8, n-8);
-    return 0;
-}
-
-// return size of next instuciton, -1 is unknown
-// not all instrction are setup
-int next_instruction(dynarec_arm_t *dyn, uintptr_t addr)
-{
-    uint8_t opcode = PK(0);
-    uint8_t nextop;
-    switch (opcode) {
-        case 0x66:
-            opcode = PK(1);
-            switch(opcode) {
-                case 0x90:
-                    return 2;
-            }
-            break;
-        case 0x81:
-            nextop = PK(1);
-            return fakeed(dyn, addr+2, 0, nextop)-addr + 4;
-        case 0x83:
-            nextop = PK(1);
-            return fakeed(dyn, addr+2, 0, nextop)-addr + 1;
-        case 0x84:
-        case 0x85:
-        case 0x88:
-        case 0x89:
-        case 0x8A:
-        case 0x8B:
-        case 0x8C:
-        case 0x8D:
-        case 0x8E:
-        case 0x8F:
-            nextop = PK(1);
-            return fakeed(dyn, addr+2, 0, nextop)-addr;
-        case 0x50:
-        case 0x51:
-        case 0x52:
-        case 0x53:
-        case 0x54:
-        case 0x55:
-        case 0x56:
-        case 0x57:
-        case 0x58:
-        case 0x59:
-        case 0x5A:
-        case 0x5B:
-        case 0x5C:
-        case 0x5D:
-        case 0x5E:
-        case 0x5F:
-        case 0x90:
-        case 0x91:
-        case 0x92:
-        case 0x93:
-        case 0x94:
-        case 0x95:
-        case 0x96:
-        case 0x97:
-        case 0x98:
-        case 0x99:
-        case 0x9B:
-        case 0x9C:
-        case 0x9D:
-        case 0x9E:
-        case 0x9F:
-            return 1;
-        case 0xA0:
-        case 0xA1:
-        case 0xA2:
-        case 0xA3:
-            return 5;
-        case 0xB0:
-        case 0xB1:
-        case 0xB2:
-        case 0xB3:
-        case 0xB4:
-        case 0xB5:
-        case 0xB6:
-        case 0xB7:
-            return 2;
-        case 0xB8:
-        case 0xB9:
-        case 0xBA:
-        case 0xBB:
-        case 0xBC:
-        case 0xBD:
-        case 0xBE:
-        case 0xBF:
-            return 5;
-        case 0xFF:
-            nextop = PK(1);
-            switch((nextop>>3)&7) {
-                case 0: // INC Ed
-                case 1: //DEC Ed
-                case 2: // CALL Ed
-                case 4: // JMP Ed
-                case 6: // Push Ed
-                    return fakeed(dyn, addr+2, 0, nextop)-addr;
-            }
-            break;
-        default:
-            break;
-    }
-    return -1;
-}
-#undef PK
-
-int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n)
-{
-    int i = 0;
-    while(i<n) {
-        int j=next_instruction(dyn, addr+i);
-        if(j<=0) return 0;
-        i+=j;
-    }
-    return (i==n)?1:0;
-}
-
-uint32_t needed_flags(dynarec_arm_t *dyn, int ninst, uint32_t setf, int recurse)
-{
-    if(recurse == 10)
-        return X_PEND;
-    if(ninst == dyn->size)
-        return X_PEND; // no more instructions, or too many jmp loop, stop
-    
-    uint32_t needed = dyn->insts[ninst].x64.use_flags;
-    if(needed) {
-        setf &= ~needed;
-        if(!setf)   // all flags already used, no need to continue
-            return needed;
-    }
-
-    if(!needed && !dyn->insts[ninst].x64.set_flags && !dyn->insts[ninst].x64.jmp_insts) {
-        int start = ninst;
-        int end = ninst;
-        while(end<dyn->size && !dyn->insts[end].x64.use_flags && !dyn->insts[end].x64.set_flags && !dyn->insts[end].x64.jmp_insts)
-            ++end;
-        needed = needed_flags(dyn, end, setf, recurse);
-        for(int i=start; i<end; ++i)
-            dyn->insts[i].x64.need_flags = needed;
-        return needed;
-    }
-
-    if(dyn->insts[ninst].x64.set_flags && (dyn->insts[ninst].x64.state_flags!=SF_MAYSET)) {
-        if((setf & ~dyn->insts[ninst].x64.set_flags) == 0)
-            return needed;    // all done, gives all the flags needed
-        setf |= dyn->insts[ninst].x64.set_flags;    // add new flags to continue
-    }
-
-    int jinst = dyn->insts[ninst].x64.jmp_insts;
-    if(dyn->insts[ninst].x64.jmp) {
-        dyn->insts[ninst].x64.need_flags = (jinst==-1)?X_PEND:needed_flags(dyn, jinst, setf, recurse+1);
-        if(dyn->insts[ninst].x64.use_flags)  // conditionnal jump
-             dyn->insts[ninst].x64.need_flags |= needed_flags(dyn, ninst+1, setf, recurse);
-    } else
-        dyn->insts[ninst].x64.need_flags = needed_flags(dyn, ninst+1, setf, recurse);
-    if(dyn->insts[ninst].x64.state_flags==SF_MAYSET)
-        needed |= dyn->insts[ninst].x64.need_flags;
-    else
-        needed |= (dyn->insts[ninst].x64.need_flags & ~dyn->insts[ninst].x64.set_flags);
-    if(needed == (X_PEND|X_ALL))
-        needed = X_ALL;
-    return needed;
-}
-
-instsize_t* addInst(instsize_t* insts, size_t* size, size_t* cap, int x64_size, int arm_size)
-{
-    // x64 instruction is <16 bytes
-    int toadd;
-    if(x64_size>arm_size)
-        toadd = 1 + x64_size/15;
-    else
-        toadd = 1 + arm_size/15;
-    if((*size)+toadd>(*cap)) {
-        *cap = (*size)+toadd;
-        insts = (instsize_t*)realloc(insts, (*cap)*sizeof(instsize_t));
-    }
-    while(toadd) {
-        if(x64_size>15)
-            insts[*size].x64 = 15;    
-        else
-            insts[*size].x64 = x64_size;
-        x64_size -= insts[*size].x64;
-        if(arm_size>15)
-            insts[*size].nat = 15;
-        else
-            insts[*size].nat = arm_size;
-        arm_size -= insts[*size].nat;
-        ++(*size);
-        --toadd;
-    }
-    return insts;
-}
-
-// add a value to table64 (if needed) and gives back the imm19 to use in LDR_literal
-int Table64(dynarec_arm_t *dyn, uint64_t val)
-{
-    // find the value if already present
-    int idx = -1;
-    for(int i=0; i<dyn->table64size && (idx==-1); ++i)
-        if(dyn->table64[i] == val)
-            idx = i;
-    // not found, add it
-    if(idx==-1) {
-        if(dyn->table64size == dyn->table64cap) {
-            dyn->table64cap+=4;
-            dyn->table64 = (uint64_t*)realloc(dyn->table64, dyn->table64cap * sizeof(uint64_t));
-        }
-        idx = dyn->table64size++;
-        dyn->table64[idx] = val;
-    }
-    // calculate offset
-    int delta = dyn->tablestart + idx*sizeof(uint64_t) - (uintptr_t)dyn->block;
-    return delta;
-}
-
-
-uintptr_t arm_pass0(dynarec_arm_t* dyn, uintptr_t addr);
-uintptr_t arm_pass1(dynarec_arm_t* dyn, uintptr_t addr);
-uintptr_t arm_pass2(dynarec_arm_t* dyn, uintptr_t addr);
-uintptr_t arm_pass3(dynarec_arm_t* dyn, uintptr_t addr);
-
-void* FillBlock64(dynablock_t* block, uintptr_t addr) {
-    if(addr>=box64_nodynarec_start && addr<box64_nodynarec_end) {
-        block->done = 1;
-        return (void*)block;
-    }
-    // init the helper
-    dynarec_arm_t helper = {0};
-    helper.start = addr;
-    uintptr_t start = addr;
-    uintptr_t end = arm_pass0(&helper, addr);
-    if(!helper.size) {
-        dynarec_log(LOG_DEBUG, "Warning, null-sized dynarec block (%p)\n", (void*)addr);
-        block->done = 1;
-        return (void*)block;
-    }
-    helper.cap = helper.size+3; // needs epilog handling
-    helper.insts = (instruction_arm64_t*)calloc(helper.cap, sizeof(instruction_arm64_t));
-    // already protect the block and compute hash signature
-    protectDB(addr, end-addr);  //end is 1byte after actual end
-    uint32_t hash = X31_hash_code((void*)addr, end-addr);
-    // pass 1, addresses, x64 jump addresses, flags
-    arm_pass1(&helper, addr);
-    // calculate barriers
-    for(int i=0; i<helper.size; ++i)
-        if(helper.insts[i].x64.jmp) {
-            uintptr_t j = helper.insts[i].x64.jmp;
-            if(j<start || j>=end)
-                helper.insts[i].x64.jmp_insts = -1;
-            else {
-                // find jump address instruction
-                int k=-1;
-                for(int i2=0; i2<helper.size && k==-1; ++i2) {
-                    if(helper.insts[i2].x64.addr==j)
-                        k=i2;
-                }
-                if(k!=-1)   // -1 if not found, mmm, probably wrong, exit anyway
-                    helper.insts[k].x64.barrier = 1;
-                helper.insts[i].x64.jmp_insts = k;
-            }
-        }
-    for(int i=0; i<helper.size; ++i)
-        if(helper.insts[i].x64.set_flags && !helper.insts[i].x64.need_flags) {
-            helper.insts[i].x64.need_flags = needed_flags(&helper, i+1, helper.insts[i].x64.set_flags, 0);
-            if((helper.insts[i].x64.need_flags&X_PEND) && (helper.insts[i].x64.state_flags==SF_MAYSET))
-                helper.insts[i].x64.need_flags = X_ALL;
-        }
-    
-    // pass 2, instruction size
-    arm_pass2(&helper, addr);
-    // ok, now allocate mapped memory, with executable flag on
-    size_t sz = helper.arm_size + helper.table64size*sizeof(uint64_t);
-    void* p = (void*)AllocDynarecMap(block, sz);
-    if(p==NULL) {
-        dynarec_log(LOG_DEBUG, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
-        free(helper.insts);
-        free(helper.next);
-        free(helper.table64);
-        return NULL;
-    }
-    helper.block = p;
-    helper.arm_start = (uintptr_t)p;
-    helper.tablestart = helper.arm_start + helper.arm_size;
-    if(helper.sons_size) {
-        helper.sons_x64 = (uintptr_t*)calloc(helper.sons_size, sizeof(uintptr_t));
-        helper.sons_arm = (void**)calloc(helper.sons_size, sizeof(void*));
-    }
-    // pass 3, emit (log emit arm opcode)
-    if(box64_dynarec_dump) {
-        dynarec_log(LOG_NONE, "%s%04d|Emitting %zu bytes for %u x64 bytes", (box64_dynarec_dump>1)?"\e[01;36m":"", GetTID(), helper.arm_size, helper.isize); 
-        printFunctionAddr(helper.start, " => ");
-        dynarec_log(LOG_NONE, "%s\n", (box64_dynarec_dump>1)?"\e[m":"");
-    }
-    int oldtable64size = helper.table64size;
-    size_t oldarmsize = helper.arm_size;
-    helper.arm_size = 0;
-    helper.table64size = 0; // reset table64 (but not the cap)
-    arm_pass3(&helper, addr);
-    if((oldarmsize!=helper.arm_size) || (oldtable64size<helper.table64size)) {
-        printf_log(LOG_NONE, "BOX64: Warning, size difference in block between pass2 (%zu) & pass3 (%zu)!\n", sz, helper.arm_size+helper.table64size*8);
-        uint8_t *dump = (uint8_t*)helper.start;
-        printf_log(LOG_NONE, "Dump of %d x64 opcodes:\n", helper.size);
-        for(int i=0; i<helper.size; ++i) {
-            printf_log(LOG_NONE, "%p:", dump);
-            for(; dump<(uint8_t*)helper.insts[i+1].x64.addr; ++dump)
-                printf_log(LOG_NONE, " %02X", *dump);
-            printf_log(LOG_NONE, "\t%d -> %d\n", helper.insts[i].size2, helper.insts[i].size);
-        }
-        printf_log(LOG_NONE, "Table64 \t%d -> %d\n", oldtable64size*8, helper.table64size*8);
-        printf_log(LOG_NONE, " ------------\n");
-        //TODO: Cancel block and return empty one
-    }
-    // add table64 if needed
-    if(helper.table64size) {
-        memcpy((void*)helper.tablestart, helper.table64, helper.table64size*8);
-    }
-    // all done...
-    __clear_cache(p, p+sz);   // need to clear the cache before execution...
-    // keep size of instructions for signal handling
-    {
-        size_t cap = 1;
-        for(int i=0; i<helper.size; ++i)
-            cap += 1 + ((helper.insts[i].x64.size>helper.insts[i].size)?helper.insts[i].x64.size:helper.insts[i].size)/15;
-        size_t size = 0;
-        block->instsize = (instsize_t*)calloc(cap, sizeof(instsize_t));
-        for(int i=0; i<helper.size; ++i)
-            block->instsize = addInst(block->instsize, &size, &cap, helper.insts[i].x64.size, helper.insts[i].size/4);
-        block->instsize = addInst(block->instsize, &size, &cap, 0, 0);    // add a "end of block" mark, just in case
-    }
-    // ok, free the helper now
-    free(helper.insts);
-    free(helper.next);
-    free(helper.table64);
-    block->size = sz;
-    block->isize = helper.size;
-    block->block = p;
-    block->need_test = 0;
-    //block->x64_addr = (void*)start;
-    block->x64_size = end-start;
-    block->hash = X31_hash_code(block->x64_addr, block->x64_size);
-    // Check if something changed, to abbort if it as
-    if(block->hash != hash) {
-        dynarec_log(LOG_INFO, "Warning, a block changed while beeing processed hash(%p:%ld)=%x/%x\n", block->x64_addr, block->x64_size, block->hash, hash);
-        free(helper.sons_x64);
-        free(helper.sons_arm);
-        FreeDynarecMap(block, (uintptr_t)p, sz);
-        return NULL;
-    }    // fill sons if any
-    dynablock_t** sons = NULL;
-    int sons_size = 0;
-    if(helper.sons_size) {
-        sons = (dynablock_t**)calloc(helper.sons_size, sizeof(dynablock_t*));
-        for (int i=0; i<helper.sons_size; ++i) {
-            int created = 1;
-            dynablock_t *son = AddNewDynablock(block->parent, helper.sons_x64[i], &created);
-            if(created) {    // avoid breaking a working block!
-                son->block = helper.sons_arm[i];
-                son->x64_addr = (void*)helper.sons_x64[i];
-                son->x64_size = end-helper.sons_x64[i];
-                if(!son->x64_size) {printf_log(LOG_NONE, "Warning, son with null x64 size! (@%p / ARM=%p)", son->x64_addr, son->block);}
-                son->father = block;
-                son->size = sz + son->block - block->block; // update size count, for debugging
-                //son->done = 1;
-                if(!son->parent)
-                    son->parent = block->parent;
-                sons[sons_size] = son;
-                ++sons_size;
-            }
-        }
-        if(sons_size) {
-            block->sons = sons;
-            block->sons_size = sons_size;
-        } else
-            free(sons);
-    }
-    free(helper.sons_x64);
-    free(helper.sons_arm);
-    //block->done = 1;
-    return (void*)block;
-}
@@ -1,167 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <stddef.h>
-#include <pthread.h>
-#include <errno.h>
-
-#include "debug.h"
-#include "box64context.h"
-#include "dynarec.h"
-#include "emu/x64emu_private.h"
-#include "emu/x64run_private.h"
-#include "x64run.h"
-#include "x64emu.h"
-#include "box64stack.h"
-#include "callback.h"
-#include "emu/x64run_private.h"
-#include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
-
-#include "dynarec_arm64_helper.h"
-#include "dynarec_arm64_functions.h"
-
-#define GETGX(a)                        \
-    gd = ((nextop&0x38)>>3)+(rex.r<<3); \
-    a = sse_get_reg(dyn, ninst, x1, gd)
-
-uintptr_t dynarec64_67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
-{
-    (void)ip; (void)need_epilog;
-
-    uint8_t opcode = F8;
-    uint8_t nextop;
-    uint8_t gd, ed;
-    int64_t fixedaddress;
-    int8_t  i8;
-    int32_t i32;
-    int64_t j64;
-    int v0, s0;
-    MAYUSE(i32);
-    MAYUSE(j64);
-    MAYUSE(v0);
-    MAYUSE(s0);
-
-    // REX prefix before the 67 are ignored
-    rex.rex = 0;
-    while(opcode>=0x40 && opcode<=0x4f) {
-        rex.rex = opcode;
-        opcode = F8;
-    }
-    rep = 0;
-    while((opcode==0xF2) || (opcode==0xF3)) {
-        rep = opcode-0xF1;
-        opcode = F8;
-    }
-
-    switch(opcode) {
-
-        case 0x0F:
-            opcode=F8;
-            switch(opcode) {
-
-                case 0x2E:
-                    // no special check...
-                case 0x2F:
-                    if(rep) {
-                        DEFAULT;
-                    } else {
-                        if(opcode==0x2F) {INST_NAME("COMISS Gx, Ex");} else {INST_NAME("UCOMISS Gx, Ex");}
-                        SETFLAGS(X_ALL, SF_SET);
-                        nextop = F8;
-                        GETGX(v0);
-                        if(MODREG) {
-                            s0 = sse_get_reg(dyn, ninst, x1, (nextop&7) + (rex.b<<3));
-                        } else {
-                            s0 = fpu_get_scratch(dyn);
-                            addr = geted32(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                            VLDR32_U12(s0, ed, fixedaddress);
-                        }
-                        FCMPS(v0, s0);
-                        FCOMI(x1, x2);
-                    }
-                    break;
-
-                default:
-                    DEFAULT;
-            }
-            break;
-
-        case 0x8D:
-            INST_NAME("LEA Gd, Ed");
-            nextop=F8;
-            GETGD;
-            if(MODREG) {   // reg <= reg? that's an invalid operation
-                DEFAULT;
-            } else {                    // mem <= reg
-                // should a geted32 be created, to use 32bits regs instead of 64bits?
-                addr = geted32(dyn, addr, ninst, nextop, &ed, gd, &fixedaddress, 0, 0, rex, 0, 0);
-                if(ed!=gd) {
-                    MOVw_REG(gd, ed);
-                }
-            }
-            break;
-
-        #define GO(NO, YES)   \
-            BARRIER(2); \
-            JUMP(addr+i8);\
-            if(dyn->insts) {    \
-                if(dyn->insts[ninst].x64.jmp_insts==-1) {   \
-                    /* out of the block */                  \
-                    i32 = dyn->insts[ninst+1].address-(dyn->arm_size); \
-                    Bcond(NO, i32);     \
-                    jump_to_next(dyn, addr+i8, 0, ninst); \
-                } else {    \
-                    /* inside the block */  \
-                    i32 = dyn->insts[dyn->insts[ninst].x64.jmp_insts].address-(dyn->arm_size);    \
-                    Bcond(YES, i32);    \
-                }   \
-            }
-        case 0xE0:
-            INST_NAME("LOOPNZ (32bits)");
-            READFLAGS(X_ZF);
-            i8 = F8S;
-            MOVw_REG(x1, xRCX);
-            SUBSw_U12(x1, x1, 1);
-            BFIx(xRCX, x1, 0, 32);
-            B_NEXT(cEQ);    // ECX is 0, no LOOP
-            TSTw_mask(xFlags, 0b011010, 0); //mask=0x40
-            GO(cNE, cEQ);
-            break;
-        case 0xE1:
-            INST_NAME("LOOPZ (32bits)");
-            READFLAGS(X_ZF);
-            i8 = F8S;
-            MOVw_REG(x1, xRCX);
-            SUBSw_U12(x1, x1, 1);
-            BFIx(xRCX, x1, 0, 32);
-            B_NEXT(cEQ);    // ECX is 0, no LOOP
-            TSTw_mask(xFlags, 0b011010, 0); //mask=0x40
-            GO(cEQ, cNE);
-            break;
-        case 0xE2:
-            INST_NAME("LOOP (32bits)");
-            i8 = F8S;
-            MOVw_REG(x1, xRCX);
-            SUBSw_U12(x1, x1, 1);
-            BFIx(xRCX, x1, 0, 32);
-            GO(cEQ, cNE);
-            break;
-        case 0xE3:
-            INST_NAME("JECXZ");
-            i8 = F8S;
-            MOVw_REG(x1, xRCX);
-            TSTw_REG(x1, x1);
-            GO(cNE, cEQ);
-            break;
-        #undef GO
-
-        case 0xE8:
-            return dynarec64_00(dyn, addr-1, ip, ninst, rex, rep, ok, need_epilog); // addr-1, to "put back" opcode)
-
-        default:
-            DEFAULT;
-    }
-    return addr;
-}
@@ -1,232 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <stddef.h>
-#include <pthread.h>
-#include <errno.h>
-
-#include "debug.h"
-#include "box64context.h"
-#include "dynarec.h"
-#include "emu/x64emu_private.h"
-#include "emu/x64run_private.h"
-#include "x64run.h"
-#include "x64emu.h"
-#include "box64stack.h"
-#include "callback.h"
-#include "emu/x64run_private.h"
-#include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
-#include "emu/x87emu_private.h"
-
-#include "dynarec_arm64_helper.h"
-#include "dynarec_arm64_functions.h"
-
-
-uintptr_t dynarec64_D8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, rex_t rex, int rep, int* ok, int* need_epilog)
-{
-    (void)ip; (void)rep; (void)need_epilog;
-
-    uint8_t nextop = F8;
-    uint8_t ed;
-    int64_t fixedaddress;
-    int v1, v2;
-    int s0;
-
-    MAYUSE(s0);
-    MAYUSE(v2);
-    MAYUSE(v1);
-
-    switch(nextop) {
-        case 0xC0:
-        case 0xC1:
-        case 0xC2:
-        case 0xC3:
-        case 0xC4:
-        case 0xC5:
-        case 0xC6:
-        case 0xC7:
-            INST_NAME("FADD ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FADDD(v1, v1, v2);
-            break;
-        case 0xC8:
-        case 0xC9:
-        case 0xCA:
-        case 0xCB:
-        case 0xCC:
-        case 0xCD:
-        case 0xCE:
-        case 0xCF:
-            INST_NAME("FMUL ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FMULD(v1, v1, v2);
-            break;
-        case 0xD0:
-        case 0xD1:
-        case 0xD2:
-        case 0xD3:
-        case 0xD4:
-        case 0xD5:
-        case 0xD6:
-        case 0xD7:
-            INST_NAME("FCOM ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
-            FCOM(x1, x2, x3);
-            break;
-        case 0xD8:
-        case 0xD9:
-        case 0xDA:
-        case 0xDB:
-        case 0xDC:
-        case 0xDD:
-        case 0xDE:
-        case 0xDF:
-            INST_NAME("FCOMP ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FCMPD(v1, v2);
-            FCOM(x1, x2, x3);
-            x87_do_pop(dyn, ninst);
-            break;
-        case 0xE0:
-        case 0xE1:
-        case 0xE2:
-        case 0xE3:
-        case 0xE4:
-        case 0xE5:
-        case 0xE6:
-        case 0xE7:
-            INST_NAME("FSUB ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FSUBD(v1, v1, v2);
-            break;
-        case 0xE8:
-        case 0xE9:
-        case 0xEA:
-        case 0xEB:
-        case 0xEC:
-        case 0xED:
-        case 0xEE:
-        case 0xEF:
-            INST_NAME("FSUBR ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FSUBD(v1, v2, v1);
-            break;
-        case 0xF0:
-        case 0xF1:
-        case 0xF2:
-        case 0xF3:
-        case 0xF4:
-        case 0xF5:
-        case 0xF6:
-        case 0xF7:
-            INST_NAME("FDIV ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FDIVD(v1, v1, v2);
-            break;
-        case 0xF8:
-        case 0xF9:
-        case 0xFA:
-        case 0xFB:
-        case 0xFC:
-        case 0xFD:
-        case 0xFE:
-        case 0xFF:
-            INST_NAME("FDIVR ST0, STx");
-            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            FDIVD(v1, v2, v1);
-            break;
-      
-        default:
-            switch((nextop>>3)&7) {
-                case 0:
-                    INST_NAME("FADD ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FADDD(v1, v1, s0);
-                    break;
-                case 1:
-                    INST_NAME("FMUL ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FMULD(v1, v1, s0);
-                    break;
-                case 2:
-                    INST_NAME("FCOM ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FCMPD(v1, s0);
-                    FCOM(x1, x2, x3);
-                    break;
-                case 3:
-                    INST_NAME("FCOMP ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FCMPD(v1, s0);
-                    FCOM(x1, x2, x3);
-                    x87_do_pop(dyn, ninst);
-                    break;
-                case 4:
-                    INST_NAME("FSUB ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FSUBD(v1, v1, s0);
-                    break;
-                case 5:
-                    INST_NAME("FSUBR ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FSUBD(v1, s0, v1);
-                    break;
-                case 6:
-                    INST_NAME("FDIV ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FDIVD(v1, v1, s0);
-                    break;
-                case 7:
-                    INST_NAME("FDIVR ST0, float[ED]");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    s0 = fpu_get_scratch(dyn);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0xfff<<2, 3, rex, 0, 0);
-                    VLDR32_U12(s0, ed, fixedaddress);
-                    FCVT_D_S(s0, s0);
-                    FDIVD(v1, s0, v1);
-                    break;
-                default:
-                    DEFAULT;
-            }
-    }
-    return addr;
-}
@@ -1,465 +0,0 @@
-#define _GNU_SOURCE
-#include <stdio.h>
-#include <stdlib.h>
-#include <pthread.h>
-#include <errno.h>
-#include <string.h>
-#include <math.h>
-#include <signal.h>
-#include <sys/types.h>
-#include <unistd.h>
-
-#include "debug.h"
-#include "box64context.h"
-#include "dynarec.h"
-#include "emu/x64emu_private.h"
-#include "tools/bridge_private.h"
-#include "x64run.h"
-#include "x64emu.h"
-#include "box64stack.h"
-#include "callback.h"
-#include "emu/x64run_private.h"
-#include "emu/x87emu_private.h"
-#include "x64trace.h"
-#include "signals.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "dynarec_arm64_functions.h"
-#include "custommem.h"
-
-void arm_fstp(x64emu_t* emu, void* p)
-{
-    if(ST0.q!=STld(0).uref)
-        D2LD(&ST0.d, p);
-    else
-        memcpy(p, &STld(0).ld, 10);
-}
-
-void arm_print_armreg(x64emu_t* emu, uintptr_t reg, uintptr_t n)
-{
-    (void)emu;
-    dynarec_log(LOG_DEBUG, "R%lu=0x%lx (%lu)\n", n, reg, reg);
-}
-
-void arm_f2xm1(x64emu_t* emu)
-{
-    ST0.d = exp2(ST0.d) - 1.0;
-}
-void arm_fyl2x(x64emu_t* emu)
-{
-    ST(1).d = log2(ST0.d)*ST(1).d;
-}
-void arm_ftan(x64emu_t* emu)
-{
-    ST0.d = tan(ST0.d);
-    emu->sw.f.F87_C2 = 0;
-}
-void arm_fpatan(x64emu_t* emu)
-{
-    ST1.d = atan2(ST1.d, ST0.d);
-}
-void arm_fxtract(x64emu_t* emu)
-{
-    int32_t tmp32s = (ST1.q&0x7ff0000000000000LL)>>52;
-    tmp32s -= 1023;
-    ST1.d /= exp2(tmp32s);
-    ST0.d = tmp32s;
-}
-void arm_fprem(x64emu_t* emu)
-{
-    int32_t tmp32s = ST0.d / ST1.d;
-    ST0.d -= ST1.d * tmp32s;
-    emu->sw.f.F87_C2 = 0;
-    emu->sw.f.F87_C0 = (tmp32s&1);
-    emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
-    emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
-}
-void arm_fyl2xp1(x64emu_t* emu)
-{
-    ST(1).d = log2(ST0.d + 1.0)*ST(1).d;
-}
-void arm_fsincos(x64emu_t* emu)
-{
-    sincos(ST1.d, &ST1.d, &ST0.d);
-    emu->sw.f.F87_C2 = 0;
-}
-void arm_frndint(x64emu_t* emu)
-{
-    ST0.d = fpu_round(emu, ST0.d);
-}
-void arm_fscale(x64emu_t* emu)
-{
-    if(ST0.d!=0.0)
-        ST0.d *= exp2(trunc(ST1.d));
-}
-void arm_fsin(x64emu_t* emu)
-{
-    ST0.d = sin(ST0.d);
-    emu->sw.f.F87_C2 = 0;
-}
-void arm_fcos(x64emu_t* emu)
-{
-    ST0.d = cos(ST0.d);
-    emu->sw.f.F87_C2 = 0;
-}
-
-void arm_fbld(x64emu_t* emu, uint8_t* ed)
-{
-    fpu_fbld(emu, ed);
-}
-
-void arm_fild64(x64emu_t* emu, int64_t* ed)
-{
-    int64_t tmp;
-    memcpy(&tmp, ed, sizeof(tmp));
-    ST0.d = tmp;
-    STll(0).sq = tmp;
-    STll(0).sref = ST0.sq;
-}
-
-void arm_fbstp(x64emu_t* emu, uint8_t* ed)
-{
-    fpu_fbst(emu, ed);
-}
-
-void arm_fistp64(x64emu_t* emu, int64_t* ed)
-{
-    // used of memcpy to avoid aligments issues
-    if(STll(0).sref==ST(0).sq) {
-        memcpy(ed, &STll(0).sq, sizeof(int64_t));
-    } else {
-        int64_t tmp;
-        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL) || !isfinite(ST0.d))
-            tmp = 0x8000000000000000LL;
-        else
-            tmp = fpu_round(emu, ST0.d);
-        memcpy(ed, &tmp, sizeof(tmp));
-    }
-}
-
-void arm_fistt64(x64emu_t* emu, int64_t* ed)
-{
-    // used of memcpy to avoid aligments issues
-    int64_t tmp = ST0.d;
-    memcpy(ed, &tmp, sizeof(tmp));
-}
-
-void arm_fld(x64emu_t* emu, uint8_t* ed)
-{
-    memcpy(&STld(0).ld, ed, 10);
-    LD2D(&STld(0), &ST(0).d);
-    STld(0).uref = ST0.q;
-}
-
-void arm_ud(x64emu_t* emu)
-{
-    emit_signal(emu, SIGILL, (void*)R_RIP, 0);
-}
-
-void arm_fsave(x64emu_t* emu, uint8_t* ed)
-{
-    fpu_savenv(emu, (char*)ed, 0);
-
-    uint8_t* p = ed;
-    p += 28;
-    for (int i=0; i<8; ++i) {
-        LD2D(p, &ST(i).d);
-        p+=10;
-    }
-}
-void arm_frstor(x64emu_t* emu, uint8_t* ed)
-{
-    fpu_loadenv(emu, (char*)ed, 0);
-
-    uint8_t* p = ed;
-    p += 28;
-    for (int i=0; i<8; ++i) {
-        D2LD(&ST(i).d, p);
-        p+=10;
-    }
-
-}
-
-void arm_fprem1(x64emu_t* emu)
-{
-    // simplified version
-    int32_t tmp32s = round(ST0.d / ST1.d);
-    ST0.d -= ST1.d*tmp32s;
-    emu->sw.f.F87_C2 = 0;
-    emu->sw.f.F87_C0 = (tmp32s&1);
-    emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
-    emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
-}
-
-static uint8_t ff_mult(uint8_t a, uint8_t b)
-{
-	int retval = 0;
-
-	for(int i = 0; i < 8; i++) {
-		if((b & 1) == 1) 
-			retval ^= a;
-		
-		if((a & 0x80)) {
-			a <<= 1;
-			a  ^= 0x1b;
-		} else {
-			a <<= 1;
-		}
-		
-		b >>= 1;
-	}
-	
-	return retval;
-}
-
-void arm_aesimc(x64emu_t* emu, int xmm)
-{
-    sse_regs_t eax1 = emu->xmm[xmm];
-
-    for(int j=0; j<4; ++j) {
-        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x0E, eax1.ub[0+j*4]) ^ ff_mult(0x0B, eax1.ub[1+j*4]) ^ ff_mult(0x0D, eax1.ub[2+j*4]) ^ ff_mult(0x09, eax1.ub[3+j*4]);
-        emu->xmm[xmm].ub[1+j*4] = ff_mult(0x09, eax1.ub[0+j*4]) ^ ff_mult(0x0E, eax1.ub[1+j*4]) ^ ff_mult(0x0B, eax1.ub[2+j*4]) ^ ff_mult(0x0D, eax1.ub[3+j*4]);
-        emu->xmm[xmm].ub[2+j*4] = ff_mult(0x0D, eax1.ub[0+j*4]) ^ ff_mult(0x09, eax1.ub[1+j*4]) ^ ff_mult(0x0E, eax1.ub[2+j*4]) ^ ff_mult(0x0B, eax1.ub[3+j*4]);
-        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x0B, eax1.ub[0+j*4]) ^ ff_mult(0x0D, eax1.ub[1+j*4]) ^ ff_mult(0x09, eax1.ub[2+j*4]) ^ ff_mult(0x0E, eax1.ub[3+j*4]);
-    }
-}
-void arm_aesmc(x64emu_t* emu, int xmm)
-{
-    sse_regs_t eax1 = emu->xmm[xmm];
-
-    for(int j=0; j<4; ++j) {
-        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x02, eax1.ub[0+j*4]) ^ ff_mult(0x03, eax1.ub[1+j*4]) ^               eax1.ub[2+j*4]  ^               eax1.ub[3+j*4] ;
-        emu->xmm[xmm].ub[1+j*4] =               eax1.ub[0+j*4]  ^ ff_mult(0x02, eax1.ub[1+j*4]) ^ ff_mult(0x03, eax1.ub[2+j*4]) ^               eax1.ub[3+j*4] ;
-        emu->xmm[xmm].ub[2+j*4] =               eax1.ub[0+j*4]  ^               eax1.ub[1+j*4]  ^ ff_mult(0x02, eax1.ub[2+j*4]) ^ ff_mult(0x03, eax1.ub[3+j*4]);
-        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x03, eax1.ub[0+j*4]) ^               eax1.ub[1+j*4]  ^               eax1.ub[2+j*4]  ^ ff_mult(0x02, eax1.ub[3+j*4]);
-    }
-}
-void arm_aesdlast(x64emu_t* emu, int xmm)
-{
-                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
-                            //   A  N  K  H  E  B  O  L  I  F  C  P  M  J  G  D
-    const uint8_t invshiftrows[] = {0,13,10, 7, 4, 1,14,11, 8, 5, 2,15,12, 9, 6, 3};
-    const uint8_t invsubbytes[256] = {
-        0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38, 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
-        0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87, 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
-        0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d, 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
-        0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2, 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
-        0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
-        0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda, 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
-        0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a, 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
-        0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02, 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
-        0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea, 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
-        0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85, 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
-        0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89, 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
-        0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20, 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
-        0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31, 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
-        0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d, 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
-        0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0, 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
-        0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26, 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
-    };
-
-    sse_regs_t eax1;
-    for(int i=0; i<16; ++i)
-        eax1.ub[i] = emu->xmm[xmm].ub[invshiftrows[i]];
-    //STATE ← InvSubBytes( STATE );
-    for(int i=0; i<16; ++i)
-        emu->xmm[xmm].ub[i] = invsubbytes[eax1.ub[i]];
-
-}
-void arm_aeselast(x64emu_t* emu, int xmm)
-{
-                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
-                            //   A  F  K  P  E  J  O  D  I  N  C  H  M  B  G  L
-    const uint8_t shiftrows[] = {0, 5,10,15, 4, 9,14, 3, 8,13, 2, 7,12, 1, 6,11};
-    const uint8_t subbytes[256] = {
-        0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
-        0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
-        0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
-        0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
-        0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
-        0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
-        0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
-        0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
-        0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
-        0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
-        0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
-        0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
-        0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
-        0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
-        0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
-        0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
-    };
-
-    sse_regs_t eax1;
-    for(int i=0; i<16; ++i)
-        eax1.ub[i] = emu->xmm[xmm].ub[shiftrows[i]];
-    //STATE ← SubBytes( STATE );
-    for(int i=0; i<16; ++i)
-        emu->xmm[xmm].ub[i] = subbytes[eax1.ub[i]];
-}
-void arm_aesd(x64emu_t* emu, int xmm)
-{
-    arm_aesdlast(emu, xmm);
-    arm_aesimc(emu, xmm);
-}
-void arm_aese(x64emu_t* emu, int xmm)
-{
-    arm_aeselast(emu, xmm);
-    arm_aesmc(emu, xmm);
-}
-
-
-#define XMM0    0
-#define XMM8    16
-#define X870    8
-#define EMM0    8
-#define SCRATCH0    24
-
-// Get a FPU scratch reg
-int fpu_get_scratch(dynarec_arm_t* dyn)
-{
-    return SCRATCH0 + dyn->fpu_scratch++;  // return an Sx
-}
-// Reset scratch regs counter
-void fpu_reset_scratch(dynarec_arm_t* dyn)
-{
-    dyn->fpu_scratch = 0;
-}
-// Get a x87 double reg
-int fpu_get_reg_x87(dynarec_arm_t* dyn)
-{
-    int i=X870;
-    while (dyn->fpuused[i]) ++i;
-    dyn->fpuused[i] = 1;
-    return i; // return a Dx
-}
-// Free a FPU double reg
-void fpu_free_reg(dynarec_arm_t* dyn, int reg)
-{
-    // TODO: check upper limit?
-    dyn->fpuused[reg] = 0;
-}
-// Get an MMX double reg
-int fpu_get_reg_emm(dynarec_arm_t* dyn, int emm)
-{
-    dyn->fpuused[EMM0 + emm] = 1;
-    return EMM0 + emm;
-}
-// Get an XMM quad reg
-int fpu_get_reg_xmm(dynarec_arm_t* dyn, int xmm)
-{
-    if(xmm>7) {
-        dyn->fpuused[XMM8 + xmm - 8] = 1;
-        return XMM8 + xmm - 8;
-    } else {
-        dyn->fpuused[XMM0 + xmm] = 1;
-        return XMM0 + xmm;
-    }
-}
-// Reset fpu regs counter
-void fpu_reset_reg(dynarec_arm_t* dyn)
-{
-    dyn->fpu_reg = 0;
-    for (int i=0; i<32; ++i)
-        dyn->fpuused[i]=0;
-}
-
-#define F8      *(uint8_t*)(addr++)
-#define F32     *(uint32_t*)(addr+=4, addr-4)
-#define F32S64  (uint64_t)(int64_t)*(int32_t*)(addr+=4, addr-4)
-// Get if ED will have the correct parity. Not emiting anything. Parity is 2 for DWORD or 3 for QWORD
-int getedparity(dynarec_arm_t* dyn, int ninst, uintptr_t addr, uint8_t nextop, int parity, int delta)
-{
-    (void)dyn; (void)ninst;
-
-    uint32_t tested = (1<<parity)-1;
-    if((nextop&0xC0)==0xC0)
-        return 0;   // direct register, no parity...
-    if(!(nextop&0xC0)) {
-        if((nextop&7)==4) {
-            uint8_t sib = F8;
-            int sib_reg = (sib>>3)&7;
-            if((sib&0x7)==5) {
-                uint64_t tmp = F32S64;
-                if (sib_reg!=4) {
-                    // if XXXXXX+reg<<N then check parity of XXXXX and N should be enough
-                    return ((tmp&tested)==0 && (sib>>6)>=parity)?1:0;
-                } else {
-                    // just a constant...
-                    return (tmp&tested)?0:1;
-                }
-            } else {
-                if(sib_reg==4 && parity<3)
-                    return 0;   // simple [reg]
-                // don't try [reg1 + reg2<<N], unless reg1 is ESP
-                return ((sib&0x7)==4 && (sib>>6)>=parity)?1:0;
-            }
-        } else if((nextop&7)==5) {
-            uint64_t tmp = F32S64;
-            tmp+=addr+delta;
-            return (tmp&tested)?0:1;
-        } else {
-            return 0;
-        }
-    } else {
-        return 0; //Form [reg1 + reg2<<N + XXXXXX]
-    }
-}
-
-// Do the GETED, but don't emit anything...
-uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop) 
-{
-    (void)dyn; (void)addr; (void)ninst;
-
-    if((nextop&0xC0)==0xC0)
-        return addr;
-    if(!(nextop&0xC0)) {
-        if((nextop&7)==4) {
-            uint8_t sib = F8;
-            if((sib&0x7)==5) {
-                addr+=4;
-            }
-        } else if((nextop&7)==5) {
-            addr+=4;
-        }
-    } else {
-        if((nextop&7)==4) {
-            ++addr;
-        }
-        if(nextop&0x80) {
-            addr+=4;
-        } else {
-            ++addr;
-        }
-    }
-    return addr;
-}
-#undef F8
-#undef F32
-
-int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn)
-{
-    (void)dyn;
-
-#define PK(a)       *(uint8_t*)(addr+a)
-#define PK32(a)     *(int32_t*)(addr+a)
-
-    if(!addr || !getProtection(addr))
-        return 0;
-    if(PK(0)==0xff && PK(1)==0x25) {            // "absolute" jump, maybe the GOT (well, RIP relative in fact)
-        uintptr_t a1 = addr + 6 + (PK32(2));    // need to add a check to see if the address is from the GOT !
-        addr = *(uintptr_t*)a1; 
-    }
-    if(!addr || !getProtection(addr))
-        return 0;
-    onebridge_t *b = (onebridge_t*)(addr);
-    if(b->CC==0xCC && b->S=='S' && b->C=='C' && b->w!=(wrapper_t)0 && b->f!=(uintptr_t)PltResolver) {
-        // found !
-        if(retn) *retn = (b->C3==0xC2)?b->N:0;
-        if(calladdress) *calladdress = addr+1;
-        return 1;
-    }
-    return 0;
-#undef PK32
-#undef PK
-}
@@ -1,140 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <stddef.h>
-#include <pthread.h>
-#include <errno.h>
-
-#include "debug.h"
-#include "box64context.h"
-#include "dynarec.h"
-#include "emu/x64emu_private.h"
-#include "emu/x64run_private.h"
-#include "x64run.h"
-#include "x64emu.h"
-#include "box64stack.h"
-#include "emu/x64run_private.h"
-#include "x64trace.h"
-#include "dynarec_arm64.h"
-#include "dynarec_arm64_private.h"
-#include "arm64_printer.h"
-#include "dynarec_arm64_functions.h"
-#include "dynarec_arm64_helper.h"
-#include "custommem.h"
-
-#ifndef STEP
-#error No STEP defined
-#endif
-
-uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
-{
-    int ok = 1;
-    int ninst = 0;
-    uintptr_t ip = addr;
-    uintptr_t init_addr = addr;
-    rex_t rex;
-    int rep;    // 0 none, 1=F2 prefix, 2=F3 prefix
-    int need_epilog = 1;
-    dyn->sons_size = 0;
-    // Clean up (because there are multiple passes)
-    dyn->state_flags = 0;
-    dyn->dfnone = 0;
-    dyn->last_ip = ip;  // RIP is always set at start of block!
-    MAYUSE(init_addr);
-    fpu_reset(dyn, ninst);
-    // ok, go now
-    INIT;
-    while(ok) {
-        ip = addr;
-        if(dyn->insts && (dyn->insts[ninst].x64.barrier==1)) {
-            dyn->last_ip = 0;
-            NEW_BARRIER_INST;
-        }
-        NEW_INST;
-        fpu_reset_scratch(dyn);
-#ifdef HAVE_TRACE
-        if(my_context->dec && box64_dynarec_trace) {
-        if((trace_end == 0) 
-            || ((ip >= trace_start) && (ip < trace_end)))  {
-                MESSAGE(LOG_DUMP, "TRACE ----\n");
-                fpu_reflectcache(dyn, ninst, x1, x2, x3);
-                GETIP(ip);
-                MOVx_REG(x1, xRIP);
-                STORE_XEMU_CALL(xRIP);
-                MOV32w(x2, 1);
-                CALL(PrintTrace, -1);
-                LOAD_XEMU_CALL(xRIP);
-                MESSAGE(LOG_DUMP, "----------\n");
-            }
-        }
-#endif
-
-        rep = 0;
-        uint8_t pk = PK(0);
-        while((pk==0xF2) || (pk==0xF3)) {
-            rep = pk-0xF1;
-            ++addr;
-            pk = PK(0);
-        }
-        while(pk==0x3E) {   //Branch Taken Hint ignored
-            ++addr;
-            pk = PK(0);
-        }
-        rex.rex = 0;
-        while(pk>=0x40 && pk<=0x4f) {
-            rex.rex = pk;
-            ++addr;
-            pk = PK(0);
-        }
-
-        addr = dynarec64_00(dyn, addr, ip, ninst, rex, rep, &ok, &need_epilog);
-
-        INST_EPILOG;
-
-        if(dyn->insts && dyn->insts[ninst+1].x64.barrier) {
-            fpu_purgecache(dyn, ninst, x1, x2, x3);
-            if(dyn->insts[ninst+1].x64.barrier!=2) {
-                dyn->state_flags = 0;
-                dyn->dfnone = 0;
-            }
-        }
-        if(!ok && !need_epilog && dyn->insts && (addr < (dyn->start+dyn->isize))) {
-            ok = 1;
-        }
-        if(!ok && !need_epilog && !dyn->insts && getProtection(addr+3)&~PROT_CUSTOM)
-            if(*(uint32_t*)addr!=0) {   // check if need to continue (but is next 4 bytes are 0, stop)
-                uintptr_t next = get_closest_next(dyn, addr);
-                if(next && (
-                    (((next-addr)<15) && is_nops(dyn, addr, next-addr)) 
-                    ||(((next-addr)<30) && is_instructions(dyn, addr, next-addr)) ))
-                {
-                    dynarec_log(LOG_DEBUG, "Extend block %p, %p -> %p (ninst=%d)\n", dyn, (void*)addr, (void*)next, ninst);
-                    ok = 1;
-                } else if(next && (next-addr)<30) {
-                    dynarec_log(LOG_DEBUG, "Cannot extend block %p -> %p (%02X %02X %02X %02X %02X %02X %02X %02x)\n", (void*)addr, (void*)next, PK(0), PK(1), PK(2), PK(3), PK(4), PK(5), PK(6), PK(7));
-                }
-            }
-        if(ok<0)  {ok = 0; need_epilog=1;}
-        ++ninst;
-        #if STEP == 0
-        if(ok && !isJumpTableDefault64((void*)addr))
-        #else
-        if(ok && dyn->insts && (ninst==dyn->size))
-        #endif
-        {
-            #if STEP == 3
-            dynarec_log(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size); 
-            #endif
-            BARRIER(2);
-            fpu_purgecache(dyn, ninst, x1, x2, x3);
-            jump_to_next(dyn, addr, 0, ninst);
-            ok=0; need_epilog=0;
-        }
-    }
-    if(need_epilog) {
-        fpu_purgecache(dyn, ninst, x1, x2, x3);
-        jump_to_epilog(dyn, ip, 0, ninst);  // no linker here, it's an unknow instruction
-    }
-    FINI;
-    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d, %d sons)\n", dyn->size, dyn->sons_size);
-    return addr;
-}
@@ -1,25 +0,0 @@
-
-#define INIT    uintptr_t sav_addr=addr
-#define FINI    dyn->isize = addr-sav_addr
-#define MESSAGE(A, ...)  
-#define SETFLAGS(A, B)  
-#define READFLAGS(A)    
-#define EMIT(A)     
-#define JUMP(A)         add_next(dyn, (uintptr_t)A)
-#define NEW_INST        ++dyn->size
-#define INST_EPILOG         
-#define INST_NAME(name) 
-#define DEFAULT                         \
-        --dyn->size;                    \
-        *ok = -1;                       \
-        if(box64_dynarec_log>=LOG_INFO) {\
-        dynarec_log(LOG_NONE, "%p: Dynarec stopped because of Opcode %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X", \
-        (void*)ip, PKip(0),             \
-        PKip(1), PKip(2), PKip(3),      \
-        PKip(4), PKip(5), PKip(6),      \
-        PKip(7), PKip(8), PKip(9),      \
-        PKip(10),PKip(11),PKip(12),     \
-        PKip(13),PKip(14));             \
-        printFunctionAddr(ip, " => ");  \
-        dynarec_log(LOG_NONE, "\n");    \
-        }
@@ -1,17 +0,0 @@
-#define INIT    
-#define FINI     \
-    dyn->insts[ninst].x64.addr = addr; \
-    if(ninst) dyn->insts[ninst-1].x64.size = dyn->insts[ninst].x64.addr - dyn->insts[ninst-1].x64.addr;
-#define MESSAGE(A, ...)  
-#define EMIT(A)     
-#define READFLAGS(A)    dyn->insts[ninst].x64.use_flags = A
-#define SETFLAGS(A,B)   {dyn->insts[ninst].x64.set_flags = A; dyn->insts[ninst].x64.state_flags = B;}
-#define JUMP(A)         dyn->insts[ninst].x64.jmp = A
-#define BARRIER(A)      dyn->insts[ninst].x64.barrier = A
-#define BARRIER_NEXT(A) if(ninst<dyn->size) dyn->insts[ninst+1].x64.barrier = A
-
-#define NEW_INST \
-    dyn->insts[ninst].x64.addr = ip; \
-    if(ninst) dyn->insts[ninst-1].x64.size = dyn->insts[ninst].x64.addr - dyn->insts[ninst-1].x64.addr;
-#define INST_EPILOG 
-#define INST_NAME(name)  
@@ -1,11 +0,0 @@
-#define INIT        dyn->arm_size = 0
-#define FINI        if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
-
-#define MESSAGE(A, ...)  
-#define EMIT(A)     dyn->insts[ninst].size+=4; dyn->arm_size+=4
-#define NEW_INST    if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
-#define INST_EPILOG dyn->insts[ninst].epilog = dyn->arm_size; 
-#define INST_NAME(name) 
-#define NEW_BARRIER_INST    if(ninst) ++dyn->sons_size
-#define TABLE64(A, V)   {Table64(dyn, (V)); EMIT(0);}
-#define FTABLE64(A, V)  {mmx87_regs_t v = {.d = V}; Table64(dyn, v.q); EMIT(0);}
@@ -1,36 +0,0 @@
-#define INIT    
-#define FINI
-#define EMIT(A)     \
-    if(box64_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (uint32_t)(A), arm64_print(A, (uintptr_t)dyn->block));} \
-    *(uint32_t*)(dyn->block) = (uint32_t)(A);       \
-    dyn->block += 4; dyn->arm_size += 4;\
-    dyn->insts[ninst].size2 += 4
-
-#define MESSAGE(A, ...)  if(box64_dynarec_dump) dynarec_log(LOG_NONE, __VA_ARGS__)
-#define NEW_INST        
-#define INST_EPILOG     
-#define INST_NAME(name) \
-    if(box64_dynarec_dump) {\
-        printf_x64_instruction(my_context->dec, &dyn->insts[ninst].x64, name); \
-        dynarec_log(LOG_NONE, "%s%p: %d emited opcodes, state=%d/%d, set=%X, use=%X, need=%X%s\n", \
-            (box64_dynarec_dump>1)?"\e[32m":"", \
-            (void*)(dyn->arm_start+dyn->insts[ninst].address),  \
-            dyn->insts[ninst].size/4,           \
-            dyn->insts[ninst].x64.state_flags,  \
-            dyn->state_flags,                   \
-            dyn->insts[ninst].x64.set_flags,    \
-            dyn->insts[ninst].x64.use_flags,    \
-            dyn->insts[ninst].x64.need_flags,   \
-            (box64_dynarec_dump>1)?"\e[m":"");  \
-    }
-
-#define NEW_BARRIER_INST                            \
-    if(ninst) {                                     \
-    dyn->sons_x64[dyn->sons_size] = (uintptr_t)ip;  \
-    dyn->sons_arm[dyn->sons_size] = dyn->block;     \
-    MESSAGE(LOG_DUMP, "----> potential Son here\n");\
-    ++dyn->sons_size;                               \
-    }
-
-#define TABLE64(A, V)   {int val64offset = Table64(dyn, (V)); MESSAGE(LOG_DUMP, "  Table64: 0x%lx\n", (V)); LDRx_literal(A, val64offset);}
-#define FTABLE64(A, V)  {mmx87_regs_t v = {.d = V}; int val64offset = Table64(dyn, v.q); MESSAGE(LOG_DUMP, "  FTable64: %g\n", v.d); VLDR64_literal(A, val64offset);}
@@ -1,62 +0,0 @@
-#ifndef __DYNAREC_ARM_PRIVATE_H_
-#define __DYNAREC_ARM_PRIVATE_H_
-
-#include "dynarec_private.h"
-
-typedef struct x64emu_s x64emu_t;
-
-typedef struct instruction_arm_s {
-    instruction_x64_t   x64;
-    uintptr_t           address;    // (start) address of the arm emited instruction
-    uintptr_t           epilog;     // epilog of current instruction (can be start of next, of barrier stuff)
-    int                 size;       // size of the arm emited instruction
-    int                 size2;      // size of the arm emited instrucion after pass2
-    uintptr_t           mark, mark2, mark3;
-    uintptr_t           markf;
-    uintptr_t           markseg;
-    uintptr_t           marklock;
-    int                 pass2choice;// value for choices that are fixed on pass2 for pass3
-    uintptr_t           natcall;
-    int                 retn;
-} instruction_arm64_t;
-
-typedef struct dynarec_arm_s {
-    instruction_arm64_t *insts;
-    int32_t             size;
-    int32_t             cap;
-    uintptr_t           start;      // start of the block
-    uint32_t            isize;      // size in byte of x64 instructions included
-    void*               block;      // memory pointer where next instruction is emited
-    uintptr_t           arm_start;  // start of the arm code
-    size_t              arm_size;   // size of emitted arm code
-    int                 state_flags;// actual state for on-demand flags
-    uintptr_t           last_ip;    // last set IP in RIP (or NULL if unclean state)
-    int8_t              x87cache[8];// cache status for the 8 x87 register behind the fpu stack
-    int8_t              x87reg[8];  // reg used for x87cache entry
-    int8_t              mmxcache[8];// cache status for the 8 MMX registers
-    int8_t              ssecache[16];// cache status for the 16 SSE(2) registers
-    int8_t              fpuused[32];// all 8..31 Q reg from fpu, used by x87, sse and mmx
-    int                 x87stack;   // cache stack counter
-    int                 fpu_scratch;// scratch counter
-    int                 fpu_reg;    // x87/sse/mmx reg counter
-    int                 dfnone;     // if defered flags is already set to df_none
-    uint64_t            *table64;   // table of 64bits value
-    int                 table64size;// size of table (will be appended at end of executable code)
-    int                 table64cap;
-    uintptr_t           tablestart;
-    uintptr_t*          next;       // variable array of "next" jump address
-    int                 next_sz;
-    int                 next_cap;
-    uintptr_t*          sons_x64;   // the x64 address of potential dynablock sons
-    void**              sons_arm;   // the arm address of potential dynablock sons
-    int                 sons_size;  // number of potential dynablock sons
-} dynarec_arm_t;
-
-void add_next(dynarec_arm_t *dyn, uintptr_t addr);
-uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr);
-int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n);
-int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n);
-
-int Table64(dynarec_arm_t *dyn, uint64_t val);  // add a value to etable64 (if needed) and gives back the imm19 to use in LDR_literal
-
-#endif //__DYNAREC_ARM_PRIVATE_H_
@@ -0,0 +1,12 @@
+#ifndef __DYNAREC_HELPER__H_
+#define __DYNAREC_HELPER__H_
+
+#ifdef ARM64
+#include "arm64/dynarec_arm64_helper.h"
+#elif defined(LA464)
+#include "la464/dynarec_la464_helper.h"
+#else
+#error Unsupported architecture
+#endif
+
+#endif //__DYNAREC_HELPER__H_
@@ -0,0 +1,599 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+#include <string.h>
+#include <assert.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "custommem.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "tools/bridge_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "callback.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "dynablock.h"
+#include "dynablock_private.h"
+#include "elfloader.h"
+
+#include "dynarec_native.h"
+#include "dynarec_arch.h"
+#include "dynarec_next.h"
+
+void printf_x64_instruction(zydis_dec_t* dec, instruction_x64_t* inst, const char* name) {
+    uint8_t *ip = (uint8_t*)inst->addr;
+    if(ip[0]==0xcc && ip[1]=='S' && ip[2]=='C') {
+        uintptr_t a = *(uintptr_t*)(ip+3);
+        if(a==0) {
+            dynarec_log(LOG_NONE, "%s%p: Exit x64emu%s\n", (box64_dynarec_dump>1)?"\e[01;33m":"", (void*)ip, (box64_dynarec_dump>1)?"\e[m":"");
+        } else {
+            dynarec_log(LOG_NONE, "%s%p: Native call to %p%s\n", (box64_dynarec_dump>1)?"\e[01;33m":"", (void*)ip, (void*)a, (box64_dynarec_dump>1)?"\e[m":"");
+        }
+    } else {
+        if(dec) {
+            dynarec_log(LOG_NONE, "%s%p: %s", (box64_dynarec_dump>1)?"\e[01;33m":"", ip, DecodeX64Trace(dec, inst->addr));
+        } else {
+            dynarec_log(LOG_NONE, "%s%p: ", (box64_dynarec_dump>1)?"\e[01;33m":"", ip);
+            for(int i=0; i<inst->size; ++i) {
+                dynarec_log(LOG_NONE, "%02X ", ip[i]);
+            }
+            dynarec_log(LOG_NONE, " %s", name);
+        }
+        // print Call function name if possible
+        if(ip[0]==0xE8 || ip[0]==0xE9) { // Call / Jmp
+            uintptr_t nextaddr = (uintptr_t)ip + 5 + *((int32_t*)(ip+1));
+            printFunctionAddr(nextaddr, "=> ");
+        } else if(ip[0]==0xFF) {
+            if(ip[1]==0x25) {
+                uintptr_t nextaddr = (uintptr_t)ip + 6 + *((int32_t*)(ip+2));
+                printFunctionAddr(nextaddr, "=> ");
+            }
+        }
+        // end of line and colors
+        dynarec_log(LOG_NONE, "%s\n", (box64_dynarec_dump>1)?"\e[m":"");
+    }
+}
+
+void add_next(dynarec_native_t *dyn, uintptr_t addr) {
+    if(!box64_dynarec_bigblock)
+        return;
+    // exist?
+    for(int i=0; i<dyn->next_sz; ++i)
+        if(dyn->next[i]==addr)
+            return;
+    // put in a free slot
+    for(int i=0; i<dyn->next_sz; ++i)
+        if(!dyn->next[i]) {
+            dyn->next[i] = addr;
+            return;
+        }
+    // add slots
+    if(dyn->next_sz == dyn->next_cap) {
+        dyn->next_cap += 64;
+        dyn->next = (uintptr_t*)customRealloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
+    }
+    dyn->next[dyn->next_sz++] = addr;
+}
+uintptr_t get_closest_next(dynarec_native_t *dyn, uintptr_t addr) {
+    // get closest, but no addresses befores
+    uintptr_t best = 0;
+    int i = 0;
+    while((i<dyn->next_sz) && (best!=addr)) {
+        if(dyn->next[i]<addr) { // remove the address, it's before current address
+            dyn->next[i] = 0;
+        } else {
+            if((dyn->next[i]<best) || !best)
+                best = dyn->next[i];
+        }
+        ++i;
+    }
+    return best;
+}
+#define PK(A) (*((uint8_t*)(addr+(A))))
+int is_nops(dynarec_native_t *dyn, uintptr_t addr, int n)
+{
+    if(!n)
+        return 1;
+    if(PK(0)==0x90)
+        return is_nops(dyn, addr+1, n-1);
+    if(n>1 && PK(0)==0x66)  // if opcode start with 0x66, and there is more after, than is *can* be a NOP
+        return is_nops(dyn, addr+1, n-1);
+    if(n>2 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x00)
+        return is_nops(dyn, addr+3, n-3);
+    if(n>2 && PK(0)==0x8d && PK(1)==0x76 && PK(2)==0x00)    // lea esi, [esi]
+        return is_nops(dyn, addr+3, n-3);
+    if(n>3 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x40 && PK(3)==0x00)
+        return is_nops(dyn, addr+4, n-4);
+    if(n>3 && PK(0)==0x8d && PK(1)==0x74 && PK(2)==0x26 && PK(3)==0x00)
+        return is_nops(dyn, addr+4, n-4);
+    if(n>4 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x44 && PK(3)==0x00 && PK(4)==0x00)
+        return is_nops(dyn, addr+5, n-5);
+    if(n>5 && PK(0)==0x8d && PK(1)==0xb6 && PK(2)==0x00 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00)
+        return is_nops(dyn, addr+6, n-6);
+    if(n>6 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x80 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00)
+        return is_nops(dyn, addr+7, n-7);
+    if(n>6 && PK(0)==0x8d && PK(1)==0xb4 && PK(2)==0x26 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00) // lea esi, [esi+0]
+        return is_nops(dyn, addr+7, n-7);
+    if(n>7 && PK(0)==0x0f && PK(1)==0x1f && PK(2)==0x84 && PK(3)==0x00 && PK(4)==0x00 && PK(5)==0x00 && PK(6)==0x00 && PK(7)==0x00)
+        return is_nops(dyn, addr+8, n-8);
+    return 0;
+}
+
+// return size of next instuciton, -1 is unknown
+// not all instrction are setup
+int next_instruction(dynarec_native_t *dyn, uintptr_t addr)
+{
+    uint8_t opcode = PK(0);
+    uint8_t nextop;
+    switch (opcode) {
+        case 0x66:
+            opcode = PK(1);
+            switch(opcode) {
+                case 0x90:
+                    return 2;
+            }
+            break;
+        case 0x81:
+            nextop = PK(1);
+            return fakeed(dyn, addr+2, 0, nextop)-addr + 4;
+        case 0x83:
+            nextop = PK(1);
+            return fakeed(dyn, addr+2, 0, nextop)-addr + 1;
+        case 0x84:
+        case 0x85:
+        case 0x88:
+        case 0x89:
+        case 0x8A:
+        case 0x8B:
+        case 0x8C:
+        case 0x8D:
+        case 0x8E:
+        case 0x8F:
+            nextop = PK(1);
+            return fakeed(dyn, addr+2, 0, nextop)-addr;
+        case 0x50:
+        case 0x51:
+        case 0x52:
+        case 0x53:
+        case 0x54:
+        case 0x55:
+        case 0x56:
+        case 0x57:
+        case 0x58:
+        case 0x59:
+        case 0x5A:
+        case 0x5B:
+        case 0x5C:
+        case 0x5D:
+        case 0x5E:
+        case 0x5F:
+        case 0x90:
+        case 0x91:
+        case 0x92:
+        case 0x93:
+        case 0x94:
+        case 0x95:
+        case 0x96:
+        case 0x97:
+        case 0x98:
+        case 0x99:
+        case 0x9B:
+        case 0x9C:
+        case 0x9D:
+        case 0x9E:
+        case 0x9F:
+            return 1;
+        case 0xA0:
+        case 0xA1:
+        case 0xA2:
+        case 0xA3:
+            return 5;
+        case 0xB0:
+        case 0xB1:
+        case 0xB2:
+        case 0xB3:
+        case 0xB4:
+        case 0xB5:
+        case 0xB6:
+        case 0xB7:
+            return 2;
+        case 0xB8:
+        case 0xB9:
+        case 0xBA:
+        case 0xBB:
+        case 0xBC:
+        case 0xBD:
+        case 0xBE:
+        case 0xBF:
+            return 5;
+        case 0xFF:
+            nextop = PK(1);
+            switch((nextop>>3)&7) {
+                case 0: // INC Ed
+                case 1: //DEC Ed
+                case 2: // CALL Ed
+                case 4: // JMP Ed
+                case 6: // Push Ed
+                    return fakeed(dyn, addr+2, 0, nextop)-addr;
+            }
+            break;
+        default:
+            break;
+    }
+    return -1;
+}
+#undef PK
+
+int is_instructions(dynarec_native_t *dyn, uintptr_t addr, int n)
+{
+    int i = 0;
+    while(i<n) {
+        int j=next_instruction(dyn, addr+i);
+        if(j<=0) return 0;
+        i+=j;
+    }
+    return (i==n)?1:0;
+}
+
+instsize_t* addInst(instsize_t* insts, size_t* size, size_t* cap, int x64_size, int native_size)
+{
+    // x64 instruction is <16 bytes
+    int toadd;
+    if(x64_size>native_size)
+        toadd = 1 + x64_size/15;
+    else
+        toadd = 1 + native_size/15;
+    if((*size)+toadd>(*cap)) {
+        *cap = (*size)+toadd;
+        insts = (instsize_t*)customRealloc(insts, (*cap)*sizeof(instsize_t));
+    }
+    while(toadd) {
+        if(x64_size>15)
+            insts[*size].x64 = 15;    
+        else
+            insts[*size].x64 = x64_size;
+        x64_size -= insts[*size].x64;
+        if(native_size>15)
+            insts[*size].nat = 15;
+        else
+            insts[*size].nat = native_size;
+        native_size -= insts[*size].nat;
+        ++(*size);
+        --toadd;
+    }
+    return insts;
+}
+
+// add a value to table64 (if needed) and gives back the imm19 to use in LDR_literal
+int Table64(dynarec_native_t *dyn, uint64_t val)
+{
+    // find the value if already present
+    int idx = -1;
+    for(int i=0; i<dyn->table64size && (idx==-1); ++i)
+        if(dyn->table64[i] == val)
+            idx = i;
+    // not found, add it
+    if(idx==-1) {
+        if(dyn->table64size == dyn->table64cap) {
+            dyn->table64cap+=16;
+            dyn->table64 = (uint64_t*)customRealloc(dyn->table64, dyn->table64cap * sizeof(uint64_t));
+        }
+        idx = dyn->table64size++;
+        dyn->table64[idx] = val;
+    }
+    // calculate offset
+    int delta = dyn->tablestart + idx*sizeof(uint64_t) - (uintptr_t)dyn->block;
+    return delta;
+}
+
+static void fillPredecessors(dynarec_native_t* dyn)
+{
+    int pred_sz = 1;    // to be safe
+    // compute total size of predecessor to alocate the array
+    // first compute the jumps
+    for(int i=0; i<dyn->size; ++i) {
+        if(dyn->insts[i].x64.jmp && (dyn->insts[i].x64.jmp_insts!=-1)) {
+            pred_sz++;
+            dyn->insts[dyn->insts[i].x64.jmp_insts].pred_sz++;
+        }
+    }
+    // second the "has_next"
+    for(int i=0; i<dyn->size-1; ++i) {
+        if(dyn->insts[i].x64.has_next) {
+            pred_sz++;
+            dyn->insts[i+1].pred_sz++;
+        }
+    }
+    dyn->predecessor = (int*)customMalloc(pred_sz*sizeof(int));
+    // fill pred pointer
+    int* p = dyn->predecessor;
+    for(int i=0; i<dyn->size; ++i) {
+        dyn->insts[i].pred = p;
+        p += dyn->insts[i].pred_sz;
+        dyn->insts[i].pred_sz=0;  // reset size, it's reused to actually fill pred[]
+    }
+    // fill pred
+    for(int i=0; i<dyn->size; ++i) {
+        if((i!=dyn->size-1) && dyn->insts[i].x64.has_next)
+            dyn->insts[i+1].pred[dyn->insts[i+1].pred_sz++] = i;
+        if(dyn->insts[i].x64.jmp && (dyn->insts[i].x64.jmp_insts!=-1)) {
+            int j = dyn->insts[i].x64.jmp_insts;
+            dyn->insts[j].pred[dyn->insts[j].pred_sz++] = i;
+        }
+    }
+
+}
+
+// updateNeed goes backward, from last intruction to top
+static int updateNeed(dynarec_arm_t* dyn, int ninst, uint8_t need) {
+    while (ninst>=0) {
+        // need pending but instruction is only a subset: remove pend and use an X_ALL instead
+        need |= dyn->insts[ninst].x64.need_after;
+        if((need&X_PEND) && (dyn->insts[ninst].x64.state_flags==SF_SUBSET)) {
+            need &=~X_PEND;
+            need |= X_ALL;
+        }
+        if((need&X_PEND) && (dyn->insts[ninst].x64.state_flags==SF_SET)) {
+            need &=~X_PEND;
+            need |= dyn->insts[ninst].x64.set_flags;    // SF_SET will compute all flags, it's not SUBSET!
+        }
+        if((need&X_PEND) && dyn->insts[ninst].x64.state_flags==SF_SUBSET_PENDING) {
+            need |= X_ALL&~(dyn->insts[ninst].x64.set_flags);
+        }
+        dyn->insts[ninst].x64.gen_flags = need&dyn->insts[ninst].x64.set_flags;
+        if((need&X_PEND) && (dyn->insts[ninst].x64.state_flags&SF_PENDING))
+            dyn->insts[ninst].x64.gen_flags |= X_PEND;
+        dyn->insts[ninst].x64.need_after = need;
+        need = dyn->insts[ninst].x64.need_after&~dyn->insts[ninst].x64.gen_flags;
+        if(dyn->insts[ninst].x64.may_set)
+            need |= dyn->insts[ninst].x64.gen_flags;    // forward the flags
+        // Consume X_PEND if relevant
+        if((need&X_PEND) && (dyn->insts[ninst].x64.set_flags&SF_PENDING))
+            need &=~X_PEND;
+        need |= dyn->insts[ninst].x64.use_flags;
+        if(dyn->insts[ninst].x64.need_before == need)
+            return ninst - 1;
+        dyn->insts[ninst].x64.need_before = need;
+        if(dyn->insts[ninst].x64.barrier&BARRIER_FLAGS) {
+            need = need?X_PEND:0;
+        }
+        int ok = 0;
+        for(int i=0; i<dyn->insts[ninst].pred_sz; ++i) {
+            if(dyn->insts[ninst].pred[i] == ninst-1)
+                ok = 1;
+            else
+                updateNeed(dyn, dyn->insts[ninst].pred[i], need);
+        }
+        if(!ok)
+            return ninst - 1;
+        --ninst;
+    }
+    return ninst;
+}
+
+__thread void* current_helper = NULL;
+
+void CancelBlock64()
+{
+    dynarec_native_t* helper = (dynarec_native_t*)current_helper;
+    current_helper = NULL;
+    if(!helper)
+        return;
+    customFree(helper->next);
+    customFree(helper->insts);
+    customFree(helper->instsize);
+    customFree(helper->predecessor);
+    customFree(helper->table64);
+    if(helper->dynablock && helper->dynablock->actual_block)
+        FreeDynarecMap(helper->dynablock, (uintptr_t)helper->dynablock->actual_block, helper->dynablock->size);
+    else if(helper->dynablock && helper->block)
+        FreeDynarecMap(helper->dynablock, (uintptr_t)helper->block-sizeof(void*), helper->dynablock->size);
+}
+
+uintptr_t native_pass0(dynarec_native_t* dyn, uintptr_t addr);
+uintptr_t native_pass1(dynarec_native_t* dyn, uintptr_t addr);
+uintptr_t native_pass2(dynarec_native_t* dyn, uintptr_t addr);
+uintptr_t native_pass3(dynarec_native_t* dyn, uintptr_t addr);
+
+void* CreateEmptyBlock(dynablock_t* block, uintptr_t addr) {
+    block->isize = 0;
+    block->done = 0;
+    size_t sz = 4*sizeof(void*);
+    void* actual_p = (void*)AllocDynarecMap(block, sz);
+    void* p = actual_p + sizeof(void*);
+    if(actual_p==NULL) {
+        dynarec_log(LOG_INFO, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
+        CancelBlock64();
+        return NULL;
+    }
+    block->size = sz;
+    block->actual_block = actual_p;
+    block->block = p;
+    block->jmpnext = p;
+    *(dynablock_t**)actual_p = block;
+    *(void**)(p+2*sizeof(void*)) = native_epilog;
+    CreateJmpNext(block->jmpnext, p+2*sizeof(void*));
+    block->need_test = 0;
+    // all done...
+    __clear_cache(actual_p, actual_p+sz);   // need to clear the cache before execution...
+    return block;
+}
+
+void* FillBlock64(dynablock_t* block, uintptr_t addr) {
+    if(IsInHotPage(addr)) {
+        dynarec_log(LOG_DEBUG, "Cancelling dynarec FillBlock on hotpage for %p\n", (void*)addr);
+        return NULL;
+    }
+    if(addr>=box64_nodynarec_start && addr<box64_nodynarec_end) {
+        dynarec_log(LOG_INFO, "Create empty block in no-dynarec zone\n");
+        return CreateEmptyBlock(block, addr);
+    }
+    // protect the 1st page
+    protectDB(addr, 1);
+    // init the helper
+    dynarec_native_t helper = {0};
+    current_helper = &helper;
+    helper.dynablock = block;
+    helper.start = addr;
+    uintptr_t start = addr;
+    helper.cap = 128;
+    helper.insts = (instruction_native_t*)customCalloc(helper.cap, sizeof(instruction_native_t));
+    // pass 0, addresses, x64 jump addresses, overall size of the block
+    uintptr_t end = native_pass0(&helper, addr);
+    // no need for next anymore
+    customFree(helper.next);
+    helper.next_sz = helper.next_cap = 0;
+    helper.next = NULL;
+    // basic checks
+    if(!helper.size) {
+        dynarec_log(LOG_INFO, "Warning, null-sized dynarec block (%p)\n", (void*)addr);
+        CancelBlock64();
+        return CreateEmptyBlock(block, addr);;
+    }
+    if(!isprotectedDB(addr, 1)) {
+        dynarec_log(LOG_INFO, "Warning, write on current page on pass0, aborting dynablock creation (%p)\n", (void*)addr);
+        CancelBlock64();
+        return NULL;
+    }
+    // protect the block of it goes over the 1st page
+    if((addr&~0xfff)!=(end&~0xfff)) // need to protect some other pages too
+        protectDB(addr, end-addr);  //end is 1byte after actual end
+    // compute hash signature
+    uint32_t hash = X31_hash_code((void*)addr, end-addr);
+    // calculate barriers
+    for(int i=0; i<helper.size; ++i)
+        if(helper.insts[i].x64.jmp) {
+            uintptr_t j = helper.insts[i].x64.jmp;
+            if(j<start || j>=end) {
+                helper.insts[i].x64.jmp_insts = -1;
+                helper.insts[i].x64.need_after |= X_PEND;
+            } else {
+                // find jump address instruction
+                int k=-1;
+                for(int i2=0; i2<helper.size && k==-1; ++i2) {
+                    if(helper.insts[i2].x64.addr==j)
+                        k=i2;
+                }
+                if(k!=-1 && !helper.insts[i].barrier_maybe)
+                    helper.insts[k].x64.barrier |= BARRIER_FULL;
+                helper.insts[i].x64.jmp_insts = k;
+            }
+        }
+    // fill predecessors with the jump address
+    fillPredecessors(&helper);
+
+    int pos = helper.size;
+    while (pos>=0)
+        pos = updateNeed(&helper, pos, 0);
+
+    // pass 1, float optimisations, first pass for flags
+    native_pass1(&helper, addr);
+    
+    // pass 2, instruction size
+    native_pass2(&helper, addr);
+    // keep size of instructions for signal handling
+    size_t insts_rsize = 0;
+    {
+        size_t insts_size = 0;
+        size_t cap = 1;
+        for(int i=0; i<helper.size; ++i)
+            cap += 1 + ((helper.insts[i].x64.size>helper.insts[i].size)?helper.insts[i].x64.size:helper.insts[i].size)/15;
+        helper.instsize = (instsize_t*)customCalloc(cap, sizeof(instsize_t));
+        for(int i=0; i<helper.size; ++i)
+            helper.instsize = addInst(helper.instsize, &insts_size, &cap, helper.insts[i].x64.size, helper.insts[i].size/4);
+        helper.instsize = addInst(helper.instsize, &insts_size, &cap, 0, 0);    // add a "end of block" mark, just in case
+        insts_rsize = insts_size*sizeof(instsize_t);
+    }
+    insts_rsize = (insts_rsize+7)&~7;   // round the size...
+    // ok, now allocate mapped memory, with executable flag on
+    size_t sz = sizeof(void*) + helper.native_size + helper.table64size*sizeof(uint64_t) + 4*sizeof(void*) + insts_rsize;
+    //           dynablock_t*     block (arm insts)            table64                       jmpnext code       instsize
+    void* actual_p = (void*)AllocDynarecMap(block, sz);
+    void* p = actual_p + sizeof(void*);
+    void* next = p + helper.native_size + helper.table64size*sizeof(uint64_t);
+    void* instsize = next + 4*sizeof(void*);
+    if(actual_p==NULL) {
+        dynarec_log(LOG_INFO, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
+        CancelBlock64();
+        return NULL;
+    }
+    helper.block = p;
+    helper.native_start = (uintptr_t)p;
+    helper.tablestart = helper.native_start + helper.native_size;
+    *(dynablock_t**)actual_p = block;
+    // pass 3, emit (log emit native opcode)
+    if(box64_dynarec_dump) {
+        dynarec_log(LOG_NONE, "%s%04d|Emitting %zu bytes for %u x64 bytes", (box64_dynarec_dump>1)?"\e[01;36m":"", GetTID(), helper.native_size, helper.isize); 
+        printFunctionAddr(helper.start, " => ");
+        dynarec_log(LOG_NONE, "%s\n", (box64_dynarec_dump>1)?"\e[m":"");
+    }
+    int oldtable64size = helper.table64size;
+    size_t oldnativesize = helper.native_size;
+    helper.native_size = 0;
+    helper.table64size = 0; // reset table64 (but not the cap)
+    native_pass3(&helper, addr);
+    if((oldnativesize!=helper.native_size) || (oldtable64size<helper.table64size)) {
+        printf_log(LOG_NONE, "BOX64: Warning, size difference in block between pass2 (%zu) & pass3 (%zu)!\n", sz, helper.native_size+helper.table64size*8);
+        uint8_t *dump = (uint8_t*)helper.start;
+        printf_log(LOG_NONE, "Dump of %d x64 opcodes:\n", helper.size);
+        for(int i=0; i<helper.size; ++i) {
+            printf_log(LOG_NONE, "%p:", dump);
+            for(; dump<(uint8_t*)helper.insts[i+1].x64.addr; ++dump)
+                printf_log(LOG_NONE, " %02X", *dump);
+            printf_log(LOG_NONE, "\t%d -> %d\n", helper.insts[i].size2, helper.insts[i].size);
+        }
+        printf_log(LOG_NONE, "Table64 \t%d -> %d\n", oldtable64size*8, helper.table64size*8);
+        printf_log(LOG_NONE, " ------------\n");
+        //TODO: Cancel block and return empty one
+    }
+    // add table64 if needed
+    if(helper.table64size) {
+        memcpy((void*)helper.tablestart, helper.table64, helper.table64size*8);
+    }
+    // keep size of instructions for signal handling
+    memcpy(instsize, helper.instsize, insts_rsize);
+    block->instsize = instsize;
+    // ok, free the helper now
+    customFree(helper.insts);
+    helper.insts = NULL;
+    customFree(helper.table64);
+    helper.table64 = NULL;
+    customFree(helper.instsize);
+    helper.instsize = NULL;
+    customFree(helper.predecessor);
+    helper.predecessor = NULL;
+    block->size = sz;
+    block->isize = helper.size;
+    block->actual_block = actual_p;
+    block->block = p;
+    block->jmpnext = next+sizeof(void*);
+    *(dynablock_t**)next = block;
+    *(void**)(next+2*sizeof(void*)) = native_next;
+    CreateJmpNext(block->jmpnext, next+2*sizeof(void*));
+    block->need_test = 0;
+    //block->x64_addr = (void*)start;
+    block->x64_size = end-start;
+    // all done...
+    __clear_cache(actual_p, actual_p+sz);   // need to clear the cache before execution...
+    block->hash = X31_hash_code(block->x64_addr, block->x64_size);
+    // Check if something changed, to abbort if it as
+    if((block->hash != hash)) {
+        dynarec_log(LOG_INFO, "Warning, a block changed while beeing processed hash(%p:%ld)=%x/%x\n", block->x64_addr, block->x64_size, block->hash, hash);
+        CancelBlock64();
+        return NULL;
+    }
+    if(!isprotectedDB(addr, end-addr)) {
+        dynarec_log(LOG_INFO, "Warning, block unprotected while beeing processed %p:%ld, cancelling\n", block->x64_addr, block->x64_size);
+        CancelBlock64();
+        return NULL;
+        //protectDB(addr, end-addr);
+    }
+    current_helper = NULL;
+    //block->done = 1;
+    return (void*)block;
+}
@@ -0,0 +1,252 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box64context.h"
+#include "dynarec.h"
+#include "emu/x64emu_private.h"
+#include "emu/x64run_private.h"
+#include "x64run.h"
+#include "x64emu.h"
+#include "box64stack.h"
+#include "emu/x64run_private.h"
+#include "x64trace.h"
+#include "dynablock.h"
+#include "dynarec_native.h"
+#include "custommem.h"
+#include "elfloader.h"
+
+#include "dynarec_arch.h"
+#include "dynarec_helper.h"
+
+#ifndef STEP
+#error No STEP defined
+#endif
+
+uintptr_t native_pass(dynarec_native_t* dyn, uintptr_t addr)
+{
+    int ok = 1;
+    int ninst = 0;
+    int j64;
+    uintptr_t ip = addr;
+    uintptr_t init_addr = addr;
+    rex_t rex;
+    int rep;    // 0 none, 1=F2 prefix, 2=F3 prefix
+    int need_epilog = 1;
+    // Clean up (because there are multiple passes)
+    dyn->f.pending = 0;
+    dyn->f.dfnone = 0;
+    fpu_reset(dyn);
+    int reset_n = -1;
+    dyn->last_ip = (dyn->insts && dyn->insts[0].pred_sz)?0:ip;  // RIP is always set at start of block unless there is a predecessor!
+    int stopblock = 2+(FindElfAddress(my_context, addr)?0:1); // if block is in elf_memory, it can be extended with bligblocks==2, else it needs 3
+    // ok, go now
+    INIT;
+    while(ok) {
+        ip = addr;
+        if (reset_n!=-1) {
+            dyn->last_ip = 0;
+            if(reset_n==-2) {
+                MESSAGE(LOG_DEBUG, "Reset Caches to zero\n");
+                dyn->f.dfnone = 0;
+                dyn->f.pending = 0;
+                fpu_reset(dyn);
+            } else {
+                MESSAGE(LOG_DEBUG, "Reset Caches with %d\n",reset_n);
+                #if STEP > 1
+                // for STEP 2 & 3, just need to refrest with current, and undo the changes (push & swap)
+                dyn->n = dyn->insts[ninst].n;
+                neoncacheUnwind(&dyn->n);
+                #ifdef HAVE_TRACE
+                if(box64_dynarec_dump)
+                    if(memcmp(&dyn->n, &dyn->insts[reset_n].n, sizeof(neon_cache_t))) {
+                        MESSAGE(LOG_DEBUG, "Warning, difference in neoncache: reset=");
+                        for(int i=0; i<24; ++i)
+                            if(dyn->insts[reset_n].n.neoncache[i].v)
+                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[reset_n].n.neoncache[i].t, dyn->insts[reset_n].n.neoncache[i].n));
+                        if(dyn->insts[reset_n].n.combined1 || dyn->insts[reset_n].n.combined2)
+                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[reset_n].n.swapped?"SWP":"CMB", dyn->insts[reset_n].n.combined1, dyn->insts[reset_n].n.combined2);
+                        if(dyn->insts[reset_n].n.stack_push || dyn->insts[reset_n].n.stack_pop)
+                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[reset_n].n.stack_push, -dyn->insts[reset_n].n.stack_pop);
+                        MESSAGE(LOG_DEBUG, " ==> ");
+                        for(int i=0; i<24; ++i)
+                            if(dyn->insts[ninst].n.neoncache[i].v)
+                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[ninst].n.neoncache[i].t, dyn->insts[ninst].n.neoncache[i].n));
+                        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
+                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
+                        if(dyn->insts[ninst].n.stack_push || dyn->insts[ninst].n.stack_pop)
+                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
+                        MESSAGE(LOG_DEBUG, " -> ");
+                        for(int i=0; i<24; ++i)
+                            if(dyn->n.neoncache[i].v)
+                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->n.neoncache[i].t, dyn->n.neoncache[i].n));
+                        if(dyn->n.combined1 || dyn->n.combined2)
+                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->n.swapped?"SWP":"CMB", dyn->n.combined1, dyn->n.combined2);
+                        if(dyn->n.stack_push || dyn->n.stack_pop)
+                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->n.stack_push, -dyn->n.stack_pop);
+                        MESSAGE(LOG_DEBUG, "\n");
+                    }
+                #endif //HAVE_TRACE
+                #else
+                dyn->n = dyn->insts[reset_n].n;
+                #endif
+                dyn->f = dyn->insts[reset_n].f_exit;
+                if(dyn->insts[ninst].x64.barrier&BARRIER_FLOAT) {
+                    MESSAGE(LOG_DEBUG, "Apply Barrier Float\n");
+                    fpu_reset(dyn);
+                }
+                if(dyn->insts[ninst].x64.barrier&BARRIER_FLAGS) {
+                    MESSAGE(LOG_DEBUG, "Apply Barrier Flags\n");
+                    dyn->f.dfnone = 0;
+                    dyn->f.pending = 0;
+                }
+            }
+            reset_n = -1;
+        } else if(ninst && (dyn->insts[ninst].pred_sz!=1 || dyn->insts[ninst].pred[0]!=ninst-1))
+            dyn->last_ip = 0;   // reset IP if some jump are comming here
+        // propagate ST stack state, especial stack pop that are defered
+        if(dyn->n.stack_pop) {
+            for(int j=0; j<24; ++j)
+                if((dyn->n.neoncache[j].t == NEON_CACHE_ST_D || dyn->n.neoncache[j].t == NEON_CACHE_ST_F)) {
+                    if(dyn->n.neoncache[j].n<dyn->n.stack_pop)
+                        dyn->n.neoncache[j].v = 0;
+                    else
+                        dyn->n.neoncache[j].n-=dyn->n.stack_pop;
+                }
+            dyn->n.stack_pop = 0;
+        }
+        dyn->n.stack = dyn->n.stack_next;
+        dyn->n.news = 0;
+        dyn->n.stack_push = 0;
+        dyn->n.swapped = 0;
+        NEW_INST;
+        fpu_reset_scratch(dyn);
+        if((dyn->insts[ninst].x64.need_before&~X_PEND) && !dyn->insts[ninst].pred_sz) {
+            READFLAGS(dyn->insts[ninst].x64.need_before&~X_PEND);
+        }
+#ifdef HAVE_TRACE
+        if(my_context->dec && box64_dynarec_trace) {
+        if((trace_end == 0) 
+            || ((ip >= trace_start) && (ip < trace_end)))  {
+                MESSAGE(LOG_DUMP, "TRACE ----\n");
+                fpu_reflectcache(dyn, ninst, x1, x2, x3);
+                GO_TRACE();
+                MESSAGE(LOG_DUMP, "----------\n");
+            }
+        }
+#endif
+
+        rep = 0;
+        uint8_t pk = PK(0);
+        while((pk==0xF2) || (pk==0xF3)) {
+            rep = pk-0xF1;
+            ++addr;
+            pk = PK(0);
+        }
+        while(pk==0x3E || pk==0x26) {   //Branch Taken Hint ignored, same for ES: prefix
+            ++addr;
+            pk = PK(0);
+        }
+        rex.rex = 0;
+        while(pk>=0x40 && pk<=0x4f) {
+            rex.rex = pk;
+            ++addr;
+            pk = PK(0);
+        }
+
+        addr = dynarec64_00(dyn, addr, ip, ninst, rex, rep, &ok, &need_epilog);
+
+        INST_EPILOG;
+
+        int next = ninst+1;
+        #if STEP > 0
+        if(!dyn->insts[ninst].x64.has_next && dyn->insts[ninst].x64.jmp && dyn->insts[ninst].x64.jmp_insts!=-1)
+            next = dyn->insts[ninst].x64.jmp_insts;
+        #endif
+        if(dyn->insts[ninst].x64.has_next && dyn->insts[next].x64.barrier) {
+            if(dyn->insts[next].x64.barrier&BARRIER_FLOAT)
+                fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+            if(dyn->insts[next].x64.barrier&BARRIER_FLAGS) {
+                dyn->f.pending = 0;
+                dyn->f.dfnone = 0;
+                dyn->last_ip = 0;
+            }
+        }
+        #if STEP != 0
+        if(!ok && !need_epilog && (addr < (dyn->start+dyn->isize))) {
+            ok = 1;
+            // we use the 1st predecessor here
+            int ii = ninst+1;
+            while(ii<dyn->size && !dyn->insts[ii].pred_sz) 
+                ++ii;
+            if((dyn->insts[ii].x64.barrier&BARRIER_FULL)==BARRIER_FULL)
+                reset_n = -2;    // hack to say Barrier!
+            else {
+                reset_n = getNominalPred(dyn, ii);  // may get -1 if no predecessor are availble
+                if(reset_n==-1) {
+                    reset_n = -2;
+                    MESSAGE(LOG_DEBUG, "Warning, Reset Caches mark not found\n");
+                }
+            }
+        }
+        #else
+        if(!ok && !need_epilog && box64_dynarec_bigblock && getProtection(addr+3)&~PROT_CUSTOM)
+            if(*(uint32_t*)addr!=0) {   // check if need to continue (but is next 4 bytes are 0, stop)
+                uintptr_t next = get_closest_next(dyn, addr);
+                if(next && (
+                    (((next-addr)<15) && is_nops(dyn, addr, next-addr)) 
+                    /*||(((next-addr)<30) && is_instructions(dyn, addr, next-addr))*/ ))
+                {
+                    ok = 1;
+                    // need to find back that instruction to copy the caches, as previous version cannot be used anymore
+                    reset_n = -2;
+                    for(int ii=0; ii<ninst; ++ii)
+                        if(dyn->insts[ii].x64.jmp == next) {
+                            reset_n = ii;
+                            ii=ninst;
+                        }
+                    if(box64_dynarec_dump) dynarec_log(LOG_NONE, "Extend block %p, %p -> %p (ninst=%d, jump from %d)\n", dyn, (void*)addr, (void*)next, ninst, reset_n);
+                } else if(next && (next-addr)<30) {
+                    if(box64_dynarec_dump) dynarec_log(LOG_NONE, "Cannot extend block %p -> %p (%02X %02X %02X %02X %02X %02X %02X %02x)\n", (void*)addr, (void*)next, PK(0), PK(1), PK(2), PK(3), PK(4), PK(5), PK(6), PK(7));
+                }
+            }
+        #endif
+        if(ok<0)  {ok = 0; need_epilog=1;}
+        ++ninst;
+        #if STEP == 0
+        if(ok && (((box64_dynarec_bigblock<stopblock) && !isJumpTableDefault64((void*)addr)) 
+            || (addr>=box64_nodynarec_start && addr<box64_nodynarec_end)))
+        #else
+        if(ok && (ninst==dyn->size))
+        #endif
+        {
+            int j32;
+            MAYUSE(j32);
+            MESSAGE(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size);
+            if(!box64_dynarec_dump && addr>=box64_nodynarec_start && addr<box64_nodynarec_end)
+                dynarec_log(LOG_INFO, "Stopping block in no-dynarec zone\n");
+            --ninst;
+            if(!dyn->insts[ninst].x64.barrier) {
+                BARRIER(BARRIER_FLOAT);
+            }
+            #if STEP == 0
+            dyn->insts[ninst].x64.need_after |= X_PEND;
+            #endif
+            ++ninst;
+            fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+            jump_to_next(dyn, addr, 0, ninst);
+            ok=0; need_epilog=0;
+        }
+    }
+    if(need_epilog) {
+        fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
+        jump_to_epilog(dyn, ip, 0, ninst);  // no linker here, it's an unknow instruction
+    }
+    FINI;
+    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d)\n", dyn->size);
+    return addr;
+}
@@ -0,0 +1,22 @@
+#ifndef __DYNAREC_NEXT_H__
+#define __DYNAREC_NEXT_H__
+
+#ifdef ARM64
+void arm64_next(void) EXPORTDYN;
+void arm64_prolog(x64emu_t* emu, void* addr) EXPORTDYN;
+void arm64_epilog() EXPORTDYN;
+#define native_next         arm64_next
+#define native_prolog       arm64_prolog
+#define native_epilog       arm64_epilog
+#elif defined(LA464)
+void la464_next(void) EXPORTDYN;
+void la464_prolog(x64emu_t* emu, void* addr) EXPORTDYN;
+void la464_epilog() EXPORTDYN;
+#define native_next         la464_next
+#define native_prolog       la464_prolog
+#define native_epilog       la464_epilog
+#else
+#error Unsupported architecture
+#endif
+
+#endif //__DYNAREC_NEXT_H__
@@ -1,6 +1,11 @@
 #ifndef __DYNAREC_PRIVATE_H_
 #define __DYNAREC_PRIVATE_H_

+#define BARRIER_NONE    0
+#define BARRIER_FLAGS   1
+#define BARRIER_FLOAT   2
+#define BARRIER_FULL    3
+
 // all flags for the use_flags field
 #define X_CF    (1<<0)
 #define X_PF    (1<<1)
@@ -9,26 +14,32 @@
 #define X_SF    (1<<4)
 #define X_OF    (1<<5)
 #define X_ALL   ((1<<6)-1)
-#define X_PEND  (0x1000)
+#define X_PEND  (0x80)

 // all state flags
 #define SF_UNKNOWN  0
 #define SF_SET      1
 #define SF_PENDING  2
-#define SF_SET_PENDING      3
-#define SF_MAYSET   4
-#define SF_SUBSET   5
+#define SF_SET_PENDING (SF_SET|SF_PENDING)
+#define SF_SUB      4
+#define SF_SUBSET   (SF_SUB|SF_SET)
+#define SF_SUBSET_PENDING   (SF_SUBSET|SF_PENDING)

 typedef struct instruction_x64_s {
-    uintptr_t   addr;   //address of the instruction
-    int32_t     size;   // size of the instruction
-    int         barrier; // next instruction is a jump point, so no optim allowed
-    uintptr_t   jmp;    // offset to jump to, even if conditionnal (0 if not), no relative offset here
+    uintptr_t   addr;       //address of the instruction
+    int32_t     size;       // size of the instruction
+    uintptr_t   jmp;        // offset to jump to, even if conditionnal (0 if not), no relative offset here
    int         jmp_insts;  // instuction to jump to (-1 if out of the block)
-    uint32_t    use_flags;  // 0 or combination of X_?F
-    uint32_t    set_flags;  // 0 or combination of X_?F
-    uint32_t    need_flags; // calculated
-    int         state_flags; // One of SF_XXX state
+    uint8_t     jmp_cond;   // 1 of conditionnal jump
+    uint8_t     has_next;   // does this opcode can continue to the next?
+    uint8_t     barrier;    // next instruction is a jump point, so no optim allowed
+    uint8_t     state_flags;// One of SF_XXX state
+    uint8_t     use_flags;  // 0 or combination of X_?F
+    uint8_t     set_flags;  // 0 or combination of X_?F
+    uint8_t     may_set;    // 1 if the flags may not be set
+    uint8_t     gen_flags;  // calculated
+    uint8_t     need_before;// calculated
+    uint8_t     need_after; // calculated
 } instruction_x64_t;

 void printf_x64_instruction(zydis_dec_t* dec, instruction_x64_t* inst, const char* name);
@@ -0,0 +1,26 @@
+#ifndef __NATIVE_LOCK__H__
+#define __NATIVE_LOCK__H__
+
+#ifdef ARM64
+#include "arm64/arm64_lock.h"
+
+#define native_lock_read_b(A)               arm64_lock_read_b(A)
+#define native_lock_write_b(A, B)           arm64_lock_write_b(A, B)
+#define native_lock_read_h(A)               arm64_lock_read_h(A)
+#define native_lock_write_h(A, B)           arm64_lock_write_h(A, B)
+#define native_lock_read_d(A)               arm64_lock_read_d(A)
+#define native_lock_write_d(A, B)           arm64_lock_write_d(A, B)
+#define native_lock_read_dd(A)              arm64_lock_read_dd(A)
+#define native_lock_write_dd(A, B)          arm64_lock_write_dd(A, B)
+#define native_lock_read_dq(A, B, C)        arm64_lock_read_dq(A, B, C)
+#define native_lock_write_dq(A, B, C)       arm64_lock_write_dq(A, B, C)
+#define native_lock_xchg(A, B)              arm64_lock_xchg(A, B)
+#define native_lock_xchg_d(A, B)            arm64_lock_xchg_d(A, B)
+#define native_lock_storeifref(A, B, C)     arm64_lock_storeifref(A, B, C)
+#define native_lock_storeifnull(A, B)       arm64_lock_storeifnull(A, B)
+
+#else
+#error Unsupported architecture
+#endif
+
+#endif //#define __NATIVE_LOCK__H__
@@ -0,0 +1,721 @@
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "debug.h"
+#include "elfloader_private.h"
+#include "elfdwarf_private.h"
+
+typedef struct dwarf_unwind_constr_s {
+    uint8_t reg_count;
+    uint64_t *table;
+    uint8_t *statuses;
+#define GET_STATUS(n,i) (((n).statuses[(i) >> 1] >> (((i) & 1) << 2)) & 0xF)
+#define SET_STATUS(n,i,v) (n).statuses[(i) >> 1] = ((n).statuses[(i) >> 1] & (0xF0 >> (((i) & 1) << 2))) | (v << (((i) & 1) << 2))
+#define REGSTATUS_undefined     0b0000
+#define REGSTATUS_same_val      0b0101
+#define REGSTATUS_offset        0b0010
+#define REGSTATUS_val_offset    0b0011
+#define REGSTATUS_register      0b0100
+} dwarf_unwind_constr_t;
+
+#define DW_EH_PE_uptr       0x00
+#define DW_EH_PE_uleb128    0x01
+#define DW_EH_PE_udata2     0x02
+#define DW_EH_PE_udata4     0x03
+#define DW_EH_PE_udata8     0x04
+#define DW_EH_PE_SIZE_MASK  0x07
+#define DW_EH_PE_signed     0x08
+#define DW_EH_PE_absptr     0x00
+#define DW_EH_PE_pcrel      0x10
+#define DW_EH_PE_textrel    0x20 // Unsupported
+#define DW_EH_PE_datarel    0x30 // Unsupported
+#define DW_EH_PE_indirect   0x80 // May crash
+#define DW_EH_PE_omit       0xff
+
+#define READ_U1(var, ptr) do { (var) = *( uint8_t*)(ptr); (ptr) += 1; } while (0)
+#define READ_S1(var, ptr) do { (var) = *(  int8_t*)(ptr); (ptr) += 1; } while (0)
+#define READ_U2(var, ptr) do { (var) = *(uint16_t*)(ptr); (ptr) += 2; } while (0)
+#define READ_S2(var, ptr) do { (var) = *( int16_t*)(ptr); (ptr) += 2; } while (0)
+#define READ_U4(var, ptr) do { (var) = *(uint32_t*)(ptr); (ptr) += 4; } while (0)
+#define READ_S4(var, ptr) do { (var) = *( int32_t*)(ptr); (ptr) += 4; } while (0)
+#define READ_U8(var, ptr) do { (var) = *(uint64_t*)(ptr); (ptr) += 8; } while (0)
+#define READ_S8(var, ptr) do { (var) = *( int64_t*)(ptr); (ptr) += 8; } while (0)
+#define READ_ULEB128(var, ptr) readULEB(&(var), &(ptr))
+#define READ_SLEB128(var, ptr) readSLEB((int64_t*)&(var), &(ptr))
+// Seems like DW_EH_PE_signed is actually an invalid encoding
+#define READ_ENCODED(var, ptr, enc, isptr) do { uintptr_t optr = (uintptr_t)(ptr); \
+         if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uleb128) &&  ((enc) & DW_EH_PE_signed)) READ_SLEB128(var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uleb128) && !((enc) & DW_EH_PE_signed)) READ_ULEB128(var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata2 ) &&  ((enc) & DW_EH_PE_signed)) READ_S2     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata2 ) && !((enc) & DW_EH_PE_signed)) READ_U2     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata4 ) &&  ((enc) & DW_EH_PE_signed)) READ_S4     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata4 ) && !((enc) & DW_EH_PE_signed)) READ_U4     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata8 ) &&  ((enc) & DW_EH_PE_signed)) READ_S8     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata8 ) && !((enc) & DW_EH_PE_signed)) READ_U8     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uptr   ) &&  ((enc) & DW_EH_PE_signed)) READ_S8     (var, ptr); \
+    else if ((((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uptr   ) && !((enc) & DW_EH_PE_signed)) READ_U8     (var, ptr); \
+    else { printf_log(LOG_DEBUG, "Invalid encoding 0x%02X\n", (enc)); (var) = 0; } \
+    if (((enc) & DW_EH_PE_pcrel) && isptr) (var) += optr; \
+    if ((enc) & DW_EH_PE_indirect) (var) = *(uint64_t*)(var); \
+} while (0)
+#define SKIP_1(ptr) (ptr) += 1
+#define SKIP_2(ptr) (ptr) += 2
+#define SKIP_4(ptr) (ptr) += 4
+#define SKIP_8(ptr) (ptr) += 8
+#define SKIP_LEB128(ptr) do { ++(ptr); } while (((*(((unsigned char*)(ptr))-1))) & (1 << 7))
+#define SKIP_ENCODED(ptr, enc) \
+    do { if (((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uleb128) SKIP_LEB128(ptr); \
+    else if (((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata2 ) SKIP_2     (ptr); \
+    else if (((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata4 ) SKIP_4     (ptr); \
+    else if (((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_udata8 ) SKIP_8     (ptr); \
+    else if (((enc) & DW_EH_PE_SIZE_MASK) == DW_EH_PE_uptr   ) SKIP_8     (ptr); \
+    else printf_log(LOG_DEBUG, "Invalid encoding 0x%02X\n", (enc)); \
+} while (0)
+
+// LEB128 wil only work on 64 bits numbers (unsigned long, like the Linux kernel)
+void readULEB(uint64_t *var, unsigned char **ptr) {
+    *var = 0;
+    int shift = 0;
+    while (1) {
+        unsigned char byte = **ptr; ++*ptr;
+        *var |= ((byte & 0x7F) << shift);
+        if (!(byte & (1 << 7))) break;
+        shift += 7;
+    }
+}
+void readSLEB(int64_t *var, unsigned char **ptr) {
+    uint64_t tmp = 0;
+    int shift = 0;
+    unsigned char byte;
+    while (1) {
+        byte = **ptr; ++*ptr;
+        tmp |= ((byte & 0x7F) << shift);
+        if (!(byte & (1 << 7))) break;
+        shift += 7;
+    }
+    // Sign extend
+    if (byte & (1 << 6)) {
+        *(uint64_t*)var = tmp | (0xFFFFFFFFFFFFFFFFull << (shift + 7));
+    } else {
+        *var = tmp;
+    }
+}
+
+uintptr_t get_parent_registers(dwarf_unwind_t *unwind, const elfheader_t *ehdr, uintptr_t addr, char *success) {
+    if (!ehdr) {
+        *success = 0;
+        return 0;
+    }
+    if(!IsAddressInElfSpace(ehdr, ehdr->ehframehdr)) {
+        *success = 0;
+        return 0;
+    }
+    unsigned char ehfh_version = *(unsigned char*)ehdr->ehframehdr;
+    if (ehfh_version != 1) {
+        *success = 0;
+        return 0;
+    }
+
+    // printf_log(LOG_NONE, "Address: 0x%016lX\nEH frame address: 0x%016lX\n", addr, ehdr->ehframe);
+
+    // Not using the binary search table (for now)
+
+    unsigned char *cur_addr = (unsigned char*)ehdr->ehframe;
+    unsigned char *end_addr = (unsigned char*)ehdr->ehframe_end;
+
+#define AUG_EHDATA      (1 << 0)
+#define AUG_AUGDATA     (1 << 1)
+#define AUG_LSDA        (1 << 2)
+#define AUG_PARG        (1 << 3)
+#define AUG_FDEENC      (1 << 4)
+#define AUG_SIGHDLER    (1 << 5)
+    unsigned short aug_fields = 0;
+    unsigned char lsda_encoding = DW_EH_PE_omit;
+    unsigned char fde_encoding  = DW_EH_PE_omit;
+    uint64_t eh_data = 0;               // Note: these are never reset, so use aug_fields & AUG_EHDATA
+    uint64_t code_alignment_factor = 0;
+    int64_t data_alignment_factor = 0;
+    uint64_t cie_aug_data_len = 0;      // Note: these are never reset, so use aug_fields & AUG_AUGDATA
+    unsigned char *cie_aug_data = NULL; // Note: these are never reset, so use aug_fields & AUG_AUGDATA
+    unsigned char *initCIEinstr = NULL; // Never NULL if initialized
+    unsigned char *endCIEinstr = NULL;  // Never NULL if initialized
+    uint64_t return_addr_reg = 0;       // Description absent in the LSBCS 5.0, but still present
+    while (cur_addr < end_addr) {
+        // Length
+        uint64_t len; int extended = 0;
+        READ_U4(len, cur_addr);
+        if (!len) {
+            // Terminator, end parsing
+            *success = 0;
+            return 0;
+        } else if (len == 0xFFFFFFFF) {
+            // Extended Length (optional)
+            extended = 1;
+            READ_U8(len, cur_addr);
+        }
+        (void)extended;
+        // printf_log(LOG_NONE, "[???] Length          %02X\n", len);
+        // printf_log(LOG_NONE, "[???] Extended        %c\n", extended ? 'Y' : 'N');
+        unsigned char *next_addr = cur_addr + len;
+
+        // CIE ID (always 0) or CIE Pointer (never 0)
+        uint32_t cie_ptr;
+        READ_U4(cie_ptr, cur_addr);
+        if (cie_ptr == 0) {
+            // Current block is a CIE
+            // printf_log(LOG_NONE, "[CIE] ID              %02X\n", cie_ptr);
+
+            // Version (always 1 or 3)
+            uint8_t cie_version;
+            READ_U1(cie_version, cur_addr);
+            // printf_log(LOG_NONE, "[CIE] Version         %01X\n", cie_version);
+            if ((cie_version != 1) && (cie_version != 3)) { // Only 1 is in the LSBCS 5.0 but I found 3 exists as well
+                // Invalid CIE version
+                printf_log(LOG_DEBUG, "Invalid CIE version %d (should be 1 or 3), stopping parsing\n", cie_version);
+                // Failed to process CIE
+                *success = 0;
+                return 0;
+            }
+
+            // Augmentation String
+            char *aug_str = (char*)cur_addr; cur_addr += strlen(aug_str) + 1;
+            // printf_log(LOG_NONE, "[CIE] Augmentation st %s\n", aug_str);
+            aug_fields = (aug_str[0] == 'z') ? AUG_AUGDATA : 0;
+            for (char *aug_str2 = aug_str + ((aug_fields & AUG_AUGDATA) ? 1 : 0); *aug_str2; ++aug_str2) {
+                if ((aug_str2[0] == 'e') && (aug_str2[1] == 'h')) {
+                    // Use has also been left untold in the Linux Standard Base Core Specification 3.0RC1
+                    printf_log(LOG_DEBUG, "Warning: EH data detected but this was removed from the standard\n");
+                    aug_fields |= AUG_EHDATA; ++aug_str2;
+                } else if (aug_fields & AUG_AUGDATA) {
+                    if (aug_str2[0] == 'L') {
+                        aug_fields |= AUG_LSDA;
+                    } else if (aug_str2[0] == 'P') {
+                        aug_fields |= AUG_PARG;
+                        printf_log(LOG_DEBUG, "Warning: augmentation string attribute 'P' unsupported\n");
+                    } else if (aug_str2[0] == 'S') {
+                        aug_fields |= AUG_SIGHDLER;
+                        printf_log(LOG_DEBUG, "Warning: augmentation string attribute 'S' ignored\n");
+                    } else if (aug_str2[0] == 'R') {
+                        aug_fields |= AUG_FDEENC;
+                    } else {
+                        printf_log(LOG_DEBUG, "Error: invalid augmentation string %s\n", aug_str);
+                        // Failed to process augmentation string
+                        break;
+                    }
+                } else {
+                    printf_log(LOG_DEBUG, "Error: invalid augmentation string %s\n", aug_str);
+                    // Failed to process augmentation string
+                    break;
+                }
+            }
+            // printf_log(LOG_NONE, "[CIE] Augmentation fs %02x\n", aug_fields);
+
+            // Code Alignment Factor
+            READ_ULEB128(code_alignment_factor, cur_addr);
+            // printf_log(LOG_NONE, "[CIE] Code AF         %d\n", code_alignment_factor);
+            // Data Alignment Factor
+            READ_SLEB128(data_alignment_factor, cur_addr);
+            // printf_log(LOG_NONE, "[CIE] Data AF         %d\n", data_alignment_factor);
+
+            if (cie_version == 1) {
+                READ_U1(return_addr_reg, cur_addr);
+            } else {
+                READ_ULEB128(return_addr_reg, cur_addr);
+            }
+            // printf_log(LOG_NONE, "[CIE] Return addr reg %01X\n", return_addr_reg);
+
+            lsda_encoding = DW_EH_PE_omit;
+            fde_encoding  = DW_EH_PE_absptr;
+            if (aug_fields & AUG_AUGDATA) {
+                // Augmentation Data Length (optional)
+                READ_ULEB128(cie_aug_data_len, cur_addr);
+                // printf_log(LOG_NONE, "[CIE] Aug data len    %01X\n", cie_aug_data_len);
+                // Augmentation Data (optional)
+                cie_aug_data = cur_addr; cur_addr += cie_aug_data_len;
+                char *cie_aug_data2 = (char*)cie_aug_data;
+                for (char *aug_str2 = aug_str + ((aug_fields & AUG_AUGDATA) ? 1 : 0); *aug_str2; ++aug_str2) {
+                    if ((aug_str2[0] == 'e') && (aug_str2[1] == 'h')) {
+                        break; // Unknown usage
+                    } else if (aug_fields & AUG_AUGDATA) {
+                        if (aug_str2[0] == 'L') {
+                            READ_U1(lsda_encoding, cie_aug_data2);
+                            if (lsda_encoding != DW_EH_PE_omit) printf_log(LOG_DEBUG, "Warning: LSDA unsupported\n");
+                        } else if (aug_str2[0] == 'P') {
+                            unsigned char pencoding;
+                            READ_U1(pencoding, cie_aug_data2);
+                            SKIP_ENCODED(cie_aug_data2, pencoding);
+                        } else if (aug_str2[0] == 'S') {
+                        } else if (aug_str2[0] == 'R') {
+                            READ_U1(fde_encoding, cie_aug_data2);
+                            // printf_log(LOG_NONE, "[CIE] FDE encoding    %02X\n", fde_encoding);
+                            if (fde_encoding == DW_EH_PE_omit) printf_log(LOG_DEBUG, "Error: FDE encoding set to 'omit'\n");
+                        } else {
+                            break;
+                        }
+                    } else {
+                        break;
+                    }
+                }
+            }
+        
+            // Initial Instructions
+            initCIEinstr = cur_addr;
+            endCIEinstr = next_addr;
+        } else if (initCIEinstr) {
+            // Current block is a FDE
+            // printf_log(LOG_NONE, "[FDE] Ptr             %02X\n", cie_ptr);
+
+            // PC Begin
+            uintptr_t pc_begin;
+            READ_ENCODED(pc_begin, cur_addr, fde_encoding, 1);
+            // printf_log(LOG_NONE, "[FDE] PC begin        %016lX\n", pc_begin);
+            // PC Range
+            uint64_t pc_range;
+            READ_ENCODED(pc_range, cur_addr, fde_encoding, 0);
+            // printf_log(LOG_NONE, "[FDE] PC end          %016lX\n", pc_begin + pc_range);
+            
+            uint64_t aug_data_len;
+            unsigned char *aug_data;
+            if (aug_fields & AUG_AUGDATA) {
+                // Augmentation Data Length (optional)
+                READ_ULEB128(aug_data_len, cur_addr);
+                // printf_log(LOG_NONE, "[FDE] Aug data len    %01X\n", aug_data_len);
+                // Augmentation Data (optional)
+                aug_data = cur_addr; cur_addr += aug_data_len;
+                (void)aug_data;
+            }
+
+            if (lsda_encoding != DW_EH_PE_omit) {
+                SKIP_ENCODED(cur_addr, lsda_encoding);
+            }
+
+            // Call Frame Instructions
+            if ((pc_begin <= addr) && (addr < pc_begin + pc_range)) {
+                // Corresponding Frame Descriptor Entry found!
+
+#define setmax(a, b) a = (a < (b) ? (b) : a)
+                dwarf_unwind_constr_t unwind_constr;
+                unwind_constr.reg_count = return_addr_reg;
+                uint64_t maxstacksize = 0;
+                uint64_t curstacksize = 0;
+                unsigned char *cur_inst = initCIEinstr;
+                uint64_t nreg;
+#define PARSE_INST \
+        switch (inst >> 6) {                                        \
+        case 0b01:                                                  \
+            break;                                                  \
+        case 0b10:                                                  \
+            setmax(unwind_constr.reg_count, inst & 0x3F);           \
+            SKIP_LEB128(cur_inst);                                  \
+            break;                                                  \
+        case 0b11:                                                  \
+            setmax(unwind_constr.reg_count, inst & 0x3F);           \
+            break;                                                  \
+        case 0b00:                                                  \
+            switch (inst & 0x3F) {                                  \
+            case 0b000000:                                          \
+                break;                                              \
+            case 0b001010:                                          \
+                ++curstacksize;                                     \
+                setmax(maxstacksize, curstacksize);                 \
+                break;                                              \
+            case 0b001011:                                          \
+                if (!curstacksize) {                                \
+                    printf_log(LOG_DEBUG, "Negative stack size during CFI execution\n"); \
+                    *success = 0;                                   \
+                    return 0;                                       \
+                }                                                   \
+                --curstacksize;                                     \
+                break;                                              \
+            case 0b000010:                                          \
+                SKIP_1(cur_inst);                                   \
+                break;                                              \
+            case 0b000011:                                          \
+                SKIP_2(cur_inst);                                   \
+                break;                                              \
+            case 0b000100:                                          \
+                SKIP_4(cur_inst);                                   \
+                break;                                              \
+            case 0b000001:                                          \
+                SKIP_8(cur_inst);                                   \
+                break;                                              \
+            case 0b001110:                                          \
+            case 0b010011:                                          \
+                SKIP_LEB128(cur_inst);                              \
+                break;                                              \
+            case 0b000110:                                          \
+            case 0b000111:                                          \
+            case 0b001000:                                          \
+            case 0b001101:                                          \
+                READ_ULEB128(nreg, cur_inst);                       \
+                setmax(unwind_constr.reg_count, nreg);              \
+                break;                                              \
+            case 0b000101:                                          \
+            case 0b001100:                                          \
+            case 0b010001:                                          \
+            case 0b010010:                                          \
+            case 0b010100:                                          \
+            case 0b010101:                                          \
+                READ_ULEB128(nreg, cur_inst);                       \
+                setmax(unwind_constr.reg_count, nreg);              \
+                SKIP_LEB128(cur_inst);                              \
+                break;                                              \
+            case 0b001001:                                          \
+                READ_ULEB128(nreg, cur_inst);                       \
+                setmax(unwind_constr.reg_count, nreg);              \
+                READ_ULEB128(nreg, cur_inst);                       \
+                setmax(unwind_constr.reg_count, nreg);              \
+                break;                                              \
+            default:                                                \
+                /* Contains undefined, user and expression CFIs */  \
+                printf_log(LOG_DEBUG, "Unknown CFI 0x%02X\n", inst); \
+                *success = 0;                                       \
+                return 0;                                        \
+            }                                                       \
+            break;                                                  \
+        }
+                while (cur_inst < endCIEinstr) {
+                    unsigned char inst; READ_U1(inst, cur_inst);
+                    PARSE_INST
+                }
+                cur_inst = cur_addr;
+                while (cur_inst < next_addr) {
+                    unsigned char inst; READ_U1(inst, cur_inst);
+                    PARSE_INST
+                }
+#undef PARSE_INST
+
+                uint64_t cfa_reg = -1;
+                unsigned char cfa_signed = 0;
+                union { uint64_t uoff; int64_t soff; } cfa_offset = { .uoff = 0 };
+                ++unwind_constr.reg_count;
+                size_t tablelen = unwind_constr.reg_count * sizeof(uint64_t);
+                size_t statuseslen = ((unwind_constr.reg_count+1) >> 1) * sizeof(uint8_t);
+                unwind_constr.table = (uint64_t*)box_malloc(tablelen);
+                unwind_constr.statuses = (uint8_t*)box_calloc((unwind_constr.reg_count+1) >> 1, sizeof(uint8_t));
+                // ~~undefined is 0: no initialization needed~~ still initialize the first 17 to same_val
+                for (int i = 0; (i < 17) && (i <= unwind_constr.reg_count); ++i) {
+                    SET_STATUS(unwind_constr, i, REGSTATUS_same_val);
+                }
+
+                curstacksize = 0;
+                uint64_t **table_stack = (uint64_t**)box_malloc(maxstacksize * sizeof(uint64_t*));
+                for (uint64_t i = 0; i < maxstacksize; ++i) {
+                    table_stack[i] = (uint64_t*)box_malloc(tablelen);
+                }
+                cur_inst = initCIEinstr;
+                uintptr_t cur_pointed_addr = pc_begin;
+                // Missing:
+                /* DW_CFA_def_cfa_expression    0   0x0f      BLOCK                              */
+                /* DW_CFA_expression            0   0x10      ULEB128 register  BLOCK            */
+                /* DW_CFA_val_expression        0   0x16      ULEB128 BLOCK                      */
+                /* DW_CFA_lo_user               0   0x1c                                         */
+                /* DW_CFA_GNU_args_size         0   0x2e      ULEB128 argsize                    */
+                /* DW_CFA_GNU_negative_offset_extended 0 0x2f ULEB128 register  ULEB128 offset (obsoleted by DW_CFA_offset_extended_sf) */
+                /* DW_CFA_hi_user               0   0x3f                                         */
+                // Known "bug": DW_CFA_set_loc can go backwards, this is ignored
+                uint64_t tmpreg;
+                uint64_t tmpuval;
+                int64_t tmpsval;
+#define PARSE_INST \
+        switch (inst >> 6) {                                                                                    \
+        case 0b01:         /* DW_CFA_advance_loc           0x1 delta                                        */  \
+            cur_pointed_addr += (inst & 0x3F) * code_alignment_factor;                                          \
+            break;                                                                                              \
+        case 0b10:         /* DW_CFA_offset                0x2 register  ULEB128 offset                     */  \
+            READ_ULEB128(tmpuval, cur_inst);                                                                    \
+            SET_STATUS(unwind_constr, inst & 0x3F, REGSTATUS_offset);                                           \
+            unwind_constr.table[inst & 0x3F] = (int64_t)tmpuval * data_alignment_factor;                        \
+            break;                                                                                              \
+        case 0b11:         /* DW_CFA_restore               0x3 register                                     */  \
+            RESTORE_REG(inst & 0x3F)                                                                            \
+            break;                                                                                              \
+        case 0b00:                                                                                              \
+            switch (inst & 0x3F) {                                                                              \
+            case 0b000001: /* DW_CFA_set_loc               0   0x01      address                            */  \
+                READ_U8(cur_pointed_addr, cur_inst);                                                            \
+                break;                                                                                          \
+            case 0b000010: /* DW_CFA_advance_loc1          0   0x02      1-byte delta                       */  \
+                READ_U1(tmpuval, cur_inst);                                                                     \
+                cur_pointed_addr += tmpuval * code_alignment_factor;                                            \
+                break;                                                                                          \
+            case 0b000011: /* DW_CFA_advance_loc2          0   0x03      2-byte delta                       */  \
+                READ_U2(tmpuval, cur_inst);                                                                     \
+                cur_pointed_addr += tmpuval * code_alignment_factor;                                            \
+                break;                                                                                          \
+            case 0b000100: /* DW_CFA_advance_loc4          0   0x04      4-byte delta                       */  \
+                READ_U4(tmpuval, cur_inst);                                                                     \
+                cur_pointed_addr += tmpuval * code_alignment_factor;                                            \
+                break;                                                                                          \
+            \
+            case 0b001100: /* DW_CFA_def_cfa               0   0x0c      ULEB128 register  ULEB128 offset   */  \
+                READ_ULEB128(cfa_reg, cur_inst);                                                                \
+                cfa_signed = 0;                                                                                 \
+                READ_ULEB128(cfa_offset.uoff, cur_inst);                                                        \
+                break;                                                                                          \
+            case 0b010010: /* DW_CFA_def_cfa_sf            0   0x12      ULEB128 register  SLEB128 offset   */  \
+                READ_ULEB128(cfa_reg, cur_inst);                                                                \
+                cfa_signed = 1;                                                                                 \
+                READ_SLEB128(cfa_offset.soff, cur_inst);                                                        \
+                cfa_offset.soff *= data_alignment_factor;                                                       \
+                break;                                                                                          \
+            case 0b001101: /* DW_CFA_def_cfa_register      0   0x0d      ULEB128 register                   */  \
+                READ_ULEB128(cfa_reg, cur_inst);                                                                \
+                break;                                                                                          \
+            case 0b001110: /* DW_CFA_def_cfa_offset        0   0x0e      ULEB128 offset                     */  \
+                cfa_signed = 0;                                                                                 \
+                READ_ULEB128(cfa_offset.uoff, cur_inst);                                                        \
+                break;                                                                                          \
+            case 0b010011: /* DW_CFA_def_cfa_offset_sf     0   0x13      SLEB128 offset                     */  \
+                READ_SLEB128(cfa_offset.soff, cur_inst);                                                        \
+                cfa_offset.soff *= data_alignment_factor;                                                       \
+                break;                                                                                          \
+            /* DW_CFA_def_cfa_expression */ \
+            \
+            case 0b000111: /* DW_CFA_undefined             0   0x07      ULEB128 register                   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_undefined);                                         \
+                break;                                                                                          \
+            case 0b001000: /* DW_CFA_same_value            0   0x08      ULEB128 register                   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_same_val);                                          \
+                break;                                                                                          \
+            case 0b000101: /* DW_CFA_offset_extended       0   0x05      ULEB128 register  ULEB128 offset   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                READ_ULEB128(tmpuval, cur_inst);                                                                \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_offset);                                            \
+                unwind_constr.table[tmpreg] = (int64_t)tmpuval * data_alignment_factor;                         \
+                break;                                                                                          \
+            case 0b010001: /* DW_CFA_offset_extended_sf    0   0x11      ULEB128 register  SLEB128 offset   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                READ_SLEB128(tmpsval, cur_inst);                                                                \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_offset);                                            \
+                unwind_constr.table[tmpreg] = tmpsval * data_alignment_factor;                                  \
+                break;                                                                                          \
+            case 0b010100: /* DW_CFA_val_offset            0   0x14      ULEB128 register  ULEB128 offset   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                READ_ULEB128(tmpuval, cur_inst);                                                                \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_val_offset);                                        \
+                unwind_constr.table[tmpreg] = (int64_t)tmpuval * data_alignment_factor;                         \
+                break;                                                                                          \
+            case 0b010101: /* DW_CFA_val_offset_sf         0   0x15      ULEB128 register  SLEB128 offset   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                READ_SLEB128(tmpsval, cur_inst);                                                                \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_val_offset);                                        \
+                unwind_constr.table[tmpreg] = tmpsval * data_alignment_factor;                                  \
+                break;                                                                                          \
+            case 0b001001: /* DW_CFA_register              0   0x09      ULEB128 register  ULEB128 register */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                READ_ULEB128(tmpuval, cur_inst);                                                                \
+                SET_STATUS(unwind_constr, tmpreg, REGSTATUS_register);                                          \
+                unwind_constr.table[tmpreg] = tmpuval;                                                          \
+                break;                                                                                          \
+            /* DW_CFA_expression */ \
+            /* DW_CFA_val_expression */ \
+            case 0b000110: /* DW_CFA_restore_extended      0   0x06      ULEB128 register                   */  \
+                READ_ULEB128(tmpreg, cur_inst);                                                                 \
+                RESTORE_REG(tmpreg)                                                                             \
+                break;                                                                                          \
+            \
+            case 0b001010: /* DW_CFA_remember_state        0   0x0a                                         */  \
+                memcpy(table_stack[curstacksize], unwind_constr.table, tablelen);                               \
+                ++curstacksize;                                                                                 \
+                break;                                                                                          \
+            case 0b001011: /* DW_CFA_restore_state         0   0x0b                                         */  \
+                --curstacksize;                                                                                 \
+                memcpy(unwind_constr.table, table_stack[curstacksize], tablelen);                               \
+                break;                                                                                          \
+            \
+            case 0b000000: /* DW_CFA_nop                   0   0                                            */  \
+                break;                                                                                          \
+            default:                                                                                            \
+                /* Contains undefined, user and expression CFIs */                                              \
+                printf_log(LOG_DEBUG, "Unknown CFI 0x%02X\n", inst);                                             \
+                FAILED                                                                                          \
+            }                                                                                                   \
+            break;                                                                                              \
+        }
+#define RESTORE_REG(reg) \
+    printf_log(LOG_DEBUG, "Trying to restore register 0x%02lX while in the intial CFIs\n", (uint64_t)reg); \
+    FAILED
+#define FAILED \
+    box_free(unwind_constr.statuses);                                       \
+    box_free(unwind_constr.table);                                          \
+    for (uint64_t i = 0; i < maxstacksize; ++i) box_free(table_stack[i]);   \
+    box_free(table_stack);                                                  \
+    *success = 0; return 0;
+                while (cur_inst < endCIEinstr) {
+                    unsigned char inst; READ_U1(inst, cur_inst);
+                    // printf_log(LOG_NONE, "Executing pre 0x%02X\n", inst);
+                    PARSE_INST
+                }
+#undef FAILED
+#undef RESTORE_REG
+                uint64_t *init_table = (uint64_t*)box_malloc(tablelen);
+                memcpy(init_table, unwind_constr.table, tablelen);
+                uint8_t *init_statuses = (uint8_t*)box_malloc(statuseslen);
+                memcpy(init_statuses, unwind_constr.statuses, statuseslen);
+                cur_inst = cur_addr;
+#define RESTORE_REG(reg) \
+    unwind_constr.table[reg] = init_table[reg]; \
+    SET_STATUS(unwind_constr, (reg), ((init_statuses[(reg) >> 1] >> (((reg) & 1) << 2)) & 0xF));
+#define FAILED \
+    box_free(init_statuses);                                                \
+    box_free(init_table);                                                   \
+    box_free(unwind_constr.statuses);                                       \
+    box_free(unwind_constr.table);                                          \
+    for (uint64_t i = 0; i < maxstacksize; ++i) box_free(table_stack[i]);   \
+    box_free(table_stack);                                                  \
+    *success = 0; return 0;
+                while ((cur_inst < next_addr) && (cur_pointed_addr <= addr)) {
+                    unsigned char inst; READ_U1(inst, cur_inst);
+                    // printf_log(LOG_NONE, "Executing post 0x%02X\n", inst);
+                    PARSE_INST
+                }
+#undef FAILED
+#undef RESTORE_REG
+#undef PARSE_INST
+                box_free(init_statuses);
+                box_free(init_table);
+                for (int i = 0; i < maxstacksize; ++i) {
+                    box_free(table_stack[i]);
+                }
+                box_free(table_stack);
+
+                dwarf_unwind_t new_unwind;
+                new_unwind.reg_count = unwind_constr.reg_count;
+                new_unwind.regs = box_calloc(unwind_constr.reg_count, sizeof(uint64_t));
+                uintptr_t cfa = unwind->regs[cfa_reg];
+                if (cfa_signed) cfa += cfa_offset.soff;
+                else            cfa += cfa_offset.uoff;
+                
+                // printf_log(LOG_NONE, "Done, rewriting registers (CFA at r%d(0x%016lX) + %lld (%c) -> 0x%016lX\n", cfa_reg, unwind->regs[cfa_reg], cfa_offset.soff, cfa_signed ? 's' : 'u', cfa);
+                for (uint64_t i = 0; i < unwind_constr.reg_count; ++i) {
+                    switch (GET_STATUS(unwind_constr, i)) {
+                    case REGSTATUS_undefined:
+                        // printf_log(LOG_NONE, "Register %02lX: (undefined)\n", i);
+                        break;
+                    case REGSTATUS_same_val:
+                        if (i >= unwind->reg_count) {
+                            printf_log(LOG_DEBUG, "Invalid register status (value copied from register 0x%02lX)\n", i);
+                            box_free(unwind_constr.statuses);
+                            box_free(unwind_constr.table);
+                            box_free(new_unwind.regs);
+                            *success = 0;
+                            return 0;
+                        }
+                        new_unwind.regs[i] = unwind->regs[i];
+                        // printf_log(LOG_NONE, "Register %02lX: copy   %016lX\n", i, new_unwind.regs[i]);
+                        break;
+                    case REGSTATUS_offset:
+                        new_unwind.regs[i] = *(uint64_t*)(cfa + (int64_t)unwind_constr.table[i]);
+                        // printf_log(LOG_NONE, "Register %02lX: offset %016lX [%016lX + %lld]\n", i, new_unwind.regs[i], cfa, (int64_t)unwind_constr.table[i]);
+                        break;
+                    case REGSTATUS_val_offset:
+                        new_unwind.regs[i] = (uint64_t)(cfa + (int64_t)unwind_constr.table[i]);
+                        // printf_log(LOG_NONE, "Register %02lX: voff   %016lX\n", i, new_unwind.regs[i]);
+                        break;
+                    case REGSTATUS_register:
+                        if (unwind_constr.table[i] >= unwind->reg_count) {
+                            printf_log(LOG_DEBUG, "Invalid register status (value copied from register 0x%02lX)\n", unwind_constr.table[i]);
+                            box_free(unwind_constr.statuses);
+                            box_free(unwind_constr.table);
+                            box_free(new_unwind.regs);
+                            *success = 0;
+                            return 0;
+                        }
+                        new_unwind.regs[i] = unwind->regs[unwind_constr.table[i]];
+                        // printf_log(LOG_NONE, "Register %02lX: reg    %016lX\n", i, new_unwind.regs[i]);
+                        break;
+                    }
+                }
+                *success = (GET_STATUS(unwind_constr, return_addr_reg) == REGSTATUS_undefined) ? 0 : ((aug_fields & AUG_SIGHDLER) ? 2 : 1);
+                box_free(unwind_constr.statuses);
+                box_free(unwind_constr.table);
+                
+                box_free(unwind->regs);
+                unwind->reg_count = new_unwind.reg_count;
+                unwind->regs = new_unwind.regs;
+
+                // Maybe?
+                unwind->regs[cfa_reg] = cfa;
+
+                // printf_log(LOG_NONE, "Returning %016lX\n", unwind->regs[return_addr_reg]);
+                return unwind->regs[return_addr_reg];
+            }
+        } else {
+            // printf_log(LOG_NONE, "[FDE] Ptr             %02X\n", cie_ptr);
+            printf_log(LOG_DEBUG, "Unexpected FDE, corresponding CIE missing\n");
+            return 0;
+        }
+
+        cur_addr = next_addr;
+    }
+    *success = 0;
+    return 0;
+}
+
+dwarf_unwind_t *init_dwarf_unwind_registers(x64emu_t *emu) {
+    dwarf_unwind_t *unwind_struct = (dwarf_unwind_t*)box_malloc(sizeof(dwarf_unwind_t));
+    unwind_struct->reg_count = 17;
+    unwind_struct->regs = (uint64_t*)box_malloc(17*sizeof(uint64_t));
+    /* x86_64-abi-0.99.pdf
+     * Register Name                    | Number | Abbreviation
+     * General Purpose Register RAX     | 0      | %rax
+     * General Purpose Register RDX     | 1      | %rdx
+     * General Purpose Register RCX     | 2      | %rcx
+     * General Purpose Register RBX     | 3      | %rbx
+     * General Purpose Register RSI     | 4      | %rsi
+     * General Purpose Register RDI     | 5      | %rdi
+     * Frame Pointer Register   RBP     | 6      | %rbp
+     * Stack Pointer Register   RSP     | 7      | %rsp
+     * Extended Integer Registers 8-15  | 8-15   | %r8-%r15
+     * Return Address RA                | 16     |
+     * Vector Registers 0-7             | 17-24  | %xmm0-%xmm7
+     * Extended Vector Registers 8-15   | 25-32  | %xmm8-%xmm15
+     * Floating Point Registers 0-7     | 33-40  | %st0-%st7
+     * MMX Registers 0-7                | 41-48  | %mm0-%mm7
+     * Flag Register                    | 49     | %rFLAGS
+     * Segment Register ES              | 50     | %es
+     * Segment Register CS              | 51     | %cs
+     * Segment Register SS              | 52     | %ss
+     * Segment Register DS              | 53     | %ds
+     * Segment Register FS              | 54     | %fs
+     * Segment Register GS              | 55     | %gs
+     * Reserved                         | 56-57  |
+     * FS Base address                  | 58     | %fs.base
+     * GS Base address                  | 59     | %gs.base
+     * Reserved                         | 60-61  |
+     * Task Register                    | 62     | %tr
+     * LDT Register                     | 63     | %ldtr
+     * 128-bit Media Control and Status | 64     | %mxcsr
+     * x87 Control Word                 | 65     | %fcw
+     * x87 Status Word                  | 66     | %fsw
+     */
+    unwind_struct->regs[ 0] = emu->regs[_RAX].q[0];
+    unwind_struct->regs[ 1] = emu->regs[_RDX].q[0];
+    unwind_struct->regs[ 2] = emu->regs[_RCX].q[0];
+    unwind_struct->regs[ 3] = emu->regs[_RBX].q[0];
+    unwind_struct->regs[ 4] = emu->regs[_RSI].q[0];
+    unwind_struct->regs[ 5] = emu->regs[_RDI].q[0];
+    unwind_struct->regs[ 6] = emu->regs[_RBP].q[0];
+    unwind_struct->regs[ 7] = emu->regs[_RSP].q[0] + 8;
+    unwind_struct->regs[ 8] = emu->regs[_R8 ].q[0];
+    unwind_struct->regs[ 9] = emu->regs[_R9 ].q[0];
+    unwind_struct->regs[10] = emu->regs[_R10].q[0];
+    unwind_struct->regs[11] = emu->regs[_R11].q[0];
+    unwind_struct->regs[12] = emu->regs[_R12].q[0];
+    unwind_struct->regs[13] = emu->regs[_R13].q[0];
+    unwind_struct->regs[14] = emu->regs[_R14].q[0];
+    unwind_struct->regs[15] = emu->regs[_R15].q[0];
+    unwind_struct->regs[16] = emu->ip.q[0];
+    return unwind_struct;
+}
+
+void free_dwarf_unwind_registers(dwarf_unwind_t **unwind_struct) {
+    box_free((*unwind_struct)->regs);
+    box_free(*unwind_struct);
+    *unwind_struct = NULL;
+}
@@ -0,0 +1,19 @@
+#ifndef __ELFDWARF_PRIVATE_H_
+#define __ELFDWARF_PRIVATE_H_
+
+#include "emu/x64emu_private.h"
+
+typedef struct dwarf_unwind_s {
+    uint8_t reg_count;
+    uint64_t *regs;
+} dwarf_unwind_t;
+typedef struct elfheader_s elfheader_t;
+
+dwarf_unwind_t *init_dwarf_unwind_registers(x64emu_t *emu);
+void free_dwarf_unwind_registers(dwarf_unwind_t **unwind_struct);
+
+// Returns the callee's address on success or NULL on failure (may be NULL regardless).
+// If success equals 2, the frame is a signal frame.
+uintptr_t get_parent_registers(dwarf_unwind_t *emu, const elfheader_t *ehdr, uintptr_t addr, char *success);
+
+#endif // __ELFDWARF_PRIVATE_H_