Bumped version to v0.2.2

[DYNAREC] Forgot this file on previous commit
[DYNAREC] Added 65 69/81/83 opcodes
2026-11-08 17:39:19 -10:00 · 2021-05-24 10:44:00 +02:00 · 2021-05-24 10:30:56 +02:00 · 2021-05-24 09:53:52 +02:00 · 2021-05-23 21:58:55 +02:00 · 2021-05-23 21:35:05 +02:00
495 changed files with 82618 additions and 20008 deletions
@@ -0,0 +1,12 @@
+# These are supported funding model platforms
+
+github: ptitSeb
+patreon: # Replace with a single Patreon username
+open_collective: # Replace with a single Open Collective username
+ko_fi: # Replace with a single Ko-fi username
+tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
+community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
+liberapay: # Replace with a single Liberapay username
+issuehunt: # Replace with a single IssueHunt username
+otechie: # Replace with a single Otechie username
+custom: paypal.me/0ptitSeb
@@ -1,4 +1,8 @@
-build
-build*
+build/
+build*/
 .vscode
 .gdb_history
+src/git_head.h
+backup/
+box86_update.sh
+restore_backup.sh
@@ -1,40 +0,0 @@
-v0.1.0
-=======
-* Dynarec!!! Only for ARM (note that dynarec doesn't support JITed code for now)
-* Added real support for getcontext/set/context/makecontext/swapcontext
-* Preliminary signal handling
-* Fixes to SDL(1/2) Image and SDL1 Mixer, and to SDL1.2 RWops usage
-* Fixed numerous issues in opcodes (both interpretor and dynarec). FTL works fine now, among many others
-* Added wrapped gtk support (still ongoing, many libs involved)
-* Make loading of libs more configurable
-* If a wrapped native libs is not found, try to use emulated one
-* Add en env. var. to force use of emulated lib for certain libs
-* Add an env. var. to define wich libGL to use
-* Added Install / Uninstall target (using systemd binfmt)
-* Added more hardware target (RPis, GameShell...)
-* Wrapped more libs (including FreeType, smpeg, ncurses, sndfile...)
-
-v0.0.4
-=======
-* Improved the Init of dependant Libs. More things work now.
-* Added a lot of wrapped functions.
-* Added a few wrapped libs (like libz or some other x11 related libs)
-* For trace enabled build, Trace can be enabled only after a certain amount of opcodes (still, a debugger would be better)
-* Some fixes in a few opcodes, and implemented x87 rounding (SuperMeatBoy behaves better now)
-* FTL 1.6.9 still have corrupted music, but older 1.5.13 seems fine (different set of libs)
-
-v0.0.2
-=======
-* A full commercial games runs fine on the Pandora: Airline Tycoon Deluxe
-* Implemented all planed subpart of Box86 except JIT
-* CPU Emulation is at 75%, roughly, including x87 and SSE/SSE2. MMX is barely implemented, but barely used anyway
-* x87 emulation is simplified, no real x87 State handling (but should not be mandatory, as native libm is used)
-* No Signal handling yet
-* ELF Loader is crude and probably full of bug. Also, the Init of libs is defered after all symbols are resolved for now.
-* Wrapped lib include libc, libm, rt, pthread, libdl, dllinux, libasound, GL, GLU, SDL1/mixer/image, SDL2/mixer/image/smpeg, OpenAL/ALUT, libz, libpng16, vorbisfile, x11/xrandr/xxf86vm.
-* Most wrapped libs are still partially implemented (SDL1 & GL should be complete)
-* Implemented specific mecanism for SDL(1/2) RWops, to be able to used them both in Native and x86 world
-* WorldOfGoo works, but painfully slow on th Pandora (too much double math, and lack of JIT)
-* FTL works, but sound is broken (issue with thread? asound? or CPU core?)
-* Limbo launch but crash before main menu
-
@@ -1,83 +1,135 @@
-cmake_minimum_required(VERSION 2.6)
-
-project(box86)
-
-enable_testing()
-enable_language(C)
+cmake_minimum_required(VERSION 3.0)

 option(PANDORA "Set to ON if targeting an OpenPandora device" ${PANDORA})
 option(PYRA "Set to ON if targeting an OpenPandora device" ${PYRA})
 option(RPI2 "Set to ON if targeting an RaspberryPI2 device" ${RPI2})
 option(RPI3 "Set to ON if targeting an RaspberryPI3 device" ${RPI3})
 option(RPI4 "Set to ON if targeting an RaspberryPI4 device" ${RPI4})
-option(GAMESHELL "Set to ON if targeting a Gameshell device" ${GAMESHELL})
+option(RPI4ARM64 "Set to ON if targeting an RaspberryPI4 device with multiarch arm64 and armhf" ${RPI4ARM64})
+option(RK3399 "Set to ON if targeting an Rockchip RK3399 based device" ${RK3399})
+option(GAMESHELL "Set to ON if targeting a GameShell device" ${GAMESHELL})
+option(ODROID "Set to ON if targeting an ODroid device" ${ODROID})
+option(POWER9 "Set to ON if targeting a POWER9 processor" ${POWER9})
+option(ANDROID "Set to ON if targeting an Android device" ${ANDROID})
 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 Box86" ${NO_LOADAADR})
+option(NOGETCLOCK "Set to ON to avoid using clock_gettime with CLOCK_MONOTONIC_COARSE for RDTSC opcode (use gltimeofday instead)" ${NOGETCLOCK})
+option(NOGIT "Set to ON if not building from a git clone repo (like when building from a zip download from github)" ${NOGIT})
+if(PANDORA OR PYRA OR RPI2 OR RPI3 OR RPI4 OR GAMESHELL OR ODROID OR RK3399 OR RPI4ARM64)
+    set(LD80BITS OFF CACHE BOOL "")
+    set(NOALIGN OFF CACHE BOOL "")
+    set(ARM_DYNAREC ON CACHE BOOL "")
+endif()
+if(ANDROID)
+    set(NOLOADADDR 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(HAVE_TRACE "Set to ON to have Trace ability (needs ZydisInfo library)" ${HAVE_TRACE})
-option(USE_FLOAT "Set to ON to use only float, no double, in all x87 Emulation" ${USE_FLOAT})
-option(NOLOADADDR "Set to ON to avoid fixing the load address of Box86" ${NO_LOADAADR})
-option(ARM_DYNAREC "Set to ON to use ARM Dynamic Recompilation (WIP, don't use yet)" ${ARM_DYNAREC})
-option(NOGETCLOCK "Set to ON to avoid using clock_gettime with CLOCK_MONOTONIC_COARSE for RDTSC opcode (use gltimeofday instead)" ${NOGETCLOCK})
+option(ARM_DYNAREC "Set to ON to use ARM Dynamic Recompilation" ${ARM_DYNAREC})
+option(POWERPCLE "Set to ON if targeting a little-endian PowerPC device" ${POWERPCLE})
+option(SYSV_VARARG "Set to ON if host device uses SysVR4 vararg calling convention" ${SYSV_VARARG})

-find_package(PythonInterp)
-if(NOT PYTHONINTERP_FOUND)
-    message( FATAL_ERROR "You need a Python interpretor, CMake will exit." )
+if(${CMAKE_VERSION} VERSION_LESS "3.12.2")
+    find_package(PythonInterp 3)
+    if(NOT PYTHONINTERP_FOUND)
+        message( FATAL_ERROR "You need a Python interpretor, CMake will exit." )
+    endif()
+    if(${PYTHON_VERSION_MAJOR} LESS 3)
+        message( FATAL_ERROR "You need a Python 3 interpretor, CMake will exit." )
+    endif()
+else()
+    find_package(Python3)
+    if(NOT Python3_Interpreter_FOUND)
+        message( FATAL_ERROR "You need a Python interpretor, CMake will exit." )
+    endif()
+    set(PYTHON_EXECUTABLE ${Python3_EXECUTABLE} CACHE INTERNAL "The Python3 executable" FORCE)
+endif()
+
+if(RK3399 OR RPI4ARM64)
+    set(CMAKE_C_COMPILER arm-linux-gnueabihf-gcc)
+    #set(CMAKE_ASM_COMPILER arm-linux-gnueabihf-as) #will use gcc in fact
+    set(CMAKE_C_COMPILER_TARGET arm-linux-gnueabihf)
+endif()
+
+project(box86 C ASM)
+
+enable_testing()
+
+set(default_build_type "RelwithDebInfo")
+if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
+  message(STATUS "Setting build type to '${default_build_type}' as none was specified.")
+  set(CMAKE_BUILD_TYPE "${default_build_type}" CACHE STRING "Choose the type of build." FORCE)
+  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
 endif()

-# Pandora
 if(PANDORA)
    add_definitions(-DPANDORA)
    #add_definitions(-mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp -ftree-vectorize -fsingle-precision-constant -ffast-math)
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
-
-# Gameshell
 if(GAMESHELL)
+    add_definitions(-DGAMEHELL)
    add_definitions(-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard")
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
-
 if(PYRA)
+    add_definitions(-DPYRA)
    add_definitions(-mcpu=cortex-a15 -mfpu=neon -mfloat-abi=hard -marm)
    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a15 -mfpu=neon -mfloat-abi=hard")
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
-
 if(RPI2)
    add_definitions(-DRPI)
    add_definitions(-DRPI2)
    add_definitions(-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard")
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
 if(RPI3)
    add_definitions(-DRPI)
    add_definitions(-DRPI3)
    add_definitions(-marm -march=armv8-a+crc -mtune=cortex-a53 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-marm -march=armv8-a+crc -mtune=cortex-a53 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
 if(RPI4)
    add_definitions(-DRPI)
-    add_definitions(-DRPI3)
+    add_definitions(-DRPI4)
    add_definitions(-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
-    set(ARM_DYNAREC ON)
-    set(LD80BITS OFF)
-    set(NOALIGN OFF)
 endif()
+if(RPI4ARM64)
+    add_definitions(-DRPI)
+    add_definitions(-DRPI4ARM64)
+    add_definitions(-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
+endif()
+if(ODROID)
+    add_definitions(-DODROID)
+    add_definitions(-mcpu=cortex-a9 -mfpu=neon-vfpv4 -mfloat-abi=hard -marm)
+    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a9 -mfpu=neon-vfpv4 -mfloat-abi=hard")
+endif()
+if(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")
+endif()
+
+if(POWER9)
+  add_definitions(-mcpu=power9 -mtune=power9)
+endif()
+
+if(NOGIT)
+    add_definitions(-DNOGIT)
+endif()
+
+if(BOX86LIB)
+    set(BOX86 box86)
+elseif(HAVE_TRACE)
+    set(BOX86 box86t)
+else()
+    set(BOX86 box86)
+endif()
+
+
+set(BOX86_ELF_ADDRESS "0x62800000")

 if(LD80BITS)
    add_definitions(-DHAVE_LD80BITS)
@@ -91,10 +143,6 @@ if(HAVE_TRACE)
    add_definitions(-DHAVE_TRACE)
 endif()

-if(USE_FLOAT)
-    add_definitions(-DUSE_FLOAT)
-endif()
-
 if(ARM_DYNAREC)
    add_definitions(-DDYNAREC)
    add_definitions(-DARM)
@@ -102,22 +150,33 @@ if(ARM_DYNAREC)
    enable_language(ASM)
 endif()

+if(POWERPCLE)
+  add_definitions(-DPOWERPCLE)
+endif()
+
+if(SYSV_VARARG)
+  add_definitions(-DSYSV_VARARG)
+endif()
+
 if(NOGETCLOCK)
    add_definitions(-DNOGETCLOCK)
 endif()

 if(BOX86LIB)
    add_definitions(-DBUILD_LIB)
+    if(BOX86DYNAMIC)
+        add_definitions(-DBUILD_DYNAMIC)
+    endif()
    add_definitions(-fPIC)
    if(NOT BOX86_ROOT)
        message(FATAL_ERROR "BOX86_ROOT not set when build as lib")
    endif()
 else()
-    set(BOX86_ROOT ${CMAKE_HOME_DIRECTORY})
+    set(BOX86_ROOT ${CMAKE_SOURCE_DIR})
 endif()


-add_definitions(-g -std=gnu99 -funwind-tables -O3 -fvisibility=hidden)
+add_definitions(-std=gnu11 -funwind-tables -fvisibility=hidden)

 if(USE_CCACHE)
    find_program(CCACHE_FOUND ccache)
@@ -133,15 +192,25 @@ include_directories(
    "${BOX86_ROOT}/src/wrapped/generated"
 )

+# git_head.h is a generated file
+set_source_files_properties(
+    "${BOX86_ROOT}/src/git_head.h"
+    PROPERTIES GENERATED TRUE
+    HEADER_FILE_ONLY TRUE)
+
 set(ELFLOADER_SRC
    "${BOX86_ROOT}/src/main.c"
    "${BOX86_ROOT}/src/box86context.c"
+    "${BOX86_ROOT}/src/build_info.c"
+    "${BOX86_ROOT}/src/custommem.c"
    "${BOX86_ROOT}/src/tools/bridge.c"
    "${BOX86_ROOT}/src/tools/pathcoll.c"
    "${BOX86_ROOT}/src/tools/fileutils.c"
    "${BOX86_ROOT}/src/tools/callback.c"
    "${BOX86_ROOT}/src/tools/box86stack.c"
+    "${BOX86_ROOT}/src/tools/my_cpuid.c"
    "${BOX86_ROOT}/src/tools/gtkclass.c"
+    "${BOX86_ROOT}/src/tools/wine_tools.c"
    "${BOX86_ROOT}/src/elfs/elfloader.c"
    "${BOX86_ROOT}/src/elfs/elfparser.c"
    "${BOX86_ROOT}/src/elfs/elfload_dump.c"
@@ -150,22 +219,30 @@ set(ELFLOADER_SRC
    "${BOX86_ROOT}/src/emu/x86emu.c"
    "${BOX86_ROOT}/src/emu/x86run.c"
    "${BOX86_ROOT}/src/emu/x86run66.c"
+    "${BOX86_ROOT}/src/emu/x86run660f.c"
    "${BOX86_ROOT}/src/emu/x86run_private.c"
    "${BOX86_ROOT}/src/emu/x86syscall.c"
    "${BOX86_ROOT}/src/emu/x86primop.c"
    "${BOX86_ROOT}/src/emu/x86trace.c"
    "${BOX86_ROOT}/src/emu/x86int3.c"
+    "${BOX86_ROOT}/src/emu/x86tls.c"
    "${BOX86_ROOT}/src/emu/x87emu_private.c"
    "${BOX86_ROOT}/src/emu/x87run.c"
+    "${BOX86_ROOT}/src/libtools/auxval.c"
    "${BOX86_ROOT}/src/libtools/myalign.c"
+    "${BOX86_ROOT}/src/libtools/myalign64.c"
    "${BOX86_ROOT}/src/libtools/myfts.c"
    "${BOX86_ROOT}/src/libtools/sdl1rwops.c"
    "${BOX86_ROOT}/src/libtools/sdl2rwops.c"
    "${BOX86_ROOT}/src/libtools/signals.c"
    "${BOX86_ROOT}/src/libtools/threads.c"
+    "${BOX86_ROOT}/src/libtools/vkalign.c"
    "${BOX86_ROOT}/src/dynarec/dynarec.c"
    "${BOX86_ROOT}/src/wrapped/generated/wrapper.c"
 )
+if(NOT ANDROID)
+    list(APPEND ELFLOADER_SRC "${BOX86_ROOT}/src/libtools/obstack.c")
+endif()

 set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibc.c"
@@ -179,11 +256,15 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibm.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl1.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl1mixer.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1net.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl1image.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1sound.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl1ttf.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl2.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl2mixer.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl2net.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsdl2image.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl2ttf.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsmpeg.c"
    "${BOX86_ROOT}/src/wrapped/wrappedsmpeg2.c"
    "${BOX86_ROOT}/src/wrapped/wrappedvorbisfile.c"
@@ -198,6 +279,7 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibxfixes.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxcursor.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxrender.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxft.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxi.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxss.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxpm.c"
@@ -205,18 +287,34 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibxdmcp.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibx11xcb.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxcb.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbxfixes.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbshape.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbshm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbrandr.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbimage.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbkeysyms.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbxtest.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbdri2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbdri3.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcbpresent.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxtst.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibxt.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcomposite.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxdamage.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxmu.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibz.c"
    "${BOX86_ROOT}/src/wrapped/wrappedopenal.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedalure.c"
    "${BOX86_ROOT}/src/wrapped/wrappedalut.c"
-    "${BOX86_ROOT}/src/wrapped/wrappedlibjpeg.c"
    "${BOX86_ROOT}/src/wrapped/wrappedcurl.c"
    "${BOX86_ROOT}/src/wrapped/wrappedudev0.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedudev1.c"
    "${BOX86_ROOT}/src/wrapped/wrappeddbus.c"
    "${BOX86_ROOT}/src/wrapped/wrappediconv.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibdrm.c"
    "${BOX86_ROOT}/src/wrapped/wrappedcrypto.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibssl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibcrypt.c"
    "${BOX86_ROOT}/src/wrapped/wrappedutil.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibuuid.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibresolv.c"
@@ -226,14 +324,20 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedfontconfig.c"
    "${BOX86_ROOT}/src/wrapped/wrappedfreetype.c"
    "${BOX86_ROOT}/src/wrapped/wrappedbz2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedexpat.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibsm.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibice.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibusb1.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibncursesw.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibformw.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibncurses.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibtinfo.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibncurses6.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibtinfo6.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgtkx112.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgdkx112.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgtk3.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgdk3.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgdkpixbuf2.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgio2.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgmodule2.c"
@@ -242,7 +346,10 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedgstreamer010.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgstinterfaces010.c"
    "${BOX86_ROOT}/src/wrapped/wrappeddbusmenugtk.c"
+    "${BOX86_ROOT}/src/wrapped/wrappeddbusmenuglib.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedappindicator.c"
    "${BOX86_ROOT}/src/wrapped/wrappedatk.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedatkbridge.c"
    "${BOX86_ROOT}/src/wrapped/wrappedpangoft2.c"
    "${BOX86_ROOT}/src/wrapped/wrappedpangocairo.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgthread2.c"
@@ -250,19 +357,67 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedpango.c"
    "${BOX86_ROOT}/src/wrapped/wrappeddbusglib1.c"
    "${BOX86_ROOT}/src/wrapped/wrappedgudev1.c"
-#    "${BOX86_ROOT}/src/wrapped/wrappedmonounity.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibnm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedtcmallocminimal.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedmpg123.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgnutls.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibcups.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedd3dadapter9.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedvulkan.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedwaylandclient.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedxml2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedxslt.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedldapr.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlber.c"
+    "${BOX86_ROOT}/src/wrapped/wrappednsl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlcms2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedkrb5.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgssapikrb5.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibtiff.c"
    "${BOX86_ROOT}/src/wrapped/wrappedldlinux.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedcrashhandler.c"
 )
+if(NOT ANDROID)
+    list(APPEND WRAPPEDS
+        "${BOX86_ROOT}/src/wrapped/wrappedlibjpeg.c"
+        "${BOX86_ROOT}/src/wrapped/wrappedlibjpeg62.c"
+        "${BOX86_ROOT}/src/wrapped/wrappedturbojpeg.c"
+    )
+endif()

+# If BOX86_ROOT contains a ".c", the build breaks...
+string(REPLACE ".c" "_private.h" MODROOT ${BOX86_ROOT})
 set(WRAPPEDS_HEAD "")
 foreach(A ${WRAPPEDS})
-    string(REPLACE ".c" "_private.h" B ${A})
+    string(REPLACE ".c" "_private.h" C ${A})
+    string(REPLACE "${MODROOT}" "${BOX86_ROOT}" B ${C})
    set(WRAPPEDS_HEAD ${WRAPPEDS_HEAD} ${B})
+    set_source_files_properties(${A} PROPERTIES OBJECT_DEPENDS ${B})
 endforeach()

 set(WRAPPER "${BOX86_ROOT}/src/wrapped/generated/wrapper.c" "${BOX86_ROOT}/src/wrapped/generated/wrapper.h")

-add_custom_command(OUTPUT ${WRAPPER} COMMAND "${PYTHON_EXECUTABLE}" "${BOX86_ROOT}/rebuild_wrappers.py" "${BOX86_ROOT}" "PANDORA" "HAVE_LD80BITS" "NOALIGN" "HAVE_TRACE" "USE_FLOAT" MAIN_DEPENDENCY "${BOX86_ROOT}/rebuild_wrappers.py" DEPENDS ${WRAPPEDS} ${WRAPPEDS_HEAD})
+add_custom_command(
+    OUTPUT "${BOX86_ROOT}/src/wrapped/generated/functions_list.txt"
+    COMMAND "${PYTHON_EXECUTABLE}" "${BOX86_ROOT}/rebuild_wrappers.py" 
+    "${BOX86_ROOT}" 
+    "PANDORA" "HAVE_LD80BITS" "NOALIGN" "HAVE_TRACE" "POWERPCLE" "ANDROID" "--" 
+    ${WRAPPEDS_HEAD}
+    MAIN_DEPENDENCY "${BOX86_ROOT}/rebuild_wrappers.py"
+    DEPENDS ${WRAPPEDS} ${WRAPPEDS_HEAD}
+    BYPRODUCTS ${WRAPPER}
+)
+
+add_custom_command(
+    OUTPUT "${BOX86_ROOT}/src/dynarec/last_run.txt"
+    COMMAND "${PYTHON_EXECUTABLE}" "${BOX86_ROOT}/rebuild_printer.py" "${BOX86_ROOT}"
+    MAIN_DEPENDENCY "${BOX86_ROOT}/rebuild_printer.py"
+    DEPENDS "${BOX86_ROOT}/src/dynarec/arm_instructions.txt"
+    BYPRODUCTS "${BOX86_ROOT}/src/dynarec/arm_printer.c"
+)
+
+add_custom_target(WRAPPERS DEPENDS "${BOX86_ROOT}/src/wrapped/generated/functions_list.txt")
+add_custom_target(PRINTER DEPENDS "${BOX86_ROOT}/src/dynarec/last_run.txt")

 if(ARM_DYNAREC)
    set(DYNAREC_SRC
@@ -274,15 +429,20 @@ if(ARM_DYNAREC)

        "${BOX86_ROOT}/src/dynarec/arm_prolog.S"
        "${BOX86_ROOT}/src/dynarec/arm_epilog.S"
-        "${BOX86_ROOT}/src/dynarec/arm_linker.S"
-        "${BOX86_ROOT}/src/dynarec/arm_table.S"
+        "${BOX86_ROOT}/src/dynarec/arm_next.S"
+        "${BOX86_ROOT}/src/dynarec/arm_lock_helper.S"
    )

    set(DYNAREC_PASS
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_helper.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_emit_tests.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_emit_math.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_emit_logic.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_emit_shift.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_pass.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_00.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_0f.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_64.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_65.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_66.c"
    "${BOX86_ROOT}/src/dynarec/dynarec_arm_67.c"
@@ -312,6 +472,10 @@ if(ARM_DYNAREC)
    set_target_properties(arm_pass2 PROPERTIES COMPILE_FLAGS "-DSTEP=2")
    add_library(arm_pass3 OBJECT ${DYNAREC_PASS})
    set_target_properties(arm_pass3 PROPERTIES COMPILE_FLAGS "-DSTEP=3")
+    add_dependencies(arm_pass0 WRAPPERS)
+    add_dependencies(arm_pass1 WRAPPERS)
+    add_dependencies(arm_pass2 WRAPPERS)
+    add_dependencies(arm_pass3 WRAPPERS)

    add_library(dynarec STATIC 
        $<TARGET_OBJECTS:dynarec_arm> 
@@ -320,30 +484,74 @@ if(ARM_DYNAREC)
        $<TARGET_OBJECTS:arm_pass2>
        $<TARGET_OBJECTS:arm_pass3>
    )
+
+endif()
+
+# creates git_head.h
+if(ARM_DYNAREC)
+    add_custom_command(
+        OUTPUT "${BOX86_ROOT}/src/git_head.h"
+        COMMAND bash -c "echo \\\#define GITREV \\\"$(git rev-parse --short HEAD)\\\">\"${BOX86_ROOT}/src/git_head.h\""
+        DEPENDS dynarec ${ELFLOADER_SRC} ${WRAPPEDS}
+        VERBATIM)
+else()
+    add_custom_command(
+        OUTPUT "${BOX86_ROOT}/src/git_head.h"
+        COMMAND bash -c "echo \\\#define GITREV \\\"$(git rev-parse --short HEAD)\\\">\"${BOX86_ROOT}/src/git_head.h\""
+        DEPENDS ${ELFLOADER_SRC} ${WRAPPEDS}
+        VERBATIM)
 endif()

 if(BOX86LIB)
-#build a static lib...
-add_library(box86 OBJECT ${ELFLOADER_SRC} ${WRAPPEDS})
+#build a static or dynamic lib...
+if(BOX86DYNAMIC)
+    add_library(box86 SHARED ${ELFLOADER_SRC} ${WRAPPEDS} "${BOX86_ROOT}/src/git_head.h")
+else()
+    add_library(${BOX86} OBJECT ${ELFLOADER_SRC} ${WRAPPEDS} "${BOX86_ROOT}/src/git_head.h")
+endif()
+if(ARM_DYNAREC)
+    if(ANDROID)
+        target_link_libraries(${BOX86} dynarec m dl)
+    else()
+        target_link_libraries(${BOX86} dynarec m dl rt pthread)
+    endif()
+endif()
+if(${CMAKE_VERSION} VERSION_LESS "3.13")
+    set_target_properties(${BOX86} PROPERTIES LINK_FLAGS -rdynamic)
+else()
+    target_link_options(${BOX86} PUBLIC -rdynamic)
+endif()
+add_dependencies(${BOX86} WRAPPERS)
+add_dependencies(${BOX86} PRINTER)
+
+add_custom_target(buildinfo ${BOX86}
+    DEPENDS dynarec ${BOX86} )

 else(BOX86LIB)

-if(USE_FLOAT)
-    set(BOX86 box86f)
-elseif(HAVE_TRACE)
-    set(BOX86 box86t)
+add_executable(${BOX86} ${ELFLOADER_SRC} ${WRAPPEDS} "${BOX86_ROOT}/src/git_head.h")
+add_dependencies(${BOX86} WRAPPERS)
+add_dependencies(${BOX86} PRINTER)
+if(ANDROID)
+    target_link_libraries(${BOX86} m dl)
 else()
-    set(BOX86 box86)
+    target_link_libraries(${BOX86} m dl rt pthread)
 endif()
-
-add_executable(${BOX86} ${ELFLOADER_SRC} ${WRAPPEDS})
-target_link_libraries(${BOX86} m dl rt pthread)
 if(ARM_DYNAREC)
    target_link_libraries(${BOX86} dynarec)
 endif()

-if(NOT NOLOADADDR)
-    target_link_libraries(${BOX86} -Wl,-Ttext-segment,0x03800000)
+if(${CMAKE_VERSION} VERSION_LESS "3.13")
+    if(NOT NOLOADADDR)
+        set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-rdynamic -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS}")
+    else()
+        set_target_properties(${BOX86} PROPERTIES LINK_FLAGS -rdynamic)
+    endif()
+else()
+    target_link_options(${BOX86} PUBLIC -rdynamic)
+    if(NOT NOLOADADDR)
+        target_link_options(${BOX86} PUBLIC -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS})
+    endif()
 endif()

 string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "i686"  _x86)
@@ -352,8 +560,10 @@ string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "x86_64"  _x86_64)
 if(NOT _x86 AND NOT _x86_64)
  install(TARGETS ${BOX86}
    RUNTIME DESTINATION bin)
-  install(FILES ${CMAKE_SOURCE_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
+  configure_file(system/box86.conf.cmake system/box86.conf)
+  install(FILES ${CMAKE_BINARY_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.6 DESTINATION /usr/lib/i386-linux-gnu/)
+  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.5 DESTINATION /usr/lib/i386-linux-gnu/)
  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libgcc_s.so.1 DESTINATION /usr/lib/i386-linux-gnu/)
 endif()

@@ -367,72 +577,88 @@ if(NOT TARGET uninstall)
        COMMAND ${CMAKE_COMMAND} -P ${CMAKE_CURRENT_BINARY_DIR}/cmake_uninstall.cmake)
 endif()

-add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+add_test(NAME bootSyscall COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test01 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME bootSyscallC COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test02 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME printf COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test03 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME args COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test04 -D TEST_ARGS2=yeah -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME maths1 COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test05 -D TEST_ARGS2=7 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME threadsStart COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test06 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME trig COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test07 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME pi COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test08 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME fork COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test09 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME cppThreads COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test10 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME tlsData COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test11 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME fpu COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test12 -D TEST_OUTPUT=tmpfile.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}/${BOX86} 
+add_test(NAME contexts COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test13 -D TEST_OUTPUT=tmpfile.txt 
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref13.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-    file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")
+add_test(NAME conditionalThreads COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test14 -D TEST_OUTPUT=tmpfile.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref14.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+# WIP
+#add_test(NAME linkingIndirectNoversion COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+#    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test15 -D TEST_OUTPUT=tmpfile.txt 
+#    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref15.txt
+#    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+#
+#add_test(NAME linkingIndirectVersion COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+#    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test16 -D TEST_OUTPUT=tmpfile.txt 
+#    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref16.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}/${BOX86}
@@ -1,54 +0,0 @@
-Compiling
----
-#### for Pandora
-
- `mkdir build; cd build; cmake .. -DPANDORA=1; make`
-
-#### for Pyra
-
- `mkdir build; cd build; cmake .. -DPYRA=1; make`
-
-#### for Gameshell
-
-`mkdir build; cd build; cmake .. -DGAMESHELL=1; make`
-
-#### for RaspberryPI
-
-  _a build for model 2, 3 and 4 can be done. Model 1 and 0 cannot (at least not with Dynarec, as they lack NEON support)_
- 
-`mkdir build; cd build; cmake .. -DRPI4=1; make`
- 
-  _For Pi4. Change to RPI2 or RPI3 for other models._
-
-#### for Other ARM Linux
-
- `mkdir build; cd build; cmake .. ; make`
-
-#### for x86 Linux
-
- `mkdir build; cd build; cmake .. -DLD80BITS=1 -DNOALIGN=1; make`
-
-#### use ccmake
-
-Alternatively, you can use the curses-bases ccmake (or any other gui frontend for cmake) to select wich platform to use interactively.
-
-#### Customize your build
-
-*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 a Trace Enabled build*
-To have a trace enabled build (warning, it will be 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 lib folders.
-
-*to have ARM Dynarec*
-The Dynarec is only avaiable on ARM Cpu. Notes also that VFPv3 and NEON are required for the Dynarec. Activate it by using `-DARM_DYNAREC=1`. Also, be sure to use `-marm` in compilation flags (because many compileur use Thumb as default, and the dynarec will not work in this mode).
-
----
-
-Testing
----
-A few tests are included.
-They can be launched with `ctest`
-They are very basic and don't test much for now.
-
@@ -1,72 +0,0 @@
-# box86
-
-Linux Userspace x86 Emulator with a twist
-
-[![Donate](https://img.shields.io/badge/Donate-PayPal-green.svg)](https://www.paypal.com/cgi-bin/webscr?cmd=_donations&business=LU8Y2T62ZWFHU)
-
-Box86 will let you run x86 Linux programs (games) on non-x86 Linux, like ARM (host system needs to be 32bit little-endian).
-
-Because Box86 uses the native version for some "system" libraries, like libc, libm, or SDL and OpenGL, it's easy to integrate and use, and performances can be surprinsigly high in some cases.
-
-Most x86 Games need OpenGL, so on ARM platforms, a solution like [gl4es](https://github.com/ptitSeb/gl4es) is probably needed.
-
-Box86 now integrate a DynaRec for ARM platform, providing a speed boost between 5 to 10 times compared to only the interpretor.
-
-Many games already work, like for example: WorldOfGoo, Airline Tycoon Deluxe or FTL. Many of the GameMaker linux games also run fine (there a long list, among them are UNDERTALE, A Risk of Rain, and Cook Server Delicious)
-
-If you are serious about developing Box86, you should install ccache and activate it's support in the cmake project (use ccmake for example)
-To have TRACE enabled (i.e. dumping to stdout all individual x86 instructions executed, with dump of registers), you'll also need [Zydis library](https://github.com/zyantific/zydis) accessible on your system.
-
-Some x86 internal opcodes use parts of "Realmode X86 Emulator Library", see [x86primop.c](src/x86primop.c) for copyright details
-
-Here are 6 videos, the first 2 of "Airline Tycoon Deluxe" and "Heretic 2" running on a gigahertz OpenPandora (the second one use the dynarec), and the next 2 of "Bit.Trip.Runner" and "Neverwinter Night" running on an ODroid XU4 (without dynarec), and the last 2 on a Pi4: Shovel Knight (video from @ITotalJustice) and Freedom Planet (video from @djazz), also without dynarec.
-
-[![Play on Youtube](https://img.youtube.com/vi/bLt0hMoFDLk/3.jpg)](https://www.youtube.com/watch?v=bLt0hMoFDLk) [![Play on Youtube](https://img.youtube.com/vi/MM7kWYts7IA/3.jpg)](https://www.youtube.com/watch?v=MM7kWYts7IA) [![Play on Youtube](https://img.youtube.com/vi/8hr71S029Hg/1.jpg)](https://www.youtube.com/watch?v=8hr71S029Hg) [![Play on Youtube](https://img.youtube.com/vi/B4YN37z3-ws/1.jpg)](https://www.youtube.com/watch?v=B4YN37z3-ws) [![Play on Youtube](https://img.youtube.com/vi/xk8Q30mxqPg/1.jpg)](https://www.youtube.com/watch?v=xk8Q30mxqPg) [![Play on Youtube](https://img.youtube.com/vi/_QMRMVvYrqU/1.jpg)](https://www.youtube.com/watch?v=_QMRMVvYrqU)
-
----
-
-Compiling
----
-How to compile can be found [here](COMPILE.md)
-
----
-
-Usage
----
-
-There are a few environment variables to control Box86 behaviour.
-
-See [here](USAGE.md) for all variables and what they do.
-
----
-
-Version history
----
-
-The change log is [here](CHANGELOG.md)
-
----
-
-A note about 64bits platform
----
-
-Because box86 works by directly translating function calls from x86 to host system, the host system (the one box86 is running on) needs to have 32bits library. Box86 doesn't include any 32bits <-> 64bits translation. So basically, to run box86 on, for example, an ARM64 platform, you will need to build box86 for arm 32bits, and also need to have a chroot with 32bits library.
-
-Also note that, even if, on day, there is a box86_64, this one will only be able to run x86_64 binary on 64bits platform. You will still need box86 (and see 32bits chroot) to run x86 binary (in fact, like it is the case on actual x86_64 linux)
-
----
-
-A note about Unity game emulation
----
-
-Running Unity games is not possible for now. Mono itself uses signals that are not well emulated enough. So the solution is to use a native version of the libmono used by Unity. It can be found here: https://github.com/Unity-Technologies/mono and it needs to be built from source. When built copy `libmonosgen-2.0.so` to `libmonounity.so` and put it somewhere it can be dlopen'd (so in `usr/lib` or friend or somewhere in your `LD_LIBRARY_PATH`).
-Note that libmonounity is not completely wrapped yet, and the mechanism to call x86 library from libmonounity is not done yet, so the use of libmonounity is not enable for now. 
-
-TL;DR: mono games are not working for now anyway.
-
----
-
-Final words
----
-
-(If you use Box86 in your project, please don't forget to mention Box86)
@@ -1,112 +0,0 @@
-Usage
----
-
-There are many environment variable to control Box86 behaviour. 
-
-##### BOX86_LOG
-Controls the Verbose level of the log
- * 0
- * NONE : No message (except some fatal error)
- * 1
- * INFO : Current default. Show some log
- * 2
- * DEBUG : Verbose a lot of stuffs (like relocations or function called)
- * 3
- * DUMP : All DEBUG plus DUMP of all ELF Info
-
-#### BOX86_LD_LIBRARY_PATH
-Controls the folder scanned to look for x86 libs. Default is current folder and `lib` in current folder.
-Also `/usr/lib/i386-linux-gnu` and `/lib/i386-linux-gnu` are added if they exists
-
-#### BOX86_PATH
-Controls the folder scanned to look for x86 executable. Default is current folder and `bin` in current folder.
-
-#### BOX86_DLSYM_ERROR
-* 0 : default. Don't log `dlsym` error
-* 1 : Log dlsym error
-
-#### BOX86_TRACE_FILE
-Send all log and trace to a file instead of `stdout`
-
-#### BOX86_TRACE
-Only on build with trace enabled. Trace allow the logging of all instruction execute, along with register dump
-* 0 : No trace
-* 1 : Trace enabled. The trace start after the init of all depending libs is done
-* symbolname : Trace only `symbolname` (trace is disable if the symbol is not found)
-* 0xXXXXXXX-0xYYYYYYY : trace only between the 2 addresses
-
-#### BOX86_TRACE_INIT
-Use BOX86_TRACE_INIT instead of BOX_TRACE to start trace before init of Libs and main program
-* 0 : No trace
-* 1 : Trace enabled. The trace start with the init of all depending libs is done
-
-#### BOX86_TRACE_START
-Only on build with trace enabled.
-* NNNNNNN : Start trace only after NNNNNNNN opcode execute (number is an `uint64_t`)
-
-#### BOX86_TRACE_XMM
-Only on build with trace enabled.
-* 0 : Default, the XMM (i.e. SSE/SSE2) register will not be logged with the general and x87 registers
-* 1 : Dump the XMM registers
-
-#### BOX86_LOAD_ADDR
-Try to load at 0xXXXXXX main binaray (if binary is a PIE)
-* 0xXXXXXXXX the load address (only active on PIE programs)
-
-#### BOX86_NOSIGSEGV
-To disable handling of SigSEGV (to ease debugging mainly)
-* 0 : default, let x86 program set sighandler for SEGV
-* 1 : disable handling of SigSEGV
-
-#### BOX86_X11COLOR16
-PANDORA only: to try convert X11 color from 32 bits to 16 bits (to avoid light green on light cyan windows
-* 0 : default, don't touch X11 colors
-* 1 : Change colors arguments in XSetForeground, XSetBackground and XCreateGC
-
-#### BOX86_X11THREADS
-To call XInitThreads when loading X11 (for old Loki games with Loki_Compat lib)
-* 0 : default, don't force call XInitThreads
-* 1 : Call XInitThreads as soon as libX11 is loaded
-
-#### BOX86_DYNAREC_LOG
-Set level of DynaRec log
- * 0 :
- * NONE : No Log for DynaRec
- * 1 :
- * INFO : Minimum Dynarec Log (only unimplemented OpCode)
- * 2 :
- * DEBUG : Debug Log for Dynarec (with detail on block created / executed)
- * 3 :
- * DUMP : All above plus dump of the x86 -> native opcode conversions
-
-#### BOX86_DYNAREC
- * 0 : Disable Dynarec
- * 1 : Enable Dynarec (default)
-
-#### BOX86_DYNAREC_LINKER
- * 0 : Disable Dynarec Linker (use that on debug, with dynarec log >= 2, to have detail on wich block get executed)
- * 1 : Enable Dynarec Linker (default)
-
-#### BOX86_DYNAREC_TRACE
- * 0 : Disable trace for generated code (default)
- * 1 : Enable trace for generated code (like regular Trace, this will slow down a lot and generate huge logs)
-
-#### BOX86_LIBGL
- * libXXXX set the name for libGL (default to libGL.so.1)
- * /PATH/TO/libGLXXX set the name and path for libGL
- You can also use SDL_VIDEO_GL_DRIVER
-
-#### BOX86_LD_PRELOAD
- * XXXX[:YYYYY] force loading XXXX (and YYYY...) libraries with the binary
- PreLoaded libs can be emulated or native, and are treated the same way as if they were comming from the binary
- 
-#### BOX86_EMULATED_LIBS
- * XXXX[:YYYYY] force lib XXXX (and YYYY...) to be emulated (and not wrapped)
-Some games use old version of some lib, with an ABI incompatible with native version.
-Note that LittleInforno for example is auto detected, and libvorbis.so.0 is automatical added to emulated libs
-
-#### BOX86_ALLOWMISSINGLIBS
-Allow box86 to continue even if a lib is missing
- * 0 : default, stop if a lib cannot be loaded
- * 1 : continue even if a needed lib cannot be loaded. Unadvised, this will, in most cases, crash later on.
-
@@ -0,0 +1 @@
+theme: jekyll-theme-hacker
@@ -0,0 +1,44 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-arm-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${_pkgname%-arm-git}-git" "${_pkgname%-arm-git}")
+conflicts=("${pkgname%-git}" "${_pkgname%-arm-git}-git" "${_pkgname%-arm-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-arm-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-arm-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-arm-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-arm-git}/build"
+	make DESTDIR="${pkgdir}/" install
+	# /usr/local/bin isn't in PATH by the default,
+	# we should move it to /usr/bin
+	cd ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,44 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-gameshell-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-gameshell-git}-git" "${pkgname%-gameshell-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-gameshell-git}-git" "${pkgname%-gameshell-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-gameshell-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-gameshell-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DGAMESHELL=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-gameshell-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-gameshell-git}/build"
+	make DESTDIR="${pkgdir}/" install
+	# /usr/local/bin isn't in PATH by the default,
+	# we should move it to /usr/bin
+	cd ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,44 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-odroid-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${_pkgname%-odroid-git}-git" "${_pkgname%-odroid-git}")
+conflicts=("${pkgname%-git}" "${_pkgname%-odroid-git}-git" "${_pkgname%-odroid-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-odroid-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-odroid-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DODROID=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-odroid-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-odroid-git}/build"
+	make DESTDIR="${pkgdir}/" install
+	# /usr/local/bin isn't in PATH by the default,
+	# we should move it to /usr/bin
+	cd ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,45 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-pandora-git
+_pkgname=box86-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-pandora-git}"-git "${pkgname%-pandora-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-pandora-git}"-git "${pkgname%-pandora-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-pandora-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-pandora-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DPANDORA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-pandora-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-git}/build"
+	make DESTDIR="${pkgdir}/" install
+	# /usr/local/bin isn't in PATH by the default,
+	# we should move it to /usr/bin
+	cd ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,45 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-pyra-git
+_pkgname=box86-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-pyra-git}-git" "${pkgname%-pyra-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-pyra-git}-git" "${pkgname%-pyra-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-pyra-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-pyra-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DPYRA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-pyra-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-pyra-git}/build"
+	make DESTDIR="${pkgdir}/" install
+	# /usr/local/bin isn't in PATH by the default,
+	# we should move it to /usr/bin
+	cd ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,60 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-rpi-git
+_pkgname=box86-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-rpi-git}-git" "${pkgname%--rpi-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-rpi-git}-git" "${pkgname%--rpi-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+_pimodel=`tr -d '\0' < /sys/firmware/devicetree/base/model`
+
+pkgver() {
+	cd "$srcdir/${pkgname%-rpi-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-rpi-git}"
+	# Check for Pi version number
+	if [[ ${_pimodel} =~ "Raspberry Pi 4" ]]; then
+		_piversion=4
+	elif [[ ${_pimodel} =~ "Raspberry Pi 3" ]]; then
+		_piversion=3
+	elif [[ ${_pimodel} =~ "Raspberry Pi 2" ]]; then
+		_piversion=2
+	else
+		# This RPI is not supported or this device is not a pi.
+		# We should cancel then the build.
+		echo -e "\033[01m\033[01;31m==> ERROR:\033[00;01m Your device is not supported! This script is for RPi model 4/3/2 (1/0 isn't supported yet).\033[00m"
+		exit 1
+	fi
+	echo -e "\033[01m\033[01;34m==> INFO:\033[00;01m Your detected Pi verion number is: \033[01;31m${_piversion}\033[01m.\033[00m"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DRPI${_piversion}=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-rpi-git}/build"
+	make -j$(nproc)
+}
+
+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 ${pkgdir}
+	mv usr/local/bin/ usr/bin/
+	# cleanup when dir is empty
+	rmdir usr/local || exit 0
+}
@@ -0,0 +1,39 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-rk3399-git
+_pkgname=box86-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-rk3399-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-rk3399-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-rk3399-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-rk3399-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,44 @@
+# Maintainer: SpacingBat3 <[email protected]>
+# Contributor: Sebastien Chevalier <[email protected]>
+pkgname=box86-x86-git
+_pkgname=box86-git
+pkgver=r2798.0edc0374
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('i386')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-x86-git}")
+conflicts=("${pkgname%-x86-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-x86-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-x86-git}"
+	if [[ ! -d ./build ]]; then
+		mkdir build && cd build
+		cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+	fi
+}
+
+build() {
+	cd "$srcdir/${pkgname%-x86-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "$srcdir/${pkgname%-x86-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 +1,21 @@
+# The directory containing all PKGBUILDs for BOX86
+## FAQ
+### 1. How to use it?
+Just download the `PKGBUILD-<your-platform>` (no need to download the whole repository), go to directory where you downloaded and type:
+```sh
+mv ./PKGBUILD-<your-platform> ./PKGBUILD
+cd ..
+makepkg -si
+```
+This will copy the PKGBUILD for your platform (where you type your platform name instead of `<your-platform>`), build it and install the PKGBUILD on 32-bit Archlinux ARM.
+
+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).
+
+### 3. Any advantages with that sollution?
+This might give you some advantages like those:
+- 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 `box86` installation is recognized as a package (where it wasn't when installing with `make install`).
@@ -0,0 +1,131 @@
+v0.2.2
+======
+* Reworked some elfloader relocation mecanism
+* Removed old "smart linker" and replace with a simple "jump table"
+* Also removed CALL/RET optomisation from Dynarec, simple Jump Table is faster
+* Reworked all memory protection tracking
+* A few more wrapped libs and wrapped functions
+* A few more opcodes
+* A few more dynarec opcodes
+* Some optimisation in Dynarec around ZF and CF flags handlings, and LL/GE/LE/G conditionnal opcodes
+* Some changes to the "cancel threads" handling
+* Some changes to the "thread once" handling
+
+v0.2.0
+======
+* Improvements on x86 Flags handling.
+* A few more opcode has been added.
+* A few potential BusError are now fixed.
+* Added the BOX86_NOVULKAN en. var. . This is used to disable the wrapping of vulkan libraries.
+* Improvments on libc O_XXXX flags handling.
+* Box86 now uses a custom allocator for Dynarec and Hash tables.
+* Improved the wrapping of pulse audio.
+* Optimisation to a few Dynarec opcode (like SHRD/SHLD).
+* Improved the tracking of memory protection.
+* A few more wrapping of libraries (like libgssapi_krb5).
+* More function wrapping on libturbojpeg
+* Preliminary support for POWER9 (ppcle) build
+* Many contributions to remove typos and rephrase the README, COMPILE and USAGE documents
+
+v0.1.8
+======
+* Fixes on some float to int x86 convertions opcodes.
+* Reworked all callback mecanisms.
+* Added libturbojpeg wrapping (and a hack for zoom to force using native one).
+* Added the BOX86_SAFEMMAP env. var.
+* Reworked Dynarec's memory manager.
+* Added a few opcodes.
+* Improve elfloader to not force PltResolver all the time (fixing SuperHexagon and maybe other)
+* Reworked Exet and Cancel Thread mecanism.
+* Added wrapped libldap_r and liblber library (used by wine)
+* Reworked Dynarec block handling, and remove the "AddMark" mecanism that wasn't efficient enough
+* Added TokiTori 2+ detection and runtime patch on the Raspberry Pi platform
+* Fixes on elfloader with TLS object.
+
+v0.1.6
+======
+* Changes in Dynarec to make flags optimizations before CALL and RET opcode less aggressive.
+* Added a Vulkan wrapper.
+* Improved wrapping of SDL_mixer/ SDL2_mixer libraries.
+* Improved wrapping of some GTK structures.
+* Added a quick and dirty wrapping of GTK3 (based on current GTK2).
+* Improved the signal handling, should be more stable now (the signal handler from syscall still need some works).
+* Added the RK3399 profile, and some hints to build 32bits box86 on 64bits OS.
+* Fixed some wrapped printf formating not handled correctly.
+* Fixed some buserror with the new LOCK mecanism handling.
+* A few more Dynarec opcodes added.
+* If winedbg is tried to be launched, exit without launching it (it doesn't work anyways).
+
+v0.1.4
+======
+* Change in Dynarec's memory handling, to simplify it (and hopefully optimized it).
+* Even more opcodes added.
+* And some more Dynarec opcodes.
+* Fixed some issue were PltResolver was injected but should not be.
+* Fixed many Dynarec and non dynarec opcodes.
+* Improved Signal handling.
+* Added a few more wrapped libraries (like curl).
+* Gallium9 is now wrapped (thx @icecream95).
+* Fixed and simplied many wrapped functions (especialy the one where a callback is involved).
+* Fixed Dynarec "freeing" a Dynablock sometimes causing a corruption of the heap (generally happens in case of JIT code).
+* Optimized the way LOCK prefix work when using Dynarec on ARM.
+* improvements to the ARM opcode Printer (for dumping Dynarec blocks) improvement (thx @rajdakin).
+
+v0.1.2
+======
+* The Dynarec now handle JIT code
+* Added support for Unity games (not perfect yet).
+* Added support for Wine (not perfect yet). You need an x86 build of Wine to use it.
+* Added support for Steam (not perfect yet). Note that Steam have limited functionalities on 32bits (only mini-mode is available).
+* More wrapped libs
+* Added support for the "PltResolver". This makes the order of libraries less important and many symbols are now resolved at runtime.
+* Added an option to build Box86 as a library (to wrap dynamic library).
+* Better Signal handling (not perfect yet).
+* More opcodes added, more opcode fixes.
+* More Dynarec opcodes.
+* Added support for FS:, and creating custom selector (needed by Wine).
+* There is now 1 x86emu_t structure per thread (simplifying/optimising many callback handling).
+* Box86 now has a logo!
+* Added options handling (only version and help for now). Now it's much usefull exept for version printing.
+
+
+v0.1.0
+=======
+* Dynarec!!! Only for ARM (note that Dynarec doesn't support JITed code for now).
+* Added real support for getcontext/set/context/makecontext/swapcontext.
+* Preliminary signal handling.
+* Fixes to SDL(1/2) Image and SDL1 Mixer, and to SDL1.2 RWops usage.
+* Fixed numerous issues in opcodes (both interpretor and dynarec). FTL works fine now, among many others.
+* Added wrapped GTK support (still ongoing, many libraries involved).
+* Make loading of libraries more configurable.
+* If a wrapped native library is not found, try to use an emulated one.
+* Added an env. var. to force the use of emulated libraries for certain libraries.
+* Added an env. var. to precise which libGL to use.
+* Added Install / Uninstall target (using systemd binfmt).
+* Added more hardware targets (RPis, GameShell...).
+* Wrapped more libraries (including FreeType, smpeg, ncurses, sndfile...).
+
+v0.0.4
+=======
+* Improved the initialisation of dependent libraries. More things work now.
+* Added a lot of wrapped functions.
+* Added a few wrapped libraries (like libz or some other x11 related library).
+* For trace enabled build, Trace can be enabled only after a certain amount of opcodes (still, a debugger would be better).
+* Some fixes in a few opcodes, and implemented x87 rounding (SuperMeatBoy behaves better now).
+* FTL 1.6.9 still has corrupted music, but older 1.5.13 seems fine (different set of libraries).
+
+v0.0.2
+=======
+* A full commercial games runs fine on the Pandora platform: Airline Tycoon Deluxe.
+* Implemented all planned subparts of Box86 except JIT support.
+* CPU Emulation is at roughly 75%. This includes x87 and SSE/SSE2. MMX is barely implemented (but barely used anyway).
+* x87 emulation is simplified, no real x87 State handling (but should not be mandatory, as the native libm library is used)
+* No Signal handling yet
+* ELF Loader is crude and probably full of bugs. Also, the initialisation of libraries are defered after all symbols are resolved for now.
+* Wrapped libraries include libc, libm, rt, pthread, libdl, dllinux, libasound, GL, GLU, SDL1/mixer/image, SDL2/mixer/image/smpeg, OpenAL/ALUT, libz, libpng16, vorbisfile, x11/xrandr/xxf86vm.
+* Most wrapped libraries are still partially implemented ( the SDL1 & GL  libraries should be complete).
+* Implemented specific mecanism for SDL(1/2) RWops, to be able to used them both in Native and x86 world.
+* WorldOfGoo works, but is painfully slow on the Pandora platform (there is too much double math, and lack of JIT).
+* FTL works, but sound is broken (issue with thread? asound? or CPU core?)
+* Limbo launches but crashes before main menu.
+
@@ -0,0 +1,81 @@
+Compiling
+----
+#### for Pandora
+
+ `mkdir build; cd build; cmake .. -DPANDORA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make`
+
+#### for Pyra
+
+ `mkdir build; cd build; cmake .. -DPYRA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make`
+
+#### for Gameshell
+
+`mkdir build; cd build; cmake .. -DGAMESHELL=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make`
+
+#### for Raspberry Pi
+
+  _a build for model 2, 3 and 4 can be done. Model 1 and 0 cannot (at least not with Dynarec, as they lack NEON support)_
+ 
+```
+git clone https://github.com/ptitSeb/box86
+cd box86
+mkdir build; cd build; cmake .. -DRPI4=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+make -j4
+sudo make install
+sudo systemctl restart systemd-binfmt
+```
+ 
+  _For Pi4. Change to RPI2 or RPI3 for other models.  Change to RPI4ARM64 for compiling on arm64. (armhf multiarch or chroot required alongside armhf gcc. Install it with 'sudo apt install gcc-arm-linux-gnueabihf'.)_
+
+#### for ODROID
+
+`mkdir build; cd build; cmake .. -DODROID=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j$(nproc)`
+
+#### for RK3399
+
+`mkdir build; cd build; cmake .. -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j4`
+
+As most RK3399 device run AARCH64 OS, you'll need an `armhf` multiarch environment, and an armhf gcc: On debian, install it with `sudo apt install gcc-arm-linux-gnueabihf`.
+
+#### for Other ARM Linux platforms
+
+ `mkdir build; cd build; cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j$(nproc)`
+
+#### for x86 Linux
+
+ `mkdir build; cd build; cmake .. -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j$(nproc)`
+
+#### use ccmake
+
+Alternatively, you can use the curses-bases ccmake (or any other gui frontend for cmake) to select wich platform to use interactively.
+
+#### Customize your build
+
+*use ccache if you have it* 
+
+Add `-DUSE_CCACHE=1` if you have ccache (it's better if you plan to touch the sources)
+
+*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 ( ***it will be 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 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`. Also, be sure to use `-marm` in compilation flags (because many compileur use Thumb as default, and the dynarec will not work in this mode).
+
+*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 box86 print).
+
+----
+
+Testing
+----
+A few tests are included.
+They can be launched with `ctest`
+They are very basic and don't test much for now.
+
@@ -1,4 +1,4 @@
-Copyright (c) 2018-2019 Sebastien Chevalier ("ptitSeb")
+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
@@ -0,0 +1,114 @@
+# box86
+
+![Official logo](img/Box86Logo.png "Official Logo")
+Linux Userspace x86 Emulator with a twist
+
+----
+
+Compiling/Installation
+----
+> Compilation instructions can be found [here](COMPILE.md)  
+> Instructions for installing Wine for Box86 can be found [here](X86WINE.md)  
+
+Box86 lets you run x86 Linux programs (such as games) on non-x86 Linux systems, like ARM (host system needs to be 32bit little-endian).
+
+You *NEED* a 32-bit subsystem to run and build Box86. Box86 is useless on 64-bit only systems. Also, you *NEED* a 32-bit toolchain to build Box86. A toolchain that only supports 64-bit will not compile Box86, and you'll get errors (typically on aarch64, you get "-marm" not recognized, and you'll need a multiarch or chroot environnement).
+
+Because Box86 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 some cases.
+
+Most x86 Games need OpenGL, so on ARM platforms an solution like [gl4es](https://github.com/ptitSeb/gl4es) is usually necessary. (Most ARM platforms only support OpenGL ES and/or their OpenGL implementation is dodgy (see OpenGL on Android).)
+
+Box86 now integrates a DynaRec (dynamic recompiler) for the ARM platform, providing a speed boost between 5 to 10 times faster than only using the interpreter.
+
+Many games already work without much tweaking, for example: WorldOfGoo, Airline Tycoon Deluxe, and FTL. Many of the GameMaker Linux games also run fine. (there's a long list, among them are UNDERTALE, A Risk of Rain, and Cook Serve Delicious)
+
+If you are serious about developing Box86, you should install ccache and build Box86 with it. (Use ccmake for example.)
+To enable TRACE (i.e. dumping to stdout all individual x86 instructions executed, with dump of registers), you'll also need [Zydis library](https://github.com/zyantific/zydis) available on your system.
+
+Some x86 internal opcodes use parts of "Realmode X86 Emulator Library", see [x86primop.c](../src/emu/x86primop.c) for copyright details
+
+Here's  6 videos, the first 2 videos are videos of "Airline Tycoon Deluxe" and "Heretic 2" running on a GigaHertz OpenPandora (the second one is using  the dynarec), and the next 2 videos are videos of of "Bit.Trip.Runner" and "Neverwinter Night" running on an ODroid XU4 (without dynarec), and the last 2  videos are on on a Pi4: Shovel Knight (video from @ITotalJustice) and Freedom Planet (video from @djazz), also without dynarec.
+
+[![Play on Youtube](https://img.youtube.com/vi/bLt0hMoFDLk/3.jpg)](https://www.youtube.com/watch?v=bLt0hMoFDLk) [![Play on Youtube](https://img.youtube.com/vi/MM7kWYts7IA/3.jpg)](https://www.youtube.com/watch?v=MM7kWYts7IA) [![Play on Youtube](https://img.youtube.com/vi/8hr71S029Hg/1.jpg)](https://www.youtube.com/watch?v=8hr71S029Hg) [![Play on Youtube](https://img.youtube.com/vi/B4YN37z3-ws/1.jpg)](https://www.youtube.com/watch?v=B4YN37z3-ws) [![Play on Youtube](https://img.youtube.com/vi/xk8Q30mxqPg/1.jpg)](https://www.youtube.com/watch?v=xk8Q30mxqPg) [![Play on Youtube](https://img.youtube.com/vi/_QMRMVvYrqU/1.jpg)](https://www.youtube.com/watch?v=_QMRMVvYrqU)
+
+You can find many more Box86 video on Youtube in the [Pi Labs Channel](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g) or [The Byteman Channel](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA)
+
+Compatibility list is there: https://github.com/ptitSeb/box86-compatibility-list/issues
+
+<img src="img/Box86Icon.png" width="96" height="96">
+
+Logo and Icon made by @grayduck, thanks!
+
+Note that this project is not to be mistaken with [86box](https://github.com/86Box/86Box), a nice "Full system" emulator specialized in early (to fairly recent) PC hardware.
+
+----
+
+Discord
+----
+
+If you want to discuss (or have any problems) about box86, feel free to join the [Pi Labs Discord Server](https://discord.gg/Fh8sjmu)!
+
+----
+
+Usage
+----
+
+There are a few environment variables to control the behaviour of Box86.
+
+See [here](USAGE.md) for all environment variables and what they do.
+
+Note: Box86'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.
+
+----
+
+Version history
+----
+
+The change log is available [here](CHANGELOG.md)
+
+----
+
+Notes about 64-bit platforms
+----
+
+Because Box86 works by directly translating function calls from x86 to host system, the host system (the one Box86 is running on) needs to have 32-bit libraries. Box86 doesn't include any 32-bit <-> 64-bit translation. So basically, to run Box86 on, for example, an ARM64 platform, you will need to build Box86 for ARM 32-bit, and also need to have a chroot with 32-bit libraries.
+
+Also note that, even if, on day, there is a Box86_64, this one will only be able to run x86_64 binaries on 64-bit platforms. You will still need Box86 (and a 32-bit chroot) to run x86 binaries (in fact, the same is the case on actual x86_64 Linux).
+
+----
+
+A note about Unity game emulation
+----
+
+Running Unity games is a hit or miss for now. Unity uses Mono (which uses signals that are not well emulated enough), and a runtime embedded in the main binary. A solution would be to use a native version of the libmono library used by Unity (it can be found here: https://github.com/Unity-Technologies/mono and needs to be built from source). But the wrapping of this lib is tricky, and not done for now, so the only solution is to emulate everything. The tricky part is to emulate the "JIT" code emitted by Mono, however with the new "protected memory" mechanism implemented it should be running with correct performance now.
+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.
+
+TL;DR: Not all Unity games work and can require a high OpenGL profile, but the speed, for the ones running, should be correct now.
+
+----
+
+Notes about Steam
+----
+
+Linux Steam's can run now with box86. But it's still a bit unstable, and not everything works:
+- First problem is Steam crashing after the sign-in window, if you encounter this issue, you may need to add libappindicator. To install it on Debian, run `sudo apt install libappindicator1`.
+- If you select to "Remember password", Steam is crashing on subsequent starts, unless you have libnm intalled. To install it on Debian, run `sudo apt install libnm0`.
+- Once open, Steam will only work on "Small Mode" and in "Big Picture", not in the regular "Large Mode". This is because some Steam components used in the browser view are only 64-bit now. So go in the "View" menu and switch to "Small view", else the list will stay empty. Alternatively, Steam can be started in small mode directly by using `+open steam://open/minigameslist` command line arguments.
+- To avoid the "libc.so.6 is absent" message, you can use `STEAMOS=1` and `STEAM_RUNTIME=1` as environment variables. 
+- Some Steam games (most Source engine games, like "Portal" or "Half-Life 2") use libtcmalloc. Box86 will detect it and will try to LD_PRELOAD it, for better compatibility. While it should work without the aformentionned feature, it is safer to add it to your system if you intend to play those game. To install it on Debian, run `sudo apt install libtcmalloc-minimal4`.
+
+Steam for Windows installs fine but doesn't work yet.
+
+----
+
+Notes about Wine
+----
+
+Wine is now partly supported. Wine integrated program all runs, and some windows programs and games now runs fine. Don't forget most Windows games use Direct3D, this may require a complete OpenGL driver and as high profile as possible (and gl4es with ES2 backend have issue with Wine for now). Also, vulkan is not wrapped on box86, so vk3d is not usable yet, even if supported by the hardware.
+Note: if you plan to use box86 with Wine on Raspberry Pi 3 or earlier, those models use a default OS that have a kernel with a 2/2 Split (meaning 2G of space for user program, and 2G of space for the Kernel). This is not compatible with Wine programs that needs to access memory > 2Gb address. So you'll need to reconfigure your kernel for a 3G/1G split.
+----
+
+Final words
+----
+
+(If you use Box86 in your project, please don't forget to mention Box86!)
@@ -0,0 +1,161 @@
+
+Usage
+----
+
+There are many environment variables to control Box86 behaviour. 
+
+#### BOX86_LOG
+Controls the Verbosity level of the logs
+ * 0: NONE : No message (except some fatal error). (Default.)
+ * 1: INFO : Show some minimum log (Example: librairies not found)
+ * 2: DEBUG : Details a lot of stuff (Example: relocations or functions called).
+ * 3: DUMP : All DEBUG plus DUMP of all ELF Info.
+
+#### BOX86_NOBANNER
+Disables Box86 printing its version and build
+ * 0 : Enable printing its banner. (Default.)
+ * 1 : Disable printing its banner. 
+
+#### BOX86_LD_LIBRARY_PATH
+Path to look for x86 libraries. Default is current folder and `lib` in current folder.
+Also, `/usr/lib/i386-linux-gnu` and `/lib/i386-linux-gnu` are added if they exist.
+
+#### BOX86_PATH
+Path to look for x86 executable. Default is current folder and `bin` in current folder.
+
+#### BOX86_DLSYM_ERROR
+Enables/Disables the logging of `dlsym` errors.
+ * 0 : Don't log `dlsym` errors. (Default.)
+ * 1 : Log dlsym errors.
+
+#### BOX86_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 box86 instance
+
+#### BOX86_TRACE
+Only on build with trace enabled. Trace allow the logging of all instruction executed, along with register dump
+ * 0 : No trace. (Default.) 
+ * 1 : Trace enabled. Trace start after the initialisation of all depending libraries is done.
+ * symbolname : Trace only `symbolname` (trace is disable if the symbol is not found).
+ * 0xXXXXXXX-0xYYYYYYY : Trace only between the 2 addresses.
+
+#### BOX86_TRACE_INIT
+Use BOX86_TRACE_INIT instead of BOX_TRACE to start trace before the initialisation of libraries and the running program
+ * 0 : No trace. (Default.)
+ * 1 : Trace enabled. The trace start with the initialisation of all depending libraries is done.
+
+#### BOX86_TRACE_START
+Only on builds with trace enabled.
+ * NNNNNNN : Start trace only after NNNNNNNN opcode execute (number is an `uint64_t`).
+
+#### BOX86_TRACE_XMM
+Only on builds with trace enabled.
+ * 0 : The XMM (i.e. SSE/SSE2) register will not be logged with the general and x86 registers. (Default.)
+ * 1 : Dump the XMM registers.
+
+#### BOX86_LOAD_ADDR
+Try to load at 0xXXXXXX main binary (if binary is a PIE)
+ * 0xXXXXXXXX : The load address . (Only active on PIE programs.)
+
+#### BOX86_NOSIGSEGV
+Disable handling of SigSEGV. (Very useful for debugging.)
+ * 0 : Let the x86 program set sighandler for SEGV (Default.)
+ * 1 : Disable the handling of SigSEGV.
+
+#### BOX86_NOSIGILL
+Disable handling of SigILL (to ease debugging mainly).
+ * 0 : Let x86 program set sighandler for Illegal Instruction
+ * 1 : Disables the handling of SigILL 
+
+#### BOX86_X11COLOR16
+PANDORA only: Try converting X11 color from 32 bits to 16 bits (to avoid light green on light cyan windows).
+ * 0 : Don't touch X11 colors. (Default.)
+ * 1 : Change colors arguments in XSetForeground, XSetBackground and XCreateGC.
+
+#### BOX86_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.)
+ * 1 : Call XInitThreads as soon as libX11 is loaded.
+
+#### BOX86_X11GLX
+Force libX11's GLX extension to be present.
+* 0 : Do not force libX11's GLX extension to be present. 
+* 1 : GLX will always be present when using XQueryExtension. (Default.)
+
+#### BOX86_DYNAREC_DUMP
+Enables/disables Box86's Dynarec's dump.
+ * 0 : Disable Dynarec's blocks dump. (Default.)
+ * 1 : Enable Dynarec's blocks dump.
+ * 2 : Enable Dynarec's blocks dump with some colors.
+
+#### BOX86_DYNAREC_LOG
+Set the level of DynaRec's logs.
+ * 0 : NONE : No Logs for DynaRec. (Default.)
+ * 1 :INFO : Minimum Dynarec Logs (only unimplemented OpCode).
+ * 2 : DEBUG : Debug Logs for Dynarec (with details on block created / executed).
+ * 3 : VERBOSE : All of the above plus more.
+
+#### BOX86_DYNAREC
+Enables/Disables Box86's Dynarec.
+ * 0 : Disables Dynarec.
+ * 1 : Enable Dynarec. (Default.)
+
+#### BOX86_DYNAREC_TRACE
+Enables/Disables trace for generated code.
+ * 0 : Disable trace for generated code. (Default.)
+ * 1 : Enable trace for generated code (like regular Trace, this will slow down the program a lot and generate huge logs).
+
+#### BOX86_NODYNAREC 
+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)
+
+#### BOX86_LIBGL
+ * libXXXX set the name for libGL (defaults to libGL.so.1).
+ * /PATH/TO/libGLXXX : Sets the name and path for libGL
+ You can also use SDL_VIDEO_GL_DRIVER
+
+#### BOX86_LD_PRELOAD
+ * XXXX[:YYYYY] force loading XXXX (and YYYY...) libraries with the binary
+ PreLoaded libs can be emulated or native, and are treated the same way as if they were comming from the binary
+ 
+#### BOX86_EMULATED_LIBS
+ * XXXX[:YYYYY] force lib XXXX (and YYYY...) to be emulated (and not wrapped)
+Some games uses an old version of some libraries, with an ABI incompatible with native version.
+Note that LittleInferno for example is auto detected, and libvorbis.so.0 is automatical added to emulated libs, and same for Don't Starve (and Together / Server variant) that use an old SDL2 too
+
+#### BOX86_ALLOWMISSINGLIBS
+Allow Box86 to continue even if a library is missing.
+ * 0 : Box86 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.
+
+#### BOX86_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
+
+#### BOX86_NOGTK
+Disables the loading of wrapped GTK libraries.
+ * 0 : Load wrapped GTK libraries if found. (Default.)
+ * 1 : Disables loading wrapped GTK libraries (can be used with Steam, along with STEAM_RUNTIME=1 to use the i386 version of GTK).
+
+#### BOX86_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.)
+
+#### BOX86_FIX_64BIT_INODES
+Fix/Don't fix 64bit inodes
+ * 0 : Don't fix 64bit inodes. (Default.)
+ * 1 : Fix 64bit inodes. Helps when running on Filesystems with 64bit inodes. Is useful when a program uses API functions which doesn't support it and the program doesn't use inodes information.
+
+#### BOX86_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_box86signalhandler` in is stack)
+ then probably need to `finish` 1 or 2 functions (inside `usleep(..)`) and then you'll be in `my_box86signalhandler`, 
+ just before the printf of the Segfault message. Then simply 
+ `set waiting=0` to exit the infinite loop.
+ * 2 : Launch `gdbserver` when a segfault, bus error or illegal instruction signal is trapped, attached to the offending process, and go in an endless loop, waiting.
+ Use `gdb /PATH/TO/box86` and then `target remote 127.0.0.1:1234` to connect to the gdbserver (or use actual IP if not on the machine). After that, the procedure is the same as with ` BOX86_JITGDB=1`.
+ This mode can be usefullwhen programs redirect all console output to a file (like Unity3D Games)
@@ -0,0 +1,143 @@
+# Installing Wine (and winetricks)
+Using Wine with Box86 allows (x86) Windows programs to run on ARM Linux computers (x64 is not yet implemented).  
+See installation steps below (in the [Examples](#examples) section).
+
+_TwisterOS users, take note: Wine, winetricks, and Box86 are pre-installed in TwisterOS. TwisterOS users do not have to install anything._
+
+Box86 actually needs `wine-i386` to be installed **manually** on ARM devices.  Even though `wine-armhf` is available in many repo's on ARM devices (ie using _apt-get_ will attempt to install `wine-armhf` by default), `wine-armhf` will not work with Box86.  Note that manual installation is required since using `multiarch` will result in your ARM device thinking it needs to install lots of i386 dependencies to make `wine-i386` work. The "twist" in Box86 is that Box86 "wraps" many Windows and Wine i386 libraries (`.so` or `.dll` files) so that they will work with your ARM device's libraries. Also note that wrapping libraries is an ongoing process throughout Box86 development and that some programs may not run properly until all of their i386 library dependencies are wrapped.
+
+Installation files for Wine can be found in the [WineHQ repository](https://dl.winehq.org/wine-builds/debian/dists/buster/main/binary-i386/), the [TwisterOS FAQ](https://twisteros.com/faq.html) page, or the [PlayOnLinux website repository](https://www.playonlinux.com/wine/). Box86 requires the "i386" (x86) versions of Wine install files (even though we are installing it on an ARM processor). Box86 "wraps" many of Wine's core Linux i386 libraries so that their calls are interpretable by Linux ARM libraries. Below are examples of how to install Wine from each of those places.
+
+_Raspberry Pi users: Wine requires a 3G/1G split memory kernel. Raspberry Pi OS for the Pi 4 has a 3G/1G split kernel, but **the Pi 3 and earlier models have a 2G/2G kernel by default and will need you to install a custom-compiled 3G/1G kernel to get Wine to work.**
+
+## Overview
+The general procedure for installing Wine for Box86 is to...
+ - Download all the install files (.deb, .zip, or even .pol 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 (in TwisterOS, this is `~/wine/` by default)
+ - Go to `/usr/local/bin` and make symlinks or scripts that will point to your main wine binaries (`wine`, `winecfg`, and `wineserver`).
+ - Boot wine to create a new wineprefix (`wine wineboot`).
+ - Download winetricks (which is simply a very complicated bash script), make it executable, then copy it to `/usr/local/bin`.
+
+## Examples
+_These examples were all tested on Raspberry Pi 4.  Note the similarities between the different install methods._
+
+### Installing wine-devel, wine-stable, or wine-staging for Raspberry Pi OS Buster from WineHQ deb files
+_Links from the [WineHQ repo](https://dl.winehq.org/wine-builds/debian/dists/buster/main/binary-i386/)_
+```
+# Backup any old wine installations
+sudo mv ~/wine ~/wine-old
+sudo mv ~/.wine ~/.wine-old
+sudo mv /usr/local/bin/wine /usr/local/bin/wine-old
+sudo mv /usr/local/bin/wineboot /usr/local/bin/wineboot-old
+sudo mv /usr/local/bin/winecfg /usr/local/bin/winecfg-old
+sudo mv /usr/local/bin/wineserver /usr/local/bin/wineserver-old
+
+# Download, extract wine, and install wine
+# (Replace the links/versions below with links/versions from the WineHQ site for the version of wine you wish to install. Note that we need the i386 version for Box86 even though we're installing it on our ARM processor.)
+# (Pick an i386 version of wine-devel, wine-staging, or wine-stable)
+cd ~/Downloads
+wget https://dl.winehq.org/wine-builds/debian/dists/buster/main/binary-i386/wine-devel-i386_5.21~buster_i386.deb # NOTE: Replace this link with the version you want
+wget https://dl.winehq.org/wine-builds/debian/dists/buster/main/binary-i386/wine-devel_5.21~buster_i386.deb  # NOTE: Also replace this link with the version you want
+dpkg-deb -xv wine-devel-i386_5.21~buster_i386.deb wine-installer # NOTE: Make sure these dpkg command matches the filename of the deb package you just downloaded
+dpkg-deb -xv wine-devel_5.21~buster_i386.deb wine-installer
+mv ~/Downloads/wine-installer/opt/wine* ~/wine
+rm wine*.deb # clean up
+rm -rf wine-installer # clean up
+
+# Install shortcuts (make 32bit launcher & symlinks. Credits: grayduck, Botspot)
+echo -e '#!/bin/bash\nsetarch linux32 -L '"$HOME/wine/bin/wine "'"$@"' | sudo tee -a /usr/local/bin/wine >/dev/null # Create a script to launch wine programs as 32bit only
+#sudo ln -s ~/wine/bin/wine /usr/local/bin/wine # You could aslo just make a symlink, but box86 only works for 32bit apps at the moment
+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/wineboot /usr/local/bin/winecfg /usr/local/bin/wineserver
+
+# These packages are needed for running wine-staging on RPi 4 (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
+
+# Boot wine (make fresh wineprefix in ~/.wine )
+wine wineboot
+```
+
+### Installing Wine for Raspberry Pi OS from the Twister OS FAQ page
+_Links from the [TwisterOS FAQ](https://twisteros.com/faq.html)_
+```
+# Backup any old wine installations
+sudo mv ~/wine ~/wine-old
+sudo mv ~/.wine ~/.wine-old
+sudo mv /usr/local/bin/wine /usr/local/bin/wine-old
+sudo mv /usr/local/bin/wineboot /usr/local/bin/wineboot-old
+sudo mv /usr/local/bin/winecfg /usr/local/bin/winecfg-old
+sudo mv /usr/local/bin/wineserver /usr/local/bin/wineserver-old
+
+# Download, extract wine, and install wine (last I checked, the Twister OS FAQ page had Wine 5.13-devel)
+wget https://twisteros.com/wine.tgz -O ~/wine.tgz
+tar -xzvf ~/wine.tgz
+rm ~/wine.tgz # clean up
+
+# Install shortcuts (make launcher & symlinks. Credits: grayduck, Botspot)
+echo -e '#!/bin/bash\nsetarch linux32 -L '"$HOME/wine/bin/wine "'"$@"' | sudo tee -a /usr/local/bin/wine >/dev/null # Create a script to launch wine programs as 32bit only
+#sudo ln -s ~/wine/bin/wine /usr/local/bin/wine # You could aslo just make a symlink, but box86 only works for 32bit apps at the moment
+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/wineboot /usr/local/bin/winecfg /usr/local/bin/wineserver
+
+# Boot wine (make fresh wineprefix in ~/.wine )
+wine wineboot
+```
+## Installing winetricks
+Winetricks is a script which makes it easier to install & configure any desired Windows core system software packages which may be dependencies for certain Windows programs.
+
+```
+# Backup old winetricks
+sudo mv /usr/local/bin/winetricks /usr/local/bin/winetricks-old
+
+# Download
+cd ~/Downloads
+wget https://raw.githubusercontent.com/Winetricks/winetricks/master/src/winetricks
+
+# Install
+sudo chmod +x winetricks
+sudo mv winetricks /usr/local/bin/
+
+# winetricks needs this installed
+sudo apt-get install cabextract -y
+```
+
+If you have a desktop menu itm for wintricks, you may need to edit it to add the nobanner env var. Do so by edition `/usr/share/applications/winetricks.desktop` with your favorite text editor (not that you'll need `sudo` to save the file) and change from `Exec=winetricks --gui` to `Exec=env BOX86_NOBANNER=1 winetricks --gui`
+
+## Example commands
+Here is an example of how we should run a winetricks command with box86:  
+`BOX86_NOBANNER=1 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._
+
+Whenever we run winetricks, we must suppress Box86's banner by typing `BOX86_NOBANNER=1` to prevent winetricks from crashing.  Invoking Box86's logging features (with `BOX86_LOG=1` or similar) will also cause winetricks to crash.  (If Box86 logging is necessary though, we can patch winetricks to avoid these crashes - see the *Troubleshooting* section) # future work.
+
+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)
+The first time Wine is run (`wine wineboot`), it will create a fresh wineprefix for you (by default, located in the hidden folder `~/.wine`).  Think of a wineprefix as Wine's virtual 'harddrive' where it installs software and saves settings. Wineprefixes are portable and deletable.
+
+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 (side-loading)
+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). For more Wine documentation, see [WineHQ](https://www.winehq.org/documentation).
+
+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 make a fresh wineprefix with your new version of Wine. You can check which version of Wine you're running with the `wine --version` command.
@@ -0,0 +1 @@
+theme: jekyll-theme-hacker
@@ -1,19 +1,70 @@
 #!/usr/bin/env python

 import os
-import glob
 import sys

-values = ['E', 'e', 'v', 'c', 'w', 'i', 'I', 'C', 'W', 'u', 'U', 'f', 'd', 'D', 'K', 'l', 'L', 'p', 'V', "O", "S"]
+"""
+Generates all files in src/wrapped/generated
+===
+
+TL;DR: Automagically creates type definitions (/.F.+/ functions/typedefs...).
+       All '//%' in the headers are used by the script.
+
+Reads each lines of each "_private.h" headers (plus wrappedd3dadapter9_gen.h as a special case).
+For each of them:
+- If if starts with a #ifdef, #else, #ifndef, #endif, it memorizes which definition is required
+- If it starts with a "GO", it will do multiple things:
+  - It memorizes the type used by the function (second macro argument)
+  - It memorizes the type it is mapped to, if needed (eg, iFEv is mapped to iFE: the second argument is dropped)
+  - It checks if the type given (both original and )
+  - If the signature contains a 'E', it will memorize the original function's signature:
+    - If it is the special file case, or if the line contains "//%%", it stops there.
+    - Otherwise, if it starts with "GOS":
+      - It will search for the special marker "//%".
+        That special marker is used to detect the following step:
+        - If the marker is followed by a registered structure ID (see later), it transforms the assembly function signature
+          into the original signature (eg, transforms pFEp into XFv if //%X was used)
+        - If the marker is followed by a '{', it will not attempt to alter the given signature, and instead use the
+          signature between '{' and '}'
+    - Otherwise, if the line contains '//%{', like before it will use the signature given and not the original one.
+    - Finally, if it still didn't stop, it will use the assembly signature (without the emulator).
+- If the line starts with a '//%'
+
+`gbl` contains the first list, `redirects` the second, `mytypedefs` the third.
+`mystructs` and `mystructs_vals` constains data about the structures.
+
+After sorting the data, it generates:
+
+wrapper.c
+---------
+(Private) type definitions (/.F.+_t/)
+Function definitions (/.F.+/ functions, that actually execute the function given as argument)
+isSimpleWrapper definition
+
+wrapper.h
+---------
+Generic "wrapper_t" type definition
+Function declarations (/.F.+/ functions)
+
+*types.h
+--------
+Local types definition, for the original signatures (eg, XFv for the earlier example)
+The SUPER() macro definition, used to generate and initialize the `*_my_t` library structure
+(TODO: also automate this declaration/definition? It would require more metadata,
+ and may break sometime in the future due to the system changing...)
+"""
+
+# Free values:
+# AB   F H J      QR T   XYZab    gh jk mno qrst   xyz
+values = ['E', 'e', 'v', 'c', 'w', 'i', 'I', 'C', 'W', 'u', 'U', 'f', 'd', 'D', 'K', 'l', 'L', 'p', 'V', 'O', 'S', '2', 'P', 'G', 'N', 'M']
 def splitchar(s):
-	ret = [len(s)]
-	i = 0
-	for c in s:
-		i = i + 1
-		if i == 2:
-			continue
-		ret.append(values.index(c))
-	return ret
+	try:
+		ret = [len(s), values.index(s[0])]
+		for c in s[2:]:
+			ret.append(values.index(c))
+		return ret
+	except ValueError as e:
+		raise ValueError("Value is " + s) from e

 def value(define):
 	return define[9:-1] if define.startswith("!") else define[8:-1]
@@ -36,19 +87,30 @@ def splitdef(dnf, defines):
 def invert(define):
 	return define[1:] if define.startswith("!") else ("!" + define)

-def main(root, defines, ver):
-	# Initialize variables: gbl for all values, vals for file-per-file values, redirects for redirections
-	gbl = {}
-	vals = {}
-	redirects = {}
+def main(root, defines, files, ver):
+	global values
+	
+	# Initialize variables: gbl for all values, redirects for redirections
+	# mytypedefs is a list of all functions per "*FE*" types per filename, mystructs of structures per filename
+	gbl        = {}
+	redirects  = {}
+	mytypedefs = {}
+	mystructs  = {}
+	mystructs_vals = {}
 	
 	# First read the files inside the headers
-	for filepath in glob.glob(os.path.join(root, "src/wrapped/", "wrapped*_private.h")):
+	# TODO: remove the special case...
+	for filepath in files + [os.path.join(root, "src", "wrapped", "wrappedd3dadapter9_gen.h")]:
 		filename = filepath.split("/")[-1]
-		locval = {}
 		dependants = []
 		with open(filepath, 'r') as file:
 			for line in file:
+				def fail(s, causedby=None):
+					if causedby:
+						raise NotImplementedError(s + " ({0}:{1})".format(filename, line[:-1])) from causedby
+					else:
+						raise NotImplementedError(s + " ({0}:{1})".format(filename, line[:-1]))
+				
 				ln = line.strip()
 				# If the line is a `#' line (#ifdef LD80BITS/#ifndef LD80BITS/header)
 				if ln.startswith("#"):
@@ -60,6 +122,8 @@ def main(root, defines, ver):
 							continue #if !(defined(GO) && defined(GOM)...)
 						elif preproc_cmd.startswith("error"):
 							continue #error meh!
+						elif preproc_cmd.startswith("include"):
+							continue #inherit other library
 						elif preproc_cmd.startswith("endif"):
 							if dependants != []:
 								dependants.pop()
@@ -68,52 +132,156 @@ def main(root, defines, ver):
 								raise KeyError(preproc_cmd[5:].strip())
 							dependants.append("defined(" + preproc_cmd[5:].strip() + ")")
 						elif preproc_cmd.startswith("ifndef"):
-							if preproc_cmd[5:].strip() not in defines:
-								raise KeyError(preproc_cmd[5:].strip())
+							if preproc_cmd[6:].strip() not in defines:
+								raise KeyError(preproc_cmd[6:].strip())
 							dependants.append("!defined(" + preproc_cmd[6:].strip() + ")")
 						elif preproc_cmd.startswith("else"):
 							dependants[-1] = invert(dependants[-1])
 						else:
-							raise NotImplementedError("Unknown preprocessor directive: {0} ({1}:{2})".format(
-								preproc_cmd.split(" ")[0], filename, line[:-1]
-							))
+							fail("Unknown preprocessor directive: {0}".format(preproc_cmd.split(" ")[0]))
 					except KeyError as k:
-						raise NotImplementedError("Unknown key: {0} ({1}:{2})".format(
-							k.args[0], filename, line[:-1]
-						), k)
+						fail("Unknown key: {0}".format(k.args[0]), k)
 				# If the line is a `GO...' line (GO/GOM/GO2/...)...
 				elif ln.startswith("GO"):
 					# ... then look at the second parameter of the line
+					gotype = ln.split("(")[0].strip()
+					funname = ln.split(",")[0].split("(")[1].strip()
 					ln = ln.split(",")[1].split(")")[0].strip()
 					
-					if ln[1] not in ["F"]:
-						raise NotImplementedError("Bad middle letter {0} ({1}:{2})".format(ln[1], filename, line[:-1]))
+					if len(ln) < 3:
+						fail("Type {0} too short".format(ln))
+					if "E" in ln:
+						if ("E" in ln[:2]) or ("E" in ln[3:]):
+							fail("emu64_t* not as the first parameter")
+						if len(ln) < 4:
+							fail("Type {0} too short".format(ln))
+					if ln[0] not in values:
+						fail("Invalid return type {0}".format(ln[0]))
+					
+					if ln[1] != "F":
+						fail("Bad middle letter {0}".format(ln[1]))
 					if any(c not in values for c in ln[2:]) or (('v' in ln[2:]) and (len(ln) > 3)):
 						old = ln
 						# This needs more work
-						acceptables = ['v', 'o', '0', '1'] + values
+						acceptables = ['v', '0', '1'] + values
 						if any(c not in acceptables for c in ln[2:]):
-							raise NotImplementedError("{0} ({1}:{2})".format(ln[2:], filename, line[:-1]))
-						# Ok, this is acceptable: there is 0, 1, stdout and void
-						ln = (ln
+							fail("Invalid type {0}".format(ln[2:]))
+						# Ok, this is acceptable: there is 0, 1 and/or void
+						ln = ln[:2] + (ln[2:]
 							.replace("v", "")   # void   -> nothing
-							.replace("o", "p")  # stdout -> pointer
 							.replace("0", "p")  # 0      -> pointer
 							.replace("1", "i")) # 1      -> integer
 						redirects.setdefault(" && ".join(dependants), {})
 						redirects[" && ".join(dependants)][old] = ln
 					# Simply append the function name if it's not yet existing
-					locval.setdefault(" && ".join(dependants), [])
 					gbl.setdefault(" && ".join(dependants), [])
-					if ln not in locval[" && ".join(dependants)]:
-						locval[" && ".join(dependants)].append(ln)
 					if ln not in gbl[" && ".join(dependants)]:
 						gbl[" && ".join(dependants)].append(ln)
-		
-		# Sort the file local values and add it to the dictionary
-		for k in locval:
-			locval[k].sort(key=splitchar)
-		vals[filename] = locval
+					
+					if filename == "wrappedd3dadapter9_gen.h":
+						pass # Special case...
+					elif ln[2] == 'E':
+						if "//%%" in line:
+							# Do not dlsym functions containing "//%%" as metadata
+							pass
+						elif gotype == "GOS":
+							# Scan the rest of the line to extract the return structure ID
+							if filename[:-10] not in mystructs:
+								fail("No structure info in the file")
+							if "//%" not in line:
+								fail("Invalid GOS (missing structure ID info)")
+							if ln[0] != 'p':
+								fail("Invalid GOS return type ('{0}' and not 'p')".format(ln[0]))
+							#if (ln[2] != 'p') and ((ln[2] != 'E') or (ln[3] != 'p')): -> only allow pFEp for now
+							#	fail("Invalid GOS first parameter ('{0}' and not 'p' or 'Ep')".format(ln[2:4]))
+							if (ln[2] != 'E') or (ln[3] != 'p'):
+								fail("Invalid GOS first parameter ('{0}' and not 'Ep')".format(ln[2:4]))
+							
+							sid = line.split("//%")[1].split(" ")[0].strip()
+							if sid[0] == '{':
+								# Change type completely, not just the return type...
+								if sid[-1] != '}':
+									fail("Invalid type (EOL or space met, expected '}')")
+								if len(sid) < 5:
+									fail("Invalid type (Type {0} too short)".format(sid[1:-1]))
+								if sid[2] != "F":
+									fail("Invalid type (Bad middle letter {0})".format(sid[2]))
+								inval_char = lambda c:\
+									(c not in mystructs[filename[:-10]]) and (c not in acceptables) \
+									or (c == 'E') or (c == 'e')
+								if inval_char(sid[1]):
+									fail("Invalid type (Invalid return type {0})".format(sid[1]))
+								if any(map(inval_char, sid[3:-1])):
+									fail("Invalid type (Invalid type {0})".format(sid[3:-1]))
+								mytypedefs.setdefault(filename[:-10], {})
+								mytypedefs[filename[:-10]].setdefault(sid[1:-1], [])
+								mytypedefs[filename[:-10]][sid[1:-1]].append((0, funname))
+							else:
+								if len(sid) != 1:
+									fail("Invalid structure ID {0} (length is too big)".format(sid))
+								if sid not in mystructs[filename[:-10]]:
+									fail("Invalid structure ID {0} (unknown ID)".format(sid))
+								mytypedefs.setdefault(filename[:-10], {})
+								mytypedefs[filename[:-10]].setdefault(sid + "F" + ln[4:], [])
+								mytypedefs[filename[:-10]][sid + "F" + ln[4:]].append((1, funname))
+						elif "//%{" in line:
+							# Change type completely...
+							# ...Maybe?
+							if filename[:-10] not in mystructs:
+								fail("No structure info in the file")
+							
+							newtype = line.split("//%")[1].split(" ")[0].strip()
+							if newtype[-1] != '}':
+								fail("Invalid type (EOL or space met, expected '}')")
+							if len(newtype) < 5:
+								fail("Invalid type (Type {0} too short)".format(newtype[1:-1]))
+							if newtype[2] != "F":
+								fail("Invalid type (Bad middle letter {0})".format(newtype[2]))
+							inval_char = lambda c:\
+								(c not in mystructs[filename[:-10]]) and (c not in acceptables) \
+								or (c == 'E') or (c == 'e')
+							if inval_char(newtype[1]):
+								fail("Invalid type (Invalid return type {0})".format(newtype[1]))
+							if any(map(inval_char, newtype[3:-1])):
+								fail("Invalid type (Invalid type {0})".format(newtype[3:-1]))
+							mytypedefs.setdefault(filename[:-10], {})
+							mytypedefs[filename[:-10]].setdefault(newtype[1:-1], [])
+							mytypedefs[filename[:-10]][newtype[1:-1]].append((2, funname))
+						else:
+							# filename isn't stored with the '_private.h' part
+							if not filename.endswith('_private.h'):
+								fail("??? {0}".format(filename))
+							if len(ln) > 3:
+								ln = ln[:2] + ln[3:]
+							else:
+								ln = ln[:2] + "v"
+							mytypedefs.setdefault(filename[:-10], {})
+							mytypedefs[filename[:-10]].setdefault(ln, [])
+							mytypedefs[filename[:-10]][ln].append((3, funname))
+				elif ln.startswith("//%S"):
+					# Extract a structure ID-name pair
+					data = [s for s in map(lambda s: s.strip(), ln.split(" ")) if s != ""]
+					if len(data) != 3:
+						fail("Too much data ({0})".format(len(data)))
+					if not filename.endswith('_private.h'):
+						fail("??? {0}".format(filename))
+					if (data[0] != "//%S") or (len(data[1]) != 1):
+						fail("Invalid structure data {0} {1}".format(data[0], data[1]))
+					if data[1] in values:
+						fail("{0} cannot be used as a structure type".format(data[1]))
+					mystructs.setdefault(filename[:-10], {})
+					if data[1] in mystructs[filename[:-10]]:
+						fail("Duplicate structure ID {0} ({1}/{2})".format(data[1], mystructs[filename[:-10]], data[2]))
+					mystructs[filename[:-10]][data[1]] = data[2]
+					mystructs_vals.setdefault(filename[:-10], [])
+					mystructs_vals[filename[:-10]].append(data[1])
+	
+	if ("" not in gbl) or ("" not in redirects):
+		print("\033[1;31mThere is suspiciously not many types...\033[m")
+		print("Check the CMakeLists.txt file. If you are SURE there is nothing wrong"
+			  " (as a random example, `set()` resets the variable...), then comment out the following return.")
+		print("(Also, the program WILL crash later if you proceed.)")
+		return 2 # Check what you did, not proceeding
 	
 	gbl_vals = {}
 	for k in gbl:
@@ -180,30 +348,68 @@ def main(root, defines, ver):
 	redirects_idxs = []
 	for k, v in redirects_vals:
 		key = "(" + (") || (".join(redirects_vals[(k, v)])) + ")"
-		val = redirects.get(key, {})
-		val[k] = v
-		redirects[key] = val
+		if key in redirects:
+			redirects[key].append([k, v])
+		else:
+			redirects[key] = [[k, v]]
 		if (key not in redirects_idxs) and (key != "()"):
 			redirects_idxs.append(key)
 	redirects_idxs.sort(key=lambda v: splitdef(v, defines))
 	
-	# Sort the table
+	# Sort the tables
 	for k in gbl:
 		gbl[k].sort(key=lambda v: splitchar(v))
+	values = values + ['0', '1']
+	for k in redirects:
+		redirects[k].sort(key=lambda v: splitchar(v[0]) + [-1] + splitchar(v[1]))
+	values = values[:-2]
+	mytypedefs_vals = {}
+	for fn in mytypedefs:
+		if fn in mystructs:
+			values = values + list(mystructs[fn].keys())
+		mytypedefs_vals[fn] = sorted(mytypedefs[fn].keys(), key=lambda v: splitchar(v))
+		if fn in mystructs:
+			values = values[:-len(mystructs[fn])]
+		for v in mytypedefs_vals[fn]:
+			mytypedefs[fn][v].sort()
 	
-	def length(s):
-		l = len(s)
-		if l < 10:
-			ret = "0" + str(l)
-		else:
-			ret = str(l)
-		return ret
+	# Check if there was any new functions
+	functions_list = ""
+	for k in ["()"] + gbl_idxs:
+		for v in gbl[k]:
+			functions_list = functions_list + "#" + k + " " + v + "\n"
+	for k in ["()"] + redirects_idxs:
+		for v in redirects[k]:
+			functions_list = functions_list + "#" + k + " " + v[0] + " -> " + v[1] + "\n"
+	for fn in sorted(mystructs.keys()):
+		for t in mystructs_vals[fn]:
+			# Structure Definition
+			functions_list = functions_list + fn + "/Sd" + t + mystructs[fn][t] + "\n"
+	for fn in sorted(mytypedefs.keys()):
+		for t in mytypedefs_vals[fn]:
+			# Structure Usage
+			for tnum, f in mytypedefs[fn][t]:
+				functions_list = functions_list + fn + "/Su" + t + str(tnum) + f + "\n"
+	
+	# functions_list is a unique string, compare it with the last run
+	try:
+		last_run = ""
+		with open(os.path.join(root, "src", "wrapped", "generated", "functions_list.txt"), 'r') as file:
+			last_run = file.read()
+		if last_run == functions_list:
+			# Mark as OK for CMake
+			with open(os.path.join(root, "src", "wrapped", "generated", "functions_list.txt"), 'w') as file:
+				file.write(functions_list)
+			return 0
+	except IOError:
+		# The file does not exist yet, first run
+		pass
 	
 	# Now the files rebuilding part
-	# File headers and guards
-	files_headers = {
+	# Files header and guard
+	files_header = {
 		"wrapper.c": """/*****************************************************************
- * File automatically generated by rebuild_wrappers.py (v%s)
+ * File automatically generated by rebuild_wrappers.py (v{version})
 *****************************************************************/
 #include <stdio.h>
 #include <stdlib.h>
@@ -215,18 +421,23 @@ def main(root, defines, ver):
 #include "regs.h"
 #include "x86emu.h"

-typedef union ui64_s {
+typedef union ui64_s {lbr}
    int64_t     i;
    uint64_t    u;
    uint32_t    d[2];
-} ui64_t;
+{rbr} ui64_t;
+
+typedef struct _2uint_struct_s {lbr}
+	uint32_t	a;
+	uint32_t	b;
+{rbr} _2uint_struct_t;

 extern void* my__IO_2_1_stderr_;
 extern void* my__IO_2_1_stdin_ ;
 extern void* my__IO_2_1_stdout_;

 static void* io_convert(void* v)
-{
+{lbr}
 	if(!v)
 		return v;
 	if(v==my__IO_2_1_stderr_)
@@ -236,25 +447,47 @@ static void* io_convert(void* v)
 	if(v==my__IO_2_1_stdout_)
 		return stdout;
 	return v;
-}
+{rbr}
+
+typedef struct my_GValue_s
+{lbr}
+  int         g_type;
+  union {lbr}
+    int        v_int;
+    int64_t    v_int64;
+    uint64_t   v_uint64;
+    float      v_float;
+    double     v_double;
+    void*      v_pointer;
+  {rbr} data[2];
+{rbr} my_GValue_t;
+
+static void alignGValue(my_GValue_t* v, void* value)
+{lbr}
+    v->g_type = *(int*)value;
+    memcpy(v->data, value+4, 2*sizeof(double));
+{rbr}
+static void unalignGValue(void* value, my_GValue_t* v)
+{lbr}
+    *(int*)value = v->g_type;
+    memcpy(value+4, v->data, 2*sizeof(double));
+{rbr}
+
+void* VulkanFromx86(void* src, void** save);
+void VulkanTox86(void* src, void* save);

-#ifdef USE_FLOAT
-#define ST0val ST0.f
-#else
 #define ST0val ST0.d
-#endif
-#ifndef NOALIGN
-int of_convert(int flag) { return (flag&~0740000) | ((flag&0140000)<<2) | ((flag&0600000)>>2);}
-#else
-int of_convert(int flag) { return flag; }
-#endif
+
+int of_convert(int);
+
 """,
 		"wrapper.h": """/*****************************************************************
- * File automatically generated by rebuild_wrappers.py (v%s)
+ * File automatically generated by rebuild_wrappers.py (v{version})
 *****************************************************************/
 #ifndef __WRAPPER_H_
 #define __WRAPPER_H_
 #include <stdint.h>
+#include <string.h>

 typedef struct x86emu_s x86emu_t;

@@ -274,71 +507,128 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 // O = libc O_ flags bitfield
 // S = _IO_2_1_stdXXX_ pointer (or FILE*)
 // Q = ...
-// S8 = struct, 8 bytes
+// 2 = struct of 2 uint
+// P = Vulkan struture pointer
+// G = a single GValue pointer
+// N = ... automatically sending 1 arg
+// M = ... automatically sending 2 args
+
+""",
+		"fntypes.h": """/*****************************************************************
+ * File automatically generated by rebuild_wrappers.py (v{version})
+ *****************************************************************/
+#ifndef __{filename}TYPES_H_
+#define __{filename}TYPES_H_
+
+#ifndef LIBNAME
+#error You should only #include this file inside a wrapped*.c file
+#endif
+#ifndef ADDED_FUNCTIONS
+#define ADDED_FUNCTIONS() 
+#endif

 """
 	}
-	files_guards = {"wrapper.c": """""",
+	files_guard = {"wrapper.c": """""",
 		"wrapper.h": """
-#endif //__WRAPPER_H_
+#endif // __WRAPPER_H_
+""",
+		"fntypes.h": """
+#endif // __{filename}TYPES_H_
 """
 	}
 	
-	# Transform strings into arrays
-	for k in gbl:
-		gbl[k] = [[c for c in v] for v in gbl[k]]
-	
-	# Rewrite the wrapper.h file:
-	with open(os.path.join(root, "src/wrapped/generated/", "wrapper.h"), 'w') as file:
-		file.write(files_headers["wrapper.h"] % ver)
-		for v in gbl["()"]:
-			file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
-		for k in gbl_idxs:
-			if k != "()":
-				file.write("\n#if " + k + "\n")
-				for v in gbl[k]:
-					file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
-				file.write("#endif\n")
-		file.write("\n")
-		for v in redirects["()"]:
-			file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
-		for k in redirects_idxs:
-			if k != "()":
-				file.write("\n#if " + k + "\n")
-				for v in redirects[k]:
-					file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
-				file.write("#endif\n")
-		file.write(files_guards["wrapper.h"])
+	#           E            e             v       c         w          i          I          C          W           u           U           f        d         D              K         l           L            p        V        O          S        2                  P        G        N      M
+	tdtypes = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "intptr_t", "uintptr_t", "void*", "void*", "int32_t", "void*", "_2uint_struct_t", "void*", "void*", "...", "..."]
+	if len(values) != len(tdtypes):
+		raise NotImplementedError("len(values) = {lenval} != len(tdtypes) = {lentypes}".format(lenval=len(values), lentypes=len(tdtypes)))
+	def generate_typedefs(arr, file):
+		for v in arr:
+			file.write("typedef " + tdtypes[values.index(v[0])] + " (*" + v + "_t)"
+							+ "(" + ', '.join(tdtypes[values.index(t)] for t in v[2:]) + ");\n")
 	
 	# Rewrite the wrapper.c file:
-	with open(os.path.join(root, "src/wrapped/generated/", "wrapper.c"), 'w') as file:
-		file.write(files_headers["wrapper.c"] % ver)
+	with open(os.path.join(root, "src", "wrapped", "generated", "wrapper.c"), 'w') as file:
+		file.write(files_header["wrapper.c"].format(lbr="{", rbr="}", version=ver))
 		
 		# First part: typedefs
-		for v in gbl["()"]:
-			#         E            e             v       c         w          i          I          C          W           u           U           f        d         D              K         l           L            p        V			O        S
-			types = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "intptr_t", "uintptr_t", "void*", "void*", "int32_t", "void*"]
-			if len(values) != len(types):
-					raise NotImplementedError("len(values) = {lenval} != len(types) = {lentypes}".format(lenval=len(values), lentypes=len(types)))
-			
-			file.write("typedef " + types[values.index(v[0])] + " (*" + ''.join(v) + "_t)"
-				+ "(" + ', '.join(types[values.index(t)] for t in v[2:]) + ");\n")
+		generate_typedefs(gbl["()"], file)
 		for k in gbl_idxs:
-			if k != "()":
-				file.write("\n#if " + k + "\n")
-				for v in gbl[k]:
-					#         E            e             v       c         w          i          I          C          W           u           U           f        d         D              K         l           L            p        V			O        S
-					types = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "intptr_t", "uintptr_t", "void*", "void*", "int32_t", "void*"]
-					if len(values) != len(types):
-							raise NotImplementedError("len(values) = {lenval} != len(types) = {lentypes}".format(lenval=len(values), lentypes=len(types)))
-					
-					file.write("typedef " + types[values.index(v[0])] + " (*" + ''.join(v) + "_t)"
-						+ "(" + ', '.join(types[values.index(t)] for t in v[2:]) + ");\n")
-				file.write("#endif\n")
+			file.write("\n#if " + k + "\n")
+			generate_typedefs(gbl[k], file)
+			file.write("#endif\n")
 		
 		file.write("\n")
+		
 		# Next part: function definitions
 		
+		# Helper variables
+		arg = [
+			"emu, ",                                  # E
+			"&emu, ",                                 # e
+			"",                                       # v
+			"*(int8_t*)(R_ESP + {p}), ",              # c
+			"*(int16_t*)(R_ESP + {p}), ",             # w
+			"*(int32_t*)(R_ESP + {p}), ",             # i
+			"*(int64_t*)(R_ESP + {p}), ",             # I
+			"*(uint8_t*)(R_ESP + {p}), ",             # C
+			"*(uint16_t*)(R_ESP + {p}), ",            # W
+			"*(uint32_t*)(R_ESP + {p}), ",            # u
+			"*(uint64_t*)(R_ESP + {p}), ",            # U
+			"*(float*)(R_ESP + {p}), ",               # f
+			"*(double*)(R_ESP + {p}), ",              # d
+			"LD2localLD((void*)(R_ESP + {p})), ",     # D
+			"FromLD((void*)(R_ESP + {p})), ",         # K
+			"*(intptr_t*)(R_ESP + {p}), ",            # l
+			"*(uintptr_t*)(R_ESP + {p}), ",           # L
+			"*(void**)(R_ESP + {p}), ",               # p
+			"(void*)(R_ESP + {p}), ",                 # V
+			"of_convert(*(int32_t*)(R_ESP + {p})), ", # O
+			"io_convert(*(void**)(R_ESP + {p})), ",   # S
+			"(_2uint_struct_t){{*(uintptr_t*)(R_ESP + {p}),*(uintptr_t*)(R_ESP + {p} + 4)}}, ",	# 2
+			"arg{p}, ",                               # P
+			"&arg{p}, ",                              # G
+			"*(void**)(R_ESP + {p}), ",				  # N
+			"*(void**)(R_ESP + {p}),*(void**)(R_ESP + {p} + 4), ",	# M
+		]
+		#         E  e  v  c  w  i  I  C  W  u  U  f  d  D   K   l  L  p  V  O  S  2  P  G  N, M
+		deltas = [0, 0, 4, 4, 4, 4, 8, 4, 4, 4, 8, 4, 8, 12, 12, 4, 4, 4, 0, 4, 4, 8, 4, 4, 0, 0]
+		vals = [
+			"\n#error Invalid return type: emulator\n",                     # E
+			"\n#error Invalid return type: &emulator\n",                    # e
+			"fn({0});",                                                     # v
+			"R_EAX=fn({0});",                                               # c
+			"R_EAX=fn({0});",                                               # w
+			"R_EAX=fn({0});",                                               # i
+			"ui64_t r; r.i=fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];",           # I
+			"R_EAX=(unsigned char)fn({0});",                                # C
+			"R_EAX=(unsigned short)fn({0});",                               # W
+			"R_EAX=(uint32_t)fn({0});",                                     # u
+			"ui64_t r; r.u=(uint64_t)fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];", # U
+			"float fl=fn({0}); fpu_do_push(emu); ST0val = fl;",             # f
+			"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # d
+			"long double ld=fn({0}); fpu_do_push(emu); ST0val = ld;",       # D
+			"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # K
+			"R_EAX=(intptr_t)fn({0});",                                     # l
+			"R_EAX=(uintptr_t)fn({0});",                                    # L
+			"R_EAX=(uintptr_t)fn({0});",                                    # p
+			"\n#error Invalid return type: va_list\n",                      # V
+			"\n#error Invalid return type: at_flags\n",                     # O
+			"\n#error Invalid return type: _io_file*\n",                    # S
+			"\n#error Invalid return type: _2uint_struct\n",                # 2
+			"\n#error Invalid return type: Vulkan Struct\n",                # P
+			"\n#error Invalid return type: GValue Pointer\n",               # G
+			"\n#error Invalid return type: ... with 1 arg\n",               # N
+			"\n#error Invalid return type: ... with 2 args\n",              # M
+		]
+		# Asserts
+		if len(values) != len(arg):
+			raise NotImplementedError("len(values) = {lenval} != len(arg) = {lenarg}".format(lenval=len(values), lenarg=len(arg)))
+		if len(values) != len(deltas):
+			raise NotImplementedError("len(values) = {lenval} != len(deltas) = {lendeltas}".format(lenval=len(values), lendeltas=len(deltas)))
+		if len(values) != len(vals):
+			raise NotImplementedError("len(values) = {lenval} != len(vals) = {lenvals}".format(lenval=len(values), lenvals=len(vals)))
+		
 		# Helper functions to write the function definitions
 		def function_args(args, d=4):
 			if len(args) == 0:
@@ -351,89 +641,99 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 			elif args[0] == "1":
 				return "1, " + function_args(args[1:], d)
 			
-			arg = [
-				"emu, ",                          # E
-				"&emu, ",                         # e
-				"",                               # v
-				"*(int8_t*)(R_ESP + {p}), ",      # c
-				"*(int16_t*)(R_ESP + {p}), ",     # w
-				"*(int32_t*)(R_ESP + {p}), ",     # i
-				"*(int64_t*)(R_ESP + {p}), ",     # I
-				"*(uint8_t*)(R_ESP + {p}), ",     # C
-				"*(uint16_t*)(R_ESP + {p}), ",    # W
-				"*(uint32_t*)(R_ESP + {p}), ",    # u
-				"*(uint64_t*)(R_ESP + {p}), ",    # U
-				"*(float*)(R_ESP + {p}), ",       # f
-				"*(double*)(R_ESP + {p}), ",      # d
-				"*(long double*)(R_ESP + {p}), ", # D
-				"FromLD((void*)(R_ESP + {p})), ", # K
-				"*(intptr_t*)(R_ESP + {p}), ",    # l
-				"*(uintptr_t*)(R_ESP + {p}), ",   # L
-				"*(void**)(R_ESP + {p}), ",       # p
-				"(void*)(R_ESP + {p}), ",         # V
-				"of_convert(*(int32_t*)(R_ESP + {p})), ",         # O
-				"io_convert(*(void**)(R_ESP + {p})), ",         # S
-			]
-			#         E  e  v  c  w  i  I  C  W  u  U  f  d  D   K   l  L  p  V  O  S
-			deltas = [0, 0, 4, 4, 4, 4, 8, 4, 4, 4, 8, 4, 8, 12, 12, 4, 4, 4, 0, 4, 4]
-			if len(values) != len(arg):
-				raise NotImplementedError("len(values) = {lenval} != len(arg) = {lenarg}".format(lenval=len(values), lenarg=len(arg)))
-			if len(values) != len(deltas):
-				raise NotImplementedError("len(values) = {lenval} != len(deltas) = {lendeltas}".format(lenval=len(values), lendeltas=len(deltas)))
 			return arg[values.index(args[0])].format(p=d) + function_args(args[1:], d + deltas[values.index(args[0])])
 		
 		def function_writer(f, N, W, rettype, args):
 			f.write("void {0}(x86emu_t *emu, uintptr_t fcn) {2} {1} fn = ({1})fcn; ".format(N, W, "{"))
-			vals = [
-				"\n#error Invalid return type: emulator\n",                     # E
-				"\n#error Invalid return type: &emulator\n",                    # e
-				"fn({0});",                                                     # v
-				"R_EAX=fn({0});",                                               # c
-				"R_EAX=fn({0});",                                               # w
-				"R_EAX=fn({0});",                                               # i
-				"ui64_t r; r.i=fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];",           # I
-				"R_EAX=(unsigned char)fn({0});",                                # C
-				"R_EAX=(unsigned short)fn({0});",                               # W
-				"R_EAX=(uint32_t)fn({0});",                                     # u
-				"ui64_t r; r.u=(uint64_t)fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];", # U
-				"float fl=fn({0}); fpu_do_push(emu); ST0val = fl;",             # f
-				"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # d
-				"long double ld=fn({0}); fpu_do_push(emu); ST0val = ld;",       # D
-				"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # K
-				"R_EAX=(intptr_t)fn({0});",                                     # l
-				"R_EAX=(uintptr_t)fn({0});",                                    # L
-				"R_EAX=(uintptr_t)fn({0});",                                    # p
-				"\n#error Invalid return type: va_list\n",                      # V
-				"\n#error Invalid return type: at_flags\n",                     # O
-				"\n#error Invalid return type: _io_file*\n",                    # S
-			]
-			if len(values) != len(vals):
-				raise NotImplementedError("len(values) = {lenval} != len(vals) = {lenvals}".format(lenval=len(values), lenvals=len(vals)))
-			f.write(vals[values.index(rettype)].format(function_args(args)[:-2]) + " }\n")
+			if any(cc in 'PG' for cc in args):
+				# Vulkan struct or GValue pointer, need to unwrap functions at the end
+				delta = 4
+				for c in args:
+					if c == 'P':
+						f.write("void* save{d}=NULL; void *arg{d} = VulkanFromx86(*(void**)(R_ESP + {d}), &save{d}); ".format(d=delta))
+					if c == 'G':
+    						f.write("my_GValue_t arg{d}; alignGValue(&arg{d}, *(void**)(R_ESP + {d})); ".format(d=delta))
+					delta = delta + deltas[values.index(c)]
+				f.write(vals[values.index(rettype)].format(function_args(args)[:-2]) + " ")
+				delta = 4
+				for c in args:
+					if c == 'P':
+						f.write("VulkanTox86(arg{d}, save{d}); ".format(d=delta))
+					if c == 'G':
+						f.write("unalignGValue(*(void**)(R_ESP + {d}), &arg{d}); ".format(d=delta))
+					delta = delta + deltas[values.index(c)]
+				f.write("}\n")
+			else:
+				# Generic function
+				f.write(vals[values.index(rettype)].format(function_args(args)[:-2]) + " }\n")
 		
 		for v in gbl["()"]:
-			function_writer(file, ''.join(v), ''.join(v) + "_t", v[0], v[2:])
+			function_writer(file, v, v + "_t", v[0], v[2:])
 		for k in gbl_idxs:
-			if k != "()":
-				file.write("\n#if " + k + "\n")
-				for v in gbl[k]:
-					function_writer(file, ''.join(v), ''.join(v) + "_t", v[0], v[2:])
-				file.write("#endif\n")
+			file.write("\n#if " + k + "\n")
+			for v in gbl[k]:
+				function_writer(file, v, v + "_t", v[0], v[2:])
+			file.write("#endif\n")
 		file.write("\n")
 		for v in redirects["()"]:
-			function_writer(file, v, redirects["()"][v] + "_t", v[0], v[2:])
+			function_writer(file, v[0], v[1] + "_t", v[0][0], v[0][2:])
 		for k in redirects_idxs:
-			if k != "()":
-				file.write("\n#if " + k + "\n")
-				for v in redirects[k]:
-					function_writer(file, v, redirects[k][v] + "_t", v[0], v[2:])
-				file.write("#endif\n")
+			file.write("\n#if " + k + "\n")
+			for v in redirects[k]:
+				function_writer(file, v[0], v[1] + "_t", v[0][0], v[0][2:])
+			file.write("#endif\n")
 		
-		file.write(files_guards["wrapper.c"])
+		file.write(files_guard["wrapper.c"].format(lbr="{", rbr="}", version=ver))
+	
+	# Rewrite the wrapper.h file:
+	with open(os.path.join(root, "src", "wrapped", "generated", "wrapper.h"), 'w') as file:
+		file.write(files_header["wrapper.h"].format(lbr="{", rbr="}", version=ver))
+		for v in gbl["()"]:
+			file.write("void " + v + "(x86emu_t *emu, uintptr_t fnc);\n")
+		for k in gbl_idxs:
+			file.write("\n#if " + k + "\n")
+			for v in gbl[k]:
+				file.write("void " + v + "(x86emu_t *emu, uintptr_t fnc);\n")
+			file.write("#endif\n")
+		file.write("\n")
+		for v in redirects["()"]:
+			file.write("void " + v[0] + "(x86emu_t *emu, uintptr_t fnc);\n")
+		for k in redirects_idxs:
+			file.write("\n#if " + k + "\n")
+			for v in redirects[k]:
+				file.write("void " + v[0] + "(x86emu_t *emu, uintptr_t fnc);\n")
+			file.write("#endif\n")
+		file.write(files_guard["wrapper.h"].format(lbr="{", rbr="}", version=ver))
+	
+	for fn in mytypedefs:
+		with open(os.path.join(root, "src", "wrapped", "generated", fn + "types.h"), 'w') as file:
+			file.write(files_header["fntypes.h"].format(lbr="{", rbr="}", version=ver, filename=fn))
+			if fn in mystructs:
+				values = values + mystructs_vals[fn]
+				tdtypes = tdtypes + [mystructs[fn][k] for k in mystructs_vals[fn]]
+			generate_typedefs(mytypedefs_vals[fn], file)
+			if fn in mystructs:
+				values = values[:-len(mystructs_vals[fn])]
+				tdtypes = tdtypes[:-len(mystructs_vals[fn])]
+			file.write("\n#define SUPER() ADDED_FUNCTIONS()")
+			for v in mytypedefs_vals[fn]:
+				for t, f in mytypedefs[fn][v]:
+					assert(t in [0, 1, 2, 3])
+					file.write(" \\\n\tGO({0}, {1}_t)".format(f, v))
+			file.write("\n")
+			file.write(files_guard["fntypes.h"].format(lbr="{", rbr="}", version=ver, filename=fn))
+	
+	# Save the string for the next iteration, writing was successful
+	with open(os.path.join(root, "src", "wrapped", "generated", "functions_list.txt"), 'w') as file:
+		file.write(functions_list)
 	
 	return 0

 if __name__ == '__main__':
-	if main(sys.argv[1], sys.argv[2:], "1.1.1.06") != 0:
+	limit = []
+	for i, v in enumerate(sys.argv):
+		if v == "--":
+			limit.append(i)
+	if main(sys.argv[1], sys.argv[2:limit[0]], sys.argv[limit[0]+1:], "1.2.0.09") != 0:
 		exit(2)
 	exit(0)
@@ -13,7 +13,15 @@ if( NOT TEST_REFERENCE )
 endif( NOT TEST_REFERENCE )

 set(ENV{BOX86_LOG} 0)
-set(ENV{LD_LIBRARY_PATH} ${CMAKE_SOURCE_DIR}/x86lib)
+set(ENV{BOX86_NOBANNER} 1)
+if( EXISTS ${CMAKE_SOURCE_DIR}/x86lib )
+  # we are inside box86 folder
+  set(ENV{LD_LIBRARY_PATH} ${CMAKE_SOURCE_DIR}/x86lib)
+else()
+  # we are inside build folder
+  set(ENV{LD_LIBRARY_PATH} ${CMAKE_SOURCE_DIR}/../x86lib)
+endif( EXISTS ${CMAKE_SOURCE_DIR}/x86lib )
+
 # run the test program, capture the stdout/stderr and the result var
 execute_process(
  COMMAND ${TEST_PROGRAM} ${TEST_ARGS} ${TEST_ARGS2}
@@ -24,17 +32,23 @@ 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 )

 # now compare the output with the reference
 execute_process(
-  COMMAND ${CMAKE_COMMAND} -E compare_files ${TEST_OUTPUT} ${TEST_REFERENCE}
+  COMMAND "${CMAKE_COMMAND}" -E compare_files "${TEST_OUTPUT}" "${TEST_REFERENCE}"
  RESULT_VARIABLE TEST_RESULT
  )

 # 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 )

@@ -16,49 +16,43 @@
 #include "callback.h"
 #include "wrapper.h"
 #include "myfts.h"
-#ifdef DYNAREC
+#include "threads.h"
+#include "x86trace.h"
+#include "signals.h"
 #include <sys/mman.h>
+#include "custommem.h"
+#ifdef DYNAREC
 #include "dynablock.h"
-
-typedef struct mmaplist_s {
-    void*         block;
-    uintptr_t     offset;           // offset in the block
-} mmaplist_t;
-
-#define MMAPSIZE (4*1024*1024)       // allocate 4Mo sized blocks
-
-uintptr_t AllocDynarecMap(box86context_t *context, int size)
-{
-    // make size 0x10 bytes aligned
-    size = (size+0x0f)&~0x0f;
-    pthread_mutex_lock(&context->mutex_mmap);
-    // look for free space
-    for(int i=0; i<context->mmapsize; ++i) {
-        if(context->mmaplist[i].offset+size < MMAPSIZE) {
-            uintptr_t ret = context->mmaplist[i].offset + (uintptr_t)context->mmaplist[i].block;
-            context->mmaplist[i].offset+=size;
-            pthread_mutex_unlock(&context->mutex_mmap);
-            return ret;
-        }
-    }
-    // no luck, add a new one !
-    int i = context->mmapsize++;    // yeah, usefull post incrementation
-    dynarec_log(LOG_DEBUG, "Ask for DynaRec Block Alloc #%d\n", context->mmapsize);
-    context->mmaplist = (mmaplist_t*)realloc(context->mmaplist, context->mmapsize*sizeof(mmaplist_t));
-    void* p = mmap(NULL, MMAPSIZE, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
-    if(p==MAP_FAILED) {
-        dynarec_log(LOG_INFO, "Cannot create memory map of %d byte for dynarec block #%d\n", MMAPSIZE, i);
-        --context->mmapsize;
-        pthread_mutex_unlock(&context->mutex_mmap);
-        return 0;
-    }
-    context->mmaplist[i].block = p;
-    context->mmaplist[i].offset=size;
-    pthread_mutex_unlock(&context->mutex_mmap);
-    return (uintptr_t)p;
-}
+#include "dynarec/arm_lock_helper.h"
 #endif

+
+
+EXPORTDYN
+void initAllHelpers(box86context_t* context)
+{
+    static int inited = 0;
+    if(inited)
+        return;
+    my_context = context;
+    init_pthread_helper();
+    init_signal_helper(context);
+    inited = 1;
+}
+
+EXPORTDYN
+void finiAllHelpers(box86context_t* context)
+{
+    static int finied = 0;
+    if(finied)
+        return;
+    fini_pthread_helper(context);
+    fini_signal_helper();
+    cleanAlternate();
+    fini_custommem_helper(context);
+    finied = 1;
+}
+
 void x86Syscall(x86emu_t *emu);

 /// maxval not inclusive
@@ -66,11 +60,11 @@ int getrand(int maxval)
 {
    if(maxval<1024) {
        return ((random()&0x7fff)*maxval)/0x7fff;
-    } else {
+    } 
        uint64_t r = random();
        r = (r*maxval) / RAND_MAX;
        return r;
-    }
+
 }

 void free_tlsdatasize(void* p)
@@ -82,14 +76,99 @@ void free_tlsdatasize(void* p)
    free(p);
 }

+int unlockMutex()
+{
+    int ret = unlockCustommemMutex();
+    int i;
+    #define GO(A, B)                    \
+        i = checkUnlockMutex(&A);       \
+        if(i) {                         \
+            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)
+    #else
+    GO(my_context->mutex_lock, 8)
+    #endif
+    GO(my_context->mutex_tls, 9)
+    GO(my_context->mutex_thread, 10)
+    #undef GO
+
+    return ret;
+}
+
+void relockMutex(int locks)
+{
+    relockCustommemMutex(locks);
+    #define GO(A, B)                    \
+        if(locks&(1<<B))                \
+            pthread_mutex_lock(&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)
+    #else
+    GO(my_context->mutex_lock, 8)
+    #endif
+    GO(my_context->mutex_tls, 9)
+    GO(my_context->mutex_thread, 10)
+    #undef GO
+}
+
+static void init_mutexes(box86context_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);
+#else
+    pthread_mutex_init(&context->mutex_dyndump, &attr);
+#endif
+    pthread_mutex_init(&context->mutex_tls, &attr);
+    pthread_mutex_init(&context->mutex_thread, &attr);
+
+    pthread_mutexattr_destroy(&attr);
+}
+
+static void atfork_child_box64context(void)
+{
+    // (re)init mutex if it was lock before the fork
+    init_mutexes(my_context);
+}
+
+EXPORTDYN
 box86context_t *NewBox86Context(int argc)
 {
+#ifdef BUILD_DYNAMIC
+    if(my_context) {
+        ++my_context->count;
+        return my_context;
+    }
+#endif
    // init and put default values
-    box86context_t *context = (box86context_t*)calloc(1, sizeof(box86context_t));
+    box86context_t *context = my_context = (box86context_t*)calloc(1, sizeof(box86context_t));

+#ifdef BUILD_LIB
+    context->deferedInit = 0;
+#else
    context->deferedInit = 1;
+#endif
+    context->sel_serial = 1;

-    context->maplib = NewLibrarian(context);
+    init_custommem_helper(context);
+
+    context->maplib = NewLibrarian(context, 1);
+    context->local_maplib = NewLibrarian(context, 1);
    context->system = NewBridge();
    // create vsyscall
    context->vsyscall = AddBridge(context->system, vFv, x86Syscall, 0);
@@ -100,132 +179,138 @@ box86context_t *NewBox86Context(int argc)
 #endif
    context->dlprivate = NewDLPrivate();

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

-    pthread_mutex_init(&context->mutex_once, NULL);
-    pthread_mutex_init(&context->mutex_once2, NULL);
-    pthread_mutex_init(&context->mutex_trace, NULL);
-    pthread_mutex_init(&context->mutex_lock, NULL);
+    init_mutexes(context);
+    pthread_atfork(NULL, NULL, atfork_child_box64context);

    pthread_key_create(&context->tlskey, free_tlsdatasize);

-#ifdef DYNAREC
-    pthread_mutex_init(&context->mutex_blocks, NULL);
-    pthread_mutex_init(&context->mutex_mmap, NULL);
-    context->dynablocks = NewDynablockList(0, 0, 0);
-#endif
    InitFTSMap(context);

    for (int i=0; i<4; ++i) context->canary[i] = 1 +  getrand(255);
-    context->canary[getrand(4)] = 0;
+    context->canary[/*getrand(4)*/0] = 0;
    printf_log(LOG_DEBUG, "Setting up canary (for Stack protector) at GS:0x14, value:%08X\n", *(uint32_t*)context->canary);

+    initAllHelpers(context);
+
    return context;
 }

+EXPORTDYN
 void FreeBox86Context(box86context_t** context)
 {
+#ifdef BUILD_DYNAMIC
+    if(--(*context)->count)
+        return;
+#endif
    if(!context)
        return;
    
    if(--(*context)->forked >= 0)
        return;

-    FreeFTSMap(*context);
+    box86context_t* ctx = *context;   // local copy to do the cleanning

-    if((*context)->maplib)
-        FreeLibrarian(&(*context)->maplib);
+    if(ctx->local_maplib)
+        FreeLibrarian(&ctx->local_maplib, NULL);
+    if(ctx->maplib)
+        FreeLibrarian(&ctx->maplib, NULL);

-#ifdef DYNAREC
-    dynarec_log(LOG_INFO, "Free global Dynarecblocks\n");
-    if((*context)->dynablocks)
-        FreeDynablockList(&(*context)->dynablocks);
-    for (int i=0; i<(*context)->mmapsize; ++i)
-        if((*context)->mmaplist[i].block)
-            munmap((*context)->mmaplist[i].block, MMAPSIZE);
-    free((*context)->mmaplist);
-    pthread_mutex_destroy(&(*context)->mutex_blocks);
-    pthread_mutex_destroy(&(*context)->mutex_mmap);
-#endif
-    
-    if((*context)->emu)
-        FreeX86Emu(&(*context)->emu);
-
-    FreeCollection(&(*context)->box86_path);
-    FreeCollection(&(*context)->box86_ld_lib);
-    FreeCollection(&(*context)->box86_emulated_libs);
-    if((*context)->zydis)
-        DeleteX86Trace(*context);
-
-    if((*context)->deferedInitList)
-        free((*context)->deferedInitList);
-
-#ifndef BUILD_LIB
-    if((*context)->box86lib)
-        dlclose((*context)->box86lib);
-#endif
-
-    FreeDLPrivate(&(*context)->dlprivate);
-
-    for(int i=0; i<(*context)->argc; ++i)
-        free((*context)->argv[i]);
-    free((*context)->argv);
-    
-    for (int i=0; i<(*context)->envc; ++i)
-        free((*context)->envv[i]);
-    free((*context)->envv);
-
-    for(int i=0; i<(*context)->elfsize; ++i) {
-        FreeElfHeader(&(*context)->elfs[i]);
+    for(int i=0; i<ctx->elfsize; ++i) {
+        FreeElfHeader(&ctx->elfs[i]);
    }
-    free((*context)->elfs);
+    free(ctx->elfs);

-    free((*context)->stack);
+    FreeFTSMap(ctx);

-    free((*context)->fullpath);
-    free((*context)->box86path);
+    FreeCollection(&ctx->box86_path);
+    FreeCollection(&ctx->box86_ld_lib);
+    FreeCollection(&ctx->box86_emulated_libs);
+    // stop trace now
+    if(ctx->dec)
+        DeleteX86TraceDecoder(&ctx->dec);
+    if(ctx->zydis)
+        DeleteX86Trace(ctx);

-    FreeBridge(&(*context)->system);
+    if(ctx->deferedInitList)
+        free(ctx->deferedInitList);

-    if((*context)->glwrappers)
-        freeProcWrapper(&(*context)->glwrappers);
-    if((*context)->alwrappers)
-        freeProcWrapper(&(*context)->alwrappers);
+    free(ctx->argv);
+    
+    for (int i=0; i<ctx->envc; ++i)
+        free(ctx->envv[i]);
+    free(ctx->envv);

-    FreeCallbackList(&(*context)->callbacks);
-
-    void* ptr;
-    if ((ptr = pthread_getspecific((*context)->tlskey)) != NULL) {
-        free_tlsdatasize(ptr);
-        pthread_setspecific((*context)->tlskey, NULL);
-    }
-    pthread_key_delete((*context)->tlskey);
-
-    if((*context)->tlsdata)
-        free((*context)->tlsdata);
-
-    pthread_mutex_destroy(&(*context)->mutex_once);
-    pthread_mutex_destroy(&(*context)->mutex_once2);
-    pthread_mutex_destroy(&(*context)->mutex_trace);
-    pthread_mutex_destroy(&(*context)->mutex_lock);
-
-    if((*context)->atfork_sz) {
-        free((*context)->atforks);
-        (*context)->atforks = NULL;
-        (*context)->atfork_sz = (*context)->atfork_cap = 0;
+    if(ctx->atfork_sz) {
+        free(ctx->atforks);
+        ctx->atforks = NULL;
+        ctx->atfork_sz = ctx->atfork_cap = 0;
    }

    for(int i=0; i<MAX_SIGNAL; ++i)
-        if((*context)->signals[i]!=0 && (*context)->signals[i]!=1) {
+        if(ctx->signals[i]!=0 && ctx->signals[i]!=1) {
            signal(i, SIG_DFL);
        }

+    *context = NULL;                // bye bye my_context

-    free(*context);
-    *context = NULL;
+    CleanStackSize(ctx);
+
+#ifndef BUILD_LIB
+    if(ctx->box86lib)
+        dlclose(ctx->box86lib);
+#endif
+
+    FreeDLPrivate(&ctx->dlprivate);
+
+//    free(ctx->stack); // don't free the stack, it's owned by main emu!
+
+    free(ctx->fullpath);
+    free(ctx->box86path);
+
+    FreeBridge(&ctx->system);
+
+    freeGLProcWrapper(ctx);
+    freeALProcWrapper(ctx);
+
+    void* ptr;
+    if ((ptr = pthread_getspecific(ctx->tlskey)) != NULL) {
+        free_tlsdatasize(ptr);
+        pthread_setspecific(ctx->tlskey, NULL);
+    }
+    pthread_key_delete(ctx->tlskey);
+
+    if(ctx->tlsdata)
+        free(ctx->tlsdata);
+
+    for(int i=0; i<3; ++i) {
+        if(ctx->segtls[i].present) {
+            // key content not handled by box86, so not doing anything with it
+            pthread_key_delete(ctx->segtls[i].key);
+        }
+    }
+
+    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);
+#else
+    pthread_mutex_destroy(&ctx->mutex_dyndump);
+#endif
+    pthread_mutex_destroy(&ctx->mutex_tls);
+    pthread_mutex_destroy(&ctx->mutex_thread);
+
+    free_neededlib(&ctx->neededlibs);
+
+    if(ctx->emu_sig)
+        FreeX86Emu(&ctx->emu_sig);
+
+    finiAllHelpers(ctx);
+
+    free(ctx);
 }

 int AddElfHeader(box86context_t* ctx, elfheader_t* head) {
@@ -242,7 +327,17 @@ int AddElfHeader(box86context_t* ctx, elfheader_t* head) {
 }

 int AddTLSPartition(box86context_t* context, int tlssize) {
+    int oldsize = context->tlssize;
    context->tlssize += tlssize;
    context->tlsdata = 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
+    if(my_context) {
+        ResetSegmentsCache(thread_get_emu());
+        if(!(++context->sel_serial))
+            ++context->sel_serial;
+    }
+
    return -context->tlssize;   // negative offset
-}
+}
@@ -2,7 +2,7 @@
 #define __BOX86_VERSION_H_

 #define BOX86_MAJOR 0
-#define BOX86_MINOR 1
-#define BOX86_REVISION 0
+#define BOX86_MINOR 2
+#define BOX86_REVISION 2

 #endif //__BOX86_VERSION_H_
@@ -0,0 +1,27 @@
+#include <stdio.h>
+#include "debug.h"
+#include "box86version.h"
+#ifdef NOGIT
+#define GITREV "nogit"
+#else
+#include "git_head.h"
+#endif
+
+void PrintBox86Version()
+{
+    printf("Box86%s%s v%d.%d.%d %s built on %s %s\n", 
+    #ifdef HAVE_TRACE
+        " with trace",
+    #else
+        "",
+    #endif
+    #ifdef DYNAREC
+        " with Dynarec",
+    #else
+        "",
+    #endif
+        BOX86_MAJOR, BOX86_MINOR, BOX86_REVISION,
+        GITREV,
+        __DATE__, __TIME__);
+}
+
@@ -0,0 +1,6 @@
+#ifndef __BUILD_INFO_H__
+#define __BUILD_INFO_H__
+
+void PrintBox86Version();
+
+#endif //__BUILD_INFO_H__
@@ -0,0 +1,774 @@
+#define _GNU_SOURCE         /* See feature_test_macros(7) */
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <dlfcn.h>
+#include <signal.h>
+
+#include "box86context.h"
+#include "elfloader.h"
+#include "debug.h"
+#include "x86trace.h"
+#include "x86emu.h"
+#include "librarian.h"
+#include "bridge.h"
+#include "library.h"
+#include "callback.h"
+#include "wrapper.h"
+#include "myfts.h"
+#include "threads.h"
+#include "x86trace.h"
+#include "signals.h"
+#include <sys/mman.h>
+#include "custommem.h"
+#include "threads.h"
+#ifdef DYNAREC
+#include "dynablock.h"
+#include "dynarec/arm_lock_helper.h"
+#include "khash.h"
+
+#define USE_MMAP
+
+// init inside dynablocks.c
+KHASH_MAP_INIT_INT(dynablocks, dynablock_t*)
+static dynablocklist_t*    dynmap[DYNAMAP_SIZE];     // 4G of memory mapped by 4K block
+static pthread_mutex_t     mutex_mmap;
+static mmaplist_t          *mmaplist;
+static int                 mmapsize;
+static kh_dynablocks_t     *dblist_oversized;      // store the list of oversized dynablocks (normal sized are inside mmaplist)
+static uintptr_t           *box86_jumptable[JMPTABL_SIZE];
+static uintptr_t           box86_jmptbl_default[1<<JMPTABL_SHIFT];
+#endif
+#define MEMPROT_SHIFT 12
+#define MEMPROT_SIZE (1<<(32-MEMPROT_SHIFT))
+static uint8_t             memprot[MEMPROT_SIZE] = {0};    // protection flags by 4K block
+static int inited = 0;
+
+typedef struct blocklist_s {
+    void*               block;
+    int                 maxfree;
+    size_t              size;
+} blocklist_t;
+
+#define MMAPSIZE (256*1024)      // allocate 256kb sized blocks
+
+static pthread_mutex_t     mutex_blocks;
+static int                 n_blocks = 0;       // number of blocks for custom malloc
+static blocklist_t*        p_blocks = NULL;    // actual blocks for custom malloc
+
+typedef union mark_s {
+    struct {
+        unsigned int    fill:1;
+        unsigned int    size:31;
+    };
+    uint32_t            x32;
+} mark_t;
+typedef struct blockmark_s {
+    mark_t  prev;
+    mark_t  next;
+} blockmark_t;
+
+
+// get first subblock free in block. Return NULL if no block, else first subblock free (mark included), filling size
+static void* getFirstBlock(void* block, int maxsize, int* size)
+{
+    // get start of block
+    blockmark_t *m = (blockmark_t*)block;
+    while(m->next.x32) {    // while there is a subblock
+        if(!m->next.fill && m->next.size>=maxsize+sizeof(blockmark_t)) {
+            *size = m->next.size-sizeof(blockmark_t);
+            return m;
+        }
+        m = (blockmark_t*)((uintptr_t)m + m->next.size);
+    }
+
+    return NULL;
+}
+
+static int getMaxFreeBlock(void* block, size_t block_size)
+{
+    // get start of block
+    blockmark_t *m = (blockmark_t*)((uintptr_t)block+block_size-sizeof(blockmark_t)); // start with the end
+    int maxsize = 0;
+    while(m->prev.x32) {    // while there is a subblock
+        if(!m->prev.fill && m->prev.size>maxsize) {
+            maxsize = m->prev.size;
+            if((uintptr_t)block+maxsize>(uintptr_t)m)
+                return (maxsize>=sizeof(blockmark_t))?(maxsize-sizeof(blockmark_t)):0; // no block large enough left...
+        }
+        m = (blockmark_t*)((uintptr_t)m - m->prev.size);
+    }
+    return (maxsize>=sizeof(blockmark_t))?(maxsize-sizeof(blockmark_t)):0;
+}
+
+static void* allocBlock(void* block, void *sub, int size)
+{
+    blockmark_t *s = (blockmark_t*)sub;
+    blockmark_t *n = (blockmark_t*)((uintptr_t)s + s->next.size);
+
+    s->next.fill = 1;
+    s->next.size = size+sizeof(blockmark_t);
+    blockmark_t *m = (blockmark_t*)((uintptr_t)s + s->next.size);   // this is new n
+    m->prev.fill = 1;
+    m->prev.size = s->next.size;
+    if(n!=m) {
+        // new mark
+        m->prev.fill = 1;
+        m->prev.size = s->next.size;
+        m->next.fill = 0;
+        m->next.size = (uintptr_t)n - (uintptr_t)m;
+        n->prev.fill = 0;
+        n->prev.size = m->next.size;
+    }
+
+    return (void*)((uintptr_t)sub + sizeof(blockmark_t));
+}
+static void freeBlock(void *block, void* sub)
+{
+    blockmark_t *m = (blockmark_t*)block;
+    blockmark_t *s = (blockmark_t*)sub;
+    blockmark_t *n = (blockmark_t*)((uintptr_t)s + s->next.size);
+    if(block!=sub)
+        m = (blockmark_t*)((uintptr_t)s - s->prev.size);
+    s->next.fill = 0;
+    n->prev.fill = 0;
+    // check if merge with previous
+    if (s->prev.x32 && !s->prev.fill) {
+        // remove s...
+        m->next.size += s->next.size;
+        n->prev.size = m->next.size;
+        s = m;
+    }
+    // check if merge with next
+    if(n->next.x32 && !n->next.fill) {
+        blockmark_t *n2 = (blockmark_t*)((uintptr_t)n + n->next.size);
+        //remove n
+        s->next.size += n->next.size;
+        n2->prev.size = s->next.size;
+    }
+}
+// return 1 if block has been expanded to new size, 0 if not
+static int expandBlock(void* block, void* sub, int newsize)
+{
+    newsize = (newsize+3)&~3;
+    blockmark_t *s = (blockmark_t*)sub;
+    blockmark_t *n = (blockmark_t*)((uintptr_t)s + s->next.size);
+    if(s->next.fill)
+        return 0;   // next block is filled
+    if(s->next.size + n->next.size < newsize)
+        return 0;   // free space too short
+    // ok, doing the alloc!
+    s->next.size = newsize+sizeof(blockmark_t);
+    blockmark_t *m = (blockmark_t*)((uintptr_t)s + s->next.size);   // this is new n
+    m->prev.fill = 1;
+    m->prev.size = s->next.size;
+    if(n!=m) {
+        // new mark
+        m->prev.fill = 1;
+        m->prev.size = s->next.size;
+        m->next.fill = 0;
+        m->next.size = (uintptr_t)n - (uintptr_t)m;
+        n->prev.fill = 0;
+        n->prev.size = m->next.size;
+    }
+    return 1;
+}
+// return size of block
+static int sizeBlock(void* sub)
+{
+    blockmark_t *s = (blockmark_t*)sub;
+    return s->next.size;
+}
+
+void* customMalloc(size_t size)
+{
+    // look for free space
+    void* sub = NULL;
+    pthread_mutex_lock(&mutex_blocks);
+    for(int i=0; i<n_blocks; ++i) {
+        if(p_blocks[i].maxfree>=size) {
+            int rsize = 0;
+            sub = getFirstBlock(p_blocks[i].block, size, &rsize);
+            if(sub) {
+                void* ret = allocBlock(p_blocks[i].block, sub, size);
+                if(rsize==p_blocks[i].maxfree)
+                    p_blocks[i].maxfree = getMaxFreeBlock(p_blocks[i].block, p_blocks[i].size);
+                pthread_mutex_unlock(&mutex_blocks);
+                return ret;
+            }
+        }
+    }
+    // add a new block
+    int i = n_blocks++;
+    p_blocks = (blocklist_t*)realloc(p_blocks, n_blocks*sizeof(blocklist_t));
+    size_t allocsize = MMAPSIZE;
+    if(size+2*sizeof(blockmark_t)>allocsize)
+        allocsize = size+2*sizeof(blockmark_t);
+    #ifdef USE_MMAP
+    void* p = mmap(NULL, allocsize, PROT_READ|PROT_WRITE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+    memset(p, 0, allocsize);
+    #else
+    void* p = calloc(1, allocsize);
+    #endif
+    p_blocks[i].block = p;
+    p_blocks[i].size = allocsize;
+    // setup marks
+    blockmark_t* m = (blockmark_t*)p;
+    m->prev.x32 = 0;
+    m->next.fill = 0;
+    m->next.size = allocsize-sizeof(blockmark_t);
+    m = (blockmark_t*)(p+allocsize-sizeof(blockmark_t));
+    m->next.x32 = 0;
+    m->prev.fill = 0;
+    m->prev.size = allocsize-sizeof(blockmark_t);
+    // alloc 1st block
+    void* ret  = allocBlock(p_blocks[i].block, p, size);
+    p_blocks[i].maxfree = getMaxFreeBlock(p_blocks[i].block, p_blocks[i].size);
+    pthread_mutex_unlock(&mutex_blocks);
+    return ret;
+}
+void* customCalloc(size_t n, size_t size)
+{
+    size_t newsize = n*size;
+    void* ret = customMalloc(newsize);
+    memset(ret, 0, newsize);
+    return ret;
+}
+void* customRealloc(void* p, size_t size)
+{
+    if(!p)
+        return customMalloc(size);
+    uintptr_t addr = (uintptr_t)p;
+    pthread_mutex_lock(&mutex_blocks);
+    for(int i=0; i<n_blocks; ++i) {
+        if ((addr>(uintptr_t)p_blocks[i].block) 
+         && (addr<((uintptr_t)p_blocks[i].block+p_blocks[i].size))) {
+            void* sub = (void*)(addr-sizeof(blockmark_t));
+            if(expandBlock(p_blocks[i].block, sub, size)) {
+                p_blocks[i].maxfree = getMaxFreeBlock(p_blocks[i].block, p_blocks[i].size);
+                pthread_mutex_unlock(&mutex_blocks);
+                return p;
+            }
+                pthread_mutex_unlock(&mutex_blocks);
+                void* newp = customMalloc(size);
+                memcpy(newp, p, sizeBlock(sub));
+                customFree(p);
+                return newp;
+            
+        }
+    }
+    pthread_mutex_unlock(&mutex_blocks);
+    if(n_blocks)
+        dynarec_log(LOG_NONE, "Warning, block %p not found in p_blocks for Realloc, Malloc'ng again without free\n", (void*)addr);
+    return customMalloc(size);
+}
+void customFree(void* p)
+{
+    if(!p)
+        return;
+    uintptr_t addr = (uintptr_t)p;
+    pthread_mutex_lock(&mutex_blocks);
+    for(int i=0; i<n_blocks; ++i) {
+        if ((addr>(uintptr_t)p_blocks[i].block) 
+         && (addr<((uintptr_t)p_blocks[i].block+p_blocks[i].size))) {
+            void* sub = (void*)(addr-sizeof(blockmark_t));
+            freeBlock(p_blocks[i].block, sub);
+            p_blocks[i].maxfree = getMaxFreeBlock(p_blocks[i].block, p_blocks[i].size);
+            pthread_mutex_unlock(&mutex_blocks);
+            return;
+        }
+    }
+    pthread_mutex_unlock(&mutex_blocks);
+    if(n_blocks)
+        dynarec_log(LOG_NONE, "Warning, block %p not found in p_blocks for Free\n", (void*)addr);
+}
+
+#ifdef DYNAREC
+typedef struct mmaplist_s {
+    void*               block;
+    int                 maxfree;
+    size_t              size;
+    kh_dynablocks_t*    dblist;
+    uint8_t*            helper;
+} mmaplist_t;
+
+uintptr_t FindFreeDynarecMap(dynablock_t* db, int size)
+{
+    // look for free space
+    void* sub = NULL;
+    for(int i=0; i<mmapsize; ++i) {
+        if(mmaplist[i].maxfree>=size) {
+            int rsize = 0;
+            sub = getFirstBlock(mmaplist[i].block, size, &rsize);
+            if(sub) {
+                uintptr_t ret = (uintptr_t)allocBlock(mmaplist[i].block, sub, size);
+                if(rsize==mmaplist[i].maxfree)
+                    mmaplist[i].maxfree = getMaxFreeBlock(mmaplist[i].block, mmaplist[i].size);
+                kh_dynablocks_t *blocks = mmaplist[i].dblist;
+                if(!blocks) {
+                    blocks = mmaplist[i].dblist = kh_init(dynablocks);
+                    kh_resize(dynablocks, blocks, 64);
+                }
+                khint_t k;
+                int r;
+                k = kh_put(dynablocks, blocks, (uintptr_t)ret, &r);
+                kh_value(blocks, k) = db;
+                for(int j=0; j<size; ++j)
+                    mmaplist[i].helper[(uintptr_t)ret-(uintptr_t)mmaplist[i].block+j] = (j<256)?j:255;
+                return ret;
+            }
+        }
+    }
+    return 0;
+}
+
+uintptr_t AddNewDynarecMap(dynablock_t* db, int size)
+{
+    int i = mmapsize++;    // yeah, useful post incrementation
+    dynarec_log(LOG_DEBUG, "Ask for DynaRec Block Alloc #%d\n", mmapsize);
+    mmaplist = (mmaplist_t*)realloc(mmaplist, mmapsize*sizeof(mmaplist_t));
+    #ifndef USE_MMAP
+    void *p = NULL;
+    if(posix_memalign(&p, box86_pagesize, MMAPSIZE)) {
+        dynarec_log(LOG_INFO, "Cannot create memory map of %d byte for dynarec block #%d\n", MMAPSIZE, i);
+        --mmapsize;
+        return 0;
+    }
+    mprotect(p, MMAPSIZE, PROT_READ | PROT_WRITE | PROT_EXEC);
+    #else
+    void* p = mmap(NULL, MMAPSIZE, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
+    if(p==(void*)-1) {
+        dynarec_log(LOG_INFO, "Cannot create memory map of %d byte for dynarec block #%d\n", MMAPSIZE, i);
+        --mmapsize;
+        return 0;
+    }
+    #endif
+    setProtection((uintptr_t)p, MMAPSIZE, PROT_READ | PROT_WRITE | PROT_EXEC);
+
+    mmaplist[i].block = p;
+    mmaplist[i].size = MMAPSIZE;
+    mmaplist[i].helper = (uint8_t*)calloc(1, MMAPSIZE);
+    // setup marks
+    blockmark_t* m = (blockmark_t*)p;
+    m->prev.x32 = 0;
+    m->next.fill = 0;
+    m->next.size = MMAPSIZE-sizeof(blockmark_t);
+    m = (blockmark_t*)(p+MMAPSIZE-sizeof(blockmark_t));
+    m->next.x32 = 0;
+    m->prev.fill = 0;
+    m->prev.size = MMAPSIZE-sizeof(blockmark_t);
+    // alloc 1st block
+    uintptr_t sub  = (uintptr_t)allocBlock(mmaplist[i].block, p, size);
+    mmaplist[i].maxfree = getMaxFreeBlock(mmaplist[i].block, mmaplist[i].size);
+    kh_dynablocks_t *blocks = mmaplist[i].dblist = kh_init(dynablocks);
+    kh_resize(dynablocks, blocks, 64);
+    khint_t k;
+    int ret;
+    k = kh_put(dynablocks, blocks, (uintptr_t)sub, &ret);
+    kh_value(blocks, k) = db;
+    for(int j=0; j<size; ++j)
+        mmaplist[i].helper[(uintptr_t)sub-(uintptr_t)mmaplist[i].block + j] = (j<256)?j:255;
+    return sub;
+}
+
+void ActuallyFreeDynarecMap(dynablock_t* db, uintptr_t addr, int size)
+{
+    if(!addr || !size)
+        return;
+    for(int i=0; i<mmapsize; ++i) {
+        if ((addr>(uintptr_t)mmaplist[i].block) 
+         && (addr<((uintptr_t)mmaplist[i].block+mmaplist[i].size))) {
+            void* sub = (void*)(addr-sizeof(blockmark_t));
+            freeBlock(mmaplist[i].block, sub);
+            mmaplist[i].maxfree = getMaxFreeBlock(mmaplist[i].block, mmaplist[i].size);
+            kh_dynablocks_t *blocks = mmaplist[i].dblist;
+            if(blocks) {
+                khint_t k = kh_get(dynablocks, blocks, (uintptr_t)sub);
+                if(k!=kh_end(blocks))
+                    kh_del(dynablocks, blocks, k);
+                for(int j=0; j<size; ++j)
+                    mmaplist[i].helper[(uintptr_t)sub-(uintptr_t)mmaplist[i].block+j] = 0;
+            }
+            return;
+        }
+    }
+    if(mmapsize)
+        dynarec_log(LOG_NONE, "Warning, block %p (size %d) not found in mmaplist for Free\n", (void*)addr, size);
+}
+
+dynablock_t* FindDynablockFromNativeAddress(void* addr)
+{
+    // look in actual list
+    for(int i=0; i<mmapsize; ++i) {
+        if ((uintptr_t)addr>=(uintptr_t)mmaplist[i].block 
+        && ((uintptr_t)addr<(uintptr_t)mmaplist[i].block+mmaplist[i].size)) {
+            if(!mmaplist[i].helper)
+                return FindDynablockDynablocklist(addr, mmaplist[i].dblist);
+            else {
+                uintptr_t p = (uintptr_t)addr - (uintptr_t)mmaplist[i].block;
+                while(mmaplist[i].helper[p]) p -= mmaplist[i].helper[p];
+                khint_t k = kh_get(dynablocks, mmaplist[i].dblist, (uintptr_t)mmaplist[i].block + p);
+                if(k!=kh_end(mmaplist[i].dblist))
+                    return kh_value(mmaplist[i].dblist, k);
+                return NULL;
+            }
+        }
+    }
+    // look in oversized
+    return FindDynablockDynablocklist(addr, dblist_oversized);
+}
+
+uintptr_t AllocDynarecMap(dynablock_t* db, int size)
+{
+    if(!size)
+        return 0;
+    if(size>MMAPSIZE-2*sizeof(blockmark_t)) {
+        #ifndef USE_MMAP
+        pthread_mutex_lock(&mutex_mmap);
+        void *p = NULL;
+        if(posix_memalign(&p, box86_pagesize, size)) {
+            dynarec_log(LOG_INFO, "Cannot create dynamic map of %d bytes\n", size);
+            return 0;
+        }
+        mprotect(p, size, PROT_READ | PROT_WRITE | PROT_EXEC);
+        #else
+        void* p = mmap(NULL, size, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+        if(p==(void*)-1) {
+            dynarec_log(LOG_INFO, "Cannot create dynamic map of %d bytes\n", size);
+            return 0;
+        }
+        #endif
+        setProtection((uintptr_t)p, size, PROT_READ | PROT_WRITE | PROT_EXEC);
+        kh_dynablocks_t *blocks = dblist_oversized;
+        if(!blocks) {
+            blocks = dblist_oversized = kh_init(dynablocks);
+            kh_resize(dynablocks, blocks, 64);
+        }
+        khint_t k;
+        int ret;
+        k = kh_put(dynablocks, blocks, (uintptr_t)p, &ret);
+        kh_value(blocks, k) = db;
+        pthread_mutex_unlock(&mutex_mmap);
+        return (uintptr_t)p;
+    }
+    
+    pthread_mutex_lock(&mutex_mmap);
+
+    uintptr_t ret = FindFreeDynarecMap(db, size);
+    if(!ret)
+        ret = AddNewDynarecMap(db, size);
+
+    pthread_mutex_unlock(&mutex_mmap);
+
+    return ret;
+}
+
+void FreeDynarecMap(dynablock_t* db, uintptr_t addr, uint32_t size)
+{
+    if(!addr || !size)
+        return;
+    if(size>MMAPSIZE-2*sizeof(blockmark_t)) {
+        pthread_mutex_lock(&mutex_mmap);
+        #ifndef USE_MMAP
+        free((void*)addr);
+        #else
+        munmap((void*)addr, size);
+        #endif
+        kh_dynablocks_t *blocks = dblist_oversized;
+        if(blocks) {
+            khint_t k = kh_get(dynablocks, blocks, addr);
+            if(k!=kh_end(blocks))
+                kh_del(dynablocks, blocks, k);
+        }
+        pthread_mutex_unlock(&mutex_mmap);
+        return;
+    }
+    pthread_mutex_lock(&mutex_mmap);
+    ActuallyFreeDynarecMap(db, addr, size);
+    pthread_mutex_unlock(&mutex_mmap);
+}
+
+dynablocklist_t* getDB(uintptr_t idx)
+{
+    return dynmap[idx];
+}
+
+// each dynmap is 64k of size
+
+void addDBFromAddressRange(uintptr_t addr, uintptr_t size)
+{
+    dynarec_log(LOG_DEBUG, "addDBFromAddressRange %p -> %p\n", (void*)addr, (void*)(addr+size-1));
+    uintptr_t idx = (addr>>DYNAMAP_SHIFT);
+    uintptr_t end = ((addr+size-1)>>DYNAMAP_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        if(!dynmap[i]) {
+            dynablocklist_t* p = NewDynablockList(i<<DYNAMAP_SHIFT, 1<<DYNAMAP_SHIFT, 0);
+            if(arm_lock_storeifnull(&dynmap[i], p)!=p)
+                FreeDynablockList(&p);
+        }
+    }
+}
+
+void cleanDBFromAddressRange(uintptr_t addr, uintptr_t size, int destroy)
+{
+    dynarec_log(LOG_DEBUG, "cleanDBFromAddressRange %p -> %p %s\n", (void*)addr, (void*)(addr+size-1), destroy?"destroy":"mark");
+    uintptr_t idx = (addr>>DYNAMAP_SHIFT);
+    uintptr_t end = ((addr+size-1)>>DYNAMAP_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        dynablocklist_t* dblist = dynmap[i];
+        if(dblist) {
+            if(destroy)
+                FreeRangeDynablock(dblist, addr, size);
+            else
+                MarkRangeDynablock(dblist, addr, size);
+        }
+    }
+}
+
+#ifdef ARM
+void arm_next(void);
+#endif
+
+void addJumpTableIfDefault(void* addr, void* jmp)
+{
+    const uintptr_t idx = ((uintptr_t)addr>>JMPTABL_SHIFT);
+    if(box86_jumptable[idx] == box86_jmptbl_default) {
+        uintptr_t* tbl = (uintptr_t*)malloc((1<<JMPTABL_SHIFT)*sizeof(uintptr_t));
+        for(int i=0; i<(1<<JMPTABL_SHIFT); ++i)
+            tbl[i] = (uintptr_t)arm_next;
+        if(arm_lock_storeifref(&box86_jumptable[idx], tbl, box86_jmptbl_default)!=tbl)
+            free(tbl);
+    }
+    const uintptr_t off = (uintptr_t)addr&((1<<JMPTABL_SHIFT)-1);
+    arm_lock_storeifref(&box86_jumptable[idx][off], jmp, arm_next);
+}
+void setJumpTableDefault(void* addr)
+{
+    const uintptr_t idx = ((uintptr_t)addr>>JMPTABL_SHIFT);
+    if(box86_jumptable[idx] == box86_jmptbl_default) {
+        return;
+    }
+    const uintptr_t off = (uintptr_t)addr&((1<<JMPTABL_SHIFT)-1);
+    box86_jumptable[idx][off] = (uintptr_t)arm_next;
+}
+int isJumpTableDefault(void* addr)
+{
+    const uintptr_t idx = ((uintptr_t)addr>>JMPTABL_SHIFT);
+    if(box86_jumptable[idx] == box86_jmptbl_default) {
+        return 1;
+    }
+    const uintptr_t off = (uintptr_t)addr&((1<<JMPTABL_SHIFT)-1);
+    return (box86_jumptable[idx][off]==(uintptr_t)arm_next)?1:0;
+}
+uintptr_t getJumpTable()
+{
+    return (uintptr_t)box86_jumptable;
+}
+
+uintptr_t getJumpTableAddress(uintptr_t addr)
+{
+    const uintptr_t idx = ((uintptr_t)addr>>JMPTABL_SHIFT);
+    if(box86_jumptable[idx] == box86_jmptbl_default) {
+        uintptr_t* tbl = (uintptr_t*)malloc((1<<JMPTABL_SHIFT)*sizeof(uintptr_t));
+        for(int i=0; i<(1<<JMPTABL_SHIFT); ++i)
+            tbl[i] = (uintptr_t)arm_next;
+        if(arm_lock_storeifref(&box86_jumptable[idx], tbl, box86_jmptbl_default)!=tbl)
+            free(tbl);
+    }
+    const uintptr_t off = (uintptr_t)addr&((1<<JMPTABL_SHIFT)-1);
+    return (uintptr_t)&box86_jumptable[idx][off];
+}
+
+// Remove the Write flag from an adress range, so DB can be executed
+// no log, as it can be executed inside a signal handler
+void protectDB(uintptr_t addr, uintptr_t size)
+{
+    dynarec_log(LOG_DEBUG, "protectDB %p -> %p\n", (void*)addr, (void*)(addr+size-1));
+    uintptr_t idx = (addr>>MEMPROT_SHIFT);
+    uintptr_t end = ((addr+size-1)>>MEMPROT_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        uint32_t prot = memprot[i];
+        if(!prot)
+            prot = PROT_READ | PROT_WRITE;    // comes from malloc & co, so should not be able to execute
+        memprot[i] = prot|PROT_DYNAREC;
+        if(!(prot&PROT_DYNAREC))
+            mprotect((void*)(i<<MEMPROT_SHIFT), 1<<MEMPROT_SHIFT, prot&~PROT_WRITE);
+    }
+}
+
+// Add the Write flag from an adress range, and mark all block as dirty
+// no log, as it can be executed inside a signal handler
+void unprotectDB(uintptr_t addr, uintptr_t size)
+{
+    dynarec_log(LOG_DEBUG, "unprotectDB %p -> %p\n", (void*)addr, (void*)(addr+size-1));
+    uintptr_t idx = (addr>>MEMPROT_SHIFT);
+    uintptr_t end = ((addr+size-1)>>MEMPROT_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        uint32_t prot = memprot[i];
+        memprot[i] = prot&~PROT_DYNAREC;
+        if(prot&PROT_DYNAREC) {
+            mprotect((void*)(i<<MEMPROT_SHIFT), 1<<MEMPROT_SHIFT, prot&~PROT_DYNAREC);
+            cleanDBFromAddressRange((i<<MEMPROT_SHIFT), 1<<MEMPROT_SHIFT, 0);
+        }
+    }
+}
+
+#endif
+
+void updateProtection(uintptr_t addr, uintptr_t size, uint32_t prot)
+{
+    const uintptr_t idx = (addr>>MEMPROT_SHIFT);
+    const uintptr_t end = ((addr+size-1)>>MEMPROT_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        uint32_t dyn=(memprot[i]&PROT_DYNAREC);
+        if(dyn && (prot&PROT_WRITE))    // need to remove the write protection from this block
+            mprotect((void*)(i<<MEMPROT_SHIFT), 1<<MEMPROT_SHIFT, prot&~PROT_WRITE);
+        memprot[i] = prot|dyn;
+    }
+}
+
+void setProtection(uintptr_t addr, uintptr_t size, uint32_t prot)
+{
+    const uintptr_t idx = (addr>>MEMPROT_SHIFT);
+    const uintptr_t end = ((addr+size-1)>>MEMPROT_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        memprot[i] = prot;
+    }
+}
+
+uint32_t getProtection(uintptr_t addr)
+{
+    const uintptr_t idx = (addr>>MEMPROT_SHIFT);
+    uint32_t ret = memprot[idx];
+    return ret;
+}
+
+int unlockCustommemMutex()
+{
+    int ret = 0;
+    int i = 0;
+    #define GO(A, B)                    \
+        i = checkUnlockMutex(&A);       \
+        if(i) {                         \
+            ret|=(1<<B);                \
+        }
+    GO(mutex_blocks, 0)
+    #ifdef DYNAREC
+    GO(mutex_mmap, 1)
+    #endif
+    #undef GO
+    return ret;
+}
+
+void relockCustommemMutex(int locks)
+{
+    #define GO(A, B)                    \
+        if(locks&(1<<B))                \
+            pthread_mutex_lock(&A);     \
+
+    GO(mutex_blocks, 0)
+    #ifdef DYNAREC
+    GO(mutex_mmap, 1)
+    #endif
+    #undef GO
+}
+
+static void init_mutexes(void)
+{
+    pthread_mutexattr_t attr;
+    pthread_mutexattr_init(&attr);
+    pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_ERRORCHECK);
+    pthread_mutex_init(&mutex_blocks, &attr);
+#ifdef DYNAREC
+    pthread_mutex_init(&mutex_mmap, &attr);
+#endif
+
+    pthread_mutexattr_destroy(&attr);
+}
+
+static void atfork_child_custommem(void)
+{
+    // (re))init mutex if it was lock before the fork
+    init_mutexes();
+}
+
+void init_custommem_helper(box86context_t* ctx)
+{
+    if(inited) // already initialized
+        return;
+    inited = 1;
+    init_mutexes();
+#ifdef DYNAREC
+#ifdef ARM
+    for(int i=0; i<(1<<JMPTABL_SHIFT); ++i)
+        box86_jmptbl_default[i] = (uintptr_t)arm_next;
+    for(int i=0; i<JMPTABL_SIZE; ++i)
+        box86_jumptable[i] = box86_jmptbl_default;
+#else
+#error Unsupported architecture!
+#endif
+#endif
+    pthread_atfork(NULL, NULL, atfork_child_custommem);
+}
+
+void fini_custommem_helper(box86context_t *ctx)
+{
+    if(!inited)
+        return;
+    inited = 0;
+#ifdef DYNAREC
+    dynarec_log(LOG_DEBUG, "Free global Dynarecblocks\n");
+    for (int i=0; i<mmapsize; ++i) {
+        if(mmaplist[i].block)
+            #ifdef USE_MMAP
+            munmap(mmaplist[i].block, mmaplist[i].size);
+            #else
+            free(mmaplist[i].block);
+            #endif
+        if(mmaplist[i].dblist) {
+            kh_destroy(dynablocks, mmaplist[i].dblist);
+            mmaplist[i].dblist = NULL;
+        }
+        if(mmaplist[i].helper) {
+            free(mmaplist[i].helper);
+            mmaplist[i].helper = NULL;
+        }
+    }
+    if(dblist_oversized) {
+        kh_destroy(dynablocks, dblist_oversized);
+        dblist_oversized = NULL;
+    }
+    mmapsize = 0;
+    dynarec_log(LOG_DEBUG, "Free dynamic Dynarecblocks\n");
+    uintptr_t idx = 0;
+    uintptr_t end = ((0xFFFFFFFF)>>DYNAMAP_SHIFT);
+    for (uintptr_t i=idx; i<=end; ++i) {
+        dynablocklist_t* dblist = dynmap[i];
+        if(dblist) {
+            uintptr_t startdb = StartDynablockList(dblist);
+            uintptr_t enddb = EndDynablockList(dblist);
+            uintptr_t startaddr = 0;
+            if(startaddr<startdb) startaddr = startdb;
+            uintptr_t endaddr = 0xFFFFFFFF;
+            if(endaddr>enddb) endaddr = enddb;
+            FreeRangeDynablock(dblist, startaddr, endaddr-startaddr+1);
+        }
+    }
+    for (uintptr_t i=idx; i<=end; ++i)
+        if(dynmap[i])
+            FreeDynablockList(&dynmap[i]);
+    pthread_mutex_destroy(&mutex_mmap);
+    free(mmaplist);
+    for (int i=0; i<DYNAMAP_SIZE; ++i)
+        if(box86_jumptable[i]!=box86_jmptbl_default)
+            free(box86_jumptable[i]);
+#endif
+    for(int i=0; i<n_blocks; ++i)
+        #ifdef USE_MMAP
+        munmap(p_blocks[i].block, p_blocks[i].size);
+        #else
+        free(p_blocks[i].block);
+        #endif
+    free(p_blocks);
+    pthread_mutex_destroy(&mutex_blocks);
+}
@@ -21,28 +21,29 @@ Op is 20-27
 #define xEBP    9
 #define xESI    10
 #define xEDI    11
-#define xEIP    12
+#define xFlags  12
+#define xEIP    14
 // scratch registers
 #define x1      1
 #define x2      2
 #define x3      3
-#define x12     12
+#define x14     14
 // emu is r0
 #define xEmu    0
 // ARM SP is r13
 #define xSP     13      

 // barel roll operations (4 possibles)
-#define brLSL(i, r) (0<<4 | 0<<5 | ((i&31)<<7) | r)
-#define brLSR(i, r) (0<<4 | 1<<5 | ((i&31)<<7) | r)
-#define brASR(i, r) (0<<4 | 2<<5 | ((i&31)<<7) | r)
-#define brROR(i, r) (0<<4 | 3<<5 | ((i&31)<<7) | r)
+#define brLSL(i, r) (0<<4 | 0<<5 | (((i)&31)<<7) | (r))
+#define brLSR(i, r) (0<<4 | 1<<5 | (((i)&31)<<7) | (r))
+#define brASR(i, r) (0<<4 | 2<<5 | (((i)&31)<<7) | (r))
+#define brROR(i, r) (0<<4 | 3<<5 | (((i)&31)<<7) | (r))
 #define brIMM(r)    (abs(r))
 // barel roll with a register
-#define brRLSL(i, r) (1<<4 | 0<<5 | ((i&15)<<8) | r)
-#define brRLSR(i, r) (1<<4 | 1<<5 | ((i&15)<<8) | r)
-#define brRASR(i, r) (1<<4 | 2<<5 | ((i&15)<<8) | r)
-#define brRROR(i, r) (1<<4 | 3<<5 | ((i&15)<<8) | r)
+#define brRLSL(i, r) (1<<4 | 0<<5 | (((i)&15)<<8) | (r))
+#define brRLSR(i, r) (1<<4 | 1<<5 | (((i)&15)<<8) | (r))
+#define brRASR(i, r) (1<<4 | 2<<5 | (((i)&15)<<8) | (r))
+#define brRROR(i, r) (1<<4 | 3<<5 | (((i)&15)<<8) | (r))


 // conditions
@@ -77,14 +78,18 @@ Op is 20-27
 // movt dst, #imm16
 #define MOVT(dst, imm16) EMIT(0xe3400000 | ((dst) << 12) | (((imm16) & 0xf000) << 4) | brIMM((imm16) & 0x0fff) )
 // pseudo insruction: mov reg, #imm with imm a 32bits value
-#define MOV32(dst, imm32)                   \
-    MOVW(dst, ((uint32_t)imm32)&0xffff);    \
-    if (((uint32_t)imm32)>>16) {            \
-        MOVT(dst, (((uint32_t)imm32)>>16)); }
+#define MOV32(dst, imm32)                       \
+    if((~(uint32_t)(imm32))<0x100) {            \
+        MVN_IMM8(dst, ~(uint32_t)(imm32), 0);   \
+    } else {                                    \
+        MOVW(dst, ((uint32_t)imm32)&0xffff);    \
+        if (((uint32_t)(imm32))>>16) {          \
+        MOVT(dst, (((uint32_t)imm32)>>16));}    \
+    }
 // pseudo insruction: mov reg, #imm with imm a 32bits value, fixed size
-#define MOV32_(dst, imm32)                   \
-    MOVW(dst, ((uint32_t)imm32)&0xffff);    \
-    MOVT(dst, (((uint32_t)imm32)>>16))
+#define MOV32_(dst, imm32)                  \
+    MOVW(dst, ((uint32_t)(imm32))&0xffff);  \
+    MOVT(dst, (((uint32_t)(imm32))>>16))
 // movw.cond dst, #imm16
 #define MOVW_COND(cond, dst, imm16) EMIT(cond | 0x03000000 | ((dst) << 12) | (((imm16) & 0xf000) << 4) | brIMM((imm16) & 0x0fff) )
 // mov dst #imm8 ror imm4*2
@@ -93,46 +98,74 @@ Op is 20-27
 #define MOV_IMM_COND(cond, dst, imm8, rimm4) EMIT(cond | 0x03a00000 | ((dst) << 12) | (imm8) | ((rimm4) << 8) )

 // mov dst, src lsl imm5
-#define MOV_REG_LSL_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (0<<5) | (imm5<<7))
+#define MOV_REG_LSL_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (0<<5) | ((imm5)<<7))
 // mov dst, src lsr imm5
-#define MOV_REG_LSR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (1<<5) | (imm5<<7))
+#define MOV_REG_LSR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (1<<5) | ((imm5)<<7))
+// mov.cond dst, src lsr imm5
+#define MOV_REG_LSR_IMM5_COND(cond, dst, src, imm5) EMIT(cond | 0x01a00000 | ((dst) << 12) | (src) | (0<<4) | (1<<5) | ((imm5)<<7))
 // mov dst, src asr imm5
-#define MOV_REG_ASR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (2<<5) | (imm5<<7))
+#define MOV_REG_ASR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (2<<5) | ((imm5)<<7))
 // mov dst, src ror imm5
-#define MOV_REG_ROR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (3<<5) | (imm5<<7))
+#define MOV_REG_ROR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (3<<5) | ((imm5)<<7))
+// mov.s dst, src lsl imm5
+#define MOVS_REG_LSL_IMM5(dst, src, imm5) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (0<<4) | (0<<5) | ((imm5)<<7))
+// mov.s dst, src lsr imm5
+#define MOVS_REG_LSR_IMM5(dst, src, imm5) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (0<<4) | (1<<5) | ((imm5)<<7))
+// mov.s dst, src asr imm5
+#define MOVS_REG_ASR_IMM5(dst, src, imm5) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (0<<4) | (2<<5) | ((imm5)<<7))
+// mov.s dst, src ror imm5
+#define MOVS_REG_ROR_IMM5(dst, src, imm5) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (0<<4) | (3<<5) | ((imm5)<<7))

+// mov.s dst, src lsl rs
+#define MOVS_REG_LSL_REG(dst, src, rs) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (1<<4) | (0<<5) | ((rs)<<8))
+// mov.s dst, src lsr rs
+#define MOVS_REG_LSR_REG(dst, src, rs) EMIT(0xe1b00000 | ((dst) << 12) | (src) | (1<<4) | (1<<5) | ((rs)<<8))
 // mov dst, src lsl rs
-#define MOV_REG_LSL_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (0<<5) | (rs<<8))
+#define MOV_REG_LSL_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (0<<5) | ((rs)<<8))
+// mov.cond dst, src lsl rs
+#define MOV_REG_LSL_REG_COND(cond, dst, src, rs) EMIT(cond | 0x01a00000 | ((dst) << 12) | (src) | (1<<4) | (0<<5) | ((rs)<<8))
 // mov dst, src lsr rs
-#define MOV_REG_LSR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (1<<5) | (rs<<8))
+#define MOV_REG_LSR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (1<<5) | ((rs)<<8))
 // mov dst, src asr rs
-#define MOV_REG_ASR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (2<<5) | (rs<<8))
+#define MOV_REG_ASR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (2<<5) | ((rs)<<8))
 // mov dst, src ror rs
-#define MOV_REG_ROR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (3<<5) | (rs<<8))
+#define MOV_REG_ROR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (3<<5) | ((rs)<<8))

 // sub dst, src, #(imm8)
 #define SUB_IMM8(dst, src, imm8) \
    EMIT(0xe2400000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
 // sub cond dst, src, #(imm8)
-#define SUB_COND_IMM8(dst, src, imm8) \
+#define SUB_COND_IMM8(cond, dst, src, imm8) \
    EMIT((cond) | 0x02400000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
 // sub.s dst, src, #(imm8)
 #define SUBS_IMM8(dst, src, imm8) \
    EMIT(0xe2500000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
-// add dst, src1, src2, lsl #imm
-#define SUB_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe0400000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
-// add.s dst, src1, src2, lsl #imm
-#define SUBS_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe0500000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// sub dst, src1, src2, lsl #imm
+#define SUB_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0400000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// sub.s dst, src1, src2, lsl #imm
+#define SUBS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0500000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )

 // sbc dst, src, #(imm8)
 #define SBC_IMM8(dst, src, imm8) \
    EMIT(0xe2c00000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// sbc.s dst, src, #(imm8)
+#define SBCS_IMM8(dst, src, imm8) \
+    EMIT(0xe2d00000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// sbc dst, src1, src2, lsl #imm
+#define SBC_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0c00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// sbc.s dst, src1, src2, lsl #imm
+#define SBCS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0d00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )

 // rsb dst, src, #(imm8)
 #define RSB_IMM8(dst, src, imm8) \
    EMIT(0xe2600000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// rsb.s dst, src, #(imm8)
+#define RSBS_IMM8(dst, src, imm8) \
+    EMIT(0xe2700000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
 // rsb cond dst, src, #(imm8)
 #define RSB_COND_IMM8(cond, dst, src, imm8) \
    EMIT((cond) | 0x02600000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
@@ -152,30 +185,56 @@ Op is 20-27
 // and.s dst, src, #(imm8)
 #define ANDS_IMM8(dst, src, imm8) \
    EMIT(0xe2100000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// and.s dst, src1, #imm ror rot*2
+#define ANDS_IMM8_ROR(src, imm8, rot) \
+    EMIT(0xe2100000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | brIMM(imm8) )
 // add dst, src, #(imm8)
 #define ADD_IMM8(dst, src, imm8) \
    EMIT(0xe2800000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add dst, src, #imm8 ror rot*2
+#define ADD_IMM8_ROR(dst, src, imm8, rot) \
+    EMIT(0xe2800000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | brIMM((imm8)) )
 // add.s dst, src, #(imm8)
 #define ADDS_IMM8(dst, src, imm8) \
    EMIT(0xe2900000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add.c dst, src, #(imm8)
+#define ADD_IMM8_cond(cond, dst, src, imm8) \
+    EMIT((cond) | 0x02800000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
 // add dst, src1, src2, lsl #imm
 #define ADD_REG_LSL_IMM5(dst, src1, src2, imm5) \
    EMIT(0xe0800000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // add.s dst, src1, src2, lsl #imm
-#define ADDS_REG_LSL_IMM5(dst, src1, src2, imm8) \
+#define ADDS_REG_LSL_IMM5(dst, src1, src2, imm5) \
    EMIT(0xe0900000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // add dst, src1, src2, lsr #imm
 #define ADD_REG_LSR_IMM5(dst, src1, src2, imm5) \
    EMIT(0xe0800000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm5, src2) )
+#define ADC_IMM8(dst, src, imm8) \
+    EMIT(0xe2a00000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add.s dst, src, #(imm8)
+#define ADCS_IMM8(dst, src, imm8) \
+    EMIT(0xe2b00000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add dst, src1, src2, lsl #imm
+#define ADC_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0a00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// add.s dst, src1, src2, lsl #imm
+#define ADCS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0b00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // cmp.s dst, src1, src2, lsl #imm
 #define CMPS_REG_LSL_IMM5(src1, src2, imm5) \
    EMIT(0xe1500000 | ((0) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// cmp.cond.s dst, src1, src2, lsl #imm
+#define CMPS_REG_LSL_IMM5_COND(cond, src1, src2, imm5) \
+    EMIT((cond) | 0x01500000 | ((0) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// cmp.s dst, src, #imm
+#define CMPS_IMM8_COND(cond, src, imm8) \
+    EMIT((cond) | 0x03500000 | ((0) << 12) | ((src) << 16) | brIMM(imm8) )
 // cmp.s dst, src, #imm
 #define CMPS_IMM8(src, imm8) \
    EMIT(0xe3500000 | ((0) << 12) | ((src) << 16) | brIMM(imm8) )
 // tst.s dst, src1, src2, lsl #imm
-#define TSTS_REG_LSL_IMM8(src1, src2, imm8) \
-    EMIT(0xe1100000 | ((0) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+#define TSTS_REG_LSL_IMM5(src1, src2, imm5) \
+    EMIT(0xe1100000 | ((0) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // tst.s dst, src1, #imm
 #define TSTS_IMM8(src, imm8) \
    EMIT(0xe3100000 | ((0) << 12) | ((src) << 16) | brIMM(imm8) )
@@ -183,26 +242,45 @@ Op is 20-27
 #define TSTS_IMM8_ROR(src, imm8, rot) \
    EMIT(0xe3100000 | ((0) << 12) | ((src) << 16) | ((rot)<<8) | brIMM(imm8) )
 // orr dst, src1, src2, lsl #imm
-#define ORR_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+#define ORR_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // orr.s dst, src1, src2, lsl #imm
-#define ORRS_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe1900000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+#define ORRS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe1900000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// orr dst, src1, src2, lsr #imm
+#define ORR_REG_LSR_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm5, src2) )
+// orr.s dst, src1, src2, lsr #imm
+#define ORRS_REG_LSR_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe1900000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm5, src2) )
 // orr dst, src1, #imm8
 #define ORR_IMM8(dst, src, imm8, rot) \
    EMIT(0xe3800000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
-// orr dst, src1, src2, lsl #rs
+// orr.s dst, src1, #imm8
+#define ORRS_IMM8(dst, src, imm8, rot) \
+    EMIT(0xe3900000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// orr.cond dst, src1, #imm8 ror rot*2
+#define ORR_IMM8_COND(cond, dst, src, imm8, rot) \
+    EMIT((cond) | 0x03800000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// orr dst, src1, src2, lsl rs
 #define ORR_REG_LSL_REG(dst, src1, src2, rs) \
    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brRLSL(rs, src2) )
 // xor dst, src1, src2, lsl #imm
-#define XOR_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+#define XOR_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // xor.s dst, src1, src2, lsl #imm
-#define XORS_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe0300000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+#define XORS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0300000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+#define XOR_REG_LSL_IMM5_COND(cond, dst, src1, src2, imm5) \
+    EMIT((cond) | 0x00200000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
 // xor dst, src, #(imm8)
 #define XOR_IMM8(dst, src, imm8) \
    EMIT(0xe2200000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+#define XORS_IMM8(dst, src, imm8) \
+    EMIT(0xe2300000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// xor dst, src, #(imm8 ror rot*2)
+#define XOR_IMM8_ROR(dst, src, imm8, rot) \
+    EMIT(0xe2200000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | brIMM(imm8) )
 // xor.cond dst, src, #(imm8)
 #define XOR_IMM8_COND(cond, dst, src, imm8) \
    EMIT(cond | 0x02200000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
@@ -212,21 +290,30 @@ Op is 20-27
 // xor dst, src1, src2, lsl #imm
 #define XOR_REG_LSR_IMM8(dst, src1, src2, imm8) \
    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm8, src2) )
+// xor.cond dst, src1, src2, lsl #imm
+#define XOR_REG_LSR_IMM8_COND(cond, dst, src1, src2, imm8) \
+    EMIT(cond | 0x00200000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm8, src2) )
 // bic dst, src1, src2, lsl #imm
-#define BIC_REG_LSL_IMM8(dst, src1, src2, imm8) \
-    EMIT(0xe1c00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
-// bic dst, src, IMM8
+#define BIC_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe1c00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// bic dst, src, IMM8 ror rot*2
 #define BIC_IMM8(dst, src, imm8, rot) \
    EMIT(0xe3c00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// bic.cond dst, src, IMM8
+#define BIC_IMM8_COND(cond, dst, src, imm8, rot) \
+    EMIT((cond) | 0x03c00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// bic.s dst, src1, #imm ror rot*2
+#define BICS_IMM8_ROR(dst, src, imm8, rot) \
+    EMIT(0xe3d00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | brIMM(imm8) )
 // bic dst, src1, src2, lsl #imm
 #define BIC_REG_LSL_REG(dst, src1, src2, rs) \
    EMIT(0xe1c00000 | ((dst) << 12) | ((src1) << 16) | brRLSL(rs, src2) )
 // mvn dst, src1, src2, lsl #imm
-#define MVN_REG_LSL_IMM8(dst, rm, imm8) \
-    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm8, rm) )
+#define MVN_REG_LSL_IMM5(dst, rm, imm5) \
+    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm5, rm) )
 // mvn dst, src, IMM8
-#define MVN_IMM8(dst, src, imm8, rot) \
-    EMIT(0xe3e00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+#define MVN_IMM8(dst, imm8, rot) \
+    EMIT(0xe3e00000 | ((dst) << 12) | (rot)<<8 | ((imm8)&0xff) )
 // mvn dst, src1, src2, lsr #imm
 #define MVN_REG_LSR_IMM8(dst, rm, imm8) \
    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | brLSR(imm8, rm) )
@@ -234,8 +321,8 @@ Op is 20-27
 #define MVN_REG_LSR_REG(dst, rm, rs) \
    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | (1<<4) | (1<<5) | (rs<<8) | (rm) )
 // mvn with condition dst, src1, src2, lsl #imm
-#define MVN_COND_REG_LSL_IMM8(cond, dst, rm, imm8) \
-    EMIT(cond | 0x01e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm8, rm) )
+#define MVN_COND_REG_LSL_IMM5(cond, dst, rm, imm5) \
+    EMIT(cond | 0x01e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm5, rm) )

 // Single data transfert construction
 #define SDT_REG(Cond, P, U, B, W, L, Rn, Rd, ShiftRm) (Cond | (0b00<<26) | (1<<25) | (P<<24) | (U<<23) | (B<<22) | (U<<23) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | ShiftRm)
@@ -264,9 +351,15 @@ Op is 20-27
 #define LDRBAI_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe6d00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
 // ldrd reg, reg+1, [addr, #+/-imm9], reg must be even, reg+1 is implicit
 #define LDRD_IMM8(reg, addr, imm8) EMIT(c__ | 0b000<<25 | 1<<24  | (((imm8)<0)?0:1)<<23 | 1<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | ((abs(imm8))&0xf0)<<(8-4) | (0b1101<<4) | ((abs(imm8))&0x0f) )
+// ldrd reg, reg+1, [addr, +rm], reg must be even, reg+1 is implicit
+#define LDRD_REG(reg, addr, rm) EMIT(c__ | 0b000<<25 | 1<<24  | 1<<23 | 0<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | 0b1101<<4 | (rm) )

 // str reg, [addr, #+/-imm9]
 #define STR_IMM9(reg, addr, imm9) EMIT(0xe5000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// str with cond reg, [addr, #+/-imm9]
+#define STR_IMM9_COND(cond, reg, addr, imm9) EMIT(cond | 0x05000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// str reg, [addr, #+/-imm9]!
+#define STR_IMM9_W(reg, addr, imm9) EMIT(0xe5200000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
 // strb reg, [addr, #+/-imm9]
 #define STRB_IMM9(reg, addr, imm9) EMIT(0xe5400000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
 // str reg, [addr], #+/-imm9
@@ -283,6 +376,8 @@ Op is 20-27
 #define STRBAI_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe6c00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
 // strd reg, reg+1, [addr, #+/-imm8], reg must be even, reg+1 is implicit
 #define STRD_IMM8(reg, addr, imm8) EMIT(c__ | 0b000<<25 | 1<<24 | (((imm8)<0)?0:1)<<23 | 1<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | ((abs(imm8))&0xf0)<<(8-4) | (0b1111<<4) | ((abs(imm8))&0x0f) )
+// strd reg, reg+1, [addr, +rm], reg must be even, reg+1 is implicit
+#define STRD_REG(reg, addr, rm) EMIT(c__ | 0b000<<25 | 1<<24 | 1<<23 | 0<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | 0b1111<<4 | (rm) )

 // bx reg
 #define BX(reg) EMIT(0xe12fff10 | (reg) )
@@ -293,33 +388,39 @@ Op is 20-27
 #define BLX(reg) EMIT(0xe12fff30 | (reg) )

 // b cond offset
-#define Bcond(C, O) EMIT(C | (0b101<<25) | (0<<24) | ((O)>>2)&0xffffff)
+#define Bcond(C, O) EMIT(C | (0b101<<25) | (0<<24) | (((O)>>2)&0xffffff))

 // bl cond offset
-#define BLcond(C, O) EMIT(C | (0b101<<25) | (1<<24) | ((O)>>2)&0xffffff)
+#define BLcond(C, O) EMIT(C | (0b101<<25) | (1<<24) | (((O)>>2)&0xffffff))

 // push reg!, {list}
 //                           all |    const    |pre-index| subs    | no PSR  |writeback| store   |   base    |reg list
-#define PUSH(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (1<<21) | (0<<20) | (reg<<16) | (list))
+#define PUSH(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (1<<21) | (0<<20) | ((reg)<<16) | (list))
+
+// push reg to xESP
+#define PUSH1(reg)   STR_IMM9_W(reg, xESP, -4)

 // pop reg!, {list}
 //                           all |    const    |postindex|  add    | no PSR  |writeback|  load   |   base    |reg list
-#define POP(reg, list)  EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (1<<21) | (1<<20) | (reg<<16) | (list))
+#define POP(reg, list)  EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (1<<21) | (1<<20) | ((reg)<<16) | (list))
+
+// pop reg from xESP
+#define POP1(reg)   LDRAI_IMM9_W(reg, xESP, 4)

 // STMDB reg, {list}
 //                            all |    const    |pre-index| subs    | no PSR  |  no wb  | store   |   base    |reg list
-#define STMDB(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (0<<21) | (0<<20) | (reg<<16) | (list))
+#define STMDB(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (0<<21) | (0<<20) | ((reg)<<16) | (list))
 // STMia reg, {list}
 //                          all |    const    |postindex|   add   | no PSR  |  no wb  |  store  |   base    |reg list
-#define STM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (0<<20) | (reg<<16) | (list))
+#define STM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (0<<20) | ((reg)<<16) | (list))
 // LDMia reg, {list}
 //                          all |    const    |postindex|   add   | no PSR  |  no wb  |  load   |   base    |reg list
-#define LDM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (1<<20) | (reg<<16) | (list))
+#define LDM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (1<<20) | ((reg)<<16) | (list))


 // Half Word and signed data transfert construction
-#define HWS_REG(Cond, P, U, W, L, Rn, Rd, S, H, Rm)     (Cond | (0b000<<25) | (P<<24) | (U<<23) | (0<<22) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | (1<<7) | (S<<6) | (H<<5) | (1<<4) | Rm)
-#define HWS_OFF(Cond, P, U, W, L, Rn, Rd, S, H, Imm8)   (Cond | (0b000<<25) | (P<<24) | (U<<23) | (1<<22) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | ((Imm8&0xf0)<<(8-4)) | (1<<7) | (S<<6) | (H<<5) | (1<<4) | (Imm8&0x0f))
+#define HWS_REG(Cond, P, U, W, L, Rn, Rd, S, H, Rm)     (Cond | 0b000<<25 | (P)<<24 | (U)<<23 | 0<<22 | (W)<<21 | (L)<<20 | (Rn)<<16 | (Rd)<<12 | 1<<7 | (S)<<6 | (H)<<5 | 1<<4 | (Rm))
+#define HWS_OFF(Cond, P, U, W, L, Rn, Rd, S, H, Imm8)   (Cond | 0b000<<25 | (P)<<24 | (U)<<23 | 1<<22 | (W)<<21 | (L)<<20 | (Rn)<<16 | (Rd)<<12 | ((Imm8)&0xf0)<<(8-4) | 1<<7 | (S)<<6 | (H)<<5 | 1<<4 | ((Imm8)&0x0f))

 #define LDRSB_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 1, addr, reg, 1, 0, abs(imm8)))
 #define LDRSH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 1, addr, reg, 1, 1, abs(imm8)))
@@ -328,17 +429,26 @@ Op is 20-27
 #define STRH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 0, addr, reg, 0, 1, abs(imm8)))
 #define STRSH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 0, addr, reg, 1, 1, abs(imm8)))

-#define LDRHAI_REG_LSL_IMM5(reg, addr, rm) EMIT(HWS_REG(c__, 0, 1, 0, 1, addr, reg, 0, 1, rm))
-#define STRHAI_REG_LSL_IMM5(reg, addr, rm) EMIT(HWS_REG(c__, 0, 1, 0, 0, addr, reg, 0, 1, rm))
+#define LDRH_REG(reg, addr, rm) EMIT(HWS_REG(c__, 1, 1, 0, 1, addr, reg, 0, 1, rm))
+#define STRH_REG(reg, addr, rm) EMIT(HWS_REG(c__, 1, 1, 0, 0, addr, reg, 0, 1, rm))
+
+#define LDRHAI_REG_LSL_IMM5(reg, addr, rm)  EMIT(HWS_REG(c__, 0, 1, 0, 1, addr, reg, 0, 1, rm))
+#define STRHAI_REG_LSL_IMM5(reg, addr, rm)  EMIT(HWS_REG(c__, 0, 1, 0, 0, addr, reg, 0, 1, rm))
+#define LDRHB_IMM8(reg, addr, imm8)         EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 1, 1, addr, reg, 0, 1, abs(imm8)))
+#define LDRHA_IMM8(reg, addr, imm8)         EMIT(HWS_OFF(c__, 0, (((imm8)<0)?0:1), 0, 1, addr, reg, 0, 1, abs(imm8)))
+#define STRHB_IMM8(reg, addr, imm8)         EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 1, 0, addr, reg, 0, 1, abs(imm8)))
+#define STRHA_IMM8(reg, addr, imm8)         EMIT(HWS_OFF(c__, 0, (((imm8)<0)?0:1), 0, 0, addr, reg, 0, 1, abs(imm8)))
+#define PUSH16(reg, addr)                   EMIT(HWS_OFF(c__, 1, 0, 1, 0, addr, reg, 0, 1, 2))
+#define POP16(reg, addr)                    EMIT(HWS_OFF(c__, 0, 1, 0, 1, addr, reg, 0, 1, 2))

 // Mul Long construction
-#define MULLONG(Cond, U, A, S, RdHi, RdLo, Rs, Rm)     (Cond | (0b00001<<23) | (U<<22) | (A<<21) | (S<<20) | (RdHi<<16) | (RdLo<<12) | (Rs<<8) | (0b1001<<4) | (Rm))
+#define MULLONG(Cond, U, A, S, RdHi, RdLo, Rs, Rm)     (Cond | 0b00001<<23 | (U)<<22 | (A)<<21 | (S)<<20 | (RdHi)<<16 | (RdLo)<<12 | (Rs)<<8 | 0b1001<<4 | (Rm))

 #define UMULL(RdHi, RdLo, Rs, Rm)   EMIT(MULLONG(c__, 0, 0, 0, RdHi, RdLo, Rs, Rm))
 #define SMULL(RdHi, RdLo, Rs, Rm)   EMIT(MULLONG(c__, 1, 0, 0, RdHi, RdLo, Rs, Rm))

 // Mul and MulA
-#define MULMULA(Cond, A, S, Rd, Rn, Rs, Rm)     (Cond | (0b000000<<22) | (A<<21) | (S<<20) | (Rd<<16) | (Rn<<12) | (Rs<<8) | (0b1001<<4) | (Rm))
+#define MULMULA(Cond, A, S, Rd, Rn, Rs, Rm)     (Cond | 0b000000<<22 | (A)<<21 | (S)<<20 | (Rd)<<16 | (Rn)<<12 | (Rs)<<8 | 0b1001<<4 | (Rm))
 #define MUL(Rd, Rm, Rn)     EMIT(MULMULA(c__, 0, 0, (Rd), 0, (Rm), (Rn)))

 #define SMUL_16_gen(cond, Rd, Rm, M, N, Rn) (cond | 0b00010110<<20 | (Rd)<<16 | (Rm)<<8 | 1<<7 | (M)<<6 | (N)<<5 | (Rn))
@@ -362,12 +472,22 @@ Op is 20-27
 // BFI_COND: Bit Field Insert with condition: copy any number of low order bit from Rn to any position of Rd
 #define BFI_COND(cond, rd, rn, lsb, width) EMIT(cond | (0b0111110<<21) | (((lsb)+(width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b001<<4) | (rn))

+// BFC: Bit Field Clear: clear any number of adjacent bit from Rd
+#define BFC(rd, lsb, width) EMIT(c__ | (0b0111110<<21) | (((lsb)+(width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b001<<4) | 0b1111)
+// BFC_COND: Bit Field Clear with condition: clear any number of adjacent bit from Rd
+#define BFC_COND(cond, rd, lsb, width) EMIT(cond | (0b0111110<<21) | (((lsb)+(width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b001<<4) | 0b1111)
+
 // REV: Reverse byte of a 32bits word
-#define REV(rd, rm) EMIT(c__ | (0b01101<<23) | (0<<22) | (0b11<<20) | (0b1111<<16) | ((rd)<<12) | (0b1111<<8) | (0b0011<<4) | (rm))
+#define REV(rd, rm)     EMIT(c__ | (0b01101<<23) | (0<<22) | (0b11<<20) | (0b1111<<16) | ((rd)<<12) | (0b1111<<8) | (0b0011<<4) | (rm))
+// REV16: Reverse byte of a 16bits word
+#define REV16(rd, rm)   EMIT(c__ | (0b01101<<23) | (0<<22) | (0b11<<20) | (0b1111<<16) | ((rd)<<12) | (0b1111<<8) | (0b1011<<4) | (rm))

 #define SSAT_gen(cond, sat_imm, Rd, imm5, sh, Rn) (cond | 0b0110101<<21 | (sat_imm)<<16 | (Rd)<<12 | (imm5)<<7 | (sh)<<6 | 0b01<<4 | (Rn))
 // Signed Staturate Rn to 2^(staturate_to-1) into Rd. Optionnaly shift left Rn before saturate
 #define SSAT_REG_LSL_IMM5(Rd, saturate_to, Rn, shift)   EMIT(SSAT_gen(c__, (saturate_to)-1, Rd, shift, 0, Rn))
+// Signed Staturate Rn to 2^(staturate_to-1) into Rd. Optionnaly shift left Rn before saturate
+#define SSAT_REG_LSL_IMM5_COND(cond, Rd, saturate_to, Rn, shift)   EMIT(SSAT_gen((cond), (saturate_to)-1, Rd, shift, 0, Rn))
+
 #define USAT_gen(cond, sat_imm, Rd, imm5, sh, Rn) (cond | 0b0110111<<21 | (sat_imm)<<16 | (Rd)<<12 | (imm5)<<7 | (sh)<<6 | 0b01<<4 | (Rn))
 // Unsigned Staturate Rn to 2^(staturate_to-1) into Rd. Optionnaly shift left Rn before saturate
 #define USAT_REG_LSL_IMM5(Rd, saturate_to, Rn, shift)   EMIT(SSAT_gen(c__, (saturate_to), Rd, shift, 0, Rn))
@@ -384,11 +504,33 @@ Op is 20-27

 #define LDREXD_gen(cond, Rn, Rt) (cond | 0b000<<25 | 0b11011<<20 | (Rn)<<16 | (Rt)<<12 | 0b1111<<8 | 0b1001<<4 | 0b1111)
 // Load Exclusive Rt/Rt+1 from Rn (tagging the memory)
-#define LDREXD(Rn, Rt)  EMIT(LDREXD_gen(c__, Rn, Rt))
+#define LDREXD(Rt, Rn)  EMIT(LDREXD_gen(c__, Rn, Rt))
+#define LDREXD_COND(cond, Rt, Rn)  EMIT(LDREXD_gen(cond, Rn, Rt))

 #define STREXD_gen(cond, Rd, Rn, Rt)   (cond | 0b000<<25 | 0b11010<<20 | (Rn)<<16 | (Rd)<<12 | 0b1111<<8 | 0b1001<<4 | (Rt))
 // Store Exclusive Rt/Rt+1 to Rn, with result in Rd if tag is ok (Rd!=Rn && Rd!=Rt && Rd!=Rt+1), Rd==1 if store failed
-#define STREXD(Rd, Rn, Rt)  EMIT(STREXD_gen(c__, Rd, Rn, Rt))
+#define STREXD(Rd, Rt, Rn)  EMIT(STREXD_gen(c__, Rd, Rn, Rt))
+#define STREXD_COND(cond, Rd, Rt, Rn)  EMIT(STREXD_gen(cond, Rd, Rn, Rt))
+
+#define LDREX_gen(cond, Rn, Rt)     (cond | 0b0001100<<21 | 1<<20 | (Rn)<<16 | (Rt)<<12 | 0b1111<<8 | 0b1001<<4 | 0b1111)
+// Load Exclusive Rt from Rn (tagging the memory)
+#define LDREX(Rt, Rn)       EMIT(LDREX_gen(c__, Rn, Rt))
+#define LDREX_COND(cond, Rt, Rn)       EMIT(LDREX_gen(cond, Rn, Rt))
+
+#define STREX_gen(cond, Rd, Rn, Rt) (cond | 0b0001100<<21 | 0<<20 | (Rn)<<16 | (Rd)<<12 | 0b1111<<8 | 0b1001<<4 | (Rt))
+// Store Exclusive Rt to Rn, with result in Rd=0 if tag is ok, Rd==1 if store failed (Rd!=Rn && Rd!=Rt)
+#define STREX(Rd, Rt, Rn)   EMIT(STREX_gen(c__, Rd, Rn, Rt))
+#define STREX_COND(cond, Rd, Rt, Rn)   EMIT(STREX_gen(cond, Rd, Rn, Rt))
+
+#define LDREXB_gen(cond, Rn, Rt)        (cond | 0b0001110<<21 | 1<<20 | (Rn)<<16 | (Rt)<<12 | 0b1111<<8 | 0b1001<<4 | 0b1111)
+// Load Exclusive Byte Rt from Rn (tagging the memory)
+#define LDREXB(Rt, Rn)      EMIT(LDREXB_gen(c__, Rn, Rt))
+#define LDREXB_COND(cond, Rt, Rn)      EMIT(LDREXB_gen(cond, Rn, Rt))
+
+#define STREXB_gen(cond, Rd, Rn, Rt)    (cond | 0b0001110<<21 | 0<<20 | (Rn)<<16 | (Rd)<<12 | 0b1111<<8 | 0b1001<<4 | (Rt))
+// Store Exclusive byte Rt to Rn, with result in Rd=0 if tag is ok, Rd==1 if store failed (Rd!=Rn && Rd!=Rt)
+#define STREXB(Rd, Rt, Rn)  EMIT(STREXB_gen(c__, Rd, Rn, Rt))
+#define STREXB_COND(cond, Rd, Rt, Rn)  EMIT(STREXB_gen(cond, Rd, Rn, Rt))

 // Count leading 0 bit of Rm, store result in Rd
 #define CLZ(Rd, Rm)  EMIT(c__ | 0b00010110<<20 | 0b1111<<16 | (Rd)<<12 | 0b1111<<8 | 0b0001<<4 | (Rm))
@@ -401,6 +543,23 @@ Op is 20-27
 // Preload Cache Rn-Rm
 #define PLDn(Rn, Rm) EMIT(PLD_gen(0, 1, Rn, 0, 0, Rm))

+#define SWP_gen(cond, B, Rn, Rt, Rt2)   (cond | 0b0001<<24 | (B)<<22 | (Rn)<<16 | (Rt)<<12 | 0b1001<<4 | (Rt2))
+// SWAP (atomic) [Rn]->Rt2 / Rt->[Rn], Rt can be same as Rt2
+#define SWP(Rt, Rt2, Rn)    EMIT(SWP_gen(c__, 0, Rn, Rt, Rt2))
+// SWAP Byte (atomic) [Rn]->Rt2 / Rt->[Rn], Rt can be same as Rt2, Byte is zero extended
+#define SWPB(Rt, Rt2, Rn)   EMIT(SWP_gen(c__, 1, Rn, Rt, Rt2))
+
+#define SDIV_gen(cond, Rd, Rm, Rn)  (cond | 0b0111<<24 | 0b0001<<20 | (Rd)<<16 | 0b1111<<12 | (Rm)<<8 | 0b0001<<4 | (Rn))
+// Signed Div Rd <- Rn/Rm
+#define SDIV(Rd, Rm, Rn)    EMIT(SDIV_gen(c__, Rd, Rm, Rn))
+
+#define UDIV_gen(cond, Rd, Rm, Rn)  (cond | 0b0111<<24 | 0b0011<<20 | (Rd)<<16 | 0b1111<<12 | (Rm)<<8 | 0b0001<<4 | (Rn))
+// Unsigned Div Rd <- Rn/Rm
+#define UDIV(Rd, Rm, Rn)    EMIT(UDIV_gen(c__, Rd, Rm, Rn))
+
+// Yield
+#define YIELD(cond) EMIT(cond | 0b00110010<<20 | 0b1111<<12 | 1)
+
 // VFPU
 #define TRANSFERT64(C, op) ((0b1100<<24) | (0b010<<21) | (0b101<<9) | ((C)<<8) | ((op)<<4))

@@ -485,21 +644,21 @@ Op is 20-27
 // Convert from int32 Sm to single Dd
 #define VCVT_F32_S32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b000<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (1<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )

-#define VCVT_16_gen(cond, D, op, U, Vd, sf, sx, i, imm4)   (cond | 0b1110<<24 | 1<<23 | (D)<<22 | 0b11<<20 | 1<<19 | (op)<<18 | 1<<17 | (U)<<16 | (Vd)<<12 | 0b101<<9 | (sf)<<8 | (sx)<<7 | 1<<6 | (i)<<5 | (imm4))
+#define VCVT_FP_gen(cond, D, op, U, Vd, sf, sx, i, imm4)   (cond | 0b1110<<24 | 1<<23 | (D)<<22 | 0b11<<20 | 1<<19 | (op)<<18 | 1<<17 | (U)<<16 | (Vd)<<12 | 0b101<<9 | (sf)<<8 | (sx)<<7 | 1<<6 | (i)<<5 | (imm4))
 // Inplace convert from F64 to S16. Rounding is not selectable
-#define VCVT_S16_F64(Dd)    EMIT(VCVT_16_gen(c__, ((Dd)>>4)&1, 1, 0, (Dd)&15, 1, 0, 16&1, (16>>1)&15))
+#define VCVT_S16_F64(Dd)    EMIT(VCVT_FP_gen(c__, ((Dd)>>4)&1, 1, 0, (Dd)&15, 1, 0, 16&1, (16>>1)&15))

 // Mutiply F64 Dd = Dn*Dm
-#define VMUL_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b10<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VMUL_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b10<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | ((((Dn)>>4)&1)<<7) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )

 // Divide F64 Dd = Dn/Dm
-#define VDIV_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b00<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VDIV_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b00<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | ((((Dn)>>4)&1)<<7) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )

 // Add F64 Dd = Dn + Dm
-#define VADD_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (0<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VADD_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | ((((Dn)>>4)&1)<<7) | (0<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )

 // Sub F64 Dd = Dn + Dm
-#define VSUB_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (1<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VSUB_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | ((((Dn)>>4)&1)<<7) | (1<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )

 // Mutiply F32 Sd = Sn*Sm
 #define VMUL_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (0<<23) | (((Sd)&1)<<22) | (0b10<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
@@ -514,7 +673,7 @@ Op is 20-27
 #define VSUB_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (0<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )

 // Cmp between 2 double Dd and Dm
-#define VCMP_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0100<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VCMP_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0100<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )
 // Cmp between 1 double Dd and 0.0
 #define VCMP_F64_0(Dd)      EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0101<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | (0<<5) | (0) )
 // Cmp between 2 single Sd and Sm
@@ -523,10 +682,10 @@ Op is 20-27
 #define VCMP_F32_0(Sd)      EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b11<<20) | (0b0101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0<<7) | (1<<6) | (0<<5) | (0) )

 // Neg F64 Dd = - Dm
-#define VNEG_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b01<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VNEG_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b01<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )

 // Sqrt F64 Dd = - Dm
-#define VSQRT_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b11<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+#define VSQRT_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b11<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )
 // Sqrt F32 Dd = - Dm
 #define VSQRT_F32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b11<<20) | (0b0001<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0b11<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )

@@ -590,12 +749,26 @@ Op is 20-27
 #define VEORQ(Dd, Dn, Dm)   EMIT(VEOR_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

 #define VMOVL_gen(U, D, imm3, Vd, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (imm3)<<19 | (Vd)<<12 | 0b1010<<8 | (M)<<5 | 1<<4 | (Vm))
+#define VMOVL_S8(Dd, Dm)    EMIT(VMOVL_gen(0, ((Dd)>>4)&1, 0b001, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+#define VMOVL_U8(Dd, Dm)    EMIT(VMOVL_gen(1, ((Dd)>>4)&1, 0b001, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
 #define VMOVL_S32(Dd, Dm)   EMIT(VMOVL_gen(0, ((Dd)>>4)&1, 0b100, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
 #define VMOVL_U32(Dd, Dm)   EMIT(VMOVL_gen(1, ((Dd)>>4)&1, 0b100, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))

 #define VMOV_gen(D,Vd, M, Vm, Q) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 0b10<<20 | (Vm)<<16 | (Vd)<<12 | 0b0001<<8 | (M)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
-#define VMOVD(Dd, Dm) EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 0))
-#define VMOVQ(Dd, Dm) EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 1))
+#define VMOVD(Dd, Dm) if((Dd)!=(Dm)) {EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 0));}
+#define VMOVQ(Dd, Dm) if((Dd)!=(Dm)) {EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 1));}
+
+#define VMOV_igen(i, D, imm3, Vd, cmode, Q, op, imm4)   (0b1111<<28 | 0b001<<25 | (i)<<24 | 1<<23 | (D)<<22 | (imm3)<<16 | (Vd)<<12 | (cmode)<<8 | (Q)<<6 | (op)<<5 | 1<<4 | (imm4))
+#define VMOV_8(Dd, imm)     EMIT(VMOV_igen(((imm)>>7)&1, ((Dd)>>4)&1, ((imm)>>4)&7, (Dd)&15, 0b1110, 0, 0, (imm)&15))
+#define VMOVQ_8(Dd, imm)    EMIT(VMOV_igen(((imm)>>7)&1, ((Dd)>>4)&1, ((imm)>>4)&7, (Dd)&15, 0b1110, 1, 0, (imm)&15))
+// Dd <= imm8 in high bits
+#define VMOV_H32(Dd, imm8)  EMIT(VMOV_igen(((imm8)>>7)&1, ((Dd)>>4)&1, ((imm8)>>4)&7, (Dd)&15, 0b0110, 0, 0, (imm8)&15))
+// Dd <= imm8 in high bits
+#define VMOV_H16(Dd, imm8)  EMIT(VMOV_igen(((imm8)>>7)&1, ((Dd)>>4)&1, ((imm8)>>4)&7, (Dd)&15, 0b1010, 0, 0, (imm8)&15))
+// Qd <= imm8 in high bits
+#define VMOVQ_H32(Dd, imm8) EMIT(VMOV_igen(((imm8)>>7)&1, ((Dd)>>4)&1, ((imm8)>>4)&7, (Dd)&15, 0b0110, 1, 0, (imm8)&15))
+// Qd <= imm8 in high bits
+#define VMOVQ_H16(Dd, imm8) EMIT(VMOV_igen(((imm8)>>7)&1, ((Dd)>>4)&1, ((imm8)>>4)&7, (Dd)&15, 0b1010, 1, 0, (imm8)&15))

 #define VLD1LANE_gen(D, Rn, Vd, size, index_align, Rm) (0b1111<<28 | 0b0100<<24 | 1<<23 | (D)<<22 | 0b10<<20 | (Rn)<<16 | (Vd)<<12 | (size)<<10 | (index_align)<<4 | (Rm))
 #define VLD1LANE_8(Dd, Rn, index)    EMIT(VLD1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 0, (index)<<1, 15))
@@ -644,6 +817,9 @@ Op is 20-27
 #define VBICD(Dd, Dn, Dm) EMIT(VBIC_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 0))
 #define VBICQ(Dd, Dn, Dm) EMIT(VBIC_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 1))

+#define VBIC_igen(i, D, imm3, Vd, cmode, Q, imm4)  (0b1111<<28 | 0b001<<25 | (i)<<24 | 1<<23 | (D)<<22 | (imm3)<<16 | (Vd)<<12 | (cmode)<<8 | (Q)<<6 | 0b11<<4 | (imm4))
+#define VBICQ_16_imm(Dd, imm)   EMIT(VBIC_16_imm(((imm)>>7)&1, (Dd>>4)&1, ((imm)>>4)&7, (Dd)&15, 0b1001, 1, (imm)&15))
+
 #define VCEQ_I_gen(size, D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
 #define VCEQ_8(Dd, Dn, Dm)     EMIT(VCEQ_I_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VCEQQ_8(Dd, Dn, Dm)    EMIT(VCEQ_I_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
@@ -657,6 +833,14 @@ Op is 20-27
 #define VCEQ_F32(Dd, Dn, Dm)    EMIT(VCEQ_F_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VCEQQ_F32(Dd, Dn, Dm)   EMIT(VCEQ_F_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

+#define VCEQ_0_gen(D, size, Vd, F, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b01<<16 | (Vd)<<12 | (F)<<10 | 0b010<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VCEQ_0_8(Dd, Dm)        EMIT(VCEQ_0_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_0_16(Dd, Dm)       EMIT(VCEQ_0_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_0_32(Dd, Dm)       EMIT(VCEQ_0_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_0_8(Dd, Dm)       EMIT(VCEQ_0_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_0_16(Dd, Dm)      EMIT(VCEQ_0_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_0_32(Dd, Dm)      EMIT(VCEQ_0_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+
 #define VCGT_I_gen(U, D, size, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0011<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
 #define VCGT_U8(Dd, Dn, Dm)     EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VCGTQ_U8(Dd, Dn, Dm)    EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
@@ -682,6 +866,15 @@ Op is 20-27
 #define VCGE_F32(Dd, Dn, Dm)    EMIT(VCGE_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VCGEQ_F32(Dd, Dn, Dm)   EMIT(VCGE_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

+#define VCLT_0_gen(D, size, Vd, F, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b01<<16 | (Vd)<<12 | (F)<<10 | 0b100<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VCLT_0_8(Dd, Dm)        EMIT(VCLT_0_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCLT_0_16(Dd, Dm)       EMIT(VCLT_0_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCLT_0_32(Dd, Dm)       EMIT(VCLT_0_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCLTQ_0_8(Dd, Dm)       EMIT(VCLT_0_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCLTQ_0_16(Dd, Dm)      EMIT(VCLT_0_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCLTQ_0_32(Dd, Dm)      EMIT(VCLT_0_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+
+
 #define VSHR_gen(U, D, imm6, Vd, L, Q, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (imm6)<<16 | (Vd)<<12 | (L)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
 #define VSHR_U8(Dd, Dm, imm3)    EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VSHR_S8(Dd, Dm, imm3)    EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
@@ -728,6 +921,11 @@ Op is 20-27
 #define VSHLQ_S32(Dd, Dm, Dn)   EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
 #define VSHLQ_S64(Dd, Dm, Dn)   EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

+#define VSHRN_gen(D, imm6, Vd, M, Vm)   (0b1111<<28 | 0b0010<<24 | 1<<23 | (D)<<22 | (imm6)<<16 | (Vd)<<12 | 1<<11 | (M)<<5 | 1<<4 | (Vm))
+#define VSHRN_16(Dd, Dm, imm3)          EMIT(VSHRN_gen(((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRN_32(Dd, Dm, imm4)          EMIT(VSHRN_gen(((Dd)>>4)&1, 0b01<<4  | (16-(imm4)), (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRN_64(Dd, Dm, imm5)          EMIT(VSHRN_gen(((Dd)>>4)&1, 0b1<<5   | (32-(imm5)), (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
 #define VTRN_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0000<<8 | 1<<7 | (Q)<<6 | (M)<<5 | (Vm))
 #define VTRN_32(Dd, Dm)     EMIT(VTRN_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))

@@ -766,6 +964,8 @@ Op is 20-27
 #define VCVTn_S32_F32(Dd, Dm)   EMIT(VCVT_NEON_gen(((Dd)>>4)&1, 2, (Dd)&15, 0b10, 0, ((Dm)>>4)&1, (Dm)&15))

 #define VMULL_NEON_gen(U, D, size, Vn, Vd, op, N, M, Vm)    (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b11<<10 | (op)<<9 | (N)<<7 | (M)<<5 | (Vm))
+#define VMULL_S16_S8(Dd, Dn, Dm)    EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
+#define VMULL_U16_U8(Dd, Dn, Dm)    EMIT(VMULL_NEON_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
 #define VMULL_S32_S16(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
 #define VMULL_U32_U16(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(1, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
 #define VMULL_S64_S32(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
@@ -778,8 +978,8 @@ Op is 20-27
 #define VMUL_16(Dd, Dn, Dm)      EMIT(VMUL_NEON_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))

 #define VEXT_gen(D, Vn, Vd, imm4, N, Q, M, Vm)  (0b1111<<28 | 0b0010<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (Vn)<<16 | (Vd)<<12 | (imm4)<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
-#define VEXT_8(Dd, Dn, Dm, imm4)    EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, imm4, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
-#define VEXTQ_8(Dd, Dn, Dm, imm4)   EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, imm4, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VEXT_8(Dd, Dn, Dm, imm4)    EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, (imm4)&0xf, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VEXTQ_8(Dd, Dn, Dm, imm4)   EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, (imm4)&0xf, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

 #define VMVN_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b1011<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
 #define VMVNQ(Dd, Dm)       EMIT(VMVN_gen(((Dd)>>4)&1, 0, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
@@ -809,6 +1009,10 @@ Op is 20-27
 #define VTBL1_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b00, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
 // Vector Table lookup, take elements from Dn/Dn+1, and using lookup table Dm, put elements in Dd
 #define VTBL2_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b01, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Vector Table lookup eXtension, take elements from Dn, and using lookup table Dm, put elements in Dd, outer element untouched
+#define VTBLX1_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b00, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Vector Table lookup eXtension, take elements from Dn/Dn+1, and using lookup table Dm, put elements in Dd, outer element untouched
+#define VTBLX2_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b01, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

 #define VRECPE_gen(D, size, Vd, F, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b11<<16 | (Vd)<<12 | 0b010<<9 | (F)<<8 | (Q)<<6 | (M)<<5 | (Vm))
 // Vector Reciprocal Estimate of Dm to Dd
@@ -818,11 +1022,20 @@ Op is 20-27
 // Vector Reciprocal Step of Dn and Dm to Dd
 #define VRECPSQ_F32(Dd, Dn, Dm) EMIT(VRECPS_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))

-#define VSWAP_gen(D, size, Vd, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0000<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VRSQRTE_gen(D, size, Vd, F, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b11<<16 | (Vd)<<12 | 0b010<<9 | (F)<<8 | 1<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Vector Reciprocal Square Root Estimate of Dm to Dd
+#define VRSQRTEQ_F32(Dd, Dm) EMIT(VRSQRTE_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VRSQRTS_gen(D, size, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 1<<21 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1111<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+// Vector Reciprocal Square Root Step of Dn and Dm to Dd
+#define VRSQRTSQ_F32(Dd, Dn, Dm) EMIT(VRSQRTS_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+
+#define VSWP_gen(D, size, Vd, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0000<<7 | (Q)<<6 | (M)<<5 | (Vm))
 // Swap Double vectors content
-#define VSWAP(Dd, Dm)   EMIT(VSWAP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15)))
+#define VSWP(Dd, Dm)   EMIT(VSWP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
 // Swap Quad vectors content
-#define VSWAPQ(Dd, Dm)  EMIT(VSWAP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15)))
+#define VSWPQ(Dd, Dm)  EMIT(VSWP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))

 #define VQADD_gen(U, D, size, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b001<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12  | 0b0000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
 // Add with saturation unsigned 8bits
@@ -903,4 +1116,145 @@ Op is 20-27
 #define VNEGN_F32(Dd, Dm)   EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
 #define VNEGNQ_F32(Dd, Dm)  EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))

+#define VMINMAXF_gen(D, op, sz, Vn, Vd, N, Q, M, Vm)    (0b1111<<28 | 0b0010<<24 | (D)<<22 | (op)<<21 | (sz)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1111<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VMINQ_F32(Dd, Dn, Dm)   EMIT(VMINMAXF_gen(((Dd)>>4)&1, 1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VMAXQ_F32(Dd, Dn, Dm)   EMIT(VMINMAXF_gen(((Dd)>>4)&1, 0, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VABD_gen(U, D, size, Vn, Vd, N, Q, M, Vm)   (0b1111<<28 | 1<<25 | (U)<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0111<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Compute absolute difference of U8
+#define VABD_U8(Dd, Dn, Dm)     EMIT(VABD_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VABDQ_U8(Dd, Dn, Dm)    EMIT(VABD_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VPADAL_gen(D, size, Vd, op, Q, M, Vm)   (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0110<<8 | (op)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Add pair of U8 from Dm, Add result in U16 in Dd (some position as U8U8)
+#define VPADAL_U8(Dd, Dm)   EMIT(VPADAL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U16 from Dm, Add result in U32 in Dd
+#define VPADAL_U16(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U32 from Dm, Add result in U64 in Dd
+#define VPADAL_U32(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S8 from Dm, Add result in S16 in Dd (some position as S8S8)
+#define VPADAL_S8(Dd, Dm)   EMIT(VPADAL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S16 from Dm, Add result in S32 in Dd
+#define VPADAL_S16(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S32 from Dm, Add result in S64 in Dd
+#define VPADAL_S32(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U8 from Dm, Add result in U16 in Dd (some position as U8U8)
+#define VPADALQ_U8(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U16 from Dm, Add result in U32 in Dd
+#define VPADALQ_U16(Dd, Dm) EMIT(VPADAL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U32 from Dm, Add result in U64 in Dd
+#define VPADALQ_U32(Dd, Dm) EMIT(VPADAL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S8 from Dm, Add result in S16 in Dd (some position as S8S8)
+#define VPADALQ_S8(Dd, Dm)  EMIT(VPADAL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S16 from Dm, Add result in S32 in Dd
+#define VPADALQ_S16(Dd, Dm) EMIT(VPADAL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S32 from Dm, Add result in S64 in Dd
+#define VPADALQ_S32(Dd, Dm) EMIT(VPADAL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VPADDL_gen(D, size, Vd, op, Q, M, Vm)   (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0010<<8 | (op)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Add pair of U8, store result in U16
+#define VPADDL_U8(Dd, Dm)   EMIT(VPADDL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U16, store result in U32
+#define VPADDL_U16(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U32, store result in U64
+#define VPADDL_U32(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S8, store result in S16
+#define VPADDL_S8(Dd, Dm)   EMIT(VPADDL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S16, store result in S32
+#define VPADDL_S16(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S32, store result in S64
+#define VPADDL_S32(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U8, store result in U16
+#define VPADDLQ_U8(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U16, store result in U32
+#define VPADDLQ_U16(Dd, Dm) EMIT(VPADDL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of U32, store result in U64
+#define VPADDLQ_U32(Dd, Dm) EMIT(VPADDL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S8, store result in S16
+#define VPADDLQ_S8(Dd, Dm)  EMIT(VPADDL_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S16, store result in S32
+#define VPADDLQ_S16(Dd, Dm) EMIT(VPADDL_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add pair of S32, store result in S64
+#define VPADDLQ_S32(Dd, Dm) EMIT(VPADDL_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VPADD_gen(D, size, Vn, Vd, N, Q, M, Vm)    (0b1111<<28 | 0b0010<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1011<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VPADD_8(Dd, Dn, Dm)     EMIT(VPADD_gen(((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VPADD_16(Dd, Dn, Dm)    EMIT(VPADD_gen(((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VPADD_32(Dd, Dn, Dm)    EMIT(VPADD_gen(((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// NO VPADDQ....
+
+#define VPADD_Fgen(D, size, Vn, Vd, N, Q, M, Vm)   (0b1111<<28 | 0b0011<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1101<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+// Add pair of F32, store result in F32
+#define VPADD_F32(Dd, Dn, Dm)   EMIT(VPADD_Fgen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMINMAX_gen(U, D, size, Vn, Vd, N, Q, M, op, Vm)    (0b1111<<28 | 0b001<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b110<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (op)<<4 | (Vm))
+#define VMIN_U8(Dd, Dn, Dm)     EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_U16(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_U32(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_U64(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_S8(Dd, Dn, Dm)     EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_S16(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_S32(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMIN_S64(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_U8(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_U16(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_U32(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_U64(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_S8(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_S16(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_S32(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMINQ_S64(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 1, (Dm)&15))
+#define VMAX_U8(Dd, Dn, Dm)     EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_U16(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_U32(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_U64(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_S8(Dd, Dn, Dm)     EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_S16(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_S32(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAX_S64(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_U8(Dd, Dn, Dm)    EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_U16(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_U32(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_U64(Dd, Dn, Dm)   EMIT(VMINMAX_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_S8(Dd, Dn, Dm)    EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_S16(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_S32(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+#define VMAXQ_S64(Dd, Dn, Dm)   EMIT(VMINMAX_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, 0, (Dm)&15))
+
+#define VRHADD_gen(U, D, size, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 1<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Dd <= (Dn + Dm + 1)>>1
+#define VRHADD_U8(Dd, Dn, Dm)   EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADD_U16(Dd, Dn, Dm)  EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADD_U32(Dd, Dn, Dm)  EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADD_S8(Dd, Dn, Dm)   EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADD_S16(Dd, Dn, Dm)  EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADD_S32(Dd, Dn, Dm)  EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_U8(Dd, Dn, Dm)  EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_U16(Dd, Dn, Dm) EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_U32(Dd, Dn, Dm) EMIT(VRHADD_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_S8(Dd, Dn, Dm)  EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_S16(Dd, Dn, Dm) EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VRHADDQ_S32(Dd, Dn, Dm) EMIT(VRHADD_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VQRDMULH_gen(D, size, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b0011 <<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1011<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Dd <= (2*Dn*Dm + 1<<(size-1))>>size
+#define VQRDMULH_S16(Dd, Dn, Dm)    EMIT(VQRDMULH_gen(((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VQRDMULH_S32(Dd, Dn, Dm)    EMIT(VQRDMULH_gen(((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VQRDMULHQ_S16(Dd, Dn, Dm)   EMIT(VQRDMULH_gen(((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VQRDMULHQ_S32(Dd, Dn, Dm)   EMIT(VQRDMULH_gen(((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VDUP_sgen(D, imm4, Vd, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (imm4)<<16 | (Vd)<<12 | 0b1100<<8 | (Q)<<6 | (M)<<5 | (Vm))
+// Dd <= Dup(Dm[x])
+#define VDUP_32(Dd, Dm, x)      EMIT(VDUP_sgen(((Dd)>>4)&1, (((x)<<3)|0b100), (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VABS_vgen(D, size, Vd, F, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b01<<16 | (Vd)<<12 | (F)<<10 | 0b110<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VABS_S32(Dd, Dm)        EMIT(VABS_vgen(((Dd)>>4)&1, 2, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VABS_S16(Dd, Dm)        EMIT(VABS_vgen(((Dd)>>4)&1, 1, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VABS_S8(Dd, Dm)         EMIT(VABS_vgen(((Dd)>>4)&1, 0, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VABS_F(Dd, Dm)          EMIT(VABS_vgen(((Dd)>>4)&1, 2, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VABSQ_S32(Dd, Dm)       EMIT(VABS_vgen(((Dd)>>4)&1, 2, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VABSQ_S16(Dd, Dm)       EMIT(VABS_vgen(((Dd)>>4)&1, 1, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VABSQ_S8(Dd, Dm)        EMIT(VABS_vgen(((Dd)>>4)&1, 0, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VABSQ_F(Dd, Dm)         EMIT(VABS_vgen(((Dd)>>4)&1, 2, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
 #endif  //__ARM_EMITTER_H__
@@ -10,9 +10,18 @@
 arm_epilog:
    //update register -> emu
    //pop     {r0}
-    stm     r0, {r4-r12}   // put back reg value in emu, including EIP (so r12 must be EIP now)
+    stm     r0, {r4-r12,r14}   // put back reg value in emu, including EIP (so r14 must be EIP now)
    //restore all used register
-    //vpop     {d8-d15}
+    vpop     {d8-d15}
+    pop     {r4-r11, pc}
+    //end, return...
+    //bx      lr
+
+
+.global arm_epilog_fast
+arm_epilog_fast:
+    //restore all used register
+    vpop     {d8-d15}
    pop     {r4-r11, pc}
    //end, return...
    //bx      lr
@@ -0,0 +1,596 @@
+; TYPE fields separated by spaces Assembler syntax
+; When there are mutiple assembler syntax (like for LDR (immediate)), separate them based on the bytes
+; E.G. for LDR (immediate), there are 3 lines with the 3 different asm syntaxes
+
+# Field options:
+# --------------
+# Syntax: ... field/coption1(coption2)* ... (inside fields)
+# Where each c correspond to the option character
+#  and optionN to the Nth option value
+# ---
+# c = '0'..'9'/'A'..'F': option=[01]
+# Force the cth bit (hex) of the field (starting from LSB) to be <option>
+# c = '=': option=[01x]{field_len}
+# Force the field to have the given bits (where x = anything, starts from MSB)
+
+# Post-fields option:
+# -------------------
+# Syntax: [fields] @ typeoption1 args1 typeoption2 args2 @ [Assembler syntax]
+# ---
+# typeoption contains '=' (l=r): args1=((.*) )?([^, ]*),([^, ]*),([^, ]*)@
+# Changes a value between two other depending on whether l == r
+
+# \2 is init (suffixed by ';', optional), \3 is set to \4 if true or \5 if false
+# (cf VMOV/VORR aka V\op\, VEOR/VBIF/VBIT/VBSL aka V\op\)
+
+# The 'true regex' is typeoption=([^= ]*)(=([^= ]*))+
+# args1=((.*) )?([^, ]*),(([^, ]*),)+([^, ]*)@
+# with the number of typeoption:\2 being equal to the number of args1:\4
+# <a>=<b>[=<c>...] <initialization> <variable-changed>,<changed-into-if-a=b>[,<changed-into-if-a=c>...],<changed-into-else>@
+
+# OR
+
+# typeoption is discarded and must be set to '=', args1=((.*) )?(@@ (( ?.*[^@])*)@ )+([^, ]*),(([^, ]*),)+([^, ]*)
+# \2 is init (suffixed by ';', optional), \6 is set to \7 if true or \8 if false
+# The difference with the simple switch is that \4 is the test (inside the `if`)
+# If there are multiple @@s, there are multiple `if`s (separated by `else`s)
+
+# The 'true regex' is [DISCARDED]typeoption==
+# args1=((.*) )?(@@ (( ?.*[^@])*)@ )+([^, ]*),([^, ]*),([^, ]*)
+# <.>=<.>[=<c>...] <initialization> @@ <test>@[ @@ <test2>@...] <variable-changed>,<changed-into-if-a=b>,<changed-into-else>
+
+# OR
+
+# args1=((.*) )?!@(( [^%]*[^%@])*)%(( ?[^%]*[^%@])*)@(( .*[^@])*)@(( .*[^@])*)@
+# Replaced by '\1; if (<equality>) { \3\7\5; } else { \3\9\5; }'
+# Meaning, <.=.> <initialization> !@ <common-left-part>%<common-right-part>@ <substitute-true>@ <substitute-false>@ <...>
+# (cf VMOV/VORR, aka V\op\ with the lastop variable)
+# You can also add more '=' with corresponding substitutes.
+# You can also add more '%' with corresponding substitutes.
+
+# ----
+# Inside the assembler syntax, you can add \var\ to mark a variable to be replaced
+# (Meaning, %s and `var` as argument to printf)
+# Full syntax: \\(%[0-9+]*.)?(.*)\\
+# If \1 != "", add \1 to the string in printf, and in any case append \2 as argument to printf
+
+! HACKS: (list of notable differences between the doc and this file)
+! ------
+! In theory, (0) and (1) bits are as important as 0 and 1:
+! > An instruction is UNPREDICTABLE if:
+! > • It is declared as UNPREDICTABLE in an instruction description or in this chapter.
+! > • The pseudocode for that encoding does not indicate that a different special case applies, and a bit marked (0)
+!     or (1) in the encoding diagram of an instruction is not 0 or 1 respectively.
+! - ARM Architecture Reference Manual ARMv7-A and ARMv7-R edition
+! In practice, they are ignored here.
+
+! SWP, SWPB, VMOV  (VFPU): Rt2 (field and assembly) -> Rn (field and assembly)
+! BKPT: must be AL, so cond (field) -> cond/=1110 (constrained field)
+! PKHBT and PKHTB: {, LSL/ASR #<imm>} (assembly) -> {, <shift>} (assembly)
+! PKHBT and PKHTB: immediate shift but shift & 1 -> ARMs (type)
+! All NEON operations (advanced SIMD): limited, so multiply the lines
+! VMOV (register): Vm ... Vm VMOV (field, assembly) -> Vn ... Vm @ m=n ...@ @ V\op\ (field, switchable var, assembly)
+! VMOV is VORR constrained -> (removed VORR)
+! VEOR, VBSL, VBIT, VBIF differ only depending on 2 bits: 0/1 0/1 ... VEOR -> @<2> ... @ ... @ V\op\ (field, switchable var, assembly)
+! Three registers of different lengths (when having L and W variants): op (field) -> Q (field)
+
+; ========== BEGINNING OF REGISTERED INSTRUCTIONS =========
+
+# This instruction changes the instruction set to Thumb! Disabled (but valid)
+#ARM_ 1 1 1 1 1 0 1 @<1> imm24 @ set imm24 = (imm24 << 2) + (param << 1)@ @ BLX <label>
+
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 op 0 N Q M 0 Vm VH\(op ? "SUB" : "ADD")\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 0 0 N Q M 1 Vm VQADD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 0 1 N Q M 0 Vm VRHADD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 0 Vn Vd 0 0 0 1 N Q M 1 Vm VAND<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 1 Vn Vd 0 0 0 1 N Q M 1 Vm VBIC<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 1 0 Vn Vd 0 0 0 1 N Q M 1 Vm @ m=n const char* op op,"MOV","ORR"@ m=n char lastop[6] !@ %@ lastop[0] = '\0'@ sprintf(lastop, ", %s", vecname[(q << 5) + 0x20 + n])@ @ V\op\<c> <Qd>, <Qm>\lastop\
+NEON 1 1 1 1 0 0 1 0 0 D 1 1 Vn Vd 0 0 0 1 N Q M 1 Vm VORN<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D @<2> Vn Vd 0 0 0 1 N Q M 1 Vm @ param=0b00=0b01=0b10 const char* op op,"EOR","BSL","BIT","BIF"@ @ V\op\<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 1 0 N Q M 1 Vm VQSUB<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 1 1 N Q M op Vm @ op=0 const char* strict strict,"T","E"@ @ VCG\strict\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 0 N Q M 0 Vm VSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 0 N Q M 1 Vm VQSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 1 N Q M 0 Vm VRSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 1 N Q M 1 Vm VQRSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 0 N Q M op Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ V\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 1 N Q M 0 Vm VABD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 1 N Q M 1 Vm VABA<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 op 0 D size Vn Vd 1 0 0 0 N Q M 0 Vm @ op=0 const char* opname opname,"ADD","SUB"@ @ V\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 0 0 N Q M 1 Vm VTST<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D size Vn Vd 1 0 0 0 N Q M 1 Vm VCEQ<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 0 1 N Q M 0 Vm @ set size = 8 << size@ @ VML\(u ? "S" : "A")\<c>.I\%dsize\ <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 0 1 N Q M 1 Vm VMUL<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 1 0 N Q M op Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ VP\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 1 1 N Q M 0 Vm VQDMULH<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D size Vn Vd 1 0 1 1 N Q M 0 Vm VQRDMULH<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 1 1 N Q/=0 M 1 Vm VPADD<c>.<dt> <Dd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 0 D U sz/=0 Vn Vd 1 1 0 0 N Q M 1 Vm @ u=0 const char* y y,"A","S"@ @ VFM\y\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D U sz/=0 Vn Vd 1 1 0 1 N Q M 0 Vm @ u=0 const char* op op,"ADD","SUB"@ @ V\op\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D 0 sz/=0 Vn Vd 1 1 0 1 N Q/=0 M 0 Vm VPADD.F32 <Dd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 1 0 D 1 sz/=0 Vn Vd 1 1 0 1 N Q M 0 Vm VABD.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D op sz/=0 Vn Vd 1 1 0 1 N Q M 1 Vm VML\(op ? "S" : "A")\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D 0 sz/=0 Vn Vd 1 1 0 1 N Q M 1 Vm VMUL.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 sz/=0 Vn Vd 1 1 1 0 N Q M 0 Vm VCEQ.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 0 N Q M 0 Vm @ op=1 const char* strict strict,"T","E"@ @ VCG\strict\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 0 N Q M 1 Vm @ op=1 const char* strict strict,"T","E"@ @ VACG\strict\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D op sz/=0 Vn Vd 1 1 1 1 N Q M 0 Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ V\opname\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 1 N Q M 0 Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ VP\opname\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 sz/=0 Vn Vd 1 1 1 1 N Q M 1 Vm VRECPS.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 1 sz/=0 Vn Vd 1 1 1 1 N Q M 1 Vm VRSQRTS.F32 <Qd>, <Qn>, <Qm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 0 <23>
+
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx0 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx1 0 Q op/=0 1 imm4 VORR<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x0 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x1 0 Q op/=0 1 imm4 VORR<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=11xx 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx0 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx1 0 Q op/=1 1 imm4 VBIC<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x0 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x1 0 Q op/=1 1 imm4 BIC<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=110x 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=1110 0 Q op/=1 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+;INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 0 0 0 <3> <4> <4>/=1111 0 <1> <1>/=1 1 <4>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 0 0 0 <11> 0 <2> 1 <4>
+
+# Special case for VMOVL (harder to do it on the documented place)
+NEON 1 1 1 1 0 0 1 0 1 D 0 0 1 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S8 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D 0 1 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D 1 0 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S32 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 0 0 1 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U8 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 0 1 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 0 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U32 <Qd>, <Dm>
+
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 0 0 L Q M 1 Vm VSHR<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 0 1 L Q M 1 Vm VSRA<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 1 0 L Q M 1 Vm VRSHR<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 1 1 L Q M 1 Vm VRSRA<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D imm6 Vd 0 1 0 0 L Q M 1 Vm VSRI<c>.<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D Imm6 Vd 0 1 0 1 L Q M 1 Vm VSHL<c>.I<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D Imm6 Vd 0 1 0 1 L Q M 1 Vm VSLI<c>.<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D Imm6 Vd 0 1 1 op L Q M 1 Vm VQSHL\((u & op) ? "U" : "")\<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D imm6 Vd 1 0 0 0 0 0 M 1 Vm @ set ++size@ @ VSHRN<c>.I<size> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D imm6 Vd 1 0 0 0 0 1 M 1 Vm VRSHRN<c>.I<size> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 1 0 0 op 0 0 M 1 Vm VQSHR\((u & op) ? "U" : "")\N<c>.<dt> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 1 0 0 op 0 1 M 1 Vm VQRSHR\((u & op) ? "U" : "")\N<c>.<dt> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 1 0 0 M 1 Vm VSHLL<c>.<dt> <Qd>, <Dm>, #\%d8 &l&l size\
+NEON 1 1 1 1 0 0 1 U 1 D Imm6 Vd 1 0 1 0 0 0 M 1 Vm VSHLL<c>.<dt> <Qd>, <Dm>, #<imm>
+; VCVT: if imm6 IN "0xxxxx" then UNDEFINED;
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 0 <9> 1 1 1 <1> 0 <2> 1 <4>
+NEON 1 1 1 1 0 0 1 0 1 D IMM6 Vd 1 1 1 0 0 Q M 1 Vm VCVT<c>.F32.S32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 0 1 D IMM6 Vd 1 1 1 1 0 Q M 1 Vm VCVT<c>.S32.F32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 1 1 D IMM6 Vd 1 1 1 0 0 Q M 1 Vm VCVT<c>.F32.U32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 1 1 D IMM6 Vd 1 1 1 1 0 Q M 1 Vm VCVT<c>.U32.F32 <Qd>, <Qm>, #<fbits>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <18> 1 <4>
+
+NEON 1 1 1 1 0 0 1 0 1 D 1 1 Vn Vd IMM4 N Q M 0 Vm VEXT<c>.8 <Qd>, <Qn>, <Qm>, #<imm>
+
+# size + param < 3
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=11 0 0 <4> 0 0 0 <2> <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=10 0 0 <4> 0 0 0 <2>/11 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=10 0 0 <4> 0 0 0 <2>/01 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=01 0 0 <4> 0 0 0 <2>/11 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=00 0 0 <4> 0 0 0 <2>/=11 <2> 0 <4>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 0 0 @<2> Q M 0 Vm VREV\%d16 &l&l param\<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 0 1 0 U Q M 0 Vm VPADDL<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 0 0 0 Q M 0 Vm VCLS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 0 0 1 Q M 0 Vm VCLZ<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 0 0 Vd 0 1 0 1 0 Q M 0 Vm VCNT<c>.8 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 0 0 Vd 0 1 0 1 1 Q M 0 Vm VMVN<c> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 0 U Q M 0 Vm VPADAL<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 1 0 Q M 0 Vm VQABS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 1 1 Q M 0 Vm VQNEG<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 0 0 Q M 0 Vm VCGT<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 0 1 Q M 0 Vm VCGE<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 1 0 Q M 0 Vm VCEQ<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 1 1 Q M 0 Vm VCLE<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 0 0 Q M 0 Vm VCLT<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 1 0 Q M 0 Vm VABS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 1 1 Q M 0 Vm VNEG<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 1 0 Vd 0 0 0 0 0 Q M 0 Vm VSWP<c> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 0 1 Q M 0 Vm VTRN<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 1 0 Q M 0 Vm VUZP<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 1 1 Q M 0 Vm VZIP<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 0 0 M 0 Vm VMOVN<c>.<dt> <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 0 1 M 0 Vm VQMOVUN<c>.S\%d16 &l&l size\ <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 1 U M 0 Vm @ set ++size@ @ VQMOVN<c>.<dt> <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 1 0 0 M 0 Vm VSHLL<c>.I\%d8 &l&l size\ <Qd>, <Dm>, #\%d8 &l&l size\
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=01 1 0 Vd 0 1 1 0 0 0 M 0 Vm VCVT<c>.F16.F32 <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=01 1 0 Vd 0 1 1 1 0 0 M 0 Vm VCVT<c>.F32.F16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 1 Vd 0 1 0 F 0 Q M 0 Vm VRECPE<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 1 Vd 0 1 0 F 1 Q M 0 Vm VRSQRTE<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 0 0 Q M 0 Vm VCVT<c>.F32.S32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 0 1 Q M 0 Vm VCVT<c>.S32.F32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 1 0 Q M 0 Vm VCVT<c>.F32.U32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 1 1 Q M 0 Vm VCVT<c>.U32.F32 <Qd>, <Qm>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <8> 0 <6> 0 <4>
+
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 Vn Vd 1 0 @<2> N op M 0 Vm @ param=0b00 char list[10] !@ sprintf(list, %)@ "{%s}", vecname[0x20 + n]@ "{%s-%s}", vecname[0x20 + n], vecname[0x20 + n + param]@ @ VTB\(op ? "X" : "L")\<c>.8 <Dd>, \list\, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 imm4 Vd 1 1 0 0 0 Q M 0 Vm VDUP<c>.\%dsize\ <Qd>, <Dm[x]>
+
+; size != 11
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 1 1 <13> 0 <1> 0 <4>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 0 0 Q N 0 M 0 Vm VADD\(q ? "W" : "L")\<c>.<dt> <qd>, <Qn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 0 1 Q N 0 M 0 Vm VSUB\(q ? "W" : "L")\<c>.<dt> <qd>, <Qn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 1 0 0 N 0 M 0 Vm @ set size = 16 << size@ @ VADDHN<c>.I\%dsize\ <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D size Vn Vd 0 1 0 0 N 0 M 0 Vm VRADDHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 1 0 1 N 0 M 0 Vm VABAL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 1 1 0 N 0 M 0 Vm VSUBHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D size Vn Vd 0 1 1 0 N 0 M 0 Vm VRSUBHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 1 1 1 N 0 M 0 Vm VABDL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 0 op 0 N 0 M 0 Vm VML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 0 op 1 N 0 M 0 Vm VQDML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 1 0 0 N 0 M 0 Vm VMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 1 0 1 N 0 M 0 Vm VQDMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U/=0 1 D size/=00 Vn Vd 1 1 1 0 N 0 M 0 Vm @ set uint8_t size = 0@ @ VMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <16> 0 <1> 0 <4>
+
+; Put after, otherwise some instructions are shadowed
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 0 op 0 F N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VML\(op ? "S" : "A")\<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 op 1 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 op 1 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 0 0 F N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VMUL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 0 1 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VMULL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 0 1 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDMULL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 1 0 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDMULH<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 1 0 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQRDMULH<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <16> 0 <1> 0 <4>
+
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+;INVALIDATE 1 1 1 1 0 1 0 0 0 <1> 0 0 <22>
+INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 0 0 <8> 1 1 <10>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> 1 1 1 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> 1 1 0 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> Rm @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+;INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 0 0 <20>
+;INVALIDATE 1 1 1 1 0 1 0 0 <2> 0 0 <20>
+
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+;INVALIDATE 1 1 1 1 0 1 0 0 0 <1> 1 0 <22>
+INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 1 0 <8> 1 1 <10>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> 1 1 1 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> 1 1 0 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> Rm @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+;INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 1 0 <20>
+;INVALIDATE 1 1 1 1 0 1 0 0 <2> 1 0 <20>
+INVALIDATE 1 1 1 1 0 1 0 0 <3> 0 <20>
+
+; ===== No more 1 1 1 1 ..., so invalidate all of them ====
+INVALIDATE 1 1 1 1 <28>
+
+
+ARMS cond 0 0 0 0 0 0 0 S Rn Rd imm5 type 0 Rm AND{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 0 0 S Rd (0) (0) (0) (0) Rm 1 0 0 1 Rn MUL{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRH<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRD<c> <Rt>, <Rt2>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm STRD<c> <Rt>, <Rt2>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRH<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSB<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSH<c> <Rt>, [<Rn>], +/-<Rm>
+ARMS cond 0 0 0 0 0 0 1 S Rn Rd imm5 type 0 Rm EOR{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 0 1 S Rd Ra Rm 1 0 0 1 Rn MLA{S}<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 0 U 0 1 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSBT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARMS cond 0 0 0 0 0 1 0 S Rn Rd imm5 type 0 Rm SUB{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 1 0 0 RdHi RdLo Rm 1 0 0 1 Rn UMAAL<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 0 1 1 imm4L STRH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 1 0 1 imm4L LDRD<c> <Rt>, <Rt2>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 1 1 1 imm4L STRD<c> <Rt>, <Rt2>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 0 1 1 imm4L LDRH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSB<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARMS cond 0 0 0 0 0 1 1 S Rn Rd imm5 type 0 Rm RSB{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 1 1 0 Rd Ra Rm 1 0 0 1 Rn MLS<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 0 U 1 1 0 Rn Rt imm4H 1 0 1 1 imm4L STRHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 0 1 1 imm4L LDRHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSBT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARMS cond 0 0 0 0 1 0 0 S Rn Rd imm5 type 0 Rm ADD{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 0 0 S RdHi RdLo Rm 1 0 0 1 Rn UMULL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 0 1 S Rn Rd imm5 type 0 Rm ADC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 0 1 S RdHi RdLo Rm 1 0 0 1 Rn UMLAL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 1 0 S Rn Rd imm5 type 0 Rm SBC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 1 0 S RdHi RdLo Rm 1 0 0 1 Rn SMULL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 1 1 S Rn Rd imm5 type 0 Rm RSC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 1 1 S RdHi RdLo Rm 1 0 0 1 Rn SMLAL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 0 0 0 Rn SMLABB<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 0 1 0 Rn SMLATB<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 1 0 0 Rn SMLABT<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 1 1 0 Rn SMLATT<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rn Rt (0) (0) (0) (0) 1 0 0 1 Rm SWP<c> <Rt>, <Rm>, [<Rn>]
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRD<c> <Rt>, <Rt2>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm STRD<c> <Rt>, <Rt2>, [<Rn>, +/-<Rm>]{!}
+ARMS cond 0 0 0 1 0 0 0 1 Rn (0) (0) (0) (0) imm5 type 0 Rm TST<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSB<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 0 1 Rm BX<c> <Rm>
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 1 0 Rm BXJ<c> <Rm>
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 1 1 Rm BLX<c> <Rm>
+ARM_ cond/=1110 0 0 0 1 0 0 1 0 imm12 0 1 1 1 imm4 BKPT #<imm16>
+ARMS cond 0 0 0 1 0 0 1 1 Rn (0) (0) (0) (0) imm5 type 0 Rm TEQ<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 0 1 0 0 Rn Rt (0) (0) (0) (0) 1 0 0 1 Rm SWPB<c> <Rt>, <Rm>, [<Rn>]
+ARM_ cond 0 0 0 1 0 1 1 0 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 0 1 Rm CLZ<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 0 1 1 imm4L STRH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 1 0 1 imm4L LDRD<c> <Rt>, <Rt2>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 1 1 1 imm4L STRD<c> <Rt>, <Rt2>, [<Rn>, #+/-<imm8>]{!}
+ARMS cond 0 0 0 1 0 1 0 1 Rn (0) (0) (0) (0) imm5 type 0 Rm CMP<c> <Rn>, <Rm>{, <shift>}
+ARMS cond 0 0 0 1 0 1 1 1 Rn (0) (0) (0) (0) imm5 type 0 Rm CMN<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 0 1 1 imm4L LDRH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSB<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARMS cond 0 0 0 1 1 0 0 S Rn Rd imm5 type 0 Rm ORR{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 0 0 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREX<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 0 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREX<c> <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd 0 0 0 0 0 0 0 0 Rm MOV{S}<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 0 0 0 Rm LSL{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 0 0 1 Rn LSL{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 0 1 0 Rm LSR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 0 1 1 Rn LSR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 1 0 0 Rm ASR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 1 0 1 Rn ASR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd 0 0 0 0 0 1 1 0 Rm RRX{S}<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 1 1 0 Rm ROR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 1 1 1 Rn ROR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXD<c> <Rd>, <Rt>, <Rt2>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 1 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXD<c> <Rt>, <Rt2>, [<Rn>]
+ARMS cond 0 0 0 1 1 1 0 S Rn Rd imm5 type 0 Rm BIC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 1 0 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXB<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 1 0 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXB<c> <Rt>, [<Rn>]
+ARMS cond 0 0 0 1 1 1 1 S (0) (0) (0) (0) Rd imm5 type 0 Rm MVN{S}<c> <Rd>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 1 1 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXH<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 1 1 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXH<c> <Rt>, [<Rn>]
+ARM_ cond 0 0 1 0 0 0 0 S Rn Rd imm12 AND{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 0 1 S Rn Rd imm12 EOR{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 1 0 S Rn Rd imm12 SUB{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 1 1 S Rn Rd imm12 RSB{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 0 0 S Rn Rd imm12 ADD{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 0 1 S Rn Rd imm12 ADC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 1 0 S Rn Rd imm12 SBC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 1 1 S Rn Rd imm12 RSC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 0 0 0 imm4 Rd imm12 MOVW<c> <Rd>, #<imm16>
+ARM_ cond 0 0 1 1 0 0 0 1 Rn (0) (0) (0) (0) imm12 TST<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 0 1 0 0 0 0 0 (1) (1) (1) (1) (0) (0) (0) (0) 0 0 0 0 0 0 0 0 NOP<c>
+ARM_ cond 0 0 1 1 0 0 1 1 Rn (0) (0) (0) (0) imm12 TEQ<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 1 0 0 imm4 Rd imm12 MOVT<c> <Rd>, #<imm16>
+ARM_ cond 0 0 1 1 0 1 0 1 Rn (0) (0) (0) (0) imm12 CMP<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 1 1 1 Rn (0) (0) (0) (0) imm12 CMN<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 0 0 S Rn Rd imm12 ORR{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 0 1 S (0) (0) (0) (0) Rd imm12 MOV{S}<c> <Rd>, #<const>
+ARM_ cond 0 0 1 1 1 1 0 S Rn Rd imm12 BIC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 1 1 S (0) (0) (0) (0) Rd imm12 MVN{S}<c> <Rd>, #<const>
+ARM_ cond 0 1 0 0 1 0 0 1 1 1 0 1 Rt 0 0 0 0 0 0 0 0 0 1 0 0 POP<c> {<Rt>}
+ARM_ cond 0 1 0 0 U 0 0 0 Rn Rt imm12 STR<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 0 1 Rn Rt imm12 LDR<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 1 0 Rn Rt imm12 STRT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 1 1 Rn Rt imm12 LDRT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 0 0 Rn Rt imm12 STRB<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 0 1 Rn Rt imm12 LDRB<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 1 0 Rn Rt imm12 STRBT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 1 1 Rn Rt imm12 LDRBT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 1 0 0 1 0 1 1 0 1 Rt 0 0 0 0 0 0 0 0 0 1 0 0 PUSH<c> {<Rt>}
+ARM_ cond 0 1 0 1 U 0 W 0 Rn Rt imm12 STR<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 0 W 1 Rn Rt imm12 LDR<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 1 W 0 Rn Rt imm12 STRB<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 1 W 1 Rn Rt imm12 LDRB<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 0 1 0 1 Rm SSAX<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 0 1 1 1 Rm SSUB16<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 1 1 1 1 Rm SSUB8<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 0 1 0 1 Rm USAX<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 0 1 1 1 Rm USUB16<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 1 1 1 1 Rm USUB8<c> <Rd>, <Rn>, <Rm>
+ARMs cond 0 1 1 0 1 0 0 0 Rn Rd imm5 0 0 1 Rm PKHBT<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARMs cond 0 1 1 0 1 0 0 0 Rn Rd imm5 1 0 1 Rm PKHTB<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 1 1 0 1 0 0 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTB16<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 0 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAB16<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 1 0 1 sat_imm Rd imm5 sh 0 1 Rn SSAT<c> <Rd>, #<imm>, <Rn>{, <shift>}
+ARM_ cond 0 1 1 0 1 0 1 0 sat_imm Rd (1) (1) (1) (1) 0 0 1 1 Rn SSAT16<c> <Rd>, #<imm>, <Rn>
+ARM_ cond 0 1 1 0 1 0 1 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTB<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAB<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 1 1 Rm REV<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 0 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 1 0 1 1 Rm REV16<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 0 1 1 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTH<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 1 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAH<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 0 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTB16<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 0 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAB16<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 1 1 1 sat_imm Rd imm5 sh 0 1 Rn USAT<c> <Rd>, #<imm>, <Rn>{, <shift>}
+ARM_ cond 0 1 1 0 1 1 1 0 sat_imm Rd (1) (1) (1) (1) 0 0 1 1 Rn USAT16<c> <Rd>, #<imm>, <Rn>
+ARM_ cond 0 1 1 0 1 1 1 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTB<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAB<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 1 1 Rm RBIT<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 1 1 1 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTH<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 1 0 1 1 Rm REVSH<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 1 1 1 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAH<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 U 0 0 0 Rn Rt imm5 type 0 Rm STR<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 0 1 Rn Rt imm5 type 0 Rm LDR<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 1 0 Rn Rt imm5 type 0 Rm STRT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 1 1 Rn Rt imm5 type 0 Rm LDRT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 0 0 Rn Rt imm5 type 0 Rm STRB<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 0 1 Rn Rt imm5 type 0 Rm LDRB<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 1 0 Rn Rt imm5 type 0 Rm STRBT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 1 1 Rn Rt imm5 type 0 Rm LDRBT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 1 U 0 W 0 Rn Rt imm5 type 0 Rm STR<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 0 W 1 Rn Rt imm5 type 0 Rm LDR<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 1 W 0 Rn Rt imm5 type 0 Rm STRB<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 1 W 1 Rn Rt imm5 type 0 Rm LDRB<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARM_ cond 0 1 1 1 1 0 0 0 Rd 1 1 1 1 Rm 0 0 0 1 Rn USAD8<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 1 1 0 0 0 Rd Ra Rm 0 0 0 1 Rn USADA8<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 1 1 1 1 1 0 msb Rd lsb 0 0 1 1 1 1 1 BFC<c> <Rd>, #<lsb>, #<width>
+ARM_ cond 0 1 1 1 1 1 0 msb Rd lsb 0 0 1 Rn BFI<c> <Rd>, <Rn>, #<lsb>, #<width>
+ARM_ cond 0 1 1 1 1 1 1 widthm1 Rd lsb 1 0 1 Rn UBFX<c> <Rd>, <Rn>, #<lsb>, #<width>
+ARM_ cond 1 0 0 0 0 0 W 0 Rn/=1101 register_list STMED<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 0 Rn register_list STMDA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 1 Rn/=1101 register_list LDMED<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 1 Rn register_list LDMDA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 W 0 Rn/=1101 register_list STMEA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 W 0 Rn register_list STM<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 0 1 Rn/=1101 register_list LDMEA<c> <Rn>, <registers>
+ARM_ cond 1 0 0 0 1 0 1 1 1 1 0 1 register_list @ = const char* op @@ (reglist != 0) && ((reglist & (reglist - 1)) == 0)@ op,"POP","LDMEA"@ @ \op\<c> <registers>
+ARM_ cond 1 0 0 0 1 0 W 1 Rn register_list LDM<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 0 0 Rn/=1101 register_list STMFD<c> <Rn>, <registers>
+ARM_ cond 1 0 0 1 0 0 1 0 1 1 0 1 register_list @ = const char* op @@ (reglist != 0) && ((reglist & (reglist - 1)) == 0)@ op,"PUSH","STMFD"@ @ \op\<c> <registers>
+ARM_ cond 1 0 0 1 0 0 W 0 Rn register_list STMDB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 W 1 Rn/=1101 register_list LDMFD<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 W 1 Rn register_list LDMDB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 0 Rn/=1101 register_list STMFA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 0 Rn register_list STMIB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 1 Rn/=1101 register_list LDMFA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 1 Rn register_list LDMIB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 1 0 imm24 B<c> <label>
+ARM_ cond 1 0 1 1 imm24 BL<c> <label>
+
+VFPU cond 1 1 0 0 0 1 0 0 Rn Rt 1 0 1 0 0 0 M 1 Vm VMOV<c> <Sm>, <Sm1>, <Rt>, <Rn>
+VFPU cond 1 1 0 0 0 1 0 1 Rn Rt 1 0 1 0 0 0 M 1 Vm VMOV<c> <Rt>, <Rn>, <Sm>, <Sm1>
+VFPU cond 1 1 0 0 0 1 0 0 Rn Rt 1 0 1 1 0 0 M 1 Vm VMOV<c> <Dm>, <Rt>, <Rn>
+VFPU cond 1 1 0 0 0 1 0 1 Rn Rt 1 0 1 1 0 0 M 1 Vm VMOV<c> <Rt>, <Rn>, <Dm>
+;INVALIDATE <4> 1 1 0 0 0 1 0 <9> 1 0 1 <9>
+
+INVALIDATE <4> 1 1 0 0 0 <11> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 0 0 <11> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 1 1 <1> 1 <9> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 1 1 <1> 1 <9> 1 0 1 <9>
+VFPU cond 1 1 0 1 0 D 1 0 1 1 0 1 Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FSTMX<c> <LIst>
+VFPU cond 1 1 0 1 0 D 1 0 1 1 0 1 Vd 1 0 1 sz imm8 VPUSH<c> <list>
+VFPU cond 1 1 0 1 U D 0 0 Rn Vd 1 0 1 sz imm8 VSTR<c> <Dd>, [<Rn>{, #+/-<imm>}]
+VFPU cond 1 1 0 @<2> D W 0 Rn Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FSTM\(param == 0b01 ? "IA" : "DB")\X<c> <Rn>{!}, <LIst>
+VFPU cond 1 1 0 @<2> D W 0 Rn Vd 1 0 1 sz imm8 VSTM\(param == 0b01 ? "IA" : "DB")\<c> <Rn>{!}, <list>
+VFPU cond 1 1 0 0 1 D 1 1 1 1 0 1 Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FLDMX<c> <LIst>
+VFPU cond 1 1 0 0 1 D 1 1 1 1 0 1 Vd 1 0 1 sz imm8 VPOP<c> <list>
+VFPU cond 1 1 0 1 0 D 0 1 1 1 1 1 Vd 1 0 1 sz 0 0 0 0 0 0 0 0 VLDR<c> <Dd>, [PC, #-0]
+VFPU cond 1 1 0 1 U D 0 1 1 1 1 1 Vd 1 0 1 sz imm8 VLDR<c> <Dd>, <label>
+VFPU cond 1 1 0 1 U D 0 1 Rn Vd 1 0 1 sz imm8 VLDR<c> <Dd>, [<Rn>{, #+/-<imm>}]
+VFPU cond 1 1 0 @<2> D W 1 Rn Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FLDM\(param == 0b01 ? "IA" : "DB")\X<c> <Rn>{!}, <LIst>
+VFPU cond 1 1 0 @<2> D W 1 Rn Vd 1 0 1 sz imm8 VLDM\(param == 0b01 ? "IA" : "DB")\<c> <Rn>{!}, <list>
+INVALIDATE <4> 1 1 0 <25>
+
+INVALIDATE <4> 1 1 1 0 0 0 <1> 0 <8> 1 0 1 1 <1> 1 0 1 (0) (0) (0) (0)
+;INVALIDATE <4> 1 1 1 0 1 <2> 0 <8> 1 0 1 1 <1> 0 <1> 1 (0) (0) (0) (0)
+VFPU cond 1 1 1 0 0 0 0 0 Vn Rt 1 0 1 0 N (0) (0) 1 (0) (0) (0) (0) VMOV<c> <Sn>, <Rt>
+VFPU cond 1 1 1 0 1 1 1 0 0 0 0 1 Rt 1 0 1 0 (0) (0) (0) 1 (0) (0) (0) (0) VMSR<c> FPSCR, <Rt>
+VFPU cond 1 1 1 0 0 @<2> 0 Vd Rt 1 0 1 1 D @<2> 1 (0) (0) (0) (0) @ set uint8_t opc = (param1_2 << 2) + param2_2@ = uint8_t shift @@ (opc & 0b1000) == 0b1000@ @@ (opc & 0b1001) == 0b0001@ @@ (opc & 0b1011) == 0b0000@ shift,0,1,2,0xFF@ set uint8_t decodedImm = (opc & 0x7) >> shift@ @ VMOV<c>.\%d8 &l&l shift\ <Dd[x]>, <Rt>
+VFPU cond 1 1 1 0 1 0 Q 0 Vd Rt 1 0 1 1 D 0 0 1 (0) (0) (0) (0) VDUP<c>.32 <Qd>, <Rt>
+VFPU cond 1 1 1 0 1 0 Q 0 Vd Rt 1 0 1 1 D 0 1 1 (0) (0) (0) (0) VDUP<c>.16 <Qd>, <Rt>
+VFPU cond 1 1 1 0 1 1 Q 0 Vd Rt 1 0 1 1 D 0 0 1 (0) (0) (0) (0) VDUP<c>.8 <Qd>, <Rt>
+VFPU cond 1 1 1 0 0 0 0 1 Vn Rt 1 0 1 0 N (0) (0) 1 (0) (0) (0) (0) VMOV<c> <Rt>, <Sn>
+VFPU cond 1 1 1 0 1 1 1 1 0 0 0 1 Rt 1 0 1 0 (0) (0) (0) 1 (0) (0) (0) (0) VMRS<c> <Rt>, FPSCR
+VFPU cond 1 1 1 0 U @<2> 1 Vn Rt 1 0 1 1 N @<2> 1 (0) (0) (0) (0) @ set uint8_t opc = (u << 4) + (param1_2 << 2) + param2_2@ = uint8_t shift, size @@ (opc & 0b01000) == 0b01000@ @@ (opc & 0b01001) == 0b00001@ @@ (opc & 0b11011) == 0b00000@ shift,0,1,2,0xFF;size,0,1,2,4@ set uint8_t decodedImm = (opc & 0x7) >> shift@ @ VMOV<c>.<dt> <Rt>, <Dn[x]>
+INVALIDATE <4> 1 1 1 0 <12> 1 0 1 <4> 1 <4>
+
+VFPU cond 1 1 1 0 0 D 0 0 Vn Vd 1 0 1 sz N op M 0 Vm VML\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 0 1 Vn Vd 1 0 1 sz N op M 0 Vm VNML\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 0 Vn Vd 1 0 1 sz N 0 M 0 Vm VMUL<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 0 Vn Vd 1 0 1 sz N 1 M 0 Vm VNMUL<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 1 Vn Vd 1 0 1 sz N 0 M 0 Vm VADD<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 1 Vn Vd 1 0 1 sz N 1 M 0 Vm VSUB<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 0 0 Vn Vd 1 0 1 sz N 0 M 0 Vm VDIV<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 0 1 Vn Vd 1 0 1 sz N op M 0 Vm VFNM\(op ? "A" : "S")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 0 Vn Vd 1 0 1 sz N op M 0 Vm VFM\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+
+!/VFPU cond 1 1 1 0 1 x 1 1 opc2 <4> 1 0 1 <1> - - <1> 0 opc4 Table A7-17
+VFPU cond 1 1 1 0 1 D 1 1 @<4> Vd 1 0 1 0 (0) 0 (0) 0 @<4> @ set int8_t sign = (param1_4 >> 3) ? -1 : 1@ set uint8_t exp = ((param1_4 & 0b111) ^ 0b100)@ set float mantissa = 16 + param2_4@ set float imm = sign * (1 << exp) * mantissa / 128.@ @ VMOV<c>.F32 <Sd>, #\%fimm\
+VFPU cond 1 1 1 0 1 D 1 1 @<4> Vd 1 0 1 1 (0) 0 (0) 0 @<4> @ set int8_t sign = (param1_4 >> 3) ? -1 : 1@ set uint8_t exp = ((param1_4 & 0b111) ^ 0b100)@ set double mantissa = 16 + param2_4@ set double imm = sign * (1 << exp) * mantissa / 128.@ @ VMOV<c>.F64 <Dd>, #\%fimm\
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 0 Vd 1 0 1 sz 0 1 M 0 Vm VMOV<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 0 Vd 1 0 1 sz 1 1 M 0 Vm VABS<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 1 Vd 1 0 1 sz 0 1 M 0 Vm VNEG<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 1 Vd 1 0 1 sz 1 1 M 0 Vm VSQRT<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 1 0 Vd 1 0 1 (0) @<1> 1 M 0 Vm VCVT\(param ? "T" : "B")\<c>.F32.F16 <Sd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 1 1 Vd 1 0 1 (0) @<1> 1 M 0 Vm VCVT\(param ? "T" : "B")\<c>.F16.F32 <Sd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 0 0 Vd 1 0 1 sz op 1 M 0 Vm VCMP\(op ? "E" : "")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 0 1 Vd 1 0 1 sz op 1 (0) 0 (0) (0) (0) (0) VCMP\(op ? "E" : "")\<c>.F\(size ? "64" : "32")\ <Dd>, #0.0
+VFPU cond 1 1 1 0 1 D 1 1 0 1 1 1 Vd 1 0 1 0 1 1 M 0 Vm VCVT<c>.F64.F32 <Dd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 1 1 Vd 1 0 1 1 1 1 M 0 Vm VCVT<c>.F32.F64 <Sd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 1 0 0 0 Vd 1 0 1 sz op 1 M 0 Vm VCVT<c>.F\(size ? "64" : "32")\.\(op ? "S" : "U")\32 <Dd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 1 0 1 U Vd 1 0 1 sz op 1 i 0 imm4 @ set uint8_t td = (u << 2) + op + 1@ td=0b001=0b010=0b101 const char* Td Td,"S16","U16","S32","U32"@ set imm5 = (16 << op) - (imm5 >> 4) - ((imm5 & 0xF) << 1)@ @ VCVT<c>.F\(size ? "64" : "32")\.\Td\ <Dd>, <Dd>, #<imm>
+VFPU cond 1 1 1 0 1 D 1 1 1 1 0 U Vd 1 0 1 sz op 1 M 0 Vm VCVT\(op ? "" : "R")\<c>.\(u ? "U" : "S")\32.F\(size ? "64" : "32")\ <Sd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 1 1 1 U Vd 1 0 1 sz op 1 i 0 imm4 @ set uint8_t td = (u << 2) + op + 1@ td=0b001=0b010=0b101 const char* Td Td,"S16","U16","S32","U32"@ set imm5 = (16 << op) - (imm5 >> 4) - ((imm5 & 0xF) << 1)@ @ VCVT<c>.\Td\.F\(size ? "64" : "32")\ <Dd>, <Dd>, #<imm>
+;INVALIDATE <4> 1 1 1 0 1 <1> 1 1 <8> 1 0 1 <4> 0 <4>
+;INVALIDATE <4> 1 1 1 0 <12> 1 0 1 <4> 0 <4>
+
+; ============= END OF REGISTERED INSTRUCTIONS ============
+INVALIDATE <32>
@@ -0,0 +1,108 @@
+// arm lock helper
+// there is 2 part: read and write
+// write return 0 on success, 1 on fail (value has been changed)
+
+.text
+.align 4
+
+.global arm_lock_read_b
+.global arm_lock_write_b
+.global arm_lock_read_h
+.global arm_lock_write_h
+.global arm_lock_read_d
+.global arm_lock_write_d
+.global arm_lock_read_dd
+.global arm_lock_write_dd
+.global arm_lock_xchg
+.global arm_lock_storeifnull
+.global arm_lock_storeifref
+
+
+arm_lock_read_b:
+    // address is r0, return is r0
+    ldrexb  r0, [r0]
+    bx      lr
+
+arm_lock_write_b:
+    // address is r0, value is r1, return is r0
+    mov     r2, r0
+    strexb  r0, r1, [r2]
+    bx      lr
+
+arm_lock_read_h:
+    // address is r0, return is r0
+    ldrexh  r0, [r0]
+    bx      lr
+
+arm_lock_write_h:
+    // address is r0, value is r1, return is r0
+    mov     r2, r0
+    strexh  r0, r1, [r2]
+    bx      lr
+
+arm_lock_read_d:
+    // address is r0, return is r0
+    // r0 needs to be aligned
+    ldrex   r0, [r0]
+    bx      lr
+
+arm_lock_write_d:
+    // address is r0, value is r1, return is r0
+    // r0 needs to be aligned
+    mov     r2, r0
+    strex   r0, r1, [r2]
+    bx      lr
+
+arm_lock_read_dd:
+    // address is r2, return is r0, r1
+    ldrexd  r2, r3, [r2]
+    str     r2, [r0]
+    str     r3, [r1]
+    bx      lr
+
+arm_lock_write_dd:
+    // address is r2, value is r0, r1, return is r0
+    // r0 needs to be aligned
+    strexd  r3, r0, r1, [r2]
+    mov     r0, r3
+    bx      lr
+
+arm_lock_xchg:
+    // address is r0, value is r1, return old value in r0
+    // r0 needs to be aligned
+    ldrex   r2, [r0]
+    strex   r3, r1, [r0]
+    cmp     r3, #1
+    beq     arm_lock_xchg
+    mov     r0, r2
+    bx      lr
+
+arm_lock_storeifnull:
+    // address is r0, value is r1, r1 store to r0 only if [r0] is 0. return new [r0] value (so r1 or old value)
+    // r0 needs to be aligned
+    ldrex   r2, [r0]
+    cmp     r2, #0
+    bne     arm_lock_storeifnull_exit
+    mov     r2, r1
+    strex   r3, r2, [r0]
+    cmp     r3, #1
+    beq     arm_lock_storeifnull
+arm_lock_storeifnull_exit:
+    mov     r0, r2
+    bx      lr
+
+arm_lock_storeifref:
+    push    {r4}
+arm_lock_storeifref_loop:
+    // address is r0, value is r1, r1 store to r0 only if [r0] is r2. return new [r0] value (so r1 or old value)
+    ldrex   r3, [r0]
+    cmp     r3, r2
+    bne     arm_lock_storeifref_exit
+    mov     r3, r1
+    strex   r4, r3, [r0]
+    cmp     r4, #1
+    beq     arm_lock_storeifref_loop
+arm_lock_storeifref_exit:
+    mov     r0, r3
+    pop     {r4}
+    bx      lr
@@ -0,0 +1,34 @@
+#ifndef __ARM_LOCK_HELPER__H__
+#define __ARM_LOCK_HELPER__H__
+#include <stdint.h>
+
+// LDREXB of ADDR
+extern uint8_t arm_lock_read_b(void* addr);
+// STREXB of ADDR, return 0 if ok, 1 if not
+extern int arm_lock_write_b(void* addr, uint8_t val);
+
+// LDREXH of ADDR
+extern uint16_t arm_lock_read_h(void* addr);
+// STREXH of ADDR, return 0 if ok, 1 if not
+extern int arm_lock_write_h(void* addr, uint16_t val);
+
+// LDREX of ADDR
+extern uint32_t arm_lock_read_d(void* addr);
+// STREX of ADDR, return 0 if ok, 1 if not
+extern int arm_lock_write_d(void* addr, uint32_t val);
+
+// LDREXD of ADDR
+extern void arm_lock_read_dd(uint32_t* a, uint32_t* b, void* addr);
+// STREX of ADDR, return 0 if ok, 1 if not
+extern int arm_lock_write_dd(uint32_t a, uint32_t b, void* addr);
+
+// Atomically exchange value at [p] with val, return old p
+extern uintptr_t arm_lock_xchg(void* p, uintptr_t val);
+
+// Atomically store value to [p] only if [p] is NULL. Return new [p] value (so val or old)
+extern void* arm_lock_storeifnull(void* p, void* val);
+
+// Atomically store value to [p] only if [p] is ref. Return new [p] value (so val or old)
+extern void* arm_lock_storeifref(void* p, void* val, void* ref);
+
+#endif  //__ARM_LOCK_HELPER__H__
@@ -6,19 +6,18 @@
 .text
 .align 4

-.extern UpdateLinkTable
+.extern LinkNext

-.global arm_linker
-arm_linker:
+.global arm_next
+arm_next:
    // emu is r0
-    // table offset is r1
-    // IP address is r12
-    push    {r0, r1, r12}
-    mov     r2, r12
+    //stm     r0, {r4-r12,r14}   // don't put put back reg value in emu, faster but more tricky to debug
+    // IP address is r1
+    push    {r0, r1, r12, r14}
    // call the function
-    bl      UpdateLinkTable
+    bl      LinkNext
    // return offset is jump address
    mov     r3, r0
-    pop     {r0, r1, r12}
+    pop     {r0, r1, r12, r14}
    bx      r3

@@ -10,10 +10,10 @@
 arm_prolog:
    //save all used register
    push     {r4-r11, lr}
-    //vpush     {d8-d15}
+    vpush     {d8-d15}
    //setup emu -> register
    //push     {r0}
-    ldm      r0, {r4-r11}    // all 8 register in direct access, no EIP (so r12 can be used as scratch)
+    ldm      r0, {r4-r12}    // all 8 register in direct access, plus flags, no EIP (so r14 can be used as scratch)
    //TODO
    //jump to function
    bx       r1
@@ -1,19 +0,0 @@
-//arm try to update linker table for dynarec
-//and address of table to as 1st parameter
-//jump addres as second and ip as third
-// don't retry if failed to write...
-
-.text
-.align 4
-
-.global arm_tableupdate
-arm_tableupdate:
-    // jump address is r0 and IP address is r1
-    // table offset is r2
-    push    {r4, r5}
-    ldrexd  r4, r5, [r2]
-    strexd  r3, r0, r1, [r2]
-    pop     {r4, r5}
-    mov     r0, r3
-    bx      lr
-
@@ -3,7 +3,6 @@
 #include <pthread.h>
 #include <errno.h>

-
 #include "debug.h"
 #include "box86context.h"
 #include "dynarec.h"
@@ -17,163 +16,409 @@
 #include "x86trace.h"
 #include "dynablock.h"
 #include "dynablock_private.h"
+#include "dynarec_private.h"
 #include "elfloader.h"
 #ifdef ARM
 #include "dynarec_arm.h"
+#include "arm_lock_helper.h"
 #else
 #error Unsupported architecture!
 #endif
-
+#include "custommem.h"
 #include "khash.h"

 KHASH_MAP_INIT_INT(dynablocks, dynablock_t*)

-
-dynablocklist_t* NewDynablockList(uintptr_t base, uintptr_t text, int textsz)
+uint32_t X31_hash_code(void* addr, int len)
 {
+    if(!len) return 0;
+    uint8_t* p = (uint8_t*)addr;
+	int32_t h = *p;
+	for (--len, ++p; len; --len, ++p) h = (h << 5) - h + (int32_t)*p;
+	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->blocks = kh_init(dynablocks);
-    ret->base = base;
    ret->text = text;
    ret->textsz = textsz;
-    pthread_rwlock_init(&ret->rwlock_blocks, NULL);
-
+    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)
+{
+    if(db) {
+        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x86=%p:%p father=%p, with %d son(s) already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size, 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->x86_addr;
+            if(addr>=startdb && addr<enddb)
+                db->parent->direct[addr-startdb] = NULL;
+        }
+        // remove jumptable
+        setJumpTableDefault(db->x86_addr);
+        // remove and free the sons
+        for (int i=0; i<db->sons_size; ++i) {
+            dynablock_t *son = (dynablock_t*)arm_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);
+    }
+}
+
 void FreeDynablockList(dynablocklist_t** dynablocks)
 {
    if(!dynablocks)
        return;
    if(!*dynablocks)
        return;
-    dynarec_log(LOG_INFO, "Free %d Blocks from Dynablocklist (with %d buckets)%s\n", kh_size((*dynablocks)->blocks), kh_n_buckets((*dynablocks)->blocks), ((*dynablocks)->direct)?" With Direct mapping enabled":"");
-    dynablock_t* db;
-    kh_foreach_value((*dynablocks)->blocks, db,
-        free(db->table);
-        free(db);
-    );
-    kh_destroy(dynablocks, (*dynablocks)->blocks);
-    if((*dynablocks)->direct)
+    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 = 0;
-
-    pthread_rwlock_destroy(&(*dynablocks)->rwlock_blocks);
+    }
+    (*dynablocks)->direct = NULL;

    free(*dynablocks);
    *dynablocks = NULL;
 }

-void ConvertHash2Direct(dynablocklist_t* dynablocks)
+void MarkDynablock(dynablock_t* db)
 {
-    if(dynablocks->textsz==0 || dynablocks->text==0)
-        return; // nothing to do
-    // create the new set
-    dynablock_t **direct = (dynablock_t**)calloc(dynablocks->textsz, sizeof(dynablock_t*));
-    kh_dynablocks_t *blocks = kh_init(dynablocks);
-    // transfert
-    int ret;
-    khint_t k;
+    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)
+        setJumpTableDefault(db->x86_addr);
+        for(int i=0; i<db->sons_size; ++i)
+            setJumpTableDefault(db->sons[i]->x86_addr);
+    }
+}
+
+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)
+        return 0;
+    return 1;
+}
+
+void MarkDirectDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+{
+    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->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1))
+                    MarkDynablock(db);
+}
+
+void FreeRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+{
+    if(!dynablocks)
+        return;
+
+    if(dynablocks->direct) {
+        dynablock_t* db;
+        int ret;
+        khint_t k;
+        kh_dynablocks_t *blocks = kh_init(dynablocks);
+        // copy in a temporary list
+        if(dynablocks->direct) {
+            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*)arm_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);
+    }
+}
+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(int 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)
+{
+    if(!dynablocks)
+        return NULL;
    dynablock_t* db;
-    uintptr_t key;
-    uintptr_t start = dynablocks->text-dynablocks->base;
-    uintptr_t end = dynablocks->text + dynablocks->textsz-dynablocks->base;
-    kh_foreach(dynablocks->blocks, key, db,
-        if(key>=start && key<end)
-            direct[key-start] = db;
-        else {
-            k = kh_put(dynablocks, blocks, key, &ret);
-            if(ret) {   // don't try to insert if already done...
-                kh_value(blocks, k) = db;
-            }
+    kh_foreach_value(dynablocks, db, 
+        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 NULL;
+}
+
+static dynablocklist_t* getDBFromAddress(uintptr_t addr)
+{
+    const int 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;
+        return NULL;
+    }
+    if((addr<dynablocks->text) || (addr>=(dynablocks->text+dynablocks->textsz))) {
+        // this should be useless
+        //dynarec_log(LOG_INFO, "Warning: Refused to create a Direct Block that is out-of-bound: dynablocks=%p (%p:%p), addr=%p\n", dynablocks, (void*)(dynablocks->text), (void*)(dynablocks->text+dynablocks->textsz), (void*)addr);
+        //*created = 0;
+        //return NULL;
+        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_DUMP, "Block already exist in Direct Map\n");
+            *created = 0;
+            return block;
        }
-    );
-    // destroy old and do the swap (should swap before destroy, but hash access is always behind mutex)
-    kh_destroy(dynablocks, dynablocks->blocks);
-    dynablocks->blocks = blocks;
-    dynablocks->direct = direct;
+    }
+    
+    if (!*created)
+        return block;
+    
+    if(!dynablocks->direct) {
+        dynablock_t** p = (dynablock_t**)calloc(dynablocks->textsz, sizeof(dynablock_t*));
+        if(arm_lock_storeifnull(&dynablocks->direct, p)!=p)
+            free(p);    // someone already create the direct array, too late...
+    }
+
+    // create and add new block
+    dynarec_log(LOG_DUMP, "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*)arm_lock_storeifnull(&dynablocks->direct[addr-dynablocks->text], block);
+    if(tmp !=  block) {
+        // a block appeard!
+        free(block);
+        *created = 0;
+        return tmp;
+    }
+
+    *created = 1;
+    return block;
 }

-#define MAGIC_SIZE 256
 /* 
    return NULL if block is not found / cannot be created. 
    Don't create if create==0
 */
-dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t* current)
+static dynablock_t* internalDBGetBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, dynablock_t* current)
 {
    // 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->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz) {
-            block = dynablocks->direct[addr-dynablocks->text];
-            if(block)
-                return block;
-        }
-        if(!(addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz))
+        dynablocks = current->parent;
+        if(dynablocks && !(addr>=dynablocks->text && addr<(dynablocks->text+dynablocks->textsz)))
            dynablocks = NULL;
    }
    // nope, lets do the long way
-    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)
-        block = dynablocks->direct[addr-dynablocks->text];
-    if(block)
-        return block;
-    // nope, put rwlock in read mode and check hash
-    pthread_rwlock_rdlock(&dynablocks->rwlock_blocks);
-    // but first, check again just in case it has been created while waiting for mutex
-    if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz)
-        block = dynablocks->direct[addr-dynablocks->text];
-    if(block) {
-        pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
-        return block;
-    }
-    // check if the block exist
-    int ret;
-    khint_t k = kh_get(dynablocks, dynablocks->blocks, addr-dynablocks->base);
-    if(k!=kh_end(dynablocks->blocks)) {
-        /*atomic_store(&dynalock, 0);*/
-        block = kh_value(dynablocks->blocks, k);
-        pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
-        return block;
-    }
-    // unlock now
-    pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
-    // Blocks doesn't exist. If creation is not allow, just return NULL
-    if(!create)
-        return block;
-    // Lock as write now!
-    pthread_rwlock_wrlock(&dynablocks->rwlock_blocks);
-    // create and add new block
-    dynarec_log(LOG_DEBUG, "Ask for DynaRec Block creation @%p\n", addr);
-    if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz)
-    {
-        block = dynablocks->direct[addr-dynablocks->text] = (dynablock_t*)calloc(1, sizeof(dynablock_t));
-    } else {
-        k = kh_put(dynablocks, dynablocks->blocks, addr-dynablocks->base, &ret);
-        if(!ret) {
-            // Ooops, block is already there? retreive it and bail out
-            block = kh_value(dynablocks->blocks, k);    
-            pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
-            return block;
+    if(!dynablocks) {
+        dynablocks = getDBFromAddress(addr);
+        if(!dynablocks) {
+            dynablocks = GetDynablocksFromAddress(emu->context, addr);
+            if(!dynablocks)
+                return NULL;
        }
-        block = kh_value(dynablocks->blocks, k) = (dynablock_t*)calloc(1, sizeof(dynablock_t));
-        // check if size of hash map == magic size, if yes, convert to direct before unlocking
-        if(!dynablocks->direct && kh_size(dynablocks->blocks)==MAGIC_SIZE)
-            ConvertHash2Direct(dynablocks);
    }
-    block->parent = dynablocks;
-    // create an empty block first, so if other thread want to execute the same block, they can, but using interpretor path
-    pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+    // 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;

+    int created = create;
+    block = AddNewDynablock(dynablocks, addr, &created);
+    if(!created)
+        return block;   // existing block...
+
+    #if 0
+    if(box86_dynarec_dump)
+        pthread_mutex_lock(&my_context->mutex_dyndump);
+    #endif
    // fill the block
-    FillBlock(emu, block, addr);
-    dynarec_log(LOG_DEBUG, " --- DynaRec Block created @%p (%p, 0x%x bytes)\n", addr, block->block, block->size);
+    block->x86_addr = (void*)addr;
+    pthread_mutex_lock(&my_context->mutex_dyndump);
+    void* ret = FillBlock(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*)arm_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 been written
+        }
+        free(block);
+        block = NULL;
+    }
+    #if 0
+    if(box86_dynarec_dump)
+        pthread_mutex_unlock(&my_context->mutex_dyndump);
+    #endif
+    // check size
+    if(block && block->x86_size) {
+        int blocksz = block->x86_size;
+        if(dynablocks->maxsz<blocksz) {
+            dynablocks->maxsz = blocksz;
+            for(int 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;
+            }
+        }
+        protectDB((uintptr_t)block->x86_addr, block->x86_size);
+        // fill-in jumptable
+        addJumpTableIfDefault(block->x86_addr, block->block);
+        for(int i=0; i<block->sons_size; ++i)
+            addJumpTableIfDefault(block->sons[i]->x86_addr, block->sons[i]->block);
+    }
+
+    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->x86_size:0)), (block)?block->block:0, (block)?block->size:0, (block)?block->sons_size:0);

    return block;
 }
+
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t** current)
+{
+    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->x86_addr, father->x86_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->x86_addr, father->x86_addr+father->x86_size, hash, father->hash, father->sons_size, (void*)addr);
+            // no more current if it gets invalidated too
+            if(*current && father->x86_addr>=(*current)->x86_addr && (father->x86_addr+father->x86_size)<(*current)->x86_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;
+            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            // fill back jumptable
+            addJumpTableIfDefault(father->x86_addr, father->block);
+            for(int i=0; i<father->sons_size; ++i)
+                addJumpTableIfDefault(father->sons[i]->x86_addr, father->sons[i]->block);
+        }
+    } 
+    return db;
+}
+
+dynablock_t* DBAlternateBlock(x86emu_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->x86_addr, father->x86_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->x86_addr, father->x86_addr+father->x86_size, hash, father->hash, father->sons_size, (void*)addr);
+            // Free father, it's now invalid!
+            FreeDynablock(father);
+            // start again... (will create a new block)
+            db = internalDBGetBlock(emu, addr, filladdr, create, NULL);
+        } else {
+            father->need_test = 0;
+            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            // fill back jumptable
+            addJumpTableIfDefault(father->x86_addr, father->block);
+            for(int i=0; i<father->sons_size; ++i)
+                addJumpTableIfDefault(father->sons[i]->x86_addr, father->sons[i]->block);
+        }
+    } 
+    return db;
+}
@@ -1,26 +1,36 @@
 #ifndef __DYNABLOCK_PRIVATE_H_
 #define __DYNABLOCK_PRIVATE_H_

-typedef struct dynablocklist_s dynablocklist_t;
+typedef struct dynablocklist_s  dynablocklist_t;
+
+typedef struct instsize_s {
+    unsigned int x86:4;
+    unsigned int nat:4;
+} instsize_t;

 typedef struct dynablock_s {
-    dynablocklist_t *parent;
-    void*       block;
-    int         size;
-    uintptr_t*  table;
-    int         tablesz;
-    int         done;
+    dynablocklist_t* parent;
+    void*           block;
+    int             size;
+    void*           x86_addr;
+    uintptr_t       x86_size;
+    uint32_t        hash;
+    uint8_t         need_test;
+    uint8_t         done;
+    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;
 } dynablock_t;

-typedef struct kh_dynablocks_s kh_dynablocks_t;
-
 typedef struct dynablocklist_s {
-    kh_dynablocks_t     *blocks;
-    pthread_rwlock_t    rwlock_blocks;
-    uintptr_t           base;
    uintptr_t           text;
    int                 textsz;
-    dynablock_t         **direct;    // direct mapping (waste of space, so not always there)
+    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_
@@ -2,6 +2,7 @@
 #include <stdlib.h>
 #include <pthread.h>
 #include <errno.h>
+#include <setjmp.h>

 #include "debug.h"
 #include "box86context.h"
@@ -14,69 +15,77 @@
 #include "callback.h"
 #include "emu/x86run_private.h"
 #include "x86trace.h"
+#include "threads.h"
 #ifdef DYNAREC
 #include "dynablock.h"
 #include "dynablock_private.h"
 #endif

 #ifdef ARM
-void arm_prolog(x86emu_t* emu, void* addr);
-void arm_epilog();
-void arm_linker();
-int arm_tableupdate(void* jump, uintptr_t addr, void** table);
+void arm_prolog(x86emu_t* emu, void* addr) EXPORTDYN;
+void arm_epilog() EXPORTDYN;
+void arm_epilog_fast() EXPORTDYN;
 #endif

 #ifdef DYNAREC
-void* UpdateLinkTable(x86emu_t* emu, void** table, uintptr_t addr)
+#ifdef HAVE_TRACE
+uintptr_t getX86Address(dynablock_t* db, uintptr_t arm_addr);
+#endif
+void* LinkNext(x86emu_t* emu, uintptr_t addr, void* x2)
 {
-    dynablock_t* block = DBGetBlock(emu, addr, 1, NULL);    // keep a copy of parent block?
-    int r;
-    if(block==0) {
-        // no block, don't try again, ever
-        #ifdef ARM
-        r = arm_tableupdate(arm_epilog, addr, table);
-        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
-        #else
-        table[0] = arm_epilog;
-        table[1] = addr;
+    #ifdef HAVE_TRACE
+    if(!addr) {
+        x2-=8;  // actual PC is 2 instructions ahead
+        dynablock_t* db = FindDynablockFromNativeAddress(x2);
+        printf_log(LOG_NONE, "Warning, jumping to NULL address from %p (db=%p, x86addr=%p)\n", x2, db, db?(void*)getX86Address(db, (uintptr_t)x2):NULL);
+    }
+    #endif
+    dynablock_t* current = NULL;
+    void * jblock;
+    dynablock_t* block = DBGetBlock(emu, addr, 1, &current);
+    if(!block) {
+        // no block, let link table as is...
+        #ifdef HAVE_TRACE
+        dynablock_t* db = FindDynablockFromNativeAddress(x2);
+        printf_log(LOG_INFO, "Warning, jumping to a no-block address from %p (db=%p, x86addr=%p)\n", x2, db, db?(void*)getX86Address(db, (uintptr_t)x2):NULL);
        #endif
+        //tableupdate(arm_epilog, addr, table);
        return arm_epilog;
    }
    if(!block->done) {
        // not finished yet... leave linker
-        #ifdef ARM
-        r = arm_tableupdate(arm_linker, addr, table);
-        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
-        #else
-        table[0] = arm_linker;
-        table[1] = addr;
-        #endif
+        //tableupdate(arm_linker, addr, table);
        return arm_epilog;
    }
-    if(!block->block) {
-        // null block, but done, go to epilog
-        #ifdef ARM
-        r = arm_tableupdate(arm_epilog, addr, table);
-        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
-        #else
-        table[0] = arm_epilog;
-        table[1] = addr;
-        #endif
+    if(!(jblock=block->block)) {
+        // null block, but done: go to epilog, no linker here
        return arm_epilog;
    }
-    #ifdef ARM
-    r = arm_tableupdate(block->block, addr, table);
-    if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
-    #else
-    table[0] = block->block;
-    table[1] = addr;
-    #endif
-    return block->block;
+    //dynablock_t *father = block->father?block->father:block;
+    return jblock;
 }
 #endif

 void DynaCall(x86emu_t* emu, uintptr_t addr)
 {
+    // prepare setjump for signal handling
+    emu_jmpbuf_t *ejb = NULL;
+    int jmpbuf_reset = 0;
+    if(emu->type == EMUTYPE_MAIN) {
+        ejb = GetJmpBuf();
+        if(!ejb->jmpbuf_ok) {
+            ejb->emu = emu;
+            ejb->jmpbuf_ok = 1;
+            jmpbuf_reset = 1;
+            int a;
+            if((a=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
+                printf_log(LOG_DEBUG, "Setjmp DynaCall %d, fs=0x%x\n", a, ejb->emu->segs[_FS]);
+                addr = R_EIP;   // not sure if it should still be inside DynaCall!
+                if(a==2)
+                    Run(emu, 1);    // "single step" next instruction that is doing auto-smc
+            }
+        }
+    }
 #ifdef DYNAREC
    if(!box86_dynarec)
 #endif
@@ -93,15 +102,17 @@ void DynaCall(x86emu_t* emu, uintptr_t addr)
        R_EIP = addr;
        emu->df = d_none;
        dynablock_t* block = NULL;
+        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_EIP, 1, block);
+            block = DBGetBlock(emu, R_EIP, 1, &current);
+            current = block;
            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, "Calling Interpretor @%p, emu=%p\n", R_EIP, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Calling Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "Calling DynaRec Block @%p (%p) emu=%p\n", R_EIP, block->block, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Calling DynaRec Block @%p (%p) of %d x86 instructions (father=%p) emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize ,block->father, emu);
                CHECK_FLAGS(emu);
                // block is here, let's run it!
                #ifdef ARM
@@ -113,6 +124,16 @@ void DynaCall(x86emu_t* emu, uintptr_t addr)
                emu->quit = 0;
                emu->fork = 0;
                emu = x86emu_fork(emu, forktype);
+                if(emu->type == EMUTYPE_MAIN) {
+                    ejb = GetJmpBuf();
+                    ejb->emu = emu;
+                    ejb->jmpbuf_ok = 1;
+                    jmpbuf_reset = 1;
+                    if(sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1)) {
+                        printf_log(LOG_DEBUG, "Setjmp inner DynaCall, fs=0x%x\n", ejb->emu->segs[_FS]);
+                        addr = R_EIP;
+                    }
+                }
            }
        }
        emu->quit = 0;  // reset Quit flags...
@@ -129,10 +150,35 @@ void DynaCall(x86emu_t* emu, uintptr_t addr)
        }
    }
 #endif
+    // clear the setjmp
+    if(ejb && jmpbuf_reset)
+        ejb->jmpbuf_ok = 0;
 }

 int DynaRun(x86emu_t* emu)
 {
+    // prepare setjump for signal handling
+    emu_jmpbuf_t *ejb = NULL;
+#ifdef DYNAREC
+    int jmpbuf_reset = 1;
+#endif
+    if(emu->type == EMUTYPE_MAIN) {
+        ejb = GetJmpBuf();
+        if(!ejb->jmpbuf_ok) {
+            ejb->emu = emu;
+            ejb->jmpbuf_ok = 1;
+            int a;
+#ifdef DYNAREC
+            jmpbuf_reset = 1;
+#endif
+            if((a=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
+                printf_log(LOG_DEBUG, "Setjmp DynaRun %d, fs=0x%x\n", a, ejb->emu->segs[_FS]);
+                if(a==2)
+                    Run(emu, 1);    // "single step" next instruction that is doing auto-smc
+
+            }
+        }
+    }
 #ifdef DYNAREC
    if(!box86_dynarec)
 #endif
@@ -140,15 +186,17 @@ int DynaRun(x86emu_t* emu)
 #ifdef DYNAREC
    else {
        dynablock_t* block = NULL;
+        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_EIP, 1, block);
+            block = DBGetBlock(emu, R_EIP, 1, &current);
+            current = block;
            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, "Running Interpretor @%p, emu=%p\n", R_EIP, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Running Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "Running DynaRec Block @%p (%p) emu=%p\n", R_EIP, block->block, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x86 insts (father=%p) emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize, block->father, emu);
                // block is here, let's run it!
                #ifdef ARM
                arm_prolog(emu, block->block);
@@ -159,9 +207,20 @@ int DynaRun(x86emu_t* emu)
                emu->quit = 0;
                emu->fork = 0;
                emu = x86emu_fork(emu, forktype);
+                if(emu->type == EMUTYPE_MAIN) {
+                    ejb = GetJmpBuf();
+                    ejb->emu = emu;
+                    ejb->jmpbuf_ok = 1;
+                    jmpbuf_reset = 1;
+                    if(sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))
+                        printf_log(LOG_DEBUG, "Setjmp inner DynaRun, fs=0x%x\n", ejb->emu->segs[_FS]);
+                }
            }
        }
    }
+    // clear the setjmp
+    if(ejb && jmpbuf_reset)
+        ejb->jmpbuf_ok = 0;
    return 0;
 #endif
 }
@@ -2,9 +2,11 @@
 #include <stdlib.h>
 #include <pthread.h>
 #include <errno.h>
+#include <string.h>

 #include "debug.h"
 #include "box86context.h"
+#include "custommem.h"
 #include "dynarec.h"
 #include "emu/x86emu_private.h"
 #include "tools/bridge_private.h"
@@ -18,51 +20,325 @@
 #include "dynablock_private.h"
 #include "dynarec_arm.h"
 #include "dynarec_arm_private.h"
+#include "dynarec_arm_functions.h"
+#include "elfloader.h"

 void printf_x86_instruction(zydis_dec_t* dec, instruction_x86_t* inst, const char* name) {
    uint8_t *ip = (uint8_t*)inst->addr;
    if(ip[0]==0xcc && ip[1]=='S' && ip[2]=='C') {
        uint32_t a = *(uint32_t*)(ip+3);
        if(a==0) {
-            dynarec_log(LOG_NONE, "%p: Exit x86emu\n", (void*)ip);
+            dynarec_log(LOG_NONE, "%s%p: Exit x86emu%s\n", (box86_dynarec_dump>1)?"\e[1m":"", (void*)ip, (box86_dynarec_dump>1)?"\e[m":"");
        } else {
-            dynarec_log(LOG_NONE, "%p: Native call to %p\n", (void*)ip, (void*)a);
+            dynarec_log(LOG_NONE, "%s%p: Native call to %p%s\n", (box86_dynarec_dump>1)?"\e[1m":"", (void*)ip, (void*)a, (box86_dynarec_dump>1)?"\e[m":"");
        }
    } else {
        if(dec) {
-            dynarec_log(LOG_NONE, "%p: %s\n", inst->addr, DecodeX86Trace(dec, inst->addr));
+            dynarec_log(LOG_NONE, "%s%p: %s", (box86_dynarec_dump>1)?"\e[1m":"", ip, DecodeX86Trace(dec, inst->addr));
        } else {
-            dynarec_log(LOG_NONE, "%p: ", inst->addr);
+            dynarec_log(LOG_NONE, "%s%p: ", (box86_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\n", name);
+            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", (box86_dynarec_dump>1)?"\e[m":"");
    }
 }

-void arm_pass0(dynarec_arm_t* dyn, uintptr_t addr);
-void arm_pass1(dynarec_arm_t* dyn, uintptr_t addr);
-void arm_pass2(dynarec_arm_t* dyn, uintptr_t addr);
-void arm_pass3(dynarec_arm_t* dyn, uintptr_t addr);
+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;
+}

-void FillBlock(x86emu_t* emu, dynablock_t* block, uintptr_t addr) {
+// 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].x86.use_flags;
+    if(needed) {
+        setf &= ~needed;
+        if(!setf)   // all flags already used, no need to continue
+            return needed;
+    }
+
+    if(!needed && !dyn->insts[ninst].x86.set_flags && !dyn->insts[ninst].x86.jmp_insts) {
+        int start = ninst;
+        int end = ninst;
+        while(end<dyn->size && !dyn->insts[end].x86.use_flags && !dyn->insts[end].x86.set_flags && !dyn->insts[end].x86.jmp_insts)
+            ++end;
+        needed = needed_flags(dyn, end, setf, recurse);
+        for(int i=start; i<end; ++i)
+            dyn->insts[i].x86.need_flags = needed;
+        return needed;
+    }
+
+    if(dyn->insts[ninst].x86.set_flags && (dyn->insts[ninst].x86.state_flags!=SF_MAYSET)) {
+        if((setf & ~dyn->insts[ninst].x86.set_flags) == 0)
+            return needed;    // all done, gives all the flags needed
+        setf |= dyn->insts[ninst].x86.set_flags;    // add new flags to continue
+    }
+
+    int jinst = dyn->insts[ninst].x86.jmp_insts;
+    if(dyn->insts[ninst].x86.jmp) {
+        dyn->insts[ninst].x86.need_flags = (jinst==-1)?X_PEND:needed_flags(dyn, jinst, setf, recurse+1);
+        if(dyn->insts[ninst].x86.use_flags)  // conditionnal jump
+             dyn->insts[ninst].x86.need_flags |= needed_flags(dyn, ninst+1, setf, recurse);
+    } else
+        dyn->insts[ninst].x86.need_flags = needed_flags(dyn, ninst+1, setf, recurse);
+    if(dyn->insts[ninst].x86.state_flags==SF_MAYSET)
+        needed |= dyn->insts[ninst].x86.need_flags;
+    else
+        needed |= (dyn->insts[ninst].x86.need_flags & ~dyn->insts[ninst].x86.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 x86_size, int arm_size)
+{
+    // x86 instruction is <16 bytes
+    int toadd;
+    if(x86_size>arm_size)
+        toadd = 1 + x86_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(x86_size>15)
+            insts[*size].x86 = 15;    
+        else
+            insts[*size].x86 = x86_size;
+        x86_size -= insts[*size].x86;
+        if(arm_size>15)
+            insts[*size].nat = 15;
+        else
+            insts[*size].nat = arm_size;
+        arm_size -= insts[*size].nat;
+        ++(*size);
+        --toadd;
+    }
+    return insts;
+}
+
+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* FillBlock(dynablock_t* block, uintptr_t addr) {
+dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
+    if(addr>=box86_nodynarec_start && addr<box86_nodynarec_end) {
+        dynarec_log(LOG_DEBUG, "Asked to fill a block in fobidden zone\n");
+        return NULL;
+    }
+    if(!isJumpTableDefault((void*)addr)) {
+        dynarec_log(LOG_DEBUG, "Asked to fill a block at %p, but JumpTable is not default\n", (void*)addr);
+        return NULL;
+    }
    // init the helper
    dynarec_arm_t helper = {0};
-    helper.emu = emu;
-    arm_pass0(&helper, addr);
+    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;
+        free(helper.next);
+        return (void*)block;
    }
    helper.cap = helper.size+3; // needs epilog handling
    helper.insts = (instruction_arm_t*)calloc(helper.cap, sizeof(instruction_arm_t));
+    // already protect the block and compute hash signature
+    protectDB(addr, end-addr+1);
+    uint32_t hash = X31_hash_code((void*)addr, end-addr+1);
    // pass 1, addresses, x86 jump addresses, flags
    arm_pass1(&helper, addr);
    // calculate barriers
-    uintptr_t start = helper.insts[0].x86.addr;
-    uintptr_t end = helper.insts[helper.size].x86.addr+helper.insts[helper.size].x86.size;
    for(int i=0; i<helper.size; ++i)
        if(helper.insts[i].x86.jmp) {
            uintptr_t j = helper.insts[i].x86.jmp;
@@ -80,44 +356,111 @@ void FillBlock(x86emu_t* emu, dynablock_t* block, uintptr_t addr) {
                helper.insts[i].x86.jmp_insts = k;
            }
        }
-    // remove useless flags calulation
    for(int i=0; i<helper.size; ++i)
-        if(helper.insts[i].x86.flags==X86_FLAGS_CHANGE) {
-            int done = 0;
-            for(int i2=i+1; i2<helper.size+1 && done==0; ++i2) {
-                if(helper.insts[i2].x86.barrier || helper.insts[i2].x86.jmp)
-                    done = 1;
-                else if(helper.insts[i2].x86.flags==X86_FLAGS_USE)
-                    done = 1;
-                else if(helper.insts[i2].x86.flags==X86_FLAGS_CHANGE) {
-                    done = 1;
-                    helper.insts[i].x86.flags=X86_FLAGS_NONE;
-                }
-            }
+        if(helper.insts[i].x86.set_flags && !helper.insts[i].x86.need_flags) {
+            helper.insts[i].x86.need_flags = needed_flags(&helper, i+1, helper.insts[i].x86.set_flags, 0);
+            if((helper.insts[i].x86.need_flags&X_PEND) && (helper.insts[i].x86.state_flags==SF_MAYSET))
+                helper.insts[i].x86.need_flags = X_ALL;
        }
+    
    // pass 2, instruction size
    arm_pass2(&helper, addr);
    // ok, now allocate mapped memory, with executable flag on
    int sz = helper.arm_size;
-    void* p = (void*)AllocDynarecMap(emu->context, sz);
+    void* p = (void*)AllocDynarecMap(block, sz);
    if(p==NULL) {
+        dynarec_log(LOG_DEBUG, "AllocDynarecMap(%p, %d) failed, cancelling block\n", block, sz);
        free(helper.insts);
-        return;
+        free(helper.next);
+        return NULL;
    }
    helper.block = p;
-    helper.tablesz = helper.tablei;
-    if(helper.tablesz)
-        helper.table = (uintptr_t*)calloc(helper.tablesz, sizeof(uintptr_t));
+    helper.arm_start = (uintptr_t)p;
+    if(helper.sons_size) {
+        helper.sons_x86 = (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)
-    dynarec_log(LOG_DEBUG, "Emitting %d bytes for %d x86 bytes\n", helper.arm_size, helper.isize);
+    if(box86_dynarec_dump) {
+        dynarec_log(LOG_NONE, "%s%04d|Emitting %d bytes for %d x86 bytes", (box86_dynarec_dump>1)?"\e[01;36m":"", GetTID(), helper.arm_size, helper.isize); 
+        printFunctionAddr(helper.start, " => ");
+        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");
+    }
    helper.arm_size = 0;
    arm_pass3(&helper, addr);
+    if(sz!=helper.arm_size) {
+        printf_log(LOG_NONE, "BOX86: Warning, size difference in block between pass2 (%d) & pass3 (%d)!\n", sz, helper.arm_size);
+        uint8_t *dump = (uint8_t*)helper.start;
+        printf_log(LOG_NONE, "Dump of %d x86 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].x86.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, " ------------\n");
+    }
    // all done...
-    __builtin___clear_cache(p, p+helper.arm_size);   // need to clear the cache before execution...
+    __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].x86.size>helper.insts[i].size)?helper.insts[i].x86.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].x86.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);
-    block->table = helper.table;
-    block->tablesz = helper.tablesz;
+    free(helper.next);
    block->size = sz;
+    block->isize = helper.size;
    block->block = p;
+    block->need_test = 0;
+    //block->x86_addr = (void*)start;
+    block->x86_size = end-start+1;
+    if(box86_dynarec_largest<block->x86_size)
+        box86_dynarec_largest = block->x86_size;
+    block->hash = X31_hash_code(block->x86_addr, block->x86_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:%d)=%x/%x\n", block->x86_addr, block->x86_size, block->hash, hash);
+        free(helper.sons_x86);
+        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_x86[i], &created);
+            if(created) {    // avoid breaking a working block!
+                son->block = helper.sons_arm[i];
+                son->x86_addr = (void*)helper.sons_x86[i];
+                son->x86_size = end-helper.sons_x86[i]+1;
+                if(!son->x86_size) {printf_log(LOG_NONE, "Warning, son with null x86 size! (@%p / ARM=%p)", son->x86_addr, son->block);}
+                son->father = block;
+                son->done = 1;
+                sons[sons_size++] = son;
+                if(!son->parent)
+                    son->parent = block->parent;
+            }
+        }
+        if(sons_size) {
+            block->sons = sons;
+            block->sons_size = sons_size;
+        } else
+            free(sons);
+    }
+    free(helper.sons_x86);
+    free(helper.sons_arm);
    block->done = 1;
+    return (void*)block;
 }
@@ -0,0 +1,323 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_helper.h"
+
+uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    int32_t i32, j32;
+    uint32_t u32;
+    uint8_t gd, ed;
+    uint8_t wback;
+    int fixedaddress;
+
+    MAYUSE(j32);
+
+    switch(opcode) {
+
+        case 0x03:
+            INST_NAME("ADD Gd, FS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_add32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x2B:
+            INST_NAME("SUB Gd, FS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_sub32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x33:
+            INST_NAME("XOR Gd, FS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_xor32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x3B:
+            INST_NAME("CMP Gd, FS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_cmp32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x67:
+            // reduced EA size...
+            opcode = F8;
+            switch(opcode) {
+                case 0x89:
+                    INST_NAME("MOV FS:Ew16, Gw");
+                    nextop = F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    GETGD;  // don't need GETGW here
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        if(ed!=gd) {
+                            BFI(ed, gd, 0, 16);
+                        }
+                    } else {
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        STRH_REG(gd, x14, ed);
+                    }
+                    break;
+
+                case 0x8B:
+                    INST_NAME("MOV Gd, FS:Ew16");
+                    nextop=F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                        MOV_REG(gd, xEAX+(nextop&7));
+                    } else {                    // mem <= reg
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        LDR_REG_LSL_IMM5(gd, ed, x14, 0);
+                    }
+                    break;
+
+                case 0x8F:
+                    INST_NAME("POP FS:Ew16");
+                    nextop=F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    if((nextop&0xC0)==0xC0) {
+                        POP1(xEAX+(nextop&7));  // 67 ignored
+                    } else {
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                        POP1(x2);
+                        STR_REG_LSL_IMM5(x2, x1, x14, 0);
+                    }
+                    break;
+
+                case 0xA1:
+                    INST_NAME("MOV EAX, FS:Ow");
+                    grab_fsdata(dyn, addr, ninst, x1);
+                    u32 = F16;
+                    if(u32) {
+                        MOV32(x2, u32);
+                        ADD_REG_LSL_IMM5(x1, x1, x2, 0);
+                    }
+                    LDR_IMM9(xEAX, x1, 0);
+                    break;
+
+                case 0xA3:
+                    INST_NAME("MOV FS:Ow, EAX");
+                    grab_fsdata(dyn, addr, ninst, x1);
+                    u32 = F16;
+                    if(u32) {
+                        MOV32(x2, u32);
+                        ADD_REG_LSL_IMM5(x1, x1, x2, 0);
+                    }
+                    STR_IMM9(xEAX, x1, 0);
+                    break;
+
+                case 0xFF:
+                    nextop = F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    switch((nextop>>3)&7) {
+                        case 6: // Push Ed
+                            INST_NAME("PUSH FS:Ew");
+                            if((nextop&0xC0)==0xC0) {   // reg
+                                DEFAULT;
+                            } else {                    // mem <= i32
+                                addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                                LDR_REG_LSL_IMM5(x3, ed, x14, 0);
+                                PUSH1(x3);
+                            }
+                            break;
+                        default:
+                            DEFAULT;
+                    }
+                    break;
+                default:
+                    DEFAULT;
+            }
+            break;
+
+        case 0x89:
+            INST_NAME("MOV FS:Ed, Gd");
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop=F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {   // reg <= reg
+                MOV_REG(xEAX+(nextop&7), gd);
+            } else {                    // mem <= reg
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                STR_REG_LSL_IMM5(gd, ed, x14, 0);
+            }
+            break;
+
+        case 0x8B:
+            INST_NAME("MOV Gd, FS:Ed");
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop=F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {   // reg <= reg
+                MOV_REG(gd, xEAX+(nextop&7));
+            } else {                    // mem <= reg
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                LDR_REG_LSL_IMM5(gd, ed, x14, 0);
+            }
+            break;
+
+        case 0x8F:
+            INST_NAME("POP FS:Ed");
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                POP1(xEAX+(nextop&7));
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                POP1(x2);
+                STR_REG_LSL_IMM5(x2, ed, x14, 0);
+            }
+            break;
+
+        case 0xA0:
+            INST_NAME("MOV AL, FS:Id");
+            grab_fsdata(dyn, addr, ninst, x1);
+            i32 = F32S;
+            if(i32>-4096 && i32<4096) {
+                LDRB_IMM9(xEAX, x1, i32);
+            } else {
+                MOV32(x2, i32);
+                LDRB_REG_LSL_IMM5(xEAX, x1, x2, 0);
+            }
+            break;
+        case 0xA1:
+            INST_NAME("MOV EAX, FS:Id");
+            grab_fsdata(dyn, addr, ninst, x1);
+            i32 = F32S;
+            if(i32>-4096 && i32<4096) {
+                LDR_IMM9(xEAX, x1, i32);
+            } else {
+                MOV32(x2, i32);
+                LDR_REG_LSL_IMM5(xEAX, x1, x2, 0);
+            }
+            break;
+        case 0xA2:
+            INST_NAME("MOV FS:Od, AL");
+            grab_fsdata(dyn, addr, ninst, x1);
+            i32 = F32S;
+            if(i32>-4096 && i32<4096) {
+                STRB_IMM9(xEAX, x1, i32);
+            } else {
+                MOV32(x2, i32);
+                STRB_REG_LSL_IMM5(xEAX, x1, x2, 0);
+            }
+            break;
+        case 0xA3:
+            INST_NAME("MOV FS:Od, EAX");
+            grab_fsdata(dyn, addr, ninst, x1);
+            i32 = F32S;
+            if(i32>-4096 && i32<4096) {
+                STR_IMM9(xEAX, x1, i32);
+            } else {
+                MOV32(x2, i32);
+                STR_REG_LSL_IMM5(xEAX, x2, x1, 0);
+            }
+            break;
+
+        case 0xC7:
+            INST_NAME("MOV FS:Ed, Id");
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop=F8;
+            if((nextop&0xC0)==0xC0) {   // reg <= i32
+                i32 = F32S;
+                ed = xEAX+(nextop&7);
+                MOV32(ed, i32);
+            } else {                    // mem <= i32
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                i32 = F32S;
+                MOV32(x3, i32);
+                STR_REG_LSL_IMM5(x3, ed, x14, 0);
+            }
+            break;
+            
+        case 0xE9:
+        case 0xEB:
+            INST_NAME("(ignored) FS:");
+            --addr;
+            break;
+
+        case 0xFF:
+            nextop = F8;
+            grab_fsdata(dyn, addr, ninst, x14);
+            switch((nextop>>3)&7) {
+                case 2: // CALL Ed
+                    INST_NAME("CALL FS:Ed");
+                    PASS2IF(ninst && dyn->insts && dyn->insts[ninst-1].x86.set_flags, 1) {
+                        READFLAGS(X_PEND);          // that's suspicious
+                    } else {
+                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
+                    }
+                    if((nextop&0xC0)==0xC0) {   // reg
+                        MOV_REG(xEIP, xEAX+(nextop&7));
+                    } else {                    // mem
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        LDR_REG_LSL_IMM5(xEIP, ed, x14, 0);
+                    }
+                    BARRIER(1);
+                    BARRIER_NEXT(1);
+                    if(!dyn->insts || ninst!=dyn->size-1) {
+                    } else {
+                        *need_epilog = 0;
+                        *ok = 0;
+                    }
+                    MOV32(x2, addr);
+                    PUSH1(x2);
+                    jump_to_next(dyn, 0, xEIP, ninst);
+                    break;
+                case 6: // Push Ed
+                    INST_NAME("PUSH FS:Ed");
+                    if((nextop&0xC0)==0xC0) {   // reg
+                        DEFAULT;
+                    } else {                    // mem <= i32
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        LDR_REG_LSL_IMM5(x3, ed, x14, 0);
+                        PUSH1(x3);
+                    }
+                    break;
+                default:
+                    DEFAULT;
+            }
+            break;
+
+        default:
+            DEFAULT;
+    }
+    return addr;
+}
+
@@ -25,53 +25,293 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
 {
    uint8_t opcode = F8;
    uint8_t nextop;
-    int32_t i32;
+    int32_t i32, j32;
    uint32_t u32;
    uint8_t gd, ed;
    uint8_t wback, wb1, wb2;
+    uint8_t u8;
    int fixedaddress;
+
+    MAYUSE(j32);
+    MAYUSE(wback);
+    MAYUSE(wb2);
+    
    switch(opcode) {
-        case 0x33:
-            grab_tlsdata(dyn, addr, ninst, x12);
-            INST_NAME("GS:XOR Gd, Ed");
+
+        case 0x03:
+            INST_NAME("ADD Gd, GS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_tlsdata(dyn, addr, ninst, x14);
            nextop = F8;
            GETGD;
-            GETEDO(x12);
-            XOR_REG_LSL_IMM8(gd, gd, ed, 0);
-            UFLAG_RES(gd);
-            UFLAG_DF(1, d_xor32);
-            UFLAGS(0);
+            GETEDO(x14);
+            emit_add32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x2B:
+            INST_NAME("SUB Gd, GS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_tlsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_sub32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x33:
+            INST_NAME("XOR Gd, GS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_tlsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_xor32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x3B:
+            INST_NAME("CMP Gd, GS:Ed");
+            SETFLAGS(X_ALL, SF_SET);
+            grab_tlsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            emit_cmp32(dyn, ninst, gd, ed, x3, x14);
+            break;
+
+        case 0x69:
+            INST_NAME("IMUL Gd, GS:Ed, Id");
+            SETFLAGS(X_ALL, SF_PENDING);
+            grab_tlsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGD;
+            GETEDO(x14);
+            i32 = F32S;
+            MOV32(x14, i32);
+            UFLAG_IF {
+                SMULL(x3, gd, x14, ed);
+                UFLAG_OP1(x3);
+                UFLAG_RES(gd);
+                UFLAG_DF(x3, d_imul32);
+            } else {
+                MUL(gd, ed, x14);
+            }
+            break;
+
+        case 0x80:
+        case 0x82:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0: //ADD
+                    INST_NAME("ADD GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_add8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 1: //OR
+                    INST_NAME("OR GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_or8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 2: //ADC
+                    INST_NAME("ADC GS:Eb, Ib");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_adc8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 3: //SBB
+                    INST_NAME("SBB GS:Eb, Ib");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_sbb8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 4: //AND
+                    INST_NAME("AND GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_and8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 5: //SUB
+                    INST_NAME("SUB GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_sub8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 6: //XOR
+                    INST_NAME("XOR GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    emit_xor8c(dyn, ninst, x1, u8, x2, x14);
+                    EBBACK;
+                    break;
+                case 7: //CMP
+                    INST_NAME("CMP GS:Eb, Ib");
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x1);
+                    GETEBO(x1);
+                    u8 = F8;
+                    if(u8) {
+                        MOVW(x2, u8);
+                        emit_cmp8(dyn, ninst, x1, x2, x3, x14);
+                    } else {
+                        emit_cmp8_0(dyn, ninst, x1, x3, x14);
+                    }
+                    break;
+            }
+            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);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_add32c(dyn, ninst, ed, i32, x3, x14);
+                    WBACK;
+                    break;
+                case 1: //OR
+                    if(opcode==0x81) {INST_NAME("OR Ed, Id");} else {INST_NAME("OR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_or32c(dyn, ninst, ed, i32, x3, x14);
+                    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);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_adc32c(dyn, ninst, ed, i32, x3, x14);
+                    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);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_sbb32c(dyn, ninst, ed, i32, x3, x14);
+                    WBACK;
+                    break;
+                case 4: //AND
+                    if(opcode==0x81) {INST_NAME("AND Ed, Id");} else {INST_NAME("AND Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_and32c(dyn, ninst, ed, i32, x3, x14);
+                    WBACK;
+                    break;
+                case 5: //SUB
+                    if(opcode==0x81) {INST_NAME("SUB Ed, Id");} else {INST_NAME("SUB Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_sub32c(dyn, ninst, ed, i32, x3, x14);
+                    WBACK;
+                    break;
+                case 6: //XOR
+                    if(opcode==0x81) {INST_NAME("XOR Ed, Id");} else {INST_NAME("XOR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    emit_xor32c(dyn, ninst, ed, i32, x3, x14);
+                    WBACK;
+                    break;
+                case 7: //CMP
+                    if(opcode==0x81) {INST_NAME("CMP Ed, Id");} else {INST_NAME("CMP Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    grab_tlsdata(dyn, addr, ninst, x14);
+                    GETEDO2(x14);
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    if(i32) {
+                        MOV32(x2, i32);
+                        emit_cmp32(dyn, ninst, ed, x2, x3, x14);
+                    } else {
+                        emit_cmp32_0(dyn, ninst, ed, x3, x14);
+                    }
+                    break;
+            }
            break;

        case 0x89:
-            grab_tlsdata(dyn, addr, ninst, x12);
            INST_NAME("MOV GS:Ed, Gd");
+            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
            if((nextop&0xC0)==0xC0) {   // reg <= reg
                MOV_REG(xEAX+(nextop&7), gd);
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
-                STR_REG_LSL_IMM5(gd, ed, x12, 0);
+                STR_REG_LSL_IMM5(gd, ed, x14, 0);
            }
            break;

        case 0x8B:
-            grab_tlsdata(dyn, addr, ninst, x12);
            INST_NAME("MOV Gd, GS:Ed");
+            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
            if((nextop&0xC0)==0xC0) {   // reg <= reg
                MOV_REG(gd, xEAX+(nextop&7));
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
-                LDR_REG_LSL_IMM5(gd, ed, x12, 0);
+                LDR_REG_LSL_IMM5(gd, ed, x14, 0);
+            }
+            break;
+
+        case 0x8F:
+            INST_NAME("POP GS:Ed");
+            grab_tlsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                POP1((xEAX+(nextop&7)));
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                POP1(x2);
+                STR_REG_LSL_IMM5(x2, ed, x14, 0);
            }
            break;

        case 0xA1:
-            grab_tlsdata(dyn, addr, ninst, x1);
            INST_NAME("MOV EAX, GS:Id");
+            grab_tlsdata(dyn, addr, ninst, x1);
            i32 = F32S;
            if(i32>-4096 && i32<4096) {
                LDR_IMM9(xEAX, x1, i32);
@@ -83,8 +323,8 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;

        case 0xA3:
-            grab_tlsdata(dyn, addr, ninst, x1);
            INST_NAME("MOV GS:Od, EAX");
+            grab_tlsdata(dyn, addr, ninst, x1);
            u32 = F32;
            MOV32(x2, u32);
            ADD_REG_LSL_IMM5(x2, x1, x2, 0);
@@ -92,8 +332,8 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;

        case 0xC7:
-            grab_tlsdata(dyn, addr, ninst, x12);
            INST_NAME("MOV GS:Ed, Id");
+            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
            if((nextop&0xC0)==0xC0) {   // reg <= i32
                i32 = F32S;
@@ -103,12 +343,36 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
                i32 = F32S;
                MOV32(x3, i32);
-                STR_REG_LSL_IMM5(x3, ed, x12, 0);
+                STR_REG_LSL_IMM5(x3, ed, x14, 0);
+            }
+            break;
+
+        case 0xE9:
+        case 0xEB:
+            INST_NAME("(ignored) GS:");
+            --addr;
+            break;
+
+        case 0xFF:
+            nextop = F8;
+            grab_tlsdata(dyn, addr, ninst, x14);
+            switch((nextop>>3)&7) {
+                case 6: // Push Ed
+                    INST_NAME("PUSH GS:Ed");
+                    if((nextop&0xC0)==0xC0) {   // reg
+                        DEFAULT;
+                    } else {                    // mem <= i32
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        LDR_REG_LSL_IMM5(x3, ed, x14, 0);
+                        PUSH1(x3);
+                    }
+                    break;
+                default:
+                    DEFAULT;
            }
            break;

        default:
-            *ok = 0;
            DEFAULT;
    }
    return addr;
@@ -21,22 +21,117 @@

 #include "dynarec_arm_helper.h"

-
 uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
-    uint8_t opcode = F8;
-    uint8_t nextop, u8;
+    uint8_t nextop, opcode;
    int8_t  i8;
+    int32_t i32, j32;
    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
    int fixedaddress;
-    switch(opcode) {
+
+    MAYUSE(i32);
+    MAYUSE(j32);
+    
+    nextop = F8;
+
+    switch(nextop) {
       

+        case 0x64:
+            // reduced EA size...
+            opcode = F8;
+            switch(opcode) {
+                case 0x89:
+                    INST_NAME("MOV FS:Ew16, Gw");
+                    nextop = F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    GETGD;  // don't need GETGW here
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        if(ed!=gd) {
+                            BFI(ed, gd, 0, 16);
+                        }
+                    } else {
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        STRH_REG(gd, x14, ed);
+                    }
+                    break;
+
+                case 0x8B:
+                    INST_NAME("MOV Gd, FS:Ew16");
+                    nextop=F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                        ed = xEAX+(nextop&7);
+                        if(ed!=gd) {
+                            BFI(gd, ed, 0, 16);
+                        }
+                    } else {                    // mem <= reg
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                        LDR_REG_LSL_IMM5(gd, x14, ed, 0);
+                    }
+                    break;
+
+                case 0x8F:
+                    INST_NAME("POP FS:Ew16");
+                    nextop=F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    if((nextop&0xC0)==0xC0) {
+                        POP1(xEAX+(nextop&7));  // 67 ignored
+                    } else {
+                        addr = geted16(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                        POP1(x2);
+                        STR_REG_LSL_IMM5(x2, x14, ed, 0);
+                    }
+                    break;
+
+                case 0xA1:
+                    INST_NAME("MOV EAX, FS:Ow");
+                    grab_fsdata(dyn, addr, ninst, x1);
+                    u32 = F16;
+                    if(u32) {
+                        MOV32(x2, u32);
+                        ADD_REG_LSL_IMM5(x1, x1, x2, 0);
+                    }
+                    LDR_IMM9(xEAX, x1, 0);
+                    break;
+
+                case 0xA3:
+                    INST_NAME("MOV FS:Ow, EAX");
+                    grab_fsdata(dyn, addr, ninst, x1);
+                    u32 = F16;
+                    if(u32) {
+                        MOV32(x2, u32);
+                        ADD_REG_LSL_IMM5(x1, x1, x2, 0);
+                    }
+                    STR_IMM9(xEAX, x1, 0);
+                    break;
+
+                case 0xFF:
+                    nextop = F8;
+                    grab_fsdata(dyn, addr, ninst, x14);
+                    switch((nextop>>3)&7) {
+                        case 6: // Push Ed
+                            INST_NAME("PUSH FS:Ew");
+                            if((nextop&0xC0)==0xC0) {   // reg
+                                DEFAULT;
+                            } else {                    // mem <= i32
+                                addr = geted16(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                                LDR_REG_LSL_IMM5(x3, ed, x14, 0);
+                                PUSH1(x3);
+                            }
+                            break;
+                        default:
+                            DEFAULT;
+                    }
+                    break;
+                default:
+                    DEFAULT;
+            }
+            break;
+            
        #define GO(NO, YES)   \
            BARRIER(2); \
            JUMP(addr+i8);\
@@ -45,7 +140,7 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    /* out of the block */                  \
                    i32 = dyn->insts[ninst+1].address-(dyn->arm_size+8); \
                    Bcond(NO, i32);     \
-                    jump_to_linker(dyn, addr+i8, 0, ninst); \
+                    jump_to_next(dyn, addr+i8, 0, ninst); \
                } else {    \
                    /* inside the block */  \
                    i32 = dyn->insts[dyn->insts[ninst].x86.jmp_insts].address-(dyn->arm_size+8);    \
@@ -54,31 +149,28 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            }
        case 0xE0:
            INST_NAME("LOOPNZ (16bits)");
-            USEFLAG(1);
+            READFLAGS(X_ZF);
            i8 = F8S;
            UXTH(x1, xECX, 0);
            SUBS_IMM8(x1, x1, 1);
            BFI(xECX, x1, 0, 16);
            B_NEXT(cEQ);    // CX is 0, no LOOP
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_ZF);
            GO(cNE, cEQ);
            break;
        case 0xE1:
            INST_NAME("LOOPZ (16bits)");
-            USEFLAG(1);
+            READFLAGS(X_ZF);
            i8 = F8S;
            UXTH(x1, xECX, 0);
            SUBS_IMM8(x1, x1, 1);
            BFI(xECX, x1, 0, 16);
            B_NEXT(cEQ);    // CX is 0, no LOOP
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_ZF);
            GO(cEQ, cNE);
            break;
        case 0xE2:
            INST_NAME("LOOP (16bits)");
-            USEFLAG(1);
            i8 = F8S;
            UXTH(x1, xECX, 0);
            SUBS_IMM8(x1, x1, 1);
@@ -87,16 +179,14 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xE3:
            INST_NAME("JCXZ");
-            USEFLAG(1);
            i8 = F8S;
            UXTH(x1, xECX, 0);
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_REG_LSL_IMM5(x1, x1, 0);
            GO(cNE, cEQ);
            break;
        #undef GO

        default:
-            *ok = 0;
            DEFAULT;
    }
    return addr;
@@ -26,17 +26,19 @@
 uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    int v1, v2, v3;
-    int s0, s1, s2;
-    int d0, d1;
+    uint8_t ed;
+    uint8_t wback;
+    int v1, v2;
+    int s0;
+    int d1;
    int fixedaddress;
+    int parity;
+
+    MAYUSE(d1);
+    MAYUSE(s0);
+    MAYUSE(v1);
+    MAYUSE(v2);
+    MAYUSE(ed);

    switch(nextop) {
        case 0xC0:
@@ -92,7 +94,7 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOM(x1, x2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xE0:
        case 0xE1:
@@ -152,30 +154,48 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 0:
                    INST_NAME("FADD ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VADD_F64(v1, v1, d1);
                    break;
                case 1:
                    INST_NAME("FMUL ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VMUL_F64(v1, v1, d1);
                    break;
                case 2:
                    INST_NAME("FCOM ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VCMP_F64(v1, d1);
                    FCOM(x1, x2);
@@ -183,57 +203,86 @@ uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 3:
                    INST_NAME("FCOMP ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VCMP_F64(v1, d1);
                    FCOM(x1, x2);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 4:
                    INST_NAME("FSUB ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VSUB_F64(v1, v1, d1);
                    break;
                case 5:
                    INST_NAME("FSUBR ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VSUB_F64(v1, d1, v1);
                    break;
                case 6:
                    INST_NAME("FDIV ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VDIV_F64(v1, v1, d1);
                    break;
                case 7:
                    INST_NAME("FDIVR ST0, float[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    GETED;
                    s0 = fpu_get_scratch_single(dyn);
                    d1 = fpu_get_scratch_double(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_F32(d1, s0);
                    VDIV_F64(v1, d1, v1);
                    break;
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -23,28 +23,29 @@
 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"

-static double d_1   = 1.0;
-static double d_l2t = L2T;
-static double d_l2e = L2E;
-static double d_pi  = PI;
-static double d_lg2 = LG2;
-static double d_ln2 = LN2;
-static double d_0   = 0.0;
+static const double d_1   = 1.0;
+static const double d_l2t = L2T;
+static const double d_l2e = L2E;
+static const double d_pi  = PI;
+static const double d_lg2 = LG2;
+static const double d_ln2 = LN2;
+static const double d_0   = 0.0;

 uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
+    uint8_t ed;
+    uint8_t wback, wb1;
    int fixedaddress;
-    int v1, v2, v3;
-    int s0, s1, s2;
+    int v1, v2;
+    int s0;
    int i1, i2, i3;
+    int parity;
+
+    MAYUSE(s0);
+    MAYUSE(v2);
+    MAYUSE(v1);
+
    switch(nextop) {

        case 0xC0:
@@ -57,7 +58,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC7:
            INST_NAME("FLD STx");
            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            v2 = x87_do_push(dyn, ninst);
+            v2 = x87_do_push(dyn, ninst, x3);
            VMOV_64(v2, v1);
            break;

@@ -107,43 +108,43 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,

        case 0xE8:
            INST_NAME("FLD1");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_1));
            VLDR_64(v1, x2, 0);
            break;
        case 0xE9:
            INST_NAME("FLDL2T");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_l2t));
            VLDR_64(v1, x2, 0);
            break;
        case 0xEA:     
            INST_NAME("FLDL2E");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_l2e));
            VLDR_64(v1, x2, 0);
            break;
        case 0xEB:
            INST_NAME("FLDPI");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_pi));
            VLDR_64(v1, x2, 0);
            break;
        case 0xEC:
            INST_NAME("FLDLG2");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_lg2));
            VLDR_64(v1, x2, 0);
            break;
        case 0xED:
            INST_NAME("FLDLN2");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_ln2));
            VLDR_64(v1, x2, 0);
            break;
        case 0xEE:
            INST_NAME("FLDZ");
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x1);
            MOV32(x2, (&d_0));
            VLDR_64(v1, x2, 0);
            break;
@@ -181,13 +182,13 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fyl2x, -1, 0);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF2:
            INST_NAME("FTAN");
            x87_forget(dyn, ninst, x1, x2, 0);
            CALL(arm_ftan, -1, 0);
-            v1 = x87_do_push(dyn, ninst);
+            v1 = x87_do_push(dyn, ninst, x3);
            MOV32(x2, (&d_1));
            VLDR_64(v1, x2, 0);
            break;
@@ -196,7 +197,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fpatan, -1, 0);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF4:
            INST_NAME("FXTRACT");
@@ -204,6 +205,28 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fxtract, -1, 0);
            break;
+        case 0xF5:
+            INST_NAME("FPREM1");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fprem1, -1, 0);
+            break;
+        case 0xF6:
+            INST_NAME("FDECSTP");
+            fpu_purgecache(dyn, ninst, x1, x2, x3);
+            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
+            SUB_IMM8(x2, x2, 1);
+            AND_IMM8(x2, x2, 7);
+            STR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
+            break;
+        case 0xF7:
+            INST_NAME("FINCSTP");
+            fpu_purgecache(dyn, ninst, x1, x2, x3);
+            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
+            ADD_IMM8(x2, x2, 1);
+            AND_IMM8(x2, x2, 7);
+            STR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
+            break;
        case 0xF8:
            INST_NAME("FPREM");
            x87_forget(dyn, ninst, x1, x2, 0);
@@ -215,7 +238,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            x87_forget(dyn, ninst, x1, x2, 0);
            x87_forget(dyn, ninst, x1, x2, 1);
            CALL(arm_fyl2xp1, -1, 0);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xFB:
            INST_NAME("FSINCOS");
@@ -259,10 +282,6 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xE6:
        case 0xE7:
        case 0xEF:
-        case 0xF5:
-        case 0xF6:
-        case 0xF7:
-            *ok = 0;
            DEFAULT;
            break;
             
@@ -270,37 +289,59 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FLD ST0, float[ED]");
-                    v1 = x87_do_push(dyn, ninst);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    v1 = x87_do_push(dyn, ninst, x1);
                    s0 = fpu_get_scratch_single(dyn);
-                    // to avoid bus error
-                    //VLDR_32(s0, ed, 0);
-                    LDR_IMM9(x2, ed, fixedaddress);
-                    VMOVtoV(s0, x2);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                        LDR_IMM9(x2, ed, fixedaddress);
+                        VMOVtoV(s0, x2);
+                    }
                    VCVT_F64_F32(v1, s0);
                    break;
                case 2:
                    INST_NAME("FST float[ED], ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
                    s0 = fpu_get_scratch_single(dyn);
                    VCVT_F32_F64(s0, v1);
-                    // to avoid bus error...
-                    //VSTR_32(s0, ed, 0);
-                    VMOVfrV(x2, s0);
-                    STR_IMM9(x2, ed, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VSTR_32(s0, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                        VMOVfrV(x2, s0);
+                        STR_IMM9(x2, ed, fixedaddress);
+                    }
                    break;
                case 3:
                    INST_NAME("FSTP float[ED], ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
                    s0 = fpu_get_scratch_single(dyn);
                    VCVT_F32_F64(s0, v1);
-                    // to avoid bus error...
-                    //VSTR_32(s0, ed, 0);
-                    VMOVfrV(x2, s0);
-                    STR_IMM9(x2, ed, fixedaddress);
-                    x87_do_pop(dyn, ninst);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VSTR_32(s0, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                        VMOVfrV(x2, s0);
+                        STR_IMM9(x2, ed, fixedaddress);
+                    }
+                    x87_do_pop(dyn, ninst, x3);
+                    break;
+                case 4:
+                    INST_NAME("FLDENV Ed");
+                    fpu_purgecache(dyn, ninst, x1, x2, x3); // maybe only x87, not SSE?
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {
+                        MOV_REG(x1, ed);
+                    }
+                    MOVW(x2, 0);
+                    CALL(fpu_loadenv, -1, 0);
                    break;
                case 5:
                    INST_NAME("FLDCW Ew");
@@ -309,6 +350,16 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    UBFX(x1, x1, 10, 2);    // extract round
                    STR_IMM9(x1, xEmu, offsetof(x86emu_t, round));
                    break;
+                case 6:
+                    INST_NAME("FNSTENV Ed");
+                    fpu_purgecache(dyn, ninst, x1, x2, x3); // maybe only x87, not SSE?
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {
+                        MOV_REG(x1, ed);
+                    }
+                    MOVW(x2, 0);
+                    CALL(fpu_savenv, -1, 0);
+                    break;
                case 7:
                    INST_NAME("FNSTCW Ew");
                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0);
@@ -318,7 +369,6 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    EWBACK;
                    break;
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -26,17 +26,21 @@
 uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
+    int32_t j32;
+    uint8_t ed;
+    uint8_t wback;
    int v1, v2;
    int d0;
    int s0;
    int fixedaddress;
+
+    MAYUSE(s0);
+    MAYUSE(d0);
+    MAYUSE(v2);
+    MAYUSE(v1);
+    MAYUSE(ed);
+    MAYUSE(j32);
+
    switch(nextop) {
        case 0xC0:
        case 0xC1:
@@ -47,11 +51,10 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC6:
        case 0xC7:
            INST_NAME("FCMOVB ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_CF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_CF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_CF);
            VMOVcond_64(cNE, v1, v2);   // F_CF==1
            break;
        case 0xC8:
@@ -63,11 +66,10 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xCE:
        case 0xCF:
            INST_NAME("FCMOVE ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_ZF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_ZF);
            VMOVcond_64(cNE, v1, v2);   // F_ZF==1
            break;
        case 0xD0:
@@ -79,12 +81,10 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xD6:
        case 0xD7:
            INST_NAME("FCMOVBE ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_CF|X_ZF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, flags[F_CF]));
-            ORRS_REG_LSL_IMM8(x1, x1, x2, 0);
+            TSTS_IMM8(xFlags, (1<<F_CF)|(1<<F_ZF));
            VMOVcond_64(cNE, v1, v2);   // F_CF==1 | F_ZF==1
            break;
        case 0xD8:
@@ -96,21 +96,20 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xDE:
        case 0xDF:
            INST_NAME("FCMOVU ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_PF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_PF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_PF);
            VMOVcond_64(cNE, v1, v2);   // F_PF==1
            break;       
        case 0xE9:
-            INST_NAME("FUCOMPP ST0, STx");
+            INST_NAME("FUCOMPP ST0, ST1");
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 1);
            VCMP_F64(v1, v2);
            FCOM(x1, x2);
-            x87_do_pop(dyn, ninst);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xE4:
@@ -123,7 +122,6 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xF8:
        case 0xF9:
        case 0xFD:
-            *ok = 0;
            DEFAULT;
            break;

@@ -170,7 +168,7 @@ uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    VCVT_F64_S32(d0, s0);
                    VCMP_F64(v1, d0);
                    FCOM(x1, x2);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 4:
                    INST_NAME("FISUB ST0, Ed");
@@ -22,22 +22,22 @@
 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"

-static double maxint32 = (double)0x7fffffff;
-
 uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    int v1, v2, v3;
-    int s0, s1, s2;
-    int d0;
+    int32_t j32;
+    uint8_t ed;
+    uint8_t wback;
+    int v1, v2;
+    int s0;
    int fixedaddress;
+    int parity;
+
+    MAYUSE(s0);
+    MAYUSE(v1);
+    MAYUSE(v2);
+    MAYUSE(j32);

    switch(nextop) {
        case 0xC0:
@@ -49,11 +49,10 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC6:
        case 0xC7:
            INST_NAME("FCMOVNB ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_CF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_CF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_CF);
            VMOVcond_64(cEQ, v1, v2);   // F_CF==0
            break;
        case 0xC8:
@@ -65,11 +64,10 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xCE:
        case 0xCF:
            INST_NAME("FCMOVNE ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_ZF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_ZF);
            VMOVcond_64(cEQ, v1, v2);   // F_ZF==0
            break;
        case 0xD0:
@@ -81,13 +79,11 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xD6:
        case 0xD7:
            INST_NAME("FCMOVNBE ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_CF|X_ZF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
-            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, flags[F_CF]));
-            ORRS_REG_LSL_IMM8(x1, x1, x2, 0);
-            VMOVcond_64(cEQ, v1, v2);   // F_CF==0 | F_ZF==0
+            TSTS_IMM8(xFlags, (1<<F_CF)|(1<<F_ZF));
+            VMOVcond_64(cEQ, v1, v2);   // F_CF==0 & F_ZF==0
            break;
        case 0xD8:
        case 0xD9:
@@ -98,17 +94,19 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xDE:
        case 0xDF:
            INST_NAME("FCMOVNU ST0, STx");
-            USEFLAG(1);
+            READFLAGS(X_PF);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
-            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_PF]));
-            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            TSTS_IMM8(xFlags, 1<<F_PF);
            VMOVcond_64(cEQ, v1, v2);   // F_PF==0
            break;
+        case 0xE1:
+            INST_NAME("FDISI8087_NOP"); // so.. NOP?
+            break;
        case 0xE2:
            INST_NAME("FNCLEX");
-            MOVW(x1, 0);
            LDRH_IMM8(x2, xEmu, offsetof(x86emu_t, sw));
+            MOVW(x1, 0);
            BFI(x2, x1, 0, 8);  // IE .. PE, SF, ES
            BFI(x2, x1, 15, 1); // B
            STRH_IMM8(x2, xEmu, offsetof(x86emu_t, sw));
@@ -127,12 +125,11 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xEE:
        case 0xEF:
            INST_NAME("FUCOMI ST0, STx");
-            UFLAGS(0);
+            SETFLAGS(X_ALL, SF_SET);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOMI(x1, x2);
-            UFLAGS(1);
            break;
        case 0xF0:  
        case 0xF1:
@@ -143,21 +140,18 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xF6:
        case 0xF7:
            INST_NAME("FCOMI ST0, STx");
-            UFLAGS(0);
+            SETFLAGS(X_ALL, SF_SET);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOMI(x1, x2);
-            UFLAGS(1);
            break;

        case 0xE0:
-        case 0xE1:
        case 0xE4:
        case 0xE5:
        case 0xE6:
        case 0xE7:
-            *ok = 0;
            DEFAULT;
            break;

@@ -165,10 +159,16 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FILD ST0, Ed");
-                    v1 = x87_do_push(dyn, ninst);
-                    GETED;
+                    v1 = x87_do_push(dyn, ninst, x1);
                    s0 = fpu_get_scratch_single(dyn);
-                    VMOVtoV(s0, ed);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 1023, 0);
+                        VLDR_32(s0, ed, fixedaddress);
+                    } else {
+                        GETED;
+                        VMOVtoV(s0, ed);
+                    }
                    VCVT_F64_S32(v1, s0);
                    break;
                case 1:
@@ -183,8 +183,8 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    }
                    s0 = fpu_get_scratch_single(dyn);
                    MSR_nzcvq_0();
-                    VMRS(x12);   // Get FPCSR reg to clear exceptions flags
-                    ORR_IMM8(x3, x12, 0b001, 6); // enable exceptions
+                    VMRS(x14);   // Get FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x14, 0b001, 6); // enable exceptions
                    BIC_IMM8(x3, x3, 0b10011111, 0);
                    VMSR(x3);
                    VCVT_S32_F64(s0, v1);
@@ -193,13 +193,13 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
                    MOV_IMM_COND(cNE, ed, 0b10, 1);   // 0x80000000
                    WBACK;
-                    VMSR(x12);  // put back values
-                    x87_do_pop(dyn, ninst);
+                    VMSR(x14);  // put back values
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 2:
                    INST_NAME("FIST Ed, ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    u8 = x87_setround(dyn, ninst, x1, x2, x12); // x1 have the modified RPSCR reg
+                    u8 = x87_setround(dyn, ninst, x1, x2, x14); // x1 have the modified RPSCR reg
                    if((nextop&0xC0)==0xC0) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
@@ -224,7 +224,7 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 3:
                    INST_NAME("FISTP Ed, ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    u8 = x87_setround(dyn, ninst, x1, x2, x12); // x1 have the modified RPSCR reg
+                    u8 = x87_setround(dyn, ninst, x1, x2, x14); // x1 have the modified RPSCR reg
                    if((nextop&0xC0)==0xC0) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
@@ -244,17 +244,32 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
                    MOV_IMM_COND(cNE, ed, 0b10, 1);   // 0x80000000
                    WBACK;
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    x87_restoreround(dyn, ninst, u8);
                    break;
                case 5:
                    INST_NAME("FLD tbyte");
-                    x87_do_push_empty(dyn, ninst, x1);
                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
-                    if(ed!=x1) {
-                        MOV_REG(x1, ed);
+                    if(PK(0)==0xDB && ((PK(1)>>3)&7)==7) {
+                        // the FLD is immediatly followed by an FSTP
+                        LDR_IMM9(x2, ed, 0);
+                        LDR_IMM9(x3, ed, 4);
+                        LDRH_IMM8(x14, ed, 8);
+                        // no persistant scratch register, so unrool both instruction here...
+                        MESSAGE(LOG_DUMP, "\tHack: FSTP tbyte\n");
+                        nextop = F8;    //0xDB
+                        nextop = F8;    //modrm
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                        STR_IMM9(x2, ed, 0);
+                        STR_IMM9(x3, ed, 4);
+                        STRH_IMM8(x14, ed, 8);
+                    } else {
+                        if(ed!=x1) {
+                            MOV_REG(x1, ed);
+                        }
+                        x87_do_push_empty(dyn, ninst, x3);
+                        CALL(arm_fld, -1, 0);
                    }
-                    CALL(arm_fld, -1, 0);
                    break;
                case 7:
                    INST_NAME("FSTP tbyte");
@@ -264,10 +279,9 @@ uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        MOV_REG(x1, ed);
                    }
                    CALL(arm_fstp, -1, 0);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -26,17 +26,16 @@
 uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    int v1, v2, v3;
-    int s0, s1;
-    int d0, d1;
+    uint8_t wback;
+    int v1, v2;
+    int d1;
    int fixedaddress;
+    int parity;
+
+    MAYUSE(d1);
+    MAYUSE(v2);
+    MAYUSE(v1);
+
    switch(nextop) {
        case 0xC0:
        case 0xC1:
@@ -91,7 +90,7 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOM(x1, x2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xE0:
        case 0xE1:
@@ -150,68 +149,132 @@ uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 0:
                    INST_NAME("FADD ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VADD_F64(v1, v1, d1);
                    break;
                case 1:
                    INST_NAME("FMUL ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VMUL_F64(v1, v1, d1);
                    break;
                case 2:
                    INST_NAME("FCOM ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VCMP_F64(v1, d1);
                    FCOM(x1, x2);
                    break;
                case 3:
                    INST_NAME("FCOMP ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VCMP_F64(v1, d1);
                    FCOM(x1, x2);
-                    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);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VSUB_F64(v1, v1, d1);
                    break;
                case 5:
                    INST_NAME("FSUBR ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VSUB_F64(v1, d1, v1);
                    break;
                case 6:
                    INST_NAME("FDIV ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VDIV_F64(v1, v1, d1);
                    break;
                case 7:
                    INST_NAME("FDIVR ST0, double[ED]");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
                    d1 = fpu_get_scratch_double(dyn);
-                    VLDR_64(d1, wback, fixedaddress);
+                    parity = getedparity(dyn, ninst, addr, nextop, 2);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                        VLDR_64(d1, wback, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, wback, fixedaddress);
+                        LDR_IMM9(x3, wback, fixedaddress+4);
+                        VMOVtoV_D(d1, x2, x3);
+                    }
                    VDIV_F64(v1, d1, v1);
                    break;
            }
@@ -27,15 +27,13 @@
 uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
    int fixedaddress;
-    int v1, v2, v3;
+    int parity;
+    uint8_t ed;
+    int v1, v2;
+
+    MAYUSE(v2);
+    MAYUSE(v1);

    switch(nextop) {
        case 0xC0:
@@ -63,7 +61,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xD8:
            INST_NAME("FSTP ST0, ST0");
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xD9:
        case 0xDA:
@@ -76,7 +74,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VMOV_64(v2, v1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xE0:
@@ -106,7 +104,7 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOM(x1, x2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

        case 0xC8:
@@ -133,7 +131,6 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xFD:
        case 0xFE:
        case 0xFF:
-            *ok = 0;
            DEFAULT;
            break;

@@ -141,15 +138,22 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FLD double");
-                    v1 = x87_do_push(dyn, ninst);
+                    v1 = x87_do_push(dyn, ninst, x1);
                    #if 0
                    // can bus error...
                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
                    VLDR_64(v1, ed, fixedaddress);
                    #else
-                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 255, 0);
-                    LDRD_IMM8(x2, ed, fixedaddress);
-                    VMOVtoV_D(v1, x2, x3);
+		            parity = getedparity(dyn, ninst, addr, nextop, 3);
+		            if (parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                        VLDR_64(v1, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                        LDR_IMM9(x2, ed, fixedaddress);
+                        LDR_IMM9(x3, ed, fixedaddress + 4);
+                        VMOVtoV_D(v1, x2, x3);
+        		    }
                    #endif
                    break;
                case 1:
@@ -157,24 +161,64 @@ uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    x87_forget(dyn, ninst, x1, x2, 0);
                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
                    if(ed!=x1) {MOV_REG(x1, ed);}
-                    CALL(arm_fistp64, -1, 0);
-                    x87_do_pop(dyn, ninst);
+                    CALL(arm_fistt64, -1, 0);
+                    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, 1023, 3);
-                    VSTR_64(v1, ed, fixedaddress); // check alignment?
+                    parity = getedparity(dyn, ninst, addr, nextop, 3);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                        VSTR_64(v1, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                        VMOVfrV_D(x2, x3, v1);
+                        STR_IMM9(x2, ed, fixedaddress);
+                        STR_IMM9(x3, ed, fixedaddress+4);
+                    }
                    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, 1023, 3);
-                    VSTR_64(v1, ed, fixedaddress); // check alignment?
-                    x87_do_pop(dyn, ninst);
+                    parity = getedparity(dyn, ninst, addr, nextop, 3);
+                    if(parity) {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                        VSTR_64(v1, ed, fixedaddress);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                        VMOVfrV_D(x2, x3, v1);
+                        STR_IMM9(x2, ed, fixedaddress);
+                        STR_IMM9(x3, ed, fixedaddress+4);
+                    }
+                    x87_do_pop(dyn, ninst, x3);
+                    break;
+                case 4: 
+                    INST_NAME("FRSTOR m108byte");
+                    fpu_purgecache(dyn, ninst, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(arm_frstor, -1, 0);
+                    break;
+                case 6: 
+                    INST_NAME("FSAVE m108byte");
+                    fpu_purgecache(dyn, ninst, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(arm_fsave, -1, 0);
+                    break;
+                case 7:
+                    INST_NAME("FNSTSW m2byte");
+                    fpu_purgecache(dyn, ninst, x1, x2, x3);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x14, &fixedaddress, 0, 0);
+                    LDR_IMM9(x1, xEmu, offsetof(x86emu_t, top));
+                    MOVW(x2, offsetof(x86emu_t, sw));
+                    ADD_REG_LSL_IMM5(x2, xEmu, x2, 0);
+                    LDRH_IMM8(x3, x2, 0);
+                    BFI(x3, x1, 11, 3); // inject TOP at bit 11 (3 bits)
+                    STRH_IMM8(x3, ed, 0);   // store whole sw flags
                    break;
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -25,16 +25,10 @@
 uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    int v1, v2, v3;
-    int s0, s1, s2;
-    int fixedaddress;
+    int v1, v2;
+
+    MAYUSE(v2);
+    MAYUSE(v1);

    switch(nextop) {
        case 0xC0:
@@ -49,7 +43,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VADD_F64(v2, v2, v1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xC8:
        case 0xC9:
@@ -63,16 +57,32 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VMUL_F64(v2, v2, v1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
-        case 0xD9:  /* FCOMPP */
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCOMP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst, x3);
+            break;
+
+        case 0xD9: 
            INST_NAME("FCOMPP ST0, ST1");
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, 1);
            VCMP_F64(v1, v2);
            FCOM(x1, x2);
-            x87_do_pop(dyn, ninst);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xE0:
        case 0xE1:
@@ -86,7 +96,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VSUB_F64(v2, v1, v2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xE8:
        case 0xE9:
@@ -100,7 +110,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VSUB_F64(v2, v2, v1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF0:
        case 0xF1:
@@ -114,7 +124,7 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VDIV_F64(v2, v1, v2);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF8:
        case 0xF9:
@@ -128,17 +138,9 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VDIV_F64(v2, v2, v1);
-            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst, x3);
            break;

-        case 0xD0:
-        case 0xD1:
-        case 0xD2:
-        case 0xD3:
-        case 0xD4:
-        case 0xD5:
-        case 0xD6:
-        case 0xD7:
        case 0xD8:
        case 0xDA:
        case 0xDB:
@@ -146,7 +148,6 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xDD:
        case 0xDE:
        case 0xDF:
-            *ok = 0;
            DEFAULT;
            break;

@@ -154,7 +155,6 @@ uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        default:
            switch((nextop>>3)&7) {
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -28,23 +28,36 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
 {
    uint8_t nextop = F8;
    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    int v1, v2, v3;
-    int s0, s1;
-    int d0;
+    int32_t j32;
+    uint8_t wback;
+    uint8_t ed;
+    int v1, v2;
+    int s0;
    int fixedaddress;

+    MAYUSE(s0);
+    MAYUSE(v2);
+    MAYUSE(v1);
+    MAYUSE(j32);
+
    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FFREEP STx");
+            // not handling Tag...
+            x87_do_pop(dyn, ninst, x3);
+            break;
+
        case 0xE0:
            INST_NAME("FNSTSW AX");
            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
            LDRH_IMM8(x1, xEmu, offsetof(x86emu_t, sw));
-            AND_IMM8(x2, x2, 7);
            BFI(x1, x2, 11, 3); // inject top
            BFI(xEAX, x1, 0, 16);
            break;
@@ -57,13 +70,12 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xEE:
        case 0xEF:
            INST_NAME("FUCOMIP ST0, STx");
-            UFLAGS(0);
+            SETFLAGS(X_ALL, SF_SET);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOMI(x1, x2);
-            x87_do_pop(dyn, ninst);
-            UFLAGS(1);
+            x87_do_pop(dyn, ninst, x3);
            break;
        case 0xF0:
        case 0xF1:
@@ -74,23 +86,14 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xF6:
        case 0xF7:
            INST_NAME("FCOMIP ST0, STx");
-            UFLAGS(0);
+            SETFLAGS(X_ALL, SF_SET);
            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
            VCMP_F64(v1, v2);
            FCOMI(x1, x2);
-            x87_do_pop(dyn, ninst);
-            UFLAGS(1);
+            x87_do_pop(dyn, ninst, x3);
            break;

-        case 0xC0:
-        case 0xC1:
-        case 0xC2:
-        case 0xC3:
-        case 0xC4:
-        case 0xC5:
-        case 0xC6:
-        case 0xC7:
        case 0xC8:
        case 0xC9:
        case 0xCA:
@@ -130,7 +133,6 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xFD:
        case 0xFE:
        case 0xFF:
-            *ok = 0;
            DEFAULT;
            break;

@@ -138,7 +140,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FILD ST0, Ew");
-                    v1 = x87_do_push(dyn, ninst);
+                    v1 = x87_do_push(dyn, ninst, x1);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0);
                    LDRSH_IMM8(x1, wback, fixedaddress);
                    s0 = fpu_get_scratch_single(dyn);
@@ -148,60 +150,69 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 1:
                    INST_NAME("FISTTP Ew, ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
                    ed = x1;
                    s0 = fpu_get_scratch_single(dyn);
-                    VCVT_S32_F64(s0, v1);
+                    MSR_nzcvq_0();
+                    VMRS(x14);               // get fpscr
+                    // x1 already have FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x14, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
+                    VCVTR_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
                    VMOVfrV(ed, s0);
-                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
-                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
-                    MRS_aspr(x3);   // check Q flag
-                    TSTS_IMM8_ROR(x3, 0b10, 3);
-                    B_MARK(cEQ);    // not saturated, continue
-                    MOV32(ed, 0xffff8000);
-                    MARK;
-                    STRH_IMM8(ed, wback, fixedaddress);
-                    x87_do_pop(dyn, ninst);
-                    x87_restoreround(dyn, ninst, u8);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    SSAT_REG_LSL_IMM5_COND(cEQ, x3, 16, ed, 0);
+                    CMPS_REG_LSL_IMM5_COND(cEQ, ed, x3, 0);
+                    MOVW_COND(cNE, x3, 0x8000); // saturated
+                    STRH_IMM8(x3, wback, fixedaddress);
+                    x87_do_pop(dyn, ninst, x3);
+                    VMSR(x14);
                    break;
                case 2:
                    INST_NAME("FIST Ew, ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
                    ed = x1;
                    s0 = fpu_get_scratch_single(dyn);
+                    MSR_nzcvq_0();
+                    // x1 already have FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x1, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
                    VCVTR_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
                    VMOVfrV(ed, s0);
-                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
-                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
-                    MRS_aspr(x3);   // check Q flag
-                    TSTS_IMM8_ROR(x3, 0b10, 3);
-                    B_MARK(cEQ);    // not saturated, continue
-                    MOV32(ed, 0xffff8000);
-                    MARK;
-                    STRH_IMM8(ed, wback, fixedaddress);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    SSAT_REG_LSL_IMM5_COND(cEQ, x3, 16, ed, 0);
+                    CMPS_REG_LSL_IMM5_COND(cEQ, ed, x3, 0);
+                    MOVW_COND(cNE, x3, 0x8000); // saturated
+                    STRH_IMM8(x3, wback, fixedaddress);
                    x87_restoreround(dyn, ninst, u8);
                    break;
                case 3:
                    INST_NAME("FISTP Ew, ST0");
                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
-                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
                    ed = x1;
                    s0 = fpu_get_scratch_single(dyn);
+                    MSR_nzcvq_0();
+                    // x1 already have FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x1, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
                    VCVTR_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
                    VMOVfrV(ed, s0);
-                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
-                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
-                    MRS_aspr(x3);   // check Q flag
-                    TSTS_IMM8_ROR(x3, 0b10, 3);
-                    B_MARK(cEQ);    // not saturated, continue
-                    MOV32(ed, 0xffff8000);
-                    MARK;
-                    STRH_IMM8(ed, wback, fixedaddress);
-                    x87_do_pop(dyn, ninst);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    SSAT_REG_LSL_IMM5_COND(cEQ, x3, 16, ed, 0);
+                    CMPS_REG_LSL_IMM5_COND(cEQ, ed, x3, 0);
+                    MOVW_COND(cNE, x3, 0x8000); // saturated
+                    STRH_IMM8(x3, wback, fixedaddress);
+                    x87_do_pop(dyn, ninst, x3);
                    x87_restoreround(dyn, ninst, u8);
                    break;
                case 4:
@@ -224,7 +235,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
                    if(ed!=x1) {MOV_REG(x1, ed);}
                    CALL(fpu_fbst, -1, 0);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                case 7: // could be inlined for most thing, but is it usefull?
                    INST_NAME("FISTP i64, ST0");
@@ -232,10 +243,9 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
                    if(ed!=x1) {MOV_REG(x1, ed);}
                    CALL(arm_fistp64, -1, 0);
-                    x87_do_pop(dyn, ninst);
+                    x87_do_pop(dyn, ninst, x3);
                    break;
                default:
-                    *ok = 0;
                    DEFAULT;
            }
    }
@@ -0,0 +1,744 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+#include "../tools/bridge_private.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+// emit OR32 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch
+void emit_or32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_or32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        ORRS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        ORR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit OR32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_or32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_or32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            ORRS_IMM8(s1, s1, c, 0);
+        } else {
+            ORR_IMM8(s1, s1, c, 0);
+        }
+    } else {
+        IFX(X_PEND) {} else {MOV32(s3, c);}
+        IFX(X_ALL) {
+            ORRS_REG_LSL_IMM5(s1, s1, s3, 0);
+        } else {
+            ORR_REG_LSL_IMM5(s1, s1, s3, 0);
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR32 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch
+void emit_xor32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_xor32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        XORS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        XOR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_xor32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_xor32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            XORS_IMM8(s1, s1, c);
+        } else {
+            XOR_IMM8(s1, s1, c);
+        }
+    } else {
+        IFX(X_PEND) {} else {MOV32(s3, c);}
+        IFX(X_ALL) {
+            XORS_REG_LSL_IMM5(s1, s1, s3, 0);
+        } else {
+            XOR_REG_LSL_IMM5(s1, s1, s3, 0);
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND32 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch
+void emit_and32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_and32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        ANDS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        AND_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_and32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_and32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            ANDS_IMM8(s1, s1, c);
+        } else {
+            AND_IMM8(s1, s1, c);
+        }
+    } else {
+        if(((~c)>=0 && (~c)<256)) {
+            IFX(X_ALL) {
+                BICS_IMM8_ROR(s1, s1, ~c, 0);
+            } else {
+                BIC_IMM8(s1, s1, ~c, 0);
+            }
+        } else {
+            IFX(X_PEND) {} else {MOV32(s3, c);}
+            IFX(X_ALL) {
+                ANDS_REG_LSL_IMM5(s1, s1, s3, 0);
+            } else {
+                AND_REG_LSL_IMM5(s1, s1, s3, 0);
+            }
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit OR8 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_or8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_or8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        ORRS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        ORR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit OR8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_or8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c&0xff);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_or8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        ORRS_IMM8(s1, s1, c, 0);
+    } else {
+        ORR_IMM8(s1, s1, c, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR8 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_xor8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_xor8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        XORS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        XOR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_xor8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c&0xff);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_xor8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        XORS_IMM8(s1, s1, c);
+    } else {
+        XOR_IMM8(s1, s1, c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND8 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_and8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_and8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        ANDS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        AND_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_and8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOV32(s3, c&0xff);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_and8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        ANDS_IMM8(s1, s1, c);
+    } else {
+        AND_IMM8(s1, s1, c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 7);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+
+// emit OR16 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_or16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_or16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        ORRS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        ORR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit OR16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_or16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_or16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            ORRS_IMM8(s1, s1, c, 0);
+        } else {
+            ORR_IMM8(s1, s1, c, 0);
+        }
+    } else {
+        IFX(X_PEND) {} else {MOVW(s3, c);}
+        IFX(X_ALL) {
+            ORRS_REG_LSL_IMM5(s1, s1, s3, 0);
+        } else {
+            ORR_REG_LSL_IMM5(s1, s1, s3, 0);
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR16 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_xor16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_xor16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        XORS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        XOR_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit XOR16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_xor16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_xor16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            XORS_IMM8(s1, s1, c);
+        } else {
+            XOR_IMM8(s1, s1, c);
+        }
+    } else {
+        IFX(X_PEND) {} else {MOVW(s3, c);}
+        IFX(X_ALL) {
+            XORS_REG_LSL_IMM5(s1, s1, s3, 0);
+        } else {
+            XOR_REG_LSL_IMM5(s1, s1, s3, 0);
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND16 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s4 can be same as s2 (and so s2 destroyed)
+void emit_and16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_and16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s3);
+    }
+    IFX(X_ALL) {
+        ANDS_REG_LSL_IMM5(s1, s1, s2, 0);
+    } else {
+        AND_REG_LSL_IMM5(s1, s1, s2, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit AND16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_and16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_and16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c>=0 && c<256) {
+        IFX(X_ALL) {
+            ANDS_IMM8(s1, s1, c);
+        } else {
+            AND_IMM8(s1, s1, c);
+        }
+    } else {
+        IFX(X_PEND) {} else {MOVW(s3, c);}
+        IFX(X_ALL) {
+            ANDS_REG_LSL_IMM5(s1, s1, s3, 0);
+        } else {
+            AND_REG_LSL_IMM5(s1, s1, s3, 0);
+        }
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
@@ -0,0 +1,462 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+#include "../tools/bridge_private.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+// emit SHL32 instruction, from s1 , shift s2, store result in s1 using s3 and s4 as scratch. s3 can be same as s2
+void emit_shl32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    int32_t j32;
+    MAYUSE(j32);
+
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_shl32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(F_OF) {
+        CMPS_IMM8(s2, 0);
+        IFX(F_OF) {
+            BIC_IMM8_COND(cEQ, xFlags, xFlags, 0b10, 0x0b);
+        }
+        IFX(X_PEND) {
+            STR_IMM9_COND(cEQ, s1, xEmu, offsetof(x86emu_t, res));
+        }
+        B_NEXT(cEQ);
+    }
+    IFX(X_CF | X_OF) {
+        RSB_IMM8(s4, s2, 32);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, F_CF, 1);
+    }
+    IFX(X_ZF) {
+        MOVS_REG_LSL_REG(s1, s1, s2);
+    } else {
+        MOV_REG_LSL_REG(s1, s1, s2);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+        BIC_IMM8_COND(cNE, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 31);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s3==1
+            IFX(X_SF) {} else {MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 31);}
+            XOR_REG_LSL_IMM5_COND(cEQ, s4, s4, xFlags, 0);  // CF is set if OF is asked
+            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
+        // else
+            BFC_COND(cNE, xFlags, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHL32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shl32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_shl32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c==0) {
+        IFX(F_OF) {
+            BIC_IMM8(xFlags, xFlags, 0b10, 0x0b);
+        }
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    IFX(X_CF) {
+        MOV_REG_LSR_IMM5(s3, s1, 32-c);
+        BFI(xFlags, s3, F_CF, 1);
+    }
+    IFX(X_ZF|X_OF) {
+        MOVS_REG_LSL_IMM5(s1, s1, c);
+    } else {
+        MOV_REG_LSL_IMM5(s1, s1, c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+        BIC_IMM8_COND(cNE, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            IFX(X_SF) {} else {MOV_REG_LSR_IMM5(s3, s1, 31);}
+            XOR_IMM8_COND(cCS, s3, s3, 1);
+            BFI(xFlags, s3, F_OF, 1);
+        } else {
+            BFC(xFlags, F_OF, 1);
+        }
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHR32 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s2 can be same as s3
+void emit_shr32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    int32_t j32;
+    MAYUSE(j32);
+
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_shr32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_ALL) {
+        CMPS_IMM8(s2, 0); //if(!c)
+            IFX(X_PEND) {
+                STR_IMM9_COND(cEQ, s1, xEmu, offsetof(x86emu_t, res));
+            }
+            B_NEXT(cEQ);
+    }
+    IFX(X_ZF|X_CF) {
+        MOVS_REG_LSR_REG(s1, s1, s2);
+    } else {
+        MOV_REG_LSR_REG(s1, s1, s2);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_CF) {
+        ORR_IMM8_COND(cCS, xFlags, xFlags, 1<<F_CF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 31);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 30);
+            XOR_REG_LSR_IMM8_COND(cEQ, s4, s4, s4, 1);
+            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHR32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shr32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_shr32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    IFX(X_ZF|X_CF) {
+        MOVS_REG_LSR_IMM5(s1, s1, c);
+    } else {
+        MOV_REG_LSR_IMM5(s1, s1, c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_CF) {
+        ORR_IMM8_COND(cCS, xFlags, xFlags, 1<<F_CF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 30);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SAR32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_sar32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_sar32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    IFX(X_ZF|X_CF) {
+        MOVS_REG_ASR_IMM5(s1, s1, c);
+    } else {
+        MOV_REG_ASR_IMM5(s1, s1, c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_CF) {
+        ORR_IMM8_COND(cCS, xFlags, xFlags, 1<<F_CF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit ROL32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_rol32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_rol32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    MOV_REG_ROR_IMM5(s1, s1, 32-c);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF) {
+        BFI(xFlags, s1, F_CF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            ADD_REG_LSR_IMM5(s3, s1, s1, 31);
+            BFI(xFlags, s3, F_OF, 1);
+        }
+    }
+}
+
+// emit ROR32 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_ror32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_ror32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    MOV_REG_ROR_IMM5(s1, s1, c);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_CF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 30);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+}
+
+// emit SHRD32 instruction, from s1, fill s2 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shrd32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int s3, int s4)
+{
+    c&=0x1f;
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        // same flags calc as shr32
+        SET_DF(s4, d_shr32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    IFX(X_CF) {
+        MOVS_REG_LSR_IMM5(s1, s1, c);
+    } else {
+        MOV_REG_LSR_IMM5(s1, s1, c);
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    IFX(X_CF) {
+        ORR_IMM8_COND(cCS, xFlags, xFlags, 1<<F_CF, 0);
+    }
+    IFX(X_ZF) {
+        ORRS_REG_LSL_IMM5(s1, s1, s2, 32-c);
+    } else {
+        ORR_REG_LSL_IMM5(s1, s1, s2, 32-c);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 30);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+void emit_shld32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int s3, int s4)
+{
+    c&=0x1f;
+    IFX(X_PEND) {
+        MOVW(s3, c);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        // same flags computation as with shl32
+        SET_DF(s4, d_shl32);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c==0) {
+        IFX(F_OF) {
+            BFC(xFlags, F_OF, 1);
+        }
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    IFX(X_CF) {
+        MOV_REG_LSR_IMM5(s3, s1, 32-c);
+        BFI(xFlags, s3, F_CF, 1);
+    }
+    IFX(X_OF) {
+        MOVS_REG_LSL_IMM5(s1, s1, c);
+    } else {
+        MOV_REG_LSL_IMM5(s1, s1, c);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            UBFX(s3, s2, 0, 1);
+            XOR_IMM8_COND(cCS, s3, s3, 1);
+            BFI(xFlags, s3, F_OF, 1);
+        } else {
+            BFC(xFlags, F_OF, 1);
+        }
+    }
+    IFX(X_ZF) {
+        ORRS_REG_LSR_IMM5(s1, s1, s2, 32-c);
+    } else {
+        ORR_REG_LSR_IMM5(s1, s1, s2, 32-c);
+    }
+
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+        BIC_IMM8_COND(cNE, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s3, s1, 31);
+        BFI(xFlags, s3, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
@@ -0,0 +1,408 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+#include "../tools/bridge_private.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+// emit CMP32 instruction, from cmp s1 , s2, using s3 and s4 as scratch
+void emit_cmp32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_cmp32);
+    } else {
+        SET_DFNONE(s4);
+    }
+    SUBS_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 - s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_OF) {
+        ORR_IMM8_COND(cVS, xFlags, xFlags, 0b10, 0x0b);
+        BIC_IMM8_COND(cVC, xFlags, xFlags, 0b10, 0x0b);
+    }
+    IFX(X_CF) {
+        // reversed carry
+        ORR_IMM8_COND(cCC, xFlags, xFlags, 1<<F_CF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 31);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // and now the tricky ones (and mostly unused), PF and AF
+    IFX(X_AF) {
+        // bc = (res & (~d | s)) | (~d & s)
+        MVN_REG_LSL_IMM5(s4, s1, 0);        // s4 = ~d
+        ORR_REG_LSL_IMM5(s4, s4, s2, 0);    // s4 = ~d | s
+        AND_REG_LSL_IMM5(s4, s4, s3, 0);    // s4 = res & (~d | s)
+        BIC_REG_LSL_IMM5(s3, s2, s1, 0);    // loosing res... s3 = s & ~d
+        ORR_REG_LSL_IMM5(s3, s4, s3, 0);    // s3 = (res & (~d | s)) | (s & ~d)
+        IFX(X_AF) {
+            MOV_REG_LSR_IMM5(s4, s3, 3);
+            BFI(xFlags, s4, F_AF, 1);    // AF: bc & 0x08
+        }
+    }
+    IFX(X_PF) {
+        // PF: (((emu->x86emu_parity_tab[(res) / 32] >> ((res) % 32)) & 1) == 0)
+        IFX(X_CF|X_AF) {
+            SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        }
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
+
+// emit CMP32 instruction, from cmp s1 , 0, using s3 and s4 as scratch
+void emit_cmp32_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        MOVW(s4, 0);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s4, xEmu, offsetof(x86emu_t, op2));
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        SET_DF(s4, d_cmp32);
+    } else {
+        SET_DFNONE(s4);
+    }
+    SUBS_IMM8(s3, s1, 0);   // res = s1 - 0
+    // and now the tricky ones (and mostly unused), PF and AF
+    // bc = (res & (~d | s)) | (~d & s) => is 0 here...
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 31);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit CMP16 instruction, from cmp s1 , s2, using s3 and s4 as scratch
+void emit_cmp16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s3, d_cmp16);
+    } else {
+        SET_DFNONE(s3);
+    }
+    SUB_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 - s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        MOVW(s4, 0xffff);
+        TSTS_REG_LSL_IMM5(s3, s4, 0);
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+        BIC_IMM8_COND(cNE, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // bc = (res & (~d | s)) | (~d & s)
+    IFX(X_CF|X_AF|X_OF) {
+        MVN_REG_LSL_IMM5(s4, s1, 0);        // s4 = ~d
+        ORR_REG_LSL_IMM5(s4, s4, s2, 0);    // s4 = ~d | s
+        AND_REG_LSL_IMM5(s4, s4, s3, 0);    // s4 = res & (~d | s)
+        BIC_REG_LSL_IMM5(s3, s2, s1, 0);    // loosing res... s3 = s & ~d
+        ORR_REG_LSL_IMM5(s3, s4, s3, 0);    // s3 = (res & (~d | s)) | (s & ~d)
+        IFX(X_CF) {
+            MOV_REG_LSR_IMM5(s4, s3, 15);
+            BFI(xFlags, s4, F_CF, 1);    // CF : bc & 0x8000
+        }
+        IFX(X_AF) {
+            MOV_REG_LSR_IMM5(s4, s3, 3);
+            BFI(xFlags, s4, F_AF, 1);    // AF: bc & 0x08
+        }
+        IFX(X_OF) {
+            MOV_REG_LSR_IMM5(s4, s3, 14);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);    // OF: ((bc >> 14) ^ ((bc>>14)>>1)) & 1
+        }
+    }
+    IFX(X_PF) {
+        IFX(X_CF|X_AF|X_OF) {
+            SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        }
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
+
+// emit CMP16 instruction, from cmp s1 , #0, using s3 and s4 as scratch
+void emit_cmp16_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        MOVW(s3, 0);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        SET_DF(s3, d_cmp16);
+    } else {
+        SET_DFNONE(s3);
+    }
+    // bc = (res & (~d | s)) | (~d & s) = 0
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF) {
+        MOVW(s4, 0xffff);
+        TSTS_REG_LSL_IMM5(s1, s4, 0);
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+// emit CMP8 instruction, from cmp s1 , s2, using s3 and s4 as scratch
+void emit_cmp8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, op2));
+        SET_DF(s4, d_cmp8);
+    } else {
+        SET_DFNONE(s4);
+    }
+    SUB_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 - s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s3, 0xff);
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+        BIC_IMM8_COND(cNE, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // bc = (res & (~d | s)) | (~d & s)
+    IFX(X_CF|X_AF|X_OF) {
+        MVN_REG_LSL_IMM5(s4, s1, 0);        // s4 = ~d
+        ORR_REG_LSL_IMM5(s4, s4, s2, 0);    // s4 = ~d | s
+        AND_REG_LSL_IMM5(s4, s4, s3, 0);    // s4 = res & (~d | s)
+        BIC_REG_LSL_IMM5(s3, s2, s1, 0);    // loosing res... s3 = s & ~d
+        ORR_REG_LSL_IMM5(s3, s4, s3, 0);    // s3 = (res & (~d | s)) | (s & ~d)
+        IFX(X_CF) {
+            MOV_REG_LSR_IMM5(s4, s3, 7);
+            BFI(xFlags, s4, F_CF, 1);    // CF : bc & 0x80
+        }
+        IFX(X_AF) {
+            MOV_REG_LSR_IMM5(s4, s3, 3);
+            BFI(xFlags, s4, F_AF, 1);    // AF: bc & 0x08
+        }
+        IFX(X_OF) {
+            MOV_REG_LSR_IMM5(s4, s3, 6);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);    // OF: ((bc >> 6) ^ ((bc>>6)>>1)) & 1
+        }
+    }
+    IFX(X_PF) {
+        IFX(X_CF|X_AF|X_OF) {
+            SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        }
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        SUB_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
+// emit CMP8 instruction, from cmp s1 , 0, using s3 and s4 as scratch
+void emit_cmp8_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
+        MOVW(s4, 0);
+        STR_IMM9(s4, xEmu, offsetof(x86emu_t, op2));
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        SET_DF(s3, d_cmp8);
+    } else {
+        SET_DFNONE(s4);
+    }
+    // bc = (res & (~d | s)) | (~d & s) = 0
+    IFX(X_CF | X_AF | X_ZF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_CF)|(1<<F_AF)|(1<<F_ZF), 0);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s1, 0xff);
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit TEST32 instruction, from test s1 , s2, using s3 and s4 as scratch
+void emit_test32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        SET_DF(s3, d_tst32);
+    } else {
+        SET_DFNONE(s4);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    ANDS_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 & s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 31);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // PF: (((emu->x86emu_parity_tab[(res) / 32] >> ((res) % 32)) & 1) == 0)
+    IFX(X_PF) {
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        AND_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
+
+// emit TEST16 instruction, from test s1 , s2, using s3 and s4 as scratch
+void emit_test16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        SET_DF(s3, d_tst16);
+    } else {
+        SET_DFNONE(s4);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    ANDS_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 & s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // PF: (((emu->x86emu_parity_tab[(res) / 32] >> ((res) % 32)) & 1) == 0)
+    IFX(X_PF) {
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        AND_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
+
+// emit TEST8 instruction, from test s1 , s2, using s3 and s4 as scratch
+void emit_test8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
+{
+    IFX_PENDOR0 {
+        SET_DF(s3, d_tst8);
+    } else {
+        SET_DFNONE(s4);
+    }
+    IFX(X_OF) {
+        BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_ZF|X_CF) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF)|(1<<F_CF), 0);
+    }
+    ANDS_REG_LSL_IMM5(s3, s1, s2, 0);   // res = s1 & s2
+    IFX_PENDOR0 {
+        STR_IMM9(s3, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 1<<F_ZF, 0);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s3, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    // PF: (((emu->x86emu_parity_tab[(res) / 32] >> ((res) % 32)) & 1) == 0)
+    IFX(X_PF) {
+        AND_IMM8(s3, s3, 0xE0); // lsr 5 masking pre-applied
+        MOV32(s4, GetParityTab());
+        LDR_REG_LSR_IMM5(s4, s4, s3, 5-2);   // x/32 and then *4 because array is integer
+        AND_REG_LSL_IMM5(s3, s1, s2, 0);
+        AND_IMM8(s3, s3, 31);
+        MVN_REG_LSR_REG(s4, s4, s3);
+        BFI(xFlags, s4, F_PF, 1);
+    }
+}
@@ -24,21 +24,480 @@

 uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
-    uint8_t nextop = PK(0); // don't increment addr
-    uint8_t u8;
-    uint32_t u32;
-    int32_t i32;
-    int16_t i16;
-    uint16_t u16;
-    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
+    uint8_t nextop, opcode = F8;
+    uint8_t wback, wb1, wb2, gb1, gb2;
+    uint8_t ed, gd, u8;
    int fixedaddress;
-    int locked = 0;
-    switch(nextop) {
-        // generic case
+    int32_t i32, j32;
+    MAYUSE(i32);
+    MAYUSE(j32);
+    MAYUSE(gb1);
+    MAYUSE(gb2);
+    MAYUSE(wb1);
+    MAYUSE(wb2);
+    switch(opcode) {
+        case 0x00:
+            INST_NAME("LOCK ADD Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGB(x2);
+            if((nextop&0xC0)==0xC0) {
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_add8(dyn, ninst, x1, x2, x14, x3, 0);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREXB(x1, wback);
+                emit_add8(dyn, ninst, x1, x2, x14, x3, (wback==x3)?1:0);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x01:
+            INST_NAME("LOCK ADD Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_add32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_add32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x08:
+            INST_NAME("LOCK OR Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                GETGB(x2);
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_or8(dyn, ninst, x1, x2, x14, x3);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                GETGB(x2);
+                LDREXB(x1, wback);
+                emit_or8(dyn, ninst, x1, x2, x14, x2);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x09:
+            INST_NAME("LOCK OR Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_or32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_or32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x10:
+            INST_NAME("LOCK ADC Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGB(x2);
+            if((nextop&0xC0)==0xC0) {
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_adc8(dyn, ninst, x1, x2, x14, x3, 0);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREXB(x1, wback);
+                emit_adc8(dyn, ninst, x1, x2, x14, x3, (wback==x3)?1:0);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x11:
+            INST_NAME("LOCK ADC Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_adc32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_adc32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x18:
+            INST_NAME("LOCK SBB Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGB(x2);
+            if((nextop&0xC0)==0xC0) {
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_sbb8(dyn, ninst, x1, x2, x14, x3, 0);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREXB(x1, wback);
+                emit_sbb8(dyn, ninst, x1, x2, x14, x3, (wback==x3)?1:0);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x19:
+            INST_NAME("LOCK SBB Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_sbb32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_sbb32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x20:
+            INST_NAME("LOCK AND Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                GETGB(x2);
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_and8(dyn, ninst, x1, x2, x14, x3);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                GETGB(x2);
+                LDREXB(x1, wback);
+                emit_and8(dyn, ninst, x1, x2, x14, x2);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x21:
+            INST_NAME("LOCK AND Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_and32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_and32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x28:
+            INST_NAME("LOCK SUB Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGB(x2);
+            if((nextop&0xC0)==0xC0) {
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_sub8(dyn, ninst, x1, x2, x14, x3, 0);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREXB(x1, wback);
+                emit_sub8(dyn, ninst, x1, x2, x14, x3, (wback==x3)?1:0);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x29:
+            INST_NAME("LOCK SUB Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_sub32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_sub32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+        case 0x38:
+            INST_NAME("LOCK XOR Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                GETGB(x2);
+                wback = (nextop&7);    
+                wb2 = (wback>>2);      
+                wback = xEAX+(wback&3);
+                UXTB(x1, wback, wb2);   
+                emit_xor8(dyn, ninst, x1, x2, x14, x3);
+                BFI(wback, ed, wb2*8, 8);
+            } else {                   
+                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                MARKLOCK;
+                GETGB(x2);
+                LDREXB(x1, wback);
+                emit_xor8(dyn, ninst, x1, x2, x14, x2);
+                STREXB(x14, x1, wback);
+                CMPS_IMM8(x14, 0);
+                B_MARKLOCK(cNE);    // write failed, try again
+            }
+            break;
+        case 0x39:
+            INST_NAME("LOCK XOR Ed, Gd");
+            SETFLAGS(X_ALL, SF_SET);
+            nextop = F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {
+                ed = xEAX+(nextop&7);
+                emit_xor32(dyn, ninst, ed, gd, x3, x14);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                MARKLOCK;
+                LDREX(x1, wback);
+                emit_xor32(dyn, ninst, x1, gd, x3, x14);
+                STREX(x3, x1, wback);
+                CMPS_IMM8(x3, 0);
+                B_MARKLOCK(cNE);
+            }
+            break;
+
+        case 0x81:
+        case 0x83:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0: //ADD
+                    if(opcode==0x81) {
+                        INST_NAME("LOCK ADD Ed, Id");
+                    } else {
+                        INST_NAME("LOCK ADD Ed, Ib");
+                    }
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_add32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        TSTS_IMM8(wback, 0x3);
+                        B_MARK(cNE);
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_add32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                        B_NEXT(c__);
+                        MARK;   // unaligned! also, not enough 
+                        LDR_IMM9(x1, wback, 0);
+                        LDREXB(x14, wback);
+                        BFI(x1, x14, 0, 8); // re-inject
+                        emit_add32c(dyn, ninst, x1, i32, x3, x14);
+                        STREXB(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARK(cNE);
+                        STR_IMM9(x1, wback, 0);    // put the whole value
+                    }
+                    break;
+                case 1: //OR
+                    if(opcode==0x81) {INST_NAME("LOCK OR Ed, Id");} else {INST_NAME("LOCK OR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_or32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_or32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+                case 2: //ADC
+                    if(opcode==0x81) {INST_NAME("LOCK ADC Ed, Id");} else {INST_NAME("LOCK ADC Ed, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_adc32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_adc32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+                case 3: //SBB
+                    if(opcode==0x81) {INST_NAME("LOCK SBB Ed, Id");} else {INST_NAME("LOCK SBB Ed, Ib");}
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_sbb32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_sbb32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+                case 4: //AND
+                    if(opcode==0x81) {INST_NAME("LOCK AND Ed, Id");} else {INST_NAME("LOCK AND Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_and32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_and32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+                case 5: //SUB
+                    if(opcode==0x81) {INST_NAME("LOCK SUB Ed, Id");} else {INST_NAME("LOCK SUB Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_sub32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        TSTS_IMM8(wback, 0x3);
+                        B_MARK(cNE);
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_sub32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                        B_NEXT(c__);
+                        MARK;   // unaligned! also, not enough 
+                        LDR_IMM9(x1, wback, 0);
+                        LDREXB(x14, wback);
+                        BFI(x1, x14, 0, 8); // re-inject
+                        emit_sub32c(dyn, ninst, x1, i32, x3, x14);
+                        STREXB(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARK(cNE);
+                        STR_IMM9(x1, wback, 0);    // put the whole value
+                    }
+                    break;
+                case 6: //XOR
+                    if(opcode==0x81) {INST_NAME("LOCK XOR Ed, Id");} else {INST_NAME("LOCK XOR Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    if((nextop&0xC0)==0xC0) {
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        ed = xEAX+(nextop&7);
+                        emit_xor32c(dyn, ninst, ed, i32, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_xor32c(dyn, ninst, x1, i32, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+                case 7: //CMP
+                    if(opcode==0x81) {INST_NAME("(LOCK) CMP Ed, Id");} else {INST_NAME("(LOCK) CMP Ed, Ib");}
+                    SETFLAGS(X_ALL, SF_SET);
+                    GETEDH(x1);
+                    // No need to LOCK, this is readonly
+                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
+                    if(i32) {
+                        MOV32(x2, i32);
+                        emit_cmp32(dyn, ninst, ed, x2, x3, x14);
+                    } else {
+                        emit_cmp32_0(dyn, ninst, ed, x3, x14);
+                    }
+                    break;
+            }
+            break;
+
+        // generic case, no lock needed, the value is only read (note that on x86 locked read is always followed wy a locked write)
        #define GO(A)   \
-        case A+0x00:    \
-        case A+0x01:    \
        case A+0x02:    \
        case A+0x03:    \
        case A+0x04:    \
@@ -51,62 +510,444 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        GO(0x28);
        GO(0x30);
        #undef GO
-
-        case 0x81:
-        case 0x83:
-            MESSAGE(LOG_DUMP, "LOCK\n");
-            locked = 1;
-            LOCK;
-            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
-            break;
-
        case 0x86:  // for this two, the lock is already done by the opcode, so just ignoring it
        case 0x87:
-            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
+            --addr;
            break;
-
+        case 0x90:  // This is xchg reg, reg, so let's ignore the LOCK prefix...
+        case 0x91:
+        case 0x92:
+        case 0x93:
+        case 0x94:
+        case 0x95:
+        case 0x96:
+        case 0x97:
+            --addr;
+            break;
+    
        case 0x0F:
-            nextop = PK(1);
+            nextop = F8;
            switch(nextop) {
                case 0xB0:
+                    INST_NAME("LOCK CMPXCHG Eb, Gb");
+                    SETFLAGS(X_ALL, SF_SET);
+                    nextop = F8;
+                    if((nextop&0xC0)==0xC0) {
+                        wback = (nextop&7);    
+                        wb2 = (wback>>2);      
+                        wback = xEAX+(wback&3);
+                        UXTB(x2, wback, wb2);   
+                        ed = x2;
+                        wb1 = 0;
+                    } else {                   
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                        MARKLOCK;
+                        LDREXB(x2, wback);
+                        ed = x2;
+                        wb1 = 1;
+                    }
+                    UXTB(x1, xEAX, 0);
+                    CMPS_REG_LSL_IMM5(x1, ed, 0);
+                    B_MARK(cNE);
+                    // AL == Eb
+                    GETGB(x1);
+                    if(wb1) {
+                        STREXB(x14, x1, wback);
+                        CMPS_IMM8(x14, 0);
+                        B_MARKLOCK(cNE);    // write failed, try again
+                        if(!(gb1==xEAX && gb2==0)) {
+                            UXTB(x1, xEAX, 0);  // readback AL if needed
+                        }
+                    } else {
+                        BFI(wback, x1, wb2*8, 8);
+                    }
+                    // done
+                    B_MARK3(c__);
+                    MARK;
+                    // AL != Eb
+                    BFI(xEAX, ed, 0, 8);
+                    MARK3;
+                    // done, do the cmp now
+                    emit_cmp8(dyn, ninst, x1, x2, x3, x14);
+                    break;
                case 0xB1:
+                    INST_NAME("LOCK CMPXCHG Ed, Gd");
+                    SETFLAGS(X_ALL, SF_SET);
+                    nextop = F8;
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        MARKLOCK;
+                        TSTS_IMM8(wback, 3);    // can be unaligned it seems
+                        LDREX_COND(cEQ, x1, wback);
+                        LDR_IMM9_COND(cNE, x1, wback, 0);
+                        LDREXB_COND(cNE, x3, wback); // dummy read, to arm the write...
+                        ed = x1;
+                    }
+                    CMPS_REG_LSL_IMM5(xEAX, ed, 0);
+                    B_MARK(cNE);
+                    // EAX == Ed
+                    if(wback) {
+                        TSTS_IMM8(wback, 3);
+                        STREX_COND(cEQ, x14, gd, wback);
+                        STREXB_COND(cNE, x14, gd, wback);
+                        CMPS_IMM8(x14, 0);
+                        B_MARKLOCK(cNE);
+                        TSTS_IMM8(wback, 3);    // anoying, all those test
+                        STR_IMM9_COND(cNE, gd, wback, 0);
+                        emit_cmp32(dyn, ninst, xEAX, ed, x1, x14);
+                    } else {
+                        emit_cmp32(dyn, ninst, xEAX, ed, x1, x14);
+                        MOV_REG(ed, gd);
+                    }
+                    // done
+                    B_MARK3(c__);   // not next, in case its called with a LOCK prefix
+                    MARK;
+                    // EAX != Ed
+                    emit_cmp32(dyn, ninst, xEAX, ed, x3, x14);
+                    MOV_REG(xEAX, ed);
+                    MARK3
+                    break;
                case 0xB3:
+                    INST_NAME("LOCK BTR Ed, Gd");
+                    SETFLAGS(X_CF, SF_SET);
+                    nextop = F8;
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                        UBFX(x1, gd, 5, 3); // r1 = (gd>>5);
+                        ADD_REG_LSL_IMM5(x3, wback, x1, 2); //(&ed)+=r1*4;
+                        wback = x3;
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        ed = x1;
+                    }
+                    AND_IMM8(x2, gd, 0x1f);
+                    MOV_REG_LSR_REG(x14, ed, x2);
+                    ANDS_IMM8(x14, x14, 1);
+                    BFI(xFlags, x14, F_CF, 1);
+                    B_MARK3(cEQ); // bit already clear, jump to end of instruction
+                    MOVW(x14, 1);
+                    XOR_REG_LSL_REG(ed, ed, x14, x2);
+                    if(wback) {
+                        STREX(x14, ed, wback);
+                        CMPS_IMM8(x14, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    MARK3;
+                    break;
+
                case 0xBA:
+                    nextop = F8;
+                    switch((nextop>>3)&7) {
+                        case 4:
+                            INST_NAME("(LOCK) BT Ed, Ib");
+                            SETFLAGS(X_CF, SF_SUBSET);
+                            gd = x2;
+                            if((nextop&0xC0)==0xC0) {
+                                ed = xEAX+(nextop&7);
+                                u8 = F8;
+                            } else {
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x3, &fixedaddress, 4095-32, 0);
+                                u8 = F8;
+                                fixedaddress+=(u8>>5)*4;
+                                LDR_IMM9(x1, ed, fixedaddress);
+                                ed = x1;
+                            }
+                            u8&=0x1f;
+                            if(u8) {
+                                MOV_REG_LSR_IMM5(x1, ed, u8);
+                                ed = x1;
+                            }
+                            BFI(xFlags, ed, F_CF, 1);
+                            break;
+                        case 5:
+                            INST_NAME("LOCK BTS Ed, Ib");
+                            SETFLAGS(X_CF, SF_SUBSET);
+                            gd = x2;
+                            if((nextop&0xC0)==0xC0) {
+                                ed = xEAX+(nextop&7);
+                                u8 = F8;
+                                MOVW(gd, u8);
+                                wback = 0;
+                            } else {
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x3, &fixedaddress, 0, 0);
+                                u8 = F8;
+                                MOVW(gd, u8);
+                                UBFX(x1, gd, 5, 3); // r1 = (gd>>5);
+                                ADD_REG_LSL_IMM5(x3, ed, x1, 2); //(&ed)+=r1*4;
+                                MARKLOCK;
+                                LDREX(x1, x3);
+                                ed = x1;
+                                wback = x3;
+                            }
+                            AND_IMM8(x2, gd, 0x1f);
+                            MOV_REG_LSR_REG(x14, ed, x2);
+                            ANDS_IMM8(x14, x14, 1);
+                            BFI(xFlags, x14, F_CF, 1);
+                            B_MARK3(cNE); // bit already set, jump to next instruction
+                            MOVW(x14, 1);
+                            XOR_REG_LSL_REG(ed, ed, x14, x2);
+                            if(wback) {
+                                STREX(x14, ed, wback);
+                                CMPS_IMM8(x14, 0);
+                                B_MARKLOCK(cNE);
+                            }
+                            MARK3;
+                            break;
+                        case 6:
+                            INST_NAME("LOCK BTR Ed, Ib");
+                            SETFLAGS(X_CF, SF_SUBSET);
+                            gd = x2;
+                            if((nextop&0xC0)==0xC0) {
+                                ed = xEAX+(nextop&7);
+                                u8 = F8;
+                                MOVW(gd, u8);
+                                wback = 0;
+                            } else {
+                                addr = geted(dyn, addr, ninst, nextop, &ed, x3, &fixedaddress, 0, 0);
+                                u8 = F8;
+                                MOVW(gd, u8);
+                                UBFX(x1, gd, 5, 3); // r1 = (gd>>5);
+                                ADD_REG_LSL_IMM5(x3, ed, x1, 2); //(&ed)+=r1*4;
+                                MARKLOCK;
+                                LDREX(x1, x3);
+                                ed = x1;
+                                wback = x3;
+                            }
+                            AND_IMM8(x2, gd, 0x1f);
+                            MOV_REG_LSR_REG(x14, ed, x2);
+                            ANDS_IMM8(x14, x14, 1);
+                            BFI(xFlags, x14, F_CF, 1);
+                            B_MARK3(cEQ); // bit already clear, jump to next instruction
+                            //MOVW(x14, 1); // already 0x01
+                            XOR_REG_LSL_REG(ed, ed, x14, x2);
+                            if(wback) {
+                                STREX(x14, ed, wback);
+                                CMPS_IMM8(x14, 0);
+                                B_MARKLOCK(cNE);
+                            }
+                            MARK3;
+                            break;
+                        default:
+                            DEFAULT;
+                    }
+                    break;
                case 0xBB:
+                    INST_NAME("LOCK BTC Ed, Gd");
+                    SETFLAGS(X_CF, SF_SET);
+                    nextop = F8;
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0);
+                        UBFX(x1, gd, 5, 3); // r1 = (gd>>5);
+                        ADD_REG_LSL_IMM5(x3, wback, x1, 2); //(&ed)+=r1*4;
+                        MARKLOCK;
+                        LDREX(x1, x3);
+                        ed = x1;
+                        wback = x3;
+                    }
+                    AND_IMM8(x2, gd, 0x1f);
+                    MOV_REG_LSR_REG(x14, ed, x2);
+                    BFI(xFlags, x14, F_CF, 1);
+                    MOVW(x14, 1);
+                    XOR_REG_LSL_REG(ed, ed, x14, x2);
+                    if(wback) {
+                        STREX(x14, ed, wback);
+                        CMPS_IMM8(x14, 0);
+                        B_MARKLOCK(cNE);
+                    }
+                    break;
+            
                case 0xC0:
+                    INST_NAME("LOCK XADD Gb, Eb");
+                    SETFLAGS(X_ALL, SF_SET);
+                    nextop = F8;
+                    GETGB(x1);
+                    if((nextop&0xC0)==0xC0) {
+                        wback = (nextop&7); 
+                        wb2 = (wback>>2); 
+                        wback = xEAX+(wback&3); 
+                        UXTB(x2, wback, wb2); 
+                        wb1 = 0;
+                        ed = x2;
+                    } else { 
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0); 
+                        MARKLOCK;
+                        LDREXB(x2, wback); 
+                        wb1 = 1;
+                        ed = x2;
+                    }
+                    BFI(gb1, ed, gb2*8, 8); // gb <- eb
+                    emit_add8(dyn, ninst, ed, gd, x14, x3, 1);
+                    if(wb1) {
+                        STREXB(x14, ed, wback);
+                        CMPS_IMM8(x14, 0);
+                        B_MARKLOCK(cNE);
+                    } else {
+                        BFI(wback, ed, wb2*8, 8);
+                    }
+                    break;
                case 0xC1:
+                    INST_NAME("LOCK XADD Gd, Ed");
+                    SETFLAGS(X_ALL, SF_SET);
+                    nextop = F8;
+                    GETGD;
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        if(gd!=ed) {
+                            XOR_REG_LSL_IMM5(gd, gd, ed, 0);    // swap gd, ed
+                            XOR_REG_LSL_IMM5(ed, gd, ed, 0);
+                            XOR_REG_LSL_IMM5(gd, gd, ed, 0);
+                        }
+                        emit_add32(dyn, ninst, ed, gd, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        TSTS_IMM8(wback, 3);
+                        B_MARK(cNE);    // unaligned
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        ADD_REG_LSL_IMM5(x14, x1, gd, 0);
+                        STREX(x3, x14, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                        B_MARK2(c__);
+                        MARK;
+                        LDR_IMM9(x1, wback, 0);
+                        LDREXB(x14, wback);
+                        BFI(x1, x14, 0, 8);
+                        ADD_REG_LSL_IMM5(x14, x1, gd, 0);
+                        STREXB(x3, x14, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARK(cNE);
+                        STR_IMM9(x14, wback, 0);
+                        MARK2;
+                        IFX(X_ALL|X_PEND) {
+                            MOV_REG(x2, x1);
+                            emit_add32(dyn, ninst, x2, gd, x3, x14);
+                        }
+                        MOV_REG(gd, x1);
+                    }
+                    break;
+
                case 0xC7:
-                    MESSAGE(LOG_DUMP, "LOCK\n");
-                    locked = 1;
-                    LOCK;
-                    addr = dynarec0F(dyn, addr+1, ip, ninst, ok, need_epilog);
+                    INST_NAME("LOCK CMPXCHG8B Gq, Eq");
+                    SETFLAGS(X_ZF, SF_SET);
+                    nextop = F8;
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 0, 0);
+                    MARKLOCK;
+                    TSTS_IMM8(wback, 7);
+                    LDREXD_COND(cEQ, x2, wback);
+                    LDREX_COND(cNE, x2, wback);
+                    LDR_IMM9_COND(cNE, x3, wback, 4);
+                    CMPS_REG_LSL_IMM5(xEAX, x2, 0);
+                    B_MARK(cNE);    // EAX != Ed[0]
+                    CMPS_REG_LSL_IMM5(xEDX, x3, 0);
+                    B_MARK(cNE);    // EDX != Ed[1]
+                    MOV_REG(x2, xEBX);
+                    MOV_REG(x3, xECX);
+                    TSTS_IMM8(wback, 7);
+                    STREXD_COND(cEQ, x14, x2, wback);
+                    STREX_COND(cNE, x14, x2, wback);
+                    CMPS_IMM8(x14, 0);
+                    B_MARKLOCK(cNE);
+                    TSTS_IMM8(wback, 7);
+                    STR_IMM9_COND(cNE, x3, wback, 4);
+                    MOVW(x1, 1);
+                    B_MARK3(c__);
+                    MARK;
+                    MOV_REG(xEAX, x2);
+                    MOV_REG(xEDX, x3);
+                    MOVW(x1, 0);
+                    MARK3;
+                    BFI(xFlags, x1, F_ZF, 1);
                    break;

                default:
-                    addr = dynarec0F(dyn, addr+1, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+                    // dafault to NO LOCK
+                    addr-=2;
            }
            break;

        case 0xFF:
-            nextop = PK(1);
+            nextop = F8;
            switch((nextop>>3)&7)
            {
-                case 0:
-                case 1:
-                    MESSAGE(LOG_DUMP, "LOCK\n");
-                    locked = 1;
-                    LOCK;
-                    addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
+                case 0: // INC Ed
+                    INST_NAME("INC Ed");
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        emit_inc32(dyn, ninst, ed, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        TSTS_IMM8(wback, 3);
+                        B_MARK(cNE);    // unaligned
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_inc32(dyn, ninst, x1, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                        B_NEXT(c__);
+                        MARK;
+                        LDR_IMM9(x1, wback, 0);
+                        LDREXB(x3, wback);
+                        BFI(x1, x3, 0, 8);
+                        emit_inc32(dyn, ninst, x1, x3, x14);
+                        STREXB(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARK(cNE);
+                        STR_IMM9(x1, wback, 0);
+                    }
+                    break;
+                case 1: //DEC Ed
+                    INST_NAME("DEC Ed");
+                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET);
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        emit_dec32(dyn, ninst, ed, x3, x14);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0);
+                        TSTS_IMM8(wback, 3);
+                        B_MARK(cNE);    // unaligned
+                        MARKLOCK;
+                        LDREX(x1, wback);
+                        emit_dec32(dyn, ninst, x1, x3, x14);
+                        STREX(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARKLOCK(cNE);
+                        B_NEXT(c__);
+                        MARK;
+                        LDR_IMM9(x1, wback, 0);
+                        LDREXB(x3, wback);
+                        BFI(x1, x3, 0, 8);
+                        emit_dec32(dyn, ninst, x1, x3, x14);
+                        STREXB(x3, x1, wback);
+                        CMPS_IMM8(x3, 0);
+                        B_MARK(cNE);
+                        STR_IMM9(x1, wback, 0);
+                    }
                    break;
                default:
-                addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+                    // default to NO LOCK
+                    addr-=2;
            }
            break;
       
        default:
-            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+            // default to NO LOCK
+            --addr;
    }
-    if(locked) {UNLOCK;}

    return addr;
 }
@@ -22,43 +22,53 @@
 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"

-// Get Ex as a double, not a quad
+// Get Ex as a double, not a quad (warning, x2 and x3 may get used)
 #define GETEX(a) \
    if((nextop&0xC0)==0xC0) { \
        a = sse_get_reg(dyn, ninst, x1, nextop&7); \
    } else {    \
-        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3); \
+        parity = getedparity(dyn, ninst, addr, nextop, 3);  \
        a = fpu_get_scratch_double(dyn);            \
-        VLDR_64(a, ed, fixedaddress);               \
+        if(parity) {                                \
+            addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3); \
+            VLDR_64(a, ed, fixedaddress);           \
+        } else {                                    \
+            addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);\
+            LDR_IMM9(x2, ed, fixedaddress+0);       \
+            LDR_IMM9(x3, ed, fixedaddress+4);       \
+            VMOVtoV_D(a, x2, x3);                   \
+        }                                           \
    }

 uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop = F8;
    uint8_t u8;
-    int32_t i32, i32_;
-    int16_t i16;
    uint32_t u32;
-    uint16_t u16;
    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    uint8_t eb1, eb2;
+    uint8_t wback;
    int v0, v1;
-    int q0, q1;
    int d0, d1;
-    int s0, s1;
+    int s0;
    int fixedaddress;
+    int parity;
+
+    MAYUSE(s0);
+    MAYUSE(d1);
+    MAYUSE(d0);
+
    switch(nextop) {
-        
+
        case 0x10:
            INST_NAME("MOVSD Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
            if((nextop&0xC0)==0xC0) {
+                v0 = sse_get_reg(dyn, ninst, x1, gd);
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
                VMOVD(v0, d0);
            } else {
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
                LDR_IMM9(x2, ed, fixedaddress+0);
                LDR_IMM9(x3, ed, fixedaddress+4);
@@ -85,18 +95,18 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVDDUP Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
            if((nextop&0xC0)==0xC0) {
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVD(v0, d0);
-                VMOVD(v0+1, d0);
            } else {
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
                LDR_IMM9(x2, ed, fixedaddress+0);
                LDR_IMM9(x3, ed, fixedaddress+4);
                VMOVtoV_D(v0, x2, x3);
-                VMOVD(v0+1, v0);
            }
+            VMOVD(v0+1, v0);
            break;

        case 0x2A:
@@ -127,7 +137,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            break;
        case 0x2D:
            INST_NAME("CVTSD2SI Gd, Ex");
-            u8 = x87_setround(dyn, ninst, x1, x2, x12);
+            u8 = sse_setround(dyn, ninst, x1, x2, x14);
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
@@ -199,7 +209,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            // MINSD: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
            VCMP_F64(v0, d0);
            VMRS_APSR();
-            VMOVcond_64(cPL, v0, d0);
+            VMOVcond_64(cCS, v0, d0); // move v0 <- d0 if v0 or d0 unordered or v0 >= d0
            break;
        case 0x5E:
            INST_NAME("DIVSD Gx, Ex");
@@ -218,7 +228,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            // MAXSD: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
            VCMP_F64(d0, v0);
            VMRS_APSR();
-            VMOVcond_64(cPL, v0, d0);
+            VMOVcond_64(cCS, v0, d0); // move v0 <- d0 if v0 or d0 is unordered or d0 >= v0
            break;

        case 0x70:
@@ -256,31 +266,74 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            }
            break;

+        case 0x7C:
+            INST_NAME("HADDPS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                v1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(v1, ed);
+            }
+            VPADD_F32(v0, v0, v0+1);
+            if(v0==v1) {
+                VMOVD(v0+1, v0);
+            } else {
+                VPADD_F32(v0+1, v1, v1+1);
+            }
+            break;
+
        case 0xC2:
            INST_NAME("CMPSD Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd);
            GETEX(d0);
-            VCMP_F64(v0, d0);
+            u8 = F8;
+            if((u8&7)==6){
+                VCMP_F64(d0, v0);
+            } else {
+                VCMP_F64(v0, d0);
+            }
            VMRS_APSR();
            MOVW(x2, 0);
-            u8 = F8;
            switch(u8&7) {
-                case 0: MVN_COND_REG_LSL_IMM8(cEQ, x2, x2, 0); break;   // Equal
-                case 1: MVN_COND_REG_LSL_IMM8(cCC, x2, x2, 0); break;   // Less than
-                case 2: MVN_COND_REG_LSL_IMM8(cLS, x2, x2, 0); break;   // Less or equal
-                case 3: MVN_COND_REG_LSL_IMM8(cVS, x2, x2, 0); break;   // NaN
-                case 4: MVN_COND_REG_LSL_IMM8(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
-                case 5: MVN_COND_REG_LSL_IMM8(cPL, x2, x2, 0); break;   // Greater or equal or unordered
-                case 6: MVN_COND_REG_LSL_IMM8(cHI, x2, x2, 0); break;   // Greater or unordered
-                case 7: MVN_COND_REG_LSL_IMM8(cVC, x2, x2, 0); break;   // not NaN
+                case 0: MVN_COND_REG_LSL_IMM5(cEQ, x2, x2, 0); break;   // Equal
+                case 1: MVN_COND_REG_LSL_IMM5(cMI, x2, x2, 0); break;   // Less than
+                case 2: MVN_COND_REG_LSL_IMM5(cLE, x2, x2, 0); break;   // Less or equal
+                case 3: MVN_COND_REG_LSL_IMM5(cVS, x2, x2, 0); break;   // NaN
+                case 4: MVN_COND_REG_LSL_IMM5(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
+                case 5: MVN_COND_REG_LSL_IMM5(cCS, x2, x2, 0); break;   // Greater or equal or unordered
+                case 6: MVN_COND_REG_LSL_IMM5(cLT, x2, x2, 0); break;   // Greater or unordered, test inverted, N!=V so unordereded or less than (inverted)
+                case 7: MVN_COND_REG_LSL_IMM5(cVC, x2, x2, 0); break;   // not NaN
            }
            VMOVtoV_D(v0, x2, x2);
            break;

+        case 0xF0:
+            INST_NAME("LDDQ Gx,Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                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, 4095-12, 0);
+                LDR_IMM9(x2, ed, fixedaddress+0);
+                LDR_IMM9(x3, ed, fixedaddress+4);
+                VMOVtoV_D(v0, x2, x3);
+                LDR_IMM9(x2, ed, fixedaddress+8);
+                LDR_IMM9(x3, ed, fixedaddress+12);
+                VMOVtoV_D(v0+1, x2, x3);
+            }
+            break;
+
        default:
-            *ok = 0;
            DEFAULT;
    }
    return addr;
@@ -34,27 +34,43 @@
        }                                           \
        a *= 2;                                     \
    } else {    \
-        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3); \
-        a = fpu_get_scratch_single(dyn);            \
-        VLDR_32(a, ed, fixedaddress);               \
+        b = fpu_get_scratch_double(dyn);            \
+        a = b*2;                                    \
+        parity = getedparity(dyn, ninst, addr, nextop, 2);  \
+        if(parity) {                                \
+            addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 0); \
+            VLDR_32(a, ed, fixedaddress);           \
+        } else {                                    \
+            GETED;                                  \
+            VMOVtoV(a, ed);                         \
+        }                                           \
    }

+#define GETGX(a) \
+    gd = (nextop&0x38)>>3;                          \
+    a = sse_get_reg(dyn, ninst, x1, gd)
+
+
 uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t opcode = F8;
    uint8_t nextop, u8;
-    int32_t i32, i32_;
+    int32_t j32;
    uint32_t u32;
-    int16_t i16;
-    uint16_t u16;
    uint8_t gd, ed;
-    uint8_t wback, wb1, wb2;
-    uint8_t eb1, eb2;
+    uint8_t wback;
    int fixedaddress;
    int v0, v1;
-    int s0, s1;
+    int s0, s1, s2;
    int d0, d1;
    int q0, q1;
+    int parity;
+
+    MAYUSE(q1);
+    MAYUSE(s1);
+    MAYUSE(s2);
+    MAYUSE(j32);
+
    switch(opcode) {

        case 0x10:
@@ -117,6 +133,49 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                STR_IMM9(x2, ed, fixedaddress);
            }
            break;
+        case 0x12:
+            INST_NAME("MOVSLDUP Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                q1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                q1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(q1, ed);
+            }
+            q0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            VDUP_32(q0+0, q1+0, 0);
+            VDUP_32(q0+1, q1+1, 0);
+            break;
+        
+        case 0x16:
+            INST_NAME("MOVSHDUP Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                q1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                q1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(q1, ed);
+            }
+            q0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            VDUP_32(q0+0, q1+0, 1);
+            VDUP_32(q0+1, q1+1, 1);
+            break;
+
+        
+        case 0x1E:
+            INST_NAME("NOP / ENDBR32 / ENDBR64");
+            nextop = F8;
+            FAKEED;
+            break;
+        case 0x1F:
+            INST_NAME("REP NOP");
+            nextop = F8;
+            FAKEED;
+            break;

        case 0x2A:
            INST_NAME("CVTSI2SS Gx, Ed");
@@ -153,15 +212,22 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                }
                VMOV_32(s0, d0*2);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
-                VLDR_32(s0, ed, fixedaddress);
+                parity = getedparity(dyn, ninst, addr, nextop, 3);
+                if(parity) {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VLDR_32(s0, ed, fixedaddress);
+                } else {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    LDR_IMM9(x2, ed, fixedaddress);
+                    VMOVtoV(s0, x2);
+                }
            }
            VCVT_S32_F32(s0, s0);
            VMOVfrV(gd, s0);
            break;
        case 0x2D:
            INST_NAME("CVTSS2SI Gd, Ex");
-            u8 = x87_setround(dyn, ninst, x1, x2, x12);
+            u8 = sse_setround(dyn, ninst, x1, x2, x14);
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
@@ -175,8 +241,15 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                }
                VMOV_32(s0, d0*2);
            } else {
-                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
-                VLDR_32(s0, ed, fixedaddress);
+                parity = getedparity(dyn, ninst, addr, nextop, 3);
+                if(parity) {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VLDR_32(s0, ed, fixedaddress);
+                } else {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    LDR_IMM9(x2, ed, fixedaddress);
+                    VMOVtoV(s0, x2);
+                }
            }
            VCVTR_S32_F32(s0, s0);
            VMOVfrV(gd, s0);
@@ -186,104 +259,118 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
        case 0x51:
            INST_NAME("SQRTSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            VSQRT_F32(d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            VSQRT_F32(s0, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x52:
            INST_NAME("RSQRTSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            s1 = fpu_get_scratch_single(dyn);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            s2 = fpu_get_scratch_single(dyn);
            // so here: F32: Imm8 = abcd efgh that gives => aBbbbbbc defgh000 00000000 00000000
            // and want 1.0f = 0x3f800000
            // so 00111111 10000000 00000000 00000000
            // a = 0, b = 1, c = 1, d = 1, efgh=0
            // 0b01110000
-            VMOV_i_32(s1, 0b01110000);
-            VDIV_F32(s0, s1, s0);
-            VSQRT_F32(d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            VMOV_i_32(s2, 0b01110000);
+            VSQRT_F32(s0, s1);
+            VDIV_F32(s0, s2, s0);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x53:
            INST_NAME("RCPSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            s1 = fpu_get_scratch_single(dyn);
-            VMOV_i_32(s1, 0b01110000);
-            VDIV_F32(d1*2, s1, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            s2 = fpu_get_scratch_single(dyn);
+            VMOV_i_32(s2, 0b01110000);
+            VDIV_F32(s0, s2, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;

        case 0x58:
            INST_NAME("ADDSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            VADD_F32(d1*2, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            VADD_F32(s0, s0, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x59:
            INST_NAME("MULSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            VMUL_F32(d1*2, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            VMUL_F32(s0, s0, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x5A:
            INST_NAME("CVTSS2SD Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETGX(v0);
            GETEX(s0, d0);
            VCVT_F64_F32(v0, s0);
            break;
@@ -304,75 +391,87 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
        case 0x5C:
            INST_NAME("SUBSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            VSUB_F32(d1*2, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            VSUB_F32(s0, s0, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x5D:
            INST_NAME("MINSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
            // MINSS: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
-            VCMP_F32(d1*2, s0);
+            VCMP_F32(s0, s1);
            VMRS_APSR();
-            VMOVcond_32(cPL, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            VMOVcond_32(cCS, s0, s1); // move s0 <- s1 if s0 or s1 unordered or s0 >= s1
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x5E:
            INST_NAME("DIVSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
-            VDIV_F32(d1*2, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            VDIV_F32(s0, s0, s1);
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;
        case 0x5F:
            INST_NAME("MAXSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
-            }
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
            // MAXSS: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
-            VCMP_F32(s0, d1*2);
+            VCMP_F32(s1, s0);
            VMRS_APSR();
-            VMOVcond_32(cPL, d1*2, s0);
-            if(v0!=d1) {
-                VMOV_64(v0, d1);
+            VMOVcond_32(cCS, s0, s1); // move s0 <- s1 if s0 or s1 is unordered or s1 >= s0
+            if(v0!=d0) {
+                VMOV_64(v0, d0);
            }
            break;

@@ -408,7 +507,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                VLD1Q_64(v1, ed);
            }
            u8 = F8;
-            // only high part need to be suffled. VTBL only handle 8bits value, so the 16bits suffles need to be changed in 8bits
+            // only high part need to be shuffled. VTBL only handle 8bits value, so the 16bits shuffles need to be changed in 8bits
            u32 = 0;
            for (int i=0; i<2; ++i) {
                u32 |= (((u8>>(i*2))&3)*2+0)<<(i*16+0);
@@ -450,8 +549,8 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            if((nextop&0xC0)==0xC0) {
-                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                v0 = sse_get_reg(dyn, ninst, x1, gd);
+                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
                v0 = sse_get_reg(dyn, ninst, x1, gd);
@@ -465,37 +564,93 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            }
            break;

+        case 0xBC:
+            INST_NAME("TZCNT Gd, Ed");
+            SETFLAGS(X_CF|X_ZF, SF_SET);
+            nextop = F8;
+            GETED;
+            GETGD;
+            SET_DFNONE(x3);
+            TSTS_REG_LSL_IMM5(ed, ed, 0);
+            MOVW_COND(cEQ, x3, 1);
+            MOVW_COND(cNE, x3, 0);
+            BFI(xFlags, x3, F_CF, 1);   // CF = is source 0?
+            RBIT(x3, ed);   // reverse
+            CLZ(gd, x3);    // x2 gets leading 0 == TZCNT
+            TSTS_REG_LSL_IMM5(gd, gd, 0);
+            MOVW_COND(cEQ, x3, 1);
+            MOVW_COND(cNE, x3, 0);
+            BFI(xFlags, x3, F_ZF, 1);   // ZF = is dest 0?
+            break;
+
        case 0xC2:
            INST_NAME("CMPSS Gx, Ex");
            nextop = F8;
-            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg(dyn, ninst, x1, gd);
-            GETEX(s0, d0);
-            if(v0<16)
-                d1 = v0;
-            else {
-                d1 = fpu_get_scratch_double(dyn);
-                VMOV_64(d1, v0);
+            GETGX(v0);
+            GETEX(s1, v1);
+            if(v0==v1)
+                d0 = s1/2;
+            else
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+            s0 = d0*2;
+            u8 = F8;
+            if((u8&7)==6){
+                VCMP_F32(s1, s0);
+            } else {
+                VCMP_F32(s0, s1);
            }
-            VCMP_F32(d1*2, s0);
            VMRS_APSR();
            MOVW(x2, 0);
-            u8 = F8;
            switch(u8&7) {
-                case 0: MVN_COND_REG_LSL_IMM8(cEQ, x2, x2, 0); break;   // Equal
-                case 1: MVN_COND_REG_LSL_IMM8(cCC, x2, x2, 0); break;   // Less than
-                case 2: MVN_COND_REG_LSL_IMM8(cLS, x2, x2, 0); break;   // Less or equal
-                case 3: MVN_COND_REG_LSL_IMM8(cVS, x2, x2, 0); break;   // NaN
-                case 4: MVN_COND_REG_LSL_IMM8(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
-                case 5: MVN_COND_REG_LSL_IMM8(cPL, x2, x2, 0); break;   // Greater or equal or unordered
-                case 6: MVN_COND_REG_LSL_IMM8(cHI, x2, x2, 0); break;   // Greater or unordered
-                case 7: MVN_COND_REG_LSL_IMM8(cVC, x2, x2, 0); break;   // not NaN
+                case 0: MVN_COND_REG_LSL_IMM5(cEQ, x2, x2, 0); break;   // Equal
+                case 1: MVN_COND_REG_LSL_IMM5(cMI, x2, x2, 0); break;   // Less than
+                case 2: MVN_COND_REG_LSL_IMM5(cLE, x2, x2, 0); break;   // Less or equal
+                case 3: MVN_COND_REG_LSL_IMM5(cVS, x2, x2, 0); break;   // NaN
+                case 4: MVN_COND_REG_LSL_IMM5(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
+                case 5: MVN_COND_REG_LSL_IMM5(cCS, x2, x2, 0); break;   // Greater or equal or unordered
+                case 6: MVN_COND_REG_LSL_IMM5(cLT, x2, x2, 0); break;   // Greater or unordered, test inverted, N!=V so unordereded or less than (inverted)
+                case 7: MVN_COND_REG_LSL_IMM5(cVC, x2, x2, 0); break;   // not NaN
            }
            VMOVtoDx_32(v0, 0, x2);
            break;
+        
+        case 0xE6:
+            INST_NAME("CVTDQ2PD Gx, Ex");
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                if(v1<16)
+                    v0 = v1;
+                else {
+                    v0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(v0, v1);    // vfpu opcode here
+                }
+            } else {
+                v0 = fpu_get_scratch_double(dyn);
+                parity = getedparity(dyn, ninst, addr, nextop, 3);
+                if(parity) {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VLDR_64(v0, ed, fixedaddress);    // vfpu opcode here
+                } else {
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-8, 0);
+                    LDR_IMM9(x2, ed, fixedaddress);
+                    LDR_IMM9(x3, ed, fixedaddress+4);
+                    VMOVtoV_D(v0, x2, x3);
+                }
+            }
+            gd = (nextop&0x38)>>3;
+            q0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            //v0 is a low reg now
+            VCVT_F64_S32(q0+1, v0*2+1);
+            VCVT_F64_S32(q0+0, v0*2+0);
+            break;

        default:
-            *ok = 0;
            DEFAULT;
    }
    return addr;
@@ -5,6 +5,9 @@
 #include <errno.h>
 #include <string.h>
 #include <math.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>

 #include "debug.h"
 #include "box86context.h"
@@ -18,20 +21,14 @@
 #include "emu/x86run_private.h"
 #include "emu/x87emu_private.h"
 #include "x86trace.h"
+#include "signals.h"
 #include "dynarec_arm.h"
 #include "dynarec_arm_private.h"
 #include "dynarec_arm_functions.h"

-void arm_popf(x86emu_t* emu, uint32_t f)
-{
-    emu->packed_eflags.x32 = (f & 0x3F7FD7) | 0x2; // mask off res2 and res3 and on res1
-    UnpackFlags(emu);
-    RESET_FLAGS(emu);
-}
-
 void arm_fstp(x86emu_t* emu, void* p)
 {
-    if(ST0.ll!=STld(0).ref)
+    if(ST0.q!=STld(0).ref)
        D2LD(&ST0.d, p);
    else
        memcpy(p, &STld(0).ld, 10);
@@ -60,7 +57,7 @@ void arm_fpatan(x86emu_t* emu)
 }
 void arm_fxtract(x86emu_t* emu)
 {
-    int32_t tmp32s = (ST1.ll&0x7ff0000000000000LL)>>52;
+    int32_t tmp32s = (ST1.q&0x7ff0000000000000LL)>>52;
    tmp32s -= 1023;
    ST1.d /= exp2(tmp32s);
    ST0.d = tmp32s;
@@ -106,9 +103,11 @@ void arm_fbld(x86emu_t* emu, uint8_t* ed)

 void arm_fild64(x86emu_t* emu, int64_t* ed)
 {
-    ST0.d = *ed;
-    STll(0).ll = *ed;
-    STll(0).ref = ST0.ll;
+    int64_t tmp;
+    memcpy(&tmp, ed, sizeof(tmp));
+    ST0.d = tmp;
+    STll(0).ll = tmp;
+    STll(0).ref = ST0.q;
 }

 void arm_fbstp(x86emu_t* emu, uint8_t* ed)
@@ -118,114 +117,74 @@ void arm_fbstp(x86emu_t* emu, uint8_t* ed)

 void arm_fistp64(x86emu_t* emu, int64_t* ed)
 {
-    if(STll(0).ref==ST(0).ll) {
-        *ed = STll(0).ll;
+    // used of memcpy to avoid aligments issues
+    if(STll(0).ref==ST(0).q) {
+        memcpy(ed, &STll(0).ll, sizeof(int64_t));
    } else {
-        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL))
-            *ed = 0x8000000000000000LL;
+        int64_t tmp;
+        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL) || !isfinite(ST0.d))
+            tmp = 0x8000000000000000LL;
        else
-            *ed = (int64_t)ST0.d;
+            tmp = fpu_round(emu, ST0.d);
+        memcpy(ed, &tmp, sizeof(tmp));
    }
 }

+void arm_fistt64(x86emu_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(x86emu_t* emu, uint8_t* ed)
 {
    memcpy(&STld(0).ld, ed, 10);
    LD2D(&STld(0), &ST(0).d);
-    STld(0).ref = ST0.ll;
+    STld(0).ref = ST0.q;
 }

-void arm_cpuid(x86emu_t* emu, uint32_t tmp32u)
+void arm_ud(x86emu_t* emu)
 {
-    switch(tmp32u) {
-        case 0x0:
-            // emulate a P4
-            R_EAX = 0x80000004;
-            // return GenuineIntel
-            R_EBX = 0x756E6547;
-            R_EDX = 0x49656E69;
-            R_ECX = 0x6C65746E;
-            break;
-        case 0x1:
-            R_EAX = 0x00000101; // familly and all
-            R_EBX = 0;          // Brand indexe, CLFlush, Max APIC ID, Local APIC ID
-            R_EDX =   1         // fpu 
-                    | 1<<8      // cmpxchg8
-                    | 1<<11     // sep (sysenter & sysexit)
-                    | 1<<15     // cmov
-                    | 1<<19     // clflush (seems to be with SSE2)
-                    | 1<<23     // mmx
-                    //| 1<<24     // fxsr (fxsave, fxrestore)
-                    | 1<<25     // SSE
-                    | 1<<26     // SSE2
-                    ;
-            R_ECX =   1<<12     // fma
-                    | 1<<13     // cx16 (cmpxchg16)
-                    ;           // also 1<<0 is SSE3 and 1<<9 is SSSE3
-            break;
-        case 0x2:   // TLB and Cache info. Sending 1st gen P4 info...
-            R_EAX = 0x665B5001;
-            R_EBX = 0x00000000;
-            R_ECX = 0x00000000;
-            R_EDX = 0x007A7000;
-            break;
-        
-        case 0x4:   // Cache info
-            switch (R_ECX) {
-                case 0: // L1 data cache
-                    R_EAX = (1 | (1<<5) | (1<<8));   //type
-                    R_EBX = (63 | (7<<22)); // size
-                    R_ECX = 63;
-                    R_EDX = 1;
-                    break;
-                case 1: // L1 inst cache
-                    R_EAX = (2 | (1<<5) | (1<<8)); //type
-                    R_EBX = (63 | (7<<22)); // size
-                    R_ECX = 63;
-                    R_EDX = 1;
-                    break;
-                case 2: // L2 cache
-                    R_EAX = (3 | (2<<5) | (1<<8)); //type
-                    R_EBX = (63 | (15<<22));    // size
-                    R_ECX = 4095;
-                    R_EDX = 1;
-                    break;
-
-                default:
-                    R_EAX = 0x00000000;
-                    R_EBX = 0x00000000;
-                    R_ECX = 0x00000000;
-                    R_EDX = 0x00000000;
-                    break;
-            }
-            break;
-        case 0x5:   //mwait info
-            R_EAX = 0;
-            R_EBX = 0;
-            R_ECX = 1 | 2;
-            R_EDX = 0;
-            break;
-        case 0x6:   // thermal
-        case 0x9:   // direct cache access
-        case 0xA:   // Architecture performance monitor
-            R_EAX = 0;
-            R_EBX = 0;
-            R_ECX = 0;
-            R_EDX = 0;
-            break;
-        case 0x7:   // extended bits...
-            /*if(R_ECX==0)    R_EAX = 0;
-            else*/ R_EAX = R_ECX = R_EBX = R_EDX = 0;
-            break;
-
-        case 0x80000000:        // max extended
-            break;              // no extended, so return same value as beeing the max value!
-        default:
-            printf_log(LOG_INFO, "Warning, CPUID command %X unsupported (ECX=%08x)\n", tmp32u, R_ECX);
-            R_EAX = 0;
-    }   
+    emit_signal(emu, SIGILL, (void*)R_EIP, 0);
 }

+void arm_fsave(x86emu_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(x86emu_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(x86emu_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);
+}
+
+
 // Get a FPU single scratch reg
 int fpu_get_scratch_single(dynarec_arm_t* dyn)
 {
@@ -241,14 +200,25 @@ int fpu_get_scratch_double(dynarec_arm_t* dyn)
 // Get a FPU quad scratch reg
 int fpu_get_scratch_quad(dynarec_arm_t* dyn)
 {
+    if(dyn->fpu_scratch>4) {
+        if(dyn->fpu_extra_qscratch) {
+            dynarec_log(LOG_NONE, "Warning, Extra QScratch slot taken and need another one!\n");
+        } else
+            dyn->fpu_extra_qscratch = fpu_get_reg_quad(dyn);
+        return dyn->fpu_extra_qscratch;
+    }
    int i = (dyn->fpu_scratch+3)&(~3);
    dyn->fpu_scratch = i+4;
    return i/2; // Return a Dx, not a Qx
 }
 // Reset scratch regs counter
-int fpu_reset_scratch(dynarec_arm_t* dyn)
+void fpu_reset_scratch(dynarec_arm_t* dyn)
 {
    dyn->fpu_scratch = 0;
+    if(dyn->fpu_extra_qscratch) {
+        fpu_free_reg_quad(dyn, dyn->fpu_extra_qscratch);
+        dyn->fpu_extra_qscratch = 0;
+    }
 }
 #define FPUFIRST    8
 // Get a FPU double reg
@@ -282,9 +252,103 @@ void fpu_free_reg_quad(dynarec_arm_t* dyn, int reg)
    dyn->fpuused[i] = dyn->fpuused[i+1] = 0;
 }
 // Reset fpu regs counter
-int fpu_reset_reg(dynarec_arm_t* dyn)
+void fpu_reset_reg(dynarec_arm_t* dyn)
 {
    dyn->fpu_reg = 0;
    for (int i=0; i<24; ++i)
        dyn->fpuused[i]=0;
 }
+
+#define F8      *(uint8_t*)(addr++)
+#define F32     *(uint32_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)
+{
+
+    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) {
+                uint32_t tmp = F32;
+                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) {
+            uint32_t tmp = F32;
+            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) 
+{
+    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)
+{
+#define PK(a)       *(uint8_t*)(addr+a)
+#define PK32(a)     *(uint32_t*)(addr+a)
+
+    if(!addr)
+        return 0;
+    if(PK(0)==0xff && PK(1)==0x25) {  // absolute jump, maybe the GOT
+        uintptr_t a1 = (PK32(2));   // need to add a check to see if the address is from the GOT !
+        addr = *(uint32_t*)a1; 
+    }
+    if(addr<0x10000)    // too low, that is suspicious
+        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
+}
+
@@ -3,7 +3,6 @@

 typedef struct x86emu_s x86emu_t;

-void arm_popf(x86emu_t* emu, uint32_t f);
 void arm_fstp(x86emu_t* emu, void* p);

 void arm_print_armreg(x86emu_t* emu, uintptr_t reg, uintptr_t n);
@@ -24,9 +23,13 @@ void arm_fbld(x86emu_t* emu, uint8_t* ed);
 void arm_fild64(x86emu_t* emu, int64_t* ed);
 void arm_fbstp(x86emu_t* emu, uint8_t* ed);
 void arm_fistp64(x86emu_t* emu, int64_t* ed);
+void arm_fistt64(x86emu_t* emu, int64_t* ed);
 void arm_fld(x86emu_t* emu, uint8_t* ed);
+void arm_fsave(x86emu_t* emu, uint8_t* ed);
+void arm_frstor(x86emu_t* emu, uint8_t* ed);
+void arm_fprem1(x86emu_t* emu);

-void arm_cpuid(x86emu_t* emu, uint32_t tmp32u);
+void arm_ud(x86emu_t* emu);

 // Get an FPU single scratch reg
 int fpu_get_scratch_single(dynarec_arm_t* dyn);
@@ -35,7 +38,7 @@ int fpu_get_scratch_double(dynarec_arm_t* dyn);
 // Get an FPU quad scratch reg
 int fpu_get_scratch_quad(dynarec_arm_t* dyn);
 // Reset scratch regs counter
-int fpu_reset_scratch(dynarec_arm_t* dyn);
+void fpu_reset_scratch(dynarec_arm_t* dyn);
 // Get an FPU double reg
 int fpu_get_reg_double(dynarec_arm_t* dyn);
 // Free a FPU double reg
@@ -45,6 +48,14 @@ int fpu_get_reg_quad(dynarec_arm_t* dyn);
 // Free a FPU quad reg
 void fpu_free_reg_quad(dynarec_arm_t* dyn, int reg);
 // Reset fpu regs counter
-int fpu_reset_reg(dynarec_arm_t* dyn);
+void fpu_reset_reg(dynarec_arm_t* dyn);
+
+// 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);
+// Do the GETED, but don't emit anything...
+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);

 #endif //__DYNAREC_ARM_FUNCTIONS_H__
@@ -69,6 +69,16 @@
                    LDR_REG_LSL_IMM5(x1, wback, O, 0);  \
                    ed = x1;                 \
                }
+//GETEDO can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI, else O is added to wback
+#define GETEDO2(O) if((nextop&0xC0)==0xC0) {            \
+                    ed = xEAX+(nextop&7);               \
+                    wback = 0;                          \
+                } else {                                \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0); \
+                    ADD_REG_LSL_IMM5(wback, wback, O, 0);\
+                    LDR_IMM9(x1, wback, 0);             \
+                    ed = x1;                            \
+                }
 #define WBACKO(O)   if(wback) {STR_REG_LSL_IMM5(ed, wback, O, 0);}
 //FAKEELike GETED, but doesn't get anything
 #define FAKEED  if((nextop&0xC0)!=0xC0) {   \
@@ -132,6 +142,21 @@
                    wb1 = 1;                \
                    ed = i;                 \
                }
+//GETEBO will use i for ed, i is also Offset, and can use r3 for wback.
+#define GETEBO(i) if((nextop&0xC0)==0xC0) {  \
+                    wback = (nextop&7);     \
+                    wb2 = (wback>>2);       \
+                    wback = xEAX+(wback&3); \
+                    UXTB(i, wback, wb2);    \
+                    wb1 = 0;                \
+                    ed = i;                 \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0); \
+                    ADD_REG_LSL_IMM5(wback, wback, i, 0);   \
+                    LDRB_IMM9(i, wback, fixedaddress);      \
+                    wb1 = 1;                \
+                    ed = i;                 \
+                }
 //GETSEB sign extend EB, will use i for ed, and can use r3 for wback.
 #define GETSEB(i) if((nextop&0xC0)==0xC0) { \
                    wback = (nextop&7);     \
@@ -164,16 +189,22 @@
 #define GBBACK   BFI(gb1, gd, gb2*8, 8);

 // Get Direction with size Z and based of F_DF flag, on register r ready for LDR/STR fetching
+// F_DF is 1<<10, so 1 ROR 11*2 (so F_OF)
 #define GETDIR(r, A)    \
-    LDR_IMM9(r, xEmu, offsetof(x86emu_t, flags[F_DF]));     \
-    CMPS_IMM8(r, 1);                                        \
-    MOVW(r, A);                                             \
-    RSB_COND_IMM8(cEQ, r, r, 0)
+    TSTS_IMM8_ROR(xFlags, 1, 0x0b);         \
+    MOVW(r, A);                             \
+    RSB_COND_IMM8(cNE, r, r, 0)

-// CALL will use x12 for the call address. Return value can be put in ret (unless ret is -1)
-#define CALL(F, ret, M) call_c(dyn, ninst, F, x12, ret, M)
+// CALL will use x14 for the call address. Return value can be put in ret (unless ret is -1)
+// R0 will not be pushed/popd if ret is -2
+#define CALL(F, ret, M) call_c(dyn, ninst, F, x14, ret, M, 1)
 // CALL_ will use x3 for the call address. Return value can be put in ret (unless ret is -1)
-#define CALL_(F, ret, M) call_c(dyn, ninst, F, x3, ret, M)
+// R0 will not be pushed/popd if ret is -2
+#define CALL_(F, ret, M) call_c(dyn, ninst, F, x3, ret, M, 1)
+// CALL_S will use x3 for the call address. Return value can be put in ret (unless ret is -1)
+// R0 will not be pushed/popd if ret is -2. Flags are not save/restored
+#define CALL_S(F, ret, M) call_c(dyn, ninst, F, x3, ret, M, 0)
+
 #define MARK    if(dyn->insts) {dyn->insts[ninst].mark = (uintptr_t)dyn->arm_size;}
 #define GETMARK ((dyn->insts)?dyn->insts[ninst].mark:(dyn->arm_size+4))
 #define MARK2   if(dyn->insts) {dyn->insts[ninst].mark2 = (uintptr_t)dyn->arm_size;}
@@ -182,94 +213,133 @@
 #define GETMARK3 ((dyn->insts)?dyn->insts[ninst].mark3:(dyn->arm_size+4))
 #define MARKF   if(dyn->insts) {dyn->insts[ninst].markf = (uintptr_t)dyn->arm_size;}
 #define GETMARKF ((dyn->insts)?dyn->insts[ninst].markf:(dyn->arm_size+4))
+#define MARKSEG if(dyn->insts) {dyn->insts[ninst].markseg = (uintptr_t)dyn->arm_size;}
+#define GETMARKSEG ((dyn->insts)?dyn->insts[ninst].markseg:(dyn->arm_size+4))
+#define MARKLOCK if(dyn->insts) {dyn->insts[ninst].marklock = (uintptr_t)dyn->arm_size;}
+#define GETMARKLOCK ((dyn->insts)?dyn->insts[ninst].marklock:(dyn->arm_size+4))

-// Branch to MARK if cond (use i32)
+// Branch to MARK if cond (use j32)
 #define B_MARK(cond)    \
-    i32 = GETMARK-(dyn->arm_size+8);    \
-    Bcond(cond, i32)
-// Branch to MARK2 if cond (use i32)
+    j32 = GETMARK-(dyn->arm_size+8);    \
+    Bcond(cond, j32)
+// Branch to MARK2 if cond (use j32)
 #define B_MARK2(cond)    \
-    i32 = GETMARK2-(dyn->arm_size+8);   \
-    Bcond(cond, i32)
-// Branch to MARK3 if cond (use i32)
+    j32 = GETMARK2-(dyn->arm_size+8);   \
+    Bcond(cond, j32)
+// Branch to MARK3 if cond (use j32)
 #define B_MARK3(cond)    \
-    i32 = GETMARK3-(dyn->arm_size+8);   \
-    Bcond(cond, i32)
-// Branch to next instruction if cond (use i32)
+    j32 = GETMARK3-(dyn->arm_size+8);   \
+    Bcond(cond, j32)
+// Branch to next instruction if cond (use j32)
 #define B_NEXT(cond)     \
-    i32 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size+8)):0; \
-    Bcond(cond, i32)
+    j32 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size+8)):0; \
+    Bcond(cond, j32)
+// Branch to MARKSEG if cond (use j32)
+#define B_MARKSEG(cond)    \
+    j32 = GETMARKSEG-(dyn->arm_size+8);   \
+    Bcond(cond, j32)
+// Branch to MARKLOCK if cond (use j32)
+#define B_MARKLOCK(cond)    \
+    j32 = GETMARKLOCK-(dyn->arm_size+8);   \
+    Bcond(cond, j32)
+
+#define IFX(A)  if(dyn->insts && (dyn->insts[ninst].x86.need_flags&(A)))
+#define IFX_PENDOR0  if(dyn->insts && (dyn->insts[ninst].x86.need_flags&(X_PEND) || !dyn->insts[ninst].x86.need_flags))
+#define IFXX(A) if(dyn->insts && (dyn->insts[ninst].x86.need_flags==(A)))
+#define IFX2X(A, B) if(dyn->insts && (dyn->insts[ninst].x86.need_flags==(A) || dyn->insts[ninst].x86.need_flags==(B) || dyn->insts[ninst].x86.need_flags==((A)|(B))))
+#define IFXN(A, B)  if(dyn->insts && (dyn->insts[ninst].x86.need_flags&(A) && !(dyn->insts[ninst].x86.need_flags&(B))))

 // Generate FCOM with s1 and s2 scratch regs (the VCMP is already done)
 #define FCOM(s1, s2)    \
    VMRS_APSR();    /* 0b0100011100000000 */                                    \
-    LDRH_IMM8(s2, xEmu, offsetof(x86emu_t, sw));   /*offset is 8bits right?*/   \
-    MOVW(s1, 0b0100011100000000);                                               \
-    BIC_REG_LSL_IMM8(s2, s2, s1, 0);                                            \
-    MOVW_COND(cVS, s1, 0b0100010100000000); /* unordered */                     \
-    MOVW_COND(cEQ, s1, 0b0100000000000000); /* zero */                          \
-    MOVW_COND(cGT, s1, 0b0000000000000000); /* greater than */                  \
-    MOVW_COND(cLO, s1, 0b0000000100000000); /* less than */                     \
-    ORR_REG_LSL_IMM8(s2, s2, s1, 0);                                            \
-    STRH_IMM8(s2, xEmu, offsetof(x86emu_t, sw))
+    LDRH_IMM8(s1, xEmu, offsetof(x86emu_t, sw));   /*offset is 8bits right?*/   \
+    BIC_IMM8(s1, s1, 0b01000111, 12);                                           \
+    ORR_IMM8_COND(cVS, s1, s1, 0b01000101, 12); /* unordered */                 \
+    ORR_IMM8_COND(cEQ, s1, s1, 0b01000000, 12); /* equal */                     \
+    ORR_IMM8_COND(cMI, s1, s1, 0b00000001, 12); /* less than */                 \
+    /* greater than leave 0 */                                                  \
+    STRH_IMM8(s1, xEmu, offsetof(x86emu_t, sw))

 // Generate FCOMI with s1 and s2 scratch regs (the VCMP is already done)
 #define FCOMI(s1, s2)    \
-    VMRS_APSR();    /* 0b111 */                             \
-    XOR_REG_LSL_IMM8(s2, s2, s2, 0);                        \
-    MOVW_COND(cVS, s1, 0b111); /* unordered */              \
-    MOVW_COND(cEQ, s1, 0b100); /* zero */                   \
-    MOVW_COND(cGT, s1, 0b000); /* greater than */           \
-    MOVW_COND(cLO, s1, 0b001); /* less than */              \
-    UBFX(s2, s1, 0, 1);                                     \
-    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_CF]));    \
-    UBFX(s2, s1, 1, 1);                                     \
-    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_PF]));    \
-    UBFX(s2, s1, 2, 1);                                     \
-    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_ZF]));    \
-    MOVW(s2, d_none);                                       \
-    STR_IMM9(s2, xEmu, offsetof(x86emu_t, df));
+    IFX(X_CF|X_PF|X_ZF|X_PEND) {                                            \
+        VMRS_APSR();    /* 0b111 */                                         \
+        BIC_IMM8(xFlags, xFlags, 0b1000101, 0);                             \
+        ORR_IMM8_COND(cVS, xFlags, xFlags, 0b01000101, 0); /* unordered */  \
+        ORR_IMM8_COND(cEQ, xFlags, xFlags, 0b01000000, 0); /* zero */       \
+        ORR_IMM8_COND(cMI, xFlags, xFlags, 0b00000001, 0); /* less than */  \
+        /* greater than leave 0 */                                          \
+    }                                                                       \
+    SET_DFNONE(s1);                                                         \
+    IFX(X_OF|X_PEND) {                                                      \
+        BFC(xFlags, F_OF, 1);                                               \
+    }                                                                       \
+    IFX(X_AF|X_PEND) {                                                      \
+        BFC(xFlags, F_AF, 1);                                               \
+    }                                                                       \
+    IFX(X_SF|X_PEND) {                                                      \
+        BFC(xFlags, F_SF, 1);                                               \
+    }                                                                       \


-#ifndef UFLAGS
-#define UFLAGS(A)  dyn->cleanflags=A
-#endif
-#ifndef USEFLAG
-// USEFLAG will check status of defered flags the call update flags if needed. x3 will be used
-#define USEFLAG(A)   \
-    if(!dyn->cleanflags) {  \
-        if(A) {             \
+#define SET_DFNONE(S)    if(!dyn->dfnone) {MOVW(S, d_none); STR_IMM9(S, xEmu, offsetof(x86emu_t, df)); dyn->dfnone=1;}
+#define SET_DF(S, N)     if(N) {MOVW(S, N); STR_IMM9(S, xEmu, offsetof(x86emu_t, df)); dyn->dfnone=0;} else SET_DFNONE(S)
+#define SET_NODF()          dyn->dfnone = 0
+#define SET_DFOK()          dyn->dfnone = 1
+
+#ifndef READFLAGS
+#define READFLAGS(A) \
+    if(((A)!=X_PEND) && dyn->state_flags!=SF_SET) {     \
+        if(dyn->state_flags!=SF_PENDING) {              \
            LDR_IMM9(x3, xEmu, offsetof(x86emu_t, df)); \
-            TSTS_REG_LSL_IMM8(x3, x3, 0);               \
-            i32 = (GETMARKF)-(dyn->arm_size+8);         \
-            Bcond(cEQ, i32);                            \
-            CALL_(UpdateFlags, -1, 0);                  \
-            MARKF;                                      \
-            dyn->cleanflags=1;                          \
-        }                   \
+            TSTS_REG_LSL_IMM5(x3, x3, 0);               \
+            j32 = (GETMARKF)-(dyn->arm_size+8);         \
+            Bcond(cEQ, j32);                            \
+        }                                               \
+        CALL_(UpdateFlags, -1, 0);                      \
+        MARKF;                                          \
+        dyn->state_flags = SF_SET;                      \
+        SET_DFOK();                                     \
    }
 #endif
+#ifndef SETFLAGS
+#define SETFLAGS(A, B)  \
+    if(dyn->state_flags!=SF_SET && B==SF_SUBSET && (dyn->insts[ninst].x86.need_flags&(~((A)|X_PEND)))) \
+        READFLAGS(dyn->insts[ninst].x86.need_flags&(~(A)));    \
+    dyn->state_flags = (B==SF_SUBSET)?SF_SET:B
+#endif
 #ifndef JUMP
 #define JUMP(A) 
 #endif
-#define BARRIER(A) if (dyn->insts && !dyn->insts[ninst].x86.barrier) dyn->insts[ninst].x86.barrier = A
-#define UFLAG_OP1(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, op1));}
-#define UFLAG_OP2(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, op2));}
-#define UFLAG_OP12(A1, A2) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A1, 0, offsetof(x86emu_t, op1));STR_IMM9(A2, 0, offsetof(x86emu_t, op2));}
-#define UFLAG_RES(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, res));}
-#define UFLAG_DF(r, A) if(dyn->insts && dyn->insts[ninst].x86.flags) {MOVW(r, A); STR_IMM9(r, 0, offsetof(x86emu_t, df));}
-#define UFLAG_IF if(dyn->insts && dyn->insts[ninst].x86.flags)
+#ifndef BARRIER
+#define BARRIER(A) 
+#endif
+#ifndef BARRIER_NEXT
+#define BARRIER_NEXT(A)
+#endif
+#define UFLAG_OP1(A) if(dyn->insts && dyn->insts[ninst].x86.need_flags) {STR_IMM9(A, xEmu, offsetof(x86emu_t, op1));}
+#define UFLAG_OP2(A) if(dyn->insts && dyn->insts[ninst].x86.need_flags) {STR_IMM9(A, xEmu, offsetof(x86emu_t, op2));}
+#define UFLAG_OP12(A1, A2) if(dyn->insts && dyn->insts[ninst].x86.need_flags) {STR_IMM9(A1, xEmu, offsetof(x86emu_t, op1));STR_IMM9(A2, 0, offsetof(x86emu_t, op2));}
+#define UFLAG_RES(A) if(dyn->insts && dyn->insts[ninst].x86.need_flags) {STR_IMM9(A, xEmu, offsetof(x86emu_t, res));}
+#define UFLAG_DF(r, A) if(dyn->insts && dyn->insts[ninst].x86.need_flags) {SET_DF(r, A)}
+#define UFLAG_IF if(dyn->insts && dyn->insts[ninst].x86.need_flags)
 #ifndef DEFAULT
-#define DEFAULT      BARRIER(2)
+#define DEFAULT      *ok = -1; BARRIER(2)
+#endif
+#ifndef NEW_BARRIER_INST
+#define NEW_BARRIER_INST
 #endif
-// Emit the LOCK mutex (x1, x2 and x3 are lost)
-#define LOCK        emit_lock(dyn, addr, ninst)
-// Emit the UNLOCK mutex (x1, x2 and x3 are lost)
-#define UNLOCK      emit_unlock(dyn, addr, ninst)

+#if STEP < 2
+#define PASS2IF(A, B) if(A)
+#elif STEP == 2
+#define PASS2IF(A, B) if(A) dyn->insts[ninst].pass2choice = B; if(dyn->insts[ninst].pass2choice == B)
+#else
+#define PASS2IF(A, B) if(dyn->insts[ninst].pass2choice == B)
+#endif

 void arm_epilog();
-void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
+void* arm_next(x86emu_t* emu, uintptr_t addr);

 #ifndef STEPNAME
 #define STEPNAME3(N,M) N##M
@@ -281,6 +351,7 @@ void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);

 #define dynarec00       STEPNAME(dynarec00)
 #define dynarec0F       STEPNAME(dynarec0F)
+#define dynarecFS       STEPNAME(dynarecFS)
 #define dynarecGS       STEPNAME(dynarecGS)
 #define dynarec66       STEPNAME(dynarec66)
 #define dynarec67       STEPNAME(dynarec67)
@@ -298,25 +369,86 @@ void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
 #define dynarecF30F     STEPNAME(dynarecF30F)

 #define geted           STEPNAME(geted_)
-#define fakeed          STEPNAME(fakeed_)
+#define geted16         STEPNAME(geted16_)
 #define jump_to_epilog  STEPNAME(jump_to_epilog_)
-#define jump_to_linker  STEPNAME(jump_to_linker_)
+#define jump_to_next    STEPNAME(jump_to_next_)
 #define ret_to_epilog   STEPNAME(ret_to_epilog_)
 #define retn_to_epilog  STEPNAME(retn_to_epilog_)
+#define iret_to_epilog  STEPNAME(iret_to_epilog_)
 #define call_c          STEPNAME(call_c_)
+#define grab_fsdata     STEPNAME(grab_fsdata_)
 #define grab_tlsdata    STEPNAME(grab_tlsdata_)
-#define isNativeCall    STEPNAME(isNativeCall_)
-#define emit_lock       STEPNAME(emit_lock)
-#define emit_unlock     STEPNAME(emit_unlock)
 #define emit_cmp8       STEPNAME(emit_cmp8)
 #define emit_cmp16      STEPNAME(emit_cmp16)
 #define emit_cmp32      STEPNAME(emit_cmp32)
 #define emit_cmp8_0     STEPNAME(emit_cmp8_0)
 #define emit_cmp16_0    STEPNAME(emit_cmp16_0)
 #define emit_cmp32_0    STEPNAME(emit_cmp32_0)
-#define emit_test8     STEPNAME(emit_test8)
+#define emit_test8      STEPNAME(emit_test8)
 #define emit_test16     STEPNAME(emit_test16)
 #define emit_test32     STEPNAME(emit_test32)
+#define emit_add32      STEPNAME(emit_add32)
+#define emit_add32c     STEPNAME(emit_add32c)
+#define emit_add8       STEPNAME(emit_add8)
+#define emit_add8c      STEPNAME(emit_add8c)
+#define emit_sub32      STEPNAME(emit_sub32)
+#define emit_sub32c     STEPNAME(emit_sub32c)
+#define emit_sub8       STEPNAME(emit_sub8)
+#define emit_sub8c      STEPNAME(emit_sub8c)
+#define emit_or32       STEPNAME(emit_or32)
+#define emit_or32c      STEPNAME(emit_or32c)
+#define emit_xor32      STEPNAME(emit_xor32)
+#define emit_xor32c     STEPNAME(emit_xor32c)
+#define emit_and32      STEPNAME(emit_and32)
+#define emit_and32c     STEPNAME(emit_and32c)
+#define emit_or8        STEPNAME(emit_or8)
+#define emit_or8c       STEPNAME(emit_or8c)
+#define emit_xor8       STEPNAME(emit_xor8)
+#define emit_xor8c      STEPNAME(emit_xor8c)
+#define emit_and8       STEPNAME(emit_and8)
+#define emit_and8c      STEPNAME(emit_and8c)
+#define emit_add16      STEPNAME(emit_add16)
+#define emit_add16c     STEPNAME(emit_add16c)
+#define emit_sub16      STEPNAME(emit_sub16)
+#define emit_sub16c     STEPNAME(emit_sub16c)
+#define emit_or16       STEPNAME(emit_or16)
+#define emit_or16c      STEPNAME(emit_or16c)
+#define emit_xor16      STEPNAME(emit_xor16)
+#define emit_xor16c     STEPNAME(emit_xor16c)
+#define emit_and16      STEPNAME(emit_and16)
+#define emit_and16c     STEPNAME(emit_and16c)
+#define emit_inc32      STEPNAME(emit_inc32)
+#define emit_inc16      STEPNAME(emit_inc16)
+#define emit_inc8       STEPNAME(emit_inc8)
+#define emit_dec32      STEPNAME(emit_dec32)
+#define emit_dec16      STEPNAME(emit_dec16)
+#define emit_dec8       STEPNAME(emit_dec8)
+#define emit_adc32      STEPNAME(emit_adc32)
+#define emit_adc32c     STEPNAME(emit_adc32c)
+#define emit_adc8       STEPNAME(emit_adc8)
+#define emit_adc8c      STEPNAME(emit_adc8c)
+#define emit_adc16      STEPNAME(emit_adc16)
+#define emit_adc16c     STEPNAME(emit_adc16c)
+#define emit_sbb32      STEPNAME(emit_sbb32)
+#define emit_sbb32c     STEPNAME(emit_sbb32c)
+#define emit_sbb8       STEPNAME(emit_sbb8)
+#define emit_sbb8c      STEPNAME(emit_sbb8c)
+#define emit_sbb16      STEPNAME(emit_sbb16)
+#define emit_sbb16c     STEPNAME(emit_sbb16c)
+#define emit_neg32      STEPNAME(emit_neg32)
+#define emit_neg16      STEPNAME(emit_neg16)
+#define emit_neg8       STEPNAME(emit_neg8)
+#define emit_shl32      STEPNAME(emit_shl32)
+#define emit_shl32c     STEPNAME(emit_shl32c)
+#define emit_shr32      STEPNAME(emit_shr32)
+#define emit_shr32c     STEPNAME(emit_shr32c)
+#define emit_sar32c     STEPNAME(emit_sar32c)
+#define emit_rol32c     STEPNAME(emit_rol32c)
+#define emit_ror32c     STEPNAME(emit_ror32c)
+#define emit_shrd32c    STEPNAME(emit_shrd32c)
+#define emit_shld32c    STEPNAME(emit_shld32c)
+
+#define emit_pf         STEPNAME(emit_pf)

 #define x87_do_push     STEPNAME(x87_do_push)
 #define x87_do_push_empty STEPNAME(x87_do_push_empty)
@@ -329,6 +461,7 @@ void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
 #define x87_stackcount  STEPNAME(x87_stackcount)
 #define x87_setround    STEPNAME(x87_setround)
 #define x87_restoreround STEPNAME(x87_restoreround)
+#define sse_setround    STEPNAME(sse_setround)
 #define mmx_get_reg     STEPNAME(mmx_get_reg)
 #define mmx_get_reg_empty STEPNAME(mmx_get_reg_empty)
 #define sse_get_reg     STEPNAME(sse_get_reg)
@@ -338,6 +471,8 @@ void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
 #define fpu_popcache    STEPNAME(fpu_popcache)
 #define fpu_reset       STEPNAME(fpu_reset)
 #define fpu_purgecache  STEPNAME(fpu_purgecache)
+#define x87_purgecache  STEPNAME(x87_purgecache)
+#define mmx_purgecache  STEPNAME(mmx_purgecache)
 #ifdef HAVE_TRACE
 #define fpu_reflectcache STEPNAME(fpu_reflectcache)
 #endif
@@ -350,17 +485,19 @@ void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
 /* 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, int* fixedaddress, uint32_t absmax, uint32_t mask);

-// Do the GETED, but don't emit anything...
-uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop);
+/* 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, int* fixedaddress, uint32_t absmax, uint32_t mask);
+

 // generic x86 helper
 void jump_to_epilog(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst);
-void jump_to_linker(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst);
+void jump_to_next(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst);
 void ret_to_epilog(dynarec_arm_t* dyn, int ninst);
 void retn_to_epilog(dynarec_arm_t* dyn, int ninst, int n);
-void call_c(dynarec_arm_t* dyn, int ninst, void* fnc, int reg, int ret, uint32_t mask);
+void iret_to_epilog(dynarec_arm_t* dyn, int ninst);
+void call_c(dynarec_arm_t* dyn, int ninst, void* fnc, int reg, int ret, uint32_t mask, int saveflags);
+void grab_fsdata(dynarec_arm_t* dyn, uintptr_t addr, int ninst, int reg);
 void grab_tlsdata(dynarec_arm_t* dyn, uintptr_t addr, int ninst, int reg);
-int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn);
 void emit_lock(dynarec_arm_t* dyn, uintptr_t addr, int ninst);
 void emit_unlock(dynarec_arm_t* dyn, uintptr_t addr, int ninst);
 void emit_cmp8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
@@ -372,16 +509,78 @@ void emit_cmp32_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
 void emit_test8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
 void emit_test16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
 void emit_test32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_add32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_add32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_add8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_add8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sub32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_sub32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sub8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_sub8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_or32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_or32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_xor32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_xor32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_and32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_and32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_or8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_or8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_xor8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_xor8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_and8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_and8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_add16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_add16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sub16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_sub16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_or16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_or16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_xor16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_xor16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_and16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_and16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_inc32(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_inc16(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_inc8(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_dec32(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_dec16(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_dec8(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_adc32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_adc32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_adc8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_adc8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_adc16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_adc16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sbb32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_sbb32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sbb8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_sbb8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sbb16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4, int save_s4);
+void emit_sbb16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_neg32(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_neg16(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_neg8(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_shl32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shl32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_shr32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shr32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sar32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_rol32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_ror32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_shrd32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int s3, int s4);
+void emit_shld32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int s3, int s4);
+
+void emit_pf(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);

 // x87 helper
 // 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);
 // 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, create the entry if needed
 int x87_get_cache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
 // get vfpu register for a x87 reg, create the entry if needed
@@ -396,12 +595,14 @@ void x87_reget_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
 int x87_setround(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 // Restore round flag
 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);

 //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
@@ -414,6 +615,8 @@ int sse_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a);
 void fpu_reset(dynarec_arm_t* dyn, int ninst);
 // purge the FPU cache (needs 3 scratch registers)
 void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+void x87_purgecache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int s1);
 #ifdef HAVE_TRACE
 void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 #endif
@@ -425,9 +628,9 @@ int fpu_get_single_reg(dynarec_arm_t* dyn, int ninst, int reg, int idx);
 // put back (if needed) the single reg in place
 void fpu_putback_single_reg(dynarec_arm_t* dyn, int ninst, int reg, int idx, int s);

-
 uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
 uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
 uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
 uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
 uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
@@ -444,4 +647,22 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
 uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
 uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);

+#if STEP < 2
+#define PASS2(A)
+#else
+#define PASS2(A)   A
+#endif
+
+#if STEP < 3
+#define PASS3(A)
+#else
+#define PASS3(A)   A
+#endif
+
+#if STEP < 3
+#define MAYUSE(A)   (void)A
+#else
+#define MAYUSE(A)   
+#endif
+
 #endif //__DYNAREC_ARM_HELPER_H__
@@ -19,50 +19,43 @@
 #include "arm_printer.h"
 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"
+#include "custommem.h"

 #ifndef STEP
 #error No STEP defined
 #endif

-#include "dynarec_arm_helper.h"
-
-void arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
+uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
 {
-    uint8_t nextop, opcode;
    int ok = 1;
    int ninst = 0;
    uintptr_t ip = addr;
-    uintptr_t natcall;
-    int retn;
-    uint8_t gd, ed;
-    int8_t i8;
-    int32_t i32, tmp;
-    uint8_t u8;
-    uint8_t gb1, gb2, eb1, eb2;
-    uint32_t u32;
+    uintptr_t init_addr = addr;
    int need_epilog = 1;
-    uint8_t wback, wb1, wb2;
-    int fixedaddress;
-    dyn->tablei = 0;
+    dyn->sons_size = 0;
    // Clean up (because there are multiple passes)
-    dyn->cleanflags = 0;
+    dyn->state_flags = 0;
+    dyn->dfnone = 0;
    fpu_reset(dyn, ninst);
    // ok, go now
    INIT;
    while(ok) {
        ip = addr;
+        if(dyn->insts && (dyn->insts[ninst].x86.barrier==1)) {
+            NEW_BARRIER_INST;
+        }
        NEW_INST;
        fpu_reset_scratch(dyn);
 #ifdef HAVE_TRACE
-        if(dyn->emu->dec && box86_dynarec_trace) {
-        if((dyn->emu->trace_end == 0) 
-            || ((ip >= dyn->emu->trace_start) && (ip <= dyn->emu->trace_end)))  {
+        if(my_context->dec && box86_dynarec_trace) {
+        if((trace_end == 0) 
+            || ((ip >= trace_start) && (ip < trace_end)))  {
                MESSAGE(LOG_DUMP, "TRACE ----\n");
                fpu_reflectcache(dyn, ninst, x1, x2, x3);
-                MOV32(1, ip);
-                STM(0, (1<<4)|(1<<5)|(1<<6)|(1<<7)|(1<<8)|(1<<9)|(1<<10)|(1<<11));
-                STR_IMM9(1, 0, offsetof(x86emu_t, ip));
-                MOV32(2, 1);
+                MOV32(x1, ip);
+                STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP));
+                STR_IMM9(x1, xEmu, offsetof(x86emu_t, ip));
+                MOVW(x2, 1);
                CALL(PrintTrace, -1, 0);
                MESSAGE(LOG_DUMP, "----------\n");
            }
@@ -75,14 +68,49 @@ void arm_pass(dynarec_arm_t* dyn, uintptr_t addr)

        if(dyn->insts && dyn->insts[ninst+1].x86.barrier) {
            fpu_purgecache(dyn, ninst, x1, x2, x3);
-            if(dyn->insts[ninst+1].x86.barrier!=2)
-                dyn->cleanflags = 0;
+            if(dyn->insts[ninst+1].x86.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))
+            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 && !isJumpTableDefault((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;
 }
@@ -2,15 +2,24 @@
 #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 UFLAGS(A)          {}
-#define USEFLAGS(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;    \
-        dynarec_log(LOG_INFO, "%p: Dynarec stopped because of Opcode %02X %02X %02X %02X %02X %02X %02X\n", \
-        ip, PKip(0),                   \
-        PKip(1), PKip(2), PKip(3),     \
-        PKip(4), PKip(5), PKip(6))
+#define DEFAULT                         \
+        --dyn->size;                    \
+        *ok = -1;                       \
+        if(box86_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");    \
+        }
@@ -1,15 +1,17 @@
 #define INIT    
 #define FINI     \
    dyn->insts[ninst].x86.addr = addr; \
-    dyn->insts[ninst].x86.flags = X86_FLAGS_USE; \
    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr;
 #define MESSAGE(A, ...)  
 #define EMIT(A)     
-#define UFLAGS(A)   dyn->insts[ninst+(A?1:0)].x86.flags = X86_FLAGS_CHANGE
-#define USEFLAG(A)  dyn->insts[ninst].x86.flags = X86_FLAGS_USE
-#define JUMP(A)     dyn->insts[ninst].x86.jmp = A
+#define READFLAGS(A)    dyn->insts[ninst].x86.use_flags = A
+#define SETFLAGS(A,B)   {dyn->insts[ninst].x86.set_flags = A; dyn->insts[ninst].x86.state_flags = B;}
+#define JUMP(A)         dyn->insts[ninst].x86.jmp = A
+#define BARRIER(A)      dyn->insts[ninst].x86.barrier = A
+#define BARRIER_NEXT(A) if(ninst<dyn->size) dyn->insts[ninst+1].x86.barrier = A
+
 #define NEW_INST \
    dyn->insts[ninst].x86.addr = ip; \
    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr;
 #define INST_EPILOG 
-#define INST_NAME(name) 
+#define INST_NAME(name)  
@@ -6,3 +6,4 @@
 #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
@@ -1,12 +1,34 @@
 #define INIT    
 #define FINI
 #define EMIT(A)     \
-    if(box86_dynarec_log>=LOG_DUMP) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (A), arm_print(A));} \
-    *(uint32_t*)(dyn->block) = A;   \
-    dyn->block += 4; dyn->arm_size += 4
+    if(box86_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (A), arm_print(A));} \
+    *(uint32_t*)(dyn->block) = A;       \
+    dyn->block += 4; dyn->arm_size += 4;\
+    dyn->insts[ninst].size2 += 4

-#define MESSAGE(A, ...)  dynarec_log(A, __VA_ARGS__);
+#define MESSAGE(A, ...)  if(box86_dynarec_dump) dynarec_log(LOG_NONE, __VA_ARGS__)
 #define NEW_INST        
 #define INST_EPILOG     
-#define INST_NAME(name) if(box86_dynarec_log>=LOG_DUMP) printf_x86_instruction(dyn->emu->dec, &dyn->insts[ninst].x86, name)
+#define INST_NAME(name) \
+    if(box86_dynarec_dump) {\
+        printf_x86_instruction(my_context->dec, &dyn->insts[ninst].x86, name); \
+        dynarec_log(LOG_NONE, "%s%p: %d emited opcodes, state=%d/%d, set=%X, use=%X, need=%X%s\n", \
+            (box86_dynarec_dump>1)?"\e[32m":"", \
+            (void*)(dyn->arm_start+dyn->insts[ninst].address),  \
+            dyn->insts[ninst].size/4,           \
+            dyn->insts[ninst].x86.state_flags,  \
+            dyn->state_flags,                   \
+            dyn->insts[ninst].x86.set_flags,    \
+            dyn->insts[ninst].x86.use_flags,    \
+            dyn->insts[ninst].x86.need_flags,   \
+            (box86_dynarec_dump>1)?"\e[m":"");  \
+    }
+
+#define NEW_BARRIER_INST                            \
+    if(ninst) {                                     \
+    dyn->sons_x86[dyn->sons_size] = (uintptr_t)ip;  \
+    dyn->sons_arm[dyn->sons_size] = dyn->block;     \
+    MESSAGE(LOG_DUMP, "----> potential Son here\n");\
+    ++dyn->sons_size;                               \
+    }

@@ -10,8 +10,14 @@ typedef struct instruction_arm_s {
    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_arm_t;

 typedef struct dynarec_arm_s {
@@ -21,21 +27,32 @@ typedef struct dynarec_arm_s {
    uintptr_t           start;      // start of the block
    uint32_t            isize;      // size in byte of x86 instructions included
    void*               block;      // memory pointer where next instruction is emited
+    uintptr_t           arm_start;  // start of the arm code
    int                 arm_size;   // size of emitted arm code
-    x86emu_t*           emu;
-    uintptr_t*          table;      // jump table
-    int                 tablesz;    // size of the jump table
-    int                 tablei;     // index
-    int                 cleanflags; // flags are clean (CMP or TST have been just executed)
+    int                 state_flags;// actual state for on-demand flags
    int                 x87cache[8];// cache status for the 8 x87 register behind the fpu stack
    int                 x87reg[8];  // reg used for x87cache entry
    int                 mmxcache[8];// cache status for the 8 MMX registers
    int                 ssecache[8];// cache status for the 8 SSE(2) registers
    int                 fpuused[24];// all 8..31 double reg from fpu, used by x87, sse and mmx
    int                 x87stack;   // cache stack counter
+    int                 x87count;   // number of x87 register used
+    int                 mmxcount;   // number of mmx register used (not both mmx and x87 at the same time)
    int                 fpu_scratch;// scratch counter
+    int                 fpu_extra_qscratch; // some opcode need an extra quad scratch register
    int                 fpu_reg;    // x87/sse/mmx reg counter
+    int                 dfnone;     // if defered flags is already set to df_none
+    uintptr_t*          next;       // variable array of "next" jump address
+    int                 next_sz;
+    int                 next_cap;
+    uintptr_t*          sons_x86;   // the x86 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);

 #endif //__DYNAREC_ARM_PRIVATE_H_
@@ -1,19 +1,35 @@
 #ifndef __DYNAREC_PRIVATE_H_
 #define __DYNAREC_PRIVATE_H_

-#define X86_FLAGS_NONE      0
-#define X86_FLAGS_CHANGE    1
-#define X86_FLAGS_USE       2
+// all flags for the use_flags field
+#define X_CF    (1<<0)
+#define X_PF    (1<<1)
+#define X_AF    (1<<2)
+#define X_ZF    (1<<3)
+#define X_SF    (1<<4)
+#define X_OF    (1<<5)
+#define X_ALL   ((1<<6)-1)
+#define X_PEND  (0x1000)
+
+// all state flags
+#define SF_UNKNOWN  0
+#define SF_SET      1
+#define SF_PENDING  2
+#define SF_MAYSET   3
+#define SF_SUBSET   4

 typedef struct instruction_x86_s {
    uintptr_t   addr;   //address of the instruction
    int32_t     size;   // size of the instruction
-    int         flags;  // flags for this instruction (see X86_FLAGS_XXXX)
    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
    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
 } instruction_x86_t;

 void printf_x86_instruction(zydis_dec_t* dec, instruction_x86_t* inst, const char* name);

-#endif //__DYNAREC_PRIVATE_H_
+#endif //__DYNAREC_PRIVATE_H_
@@ -0,0 +1,477 @@
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 op 0 N Q M 0 Vm VH\(op ? "SUB" : "ADD")\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 0 0 N Q M 1 Vm VQADD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 0 1 N Q M 0 Vm VRHADD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 0 Vn Vd 0 0 0 1 N Q M 1 Vm VAND<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 1 Vn Vd 0 0 0 1 N Q M 1 Vm VBIC<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 1 0 Vn Vd 0 0 0 1 N Q M 1 Vm @ m=n const char* op op,"MOV","ORR"@ m=n char lastop[6] !@ %@ lastop[0] = '\0'@ sprintf(lastop, ", %s", vecname[(q << 5) + 0x20 + n])@ @ V\op\<c> <Qd>, <Qm>\lastop\
+NEON 1 1 1 1 0 0 1 0 0 D 1 1 Vn Vd 0 0 0 1 N Q M 1 Vm VORN<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D @<2> Vn Vd 0 0 0 1 N Q M 1 Vm @ param=0b00=0b01=0b10 const char* op op,"EOR","BSL","BIT","BIF"@ @ V\op\<c> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 1 0 N Q M 1 Vm VQSUB<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 0 1 1 N Q M op Vm @ op=0 const char* strict strict,"T","E"@ @ VCG\strict\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 0 N Q M 0 Vm VSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 0 N Q M 1 Vm VQSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 1 N Q M 0 Vm VRSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 0 1 N Q M 1 Vm VQRSHL<c>.<dt> <Qd>, <Qm>, <Qn>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 0 N Q M op Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ V\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 1 N Q M 0 Vm VABD<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 0 1 1 1 N Q M 1 Vm VABA<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 op 0 D size Vn Vd 1 0 0 0 N Q M 0 Vm @ op=0 const char* opname opname,"ADD","SUB"@ @ V\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 0 0 N Q M 1 Vm VTST<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D size Vn Vd 1 0 0 0 N Q M 1 Vm VCEQ<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 0 1 N Q M 0 Vm @ set size = 8 << size@ @ VML\(u ? "S" : "A")\<c>.I\%dsize\ <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 0 1 N Q M 1 Vm VMUL<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 0 D size Vn Vd 1 0 1 0 N Q M op Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ VP\opname\<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 1 1 N Q M 0 Vm VQDMULH<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D size Vn Vd 1 0 1 1 N Q M 0 Vm VQRDMULH<c>.<dt> <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D size Vn Vd 1 0 1 1 N Q/=0 M 1 Vm VPADD<c>.<dt> <Dd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 0 D U sz/=0 Vn Vd 1 1 0 0 N Q M 1 Vm @ u=0 const char* y y,"A","S"@ @ VFM\y\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D U sz/=0 Vn Vd 1 1 0 1 N Q M 0 Vm @ u=0 const char* op op,"ADD","SUB"@ @ V\op\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D 0 sz/=0 Vn Vd 1 1 0 1 N Q/=0 M 0 Vm VPADD.F32 <Dd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 1 0 D 1 sz/=0 Vn Vd 1 1 0 1 N Q M 0 Vm VABD.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D op sz/=0 Vn Vd 1 1 0 1 N Q M 1 Vm VML\(op ? "S" : "A")\<c>.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D 0 sz/=0 Vn Vd 1 1 0 1 N Q M 1 Vm VMUL.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 sz/=0 Vn Vd 1 1 1 0 N Q M 0 Vm VCEQ.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 0 N Q M 0 Vm @ op=1 const char* strict strict,"T","E"@ @ VCG\strict\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 0 N Q M 1 Vm @ op=1 const char* strict strict,"T","E"@ @ VACG\strict\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D op sz/=0 Vn Vd 1 1 1 1 N Q M 0 Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ V\opname\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 0 D op sz/=0 Vn Vd 1 1 1 1 N Q M 0 Vm @ op=0 const char* opname opname,"MAX","MIN"@ @ VP\opname\.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 0 sz/=0 Vn Vd 1 1 1 1 N Q M 1 Vm VRECPS.F32 <Qd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 0 0 D 1 sz/=0 Vn Vd 1 1 1 1 N Q M 1 Vm VRSQRTS.F32 <Qd>, <Qn>, <Qm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 0 <23>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx0 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx1 0 Q op/=0 1 imm4 VORR<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x0 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x1 0 Q op/=0 1 imm4 VORR<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=11xx 0 Q op/=0 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx0 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=0xx1 0 Q op/=1 1 imm4 VBIC<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x0 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=10x1 0 Q op/=1 1 imm4 BIC<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=110x 0 Q op/=1 1 imm4 VMVN<c>.<dt> <Qd>, #<imm>
+NEON 1 1 1 1 0 0 1 i 1 D 0 0 0 imm3 Vd cmode/=1110 0 Q op/=1 1 imm4 VMOV<c>.<dt> <Qd>, #<imm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 0 0 0 <11> 0 <2> 1 <4>
+NEON 1 1 1 1 0 0 1 0 1 D 0 0 1 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S8 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D 0 1 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D 1 0 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.S32 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 0 0 1 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U8 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 0 1 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 0 0 0 0 0 Vd 1 0 1 0 0 0 M 1 Vm VMOVL<c>.U32 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 0 0 L Q M 1 Vm VSHR<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 0 1 L Q M 1 Vm VSRA<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 1 0 L Q M 1 Vm VRSHR<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 0 0 1 1 L Q M 1 Vm VRSRA<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D imm6 Vd 0 1 0 0 L Q M 1 Vm VSRI<c>.<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D Imm6 Vd 0 1 0 1 L Q M 1 Vm VSHL<c>.I<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D Imm6 Vd 0 1 0 1 L Q M 1 Vm VSLI<c>.<size> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D Imm6 Vd 0 1 1 op L Q M 1 Vm VQSHL\((u & op) ? "U" : "")\<c>.<dt> <Qd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D imm6 Vd 1 0 0 0 0 0 M 1 Vm @ set ++size@ @ VSHRN<c>.I<size> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 0 1 D imm6 Vd 1 0 0 0 0 1 M 1 Vm VRSHRN<c>.I<size> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 1 0 0 op 0 0 M 1 Vm VQSHR\((u & op) ? "U" : "")\N<c>.<dt> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 U 1 D imm6 Vd 1 0 0 op 0 1 M 1 Vm VQRSHR\((u & op) ? "U" : "")\N<c>.<dt> <Dd>, <Qm>, #<imm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 1 0 0 M 1 Vm VSHLL<c>.<dt> <Qd>, <Dm>, #\%d8 &l&l size\
+NEON 1 1 1 1 0 0 1 U 1 D Imm6 Vd 1 0 1 0 0 0 M 1 Vm VSHLL<c>.<dt> <Qd>, <Dm>, #<imm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 0 <9> 1 1 1 <1> 0 <2> 1 <4>
+NEON 1 1 1 1 0 0 1 0 1 D IMM6 Vd 1 1 1 0 0 Q M 1 Vm VCVT<c>.F32.S32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 0 1 D IMM6 Vd 1 1 1 1 0 Q M 1 Vm VCVT<c>.S32.F32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 1 1 D IMM6 Vd 1 1 1 0 0 Q M 1 Vm VCVT<c>.F32.U32 <Qd>, <Qm>, #<fbits>
+NEON 1 1 1 1 0 0 1 1 1 D IMM6 Vd 1 1 1 1 0 Q M 1 Vm VCVT<c>.U32.F32 <Qd>, <Qm>, #<fbits>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <18> 1 <4>
+NEON 1 1 1 1 0 0 1 0 1 D 1 1 Vn Vd IMM4 N Q M 0 Vm VEXT<c>.8 <Qd>, <Qn>, <Qm>, #<imm>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=11 0 0 <4> 0 0 0 <2> <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=10 0 0 <4> 0 0 0 <2>/11 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=10 0 0 <4> 0 0 0 <2>/01 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=01 0 0 <4> 0 0 0 <2>/11 <2> 0 <4>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <2>/=00 0 0 <4> 0 0 0 <2>/=11 <2> 0 <4>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 0 0 @<2> Q M 0 Vm VREV\%d16 &l&l param\<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 0 1 0 U Q M 0 Vm VPADDL<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 0 0 0 Q M 0 Vm VCLS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 0 0 1 Q M 0 Vm VCLZ<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 0 0 Vd 0 1 0 1 0 Q M 0 Vm VCNT<c>.8 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 0 0 Vd 0 1 0 1 1 Q M 0 Vm VMVN<c> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 0 U Q M 0 Vm VPADAL<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 1 0 Q M 0 Vm VQABS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 0 Vd 0 1 1 1 1 Q M 0 Vm VQNEG<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 0 0 Q M 0 Vm VCGT<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 0 1 Q M 0 Vm VCGE<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 1 0 Q M 0 Vm VCEQ<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 0 1 1 Q M 0 Vm VCLE<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 0 0 Q M 0 Vm VCLT<c>.<dt> <Qd>, <Qm>, #0
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 1 0 Q M 0 Vm VABS<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 0 1 Vd 0 F 1 1 1 Q M 0 Vm VNEG<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=00 1 0 Vd 0 0 0 0 0 Q M 0 Vm VSWP<c> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 0 1 Q M 0 Vm VTRN<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 1 0 Q M 0 Vm VUZP<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 0 1 1 Q M 0 Vm VZIP<c>.\%d8 &l&l size\ <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 0 0 M 0 Vm VMOVN<c>.<dt> <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 0 1 M 0 Vm VQMOVUN<c>.S\%d16 &l&l size\ <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 0 1 U M 0 Vm @ set ++size@ @ VQMOVN<c>.<dt> <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 0 Vd 0 0 1 1 0 0 M 0 Vm VSHLL<c>.I\%d8 &l&l size\ <Qd>, <Dm>, #\%d8 &l&l size\
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=01 1 0 Vd 0 1 1 0 0 0 M 0 Vm VCVT<c>.F16.F32 <Dd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=01 1 0 Vd 0 1 1 1 0 0 M 0 Vm VCVT<c>.F32.F16 <Qd>, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 1 Vd 0 1 0 F 0 Q M 0 Vm VRECPE<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size 1 1 Vd 0 1 0 F 1 Q M 0 Vm VRSQRTE<c>.<dt> <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 0 0 Q M 0 Vm VCVT<c>.F32.S32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 0 1 Q M 0 Vm VCVT<c>.S32.F32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 1 0 Q M 0 Vm VCVT<c>.F32.U32 <Qd>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 size/=10 1 1 Vd 0 1 1 1 1 Q M 0 Vm VCVT<c>.U32.F32 <Qd>, <Qm>
+INVALIDATE 1 1 1 1 0 0 1 1 1 <1> 1 1 <8> 0 <6> 0 <4>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 Vn Vd 1 0 @<2> N op M 0 Vm @ param=0b00 char list[10] !@ sprintf(list, %)@ "{%s}", vecname[0x20 + n]@ "{%s-%s}", vecname[0x20 + n], vecname[0x20 + n + param]@ @ VTB\(op ? "X" : "L")\<c>.8 <Dd>, \list\, <Dm>
+NEON 1 1 1 1 0 0 1 1 1 D 1 1 imm4 Vd 1 1 0 0 0 Q M 0 Vm VDUP<c>.\%dsize\ <Qd>, <Dm[x]>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <1> 1 1 <13> 0 <1> 0 <4>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 0 0 Q N 0 M 0 Vm VADD\(q ? "W" : "L")\<c>.<dt> <qd>, <Qn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 0 1 Q N 0 M 0 Vm VSUB\(q ? "W" : "L")\<c>.<dt> <qd>, <Qn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 1 0 0 N 0 M 0 Vm @ set size = 16 << size@ @ VADDHN<c>.I\%dsize\ <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D size Vn Vd 0 1 0 0 N 0 M 0 Vm VRADDHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 1 0 1 N 0 M 0 Vm VABAL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 1 1 0 N 0 M 0 Vm VSUBHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 1 1 D size Vn Vd 0 1 1 0 N 0 M 0 Vm VRSUBHN<c>.<dt> <Dd>, <Qn>, <Qm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 1 1 1 N 0 M 0 Vm VABDL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 0 op 0 N 0 M 0 Vm VML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 0 op 1 N 0 M 0 Vm VQDML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 1 0 0 N 0 M 0 Vm VMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 1 0 1 N 0 M 0 Vm VQDMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+NEON 1 1 1 1 0 0 1 U/=0 1 D size/=00 Vn Vd 1 1 1 0 N 0 M 0 Vm @ set uint8_t size = 0@ @ VMULL<c>.<dt> <Qd>, <Dn>, <Dm>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <16> 0 <1> 0 <4>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 0 op 0 F N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VML\(op ? "S" : "A")\<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 0 op 1 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 0 op 1 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDML\(op ? "S" : "A")\L<c>.<dt> <Qd>, <Dn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 0 0 F N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VMUL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 U 1 D size Vn Vd 1 0 1 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VMULL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 0 1 D size Vn Vd 1 0 1 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDMULL<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 1 0 0 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQDMULH<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+NEON 1 1 1 1 0 0 1 Q 1 D size Vn Vd 1 1 0 1 N 1 M 0 Vm @ size=0b01=0b10 unsigned int decodedImm !@ decodedImm = m >> %; m %@ 3@ &= 0x7@ 4@ &= 0xF@ 5@ = 0x21@ @ VQRDMULH<c>.<dt> <Qd>, <Qn>, <Dm[x]>
+INVALIDATE 1 1 1 1 0 0 1 <1> 1 <16> 0 <1> 0 <4>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 0 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 0 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 0 1 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 0 1 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 0 0 Rn Vd 1 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 0 0 <8> 1 1 <10>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> 1 1 1 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> 1 1 0 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 0 @<4> Rm @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VST1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VST3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 0 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VST4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\, \vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\, \vecname[0x26 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 0 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x23 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 0 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 0 1 size @<2> Rm @ param=0b00 char align[4] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 63 + param)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\, \vecname[0x24 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\-\vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 0 1 1 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 0 1 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x22 + d]\}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> 1 1 1 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> 1 1 0 1 @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 0 D 1 0 Rn Vd 1 0 1 0 size @<2> Rm @ param=0b00 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 32 << param)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\, \vecname[0x21 + d]\}, [<Rn>\align\], <Rm>
+INVALIDATE 1 1 1 1 0 1 0 0 1 <1> 1 0 <8> 1 1 <10>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> 1 1 1 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> 1 1 0 1 @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 0 @<4> Rm @ set uint8_t temp = (2 << size) - 1@ set uint8_t temp2 = param & temp@ temp2=0=param>>1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 8 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ @ VLD1<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0=1 char align[5] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", 16 << size)@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ set uint8_t temp = param & (1 << size)@ temp=0 uint8_t spacing spacing,0,1@ set uint8_t x = param >> (size + 1)@ @ VLD3<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\]}, [<Rn>], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 1 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> 1 1 0 1 @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd size 0 1 @<4> Rm @ size=0b00=0b01 uint8_t temp temp,1,1,3@ set uint8_t temp2 = param & temp@ temp2=0b0=0b1=0b10 char align[6] !@ %@ align[0] = '\0'@ sprintf(align, ":%d", (size ? 64 : 32))@ sprintf(align, ":128")@ sprintf(align, "???")@ set uint8_t x = param >> (size + 1)@ set int spacing = size ? ((param >> size) & 1) : 0@ @ VLD4<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[\%dx\], \vecname[0x21 + d + spacing]\[\%dx\], \vecname[0x22 + d + 2*spacing]\[\%dx\], \vecname[0x23 + d + 3*spacing]\[\%dx\]}, [<Rn>\align\], <Rm>
+INVALIDATE 1 1 1 1 0 1 0 0 <3> 0 <20>
+INVALIDATE 1 1 1 1 <28>
+ARMS cond 0 0 0 0 0 0 0 S Rn Rd imm5 type 0 Rm AND{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 0 0 S Rd (0) (0) (0) (0) Rm 1 0 0 1 Rn MUL{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRH<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRD<c> <Rt>, <Rt2>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 0 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm STRD<c> <Rt>, <Rt2>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRH<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSB<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 0 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSH<c> <Rt>, [<Rn>], +/-<Rm>
+ARMS cond 0 0 0 0 0 0 1 S Rn Rd imm5 type 0 Rm EOR{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 0 1 S Rd Ra Rm 1 0 0 1 Rn MLA{S}<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 0 U 0 1 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSBT<c> <Rt>, [<Rn>], +/-<Rm>
+ARM_ cond 0 0 0 0 U 0 1 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSHT<c> <Rt>, [<Rn>], +/-<Rm>
+ARMS cond 0 0 0 0 0 1 0 S Rn Rd imm5 type 0 Rm SUB{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 1 0 0 RdHi RdLo Rm 1 0 0 1 Rn UMAAL<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 0 1 1 imm4L STRH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 1 0 1 imm4L LDRD<c> <Rt>, <Rt2>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 0 Rn Rt imm4H 1 1 1 1 imm4L STRD<c> <Rt>, <Rt2>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 0 1 1 imm4L LDRH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSB<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 0 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSH<c> <Rt>, [<Rn>], #+/-<imm8>
+ARMS cond 0 0 0 0 0 1 1 S Rn Rd imm5 type 0 Rm RSB{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 0 1 1 0 Rd Ra Rm 1 0 0 1 Rn MLS<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 0 U 1 1 0 Rn Rt imm4H 1 0 1 1 imm4L STRHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 0 1 1 imm4L LDRHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSBT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARM_ cond 0 0 0 0 U 1 1 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSHT<c> <Rt>, [<Rn>], #+/-<imm8>
+ARMS cond 0 0 0 0 1 0 0 S Rn Rd imm5 type 0 Rm ADD{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 0 0 S RdHi RdLo Rm 1 0 0 1 Rn UMULL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 0 1 S Rn Rd imm5 type 0 Rm ADC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 0 1 S RdHi RdLo Rm 1 0 0 1 Rn UMLAL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 1 0 S Rn Rd imm5 type 0 Rm SBC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 1 0 S RdHi RdLo Rm 1 0 0 1 Rn SMULL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARMS cond 0 0 0 0 1 1 1 S Rn Rd imm5 type 0 Rm RSC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 0 1 1 1 S RdHi RdLo Rm 1 0 0 1 Rn SMLAL{S}<c> <RdLo>, <RdHi>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 0 0 0 Rn SMLABB<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 0 1 0 Rn SMLATB<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 1 0 0 Rn SMLABT<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rd Ra Rm 1 1 1 0 Rn SMLATT<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 0 0 1 0 0 0 0 Rn Rt (0) (0) (0) (0) 1 0 0 1 Rm SWP<c> <Rt>, <Rm>, [<Rn>]
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm STRH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRD<c> <Rt>, <Rt2>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 0 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm STRD<c> <Rt>, <Rt2>, [<Rn>, +/-<Rm>]{!}
+ARMS cond 0 0 0 1 0 0 0 1 Rn (0) (0) (0) (0) imm5 type 0 Rm TST<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 0 1 1 Rm LDRH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 1 0 1 Rm LDRSB<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 U 0 W 1 Rn Rt (0) (0) (0) (0) 1 1 1 1 Rm LDRSH<c> <Rt>, [<Rn>, +/-<Rm>]{!}
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 0 1 Rm BX<c> <Rm>
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 1 0 Rm BXJ<c> <Rm>
+ARM_ cond 0 0 0 1 0 0 1 0 (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) 0 0 1 1 Rm BLX<c> <Rm>
+ARM_ cond/=1110 0 0 0 1 0 0 1 0 imm12 0 1 1 1 imm4 BKPT #<imm16>
+ARMS cond 0 0 0 1 0 0 1 1 Rn (0) (0) (0) (0) imm5 type 0 Rm TEQ<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 0 1 0 0 Rn Rt (0) (0) (0) (0) 1 0 0 1 Rm SWPB<c> <Rt>, <Rm>, [<Rn>]
+ARM_ cond 0 0 0 1 0 1 1 0 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 0 1 Rm CLZ<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 0 1 1 imm4L STRH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 1 0 1 imm4L LDRD<c> <Rt>, <Rt2>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 0 Rn Rt imm4H 1 1 1 1 imm4L STRD<c> <Rt>, <Rt2>, [<Rn>, #+/-<imm8>]{!}
+ARMS cond 0 0 0 1 0 1 0 1 Rn (0) (0) (0) (0) imm5 type 0 Rm CMP<c> <Rn>, <Rm>{, <shift>}
+ARMS cond 0 0 0 1 0 1 1 1 Rn (0) (0) (0) (0) imm5 type 0 Rm CMN<c> <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 0 1 1 imm4L LDRH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 1 0 1 imm4L LDRSB<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARM_ cond 0 0 0 1 U 1 W 1 Rn Rt imm4H 1 1 1 1 imm4L LDRSH<c> <Rt>, [<Rn>, #+/-<imm8>]{!}
+ARMS cond 0 0 0 1 1 0 0 S Rn Rd imm5 type 0 Rm ORR{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 0 0 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREX<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 0 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREX<c> <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd 0 0 0 0 0 0 0 0 Rm MOV{S}<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 0 0 0 Rm LSL{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 0 0 1 Rn LSL{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 0 1 0 Rm LSR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 0 1 1 Rn LSR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 1 0 0 Rm ASR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 1 0 1 Rn ASR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd 0 0 0 0 0 1 1 0 Rm RRX{S}<c> <Rd>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd imm5 1 1 0 Rm ROR{S}<c> <Rd>, <Rm>, #<imm>
+ARM_ cond 0 0 0 1 1 0 1 S (0) (0) (0) (0) Rd Rm 0 1 1 1 Rn ROR{S}<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 0 0 1 1 0 1 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXD<c> <Rd>, <Rt>, <Rt2>, [<Rn>]
+ARM_ cond 0 0 0 1 1 0 1 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXD<c> <Rt>, <Rt2>, [<Rn>]
+ARMS cond 0 0 0 1 1 1 0 S Rn Rd imm5 type 0 Rm BIC{S}<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 1 0 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXB<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 1 0 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXB<c> <Rt>, [<Rn>]
+ARMS cond 0 0 0 1 1 1 1 S (0) (0) (0) (0) Rd imm5 type 0 Rm MVN{S}<c> <Rd>, <Rm>{, <shift>}
+ARM_ cond 0 0 0 1 1 1 1 0 Rn Rd (1) (1) (1) (1) 1 0 0 1 Rt STREXH<c> <Rd>, <Rt>, [<Rn>]
+ARM_ cond 0 0 0 1 1 1 1 1 Rn Rt (1) (1) (1) (1) 1 0 0 1 (1) (1) (1) (1) LDREXH<c> <Rt>, [<Rn>]
+ARM_ cond 0 0 1 0 0 0 0 S Rn Rd imm12 AND{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 0 1 S Rn Rd imm12 EOR{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 1 0 S Rn Rd imm12 SUB{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 0 1 1 S Rn Rd imm12 RSB{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 0 0 S Rn Rd imm12 ADD{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 0 1 S Rn Rd imm12 ADC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 1 0 S Rn Rd imm12 SBC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 0 1 1 1 S Rn Rd imm12 RSC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 0 0 0 imm4 Rd imm12 MOVW<c> <Rd>, #<imm16>
+ARM_ cond 0 0 1 1 0 0 0 1 Rn (0) (0) (0) (0) imm12 TST<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 0 1 0 0 0 0 0 (1) (1) (1) (1) (0) (0) (0) (0) 0 0 0 0 0 0 0 0 NOP<c>
+ARM_ cond 0 0 1 1 0 0 1 1 Rn (0) (0) (0) (0) imm12 TEQ<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 1 0 0 imm4 Rd imm12 MOVT<c> <Rd>, #<imm16>
+ARM_ cond 0 0 1 1 0 1 0 1 Rn (0) (0) (0) (0) imm12 CMP<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 0 1 1 1 Rn (0) (0) (0) (0) imm12 CMN<c> <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 0 0 S Rn Rd imm12 ORR{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 0 1 S (0) (0) (0) (0) Rd imm12 MOV{S}<c> <Rd>, #<const>
+ARM_ cond 0 0 1 1 1 1 0 S Rn Rd imm12 BIC{S}<c> <Rd>, <Rn>, #<const>
+ARM_ cond 0 0 1 1 1 1 1 S (0) (0) (0) (0) Rd imm12 MVN{S}<c> <Rd>, #<const>
+ARM_ cond 0 1 0 0 1 0 0 1 1 1 0 1 Rt 0 0 0 0 0 0 0 0 0 1 0 0 POP<c> {<Rt>}
+ARM_ cond 0 1 0 0 U 0 0 0 Rn Rt imm12 STR<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 0 1 Rn Rt imm12 LDR<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 1 0 Rn Rt imm12 STRT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 0 1 1 Rn Rt imm12 LDRT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 0 0 Rn Rt imm12 STRB<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 0 1 Rn Rt imm12 LDRB<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 1 0 Rn Rt imm12 STRBT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 0 U 1 1 1 Rn Rt imm12 LDRBT<c> <Rt>, [<Rn>], #+/-<imm12>
+ARM_ cond 0 1 0 1 0 0 1 0 1 1 0 1 Rt 0 0 0 0 0 0 0 0 0 1 0 0 PUSH<c> {<Rt>}
+ARM_ cond 0 1 0 1 U 0 W 0 Rn Rt imm12 STR<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 0 W 1 Rn Rt imm12 LDR<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 1 W 0 Rn Rt imm12 STRB<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 0 1 U 1 W 1 Rn Rt imm12 LDRB<c> <Rt>, [<Rn>, #+/-<imm12>]{!}
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 0 1 0 1 Rm SSAX<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 0 1 1 1 Rm SSUB16<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 0 0 1 Rn Rd (1) (1) (1) (1) 1 1 1 1 Rm SSUB8<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 0 1 0 1 Rm USAX<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 0 1 1 1 Rm USUB16<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 0 0 1 0 1 Rn Rd (1) (1) (1) (1) 1 1 1 1 Rm USUB8<c> <Rd>, <Rn>, <Rm>
+ARMs cond 0 1 1 0 1 0 0 0 Rn Rd imm5 0 0 1 Rm PKHBT<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARMs cond 0 1 1 0 1 0 0 0 Rn Rd imm5 1 0 1 Rm PKHTB<c> <Rd>, <Rn>, <Rm>{, <shift>}
+ARM_ cond 0 1 1 0 1 0 0 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTB16<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 0 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAB16<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 1 0 1 sat_imm Rd imm5 sh 0 1 Rn SSAT<c> <Rd>, #<imm>, <Rn>{, <shift>}
+ARM_ cond 0 1 1 0 1 0 1 0 sat_imm Rd (1) (1) (1) (1) 0 0 1 1 Rn SSAT16<c> <Rd>, #<imm>, <Rn>
+ARM_ cond 0 1 1 0 1 0 1 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTB<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAB<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 1 1 Rm REV<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 0 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 1 0 1 1 Rm REV16<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 0 1 1 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm SXTH<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 0 1 1 Rn Rd rotate (0) (0) 0 1 1 1 Rm SXTAH<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 0 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTB16<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 0 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAB16<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 1 1 1 sat_imm Rd imm5 sh 0 1 Rn USAT<c> <Rd>, #<imm>, <Rn>{, <shift>}
+ARM_ cond 0 1 1 0 1 1 1 0 sat_imm Rd (1) (1) (1) (1) 0 0 1 1 Rn USAT16<c> <Rd>, #<imm>, <Rn>
+ARM_ cond 0 1 1 0 1 1 1 0 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTB<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 0 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAB<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 0 0 1 1 Rm RBIT<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 1 1 1 1 1 1 1 Rd rotate (0) (0) 0 1 1 1 Rm UXTH<c> <Rd>, <Rm>{, <rotation>}
+ARM_ cond 0 1 1 0 1 1 1 1 (1) (1) (1) (1) Rd (1) (1) (1) (1) 1 0 1 1 Rm REVSH<c> <Rd>, <Rm>
+ARM_ cond 0 1 1 0 1 1 1 1 Rn Rd rotate (0) (0) 0 1 1 1 Rm UXTAH<c> <Rd>, <Rn>, <Rm>{, <rotation>}
+ARMs cond 0 1 1 0 U 0 0 0 Rn Rt imm5 type 0 Rm STR<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 0 1 Rn Rt imm5 type 0 Rm LDR<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 1 0 Rn Rt imm5 type 0 Rm STRT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 0 1 1 Rn Rt imm5 type 0 Rm LDRT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 0 0 Rn Rt imm5 type 0 Rm STRB<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 0 1 Rn Rt imm5 type 0 Rm LDRB<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 1 0 Rn Rt imm5 type 0 Rm STRBT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 0 U 1 1 1 Rn Rt imm5 type 0 Rm LDRBT<c> <Rt>, [<Rn>], +/-<Rm>{, <shift>}
+ARMs cond 0 1 1 1 U 0 W 0 Rn Rt imm5 type 0 Rm STR<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 0 W 1 Rn Rt imm5 type 0 Rm LDR<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 1 W 0 Rn Rt imm5 type 0 Rm STRB<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARMs cond 0 1 1 1 U 1 W 1 Rn Rt imm5 type 0 Rm LDRB<c> <Rt>, [<Rn>, +/-<Rm>{, <shift>}]{!}
+ARM_ cond 0 1 1 1 1 0 0 0 Rd 1 1 1 1 Rm 0 0 0 1 Rn USAD8<c> <Rd>, <Rn>, <Rm>
+ARM_ cond 0 1 1 1 1 0 0 0 Rd Ra Rm 0 0 0 1 Rn USADA8<c> <Rd>, <Rn>, <Rm>, <Ra>
+ARM_ cond 0 1 1 1 1 1 0 msb Rd lsb 0 0 1 1 1 1 1 BFC<c> <Rd>, #<lsb>, #<width>
+ARM_ cond 0 1 1 1 1 1 0 msb Rd lsb 0 0 1 Rn BFI<c> <Rd>, <Rn>, #<lsb>, #<width>
+ARM_ cond 0 1 1 1 1 1 1 widthm1 Rd lsb 1 0 1 Rn UBFX<c> <Rd>, <Rn>, #<lsb>, #<width>
+ARM_ cond 1 0 0 0 0 0 W 0 Rn/=1101 register_list STMED<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 0 Rn register_list STMDA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 1 Rn/=1101 register_list LDMED<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 0 0 W 1 Rn register_list LDMDA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 W 0 Rn/=1101 register_list STMEA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 W 0 Rn register_list STM<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 0 1 0 0 1 Rn/=1101 register_list LDMEA<c> <Rn>, <registers>
+ARM_ cond 1 0 0 0 1 0 1 1 1 1 0 1 register_list @ = const char* op @@ (reglist != 0) && ((reglist & (reglist - 1)) == 0)@ op,"POP","LDMEA"@ @ \op\<c> <registers>
+ARM_ cond 1 0 0 0 1 0 W 1 Rn register_list LDM<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 0 0 Rn/=1101 register_list STMFD<c> <Rn>, <registers>
+ARM_ cond 1 0 0 1 0 0 1 0 1 1 0 1 register_list @ = const char* op @@ (reglist != 0) && ((reglist & (reglist - 1)) == 0)@ op,"PUSH","STMFD"@ @ \op\<c> <registers>
+ARM_ cond 1 0 0 1 0 0 W 0 Rn register_list STMDB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 W 1 Rn/=1101 register_list LDMFD<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 0 0 W 1 Rn register_list LDMDB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 0 Rn/=1101 register_list STMFA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 0 Rn register_list STMIB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 1 Rn/=1101 register_list LDMFA<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 0 1 1 0 W 1 Rn register_list LDMIB<c> <Rn>{!}, <registers>
+ARM_ cond 1 0 1 0 imm24 B<c> <label>
+ARM_ cond 1 0 1 1 imm24 BL<c> <label>
+VFPU cond 1 1 0 0 0 1 0 0 Rn Rt 1 0 1 0 0 0 M 1 Vm VMOV<c> <Sm>, <Sm1>, <Rt>, <Rn>
+VFPU cond 1 1 0 0 0 1 0 1 Rn Rt 1 0 1 0 0 0 M 1 Vm VMOV<c> <Rt>, <Rn>, <Sm>, <Sm1>
+VFPU cond 1 1 0 0 0 1 0 0 Rn Rt 1 0 1 1 0 0 M 1 Vm VMOV<c> <Dm>, <Rt>, <Rn>
+VFPU cond 1 1 0 0 0 1 0 1 Rn Rt 1 0 1 1 0 0 M 1 Vm VMOV<c> <Rt>, <Rn>, <Dm>
+INVALIDATE <4> 1 1 0 0 0 <11> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 0 0 <11> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 1 1 <1> 1 <9> 1 0 1 <9>
+INVALIDATE <4> 1 1 0 1 1 <1> 1 <9> 1 0 1 <9>
+VFPU cond 1 1 0 1 0 D 1 0 1 1 0 1 Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FSTMX<c> <LIst>
+VFPU cond 1 1 0 1 0 D 1 0 1 1 0 1 Vd 1 0 1 sz imm8 VPUSH<c> <list>
+VFPU cond 1 1 0 1 U D 0 0 Rn Vd 1 0 1 sz imm8 VSTR<c> <Dd>, [<Rn>{, #+/-<imm>}]
+VFPU cond 1 1 0 @<2> D W 0 Rn Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FSTM\(param == 0b01 ? "IA" : "DB")\X<c> <Rn>{!}, <LIst>
+VFPU cond 1 1 0 @<2> D W 0 Rn Vd 1 0 1 sz imm8 VSTM\(param == 0b01 ? "IA" : "DB")\<c> <Rn>{!}, <list>
+VFPU cond 1 1 0 0 1 D 1 1 1 1 0 1 Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FLDMX<c> <LIst>
+VFPU cond 1 1 0 0 1 D 1 1 1 1 0 1 Vd 1 0 1 sz imm8 VPOP<c> <list>
+VFPU cond 1 1 0 1 0 D 0 1 1 1 1 1 Vd 1 0 1 sz 0 0 0 0 0 0 0 0 VLDR<c> <Dd>, [PC, #-0]
+VFPU cond 1 1 0 1 U D 0 1 1 1 1 1 Vd 1 0 1 sz imm8 VLDR<c> <Dd>, <label>
+VFPU cond 1 1 0 1 U D 0 1 Rn Vd 1 0 1 sz imm8 VLDR<c> <Dd>, [<Rn>{, #+/-<imm>}]
+VFPU cond 1 1 0 @<2> D W 1 Rn Vd 1 0 1 1 imm8/01 @ set --imm8@ @ FLDM\(param == 0b01 ? "IA" : "DB")\X<c> <Rn>{!}, <LIst>
+VFPU cond 1 1 0 @<2> D W 1 Rn Vd 1 0 1 sz imm8 VLDM\(param == 0b01 ? "IA" : "DB")\<c> <Rn>{!}, <list>
+INVALIDATE <4> 1 1 0 <25>
+INVALIDATE <4> 1 1 1 0 0 0 <1> 0 <8> 1 0 1 1 <1> 1 0 1 (0) (0) (0) (0)
+VFPU cond 1 1 1 0 0 0 0 0 Vn Rt 1 0 1 0 N (0) (0) 1 (0) (0) (0) (0) VMOV<c> <Sn>, <Rt>
+VFPU cond 1 1 1 0 1 1 1 0 0 0 0 1 Rt 1 0 1 0 (0) (0) (0) 1 (0) (0) (0) (0) VMSR<c> FPSCR, <Rt>
+VFPU cond 1 1 1 0 0 @<2> 0 Vd Rt 1 0 1 1 D @<2> 1 (0) (0) (0) (0) @ set uint8_t opc = (param1_2 << 2) + param2_2@ = uint8_t shift @@ (opc & 0b1000) == 0b1000@ @@ (opc & 0b1001) == 0b0001@ @@ (opc & 0b1011) == 0b0000@ shift,0,1,2,0xFF@ set uint8_t decodedImm = (opc & 0x7) >> shift@ @ VMOV<c>.\%d8 &l&l shift\ <Dd[x]>, <Rt>
+VFPU cond 1 1 1 0 1 0 Q 0 Vd Rt 1 0 1 1 D 0 0 1 (0) (0) (0) (0) VDUP<c>.32 <Qd>, <Rt>
+VFPU cond 1 1 1 0 1 0 Q 0 Vd Rt 1 0 1 1 D 0 1 1 (0) (0) (0) (0) VDUP<c>.16 <Qd>, <Rt>
+VFPU cond 1 1 1 0 1 1 Q 0 Vd Rt 1 0 1 1 D 0 0 1 (0) (0) (0) (0) VDUP<c>.8 <Qd>, <Rt>
+VFPU cond 1 1 1 0 0 0 0 1 Vn Rt 1 0 1 0 N (0) (0) 1 (0) (0) (0) (0) VMOV<c> <Rt>, <Sn>
+VFPU cond 1 1 1 0 1 1 1 1 0 0 0 1 Rt 1 0 1 0 (0) (0) (0) 1 (0) (0) (0) (0) VMRS<c> <Rt>, FPSCR
+VFPU cond 1 1 1 0 U @<2> 1 Vn Rt 1 0 1 1 N @<2> 1 (0) (0) (0) (0) @ set uint8_t opc = (u << 4) + (param1_2 << 2) + param2_2@ = uint8_t shift, size @@ (opc & 0b01000) == 0b01000@ @@ (opc & 0b01001) == 0b00001@ @@ (opc & 0b11011) == 0b00000@ shift,0,1,2,0xFF;size,0,1,2,4@ set uint8_t decodedImm = (opc & 0x7) >> shift@ @ VMOV<c>.<dt> <Rt>, <Dn[x]>
+INVALIDATE <4> 1 1 1 0 <12> 1 0 1 <4> 1 <4>
+VFPU cond 1 1 1 0 0 D 0 0 Vn Vd 1 0 1 sz N op M 0 Vm VML\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 0 1 Vn Vd 1 0 1 sz N op M 0 Vm VNML\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 0 Vn Vd 1 0 1 sz N 0 M 0 Vm VMUL<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 0 Vn Vd 1 0 1 sz N 1 M 0 Vm VNMUL<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 1 Vn Vd 1 0 1 sz N 0 M 0 Vm VADD<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 0 D 1 1 Vn Vd 1 0 1 sz N 1 M 0 Vm VSUB<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 0 0 Vn Vd 1 0 1 sz N 0 M 0 Vm VDIV<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 0 1 Vn Vd 1 0 1 sz N op M 0 Vm VFNM\(op ? "A" : "S")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 0 Vn Vd 1 0 1 sz N op M 0 Vm VFM\(op ? "S" : "A")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dn>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 @<4> Vd 1 0 1 0 (0) 0 (0) 0 @<4> @ set int8_t sign = (param1_4 >> 3) ? -1 : 1@ set uint8_t exp = ((param1_4 & 0b111) ^ 0b100)@ set float mantissa = 16 + param2_4@ set float imm = sign * (1 << exp) * mantissa / 128.@ @ VMOV<c>.F32 <Sd>, #\%fimm\
+VFPU cond 1 1 1 0 1 D 1 1 @<4> Vd 1 0 1 1 (0) 0 (0) 0 @<4> @ set int8_t sign = (param1_4 >> 3) ? -1 : 1@ set uint8_t exp = ((param1_4 & 0b111) ^ 0b100)@ set double mantissa = 16 + param2_4@ set double imm = sign * (1 << exp) * mantissa / 128.@ @ VMOV<c>.F64 <Dd>, #\%fimm\
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 0 Vd 1 0 1 sz 0 1 M 0 Vm VMOV<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 0 Vd 1 0 1 sz 1 1 M 0 Vm VABS<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 1 Vd 1 0 1 sz 0 1 M 0 Vm VNEG<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 0 1 Vd 1 0 1 sz 1 1 M 0 Vm VSQRT<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 1 0 Vd 1 0 1 (0) @<1> 1 M 0 Vm VCVT\(param ? "T" : "B")\<c>.F32.F16 <Sd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 0 1 1 Vd 1 0 1 (0) @<1> 1 M 0 Vm VCVT\(param ? "T" : "B")\<c>.F16.F32 <Sd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 0 0 Vd 1 0 1 sz op 1 M 0 Vm VCMP\(op ? "E" : "")\<c>.F\(size ? "64" : "32")\ <Dd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 0 1 Vd 1 0 1 sz op 1 (0) 0 (0) (0) (0) (0) VCMP\(op ? "E" : "")\<c>.F\(size ? "64" : "32")\ <Dd>, #0.0
+VFPU cond 1 1 1 0 1 D 1 1 0 1 1 1 Vd 1 0 1 0 1 1 M 0 Vm VCVT<c>.F64.F32 <Dd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 0 1 1 1 Vd 1 0 1 1 1 1 M 0 Vm VCVT<c>.F32.F64 <Sd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 1 0 0 0 Vd 1 0 1 sz op 1 M 0 Vm VCVT<c>.F\(size ? "64" : "32")\.\(op ? "S" : "U")\32 <Dd>, <Sm>
+VFPU cond 1 1 1 0 1 D 1 1 1 0 1 U Vd 1 0 1 sz op 1 i 0 imm4 @ set uint8_t td = (u << 2) + op + 1@ td=0b001=0b010=0b101 const char* Td Td,"S16","U16","S32","U32"@ set imm5 = (16 << op) - (imm5 >> 4) - ((imm5 & 0xF) << 1)@ @ VCVT<c>.F\(size ? "64" : "32")\.\Td\ <Dd>, <Dd>, #<imm>
+VFPU cond 1 1 1 0 1 D 1 1 1 1 0 U Vd 1 0 1 sz op 1 M 0 Vm VCVT\(op ? "" : "R")\<c>.\(u ? "U" : "S")\32.F\(size ? "64" : "32")\ <Sd>, <Dm>
+VFPU cond 1 1 1 0 1 D 1 1 1 1 1 U Vd 1 0 1 sz op 1 i 0 imm4 @ set uint8_t td = (u << 2) + op + 1@ td=0b001=0b010=0b101 const char* Td Td,"S16","U16","S32","U32"@ set imm5 = (16 << op) - (imm5 >> 4) - ((imm5 & 0xF) << 1)@ @ VCVT<c>.\Td\.F\(size ? "64" : "32")\ <Dd>, <Dd>, #<imm>
+INVALIDATE <32>
@@ -40,6 +40,7 @@ const char* DumpSection(Elf32_Shdr *s, char* SST) {
        GO(SHT_HIPROC);
        GO(SHT_LOUSER);
        GO(SHT_HIUSER);
+        #if defined(SHT_GNU_versym) && defined(SHT_GNU_ATTRIBUTES)
        GO(SHT_GNU_versym);
        GO(SHT_GNU_ATTRIBUTES);
        GO(SHT_GNU_HASH);
@@ -51,6 +52,7 @@ const char* DumpSection(Elf32_Shdr *s, char* SST) {
        GO(SHT_SUNW_syminfo);
        GO(SHT_GNU_verdef);
        GO(SHT_GNU_verneed);
+        #endif
        #undef GO
        default:
            sprintf(buff, "0x%X unknown type", s->sh_type);
@@ -98,6 +100,7 @@ const char* DumpDynamic(Elf32_Dyn *s) {
            GO(DT_RUNPATH, 0);
            GO(DT_FLAGS, 0);
            GO(DT_ENCODING, 0);
+            #if defined(DT_NUM) && defined(DT_TLSDESC_PLT)
            GO(DT_NUM, 0);
            GO(DT_VALRNGLO, 0);
            GO(DT_GNU_PRELINKED, 0);
@@ -133,6 +136,7 @@ const char* DumpDynamic(Elf32_Dyn *s) {
            GO(DT_VERNEEDNUM, 0);
            GO(DT_AUXILIARY, 0);
            GO(DT_FILTER, 0);
+            #endif
        #undef GO
        default:
            sprintf(buff, "0x%X unknown type", s->d_tag);
@@ -154,11 +158,13 @@ const char* DumpPHEntry(Elf32_Phdr *e)
        GO(PT_SHLIB);
        GO(PT_PHDR);
        GO(PT_TLS);
+        #ifdef PT_NUM
        GO(PT_NUM);
        GO(PT_LOOS);
        GO(PT_GNU_EH_FRAME);
        GO(PT_GNU_STACK);
        GO(PT_GNU_RELRO);
+        #endif
        #undef GO
        default: sprintf(buff, "type: %x, Off=%x vaddr=%p paddr=%p filesz=%u memsz=%u flags=%x align=%u", e->p_type, e->p_offset, (void*)e->p_vaddr, (void*)e->p_paddr, e->p_filesz, e->p_memsz, e->p_flags, e->p_align); break;
    }
@@ -187,6 +193,7 @@ const char* DumpRelType(int t)
        GO(R_386_TLS_GD_32);
        GO(R_386_TLS_DTPMOD32);
        GO(R_386_TLS_DTPOFF32);
+        GO(R_386_TLS_TPOFF32);
        #undef GO
        default: sprintf(buff, "type: 0x%x (unknown)", t); break;
    }
@@ -284,6 +291,22 @@ void DumpDynamicNeeded(elfheader_t *h)
    }
 }

+void DumpDynamicRPath(elfheader_t *h)
+{
+    if(box86_log>=LOG_DUMP && h->DynStrTab) {
+        printf_log(LOG_DUMP, "ELF Dump DT_RPATH/DT_RUNPATH=====\n");
+        for (int i=0; i<h->numDynamic; ++i) {
+            if(h->Dynamic[i].d_tag==DT_RPATH) {
+                printf_log(LOG_DUMP, "   RPATH : %s\n", h->DynStrTab+h->Dynamic[i].d_un.d_val + h->delta);
+            }
+            if(h->Dynamic[i].d_tag==DT_RUNPATH) {
+                printf_log(LOG_DUMP, " RUNPATH : %s\n", h->DynStrTab+h->Dynamic[i].d_un.d_val + h->delta);
+            }
+        }
+        printf_log(LOG_DUMP, "=====ELF Dump DT_RPATH/DT_RUNPATH\n");
+    }
+}
+
 void DumpRelTable(elfheader_t *h, int cnt, Elf32_Rel *rel, const char* name)
 {
    if(box86_log>=LOG_DUMP) {
@@ -5,10 +5,14 @@
 typedef struct dynablocklist_s dynablocklist_t;
 #endif

+typedef struct library_s library_t;
+typedef struct needed_libs_s needed_libs_t;
+
 #include <pthread.h>

 struct elfheader_s {
    char*       name;
+    char*       path;   // Resolved path to file
    int         numPHEntries;
    Elf32_Phdr  *PHEntries;
    int         numSHEntries;
@@ -26,6 +30,10 @@ struct elfheader_s {
    int         numDynamic;
    char*       DynStrTab;
    int         szDynStrTab;
+    Elf32_Half* VerSym;
+    Elf32_Verneed*  VerNeed;
+    int         szVerNeed;
+    int         e_type;

    intptr_t    delta;  // should be 0

@@ -48,7 +56,12 @@ struct elfheader_s {
    int         pltent;
    uint32_t    pltrel;
    uintptr_t   gotplt;
+    uintptr_t   gotplt_end;
+    uintptr_t   pltgot;
    uintptr_t   got;
+    uintptr_t   got_end;
+    uintptr_t   plt;
+    uintptr_t   plt_end;
    uintptr_t   text;
    int         textsz;

@@ -67,9 +80,13 @@ struct elfheader_s {
    int         fini_done;

    char*       memory; // char* and not void* to allow math on memory pointer
-#ifdef DYNAREC
-    dynablocklist_t *blocks;
-#endif
+    void**      multiblock;
+    uintptr_t*  multiblock_offs;
+    uint32_t*   multiblock_size;
+    int         multiblock_n;
+
+    library_t   *lib;
+    needed_libs_t *neededlibs;
 };

 #define R_386_NONE	0
@@ -1,3 +1,4 @@
+#define _GNU_SOURCE
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -23,7 +24,7 @@ int LoadSH(FILE *f, Elf32_Shdr *s, void** SH, const char* name, uint32_t type)
        printf_log(LOG_INFO, "Section Header \"%s\" (off=%d, size=%d) has size (not multiple of %d)\n", name, s->sh_offset, s->sh_size, sizeof(Elf32_Sym));
    }
    *SH = calloc(1, s->sh_size);
-    fseek(f, s->sh_offset ,SEEK_SET);
+    fseeko64(f, s->sh_offset ,SEEK_SET);
    if(fread(*SH, s->sh_size, 1, f)!=1) {
            printf_log(LOG_INFO, "Cannot read Section Header \"%s\" (off=%d, size=%d)\n", name, s->sh_offset, s->sh_size);
            return -1;
@@ -60,8 +61,9 @@ void LoadNamedSection(FILE *f, Elf32_Shdr *s, int size, char* SHStrTab, const ch
 elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
 {
    Elf32_Ehdr header;
+    int level = (exec)?LOG_INFO:LOG_DEBUG;
    if(fread(&header, sizeof(Elf32_Ehdr), 1, f)!=1) {
-        printf_log(LOG_INFO, "Cannot read ELF Header\n");
+        printf_log(level, "Cannot read ELF Header\n");
        return NULL;
    }
    if(memcmp(header.e_ident, ELFMAG, SELFMAG)!=0) {
@@ -69,11 +71,15 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
        return NULL;
    }
    if(header.e_ident[EI_CLASS]!=ELFCLASS32) {
-        printf_log(LOG_INFO, "Not an 32bits ELF (%d)\n", header.e_ident[EI_CLASS]);
+        if(header.e_ident[EI_CLASS]==ELFCLASS64) {
+            printf_log(LOG_INFO, "This is a 64bits ELF! box86 can only run 32bits ELF!\n");
+        } else {
+            printf_log(LOG_INFO, "Not a 32bits ELF (%d)\n", header.e_ident[EI_CLASS]);
+        }
        return NULL;
    }
    if(header.e_ident[EI_DATA]!=ELFDATA2LSB) {
-        printf_log(LOG_INFO, "Not an LittleEndian ELF (%d)\n", header.e_ident[EI_DATA]);
+        printf_log(LOG_INFO, "Not a LittleEndian ELF (%d)\n", header.e_ident[EI_DATA]);
        return NULL;
    }
    if(header.e_ident[EI_VERSION]!=EV_CURRENT) {
@@ -91,7 +97,7 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
    }

    if(header.e_machine != EM_386) {
-        printf_log(LOG_INFO, "Not an i386 ELF (%d)\n", header.e_machine);
+        printf_log(level, "Not an i386 ELF (%d)\n", header.e_machine);
        return NULL;
    }

@@ -114,12 +120,13 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
    h->numPHEntries = header.e_phnum;
    h->numSHEntries = header.e_shnum;
    h->SHIdx = header.e_shstrndx;
-    if(header.e_shentsize!=0) {
+    h->e_type = header.e_type;
+    if(header.e_shentsize && header.e_shnum) {
        // special cases for nums
        if(h->numSHEntries == 0) {
            printf_log(LOG_DEBUG, "Read number of Sections in 1st Section\n");
            // read 1st section header and grab actual number from here
-            fseek(f, header.e_shoff, SEEK_SET);
+            fseeko64(f, header.e_shoff, SEEK_SET);
            Elf32_Shdr section;
            if(fread(&section, sizeof(Elf32_Shdr), 1, f)!=1) {
                free(h);
@@ -131,7 +138,7 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
        // now read all section headers
        printf_log(LOG_DEBUG, "Read %d Section header\n", h->numSHEntries);
        h->SHEntries = (Elf32_Shdr*)calloc(h->numSHEntries, sizeof(Elf32_Shdr));
-        fseek(f, header.e_shoff ,SEEK_SET);
+        fseeko64(f, header.e_shoff ,SEEK_SET);
        if(fread(h->SHEntries, sizeof(Elf32_Shdr), h->numSHEntries, f)!=h->numSHEntries) {
                FreeElfHeader(&h);
                printf_log(LOG_INFO, "Cannot read all Section Header\n");
@@ -147,14 +154,14 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)

    printf_log(LOG_DEBUG, "Read %d Program header\n", h->numPHEntries);
    h->PHEntries = (Elf32_Phdr*)calloc(h->numPHEntries, sizeof(Elf32_Phdr));
-    fseek(f, header.e_phoff ,SEEK_SET);
+    fseeko64(f, header.e_phoff ,SEEK_SET);
    if(fread(h->PHEntries, sizeof(Elf32_Phdr), h->numPHEntries, f)!=h->numPHEntries) {
            FreeElfHeader(&h);
            printf_log(LOG_INFO, "Cannot read all Program Header\n");
            return NULL;
    }

-    if(header.e_shentsize!=0) {
+    if(header.e_shentsize && header.e_shnum) {
        if(h->SHIdx == SHN_XINDEX) {
            printf_log(LOG_DEBUG, "Read number of String Table in 1st Section\n");
            h->SHIdx = h->SHEntries[0].sh_link;
@@ -182,28 +189,79 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
        // also grab the DT_STRTAB string table
        {
            for (int i=0; i<h->numDynamic; ++i) {
-                if(h->Dynamic[i].d_tag == DT_REL)
-                    h->rel = h->Dynamic[i].d_un.d_ptr;
-                else if(h->Dynamic[i].d_tag == DT_RELSZ)
-                    h->relsz = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_RELENT)
-                    h->relent = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_RELA)
-                    h->rela = h->Dynamic[i].d_un.d_ptr;
-                else if(h->Dynamic[i].d_tag == DT_RELASZ)
-                    h->relasz = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_RELAENT)
-                    h->relaent = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_PLTREL)
-                    h->pltrel = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_PLTRELSZ)
-                    h->pltsz = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_JMPREL)
-                    h->jmprel = h->Dynamic[i].d_un.d_val;
-                else if(h->Dynamic[i].d_tag == DT_STRTAB)
-                    h->DynStrTab = (char*)(h->Dynamic[i].d_un.d_ptr);
-                else if(h->Dynamic[i].d_tag == DT_STRSZ)
-                    h->szDynStrTab = h->Dynamic[i].d_un.d_val;
+                Elf32_Dyn d = h->Dynamic[i];
+                Elf32_Word val = d.d_un.d_val;
+                Elf32_Addr ptr = d.d_un.d_ptr;
+                switch (d.d_tag) {
+                case DT_REL:
+                    h->rel = ptr;
+                    break;
+                case DT_RELSZ:
+                    h->relsz = val;
+                    break;
+                case DT_RELENT:
+                    h->relent = val;
+                    break;
+                case DT_RELA:
+                    h->rela = ptr;
+                    break;
+                case DT_RELASZ:
+                    h->relasz = val;
+                    break;
+                case DT_RELAENT:
+                    h->relaent = val;
+                    break;
+                case DT_PLTGOT:
+                    h->pltgot = val;
+                    break;
+                case DT_PLTREL:
+                    h->pltrel = val;
+                    break;
+                case DT_PLTRELSZ:
+                    h->pltsz = val;
+                    break;
+                case DT_JMPREL:
+                    h->jmprel = val;
+                    break;
+                case DT_STRTAB:
+                    h->DynStrTab = (char *)(ptr);
+                    break;
+                case DT_STRSZ:
+                    h->szDynStrTab = val;
+                    break;
+                case DT_INIT: // Entry point
+                    h->initentry = ptr;
+                    printf_log(LOG_DEBUG, "The DT_INIT is at address %p\n", (void*)h->initentry);
+                    break;
+                case DT_INIT_ARRAY:
+                    h->initarray = ptr;
+                    printf_log(LOG_DEBUG, "The DT_INIT_ARRAY is at address %p\n", (void*)h->initarray);
+                    break;
+                case DT_INIT_ARRAYSZ:
+                    h->initarray_sz = val / sizeof(Elf32_Addr);
+                    printf_log(LOG_DEBUG, "The DT_INIT_ARRAYSZ is %d\n", h->initarray_sz);
+                    break;
+                case DT_FINI: // Exit hook
+                    h->finientry = ptr;
+                    printf_log(LOG_DEBUG, "The DT_FINI is at address %p\n", (void*)h->finientry);
+                    break;
+                case DT_FINI_ARRAY:
+                    h->finiarray = ptr;
+                    printf_log(LOG_DEBUG, "The DT_FINI_ARRAY is at address %p\n", (void*)h->finiarray);
+                    break;
+                case DT_FINI_ARRAYSZ:
+                    h->finiarray_sz = val / sizeof(Elf32_Addr);
+                    printf_log(LOG_DEBUG, "The DT_FINI_ARRAYSZ is %d\n", h->finiarray_sz);
+                    break;
+                case DT_VERNEEDNUM:
+                    h->szVerNeed = val / sizeof(Elf32_Verneed);
+                    printf_log(LOG_DEBUG, "The DT_VERNEEDNUM is %d\n", h->szVerNeed);
+                    break;
+                case DT_VERNEED:
+                    h->VerNeed = (Elf32_Verneed*)ptr;
+                    printf_log(LOG_DEBUG, "The DT_VERNEED is at address %p\n", h->VerNeed);
+                    break;
+                }
            }
            if(h->rel) {
                if(h->relent != sizeof(Elf32_Rel)) {
@@ -246,38 +304,26 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
        int ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got.plt");
        if(ii) {
            h->gotplt = h->SHEntries[ii].sh_addr;
+            h->gotplt_end = h->gotplt + h->SHEntries[ii].sh_size;
            printf_log(LOG_DEBUG, "The GOT.PLT Table is at address %p\n", (void*)h->gotplt);
        }
        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got");
        if(ii) {
            h->got = h->SHEntries[ii].sh_addr;
-            printf_log(LOG_DEBUG, "The GOT Table is at address %p\n", (void*)h->got);
+            h->got_end = h->got + h->SHEntries[ii].sh_size;
+            printf_log(LOG_DEBUG, "The GOT Table is at address %p..%p\n", (void*)h->got, (void*)h->got_end);
        }
-        // look for .init entry point
-        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".init");
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".plt");
        if(ii) {
-            h->initentry = h->SHEntries[ii].sh_addr;
-            printf_log(LOG_DEBUG, "The .init is at address %p\n", (void*)h->initentry);
+            h->plt = h->SHEntries[ii].sh_addr;
+            h->plt_end = h->plt + h->SHEntries[ii].sh_size;
+            printf_log(LOG_DEBUG, "The PLT Table is at address %p..%p\n", (void*)h->plt, (void*)h->plt_end);
        }
-        // and .init_array
-        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".init_array");
+        // grab version of symbols
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".gnu.version");
        if(ii) {
-            h->initarray_sz = h->SHEntries[ii].sh_size / sizeof(Elf32_Addr);
-            h->initarray = (uintptr_t)(h->SHEntries[ii].sh_addr);
-            printf_log(LOG_DEBUG, "The .init_array is at address %p, and have %d elements\n", (void*)h->initarray, h->initarray_sz);
-        }
-        // look for .fini entry point
-        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".fini");
-        if(ii) {
-            h->finientry = h->SHEntries[ii].sh_addr;
-            printf_log(LOG_DEBUG, "The .fini is at address %p\n", (void*)h->finientry);
-        }
-        // and .fini_array
-        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".fini_array");
-        if(ii) {
-            h->finiarray_sz = h->SHEntries[ii].sh_size / sizeof(Elf32_Addr);
-            h->finiarray = (uintptr_t)(h->SHEntries[ii].sh_addr);
-            printf_log(LOG_DEBUG, "The .fini_array is at address %p, and have %d elements\n", (void*)h->finiarray, h->finiarray_sz);
+            h->VerSym = (Elf32_Half*)(h->SHEntries[ii].sh_addr);
+            printf_log(LOG_DEBUG, "The .gnu.version is at address %p\n", h->VerSym);
        }
        // grab .text for main code
        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".text");
@@ -294,3 +340,20 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
    
    return h;
 }
+
+const char* GetSymbolVersion(elfheader_t* h, int version)
+{
+    if(!h->VerNeed || (version<2))
+        return NULL;
+    Elf32_Verneed *ver = (Elf32_Verneed*)((uintptr_t)h->VerNeed + h->delta);
+    while(ver) {
+        Elf32_Vernaux *aux = (Elf32_Vernaux*)((uintptr_t)ver + ver->vn_aux);
+        for(int j=0; j<ver->vn_cnt; ++j) {
+            if(aux->vna_other==version)
+                return h->DynStr+aux->vna_name;
+            aux = (Elf32_Vernaux*)((uintptr_t)aux + aux->vna_next);
+        }
+        ver = ver->vn_next?((Elf32_Verneed*)((uintptr_t)ver + ver->vn_next)):NULL;
+    }
+    return NULL;
+}
@@ -47,10 +47,60 @@
        base+=(nextop&0x80)?(F32S):(F8S); \
        A = (T*)(base+O); \
    }
+
+#define  getecommon16(A, T)                     \
+    {                                           \
+        uint32_t m = nextop&0xC7;               \
+        uint32_t base = 0;                      \
+        switch(m&7) {                           \
+            case 0: base = R_BX+R_SI; break;    \
+            case 1: base = R_BX+R_DI; break;    \
+            case 2: base = R_BP+R_SI; break;    \
+            case 3: base = R_BP+R_DI; break;    \
+            case 4: base =      R_SI; break;    \
+            case 5: base =      R_DI; break;    \
+            case 6: base =      R_BP; break;    \
+            case 7: base =      R_BX; break;    \
+        }                                       \
+        switch((m>>6)&3) {                      \
+            case 0: if(m==6) base = F16; break; \
+            case 1: base += F8S; break;         \
+            case 2: base += F16S; break;        \
+        }                                       \
+        A =  (T*)base;                          \
+    }
+
+#define  getecommon16o(A, T, O)                 \
+    {                                           \
+        uint32_t m = nextop&0xC7;               \
+        uint32_t base = 0;                      \
+        switch(m&7) {                           \
+            case 0: base = R_BX+R_SI; break;    \
+            case 1: base = R_BX+R_DI; break;    \
+            case 2: base = R_BP+R_SI; break;    \
+            case 3: base = R_BP+R_DI; break;    \
+            case 4: base =      R_SI; break;    \
+            case 5: base =      R_DI; break;    \
+            case 6: base =      R_BP; break;    \
+            case 7: base =      R_BX; break;    \
+        }                                       \
+        switch((m>>6)&3) {                      \
+            case 0: if(m==6) base = F16; break; \
+            case 1: base += F8S; break;         \
+            case 2: base += F16S; break;        \
+        }                                       \
+        A =  (T*)(base+O);                      \
+    }
+
+
 #define geteb(A) \
    if((nextop&0xC0)==0xC0) { \
        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
    } else getecommon(A, reg32_t)
+#define getebo(A, O)          \
+    if((nextop&0xC0)==0xC0) { \
+        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
+    } else getecommono(A, reg32_t, O)
 #define geted(A) \
    if((nextop&0xC0)==0xC0) { \
        A = &emu->regs[(nextop&7)]; \
@@ -61,28 +111,41 @@
    } else getecommono(A, reg32_t, O)
 #define getem(A) \
    if((nextop&0xC0)==0xC0) { \
-        A = &emu->mmx[(nextop&7)]; \
-    } else getecommon(A, mmx_regs_t)
+        A = &emu->mmx87[(nextop&7)]; \
+    } else getecommon(A, mmx87_regs_t)
 #define getex(A) \
    if((nextop&0xC0)==0xC0) { \
        A = &emu->xmm[(nextop&7)]; \
    } else getecommon(A, sse_regs_t)
+#define getew16(A)  \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->regs[(nextop&7)]; \
+    } else getecommon16(A, reg32_t)
+#define getew16o(A, O)              \
+    if((nextop&0xC0)==0xC0) {       \
+        A = &emu->regs[(nextop&7)]; \
+    } else getecommon16o(A, reg32_t, O)
+
 // Macros for ModR/M gets
 #define GET_EB      geteb(oped)
 #define GET_ED      geted(oped)
 #define GET_ED_OFFS(o) getedo(oped, o)
+#define GET_EB_OFFS(o) getebo(oped, o)
 #define GET_EM      getem(opem)
 #define GET_EX      getex(opex)
+#define GET_EW16    getew16(oped)
+#define GET_EW16_OFFS(o)    getew16o(oped, o)
 #define EB          oped
 #define ED          oped
 #define EM          opem
 #define EX          opex
 #define GB          emu->regs[(nextop>>3)&3].byte[(nextop>>5)&0x1]
 #define GD          emu->regs[((nextop&0x38)>>3)]
-#define GM          emu->mmx[((nextop&0x38)>>3)]
+#define GM          emu->mmx87[((nextop&0x38)>>3)]
 #define GX          emu->xmm[((nextop&0x38)>>3)]

 // Alias
 #define GET_EW      GET_ED
+#define GET_EW_OFFS GET_ED_OFFS
 #define EW          ED
 #define GW          GD
@@ -1,7 +1,40 @@
    opcode = F8;
    goto *opcodes0f[opcode];
+        _0f_0x00:                       /* VERx Ed */
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 4: //VERR
+                case 5: //VERW
+                    GET_EW;
+                    if(!EW->word[0])
+                        CLEAR_FLAG(F_ZF);
+                    else
+                        SET_FLAG(F_ZF); // should test if selector is ok
+                    break;
+                default:
+                    goto _default;
+            }
+            NEXT;
+        _0f_0x01:                      
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 4:                 /* SMSW Ew */
+                    GET_ED;
+                    // dummy for now... Do I need to track CR0 state?
+                    ED->word[0] = (1<<0) | (1<<4); // only PE and ET set...
+                    break;
+                default:
+                    goto _default;
+            }
+            NEXT;
+
        _0f_0x0B:                      /* UD2 */
-            kill(getpid(), SIGILL);              // this is undefined instruction
+            emu->old_ip = R_EIP;
+            R_EIP = ip-2;
+            emit_signal(emu, SIGILL, (void*)R_EIP, 0);
+            ip = R_EIP;
+            if(emu->quit) goto fini;
+            STEP;
            NEXT;
        _0f_0x10:                      /* MOVUPS Gx,Ex */
            nextop = F8;
@@ -54,7 +87,27 @@
        _0f_0x18:                       /* PREFETCHh Ed */
            nextop = F8;
            GET_ED;
-            __builtin_prefetch((void*)ED->dword[0], 0, 0); // inoring wich level of cache (roughly (nextop>>3)&7)
+            if((nextop&0xC0)==0xC0) {
+            } else
+            switch((nextop>>3)&7) {
+                case 0: //PREFETCHnta
+                case 1: //PREFETCH1
+                case 2: //PREFETCH2
+                case 3: //PREFETCH3
+                    __builtin_prefetch((void*)ED, 0, 0); // ignoring wich level of cache
+                    break;
+                default:    //NOP
+                    break;
+            }
+            NEXT;
+
+        _0f_0x1A:                      /* NOP (multi-byte) / ignored BNDLDX */
+            nextop = F8;
+            GET_ED;
+            NEXT;
+        _0f_0x1B:                      /* NOP (multi-byte) / ignored BNDSTX */
+            nextop = F8;
+            GET_ED;
            NEXT;

        _0f_0x1F:                      /* NOP (multi-byte) */
@@ -96,7 +149,7 @@
            // rounding should be done; and indefinite integer should also be assigned if overflow or NaN/Inf
            nextop = F8;
            GET_EX;
-            switch(emu->round) {
+            switch((emu->mxcsr>>13)&3) {
                case ROUND_Nearest:
                    GM.sd[1] = floorf(EX->f[1]+0.5f);
                    GM.sd[0] = floorf(EX->f[0]+0.5f);
@@ -139,6 +192,73 @@
            R_EAX = tmp64u&0xFFFFFFFF;
            NEXT;
        
+        _0f_0x38:  // these are some SSE3 opcodes
+            opcode = F8;
+            switch(opcode) {
+                case 0x00:  /* PSHUFB */
+                    nextop = F8;
+                    GET_EM;
+                    eam1 = GM;
+                    for (int i=0; i<8; ++i) {
+                        if(EM->ub[i]&128)
+                            GM.ub[i] = 0;
+                        else
+                            GM.ub[i] = eax1.ub[EM->ub[i]&7];
+                    }
+                    break;
+                case 0x04:  /* PMADDUBSW Gm,Em */
+                    nextop = F8;
+                    GET_EM;
+                    for (int i=0; i<4; ++i) {
+                        tmp32s = (int32_t)(GM.ub[i*2+0])*EM->sb[i*2+0] + (int32_t)(GM.ub[i*2+1])*EM->sb[i*2+1];
+                        GM.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
+                    }
+                    break;
+                case 0x0B:  /* PMULHRSW Gm, Em */
+                    nextop = F8;
+                    GET_EM;
+                    for (int i=0; i<4; ++i) {
+                        tmp32s = ((((int32_t)(GM.sw[i])*(int32_t)(EM->sw[i]))>>14) + 1)>>1;
+                        GM.uw[i] = tmp32s&0xffff;
+                    }
+                    break;
+
+                case 0x1C:  /* PABSB Gm, Em */
+                    nextop = F8;
+                    GET_EM;
+                    for (int i=0; i<8; ++i) {
+                        GM.sb[i] = abs(EM->sb[i]);
+                    }
+                    break;
+                case 0x1D:  /* PABSW Gm, Em */
+                    nextop = F8;
+                    GET_EM;
+                    for (int i=0; i<4; ++i) {
+                        GM.sw[i] = abs(EM->sw[i]);
+                    }
+                    break;
+                case 0x1E:  /* PABSD Gm, Em */
+                    nextop = F8;
+                    GET_EM;
+                    for (int i=0; i<2; ++i) {
+                        GM.sd[i] = abs(EM->sd[i]);
+                    }
+                    break;
+
+                default:
+                    goto _default;
+            }
+            NEXT;
+
+        _0f_0x3F:
+            emu->old_ip = R_EIP;
+            R_EIP = ip-2;
+            emit_signal(emu, SIGILL, (void*)R_EIP, 0);
+            ip = R_EIP;
+            if(emu->quit) goto fini;
+            STEP;
+            NEXT;
+            
        #define GOCOND(BASE, PREFIX, CONDITIONAL) \
        _0f_##BASE##_0:                          \
            PREFIX                              \
@@ -372,10 +492,14 @@
            GM.sb[1] = (GM.sw[1] > 127) ? 127 : ((GM.sw[1] < -128) ? -128 : GM.sw[1]);
            GM.sb[2] = (GM.sw[2] > 127) ? 127 : ((GM.sw[2] < -128) ? -128 : GM.sw[2]);
            GM.sb[3] = (GM.sw[3] > 127) ? 127 : ((GM.sw[3] < -128) ? -128 : GM.sw[3]);
-            GM.sb[4] = (EM->sw[0] > 127) ? 127 : ((EM->sw[0] < -128) ? -128 : EM->sw[0]);
-            GM.sb[5] = (EM->sw[1] > 127) ? 127 : ((EM->sw[1] < -128) ? -128 : EM->sw[1]);
-            GM.sb[6] = (EM->sw[2] > 127) ? 127 : ((EM->sw[2] < -128) ? -128 : EM->sw[2]);
-            GM.sb[7] = (EM->sw[3] > 127) ? 127 : ((EM->sw[3] < -128) ? -128 : EM->sw[3]);
+            if(EM==&GM)
+                GM.ud[1] = GM.ud[0];
+            else {
+                GM.sb[4] = (EM->sw[0] > 127) ? 127 : ((EM->sw[0] < -128) ? -128 : EM->sw[0]);
+                GM.sb[5] = (EM->sw[1] > 127) ? 127 : ((EM->sw[1] < -128) ? -128 : EM->sw[1]);
+                GM.sb[6] = (EM->sw[2] > 127) ? 127 : ((EM->sw[2] < -128) ? -128 : EM->sw[2]);
+                GM.sb[7] = (EM->sw[3] > 127) ? 127 : ((EM->sw[3] < -128) ? -128 : EM->sw[3]);
+            }
            NEXT;
        _0f_0x64:                       /* PCMPGTB Gm,Em */
            nextop = F8;
@@ -403,33 +527,42 @@
            GET_EM;
            for(int i=0; i<4; ++i)
                GM.ub[i] = (GM.sw[i]<0)?0:((GM.sw[i]>0xff)?0xff:GM.sw[i]);
-            for(int i=0; i<4; ++i)
-                GM.ub[4+i] = (EM->sw[i]<0)?0:((EM->sw[i]>0xff)?0xff:EM->sw[i]);
+            if(EM==&GM)
+                GM.ud[1] = GM.ud[0];
+            else
+                for(int i=0; i<4; ++i)
+                    GM.ub[4+i] = (EM->sw[i]<0)?0:((EM->sw[i]>0xff)?0xff:EM->sw[i]);
            NEXT;
        _0f_0x68:                       /* PUNPCKHBW Gm,Em */
            nextop = F8;
            GET_EM;
-            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
            for(int i=0; i<4; ++i)
                GM.ub[2 * i] = GM.ub[i + 4];
-            for(int i=0; i<4; ++i)
-                GM.ub[2 * i + 1] = EM->ub[i + 4];
+            if(EM==&GM)
+                for(int i=0; i<4; ++i)
+                    GM.ub[2 * i + 1] = GM.ub[2 * i];
+            else
+                for(int i=0; i<4; ++i)
+                    GM.ub[2 * i + 1] = EM->ub[i + 4];
            NEXT;
        _0f_0x69:                       /* PUNPCKHWD Gm,Em */
            nextop = F8;
            GET_EM;
-            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
            for(int i=0; i<2; ++i)
                GM.uw[2 * i] = GM.uw[i + 2];
-            for(int i=0; i<2; ++i)
-                GM.uw[2 * i + 1] = EM->uw[i + 2];
+            if(EM==&GM)
+                for(int i=0; i<2; ++i)
+                    GM.uw[2 * i + 1] = GM.uw[2 * i];
+            else
+                for(int i=0; i<2; ++i)
+                    GM.uw[2 * i + 1] = EM->uw[i + 2];
            NEXT;
        _0f_0x6A:                       /* PUNPCKHDQ Gm,Em */
            nextop = F8;
            GET_EM;
-            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
            GM.ud[0] = GM.ud[1];
-            GM.ud[1] = EM->ud[1];
+            if(EM!=&GM)
+                GM.ud[1] = EM->ud[1];
            NEXT;
        _0f_0x6B:                       /* PACKSSDW Gm,Em */
            nextop = F8;
@@ -437,8 +570,11 @@
            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
            for(int i=0; i<2; ++i)
                GM.sw[i] = (GM.sd[i]<-32768)?-32768:((GM.sd[i]>32767)?32767:GM.sd[i]);
-            for(int i=0; i<2; ++i)
-                GM.sw[2+i] = (EM->sd[i]<-32768)?-32768:((EM->sd[i]>32767)?32767:EM->sd[i]);
+            if(EM==&GM)
+                GM.ud[1] = GM.ud[0];
+            else
+                for(int i=0; i<2; ++i)
+                    GM.sw[2+i] = (EM->sd[i]<-32768)?-32768:((EM->sd[i]>32767)?32767:EM->sd[i]);
            NEXT;

        _0f_0x6E:                      /* MOVD Gm, Ed */
@@ -571,95 +707,16 @@
            EM->q = GM.q;
            NEXT;

+        _0f_0xA0:                      /* PUSH FS */
+            Push(emu, emu->segs[_FS]);    // even if a segment is a 16bits, a 32bits push/pop is done
+            NEXT;
+        _0f_0xA1:                      /* POP FS */
+            emu->segs[_FS] = Pop(emu);    // no check, no use....
+            emu->segs_serial[_FS] = 0;
+            NEXT;
        _0f_0xA2:                      /* CPUID */
            tmp32u = R_EAX;
-            switch(tmp32u) {
-                case 0x0:
-                    // emulate a P4
-                    R_EAX = 0x80000004;
-                    // return GenuineIntel
-                    R_EBX = 0x756E6547;
-                    R_EDX = 0x49656E69;
-                    R_ECX = 0x6C65746E;
-                    break;
-                case 0x1:
-                    R_EAX = 0x00000101; // familly and all
-                    R_EBX = 0;          // Brand indexe, CLFlush, Max APIC ID, Local APIC ID
-                    R_EDX =   1         // fpu 
-                            | 1<<8      // cmpxchg8
-                            | 1<<11     // sep (sysenter & sysexit)
-                            | 1<<15     // cmov
-                            | 1<<19     // clflush (seems to be with SSE2)
-                            | 1<<23     // mmx
-                            //| 1<<24     // fxsr (fxsave, fxrestore)
-                            | 1<<25     // SSE
-                            | 1<<26     // SSE2
-                            ;
-                    R_ECX =   1<<12     // fma
-                            | 1<<13     // cx16 (cmpxchg16)
-                            ;           // also 1<<0 is SSE3 and 1<<9 is SSSE3
-                    break;
-                case 0x2:   // TLB and Cache info. Sending 1st gen P4 info...
-                    R_EAX = 0x665B5001;
-                    R_EBX = 0x00000000;
-                    R_ECX = 0x00000000;
-                    R_EDX = 0x007A7000;
-                    break;
-                
-                case 0x4:   // Cache info
-                    switch (R_ECX) {
-                        case 0: // L1 data cache
-                            R_EAX = (1 | (1<<5) | (1<<8));   //type
-                            R_EBX = (63 | (7<<22)); // size
-                            R_ECX = 63;
-                            R_EDX = 1;
-                            break;
-                        case 1: // L1 inst cache
-                            R_EAX = (2 | (1<<5) | (1<<8)); //type
-                            R_EBX = (63 | (7<<22)); // size
-                            R_ECX = 63;
-                            R_EDX = 1;
-                            break;
-                        case 2: // L2 cache
-                            R_EAX = (3 | (2<<5) | (1<<8)); //type
-                            R_EBX = (63 | (15<<22));    // size
-                            R_ECX = 4095;
-                            R_EDX = 1;
-                            break;
-
-                        default:
-                            R_EAX = 0x00000000;
-                            R_EBX = 0x00000000;
-                            R_ECX = 0x00000000;
-                            R_EDX = 0x00000000;
-                            break;
-                    }
-                    break;
-                case 0x5:   //mwait info
-                    R_EAX = 0;
-                    R_EBX = 0;
-                    R_ECX = 1 | 2;
-                    R_EDX = 0;
-                    break;
-                case 0x6:   // thermal
-                case 0x9:   // direct cache access
-                case 0xA:   // Architecture performance monitor
-                    R_EAX = 0;
-                    R_EBX = 0;
-                    R_ECX = 0;
-                    R_EDX = 0;
-                    break;
-                case 0x7:   // extended bits...
-                    /*if(R_ECX==0)    R_EAX = 0;
-                    else*/ R_EAX = R_ECX = R_EBX = R_EDX = 0;
-                    break;
-
-                case 0x80000000:        // max extended
-                    break;              // no extended, so return same value as beeing the max value!
-                default:
-                    printf_log(LOG_INFO, "Warning, CPUID command %X unsupported (ECX=%08x)\n", tmp32u, R_ECX);
-                    R_EAX = 0;
-            }
+            my_cpuid(emu, tmp32u);
            NEXT;
        _0f_0xA3:                      /* BT Ed,Gd */
            CHECK_FLAGS(emu);
@@ -684,6 +741,15 @@
            ED->dword[0] = shld32(emu, ED->dword[0], GD.dword[0], tmp8u);
            NEXT;

+        _0f_0xA8:                      /* PUSH GS */
+            Push(emu, emu->segs[_GS]);    // even if a segment is a 16bits, a 32bits push/pop is done
+            NEXT;
+        _0f_0xA9:                      /* POP GS */
+            emu->segs[_GS] = Pop(emu);
+            emu->segs_serial[_GS] = 0;
+            NEXT;
+
+
        _0f_0xAB:                      /* BTS Ed,Gd */
            CHECK_FLAGS(emu);
            nextop = F8;
@@ -725,16 +791,18 @@
                case 0:                 /* FXSAVE m512byte */
                    // should save flags & all
                    // copy MMX regs...
-                    memcpy(ED->dword+32/4, &emu->mmx[0], sizeof(emu->mmx));
+                    for(int i=0; i<8; ++i)
+                        memcpy(((void*)(ED))+32+i*16, &emu->mmx87[0], sizeof(emu->mmx87[0]));
                    // copy SSE regs
-                    memcpy(ED->dword+160/4, &emu->xmm[0], sizeof(emu->xmm));
+                    memcpy(((void*)(ED))+160, &emu->xmm[0], sizeof(emu->xmm));
                    break;
                case 1:                 /* FXRSTOR m512byte */
                    // should restore flags & all
-                    // copy MMX regs...
-                    memcpy(&emu->mmx[0], ED->dword+32/4, sizeof(emu->mmx));
+                    // copy back MMX regs...
+                    for(int i=0; i<8; ++i)
+                        memcpy(&emu->mmx87[i], ((void*)(ED))+32+i*16, sizeof(emu->mmx87[0]));
                    // copy SSE regs
-                    memcpy(&emu->xmm[0], ED->dword+160/4, sizeof(emu->xmm));
+                    memcpy(&emu->xmm[0], ((void*)(ED))+160/4, sizeof(emu->xmm));
                    break;
                case 2:                 /* LDMXCSR Md */
                    emu->mxcsr = ED->dword[0];
@@ -752,7 +820,6 @@
            GD.dword[0] = imul32(emu, GD.dword[0], ED->dword[0]);
            NEXT;
        _0f_0xB0:                      /* CMPXCHG Eb,Gb */
-            CHECK_FLAGS(emu);
            nextop = F8;
            GET_EB;
            cmp8(emu, R_AL, EB->byte[0]);
@@ -818,6 +885,22 @@
                    else
                        CLEAR_FLAG(F_CF);
                    break;
+                case 5:             /* BTS Ed, Ib */
+                    CHECK_FLAGS(emu);
+                    GET_ED;
+                    tmp8u = F8;
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                    }
+                    tmp8u&=31;
+                    if(ED->dword[0] & (1<<tmp8u)) {
+                        SET_FLAG(F_CF);
+                    } else {
+                        ED->dword[0] ^= (1<<tmp8u);
+                        CLEAR_FLAG(F_CF);
+                    }
+                    break;
                case 6:             /* BTR Ed, Ib */
                    CHECK_FLAGS(emu);
                    GET_ED;
@@ -833,6 +916,21 @@
                    } else
                        CLEAR_FLAG(F_CF);
                    break;
+                case 7:             /* BTC Ed, Ib */
+                    CHECK_FLAGS(emu);
+                    GET_ED;
+                    tmp8u = F8;
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                    }
+                    tmp8u&=31;
+                    if(ED->dword[0] & (1<<tmp8u))
+                        SET_FLAG(F_CF);
+                    else
+                        CLEAR_FLAG(F_CF);
+                    ED->dword[0] ^= (1<<tmp8u);
+                    break;

                default:
                    goto _default;
@@ -926,7 +1024,11 @@
                GX.ud[i]=(tmp8s)?0xffffffff:0;
            }
            NEXT;
-
+        _0f_0xC3:                       /* MOVNTI Ed,Gd */
+            nextop = F8;
+            GET_ED;
+            ED->dword[0] = GD.dword[0];
+            NEXT;
        _0f_0xC4:                       /* PINSRW Gm,Ew,Ib */
            nextop = F8;
            GET_ED;
@@ -1021,6 +1123,17 @@
            }
            NEXT;

+        _0f_0xD7:                   /* PMOVMSKB Gd,Em */
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                GET_EM;
+                GD.dword[0] = 0;
+                for (int i=0; i<8; ++i)
+                    if(EM->ub[i]&0x80)
+                        GD.dword[0] |= (1<<i);
+                NEXT;
+            } else
+                goto _default;
        _0f_0xD8:                   /* PSUBUSB Gm,Em */
            nextop = F8;
            GET_EM;
@@ -1037,7 +1150,12 @@
                GM.uw[i] = (tmp32s < 0) ? 0 : tmp32s;
            }
            NEXT;
-
+        _0f_0xDA:                   /* PMINUB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i=0; i<8; ++i)
+                GM.ub[i] = (GM.ub[i]<EM->ub[i])?GM.ub[i]:EM->ub[i];
+            NEXT;
        _0f_0xDB:                   /* PAND Gm,Em */
            nextop = F8;
            GET_EM;
@@ -1059,7 +1177,12 @@
                GM.uw[i] = (tmp32u>65535) ? 65535 : tmp32u;
            }
            NEXT;
-
+        _0f_0xDE:                   /* PMAXUB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i=0; i<8; ++i)
+                GM.ub[i] = (GM.ub[i]>EM->ub[i])?GM.ub[i]:EM->ub[i];
+            NEXT;
        _0f_0xDF:                   /* PANDN Gm,Em */
            nextop = F8;
            GET_EM;
@@ -1093,20 +1216,17 @@
                    GM.sd[i] >>= tmp8u;
            }
            NEXT;
-        _0f_0xE3:                   /* PSRAQ Gm, Em */
+        _0f_0xE3:                   /* PAVGW Gm, Em */
            nextop = F8;
            GET_EM;
-            if(EM->q>63)
-                tmp8u = 64;
-            else
-                tmp8u = EM->ub[0];
-            GM.sq >>= tmp8u;
+            for(int i=0; i<4; ++i)
+                GM.uw[i] = ((uint32_t)GM.uw[i]+EM->uw[i]+1)>>1;
            NEXT;
        _0f_0xE4:                   /* PMULHUW Gm, Em */
            nextop = F8;
            GET_EM;
            for(int i=0; i<4; ++i) {
-                tmp32u = (int32_t)GM.uw[i] * EM->uw[i];
+                tmp32u = (uint32_t)GM.uw[i] * EM->uw[i];
                GM.uw[i] = (tmp32u>>16)&0xffff;
            }
            NEXT;
@@ -1119,6 +1239,13 @@
            }
            NEXT;

+        _0f_0xE7:                   /* MOVNTQ Em,Gm */
+            nextop = F8;
+            if((nextop&0xC0)==0xC0)
+                goto _default;
+            GET_EM;
+            EM->q = GM.q;
+            NEXT;
        _0f_0xE8:                   /* PSUBSB Gm,Em */
            nextop = F8;
            GET_EM;
@@ -1135,7 +1262,12 @@
                GM.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
            }
            NEXT;
-
+        _0f_0xEA:                   /* PMINSW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i=0; i<4; ++i)
+                GM.sw[i] = (GM.sw[i]<EM->sw[i])?GM.sw[i]:EM->sw[i];
+            NEXT;
        _0f_0xEB:                   /* POR Gm, Em */
            nextop = F8;
            GET_EM;
@@ -1157,7 +1289,12 @@
                GM.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
            }
            NEXT;
-
+        _0f_0xEE:                   /* PMAXSW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i=0; i<4; ++i)
+                GM.sw[i] = (GM.sw[i]>EM->sw[i])?GM.sw[i]:EM->sw[i];
+            NEXT;
        _0f_0xEF:                   /* PXOR Gm, Em */
            nextop = F8;
            GET_EM;
@@ -1215,7 +1352,15 @@
                tmp32u += (GM.ub[i]>EM->ub[i])?(GM.ub[i] - EM->ub[i]):(EM->ub[i] - GM.ub[i]);
            GM.q = tmp32u;
            NEXT;
-
+        _0f_0xF7:                   /* MASKMOVQ Gm, Em */
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                GET_EM;
+                for (int i=0; i<8; ++i)
+                    if(EM->ub[i]&0x80) ((uint8_t*)(R_EDI))[i] = GM.ub[i];
+            } else
+                goto _default;
+            NEXT;
        _0f_0xF8:                   /* PSUBB Gm,Em */
            nextop = F8;
            GET_EM;
@@ -1234,7 +1379,11 @@
            for(int i=0; i<2; ++i)
                GM.sd[i] -= EM->sd[i];
            NEXT;
-
+        _0f_0xFB:                   /* PSUBQ Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.sq -= EM->sq;
+            NEXT;
        _0f_0xFC:                   /* PADDB Gm, Em */
            nextop = F8;
            GET_EM;
@@ -2,16 +2,29 @@
    goto *opcodes66[opcode];

    #define GO(B, OP)                       \
+    _66_##B##_0: \
+        nextop = F8;               \
+        GET_EB;             \
+        EB->byte[0] = OP##8(emu, EB->byte[0], GB);  \
+        NEXT;                              \
    _66_##B##_1:                               \
        nextop = F8;                \
        GET_EW;                   \
        EW->word[0] = OP##16(emu, EW->word[0], GW.word[0]); \
        NEXT;                              \
+    _66_##B##_2: \
+        nextop = F8;               \
+        GET_EB;                   \
+        GB = OP##8(emu, GB, EB->byte[0]); \
+        NEXT;                              \
    _66_##B##_3:                               \
        nextop = F8;                \
        GET_EW;                   \
        GW.word[0] = OP##16(emu, GW.word[0], EW->word[0]); \
        NEXT;                              \
+    _66_##B##_4: \
+        R_AL = OP##8(emu, R_AL, F8); \
+        NEXT;                              \
    _66_##B##_5:                               \
        R_AX = OP##16(emu, R_AX, F16); \
        NEXT;
@@ -30,17 +43,36 @@
        NEXT;
    _66_0x07:                      /* POP ES */
        emu->segs[_ES] = Pop16(emu);    // no check, no use....
+        emu->segs_serial[_ES] = 0;
        NEXT;

    _66_0x0F:                      /* 66 0f prefix */
-        #include "run660f.h"
+        emu->old_ip = R_EIP;
+        R_EIP = ip-2;   // don't count 66 0F yet
+        Run660F(emu); // implemented in Run660f.c
+        ip = R_EIP;
+        if(emu->quit) goto fini;
+        STEP
+        NEXT;
+
+    _66_0x1F:                      /* POP DS */
+        emu->segs[_DS] = Pop16(emu);    // no check, no use....
+        emu->segs_serial[_DS] = 0;
+        NEXT;
        
    _66_0x26:                      /* ES: */
        // ignored
-        NEXT;
+        opcode = F8;
+        goto *opcodes66[opcode];
    _66_0x2E:                      /* CS: */
        // ignored
-        NEXT;
+        opcode = F8;
+        goto *opcodes66[opcode];
+    _66_0x36:                      /* SS: */
+        // ignored
+        opcode = F8;
+        goto *opcodes66[opcode];
+
    _66_0x39:
        nextop = F8;
        GET_EW;
@@ -99,6 +131,34 @@
        tmp8u = opcode&7;
        emu->regs[tmp8u].word[0] = Pop16(emu);
        NEXT;
+    _66_0x60:                              /* PUSHA */
+        tmp16u = R_SP;
+        Push16(emu, R_AX);
+        Push16(emu, R_CX);
+        Push16(emu, R_DX);
+        Push16(emu, R_BX);
+        Push16(emu, tmp16u);
+        Push16(emu, R_BP);
+        Push16(emu, R_SI);
+        Push16(emu, R_DI);
+        NEXT;
+    _66_0x61:                              /* POPA */
+        R_DI = Pop16(emu);
+        R_SI = Pop16(emu);
+        R_BP = Pop16(emu);
+        R_ESP+=2;   // POP ESP
+        R_BX = Pop16(emu);
+        R_DX = Pop16(emu);
+        R_CX = Pop16(emu);
+        R_AX = Pop16(emu);
+        NEXT;
+
+    _66_0x64:
+        R_EIP = ip-2;
+        RunFS66(emu, GetFSBaseEmu(emu));
+        ip = R_EIP;
+        STEP;
+        NEXT;

    _66_0x66:
        goto _0x66; // 0x66 0x66 => can remove one 0x66
@@ -174,13 +234,14 @@
    _66_0x8C:                              /* MOV Ew,Seg */
        nextop = F8;
        GET_EW;
-        EW->word[0] = emu->segs[(nextop&38)>>3];
+        EW->word[0] = emu->segs[(nextop&0x38)>>3];
        NEXT;
    
    _66_0x8E:                               /* MOV Seg,Ew */
        nextop = F8;
        GET_EW;
        emu->segs[((nextop&0x38)>>3)] = EW->word[0];
+        emu->segs_serial[((nextop&0x38)>>3)] = 0;
        NEXT;
    _66_0x8F:                              /* POP Ew */
        nextop = F8;
@@ -190,10 +251,16 @@
    _66_0x90:                              /* NOP */
        NEXT;

-    _66_0x92:                              /* XCHG DX,AX */
+    _66_0x91:
+    _66_0x92:
+    _66_0x93:
+    _66_0x94:
+    _66_0x95:
+    _66_0x96:
+    _66_0x97:                      /* XCHG reg,EAX */
        tmp16u = R_AX;
-        R_AX = R_DX;
-        R_DX = tmp16u;
+        R_AX = emu->regs[opcode&7].word[0];
+        emu->regs[opcode&7].word[0] = tmp16u;
        NEXT;

    _66_0x98:                               /* CBW */
@@ -205,8 +272,12 @@

    _66_0x9C:                              /* PUSHFW */
        CHECK_FLAGS(emu);
-        PackFlags(emu);
-        Push16(emu, (uint16_t)emu->packed_eflags.x32);
+        Push16(emu, (uint16_t)emu->eflags.x32);
+        NEXT;
+    _66_0x9D:                              /* POPFW */
+        CHECK_FLAGS(emu);
+        emu->eflags.x32 &=0xffff0000;
+        emu->eflags.x32 |= (Pop16(emu) & 0x3F7FD7) | 0x2;
        NEXT;

    _66_0xA1:                              /* MOV AX,Ow */
@@ -284,6 +355,22 @@
        EW->word[0] = F16;
        NEXT;

+    _66_0xCB:                               /* FAR RET */
+        ip = Pop(emu);
+        emu->segs[_CS] = Pop(emu);    // no check, no use....
+        emu->segs_serial[_CS] = 0;
+        // need to check status of CS register!
+        STEP
+        NEXT;
+    _66_0xCC:                              /* INT3 */
+        emu->old_ip = R_EIP;
+        R_EIP = ip-1;
+        if(my_context->signals[SIGTRAP])
+            raise(SIGTRAP);
+        ip = R_EIP;
+        if(emu->quit) goto fini;
+        STEP
+        NEXT;
    _66_0xD1:                              /* GRP2 Ew,1  */
    _66_0xD3:                              /* GRP2 Ew,CL */
        nextop = F8;
@@ -302,21 +389,31 @@
        NEXT;
    
    _66_0xD9:
-        emu->old_ip = old_ip;
+        emu->old_ip = R_EIP;
        R_EIP = ip;
        Run66D9(emu);
        ip = R_EIP;
        if(emu->quit) goto fini;
+        STEP
        NEXT;

    _66_0xDD:
-        emu->old_ip = old_ip;
+        emu->old_ip = R_EIP;
        R_EIP = ip;
        Run66DD(emu);
        ip = R_EIP;
        if(emu->quit) goto fini;
+        STEP
        NEXT;

+    _66_0xF0:                      /* LOCK prefix */
+        emu->old_ip = R_EIP;
+        R_EIP = ip - 2;
+        RunLock66(emu);
+        ip = R_EIP;
+        if(emu->quit) goto fini;
+        STEP
+        NEXT;

    _66_0xF2:                      /* REPNZ prefix */
    _66_0xF3:                      /* REPZ prefix */
@@ -324,6 +421,14 @@
        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
        tmp32u = R_ECX;
        switch(nextop) {
+            case 0xA4:              /* REP MOVSB */
+                while(tmp32u) {
+                    --tmp32u;
+                    *(uint8_t*)R_EDI = *(uint8_t*)R_ESI;
+                    R_EDI += tmp8s;
+                    R_ESI += tmp8s;
+                }
+                break;
            case 0xA5:              /* REP MOVSW */
                while(tmp32u) {
                    --tmp32u;
@@ -405,6 +510,14 @@
                break;
        }
        NEXT;
+    _66_0xF8:                       /* CLC */
+        CHECK_FLAGS(emu);
+        CLEAR_FLAG(F_CF);
+        NEXT;
+    _66_0xF9:                       /* STC */
+        CHECK_FLAGS(emu);
+        SET_FLAG(F_CF);
+        NEXT;

    _66_0xFF:                      /* GRP 5 Ew */
        nextop = F8;
@@ -416,8 +529,11 @@
            case 1:                 /* DEC Ed */
                EW->word[0] = dec16(emu, EW->word[0]);
                break;
+            case 6:
+                Push16(emu, EW->word[0]);
+                break;
            default:
-                emu->old_ip = old_ip;
+                emu->old_ip = R_EIP;
                R_EIP = ip;
                printf_log(LOG_NONE, "Illegal Opcode 66 %02X %02X\n", opcode, nextop);
                emu->quit=1;
@@ -10,11 +10,7 @@
        case 0xC5:
        case 0xC6:
        case 0xC7:  /* FADD */
-            #ifdef USE_FLOAT
-            ST0.f += ST(nextop&7).f;
-            #else
            ST0.d += ST(nextop&7).d;
-            #endif
            break;
        case 0xC8:
        case 0xC9:
@@ -24,11 +20,7 @@
        case 0xCD:
        case 0xCE:
        case 0xCF:  /* FMUL */
-            #ifdef USE_FLOAT
-            ST0.f *= ST(nextop&7).f;
-            #else
            ST0.d *= ST(nextop&7).d;
-            #endif
            break;
        case 0xD0:
        case 0xD1:
@@ -38,11 +30,7 @@
        case 0xD5:
        case 0xD6:
        case 0xD7:  /* FCOM */
-            #ifdef USE_FLOAT
-            fpu_fcom(emu, ST(nextop&7).f);
-            #else
            fpu_fcom(emu, ST(nextop&7).d);
-            #endif
            break;
        case 0xD8:
        case 0xD9:
@@ -52,11 +40,7 @@
        case 0xDD:
        case 0xDE:
        case 0xDF:  /* FCOMP */
-            #ifdef USE_FLOAT
-            fpu_fcom(emu, ST(nextop&7).f);
-            #else
            fpu_fcom(emu, ST(nextop&7).d);
-            #endif
            fpu_do_pop(emu);
            break;
        case 0xE0:
@@ -67,11 +51,7 @@
        case 0xE5:
        case 0xE6:
        case 0xE7:  /* FSUB */
-            #ifdef USE_FLOAT
-            ST0.f -= ST(nextop&7).f;
-            #else
            ST0.d -= ST(nextop&7).d;
-            #endif
            break;
        case 0xE8:
        case 0xE9:
@@ -81,11 +61,7 @@
        case 0xED:
        case 0xEE:
        case 0xEF:  /* FSUBR */
-            #ifdef USE_FLOAT
-            ST0.f = ST(nextop&7).f - ST0.f;
-            #else
            ST0.d = ST(nextop&7).d - ST0.d;
-            #endif
            break;
        case 0xF0:
        case 0xF1:
@@ -95,11 +71,7 @@
        case 0xF5:
        case 0xF6:
        case 0xF7:  /* FDIV */
-            #ifdef USE_FLOAT
-            ST0.f /= ST(nextop&7).f;
-            #else
            ST0.d /= ST(nextop&7).d;
-            #endif
            break;
        case 0xF8:
        case 0xF9:
@@ -109,46 +81,26 @@
        case 0xFD:
        case 0xFE:
        case 0xFF:  /* FDIVR */
-            #ifdef USE_FLOAT
-            ST0.f = ST(nextop&7).f / ST0.f;
-            #else
            ST0.d = ST(nextop&7).d / ST0.d;
-            #endif
            break;
        default:
        switch((nextop>>3)&7) {
            case 0:         /* FADD ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f += *(float*)ED;
-                    #else
                    ST0.d += *(float*)ED;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f += f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d += f;
-                    #endif
                }
                break;
            case 1:         /* FMUL ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f *= *(float*)ED;
-                    #else
                    ST0.d *= *(float*)ED;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f *= f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d *= f;
-                    #endif
                }
                break;
            case 2:      /* FCOM ST0, float */
@@ -156,7 +108,7 @@
                if(!(((uintptr_t)ED)&3))
                    fpu_fcom(emu, *(float*)ED);
                else {
-                    *(uint32_t*)&f = ED->dword[0];
+                    memcpy(&f, ED, sizeof(float));
                    fpu_fcom(emu, f);
                }
                break;
@@ -165,7 +117,7 @@
                if(!(((uintptr_t)ED)&3))
                    fpu_fcom(emu, *(float*)ED);
                else {
-                    *(uint32_t*)&f = ED->dword[0];
+                    memcpy(&f, ED, sizeof(float));
                    fpu_fcom(emu, f);
                }
                fpu_do_pop(emu);
@@ -173,69 +125,37 @@
            case 4:         /* FSUB ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f -= *(float*)ED;
-                    #else
                    ST0.d -= *(float*)ED;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f -= f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d -= f;
-                    #endif
                }
                break;
            case 5:         /* FSUBR ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f = *(float*)ED - ST0.f;
-                    #else
                    ST0.d = *(float*)ED - ST0.d;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f = f - ST0.f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d = f - ST0.d;
-                    #endif
                }
                break;
            case 6:         /* FDIV ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f /= *(float*)ED;
-                    #else
                    ST0.d /= *(float*)ED;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f /= f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d /= f;
-                    #endif
                }
                break;
            case 7:         /* FDIVR ST0, float */
                GET_ED;
                if(!(((uintptr_t)ED)&3))
-                    #ifdef USE_FLOAT
-                    ST0.f = *(float*)ED / ST0.f;
-                    #else
                    ST0.d = *(float*)ED / ST0.d;
-                    #endif
                else {
-                    *(uint32_t*)&f = ED->dword[0];
-                    #ifdef USE_FLOAT
-                    ST0.f = f / ST0.f;
-                    #else
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d = f / ST0.d;
-                    #endif
                }
                break;
            default:
@@ -1,5 +1,3 @@
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstrict-aliasing"
    nextop = F8;
    switch (nextop) {
        case 0xC0:
@@ -10,9 +8,9 @@
        case 0xC5:
        case 0xC6:
        case 0xC7:  /* FLD STx */
-            ll = ST(nextop&7).ll;
+            ll = ST(nextop&7).q;
            fpu_do_push(emu);
-            ST0.ll = ll;
+            ST0.q = ll;
            break;
        case 0xC8:
        case 0xC9:
@@ -22,27 +20,19 @@
        case 0xCD:
        case 0xCE:
        case 0xCF:  /* FXCH STx */
-            ll = ST(nextop&7).ll;
-            ST(nextop&7).ll = ST0.ll;
-            ST0.ll = ll;
+            ll = ST(nextop&7).q;
+            ST(nextop&7).q = ST0.q;
+            ST0.q = ll;
            break;

        case 0xD0:  /* FNOP */
            break;

        case 0xE0:  /* FCHS */
-            #ifdef USE_FLOAT
-            ST0.f = -ST0.f;
-            #else
            ST0.d = -ST0.d;
-            #endif
            break;
        case 0xE1:  /* FABS */
-            #ifdef USE_FLOAT
-            ST0.f = fabsf(ST0.f);
-            #else
            ST0.d = fabs(ST0.d);
-            #endif
            break;
        
        case 0xE4:  /* FTST */
@@ -54,175 +44,128 @@

        case 0xE8:  /* FLD1 */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = 1.0f;
-            #else
            ST0.d = 1.0;
-            #endif
            break;
        case 0xE9:  /* FLDL2T */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = L2T;
-            #else
            ST0.d = L2T;
-            #endif
            break;
        case 0xEA:  /* FLDL2E */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = L2E;
-            #else
            ST0.d = L2E;
-            #endif
            break;
        case 0xEB:  /* FLDPI */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = PI;
-            #else
            ST0.d = PI;
-            #endif
            break;
        case 0xEC:  /* FLDLG2 */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = LG2;
-            #else
            ST0.d = LG2;
-            #endif
            break;
        case 0xED:  /* FLDLN2 */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = LN2;
-            #else
            ST0.d = LN2;
-            #endif
            break;
        case 0xEE:  /* FLDZ */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = 0.0f;
-            #else
            ST0.d = 0.0;
-            #endif
            break;

        case 0xF0:  /* F2XM1 */
-            #ifdef USE_FLOAT
-            ST0.f = exp2f(ST0.f) - 1.0f;
-            #else
            ST0.d = exp2(ST0.d) - 1.0;
-            #endif
            break;
        case 0xF1:  /* FYL2X */
-            #ifdef USE_FLOAT
-            ST(1).f = log2f(ST0.f)*ST(1).f;
-            #else
-            ST(1).d = log2(ST0.d)*ST(1).d;
-            #endif
+            ST(1).d *= log2(ST0.d);
            fpu_do_pop(emu);
            break;
-        case 0xF2:  /* FTAN */
-            #ifdef USE_FLOAT
-            ST0.f = tanf(ST0.f);
-            fpu_do_push(emu);
-            ST0.f = 1.0f;
-            #else
+        case 0xF2:  /* FPTAN */
            ST0.d = tan(ST0.d);
            fpu_do_push(emu);
            ST0.d = 1.0;
-            #endif
+            emu->sw.f.F87_C2 = 0;
            break;
        case 0xF3:  /* FPATAN */
-            #ifdef USE_FLOAT
-            ST1.f = atan2f(ST1.f, ST0.f);
-            #else
            ST1.d = atan2(ST1.d, ST0.d);
-            #endif
            fpu_do_pop(emu);
            break;
        case 0xF4:  /* FXTRACT */
-            #ifdef USE_FLOAT
-            tmp32s = (ST0.ll&0x7f700000)>>23;
-            tmp32s -= 127;
-            ST0.f /= exp2f(tmp32s);
-            fpu_do_push(emu);
-            ST0.f = tmp32s;
-            #else
-            tmp32s = (ST0.ll&0x7ff0000000000000LL)>>52;
-            tmp32s -= 1023;
-            ST0.d /= exp2(tmp32s);
+            ST0.d = frexp(ST0.d, &tmp32s);
            fpu_do_push(emu);
            ST0.d = tmp32s;
-            #endif
            break;

        case 0xF8:  /* FPREM */
-            #ifdef USE_FLOAT
-            tmp32s = ST0.f / ST1.f;
-            ST0.f -= ST1.f * tmp32s;
-            #else
-            tmp32s = ST0.d / ST1.d;
-            ST0.d -= ST1.d * tmp32s;
-            #endif
-            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);
+            {
+                int e0, e1;
+                frexp(ST0.d, &e0);
+                frexp(ST1.d, &e1);
+                tmp32s = e0 - e1;
+            }
+            if(tmp32s<64)
+            {
+                ll = (int64_t)floor(ST0.d/ST1.d);
+                ST0.d = ST0.d - (ST1.d*ll);
+                emu->sw.f.F87_C2 = 0;
+                emu->sw.f.F87_C1 = (ll&1)?1:0;
+                emu->sw.f.F87_C3 = (ll&2)?1:0;
+                emu->sw.f.F87_C0 = (ll&4)?1:0;
+            } else {
+                ll = (int64_t)(floor((ST0.d/ST1.d))/exp2(tmp32s - 32));
+                ST0.d = ST0.d - ST1.d*ll*exp2(tmp32s - 32);
+                emu->sw.f.F87_C2 = 1;
+            }
+            break;
+        case 0xF5:  /* FPREM1 */
+            // get exponant(ST(0))-exponant(ST(1)) in temp32s
+            {
+                int e0, e1;
+                frexp(ST0.d, &e0);
+                frexp(ST1.d, &e1);
+                tmp32s = e0 - e1;
+            }
+            if(tmp32s<64)
+            {
+                ll = (int64_t)round(ST0.d/ST1.d);
+                ST0.d = ST0.d - (ST1.d*ll);
+                emu->sw.f.F87_C2 = 0;
+                emu->sw.f.F87_C1 = (ll&1)?1:0;
+                emu->sw.f.F87_C3 = (ll&2)?1:0;
+                emu->sw.f.F87_C0 = (ll&4)?1:0;
+            } else {
+                ll = (int64_t)(trunc((ST0.d/ST1.d))/exp2(tmp32s - 32));
+                ST0.d = ST0.d - ST1.d*ll*exp2(tmp32s - 32);
+                emu->sw.f.F87_C2 = 1;
+            }
+            break;
+        case 0xF6:  /* FDECSTP */
+            emu->top=(emu->top-1)&7;    // this will probably break a few things
+            break;
+        case 0xF7:  /* FINCSTP */
+            emu->top=(emu->top+1)&7;    // this will probably break a few things
            break;
        case 0xF9:  /* FYL2XP1 */
-            #ifdef USE_FLOAT
-            ST(1).f = log2f(ST0.f + 1.0f)*ST(1).f;
-            #else
-            ST(1).d = log2(ST0.d + 1.0)*ST(1).d;
-            #endif
+            ST(1).d *= log2(ST0.d + 1.0);
            fpu_do_pop(emu);
            break;
        case 0xFA:  /* FSQRT */
-            #ifdef USE_FLOAT
-            ST0.f = sqrtf(ST0.f);
-            #else
            ST0.d = sqrt(ST0.d);
-            #endif
            break;
        case 0xFB:  /* FSINCOS */
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            sincosf(ST1.f, &ST1.f, &ST0.f);
-            #else
            sincos(ST1.d, &ST1.d, &ST0.d);
-            #endif
            break;
        case 0xFC:  /* FRNDINT */
-            #ifdef USE_FLOAT
-            ST0.f = fpu_round(emu, ST0.f);
-            #else
            ST0.d = fpu_round(emu, ST0.d);
-            #endif
            break;
        case 0xFD:  /* FSCALE */
            // this could probably be done by just altering the exponant part of the float...
-            #ifdef USE_FLOAT
-            ST0.f *= exp2f(truncf(ST1.f));
-            #else
            ST0.d *= exp2(trunc(ST1.d));
-            #endif
            break;
        case 0xFE:  /* FSIN */
-            #ifdef USE_FLOAT
-            ST0.f = sinf(ST0.f);
-            #else
            ST0.d = sin(ST0.d);
-            #endif
            break;
        case 0xFF:  /* FCOS */
-            #ifdef USE_FLOAT
-            ST0.f = cosf(ST0.f);
-            #else
            ST0.d = cos(ST0.d);
-            #endif
            break;


@@ -244,51 +187,36 @@
        case 0xE6:
        case 0xE7:
        case 0xEF:
-        case 0xF5:
-        case 0xF6:
-        case 0xF7:
            goto _default;
        default:
        switch((nextop>>3)&7) {
            case 0:     /* FLD ST0, Ed float */
                GET_ED;
                fpu_do_push(emu);
-                #ifdef USE_FLOAT
-                ST0.ll = ED->dword[0];
-                #else
                if(!(((uintptr_t)ED)&3))
                    ST0.d = *(float*)ED;
                else {
-                    memcpy(&f, ED->dword, sizeof(float));
+                    memcpy(&f, ED, sizeof(float));
                    ST0.d = f;
                }
-                #endif
                break;
            case 2:     /* FST Ed, ST0 */
                GET_ED;
-                #ifdef USE_FLOAT
-                *(float*)ED = ST0.f;
-                #else
                if(!(((uintptr_t)ED)&3))
                    *(float*)ED = ST0.d;
                else {
                    f = ST0.d;
-                    memcpy(ED->dword, &f, sizeof(float));
+                    memcpy(ED, &f, sizeof(float));
                }
-                #endif
                break;
            case 3:     /* FSTP Ed, ST0 */
                GET_ED;
-                #ifdef USE_FLOAT
-                ED->dword[0] = ST0.ll;
-                #else
                if(!(((uintptr_t)ED)&3))
                    *(float*)ED = ST0.d;
                else {
                    f = ST0.d;
-                    memcpy(ED->dword, &f, sizeof(float));
+                    memcpy(ED, &f, sizeof(float));
                }
-                #endif
                fpu_do_pop(emu);
                break;
            case 4:     /* FLDENV m */
@@ -306,9 +234,6 @@
                // warning, incomplete
                GET_ED;
                fpu_savenv(emu, (char*)ED, 0);
-                ED->dword[0] = emu->cw;
-                ED->dword[1] = emu->sw.x16;
-                // tagword: 2bits*8
                // intruction pointer: 48bits
                // data (operand) pointer: 48bits
                // last opcode: 11bits save: 16bits restaured (1st and 2nd opcode only)
@@ -321,4 +246,3 @@
                goto _default;
        }
    }
-#pragma GCC diagnostic pop
@@ -11,7 +11,7 @@
    case 0xC7:
        CHECK_FLAGS(emu);
        if(ACCESS_FLAG(F_CF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xC8:      /* FCMOVE ST(0), ST(i) */
    case 0xC9:
@@ -23,7 +23,7 @@
    case 0xCF:
        CHECK_FLAGS(emu);
        if(ACCESS_FLAG(F_ZF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xD0:      /* FCMOVBE ST(0), ST(i) */
    case 0xD1:
@@ -35,7 +35,7 @@
    case 0xD7:
        CHECK_FLAGS(emu);
        if(ACCESS_FLAG(F_CF) || ACCESS_FLAG(F_ZF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xD8:      /* FCMOVU ST(0), ST(i) */
    case 0xD9:
@@ -47,15 +47,11 @@
    case 0xDF:
        CHECK_FLAGS(emu);
        if(ACCESS_FLAG(F_PF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    
    case 0xE9:      /* FUCOMPP */
-        #ifdef USE_FLOAT
-        fpu_fcom(emu, ST1.f);
-        #else
        fpu_fcom(emu, ST1.d);   // bad, should handle QNaN and IA interrupt
-        #endif
        fpu_do_pop(emu);
        fpu_do_pop(emu);
        break;
@@ -76,19 +72,11 @@
        switch((nextop>>3)&7) {
            case 0:     /* FIADD ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f += ED->sdword[0];
-                #else
                ST0.d += ED->sdword[0];
-                #endif
                break;
            case 1:     /* FIMUL ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f *= ED->sdword[0];
-                #else
                ST0.d *= ED->sdword[0];
-                #endif
                break;
            case 2:     /* FICOM ST0, Ed int */
                GET_ED;
@@ -101,35 +89,19 @@
                break;
            case 4:     /* FISUB ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f -= ED->sdword[0];
-                #else
                ST0.d -= ED->sdword[0];
-                #endif
                break;
            case 5:     /* FISUBR ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f = ED->sdword[0] - ST0.f;
-                #else
-                ST0.d = ED->sdword[0] - ST0.d;
-                #endif
+                ST0.d = (double)ED->sdword[0] - ST0.d;
                break;
            case 6:     /* FIDIV ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f /= ED->sdword[0];
-                #else
                ST0.d /= ED->sdword[0];
-                #endif
                break;
            case 7:     /* FIDIVR ST0, Ed int */
                GET_ED;
-                #ifdef USE_FLOAT
-                ST0.f = ED->sdword[0] / ST0.f;
-                #else
-                ST0.d = ED->sdword[0] / ST0.d;
-                #endif
+                ST0.d = (double)ED->sdword[0] / ST0.d;
                break;
        }
    }
@@ -10,7 +10,7 @@
    case 0xC7:
        CHECK_FLAGS(emu);
        if(!ACCESS_FLAG(F_CF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xC8:      /* FCMOVNE ST(0), ST(i) */
    case 0xC9:
@@ -22,7 +22,7 @@
    case 0xCF:
        CHECK_FLAGS(emu);
        if(!ACCESS_FLAG(F_ZF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xD0:      /* FCMOVNBE ST(0), ST(i) */
    case 0xD1:
@@ -34,7 +34,7 @@
    case 0xD7:
        CHECK_FLAGS(emu);
        if(!(ACCESS_FLAG(F_CF) || ACCESS_FLAG(F_ZF)))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;
    case 0xD8:      /* FCMOVNU ST(0), ST(i) */
    case 0xD9:
@@ -46,9 +46,11 @@
    case 0xDF:
        CHECK_FLAGS(emu);
        if(!ACCESS_FLAG(F_PF))
-            ST0.ll = ST(nextop&7).ll;
+            ST0.q = ST(nextop&7).q;
        break;

+    case 0xE1:      /* FDISI8087_NOP */
+        break;
    case 0xE2:      /* FNCLEX */
        //Clears the floating-point exception flags (PE, UE, OE, ZE, DE, and IE), 
        // the exception summary status flag (ES), the stack fault flag (SF), and the busy flag (B) in the FPU status word
@@ -73,11 +75,7 @@
    case 0xED:
    case 0xEE:
    case 0xEF:
-        #ifdef USE_FLOAT
-        fpu_fcomi(emu, ST(nextop&7).f);
-        #else
        fpu_fcomi(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
-        #endif
        break;

    case 0xF0:  /* FCOMI ST0, STx */
@@ -88,14 +86,9 @@
    case 0xF5:
    case 0xF6:
    case 0xF7:
-        #ifdef USE_FLOAT
-        fpu_fcomi(emu, ST(nextop&7).f);
-        #else
        fpu_fcomi(emu, ST(nextop&7).d);
-        #endif
        break;
    case 0xE0:
-    case 0xE1:
    case 0xE4:
    case 0xE5:
    case 0xE6:
@@ -103,134 +96,49 @@
        goto _default;
    default:
        switch((nextop>>3)&7) {
-            case 0: /* FILD ST0, Gd */
+            case 0: /* FILD ST0, Ed */
                GET_ED;
                fpu_do_push(emu);
-                #ifdef USE_FLOAT
-                ST0.f = ED->sdword[0];
-                #else
                ST0.d = ED->sdword[0];
-                #endif
                break;
            case 1: /* FISTTP Ed, ST0 */
                GET_ED;
-                #ifdef USE_FLOAT
-                tmp32s = ST0.f; // TODO: Handling of FPU Exception
-                if(tmp32s==0x7fffffff && isgreater(ST0.f, (float)(int32_t)0x7fffffff))
-                    tmp32s = 0x80000000;
-                #else
                tmp32s = ST0.d; // TODO: Handling of FPU Exception
                if(tmp32s==0x7fffffff && isgreater(ST0.d, (double)(int32_t)0x7fffffff))
                    tmp32s = 0x80000000;
-                #endif
                fpu_do_pop(emu);
                ED->sdword[0] = tmp32s;
                break;
            case 2: /* FIST Ed, ST0 */
                GET_ED;
-                #ifdef USE_FLOAT
-                if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
+                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff) || !isfinite(ST0.d))
                    ED->sdword[0] = 0x80000000;
                else {
-                    switch(emu->round) {
-                        case ROUND_Nearest:
-                            ED->sdword[0] = floorf(ST0.f+0.5);
-                            break;
-                        case ROUND_Down:
-                            ED->sdword[0] = floorf(ST0.f);
-                            break;
-                        case ROUND_Up:
-                            ED->sdword[0] = ceilf(ST0.f);
-                            break;
-                        case ROUND_Chop:
-                            ED->sdword[0] = ST0.f;
-                            break;
-                    }
+                    volatile int32_t tmp = fpu_round(emu, ST0.d);    // tmp to avoid BUS ERROR
+                    ED->sdword[0] = tmp;
                }
-                #else
-                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
-                    ED->sdword[0] = 0x80000000;
-                else {
-                    switch(emu->round) {
-                        case ROUND_Nearest:
-                            ED->sdword[0] = floor(ST0.d+0.5);
-                            break;
-                        case ROUND_Down:
-                            ED->sdword[0] = floor(ST0.d);
-                            break;
-                        case ROUND_Up:
-                            ED->sdword[0] = ceil(ST0.d);
-                            break;
-                        case ROUND_Chop:
-                            ED->sdword[0] = ST0.d;
-                            break;
-                    }
-                }
-                #endif
                break;
            case 3: /* FISTP Ed, ST0 */
                GET_ED;
-                #ifdef USE_FLOAT
-                if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
+                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff) || !isfinite(ST0.d))
                    ED->sdword[0] = 0x80000000;
                else {
-                    switch(emu->round) {
-                        case ROUND_Nearest:
-                            ED->sdword[0] = floorf(ST0.f+0.5);
-                            break;
-                        case ROUND_Down:
-                            ED->sdword[0] = floorf(ST0.f);
-                            break;
-                        case ROUND_Up:
-                            ED->sdword[0] = ceilf(ST0.f);
-                            break;
-                        case ROUND_Chop:
-                            ED->sdword[0] = ST0.f;
-                            break;
-                    }
+                    volatile int32_t tmp = fpu_round(emu, ST0.d);    // tmp to avoid BUS ERROR
+                    ED->sdword[0] = tmp;
                }
-                #else
-                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
-                    ED->sdword[0] = 0x80000000;
-                else {
-                    switch(emu->round) {
-                        case ROUND_Nearest:
-                            ED->sdword[0] = floor(ST0.d+0.5);
-                            break;
-                        case ROUND_Down:
-                            ED->sdword[0] = floor(ST0.d);
-                            break;
-                        case ROUND_Up:
-                            ED->sdword[0] = ceil(ST0.d);
-                            break;
-                        case ROUND_Chop:
-                            ED->sdword[0] = ST0.d;
-                            break;
-                    }
-                }
-                #endif
                fpu_do_pop(emu);
                break;
-            case 5: /* FLD ST0, Gt */
+            case 5: /* FLD ST0, Et */
                GET_ED;
                fpu_do_push(emu);
                memcpy(&STld(0).ld, ED, 10);
-                #ifdef USE_FLOAT
-                LD2D(&STld(0), &d);
-                ST0.f = d;
-                #else
                LD2D(&STld(0), &ST(0).d);
-                #endif
-                STld(0).ref = ST0.ll;
+                STld(0).ref = ST0.q;
                break;
            case 7: /* FSTP tbyte */
                GET_ED;
-                if(ST0.ll!=STld(0).ref)
-                    #ifdef USE_FLOAT
-                    {d = ST0.f; D2LD(&d, ED);}
-                    #else
+                if(ST0.q!=STld(0).ref)
                    D2LD(&ST0.d, ED);
-                    #endif
                else
                    memcpy(ED, &STld(0).ld, 10);
                fpu_do_pop(emu);
@@ -1,5 +1,3 @@
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Wstrict-aliasing"
    nextop = F8;
    switch(nextop) {
        case 0xC0:
@@ -10,11 +8,7 @@
        case 0xC5:
        case 0xC6:
        case 0xC7:  /* FADD */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f += ST0.f;
-            #else
            ST(nextop&7).d += ST0.d;
-            #endif
            break;
        case 0xC8:
        case 0xC9:
@@ -24,11 +18,7 @@
        case 0xCD:
        case 0xCE:
        case 0xCF:  /* FMUL */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f *= ST0.f;
-            #else
            ST(nextop&7).d *= ST0.d;
-            #endif
            break;
        case 0xD0:
        case 0xD1:
@@ -38,11 +28,7 @@
        case 0xD5:
        case 0xD6:
        case 0xD7:  /* FCOM */
-            #ifdef USE_FLOAT
-            fpu_fcom(emu, ST(nextop&7).f);
-            #else
            fpu_fcom(emu, ST(nextop&7).d);
-            #endif
            break;
        case 0xD8:
        case 0xD9:
@@ -52,11 +38,7 @@
        case 0xDD:
        case 0xDE:
        case 0xDF:  /* FCOMP */
-            #ifdef USE_FLOAT
-            fpu_fcom(emu, ST(nextop&7).f);
-            #else
-            fpu_fcom(emu, ST(nextop&7).d);  // TODO: is this ok?
-            #endif
+            fpu_fcom(emu, ST(nextop&7).d);
            fpu_do_pop(emu);
            break;
        case 0xE0:
@@ -67,11 +49,7 @@
        case 0xE5:
        case 0xE6:
        case 0xE7:  /* FSUBR */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f = ST0.f -ST(nextop&7).f;
-            #else
            ST(nextop&7).d = ST0.d -ST(nextop&7).d;
-            #endif
            break;
        case 0xE8:
        case 0xE9:
@@ -81,11 +59,7 @@
        case 0xED:
        case 0xEE:
        case 0xEF:  /* FSUB */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f -= ST0.f;
-            #else
            ST(nextop&7).d -= ST0.d;
-            #endif
            break;
        case 0xF0:
        case 0xF1:
@@ -95,11 +69,7 @@
        case 0xF5:
        case 0xF6:
        case 0xF7:  /* FDIVR */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f = ST0.f / ST(nextop&7).f;
-            #else
            ST(nextop&7).d = ST0.d / ST(nextop&7).d;
-            #endif
            break;
        case 0xF8:
        case 0xF9:
@@ -109,54 +79,32 @@
        case 0xFD:
        case 0xFE:
        case 0xFF:  /* FDIV */
-            #ifdef USE_FLOAT
-            ST(nextop&7).f /=  ST0.f;
-            #else
            ST(nextop&7).d /=  ST0.d;
-            #endif
            break;
        default:
            GET_ED;
            switch((nextop>>3)&7) {
            case 0:         /* FADD ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f += *(double*)ED;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f += d;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d += *(double*)ED;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d += d;
                }
-                #endif
                break;
            case 1:         /* FMUL ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f *= *(double*)ED;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f *= d;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d *= *(double*)ED;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d *= d;
                }
-                #endif
                break;
            case 2:      /* FCOM ST0, double */
                if(!(((uintptr_t)ED)&7))
                    fpu_fcom(emu, *(double*)ED);
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    fpu_fcom(emu, d);
                }
                break;
@@ -164,78 +112,42 @@
                if(!(((uintptr_t)ED)&7))
                    fpu_fcom(emu, *(double*)ED);
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    fpu_fcom(emu, d);
                }
                fpu_do_pop(emu);
                break;
            case 4:         /* FSUB ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f -= *(double*)ED;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f -= d;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d -= *(double*)ED;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d -= d;
                }
-                #endif
                break;
            case 5:         /* FSUBR ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f = *(double*)ED - ST0.f;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f = d - ST0.f;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d = *(double*)ED - ST0.d;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d = d - ST0.d;
                }
-                #endif
                break;
            case 6:         /* FDIV ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f /= *(double*)ED;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f /= d;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d /= *(double*)ED;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d /= d;
                }
-                #endif
                break;
            case 7:         /* FDIVR ST0, double */
-                #ifdef USE_FLOAT
-                if(!(((uintptr_t)ED)&7))
-                    ST0.f = *(double*)ED / ST0.f;
-                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
-                    ST0.f = d / ST0.f;
-                }
-                #else
                if(!(((uintptr_t)ED)&7))
                    ST0.d = *(double*)ED / ST0.d;
                else {
-                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    memcpy(&d, ED, sizeof(double));
                    ST0.d = d / ST0.d;
                }
-                #endif
                break;
            default:
                goto _default;
@@ -9,7 +9,8 @@
    case 0xC5:
    case 0xC6:
    case 0xC7:
-        break;  // not handling Tag...
+        fpu_do_free(emu, nextop-0xC0);
+        break;

    case 0xD0:  /* FST ST0, STx */
    case 0xD1:
@@ -19,7 +20,7 @@
    case 0xD5:
    case 0xD6:
    case 0xD7:
-        ST(nextop&7).ll = ST0.ll;
+        ST(nextop&7).q = ST0.q;
        break;
    case 0xD8:  /* FSTP ST0, STx */
    case 0xD9:
@@ -29,7 +30,7 @@
    case 0xDD:
    case 0xDE:
    case 0xDF:
-        ST(nextop&7).ll = ST0.ll;
+        ST(nextop&7).q = ST0.q;
        fpu_do_pop(emu);
        break;
    case 0xE0:  /* FUCOM ST0, STx */
@@ -40,11 +41,7 @@
    case 0xE5:
    case 0xE6:
    case 0xE7:
-        #ifdef USE_FLOAT
-        fpu_fcom(emu, ST(nextop&7).f);
-        #else
        fpu_fcom(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
-        #endif
        break;
    case 0xE8:  /* FUCOMP ST0, STx */
    case 0xE9:
@@ -54,11 +51,7 @@
    case 0xED:
    case 0xEE:
    case 0xEF:
-        #ifdef USE_FLOAT
-        fpu_fcom(emu, ST(nextop&7).f);
-        #else
        fpu_fcom(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
-        #endif
        fpu_do_pop(emu);
        break;

@@ -93,39 +86,37 @@
            case 0: /* FLD double */
                GET_ED;
                fpu_do_push(emu);
-                #ifdef USE_FLOAT
-                *(uint64_t*)&d = *(int64_t*)ED;
-                ST0.f = d;
-                #else
-                ST0.ll = *(int64_t*)ED;
-                #endif
+                if(!(((uintptr_t)ED)&7))
+                    ST0.d = *(double*)ED;
+                else {
+                    memcpy(&ST0.d, ED, sizeof(double));
+                }
                break;
            case 1: /* FISTTP ED qword */
                GET_ED;
-                #ifdef USE_FLOAT
-                *(int64_t*)ED = ST0.f;
-                #else
-                *(int64_t*)ED = ST0.d;
-                #endif
+                if(!(((uintptr_t)ED)&7))
+                    *(int64_t*)ED = ST0.d;
+                else {
+                    int64_t i64 = ST0.d;
+                    memcpy(ED, &i64, sizeof(int64_t));
+                }
                fpu_do_pop(emu);
                break;
            case 2: /* FST double */
                GET_ED;
-                #ifdef USE_FLOAT
-                d = ST0.f;
-                *(int64_t*)ED = *(uint64_t*)&d;
-                #else
-                *(int64_t*)ED = ST0.ll;
-                #endif
+                if(!(((uintptr_t)ED)&7))
+                    *(double*)ED = ST0.d;
+                else {
+                    memcpy(ED, &ST0.d, sizeof(double));
+                }
                break;
            case 3: /* FSTP double */
                GET_ED;
-                #ifdef USE_FLOAT
-                d = ST0.f;
-                *(int64_t*)ED = *(uint64_t*)&d;
-                #else
-                *(int64_t*)ED = ST0.ll;
-                #endif
+                if(!(((uintptr_t)ED)&7))
+                    *(double*)ED = ST0.d;
+                else {
+                    memcpy(ED, &ST0.d, sizeof(double));
+                }
                fpu_do_pop(emu);
                break;
            case 4: /* FRSTOR m108byte */
@@ -136,12 +127,7 @@
                    char* p =(char*)ED;
                    p += 28;
                    for (int i=0; i<8; ++i) {
-                        #ifdef USE_FLOAT
-                        d = ST(i).f;
-                        LD2D(p, &d);
-                        #else
                        LD2D(p, &ST(i).d);
-                        #endif
                        p+=10;
                    }
                }
@@ -156,12 +142,7 @@
                    char* p =(char*)ED;
                    p += 28;
                    for (int i=0; i<8; ++i) {
-                        #ifdef USE_FLOAT
-                        D2LD(&d, p);
-                        ST(i).f = d;
-                        #else
                        D2LD(&ST(i).d, p);
-                        #endif
                        p+=10;
                    }
                }
@@ -1,55 +1,41 @@
    nextop = F8;
    switch (nextop) {
-    case 0xC1:  /* FADDP ST1, ST0 */
-        #ifdef USE_FLOAT
-        ST(1).f += ST0.f;
-        #else
-        ST(1).d += ST0.d;
-        #endif
-        fpu_do_pop(emu);
-        break;
    case 0xC0:  /* FADDP STx, ST0 */
+    case 0xC1:  /* FADDP ST1, ST0 */
    case 0xC2:
    case 0xC3:
    case 0xC4:
    case 0xC5:
    case 0xC6:
    case 0xC7:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f += ST0.f;
-        #else
        ST(nextop&7).d += ST0.d;
-        #endif
-        fpu_do_pop(emu);
-        break;
-    case 0xC9:  /* FMULP ST1, ST0 */
-        #ifdef USE_FLOAT
-        ST1.f *= ST0.f;
-        #else
-        ST1.d *= ST0.d;
-        #endif
        fpu_do_pop(emu);
        break;
    case 0xC8:  /* FMULP STx, ST0 */
+    case 0xC9:  /* FMULP ST1, ST0 */
    case 0xCA:
    case 0xCB:
    case 0xCC:
    case 0xCD:
    case 0xCE:
    case 0xCF:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f *= ST0.f;
-        #else
        ST(nextop&7).d *= ST0.d;
-        #endif
        fpu_do_pop(emu);
        break;
+    case 0xD0:
+    case 0xD1:
+    case 0xD2:
+    case 0xD3:
+    case 0xD4:
+    case 0xD5:
+    case 0xD6:
+    case 0xD7:  /* FCOMP */
+        fpu_fcom(emu, ST(nextop&7).d);
+        fpu_do_pop(emu);
+        break;
+
    case 0xD9:  /* FCOMPP */
-        #ifdef USE_FLOAT
-        fpu_fcom(emu, ST1.f);
-        #else
        fpu_fcom(emu, ST1.d);
-        #endif
        fpu_do_pop(emu);
        fpu_do_pop(emu);
        break;
@@ -65,33 +51,18 @@
    case 0xE5:
    case 0xE6:
    case 0xE7:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f = ST0.f - ST(nextop&7).f;
-        #else
        ST(nextop&7).d = ST0.d - ST(nextop&7).d;
-        #endif
-        fpu_do_pop(emu);
-        break;
-   case 0xE9:  /* FSUBP ST1, ST0 */
-        #ifdef USE_FLOAT
-        ST1.f -= ST0.f;
-        #else
-        ST1.d -= ST0.d;
-        #endif
        fpu_do_pop(emu);
        break;
    case 0xE8:  /* FSUBP STx, ST0 */
+    case 0xE9:  /* FSUBP ST1, ST0 */
    case 0xEA:
    case 0xEB:
    case 0xEC:
    case 0xED:
    case 0xEE:
    case 0xEF:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f -= ST0.f;
-        #else
        ST(nextop&7).d -= ST0.d;
-        #endif
        fpu_do_pop(emu);
        break;
    case 0xF0:  /* FDIVRP STx, ST0 */
@@ -102,11 +73,7 @@
    case 0xF5:
    case 0xF6:
    case 0xF7:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f = ST0.f / ST(nextop&7).f;
-        #else
        ST(nextop&7).d = ST0.d / ST(nextop&7).d;
-        #endif
        fpu_do_pop(emu);
        break;
    case 0xF8:  /* FDIVP STx, ST0 */
@@ -117,23 +84,11 @@
    case 0xFD:
    case 0xFE:
    case 0xFF:
-        #ifdef USE_FLOAT
-        ST(nextop&7).f /= ST0.f;
-        #else
        ST(nextop&7).d /= ST0.d;
-        #endif
        fpu_do_pop(emu);
        break;

-    case 0xD0:
-    case 0xD1:
-    case 0xD2:
-    case 0xD3:
-    case 0xD4:
-    case 0xD5:
-    case 0xD6:
-    case 0xD7:
-    case 0xD8:
+    case 0xD8:  
    case 0xDA:
    case 0xDB:
    case 0xDC:
@@ -146,51 +101,27 @@
        switch((nextop>>3)&7) {
            case 0:     /* FIADD ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f += EW->sword[0];
-                #else
                ST0.d += EW->sword[0];
-                #endif
                break;
            case 1:     /* FIMUL ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f *= EW->sword[0];
-                #else
                ST0.d *= EW->sword[0];
-                #endif
                break;
            case 4:     /* FISUB ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f -= EW->sword[0];
-                #else
                ST0.d -= EW->sword[0];
-                #endif
                break;
            case 5:     /* FISUBR ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f = EW->sword[0] - ST0.f;
-                #else
-                ST0.d = EW->sword[0] - ST0.d;
-                #endif
+                ST0.d = (double)EW->sword[0] - ST0.d;
                break;
            case 6:     /* FIDIV ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f /= EW->sword[0];
-                #else
                ST0.d /= EW->sword[0];
-                #endif
                break;
            case 7:     /* FIDIVR ST0, Ew int */
                GET_EW;
-                #ifdef USE_FLOAT
-                ST0.f = EW->sword[0] / ST0.f;
-                #else
-                ST0.d = EW->sword[0] / ST0.d;
-                #endif
+                ST0.d = (double)EW->sword[0] / ST0.d;
                break;
        default:
            goto _default;
@@ -1,5 +1,29 @@
    nextop = F8;
    switch (nextop) {
+    case 0xC0:  /* FFREEP STx */
+    case 0xC1:
+    case 0xC2:
+    case 0xC3:
+    case 0xC4:
+    case 0xC5:
+    case 0xC6:
+    case 0xC7:
+        fpu_do_free(emu, nextop-0xC0);
+        fpu_do_pop(emu);
+        break;
+
+    case 0xD0:  /* FSTP STx, ST0 */
+    case 0xD1:
+    case 0xD2:
+    case 0xD3:
+    case 0xD4:
+    case 0xD5:
+    case 0xD6:
+    case 0xD7:
+        ST(nextop&7).q = ST0.q;
+        fpu_do_pop(emu);
+        break;
+
    case 0xE0:  /* FNSTSW AX */
        emu->sw.f.F87_TOP = emu->top&7;
        R_AX = emu->sw.x16;
@@ -13,11 +37,7 @@
    case 0xED:
    case 0xEE:
    case 0xEF:
-        #ifdef USE_FLOAT
-        fpu_fcomi(emu, ST(nextop&7).f);
-        #else
        fpu_fcomi(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
-        #endif
        fpu_do_pop(emu);
        break;

@@ -29,22 +49,10 @@
    case 0xF5:
    case 0xF6:
    case 0xF7:
-        #ifdef USE_FLOAT
-        fpu_fcomi(emu, ST(nextop&7).f);
-        #else
        fpu_fcomi(emu, ST(nextop&7).d);
-        #endif
        fpu_do_pop(emu);
        break;

-    case 0xC0:
-    case 0xC1:
-    case 0xC2:
-    case 0xC3:
-    case 0xC4:
-    case 0xC5:
-    case 0xC6:
-    case 0xC7:
    case 0xC8:
    case 0xC9:
    case 0xCA:
@@ -53,14 +61,6 @@
    case 0xCD:
    case 0xCE:
    case 0xCF:
-    case 0xD0:
-    case 0xD1:
-    case 0xD2:
-    case 0xD3:
-    case 0xD4:
-    case 0xD5:
-    case 0xD6:
-    case 0xD7:
    case 0xD8:
    case 0xD9:
    case 0xDA:
@@ -92,105 +92,27 @@
            GET_EW;
            tmp16s = EW->sword[0];
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = tmp16s;
-            #else
            ST0.d = tmp16s;
-            #endif
            break;
        case 1: /* FISTTP Ew, ST0 */
            GET_EW;
-            #ifdef USE_FLOAT
-            tmp16s = ST0.f;
-            #else
            tmp16s = ST0.d;
-            #endif
            EW->sword[0] = tmp16s;
            fpu_do_pop(emu);
            break;
        case 2: /* FIST Ew, ST0 */
            GET_EW;
-            #ifdef USE_FLOAT
-            if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff) || !isfinite(ST0.d))
                EW->sword[0] = 0x8000;
-            else {
-                switch(emu->round) {
-                    case ROUND_Nearest:
-                        EW->sword[0] = floorf(ST0.f+0.5);
-                        break;
-                    case ROUND_Down:
-                        EW->sword[0] = floorf(ST0.f);
-                        break;
-                    case ROUND_Up:
-                        EW->sword[0] = ceilf(ST0.f);
-                        break;
-                    case ROUND_Chop:
-                        EW->sword[0] = ST0.f;
-                        break;
-                }
-            }
-            #else
-            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
-                EW->sword[0] = 0x8000;
-            else {
-                switch(emu->round) {
-                    case ROUND_Nearest:
-                        EW->sword[0] = floor(ST0.d+0.5);
-                        break;
-                    case ROUND_Down:
-                        EW->sword[0] = floor(ST0.d);
-                        break;
-                    case ROUND_Up:
-                        EW->sword[0] = ceil(ST0.d);
-                        break;
-                    case ROUND_Chop:
-                        EW->sword[0] = ST0.d;
-                        break;
-                }
-            }
-            #endif
+            else
+                EW->sword[0] = fpu_round(emu, ST0.d);
            break;
        case 3: /* FISTP Ew, ST0 */
            GET_EW;
-            #ifdef USE_FLOAT
-            if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff) || !isfinite(ST0.d))
                EW->sword[0] = 0x8000;
-            else {
-                switch(emu->round) {
-                    case ROUND_Nearest:
-                        EW->sword[0] = floorf(ST0.f+0.5);
-                        break;
-                    case ROUND_Down:
-                        EW->sword[0] = floorf(ST0.f);
-                        break;
-                    case ROUND_Up:
-                        EW->sword[0] = ceilf(ST0.f);
-                        break;
-                    case ROUND_Chop:
-                        EW->sword[0] = ST0.f;
-                        break;
-                }
-            }
-            #else
-            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
-                EW->sword[0] = 0x8000;
-            else {
-                switch(emu->round) {
-                    case ROUND_Nearest:
-                        EW->sword[0] = floor(ST0.d+0.5);
-                        break;
-                    case ROUND_Down:
-                        EW->sword[0] = floor(ST0.d);
-                        break;
-                    case ROUND_Up:
-                        EW->sword[0] = ceil(ST0.d);
-                        break;
-                    case ROUND_Chop:
-                        EW->sword[0] = ST0.d;
-                        break;
-                }
-            }
-            #endif
+            else
+                EW->sword[0] = fpu_round(emu, ST0.d);
            fpu_do_pop(emu);
            break;
        case 4: /* FBLD ST0, tbytes */
@@ -202,13 +124,9 @@
            GET_ED;
            tmp64s = *(int64_t*)ED;
            fpu_do_push(emu);
-            #ifdef USE_FLOAT
-            ST0.f = tmp64s;
-            #else
            ST0.d = tmp64s;
-            #endif
            STll(0).ll = tmp64s;
-            STll(0).ref = ST0.ll;
+            STll(0).ref = ST0.q;
            break;
        case 6: /* FBSTP tbytes, ST0 */
            GET_ED;
@@ -217,20 +135,27 @@
            break;
        case 7: /* FISTP i64 */
            GET_ED;
-            if(STll(0).ref==ST(0).ll) {
-                *(int64_t*)ED = STll(0).ll;
+            if((uintptr_t)ED & 0x7) {
+                // un-aligned!
+                if(STll(0).ref==ST(0).q)
+                    memcpy(ED, &STll(0).ll, sizeof(int64_t));
+                else {
+                    int64_t i64;
+                    if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(int64_t)0x7fffffffffffffffLL) || !isfinite(ST0.d))
+                        i64 = 0x8000000000000000LL;
+                    else
+                        i64 = fpu_round(emu, ST0.d);
+                    memcpy(ED, &i64, sizeof(int64_t));
+                }
            } else {
-                #ifdef USE_FLOAT
-                if(isgreater(ST0.f, (float)(int64_t)0x7fffffffffffffffLL) || isless(ST0.f, (float)(int64_t)0x8000000000000000LL))
-                    *(int64_t*)ED = 0x8000000000000000LL;
-                else
-                    *(int64_t*)ED = (int64_t)ST0.f;
-                #else
-                if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL))
-                    *(int64_t*)ED = 0x8000000000000000LL;
-                else
-                    *(int64_t*)ED = (int64_t)ST0.d;
-                #endif
+                if(STll(0).ref==ST(0).q)
+                    *(int64_t*)ED = STll(0).ll;
+                else {
+                    if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(int64_t)0x7fffffffffffffffLL) || !isfinite(ST0.d))
+                        *(int64_t*)ED = 0x8000000000000000LL;
+                    else
+                        *(int64_t*)ED = fpu_round(emu, ST0.d);
+                }
            }
            fpu_do_pop(emu);
            break;
@@ -35,7 +35,7 @@
    case 0x2D:  /* CVTSD2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        switch(emu->round) {
+        switch((emu->mxcsr>>13)&3) {
            case ROUND_Nearest:
                GD.sdword[0] = floor(EX->d[0]+0.5);
                break;
@@ -66,7 +66,6 @@
        GET_EX;
        GX.d[0] *= EX->d[0];
        break;
-
    case 0x5A:  /* CVTSD2SS Gx, Ex */
        nextop = F8;
        GET_EX;
@@ -111,6 +110,107 @@
        }
        break;

+    case 0x7C:  /* HADDPS Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.f[0] += GX.f[1];
+        GX.f[1] = GX.f[2] + GX.f[3];
+        if(EX==&GX) {
+            GX.f[2] = GX.f[0];
+            GX.f[3] = GX.f[1];
+        } else {
+            GX.f[2] = EX->f[0] + EX->f[1];
+            GX.f[3] = EX->f[2] + EX->f[3];
+        }
+        break;
+
+    #define GOCOND(BASE, PREFIX, CONDITIONAL)   \
+    case BASE+0:                                \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_OF))                   \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+1:                                \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_OF))                  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+2:                                \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_CF))                   \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+3:                                \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_CF))                  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+4:                                \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_ZF))                   \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+5:                                \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_ZF))                  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+6:                                \
+        PREFIX                                  \
+        if((ACCESS_FLAG(F_ZF) || ACCESS_FLAG(F_CF)))  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+7:                                \
+        PREFIX                                  \
+        if(!(ACCESS_FLAG(F_ZF) || ACCESS_FLAG(F_CF))) \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+8:                                \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_SF))                   \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+9:                                \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_SF))                  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xA:                              \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_PF))                   \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xB:                              \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_PF))                  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xC:                              \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_SF) != ACCESS_FLAG(F_OF))  \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xD:                              \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_SF) == ACCESS_FLAG(F_OF)) \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xE:                              \
+        PREFIX                                  \
+        if(ACCESS_FLAG(F_ZF) || (ACCESS_FLAG(F_SF) != ACCESS_FLAG(F_OF))) \
+            CONDITIONAL                         \
+        break;                                  \
+    case BASE+0xF:                              \
+        PREFIX                                  \
+        if(!ACCESS_FLAG(F_ZF) && (ACCESS_FLAG(F_SF) == ACCESS_FLAG(F_OF))) \
+            CONDITIONAL                         \
+        break;
+        
+    GOCOND(0x80
+        , tmp32s = F32S; CHECK_FLAGS(emu);
+        , ip += tmp32s;
+    )                               /* 0x80 -> 0x8F Jxx */
+        
    case 0xC2:  /* CMPSD Gx, Ex, Ib */
        nextop = F8;
        GET_EX;
@@ -129,6 +229,15 @@
        GX.q[0]=(tmp8s)?0xffffffffffffffffLL:0LL;
        break;

+    case 0xD0:  /* ADDSUBPS Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.f[0] -= EX->f[0];
+        GX.f[1] += EX->f[1];
+        GX.f[2] -= EX->f[2];
+        GX.f[3] += EX->f[3];
+        break;
+
    case 0xD6:  /* MOVDQ2Q Gm, Ex */
        nextop = F8;
        GET_EX;
@@ -138,11 +247,33 @@
    case 0xE6:  /* CVTPD2DQ Gx, Ex */
        nextop = F8;
        GET_EX;
-        GX.sd[0] = EX->d[0];
-        GX.sd[1] = EX->d[1];
+        switch((emu->mxcsr>>13)&3) {
+            case ROUND_Nearest:
+                GX.sd[0] = floor(EX->d[0]+0.5);
+                GX.sd[1] = floor(EX->d[1]+0.5);
+                break;
+            case ROUND_Down:
+                GX.sd[0] = floor(EX->d[0]);
+                GX.sd[1] = floor(EX->d[1]);
+                break;
+            case ROUND_Up:
+                GX.sd[0] = ceil(EX->d[0]);
+                GX.sd[1] = ceil(EX->d[1]);
+                break;
+            case ROUND_Chop:
+                GX.sd[0] = EX->d[0];
+                GX.sd[1] = EX->d[1];
+                break;
+        }
        GX.q[1] = 0;
        break;

+    case 0xF0:  /* LDDQU Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        memcpy(&GX, EX, 16);
+        break;
+
    default:
        goto _default;
    }
@@ -15,6 +15,28 @@
        GET_EX;
        EX->ud[0] = GX.ud[0];
        break;
+    case 0x12:  /* MOVSLDUP Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.ud[1] = GX.ud[0] = EX->ud[0];
+        GX.ud[3] = GX.ud[2] = EX->ud[2];
+        break;
+    
+    case 0x16:  /* MOVSHDUP Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.ud[1] = GX.ud[0] = EX->ud[1];
+        GX.ud[3] = GX.ud[2] = EX->ud[3];
+        break;
+    
+    case 0x1E:  /* NOP / ENDBR32 / ENDBR64 */
+        nextop = F8;
+        GET_ED;
+        break;
+    case 0x1F:  /* REP NOP */
+        nextop = F8;
+        GET_ED;
+        break;

    case 0x2A:  /* CVTSI2SS Gx, Ed */
        nextop = F8;
@@ -30,7 +52,7 @@
    case 0x2D:  /* CVTSS2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        switch(emu->round) {
+        switch((emu->mxcsr>>13)&3) {
            case ROUND_Nearest:
                GD.sdword[0] = floor(EX->f[0]+0.5);
                break;
@@ -141,7 +163,7 @@
        memcpy(EX, &GX, 16);    // unaligned...
        break;

-    case 0xBC:  /* TZCNT Ew,Gw */
+    case 0xBC:  /* TZCNT Ed,Gd */
        CHECK_FLAGS(emu);
        nextop = F8;
        GET_ED;
@@ -151,7 +173,7 @@
            tmp8u = 0;
            while(!(tmp32u&(1<<tmp8u))) ++tmp8u;
            GD.dword[0] = tmp8u;
-            CONDITIONAL_SET_FLAG(tmp8u, F_ZF);
+            CONDITIONAL_SET_FLAG(tmp8u==0, F_ZF);
            CLEAR_FLAG(F_CF);
        } else {
            CLEAR_FLAG(F_ZF);
@@ -5,6 +5,7 @@
 #include <unistd.h>
 #include <sys/syscall.h>
 #include <sys/time.h>
+#include <sys/mman.h>

 #include "debug.h"
 #include "box86stack.h"
@@ -12,9 +13,16 @@
 #include "x86emu_private.h"
 #include "x87emu_private.h"
 #include "box86context.h"
-#include "x86trace.h"
 #include "x86run.h"
 #include "x86run_private.h"
+#include "callback.h"
+#include "bridge.h"
+#ifdef HAVE_TRACE
+#include "x86trace.h"
+#endif
+#ifdef DYNAREC
+#include "custommem.h"
+#endif

 typedef struct cleanup_s {
    void*       f;
@@ -34,15 +42,20 @@ static uint32_t x86emu_parity_tab[8] =
 	0x69969669,
 };

-static uint8_t EndEmuMarker[] = {0xcc, 'S', 'C', 0, 0, 0, 0};
+uint32_t* GetParityTab()
+{
+    return x86emu_parity_tab;
+}
+
 void PushExit(x86emu_t* emu)
 {
-    Push(emu, (uint32_t)&EndEmuMarker);
+    uintptr_t endMarker = AddCheckBridge(my_context->system, NULL, NULL, 0);
+    Push(emu, endMarker);
 }

 void* GetExit()
 {
-    return &EndEmuMarker;
+    return (void*)AddCheckBridge(my_context->system, NULL, NULL, 0);
 }

 static void internalX86Setup(x86emu_t* emu, box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
@@ -53,11 +66,9 @@ static void internalX86Setup(x86emu_t* emu, box86context_t *context, uintptr_t s
        emu->sbiidx[i] = &emu->regs[i];
    emu->sbiidx[4] = &emu->zero;
    emu->x86emu_parity_tab = x86emu_parity_tab;
-    emu->packed_eflags.x32 = 0x202; // default flags?
-    UnpackFlags(emu);
+    emu->eflags.x32 = 0x202; // default flags?
    // own stack?
-    if(ownstack)
-        emu->stack = (void*)stack;
+    emu->stack2free = (ownstack)?(void*)stack:NULL;
    emu->init_stack = (void*)stack;
    emu->size_stack = stacksize;
    // set default value
@@ -66,18 +77,14 @@ static void internalX86Setup(x86emu_t* emu, box86context_t *context, uintptr_t s
    // fake init of segments...
    emu->segs[_CS] = 0x73;
    emu->segs[_DS] = emu->segs[_ES] = emu->segs[_SS] = 0x7b;
-    emu->segs[_FS] = 0;
+    emu->segs[_FS] = default_fs;
    emu->segs[_GS] = 0x33;
    // setup fpu regs
    reset_fpu(emu);
-    // if trace is activated
-    if(context->x86trace) {
-        emu->dec = InitX86TraceDecoder(context);
-        if(!emu->dec)
-            printf_log(LOG_INFO, "Failed to initialize Zydis decoder and formater, no trace activated\n");
-    }
+    emu->mxcsr = 0x1f80;  
 }

+EXPORTDYN
 x86emu_t *NewX86Emu(box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
 {
    printf_log(LOG_DEBUG, "Allocate a new X86 Emu, with EIP=%p and Stack=%p/0x%X\n", (void*)start, (void*)stack, stacksize);
@@ -98,101 +105,88 @@ x86emu_t *NewX86EmuFromStack(x86emu_t* emu, box86context_t *context, uintptr_t s
    return emu;
 }

+EXPORTDYN
 void SetupX86Emu(x86emu_t *emu)
 {
    printf_log(LOG_DEBUG, "Setup X86 Emu\n");
 }

-void SetTraceEmu(x86emu_t *emu, uintptr_t trace_start, uintptr_t trace_end)
+void SetTraceEmu(uintptr_t start, uintptr_t end)
 {
-    if(emu->dec) {
-        if (trace_end == 0) {
+    if(my_context->zydis) {
+        if (end == 0) {
            printf_log(LOG_INFO, "Setting trace\n");
        } else {
-            if(trace_end!=1) {  // 0-1 is basically no trace, so don't printf it...
-                printf_log(LOG_INFO, "Setting trace only between %p and %p\n", (void*)trace_start, (void*)trace_end);
+            if(end!=1) {  // 0-1 is basically no trace, so don't printf it...
+                printf_log(LOG_INFO, "Setting trace only between %p and %p\n", (void*)start, (void*)end);
            }
        }
    }
-    emu->trace_start = trace_start;
-    emu->trace_end = trace_end;
+    trace_start = start;
+    trace_end = end;
 }

 void AddCleanup(x86emu_t *emu, void *p)
 {
-    if(emu->clean_sz == emu->clean_cap) {
-        emu->clean_cap += 4;
-        emu->cleanups = (cleanup_t*)realloc(emu->cleanups, sizeof(cleanup_t)*emu->clean_cap);
+    if(my_context->clean_sz == my_context->clean_cap) {
+        my_context->clean_cap += 4;
+        my_context->cleanups = (cleanup_t*)realloc(my_context->cleanups, sizeof(cleanup_t)*my_context->clean_cap);
    }
-    emu->cleanups[emu->clean_sz].arg = 0;
-    emu->cleanups[emu->clean_sz++].f = p;
+    my_context->cleanups[my_context->clean_sz].arg = 0;
+    my_context->cleanups[my_context->clean_sz].a = NULL;
+    my_context->cleanups[my_context->clean_sz++].f = p;
 }

 void AddCleanup1Arg(x86emu_t *emu, void *p, void* a)
 {
-    if(emu->clean_sz == emu->clean_cap) {
-        emu->clean_cap += 4;
-        emu->cleanups = (cleanup_t*)realloc(emu->cleanups, sizeof(cleanup_t)*emu->clean_cap);
+    if(my_context->clean_sz == my_context->clean_cap) {
+        my_context->clean_cap += 4;
+        my_context->cleanups = (cleanup_t*)realloc(my_context->cleanups, sizeof(cleanup_t)*my_context->clean_cap);
    }
-    emu->cleanups[emu->clean_sz].arg = 1;
-    emu->cleanups[emu->clean_sz].a = a;
-    emu->cleanups[emu->clean_sz++].f = p;
+    my_context->cleanups[my_context->clean_sz].arg = 1;
+    my_context->cleanups[my_context->clean_sz].a = a;
+    my_context->cleanups[my_context->clean_sz++].f = p;
 }

 void CallCleanup(x86emu_t *emu, void* p)
 {
-    for(int i=emu->clean_sz-1; i>=0; --i) {
-        if(p==emu->cleanups[i].f) {
+    printf_log(LOG_DEBUG, "Calling atexit registered functions for %p mask\n", p);
+    for(int i=my_context->clean_sz-1; i>=0; --i) {
+        if(p==my_context->cleanups[i].f) {
            printf_log(LOG_DEBUG, "Call cleanup #%d\n", i);
-            if(emu->cleanups[i].arg)
-                Push(emu, (uintptr_t)emu->cleanups[i].a);
-            PushExit(emu);
-            emu->ip.dword[0] = (uintptr_t)(emu->cleanups[i].f);
-            Run(emu, 0);
-            emu->quit = 0;
+            RunFunctionWithEmu(emu, 0, (uintptr_t)(my_context->cleanups[i].f), my_context->cleanups[i].arg, my_context->cleanups[i].a );
            // now remove the cleanup
-            if(i!=emu->clean_sz-1)
-                memmove(emu->cleanups+i, emu->cleanups+i+1, (emu->clean_sz-i-1)*sizeof(cleanup_t));
-            --emu->clean_sz;
-            return;
+            if(i!=my_context->clean_sz-1)
+                memmove(my_context->cleanups+i, my_context->cleanups+i+1, (my_context->clean_sz-i-1)*sizeof(cleanup_t));
+            --my_context->clean_sz;
        }
    }
 }

 void CallAllCleanup(x86emu_t *emu)
 {
-    /* disabling for now...
-    for(int i=emu->clean_sz-1; i>=0; --i) {
+    printf_log(LOG_DEBUG, "Calling atexit registered functions\n");
+    for(int i=my_context->clean_sz-1; i>=0; --i) {
        printf_log(LOG_DEBUG, "Call cleanup #%d\n", i);
-        if(emu->cleanups[i].arg)
-            Push(emu, (uintptr_t)emu->cleanups[i].a);
-        PushExit(emu);
-        emu->ip.dword[0] = (uintptr_t)(emu->cleanups[i].f);
-        Run(emu);
-        emu->quit = 0;
+        RunFunctionWithEmu(emu, 0, (uintptr_t)(my_context->cleanups[i].f), my_context->cleanups[i].arg, my_context->cleanups[i].a );
    }
-    */
-    emu->clean_sz = 0;
+    my_context->clean_sz = 0;
+    free(my_context->cleanups);
+    my_context->cleanups = NULL;
 }

 static void internalFreeX86(x86emu_t* emu)
 {
-    // stop trace now
-    if(emu->dec)
-        DeleteX86TraceDecoder(&emu->dec);
-    emu->dec = NULL;
-    // call atexit and fini first!
-    CallAllCleanup(emu);
-    free(emu->cleanups);
-
-    free(emu->stack);
+    if(emu->stack2free)
+        munmap(emu->stack2free, emu->size_stack);
 }

+EXPORTDYN
 void FreeX86Emu(x86emu_t **emu)
 {
    if(!emu)
        return;
-    printf_log(LOG_DEBUG, "Free a X86 Emu (%p)\n", *emu);
+    printf_log(LOG_DEBUG, "%04d|Free a X86 Emu (%p)\n", GetTID(), *emu);

    internalFreeX86(*emu);

@@ -204,22 +198,20 @@ void FreeX86EmuFromStack(x86emu_t **emu)
 {
    if(!emu)
        return;
-    printf_log(LOG_DEBUG, "Free a X86 Emu from stack (%p)\n", *emu);
+    printf_log(LOG_DEBUG, "%04d|Free a X86 Emu from stack (%p)\n", GetTID(), *emu);

    internalFreeX86(*emu);
-
-    *emu = NULL;
 }

 void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
 {
 	memcpy(newemu->regs, emu->regs, sizeof(emu->regs));
    memcpy(&newemu->ip, &emu->ip, sizeof(emu->ip));
-	memcpy(&newemu->packed_eflags, &emu->packed_eflags, sizeof(emu->packed_eflags));
-    memcpy(&newemu->flags, &emu->flags, sizeof(emu->flags));
+	memcpy(&newemu->eflags, &emu->eflags, sizeof(emu->eflags));
    newemu->old_ip = emu->old_ip;
    memcpy(newemu->segs, emu->segs, sizeof(emu->segs));
-	memcpy(newemu->fpu, emu->fpu, sizeof(emu->fpu));
+    memset(newemu->segs_serial, 0, sizeof(newemu->segs_serial));
+	memcpy(newemu->mmx87, emu->mmx87, sizeof(emu->mmx87));
    memcpy(newemu->fpu_ld, emu->fpu_ld, sizeof(emu->fpu_ld));
    memcpy(newemu->fpu_ll, emu->fpu_ll, sizeof(emu->fpu_ll));
 	memcpy(newemu->p_regs, emu->p_regs, sizeof(emu->p_regs));
@@ -229,15 +221,13 @@ void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
 	newemu->top = emu->top;
    newemu->fpu_stack = emu->fpu_stack;
 	memcpy(&newemu->round, &emu->round, sizeof(emu->round));
-    memcpy(newemu->mmx, emu->mmx, sizeof(emu->mmx));
    memcpy(newemu->xmm, emu->xmm, sizeof(emu->xmm));
    newemu->mxcsr = emu->mxcsr;
    newemu->quit = emu->quit;
    newemu->error = emu->error;
-    SetTraceEmu(newemu, emu->trace_start, emu->trace_end);
-    // addapt R_ESP to new stack frame
-    uintptr_t oldst = (uintptr_t)((emu->stack)?emu->stack:emu->context->stack);
-    uintptr_t newst = (uintptr_t)((newemu->stack)?newemu->stack:newemu->context->stack);
+    // adapt R_ESP to new stack frame
+    uintptr_t oldst = (uintptr_t)((emu->init_stack)?emu->init_stack:emu->context->stack);
+    uintptr_t newst = (uintptr_t)((newemu->init_stack)?newemu->init_stack:newemu->context->stack);
    newemu->regs[_SP].dword[0] = emu->regs[_SP].dword[0] + (intptr_t)(newst - oldst);
 }

@@ -294,6 +284,14 @@ uint32_t GetESP(x86emu_t *emu)
 {
    return R_ESP;
 }
+void SetFS(x86emu_t *emu, uint16_t v)
+{
+    emu->segs[_FS] = v;
+}
+uint16_t GetFS(x86emu_t *emu)
+{
+    return emu->segs[_FS];
+}


 void ResetFlags(x86emu_t *emu)
@@ -310,7 +308,7 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
    if(trace_emm) {
        // do emm reg is needed
        for(int i=0; i<8; ++i) {
-            sprintf(tmp, "mm%d:%016llx", i, emu->mmx[i].q);
+            sprintf(tmp, "mm%d:%016llx", i, emu->mmx87[i].q);
            strcat(buff, tmp);
            if ((i&3)==3) strcat(buff, "\n"); else strcat(buff, " ");
        }
@@ -326,11 +324,7 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
    // start with FPU regs...
    if(emu->fpu_stack) {
        for (int i=0; i<emu->fpu_stack; i++) {
-            #ifdef USE_FLOAT
-            sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].f);
-            #else
-            sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].d);
-            #endif
+            sprintf(tmp, "ST%d=%f", i, emu->mmx87[(emu->top+i)&7].d);
            strcat(buff, tmp);
            int c = 10-strlen(tmp);
            if(c<1) c=1;
@@ -364,8 +358,19 @@ void StopEmu(x86emu_t* emu, const char* reason)
    emu->quit = 1;
    printf_log(LOG_NONE, "%s", reason);
    // dump stuff...
-    printf_log(LOG_NONE, "CPU Regs=%s\n", DumpCPURegs(emu, R_EIP));
-    // TODO: stack, memory/instruction around EIP, etc..
+    printf_log(LOG_NONE, "\n==== CPU Registers ====\n%s\n", DumpCPURegs(emu, R_EIP));
+    printf_log(LOG_NONE, "\n======== Stack ========\nStack is from 0x%x to 0x%0x\n", R_EBP, R_ESP);
+    if (R_EBP == R_ESP) {
+        printf_log(LOG_NONE, "EBP = ESP: leaf function detected; next 128 bytes should be either data or random.\n");
+    } else {
+        // TODO: display stack if operation should be allowed (to avoid crashes)
+        /* for (uint32_t *sp = R_EBP; sp >= R_ESP; --sp) {
+        } */
+    }
+    printf_log(LOG_NONE, "\n====== Past data ======\nOld IP: %tX\n", emu->old_ip);
+#ifdef HAVE_TRACE
+    printf_log(LOG_NONE, "%s\n", DecodeX86Trace(my_context->dec, emu->old_ip));
+#endif
 }

 void UnimpOpcode(x86emu_t* emu)
@@ -373,8 +378,8 @@ void UnimpOpcode(x86emu_t* emu)
    R_EIP = emu->old_ip;

    int tid = syscall(SYS_gettid);
-    printf_log(LOG_NONE, "%04d|%p: Unimplemented Opcode %02X %02X %02X %02X %02X %02X %02X %02X\n", 
-        tid, (void*)emu->old_ip,
+    printf_log(LOG_NONE, "%04d|%p: Unimplemented Opcode (%02X) %02X %02X %02X %02X %02X %02X %02X %02X\n", 
+        tid, (void*)emu->old_ip, Peek(emu, -1),
        Peek(emu, 0), Peek(emu, 1), Peek(emu, 2), Peek(emu, 3),
        Peek(emu, 4), Peek(emu, 5), Peek(emu, 6), Peek(emu, 7));
    emu->quit=1;
@@ -449,11 +454,11 @@ static inline void init_perfcounters (int32_t do_reset, int32_t enable_divider)
 uint64_t ReadTSC(x86emu_t* emu)
 {
    // Read the TimeStamp Counter as 64bits.
-    // this is supposed to be the number of instrunctions executed since last reset
+    // this is supposed to be the number of instructions executed since last reset
 #if defined(__i386__)
-  uint64_t ret;
-  __asm__ volatile("rdtsc" : "=A"(ret));
-  return ret;
+    uint64_t ret;
+    __asm__ volatile("rdtsc" : "=A"(ret));
+    return ret;
 #if 0
 #elif defined(__ARM_ARCH)
 #if (__ARM_ARCH >= 6)
@@ -469,14 +474,21 @@ uint64_t ReadTSC(x86emu_t* emu)
 #endif
 #endif
 #endif
-// fall back to gettime...
+    // fall back to gettime...
 #ifndef NOGETCLOCK
-  struct timespec ts;
-  clock_gettime(CLOCK_MONOTONIC_COARSE, &ts);
-  return (uint64_t)(ts.tv_sec) * 1000000000LL + ts.tv_nsec;
+    struct timespec ts;
+    clock_gettime(CLOCK_MONOTONIC_COARSE, &ts);
+    return (uint64_t)(ts.tv_sec) * 1000000000LL + ts.tv_nsec;
 #else
-  struct timeval tv;
-  gettimeofday(&tv, NULL);
-  return (uint64_t)(tv.tv_sec) * 1000000 + tv.tv_usec;
+    struct timeval tv;
+    gettimeofday(&tv, NULL);
+    return (uint64_t)(tv.tv_sec) * 1000000 + tv.tv_usec;
 #endif
 }
+
+void ResetSegmentsCache(x86emu_t *emu)
+{
+    if(!emu)
+        return;
+    memset(emu->segs_serial, 0, sizeof(emu->segs_serial));
+}
@@ -3,7 +3,6 @@

 #include "regs.h"

-typedef struct zydis_dec_s zydis_dec_t;
 typedef struct box86context_s box86context_t;
 typedef struct i386_ucontext_s i386_ucontext_t;

@@ -11,7 +10,10 @@ typedef struct i386_ucontext_s i386_ucontext_t;
 #define ERR_DIVBY0  2
 #define ERR_ILLEGAL 4

-typedef struct cleanup_s cleanup_t;
+#ifdef DYNAREC
+#define CSTACK      32
+#define CSTACKMASK  31
+#endif

 typedef struct forkpty_s {
    void*    amaster;
@@ -23,22 +25,20 @@ typedef struct forkpty_s {

 typedef struct x86emu_s {
    // cpu
-	reg32_t     regs[8],ip;
-    int         flags[F_LAST];
-	x86flags_t  packed_eflags;
+	reg32_t     regs[8];
+	x86flags_t  eflags;
+    reg32_t     ip;
    uintptr_t   old_ip;
-    // fpu
-	fpu_reg_t   fpu[9];
+    // fpu / mmx
+	mmx87_regs_t mmx87[8];
 	uint16_t    cw,cw_mask_all;
 	x87flags_t  sw;
 	uint32_t    top;        // top is part of sw, but it's faster to have it separatly
    int         fpu_stack;
 	fpu_round_t round;
-    fpu_ld_t    fpu_ld[9]; // for long double emulation / 80bits fld fst
-    fpu_ll_t    fpu_ll[9]; // for 64bits fild / fist sequence
-	fpu_p_reg_t p_regs[9];
-    // mmx
-    mmx_regs_t  mmx[8];
+    fpu_ld_t    fpu_ld[8]; // for long double emulation / 80bits fld fst
+    fpu_ll_t    fpu_ll[8]; // for 64bits fild / fist sequence
+	fpu_p_reg_t p_regs[8];
    // sse
    sse_regs_t  xmm[8];
    uint32_t    mxcsr;
@@ -52,7 +52,9 @@ typedef struct x86emu_s {
    uintptr_t   prev2_ip, prev_ip;
    #endif
    // segments
-    uint32_t    segs[6];    // only 32bits value?
+    uint32_t    segs[6];        // only 32bits value?
+    uintptr_t   segs_offs[6];   // computed offset associate with segment
+    uint32_t    segs_serial[6];  // are seg offset clean (not 0) or does they need to be re-computed (0)? For GS, serial need to be the same as context->sel_serial
    // emu control
    int         quit;
    int         error;
@@ -61,29 +63,28 @@ typedef struct x86emu_s {
    int         exit;
    int         quitonlongjmp;  // quit if longjmp is called
    int         longjmp;        // if quit because of longjmp
-    // trace
-    zydis_dec_t *dec;
-    uintptr_t   trace_start, trace_end;
    // parent context
    box86context_t *context;
    // cpu helpers
    reg32_t     zero;
    reg32_t     *sbiidx[8];
-    // atexit and fini functions
-    cleanup_t   *cleanups;
-    int         clean_sz;
-    int         clean_cap;
    // scratch stack, used for alignement of double and 64bits ints on arm. 200 elements should be enough
    uint32_t    scratch[200];
    // local stack, do be deleted when emu is freed
-    void*       stack;
+    void*       stack2free; // this is the stack to free (can be NULL)
    void*       init_stack; // initial stack (owned or not)
    uint32_t    size_stack; // stack size (owned or not)

    i386_ucontext_t *uc_link; // to handle setcontext

+    int         type;       // EMUTYPE_xxx define
+
 } x86emu_t;

+#define EMUTYPE_NONE    0
+#define EMUTYPE_MAIN    1
+#define EMUTYPE_SIGNAL  2
+
 //#define INTR_RAISE_DIV0(emu) {emu->error |= ERR_DIVBY0; emu->quit=1;}
 #define INTR_RAISE_DIV0(emu) {emu->error |= ERR_DIVBY0;} // should rise a SIGFPE and not quit