Bumped version to v0.3.4

That debian changelog was outdated a bit
Update README.md
2026-11-08 02:08:33 -10:00 · 2023-12-15 19:25:05 +01:00 · 2023-12-15 17:20:42 +01:00 · 2023-12-15 16:41:02 +01:00 · 2023-12-15 13:24:47 +01:00 · 2023-12-15 13:19:03 +01:00
875 changed files with 32813 additions and 16154 deletions
@@ -23,6 +23,9 @@ on:
        - RPI4
        - RPI4ARM64
        - SD845
+        - SD888
+        - ADLINK
+        - PHYTIUM
      build_type:
        description: 'Build type'
        required: true
@@ -27,7 +27,7 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        platform: [I386, OTHER_ARM, PANDORA, PYRA, GAMESHELL, RPI4, RPI4ARM64, ODROID, RK3399, RK3288, A64, SD845, PHYTIUM, ANDROID]
+        platform: [I386, OTHER_ARM, PANDORA, PYRA, GAMESHELL, RPI4, RPI4ARM64, ODROID, RK3399, RK3288, A64, SD845, SD888, ADLINK, PHYTIUM, ANDROID]
        type: [Release, Trace]
        os: [ubuntu-latest]
        
@@ -60,7 +60,7 @@ jobs:
              wget https://dl.google.com/android/repository/android-ndk-r23b-linux.zip
              unzip android-ndk-r23b-linux.zip
              echo "BOX86_COMPILER=$PWD/android-ndk-r23b/toolchains/llvm/prebuilt/linux-x86_64/bin/armv7a-linux-androideabi31-clang" >> $GITHUB_ENV
-              echo "BOX86_PLATFORM_MARCRO=-DANDROID=1 -DNOLOADADDR=1" >> $GITHUB_ENV
+              echo "BOX86_PLATFORM_MARCRO=-DANDROID=1 -DARM_DYNAREC=ON -DNOLOADADDR=1" >> $GITHUB_ENV
            else
              sudo apt-get -y install gcc-arm-linux-gnueabihf
              echo "BOX86_COMPILER=arm-linux-gnueabihf-gcc" >> $GITHUB_ENV
@@ -19,17 +19,23 @@ 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(SD845 "Set to ON if targeting a Snapragon 845 based device" ${SD845})
+option(SD888 "Set to ON if targeting a Snapragon 888 based device" ${SD888})
+option(ADLINK "Set to ON if targeting an ADLink AmpereAltra based device" ${ADLINK})
 option(A64 "Set to ON if targeting an Allwinner A64 based device" ${A64})
 option(ARM64 "Set to ON if targeting a generic ARM64 based device" ${ARM64})
 option(BAD_SIGNAL "Set to ON to activate the workaround for incoherent si_info on SIGSEGV" ${BAD_SIGNAL})
 option(LX2160A "Set to ON if targeting an LX2160A based device" ${LX2160A})
 option(ANDROID "Set to ON if targeting an Android device" ${ANDROID})
+option(TERMUX "Set to ON if targeting an Android device with Termux" ${TERMUX})
 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 GOA_CLONE OR RK3288 OR RK3399 OR ODROIDN2 OR TEGRAX1 OR TEGRAX1ARM32 OR RPI3ARM64 OR RPI4ARM64 OR PHYTIUM OR SD845 OR A64 OR LX2160A OR ARM64 OR RK3588)
+option(NO_LIB_INSTALL "Set ON to not install a few x86 libs that are used by many program" ${NO_LIB_INSTALL})
+option(NO_CONF_INSTALL "Set ON to not install config files" ${NO_CONF_INSTALL})
+
+if(PANDORA OR PYRA OR RPI2 OR RPI3 OR RPI4 OR GAMESHELL OR ODROID OR GOA_CLONE OR RK3288 OR RK3399 OR ODROIDN2 OR TEGRAX1 OR TEGRAX1ARM32 OR RPI3ARM64 OR RPI4ARM64 OR PHYTIUM OR SD845 OR SD888 OR ADLINK OR A64 OR LX2160A OR ARM64 OR RK3588)
    set(LD80BITS OFF CACHE BOOL "")
    set(NOALIGN OFF CACHE BOOL "")
    set(ARM_DYNAREC ON CACHE BOOL "")
@@ -37,8 +43,13 @@ if(PANDORA OR PYRA OR RPI2 OR RPI3 OR RPI4 OR GAMESHELL OR ODROID OR GOA_CLONE O
        set(BAD_SIGNAL ON CACHE BOOL "")
    endif()
 endif()
+if(TERMUX)
+    set(ANDROID ON CACHE BOOL "")
+    set(TERMUX_PATH "/data/data/com.termux/files")
+endif()
 if(ANDROID)
    set(NOLOADADDR ON CACHE BOOL "")
+    set(BAD_SIGNAL 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)" ${NOALIGN})
@@ -67,8 +78,10 @@ string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "i686"  _x86)
 string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "x86_64"  _x86_64)
 string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "aarch64"  _aarch64)
 string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "armhf"  _armhf)
+string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "arm"  _arm)
+string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "armv8l"  _armv8l)

-if(_aarch64 AND (RK3399 OR ODROIDN2 OR TEGRAX1 OR RPI3ARM64 OR RPI4ARM64 OR PHYTIUM OR SD845 OR A64 OR LX2160A OR RK3588 OR ARM64))
+if(_aarch64 AND (RK3399 OR ODROIDN2 OR TEGRAX1 OR RPI3ARM64 OR RPI4ARM64 OR PHYTIUM OR SD845 OR SD888 OR ADLINK OR A64 OR LX2160A OR RK3588 OR ARM64))
    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)
@@ -134,6 +147,14 @@ elseif(SD845)
    add_definitions(-DSD845)
    add_definitions(-marm -march=armv8.2-a+simd+crypto -mtune=cortex-a75.cortex-a55 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-marm -march=armv8.2-a+simd+crypto -mtune=cortex-a75.cortex-a55 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
+elseif(SD888)
+    add_definitions(-DSD888)
+    add_definitions(-marm -march=armv8.4-a+simd+crypto -mfpu=neon-fp-armv8 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -march=armv8.4-a+simd+crypto -mfpu=neon-fp-armv8 -mfloat-abi=hard")
+elseif(ADLINK)
+    add_definitions(-DADLINK)
+    add_definitions(-marm -mcpu=neoverse-n1 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -mcpu=neoverse-n1 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
 elseif(A64)
    add_definitions(-DA64)
    add_definitions(-marm -march=armv8-a+crc+simd+crypto -mcpu=cortex-a53+crypto -mfpu=neon-fp-armv8 -mfloat-abi=hard)
@@ -175,13 +196,18 @@ elseif(LX2160A)
    add_definitions(-DLX2160A)
    add_definitions(-pipe -march=armv8-a+crypto+crc -mcpu=cortex-a72+crypto -mfpu=neon-fp-armv8 -mfloat-abi=hard)
    set(CMAKE_ASM_FLAGS  "-pipe -march=armv8-a+crypto+crc -mcpu=cortex-a72+crypto -mfpu=neon-fp-armv8 -mfloat-abi=hard")
-elseif(ANDROID)
-    add_definitions(-DANDROID)
 elseif(ARM_DYNAREC)
    #if DYNAREC is selected alone, without any arch
-    set(CMAKE_ASM_FLAGS  "-pipe -march=armv7-a+simd -mfpu=neon")
+    if(NOT ANDROID)
+        set(CMAKE_ASM_FLAGS  "-pipe -march=armv7-a+simd -mfpu=neon")
+    endif()
+endif()
+if(ANDROID)
+    add_definitions(-DANDROID)
+endif()
+if(TERMUX)
+    add_definitions(-DTERMUX)
 endif()
-
 if(BAD_SIGNAL)
    add_definitions(-DBAD_SIGNAL)
 endif()
@@ -296,10 +322,8 @@ set(ELFLOADER_SRC
    "${BOX86_ROOT}/src/librarian/dictionnary.c"
    "${BOX86_ROOT}/src/librarian/symbols.c"
    "${BOX86_ROOT}/src/librarian/globalsymbols.c"
+    "${BOX86_ROOT}/src/emu/x86compstrings.c"
    "${BOX86_ROOT}/src/emu/x86emu.c"
-    "${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"
    "${BOX86_ROOT}/src/emu/x86primop.c"
@@ -307,7 +331,6 @@ set(ELFLOADER_SRC
    "${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"
@@ -324,10 +347,39 @@ if(NOT ANDROID)
    list(APPEND ELFLOADER_SRC "${BOX86_ROOT}/src/libtools/obstack.c")
 endif()

+set(INTERPRETER
+    "${BOX86_ROOT}/src/emu/x86run.c"
+    "${BOX86_ROOT}/src/emu/x86run0f.c"
+    "${BOX86_ROOT}/src/emu/x86run64.c"
+    "${BOX86_ROOT}/src/emu/x86run640f.c"
+    "${BOX86_ROOT}/src/emu/x86run6466.c"
+    "${BOX86_ROOT}/src/emu/x86run6467.c"
+    "${BOX86_ROOT}/src/emu/x86run66.c"
+    "${BOX86_ROOT}/src/emu/x86run660f.c"
+    "${BOX86_ROOT}/src/emu/x86run66d9.c"
+    "${BOX86_ROOT}/src/emu/x86run66dd.c"
+    "${BOX86_ROOT}/src/emu/x86run66f20f.c"
+    "${BOX86_ROOT}/src/emu/x86run67.c"
+    "${BOX86_ROOT}/src/emu/x86run6766.c"
+    "${BOX86_ROOT}/src/emu/x86rund8.c"
+    "${BOX86_ROOT}/src/emu/x86rund9.c"
+    "${BOX86_ROOT}/src/emu/x86runda.c"
+    "${BOX86_ROOT}/src/emu/x86rundb.c"
+    "${BOX86_ROOT}/src/emu/x86rundc.c"
+    "${BOX86_ROOT}/src/emu/x86rundd.c"
+    "${BOX86_ROOT}/src/emu/x86runde.c"
+    "${BOX86_ROOT}/src/emu/x86rundf.c"
+    "${BOX86_ROOT}/src/emu/x86runf0.c"
+    "${BOX86_ROOT}/src/emu/x86runf066.c"
+    "${BOX86_ROOT}/src/emu/x86runf20f.c"
+    "${BOX86_ROOT}/src/emu/x86runf30f.c"
+)
+
 set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibc.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibpthread.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibrt.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibbsd.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibgl.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibglu.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibx11.c"
@@ -387,6 +439,7 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedopenal.c"
    "${BOX86_ROOT}/src/wrapped/wrappedalure.c"
    "${BOX86_ROOT}/src/wrapped/wrappedalut.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedopencl.c"
    "${BOX86_ROOT}/src/wrapped/wrappedcurl.c"
    "${BOX86_ROOT}/src/wrapped/wrappedudev0.c"
    "${BOX86_ROOT}/src/wrapped/wrappedudev1.c"
@@ -441,6 +494,7 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappeddbusmenuglib.c"
    "${BOX86_ROOT}/src/wrapped/wrappedappindicator.c"
    "${BOX86_ROOT}/src/wrapped/wrappedappindicator3.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedayatanaappindicator3.c"
    "${BOX86_ROOT}/src/wrapped/wrappedatk.c"
    "${BOX86_ROOT}/src/wrapped/wrappedatkbridge.c"
    "${BOX86_ROOT}/src/wrapped/wrappedpangoft2.c"
@@ -492,6 +546,9 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedpcap.c"
    "${BOX86_ROOT}/src/wrapped/wrappedfaudio.c"
    "${BOX86_ROOT}/src/wrapped/wrappedtbbmallocproxy.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibfuse.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedatomic.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibpcre.c"
 )
 if(ANDROID)
    list(APPEND WRAPPEDS
@@ -521,7 +578,7 @@ 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" "--" 
+    "PANDORA" "HAVE_LD80BITS" "NOALIGN" "HAVE_TRACE" "POWERPCLE" "ANDROID" "TERMUX" "--" 
    ${WRAPPEDS_HEAD}
    MAIN_DEPENDENCY "${BOX86_ROOT}/rebuild_wrappers.py"
    DEPENDS ${WRAPPEDS} ${WRAPPEDS_HEAD}
@@ -545,7 +602,9 @@ if(ARM_DYNAREC)

        "${BOX86_ROOT}/src/dynarec/dynarec_arm.c"
        "${BOX86_ROOT}/src/dynarec/dynarec_arm_functions.c"
+        "${BOX86_ROOT}/src/dynarec/dynarec_arm_jmpnext.c"
        "${BOX86_ROOT}/src/dynarec/arm_printer.c"
+        "${BOX86_ROOT}/src/emu/x86test.c"

        "${BOX86_ROOT}/src/dynarec/arm_prolog.S"
        "${BOX86_ROOT}/src/dynarec/arm_epilog.S"
@@ -593,6 +652,8 @@ 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_library(test_interpreter OBJECT ${INTERPRETER})
+    set_target_properties(test_interpreter PROPERTIES COMPILE_FLAGS "-DTEST_INTERPRETER")
    add_dependencies(arm_pass0 WRAPPERS)
    add_dependencies(arm_pass1 WRAPPERS)
    add_dependencies(arm_pass2 WRAPPERS)
@@ -604,6 +665,7 @@ if(ARM_DYNAREC)
        $<TARGET_OBJECTS:arm_pass1>
        $<TARGET_OBJECTS:arm_pass2>
        $<TARGET_OBJECTS:arm_pass3>
+        $<TARGET_OBJECTS:test_interpreter>
    )

 endif()
@@ -613,16 +675,18 @@ 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}
+        DEPENDS dynarec ${ELFLOADER_SRC} ${INTERPRETER} ${WRAPPEDS}
        VERBATIM)
 else()
    add_custom_command(
        OUTPUT "${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}
+        DEPENDS ${ELFLOADER_SRC} ${INTERPRETER} ${WRAPPEDS}
        VERBATIM)
 endif()

+add_library(interpreter OBJECT ${INTERPRETER})
+
 if(BOX86LIB)
 #build a static or dynamic lib...
 if(BOX86DYNAMIC)
@@ -632,9 +696,13 @@ else()
 endif()
 if(ARM_DYNAREC)
    if(ANDROID)
-        target_link_libraries(${BOX86} dynarec m dl)
+        if(TERMUX)
+        target_link_libraries(${BOX86} dynarec interpreter m dl android-sysv-semaphore android-spawn)
+        else()
+        target_link_libraries(${BOX86} dynarec interpreter m dl)
+        endif()
    else()
-        target_link_libraries(${BOX86} dynarec m dl rt pthread resolv)
+        target_link_libraries(${BOX86} dynarec interpreter m dl rt pthread resolv)
    endif()
 endif()
 if(${CMAKE_VERSION} VERSION_LESS "3.13")
@@ -654,38 +722,77 @@ add_executable(${BOX86} ${ELFLOADER_SRC} ${WRAPPEDS} "${BOX86_ROOT}/src/git_head
 add_dependencies(${BOX86} WRAPPERS)
 add_dependencies(${BOX86} PRINTER)
 if(ANDROID)
+    if(TERMUX)
+    target_link_libraries(${BOX86} dynarec interpreter m dl android-complex-math android-sysv-semaphore android-spawn )
+    else()
+    target_link_libraries(${BOX86} dynarec interpreter m dl)
+    endif()
    target_link_libraries(${BOX86} c m dl)
 else()
    #target_link_libraries(${BOX86} c m dl rt pthread resolv)
    set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-Wl,--no-as-needed -lc -lm -ldl -lrt -lpthread -lresolv -Wl,--as-needed -pthread")
 endif()
 if(ARM_DYNAREC)
-    target_link_libraries(${BOX86} dynarec)
+    target_link_libraries(${BOX86} dynarec interpreter)
+else()
+    target_link_libraries(${BOX86} interpreter)
 endif()

 if(${CMAKE_VERSION} VERSION_LESS "3.13")
    if(NOT NOLOADADDR)
-        set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-rdynamic -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS}")
+        if(CMAKE_C_COMPILER_ID STREQUAL "Clang")
+            set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-rdynamic -Wl,--image-base,${BOX86_ELF_ADDRESS}")
+        else()
+            set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-rdynamic -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS}")
+        endif()
    else()
        set_target_properties(${BOX86} PROPERTIES LINK_FLAGS -rdynamic)
    endif()
 else()
    target_link_options(${BOX86} PUBLIC -rdynamic)
    if(NOT NOLOADADDR)
-        target_link_options(${BOX86} PUBLIC -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS})
+        if(CMAKE_C_COMPILER_ID STREQUAL "Clang")
+            target_link_options(${BOX86} PUBLIC -Wl,--image-base,${BOX86_ELF_ADDRESS})
+        else()
+            target_link_options(${BOX86} PUBLIC -Wl,-Ttext-segment,${BOX86_ELF_ADDRESS})
+        endif()
    endif()
 endif()

 if(NOT _x86 AND NOT _x86_64)
-  install(TARGETS ${BOX86}
+  if(NOT TERMUX)
+   install(TARGETS ${BOX86}
    RUNTIME DESTINATION bin)
-  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/)
-  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libpng12.so.0 DESTINATION /usr/lib/i386-linux-gnu/)
-  install(FILES ${CMAKE_SOURCE_DIR}/system/box86.box86rc DESTINATION /etc/)
+  else()
+   install(TARGETS ${BOX86}
+    RUNTIME DESTINATION ${TERMUX_PATH}/usr/bin)
+  endif()
+  if(NOT NO_CONF_INSTALL)
+    configure_file(system/box86.conf.cmake system/box86.conf)
+    	if(NOT TERMUX)
+    		install(FILES ${CMAKE_BINARY_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
+		install(FILES ${CMAKE_SOURCE_DIR}/system/box86.box86rc DESTINATION /etc/)
+	else()
+		#install(FILES ${CMAKE_BINARY_DIR}/system/box86.conf DESTINATION ${TERMUX_PATH}/usr/etc/binfmt.d/)
+		install(FILES ${CMAKE_SOURCE_DIR}/system/box86.box86rc DESTINATION ${TERMUX_PATH}/usr/etc/)
+	endif()
+  endif()
+  if(NOT NO_LIB_INSTALL)
+    if(NOT TERMUX)
+    	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/)
+    	install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libpng12.so.0 DESTINATION /usr/lib/i386-linux-gnu/)
+    	install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libunwind.so.8 DESTINATION /usr/lib/i386-linux-gnu/)
+    else()
+
+	    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.6 DESTINATION ${TERMUX_PATH}/usr/lib/i386-linux-gnu/)
+	    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.5 DESTINATION ${TERMUX_PATH}/usr/lib/i386-linux-gnu/)
+	    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libgcc_s.so.1 DESTINATION ${TERMUX_PATH}/usr/lib/i386-linux-gnu/)
+	    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libpng12.so.0 DESTINATION ${TERMUX_PATH}/usr/lib/i386-linux-gnu/)
+	    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libunwind.so.8 DESTINATION ${TERMUX_PATH}/usr/lib/i386-linux-gnu/)
+    endif()
+  endif()
 endif()

 if(NOT TARGET uninstall)
@@ -702,7 +809,11 @@ set(CPACK_GENERATOR "DEB")
 set(CPACK_DEBIAN_PACKAGE_MAINTAINER "ptitSeb")
 set(CPACK_PACKAGE_CONTACT "[email protected]")
 set(CPACK_PACKAGE_DESCRIPTION "Box86 - Linux Userspace x86 Emulator with a twist")
+if(TERMUX)
+set(CPACK_DEBIAN_PACKAGE_DEPENDS "libandroid-sysv-semaphore, libandroid-spawn")
+else()
 set(CPACK_DEBIAN_PACKAGE_DEPENDS "libc6")
+endif()
 set(CPACK_PACKAGE_HOMEPAGE_URL, "https://box86.org")
 file(STRINGS "${BOX86_ROOT}/src/box86version.h" TEMP_STRING REGEX "BOX86_MAJOR\\s*")
 string(REGEX REPLACE "BOX86_MAJOR" "" TEMP_STRING ${TEMP_STRING})
@@ -716,16 +827,29 @@ string(REGEX MATCH "[0-9]" BOX86_REVISION ${TEMP_STRING})
 set(CPACK_PACKAGE_VERSION_MAJOR ${BOX86_MAJOR})
 set(CPACK_PACKAGE_VERSION_MINOR ${BOX86_MINOR})
 set(CPACK_PACKAGE_VERSION_PATCH ${BOX86_REVISION})
+if(NOT TERMUX)
 set(CPACK_DEBIAN_PACKAGE_CONTROL_EXTRA "${CMAKE_CURRENT_SOURCE_DIR}/postinst")
-if(_aarch64 OR _armhf)
+endif()
+if(_aarch64 OR _armhf OR _arm OR _armv8l)
+    if(TERMUX)
+    set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE "arm")
+    else()
    set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE "armhf")
+    endif()
 elseif(_x86_64 OR _x86)
    set(CPACK_DEBIAN_PACKAGE_ARCHITECTURE "i386")
 else()
    #probably wrong...
    execute_process(COMMAND "dpkg --print-architecture" OUTPUT_VARIABLE CPACK_DEBIAN_PACKAGE_ARCHITECTURE)
 endif()
-set(CPACK_DEBIAN_FILE_NAME "${BOX86}-${BOX86_MAJOR}.${BOX86_MINOR}.${BOX86_REVISION}_Linux-${CPACK_DEBIAN_PACKAGE_ARCHITECTURE}.deb")
+if(TERMUX)
+set(CPACK_PACKAGING_INSTALL_PREFIX "${TERMUX_PATH}/usr")
+endif()
+if(NOT TERMUX)
+set(CPACK_DEBIAN_FILE_NAME "${BOX86}-${BOX86_MAJOR}.${BOX86_MINOR}.${BOX86_REVISION}_${CMAKE_SYSTEM_NAME}-${CPACK_DEBIAN_PACKAGE_ARCHITECTURE}.deb")
+else()
+set(CPACK_DEBIAN_FILE_NAME "${BOX86}-${BOX86_MAJOR}.${BOX86_MINOR}.${BOX86_REVISION}_Termux-${CPACK_DEBIAN_PACKAGE_ARCHITECTURE}.deb")
+endif()
 INCLUDE(CPack)

 add_test(NAME bootSyscall COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
@@ -832,20 +956,45 @@ add_test(NAME idiv COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref20.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")
+add_test(NAME multiple_dlopen COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test21 -D TEST_OUTPUT=tmpfile21.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref21.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(NAME aes COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test22 -D TEST_OUTPUT=tmpfile22.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref22.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")    
 foreach(file ${extension_tests})
    get_filename_component(testname "${file}" NAME_WE)
    add_test(NAME "${testname}" COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/extensions/${testname} -D TEST_OUTPUT=tmpfile-${testname}.txt
        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/extensions/${testname}.txt
        -P ${CMAKE_SOURCE_DIR}/runTest.cmake)
-endforeach()
+endforeach()        

 add_test(NAME sse_optimized COMMAND ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test17_o2 -D TEST_OUTPUT=tmpfile17_o2.txt 
    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref17_o2.txt
    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

-set_tests_properties(sse_optimized PROPERTIES ENVIRONMENT "BOX86_DYNAREC_FASTNAN=0;BOX86_DYNAREC_FASTROUND=0")
+set_tests_properties(sse_optimized PROPERTIES ENVIRONMENT "BOX86_DYNAREC_FASTNAN=0;BOX86_DYNAREC_FASTROUND=0")    
+    
+add_test(bswap ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
+        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test23 -D TEST_OUTPUT=tmpfile23.txt
+        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref23.txt
+        -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+    
+#add_test(feround ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
+#        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test24 -D TEST_OUTPUT=tmpfile24.txt
+#        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref24.txt
+#        -P ${CMAKE_SOURCE_DIR}/runTest.cmake )

+add_test(sse4_2 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86}
+        -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test25 -D TEST_OUTPUT=tmpfile25.txt
+        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref25.txt
+        -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+    
 endif(BOX86LIB)
@@ -1,3 +1,38 @@
+box86 (0.3.4) unstable; urgency=low
+
+ * X86Libs: Updated libstdc++ & libgcc_s
+ * X86Libs: Use x86 vrsion of libunwind (it's needed by wine)
+ * Dynarec: Fixed some flags propagation issues inside blocks
+ * Dynarec: Improved Strong Memory Model emulation, introducing a 3rd level
+ * CPU: Remoworked how INT, priviliged opcodes and teh Trace flags works
+ * CPU: Added full support for SSE4.2
+ * ElfLoader: rework how elf memory is handled
+ * Emulation: Better handling of Signal (also handling SIGMAX correctly)
+ * Emulation: Better support for POKEUSER/PEEKUSER (helps Windows Unity games)
+ * Wrapping: More libs and some fixes in function wrapping (gtk3, vulkan...)
+ * Wrapping: Removed faked libunwind wrapping
+ * Hardware: Added more profile, with support for ADLink Ampere Altra
+ * OS: Added profile to build for Android and Termux.
+ * => Mostly a stability update with improved elfloader again, and added SSE 4.2 support (still need to work on the Dynarec part of this)
+ * => Some work also to reduce the memory footprint
+
+-- Sebastien Chevalier <[email protected]>  Thu, 15 Dec 2023 17:30:33 -0100
+
+box86 (0.3.2) unstable; urgency=low
+
+ * Better syscall handling
+ * Fixed some wrapped functions, added some others
+ * Introduced another wrapper helper
+ * Added some more opcodes
+ * Added a cosimulation mode (to compare the behaviour of the dynarec against the interpreter)
+ * Added support of FASTNAN for some more opcodes
+ * Reduced memory footprint
+ * Added BOX86_NOSANDBOX (adds an argument to the command line)
+ * Added an ukrainian translation
+ * => Mostly a stability update with improved elfloader, fixed opcodes and refined profiles
+
+-- Sebastien Chevalier <[email protected]>  Thu, 10 Mar 2023 10:20:33 -0500
+
 box86 (0.3.0) unstable; urgency=low

 * Added rcfile handling: box86 now read "/etc/box86.box86rc" and "~/.box86rc" and apply per process parameters
@@ -6,7 +6,8 @@
 override_dh_auto_configure:
 	dh_auto_configure -- \
 	      -DCMAKE_BUILD_TYPE=RelWithDebInfo \
-	      -DARM_DYNAREC=1
+	      -DARM_DYNAREC=1 \
+		  -DNOGIT=1

 override_dh_shlibdeps:
 	dh_shlibdeps --exclude=libgcc_s.so.1 --exclude=libpng12.so.0 --exclude=libstdc++.so.5 --exclude=libstdc++.so.6 
@@ -1,3 +1,34 @@
+v0.3.4
+======
+* X86Libs: Updated libstdc++ & libgcc_s
+* X86Libs: Use x86 vrsion of libunwind (it's needed by wine)
+* Dynarec: Fixed some flags propagation issues inside blocks
+* Dynarec: Improved Strong Memory Model emulation, introducing a 3rd level
+* CPU: Remoworked how INT, priviliged opcodes and teh Trace flags works
+* CPU: Added full support for SSE4.2
+* ElfLoader: rework how elf memory is handled
+* Emulation: Better handling of Signal (also handling SIGMAX correctly)
+* Emulation: Better support for POKEUSER/PEEKUSER (helps Windows Unity games)
+* Wrapping: More libs and some fixes in function wrapping (gtk3, vulkan...)
+* Wrapping: Removed faked libunwind wrapping
+* Hardware: Added more profile, with support for ADLink Ampere Altra
+* OS: Added profile to build for Android and Termux.
+=> Mostly a stability update with improved elfloader again, and added SSE 4.2 support (still need to work on the Dynarec part of this)
+=> Some work also to reduce the memory footprint
+
+v0.3.2
+======
+* Better syscall handling
+* Fixed some wrapped functions, added some others
+* Introduced another wrapper helper
+* Added some more opcodes
+* Added a cosimulation mode (to compare the behaviour of the dynarec against the interpreter)
+* Added support of FASTNAN for some more opcodes
+* Reduced memory footprint
+* Added BOX86_NOSANDBOX (adds an argument to the command line)
+* Added an ukrainian translation
+=> Mostly a stability update with improved elfloader, fixed opcodes and refined profiles
+
 v0.3.0
 ======
 * Added rcfile handling: box86 now read `/etc/box86.box86rc` and `~/.box86rc` and apply per process parameters
@@ -6,7 +6,7 @@ You can use [@Itai-Nelken](https://github.com/Itai-Nelken)'s apt repository to i

 ```
 sudo wget https://itai-nelken.github.io/weekly-box86-debs/debian/box86.list -O /etc/apt/sources.list.d/box86.list
-wget -qO- https://itai-nelken.github.io/weekly-box86-debs/debian/KEY.gpg | sudo apt-key add -
+wget -qO- https://itai-nelken.github.io/weekly-box86-debs/debian/KEY.gpg | sudo gpg --dearmor -o /etc/apt/trusted.gpg.d/box86-debs-archive-keyring.gpg
 sudo apt update && sudo apt install box86 -y
 ```

@@ -100,11 +100,13 @@ As most Allwinner A64 devices run an AARCH64 OS, you'll need an `armhf` multiarc

 Also, on armbian, you may need to install `libc6-dev-armhf-cross` or you may have an issue with `crt1.o` and a few other files not included with box86.

-#### for Snapdragon 845
+#### for Snapdragon

 `mkdir build; cd build; cmake .. -DSD845=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j4`
+or
+`mkdir build; cd build; cmake .. -DSD888=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j4`

-As most Snapdragon 845 devices run an AARCH64 OS, you'll need an `armhf` multiarch environment, and an armhf gcc: On mobian, install it with `sudo apt install gcc-arm-linux-gnueabihf`.
+Depending how old/recent you SD is. As most Snapdragon devices run an AARCH64 OS, you'll need an `armhf` multiarch environment, and an armhf gcc: On mobian, install it with `sudo apt install gcc-arm-linux-gnueabihf`.

 #### for Phytium

@@ -115,9 +117,15 @@ As most Phytium (D2000 or FT2000/4) devices run an AARCH64 OS, you'll need an `a

 Also, on armbian, you may need to install `libc6-dev-armhf-cross` or you may have an issue with `crt1.o` and a few other files not included with box86.

+#### for ADLink AmpereAltra
+
+`mkdir build; cd build; cmake .. -DADLINK=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j`
+
+As ADLINK AmpereAltra computers run an 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 ARM64 64bits Linux platform

-`mkdir build; cd build; cmake .. -DARM64 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j3`
+`mkdir build; cd build; cmake .. -DARM64=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo; make -j3`

 You can also add `-DBAD_SIGNAL=ON` to the cmake command if you are on Linux Kernel mixed with Android, like on RK3588 or maybe Termux

@@ -194,5 +202,5 @@ NVIDIA doesn't provide armhf libraries for their GPU drivers at this time. There

 Debian Packaging
 ----
-Box86 can also be packaged into a .deb file with `DEB_BUILD_OPTIONS=nostrip dpkg-buildpackage -us -uc -nc`.
+Box86 can also be packaged into a .deb file ***using the source code zip from the releases page*** with `DEB_BUILD_OPTIONS=nostrip dpkg-buildpackage -us -uc -nc`. Configure any additional cmake options you might want in `debian/rules`.

@@ -2,7 +2,7 @@

 Linux Userspace x86 Emulator with a twist

-[View changelog](https://github.com/ptitSeb/box86/blob/master/docs/CHANGELOG.md) | [中文](https://github.com/ptitSeb/box86/blob/master/docs/README_CN.md) | [Report an error](https://github.com/ptitSeb/box86/issues/new)
+[View changelog](https://github.com/ptitSeb/box86/blob/master/docs/CHANGELOG.md) | [中文](https://github.com/ptitSeb/box86/blob/master/docs/README_CN.md) | [Українська](https://github.com/ptitSeb/box86/blob/master/docs/README_UK.md) | [Report an error](https://github.com/ptitSeb/box86/issues/new)

 ![build](https://app.travis-ci.com/ptitSeb/box86.svg?branch=master) ![stars](https://img.shields.io/github/stars/ptitSeb/box86) ![forks](https://img.shields.io/github/forks/ptitSeb/box86) ![contributors](https://img.shields.io/github/contributors/ptitSeb/box86) ![prs](https://img.shields.io/github/issues-pr/ptitSeb/box86) ![issues](https://img.shields.io/github/issues/ptitSeb/box86)

@@ -28,8 +28,8 @@ Some x86 internal opcodes use parts of "Realmode X86 Emulator Library", see [x86
 ----
 Compiling/Installation
 ----
-> Compilation instructions can be found [here](COMPILE.md)  
-> Instructions for installing Wine for Box86 can be found [here](X86WINE.md)  
+> Compilation instructions can be found [here](https://github.com/ptitSeb/box86/blob/master/docs/COMPILE.md)  
+> Instructions for installing Wine for Box86 can be found [here](https://github.com/ptitSeb/box86/blob/master/docs/X86WINE.md)  

 ----

@@ -80,15 +80,15 @@ Notes about Box86 configuration
 ----

 Box86 now have configurations files. There are 2 files loaded. `/etc/box4.box86rc` and `~/.box86rc`. Both files have the same syntax, and is basicaly an ini files. Section in square brakets define the process name, and the rest is the env. var. to set. Looke at [Usage](USAGE.md) for detail on what parameters can be put. Box86 comes with a default file that should be installed for better stability. The file in in `system/box86.box86rc` and should be installed to `/etc/box86.box86rc` If, for some reasons, you don't want to install that file here, at least copy it to `~/.box86rc` or some game may not function correctly.
-Note that the priority is: `~/.bashrc` > `/etc/box86.box86rc` > command line
-So, your settings in `~/.bashrc` may override the setting from your command line...
+Note that the priority is: `~/.box86rc` > `/etc/box86.box86rc` > command line
+So, your settings in `~/.box86rc` may override the setting from your command line...

 ----

 Notes about Unity game emulation
 ----

-Running Unity games shoudl generaly work now, but you should also note that many Unity3D games require OpenGL 3+ which can be tricky to provide on ARM SBC (single-board computers) for now.
+Running Unity games should generaly work now, but you should also note that many Unity3D games require OpenGL 3+ which can be tricky to provide on ARM SBC (single-board computers) for now.
 Hint: on Pi4, use `MESA_GL_VERSION_OVERRIDE=3.2` and with Panfrost use `PAN_MESA_DEBUG=gl3` to use higher profile if the game starts then quits before showing anything.

 ----
@@ -134,8 +134,8 @@ I want to thank everyone who has contributed to box86 development.
 There are many ways to contribute: code, financial, hardware and advertisement!
 So, in no particular order, I want to thank:
 * For their major code contribution: rajdakin, icecream95, M-HT
- * For their major financial contribution: FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/)
- * For their hardware contribution: [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [DragonBox](https://pyra-handheld.com/), [Novaspirit](https://www.youtube.com/channel/UCrjKdwxaQMSV_NDywgKXVmw), [HardKernel](https://www.hardkernel.com/), [TwisterOS team](https://twisteros.com/)
+ * For their major financial contribution: FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/), [CubeCoders Limited](http://cubecoders.com/)
+ * For their hardware contribution: [ADLink](https://www.adlinktech.com/Products/Computer_on_Modules/COM-HPC-Server-Carrier-and-Starter-Kit/Ampere_Altra_Developer_Platform?lang=en), [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [DragonBox](https://pyra-handheld.com/), [Novaspirit](https://www.youtube.com/channel/UCrjKdwxaQMSV_NDywgKXVmw), [HardKernel](https://www.hardkernel.com/), [TwisterOS team](https://twisteros.com/), [AYN](https://www.ayntec.com/)
 * For their continuous advertisement of box86 project: salva ([microLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)), [PILab](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)/[TwisterOS team](https://twisteros.com/), [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA), [NicoD](https://www.youtube.com/channel/UCpv7NFr0-9AB5xoklh3Snhg)

 And I also thank the many other people who participated even once in the project.
@@ -2,7 +2,7 @@

 Linux用户空间的x86应用模拟器 

-[更新历史](https://github.com/ptitSeb/box86/blob/master/docs/CHANGELOG.md) | [English](https://github.com/ptitSeb/box86/blob/master/docs/README.md) | [报告错误](https://github.com/ptitSeb/box86/issues/new)
+[更新历史](https://github.com/ptitSeb/box86/blob/master/docs/CHANGELOG.md) | [English](https://github.com/ptitSeb/box86/blob/master/docs/README.md) | [Українська](https://github.com/ptitSeb/box86/blob/master/docs/README_UK.md) | [报告错误](https://github.com/ptitSeb/box86/issues/new)

 ![build](https://app.travis-ci.com/ptitSeb/box86.svg?branch=master) ![stars](https://img.shields.io/github/stars/ptitSeb/box86) ![forks](https://img.shields.io/github/forks/ptitSeb/box86) ![contributors](https://img.shields.io/github/contributors/ptitSeb/box86) ![prs](https://img.shields.io/github/issues-pr/ptitSeb/box86) ![issues](https://img.shields.io/github/issues/ptitSeb/box86)

@@ -127,7 +127,7 @@ Box86已经封装了Vulkan。如果设备的系统有32位的Vulkan驱动程序
 以下感谢不分先后： 
 * 代码贡献：rajdakin, icecream95, M-HT
 * 财务帮助：FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/)
- * 硬件捐赠：[Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [DragonBox](https://pyra-handheld.com/), [Novaspirit](https://www.youtube.com/channel/UCrjKdwxaQMSV_NDywgKXVmw), [HardKernel](https://www.hardkernel.com/), [TwisterOS team](https://twisteros.com/)
+ * 硬件捐赠：[ADLink](https://www.adlinktech.com/Products/Computer_on_Modules/COM-HPC-Server-Carrier-and-Starter-Kit/Ampere_Altra_Developer_Platform?lang=en), [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [DragonBox](https://pyra-handheld.com/), [Novaspirit](https://www.youtube.com/channel/UCrjKdwxaQMSV_NDywgKXVmw), [HardKernel](https://www.hardkernel.com/), [TwisterOS team](https://twisteros.com/)
 * 为本项目持续的宣传：([microLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)), [PILab](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)/[TwisterOS team](https://twisteros.com/), [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA), [NicoD](https://www.youtube.com/channel/UCpv7NFr0-9AB5xoklh3Snhg)

 我还要感谢很多人，即使他们只参与了一次本项目。
@@ -0,0 +1,143 @@
+![Official logo](img/Box86Logo.png "Official Logo")
+
+Емулятор простіра користувача x86_64 із родзинкою
+
+[Дивитись журнал змін](https://github.com/ptitSeb/box86/blob/master/docs/CHANGELOG.md) | [中文](https://github.com/ptitSeb/box86/blob/master/docs/README_CN.md) | [Українська](https://github.com/ptitSeb/box86/blob/master/docs/README_UK.md) | [Повідомити про помилку](https://github.com/ptitSeb/box86/issues/new)
+
+![build](https://app.travis-ci.com/ptitSeb/box86.svg?branch=main) ![stars](https://img.shields.io/github/stars/ptitSeb/box86) ![forks](https://img.shields.io/github/forks/ptitSeb/box86) ![contributors](https://img.shields.io/github/contributors/ptitSeb/box86) ![prs](https://img.shields.io/github/issues-pr/ptitSeb/box86) ![issues](https://img.shields.io/github/issues/ptitSeb/box86)
+
+----
+
+Box86 дозволяє запускати x86 Linux програми (наприклад, ігри) на системах Linux відмінних від x86, наприклад ARM (хост-система має мати 32-розрядний порядок редагування).
+
+Вам *ПОТРІБНА* 32-розрядна підсистема для запуску та створення Box86. Box86 марний лише на 64-розрядних системах. Крім того, вам *ПОТРІБЕН* 32-розрядний інструментарій для створення Box86. Ланцюжок інструментів, який підтримує лише 64-розрядні версії, не компілює Box86, і ви отримаєте помилки (зазвичай на aarch64 ви отримуєте що «-marm» не розпізнається, і вам знадобиться середовище з мультиархітектурою або chroot).
+
+Оскільки Box86 використовує рідні версії деяких «системних» бібліотек, таких як libc, libm, SDL і OpenGL, його легко інтегрувати та використовувати з більшістю програм, а продуктивність у багатьох випадках може бути напрочуд високою. Подивіться на цей стендовий аналіз для прикладу [тут](https://box86.org/index.php/2021/06/game-performances/). Це також означає, що вам знадобиться 32-бітний простір користувача на 64-бітній ОС, наприклад `armhf` поверх `aarch64` 64-бітної ОС.
+
+Більшість x86-ігор потребують OpenGL, тому на платформах ARM може знадобитися таке рішення, як [gl4es](https://github.com/ptitSeb/gl4es). (Деякі платформи ARM підтримують лише OpenGL ES і/або їх реалізація OpenGL хитра. (див. OpenGL на Android))
+
+Box86 інтегрується з DynaRec (динамічний рекомпілятор) для платформ ARM64 і RV64 забезпечуючи підвищення швидкості в 5-10 разів, ніж використання лише інтерпретатора. Деякі високорівневі відомості про те, як працює DynaRec, можна знайти [тут](https://box86.org/2021/07/inner-workings-a-high%e2%80%91level-view-of-box86-and-a-low%e2%80%91level-view-of-the-dynarec/).
+
+Багато ігор просто працюють без особливих налаштувань, наприклад: WorldOfGoo, Airline Tycoon Deluxe і FTL. Багато ігор GameMaker Linux також працюють нормально. (є довгий список, серед них UNDERTALE, A Risk of Rain і Cook Serve Delicious). Ігри Unity3D також працюють добре, але вимога OpenGL може бути проблемою на деяких платформах ARM.
+
+Якщо ви серйозно ставитеся до розробки Box86, вам слід встановити ccache та створити Box86 з ним. (Використовуйте, наприклад, ccmake.)
+Щоб увімкнути TRACE (тобто скидання в stdout усіх окремих виконаних інструкцій x86 із дампом регістрів), вам також знадобиться [бібліотека Zydis](https://github.com/zyantific/zydis), доступна у вашій системі.
+
+Деякі внутрішні коди операцій x86 використовують частини «Бібліотеки емулятора Realmode X86», дивіться [x86primop.c](../src/emu/x86primop.c) для детальної інформації про авторські права.
+
+----
+Компіляція/Встановлення
+----
+> Інструкцію зі складання Box86 можна знайти [тут](COMPILE.md).\
+> Інструкцію з встановлення Wine для Box86 можна знайти [тут](X86WINE.md).
+
+----
+
+Ось 6 відео, перші 2 відео – це відео "Airline Tycoon Deluxe" і "Heretic 2", які працюють на GigaHertz OpenPandora (друге використовує Dynarec), а наступні 2 відео – це відео "Bit.Trip.Runner" і "Neverwinter Night", які працюють на ODroid XU4 (без dynarec), а останні 2 відео на Pi4: Shovel Knight (відео від @ITotalJustice) і Freedom Planet (відео від @djazz), також без 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)
+
+Ви можете знайти багато інших відео про Box86 у YouTube каналах [MicroLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ), [Pi Labs](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g) або [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA).
+
+Список сумісності доступний тут: https://github.com/ptitSeb/box86-compatibility-list/issues
+
+<img src="img/Box86Icon.png" width="96" height="96">
+
+Логотип і піктограма зроблені @grayduck, дякую!
+
+Зауважте, що цей проект не слід плутати з [86box](https://github.com/86Box/86Box), гарним емулятором "повної системи", що спеціалізується на ранньому (до відносно недавньому) апаратному забезпеченні ПК.
+
+----
+
+Використання
+----
+
+Є кілька змінних середовища для керування поведінкою Box86.
+
+Дивіться [тут](USAGE.md) для перегляду всіх змінних середовища та того, що вони роблять.
+
+Примітка: Dynarec від Box86 використовує механізм із захистом пам’яті та обробником сигналу помилки сегментів для обробки JIT коду. Простіше кажучи, якщо ви хочете використовувати GDB для налагодження запущеної програми, яка використовує JIT-код (наприклад, mono/Unity3D), ви все одно матимете багато «звичайних» помилок сегментів. Пропонується використовувати щось на зразок `handle SIGSEGV nostop` в GDB, щоб не зупинятися на кожній помилці сегмента, і, можливо, поставити точку зупину всередині `my_memprotectionhandler` в `signals.c`, якщо ви хочете перехопити помилки сегментів.
+
+----
+
+Історія версій/Журнал змін
+----
+
+Історія версій доступна [тут](docs/CHANGELOG.md).
+
+----
+
+Примітки щодо 64-розрядних платформ
+----
+
+Оскільки Box86 працює шляхом прямого перекладу викликів функцій із x86 на хост-систему, хост-система (на якій працює Box86) повинна мати 32-розрядні бібліотеки. Box86 не містить 32-біт <-> 64-біт перекладу. Отже, щоб запустити Box86, наприклад, на платформі ARM64, вам потрібно буде зібрати Box86 для 32-бітної ARM, а також потрібно мати chroot з 32-бітними бібліотеками.
+
+Якщо ви дивитесь на 64-розрядну версію box86, подивіться на [Box64](https://github.com/ptitSeb/box64): він може запускати двійкові файли x86_64 на 64-розрядних платформах. Але зауважте, що вам все одно потрібен Box86 (і 32-розрядний chroot) для запуску бінарних файлів x86 (як це також трапляється в реальному Linux x86_64, для запуску якого потрібні бібліотеки x86 і двійковий файл в multiarch).
+
+----
+
+Примітки щодо конфігурації Box86
+----
+
+Box86 тепер має конфігураційні файли. Завантажуються 2 файли: `/etc/box4.box86rc` і `~/.box86rc`. Обидва файли мають однаковий синтаксис і в основному є файлами ini. Розділ у квадратних дужках визначає назву процесу, а решта — встановлють змінну середовища. Подивіться на [використання](docs/USAGE.md), щоб дізнатися, які параметри можна використовувати. Box86 поставляється з файлом за замовчуванням, який слід інсталювати для кращої стабільності. Файл знаходиться в `system/box86.box86rc` і має бути встановлено в `/etc/box86.box86rc`. Якщо з певних причин ви не хочете встановлювати цей файл у `/etc`, принаймні скопіюйте його в `~/.box86rc` або гра може не працювати належним чином. Зауважте, що пріоритетом є: `~/.bashrc` > `/etc/box86.box86rc` > командний рядок
+Таким чином, ваші налаштування в `~/.bashrc` можуть замінити налаштування з вашого командного рядка...
+
+----
+
+Примітки щодо емуляції ігор Unity
+----
+
+Запуск ігор Unity має працювати, але ви також повинні зауважити, що для багатьох ігор Unity3D потрібен OpenGL 3+, який може бути складно забезпечити на ARM SBC (одноплатних комп’ютерах). Крім того, багато нових ігор Unity3D (наприклад, KSP) використовують стиснуті текстури BC7, які не підтримуються багатьма інтегрованими GPU ARM.
+Підказка: в Pi4 використовуйте `MESA_GL_VERSION_OVERRIDE=3.2`, а в Panfrost — `PAN_MESA_DEBUG=gl3`, щоб використовувати вищий профіль, якщо гра запускається, а потім завершується, перш ніж щось показує.
+
+----
+
+Примітки щодо Steam
+----
+
+Linux версія Steam тепер може працювати з Box86, але вам також потрібен Box64, щоб його можна було повністю використовувати.
+Рекомендується запускати Steam у small режимі, оскільки він використовує менше пам’яті, але steamwebhelper (64-бітний процес) усе одно завантажуватиметься, навіть якщо він не використовується.
+Екран входу не може працювати без steamwebhelper, і буде показувати лише порожні вікна без належного налаштування box64 у системі.
+Зауважте, що Steam використовуватиме багато пам’яті та ледве поміститься в систему з 4 ГБ оперативної пам’яті. Він більше не працюватиме в системі з меншим об’ємом пам’яті (як обхідний шлях, створіть файл підкачки, увійдіть і поставте прапорець «запам’ятати мене», використовуйте box64rc, щоб вимкнути steamwebhelper і запускати лише в малому режимі без підкачки після першого входу в систему)
+Останнє зауваження: Steam BigPicture працюватиме, але також потрібен steamwebhelper (і box64) і багато пам’яті. Він не запуститься в системі лише з 4 ГБ оперативної пам’яті без файлу підкачки.
+- Якщо у вас виникли проблеми зі встановленням Steam, ви можете знайти `install_steam.sh` у кореневій папці сховища box86. Цей простий скрипт завантажить і встановить Steam у вашу домашню папку, а потім створить ярлик steam у `/usr/local/bin` (для цього він запитає дозвіл sudo). Просто використовуйте `steam` для запуску після встановлення. Зауважте, що інсталяція, яка знаходиться в папці Home, працюватиме лише для одного користувача. Не використовуйте цей скрипт, якщо вам потрібна багатокористувацька інсталяція.
+- Щоб уникнути повідомлення «libc.so.6 відсутній», ви можете використовувати `STEAMOS=1` і `STEAM_RUNTIME=1` як змінні середовища (воно автоматично з’являється, якщо ви використовували сценарій `install_steam.sh`)
+
+Щоб вимкнути `steamwebhelper`, коли Box64 встановлений і працює, створіть (або відредагуйте) `~/.box64rc` і додайте у ньому:
+```
+[steamwebhelper]
+BOX64_EXIT=1
+```
+І тоді `steamwebhelper` буде вимкнено. Прокоментуйте `#` ці 2 рядки, щоб увімкнути його знову.
+
+----
+
+Примітки щодо Wine
+----
+
+Wine тепер підтримується. Інтегровани програми Wine працює як і багато програм та ігор для Windows. Не забувайте, що більшість ігор для Windows використовують Direct3D, для цього може знадобитися повний драйвер OpenGL і якомога більш високий профіль (і gl4es з бекендом ES2 наразі мають проблеми з Wine).
+Примітка: якщо ви плануєте використовувати box86 з Wine на Raspberry Pi 3 або ранішої версії, ці моделі використовують ОС за замовчуванням, яка має ядро з розділенням 2/2 (тобто 2G місця для програми користувача та 2 Г місця для ядра). Це несумісно з програмами Wine, яким потрібен доступ до пам’яті > 2 Гб адреси. Тож вам потрібно буде переналаштувати ядро для розділення 3G/1G.
+
+----
+
+Примітки щодо Vulkan
+----
+
+Box86 вже використовує Vulkan. Якщо у вашій системі є 32-розрядний драйвер Vulkan, Box86 використовуватиме його за потреби. Профіль 1.0, 1.1, 1.2 і 1.3, з деякими розширеннями, має бути в порядку. DXVK, включаючи 2.0, працює. Я знаю, що деякі демонстраційні версії працюють на Pi4 (демонстраційні версії Sascha Willems для x86 працюють так само, як якщо б створювати безпосередньо на armhf). Зауважте, що драйвер Vulkan для Pi4 наразі НЕ підтримує DXVK (обгортка Wine DirectX->Vulkan). Це не проблема Box86, у драйвері відсутні розширення (підтримка апаратного забезпечення) та кілька інших речей, через які DXVK не працює на pi4. З боку Panfrost, PanVK трохи молодий, і я ще не тестував DXVK з ним.
+
+----
+Останнє слово
+----
+
+Я хочу подякувати всім, хто зробив внесок у розвиток box86.
+Є багато способів зробити свій внесок: код, фінанси, апаратне забезпечення та реклама!
+Отже, без особливого порядку, я хочу подякувати:
+ * За основний внесок у код: rajdakin, icecream95, M-HT
+ * За великий фінансовий внесок: FlyingFathead, stormchaser3000, dennis1248, sll00, [libre-computer-project](https://libre.computer/)
+ * За внесок у апаратне забезпечення: [ADLink](https://www.adlinktech.com/Products/Computer_on_Modules/COM-HPC-Server-Carrier-and-Starter-Kit/Ampere_Altra_Developer_Platform?lang=en), [Radxa](https://rockpi.org/), [Pine64](https://www.pine64.org/), [DragonBox](https://pyra-handheld.com/), [Novaspirit](https://www.youtube.com/channel/UCrjKdwxaQMSV_NDywgKXVmw), [HardKernel](https://www.hardkernel.com/), [Команда TwisterOS](https://twisteros.com/)
+ * За постійну рекламу Box86: salva ([microLinux](https://www.youtube.com/channel/UCwFQAEj1lp3out4n7BeBatQ)), [PILab](https://www.youtube.com/channel/UCgfQjdc5RceRlTGfuthBs7g)/[Команда TwisterOS](https://twisteros.com/), [The Byteman](https://www.youtube.com/channel/UCEr8lpIJ3B5Ctc5BvcOHSnA), [NicoD](https://www.youtube.com/channel/UCpv7NFr0-9AB5xoklh3Snhg)
+
+І я також дякую багатьом іншим людям, які хоч раз брали участь у проекті.
+
+(Якщо ви використовуєте Box86 у своєму проекті, будь ласка, не забудьте згадати про це!)
@@ -127,6 +127,11 @@ Enables/Disables trace for generated code.
 Forbid dynablock creation in the interval specified (helpfull for debugging behaviour difference between Dynarec and Interpretor)
 * 0xXXXXXXXX-0xYYYYYYYY : define the interval where dynablock cannot start (inclusive-exclusive)

+#### BOX86_DYNAREC_TEST *
+Dynarec will compare it's execution with the interpreter (super slow, only for testing)
+ * 0 : No comparison
+ * 1 : Each opcode runs on interepter and on Dynarec, and regs and memory are compared and print if different
+
 #### BOX86_DYNAREC_BIGBLOCK *
 Enables/Disables Box86's Dynarec building BigBlock.
 * 0 : Don't try to build block as big as possible (can help program using lots of thread and a JIT, like C#/Unity) (Default when libmonobdwgc-2.0.so is loaded)
@@ -141,8 +146,9 @@ Define Box86's Dynarec max allowed forward value when building Block.
 #### BOX86_DYNAREC_STRONGMEM *
 Enable/Disable simulation of Strong Memory model
 * 0 : Don't try anything special (Default.)
-* 1 : Enable some Memory Barrier when reading from memory (on some MOV opcode) to simulate Strong Memory Model while trying to limit performance impact (Default when libmonobdwgc-2.0.so is loaded)
-* 2 : Enable some Memory Barrier when reading from memory (on some MOV opcode) to simulate Strong Memory Model
+* 1 : Enable some Memory Barrier when writting to memory (on some MOV opcode) to simulate Strong Memory Model while trying to limit performance impact (Default when libmonobdwgc-2.0.so is loaded)
+* 2 : All 1. plus a memory barrier on every write to memory using MOV
+* 3 : All 2. plus Memory Barrier when reading from memory and on some SSE/SSE2 opcodes too

 #### BOX86_DYNAREC_X87DOUBLE *
 Force the use of Double for x87 emulation
@@ -170,11 +176,31 @@ Handling of HotPage (Page beeing both executed and writen)
 * 0 : Don't track hotpage
 * 1-255 : Trak HotPage, and disable execution of a page beeing writen for N attempts (default is 16)

+#### BOX86_DYNAREC_FASTPAGE *
+Will use a faster handling of HotPage (Page being both executed and written)
+* 0 : use regular hotpage (Default)
+* 1 : Use faster hotpage, taking the risk of running obsolete JIT code (might be faster, but more prone to crash)
+
 #### BOX86_DYNAREC_BLEEDING_EDGE *
 Detect MonoBleedingEdge and apply conservative settings
 * 0 : Don't detect MonoBleedingEdge
 * 1 : Detect MonoBleedingEdge, and apply BIGBLOCK=0 STRONGMEM=1 if detected (Default)

+#### BOX86_DYNAREC_JVM *
+Detect libjvm and apply conservative settings
+* 0 : Don't detect libjvm
+* 1 : Detect libjvm, and apply BIGBLOCK=0 STRONGMEM=1 if detected (Default)
+
+#### BOX86_DYNAREC_WAIT *
+Behavior with FillBlock is not availble (FillBlock build Dynarec blocks and is not multithreaded)
+* 0 : Dynarec will not wait for FillBlock to ready and use Interpreter instead (might speedup a bit massive multithread or JIT programs)
+* 1 : Dynarec will wait for FillBlock to be ready (Default)
+
+#### BOX86_DYNAREC_MISSING *
+Dynarec print the missing opcodes
+* 0 : not print the missing opcode (Default, unless DYNAREC_LOG>=1 or DYNAREC_DUMP>=1 is used)
+* 1 : Will print the missing opcodes
+
 #### BOX86_SSE_FLUSHTO0 *
 Handling of SSE Flush to 0 flags
 * 0 : Just track the flag (Default)
@@ -185,6 +211,21 @@ Handling of x87 80bits long double
 * 0 : Try to handle 80bits long double as precise as possible (Default)
 * 1 : Handle them as double

+#### BOX86_LIBCEF *
+Detect libcef and apply malloc_hack settings
+* 0 : Don't detect libcef
+* 1 : Detect libcef, and apply MALLOC_HACK=2 if detected (Default)
+
+#### BOX86_SDL2_JGUID *
+Need a workaround for SDL_GetJoystickGUIDInfo function for wrapped SDL2
+* 0 : Don't use any workaround
+* 1 : Use a workaround for program that use the private SDL_GetJoystickGUIDInfo function with 1 missing argument
+
+#### BOX86_MUTEX_ALIGNED *
+Will mutex are used as-is or wrapped to handle unaligned used
+* 0 : Mutex will be wrapped in case unaligned mutexes are used (Default)
+* 1 : Do no wrap mutex and use them as-is (faster, but might crash with SEGBUS error)
+
 #### BOX86_LIBGL *
 * libXXXX set the name for libGL (defaults to libGL.so.1).
 * /PATH/TO/libGLXXX : Sets the name and path for libGL
@@ -234,6 +275,11 @@ 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_FUTEX_WAITV *
+Use of the new fuext_waitc syscall
+ * 0 : Do not try to use it, return unsupported (Default for BAD_SIGNAL build)
+ * 1 : let program use the syscall if the host system support it (Default for other build)
+
 #### BOX86_ENV
 * XXX=yyyy
 will add XXX=yyyy env. var.
@@ -278,6 +324,12 @@ Will use yyyy as x86_64 interpretor, to launch x64_64 binaries
 * 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)
+ * 3 : Launch `lldb` when a segfault, bus error or illegal instruction signal is trapped, attached to the offending process and go in an endless loop, waiting.

 #### BOX86_NORCFILES
 If the env var exist, no rc files (like /etc/box86.box86rc and ~/.box86rc) will be loaded
+
+#### BOX86_NOSANDBOX
+ * 0 : Nothing special
+ * 1 : Added "--no-sandbox" to command line arguments (usefull for chrome based programs)
+
@@ -42,7 +42,7 @@ 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/wine /usr/local/bin/wine # You could also 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
@@ -22,20 +22,29 @@ echo "#!/bin/bash
 export STEAMOS=1
 export STEAM_RUNTIME=1
 export DBUS_FATAL_WARNINGS=0
-~/steam/bin/steam -noreactlogin steam://open/minigameslist $@" > steam
+~/steam/bin/steam $@" > steam

 # make script executable and move
 chmod +x steam
 sudo mv steam /usr/local/bin/

-# detect if we're running on 64 bit Debian 
+# detect if we're running on 64 bit Debian (maybe this can be moved in another script then mentioned in documentation for the people that don't use steam but don't want a headache)
 MACHINE_TYPE=`uname -m`
 if [ ${MACHINE_TYPE} == 'aarch64' ] && [ -f '/etc/debian_version' ]; then
 echo "Detected 64 bit ARM Debian. Installing 32 bit libraries"
 sudo dpkg --add-architecture armhf # enable installation of armhf libraries
 sudo apt update # update package lists with the newly added arch
 # install the libraries that Steam requires
- sudo apt install libc6:armhf libncurses5:armhf libsdl2-2.0-0:armhf libsdl2-image-2.0-0:armhf libsdl2-mixer-2.0-0:armhf libsdl2-ttf-2.0-0:armhf libopenal1:armhf libpng16-16:armhf libfontconfig1:armhf libxcomposite1:armhf libbz2-1.0:armhf libxtst6:armhf
+ sudo apt install libc6:armhf libsdl2-2.0-0:armhf libsdl2-image-2.0-0:armhf libsdl2-mixer-2.0-0:armhf libsdl2-ttf-2.0-0:armhf libopenal1:armhf libpng16-16:armhf libfontconfig1:armhf libxcomposite1:armhf libbz2-1.0:armhf libxtst6:armhf libsm6:armhf libice6:armhf libgl1:armhf libxinerama1:armhf libxdamage1:armhf
+ # this one is not there all the time, so just try it...
+ sudo apt install libncurses5:armhf 
+
+ # install mesa for armhf if already installed
+ if dpkg-query -W libgl1-mesa-dri 2>/dev/null; then
+  echo "Detected mesa driver for 64 bit ARM. Installing complementary 32 bit one"
+  sudo apt install libgl1-mesa-dri:armhf
+ fi
+
 echo "Don't forget to compile/install Box64!"
 fi

@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
-pkgname=box86-rk3399-git
-pkgver=r3858.5368285
+pkgname=box86-a64-git
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -9,21 +9,28 @@ 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}")
+provides=("${pkgname%-git}" "${pkgname%-a64-git}-git" "${pkgname%-a64-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-a64-git}-git" "${pkgname%-a64-git}")
 source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-rk3399-git}"
+	cd "${srcdir}/${pkgname%-a64-git}"
 	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
 }

+prepare() {
+	cd "${srcdir}/${pkgname%-a64-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DA64=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-rk3399-git}" -DRK3399=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-a64-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-a64-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-arm-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-arm-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-arm-git}" -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-arm-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-arm-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-arm64-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('aarch64')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-arm64-git}-git" "${pkgname%-arm64-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-arm64-git}-git" "${pkgname%-arm64-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-arm64-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-arm64-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DARM64=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-arm64-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-arm64-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-gameshell-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-gameshell-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DGAMESHELL=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-gameshell-git}" -DGAMESHELL=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-gameshell-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-gameshell-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-goaclone-git
+pkgver=r4581.d698861
+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%-goaclone-git}-git" "${pkgname%-goaclone-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-goaclone-git}-git" "${pkgname%-goaclone-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-goaclone-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-goaclone-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DGOA_CLONE=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-goaclone-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-goaclone-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-lx2610a-git
+pkgver=r4581.d698861
+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%-lx2610a-git}-git" "${pkgname%-lx2610a-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-lx2610a-git}-git" "${pkgname%-lx2610a-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DLX2610A=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-odroid-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-odroid-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DODROID=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-odroid-git}" -DODROID=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-odroid-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-odroid-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-odroidn2-git
+pkgver=r4581.d698861
+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%-odroidn2-git}-git" "${pkgname%-odroidn2-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-odroidn2-git}-git" "${pkgname%-odroidn2-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DODROID=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-pandora-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-pandora-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DPANDORA=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-pandora-git}" -DPANDORA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-pandora-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-pandora-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-phytium-git
+pkgver=r4581.d698861
+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%-phytium-git}"-git "${pkgname%-phytium-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-phytium-git}"-git "${pkgname%-phytium-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-phytium-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-phytium-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DPHYTIUM=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-phytium-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-phytium-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-pyra-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-pyra-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DPYRA=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-pyra-git}" -DPYRA=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-pyra-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-pyra-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-rpi-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -16,11 +16,12 @@ md5sums=('SKIP')
 _pimodel=`tr -d '\0' < /sys/firmware/devicetree/base/model`

 pkgver() {
-	cd "$srcdir/${pkgname%-rpi-git}"
+	cd "${srcdir}/${pkgname%-rpi-git}"
 	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
 }

-build() {
+prepare() {
+	cd "${srcdir}/${pkgname%-rpi-git}"
 	# Check for Pi version number
 	if [[ ${_pimodel} =~ "Raspberry Pi 4" ]]; then
 		_piversion=4
@@ -31,15 +32,20 @@ build() {
 	else
 		# This RPI is not supported or this device is not a pi.
 		# We should cancel then the build.
-		echo -e "\033[01m\033[01;31m==> ERROR:\033[00;01m Your device is not supported! This script is for RPi model 4/3/2 (1/0 isn't supported yet).\033[00m"
+		echo -e "\033[01m\033[01;31m==> ERROR:\033[00;01m Your device is not supported! This script is for RPi model 4/3/2 (1/0 isn't supported).\033[00m"
 		exit 1
 	fi
 	echo -e "\033[01m\033[01;34m==> INFO:\033[00;01m Your detected Pi verion number is: \033[01;31m${_piversion}\033[01m.\033[00m"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DRPI${_piversion}=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}

-	cmake -B build -S "${pkgname%-rpi-git}" -DRPI${_piversion}=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+build() {
+	cd "${srcdir}/${pkgname%-rpi-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
+	cd "${srcdir}/${pkgname%-rpi-git}/build"
+	make DESTDIR="${pkgdir}/" install
 }
@@ -0,0 +1,50 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-rk3xxx-git
+pkgver=r4581.d698861
+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%-rk3xxx-git}-git" "${pkgname%-rk3xxx-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-rk3xxx-git}-git" "${pkgname%-rk3xxx-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "$srcdir/${pkgname%-rk3xxx-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-rk3xxx-git}"
+	# Check for RockChip version number
+	if [[ ${_rkmodel} =~ "rk3288"]]; then
+		_rksuffix=3288
+	elif [[ ${_rkmodel} =~ "rk3399" ]]; then
+		_rksuffix=3399	
+	elif [[ ${_rkmodel} =~ "rk3588" ]]; then
+		_rksuffix=3588	
+	else
+		# This RKxxxx is not supported or this device is not RockChip based.
+		# We should cancel then the build.
+		echo -e "\033[01m\033[01;31m==> ERROR:\033[00;01m Your device is not supported! This script is for RockChip model 3399/3588.\033[00m"
+		exit 1
+	fi
+	echo -e "\033[01m\033[01;34m==> INFO:\033[00;01m Your detected RockChip verion number is: \033[01;31m${_rksuffix}\033[01m.\033[00m"
+	[[ ! -d ./build ]] && mkdir build
+    cmake -B build -DRK${_rksuffix}=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-rk3xxx-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-rk3xxx-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-sd845-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-sd845-git}-git" "${pkgname%-sd845-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-sd845-git}-git" "${pkgname%-sd845-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-sd845-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-sd845-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DSD845=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-sd845-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-sd845-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-tegra-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-tegra-git}"-git "${pkgname%-tegra-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-tegra-git}"-git "${pkgname%-tegra-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-tegra-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-tegra-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DTEGRAX1ARM32=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-tegra-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-tegra-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-x86-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('i386')
@@ -14,15 +14,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

 pkgver() {
-	cd "$srcdir/${pkgname%-x86-git}"
+	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}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DLD80BITS=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
 build() {
-	cmake -B build -S "${pkgname%-x86-git}" -DLD80BITS=1 -DNOALIGN=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
-	make -C build
+	cd "${srcdir}/${pkgname%-x86-git}/build"
+	make -j$(nproc)
 }

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-x86-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -12,7 +12,7 @@ This will copy the PKGBUILD for your platform (where you type your platform name
 In the future, I might try to do an unified PKGBUILD for all platforms that will recognize your platform automatically.

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

 ### 3. Any advantages with that sollution?
 This might give you some advantages like those:
@@ -1834,7 +1834,7 @@ def main(root, ver, __debug_forceAllDebugging=False):
 			start = (start >> 4) + ((start & 0xF) << 1);
 		}
 		
-		char tmp[6];
+		char tmp[7];
 		for (int cur = start; cur < start + size - 1; ++cur) {
 			sprintf(tmp, "%c%d, ", regChr, cur);
 			strcat(ret, tmp);
@@ -679,12 +679,12 @@ class Clauses:
 		return isinstance(o, Clauses) and (self.clauses == o.clauses)

 JumbledFunctions     = CustOrderedDictList[Clause, Function]
-FilesSpecific        = Dict[Filename, FileSpec]
+FileSpecifics        = Dict[Filename, FileSpec]

 SortedGlobals        = CustOrderedDictList[Clauses, FunctionType]
 SortedRedirects      = CustOrderedDictList[Clauses, FunctionType]

-def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions, FilesSpecific]:
+def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions, FileSpecifics]:
 	"""
 	readFiles
 	
@@ -693,7 +693,7 @@ def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions,
 	
 	gbls:      JumbledFunctions = CustOrderedDictList()
 	redirects: JumbledFunctions = CustOrderedDictList()
-	filespecs: FilesSpecific    = {}
+	filespecs: FileSpecifics    = {}
 	
 	symbols: Dict[str, Filename] = {}
 	need_halt: bool = False
@@ -751,6 +751,8 @@ def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions,
 							 or match("gstinterfaces010","gstaudio")\
 							 or match("gstreamer010","gstreamer")	\
 							 or match("appindicator","appindicator3")\
+							 or match("appindicator","ayatanaappindicator3")\
+							 or match("appindicator3","ayatanaappindicator3")\
 							 \
 							 or match("libc",        "tcmallocminimal") \
 							 or match("libc",        "ldlinux") 	\
@@ -880,7 +882,7 @@ def readFiles(files: Iterable[str]) -> Tuple[JumbledFunctions, JumbledFunctions,
 	
 	return gbls, redirects, filespecs

-def sortArrays(gbl_funcs: JumbledFunctions, red_funcs: JumbledFunctions, filespecs: FilesSpecific) \
+def sortArrays(gbl_funcs: JumbledFunctions, red_funcs: JumbledFunctions, filespecs: FileSpecifics) \
 -> Tuple[SortedGlobals, SortedRedirects]:
 	# First sort file specific stuff
 	for fn in filespecs:
@@ -944,13 +946,13 @@ def sortArrays(gbl_funcs: JumbledFunctions, red_funcs: JumbledFunctions, filespe
 		redirects[red_vals[funtype]].append(funtype)
 	redirects.sort(key=Clauses.splitdef)
 	
-	def fail(): assert(False)
+	def fail(): assert False
 	for clauses in redirects:
 		redirects[clauses].sort(key=lambda v: fail() if v.redirected is None else v.splitchar() + v.redirected.splitchar())
 	
 	return gbls, redirects

-def checkRun(root: str, gbls: SortedGlobals, redirects: SortedRedirects, filesspec: FilesSpecific) -> Optional[str]:
+def checkRun(root: str, gbls: SortedGlobals, redirects: SortedRedirects, filespecs: FileSpecifics) -> Optional[str]:
 	# Check if there was any new functions compared to last run
 	functions_list: str = ""
 	for clauses in gbls:
@@ -960,18 +962,18 @@ def checkRun(root: str, gbls: SortedGlobals, redirects: SortedRedirects, filessp
 		for v in redirects[clauses]:
 			assert(v.redirected is not None)
 			functions_list = functions_list + "#" + str(clauses) + " " + v.orig + " -> " + v.redirected.orig + "\n"
-	for filename in sorted(filesspec.keys()):
+	for filename in sorted(filespecs.keys()):
 		functions_list = functions_list + filename + ":\n"
-		for st in filesspec[filename].structs:
-			struct = filesspec[filename].structs[st]
+		for st in filespecs[filename].structs:
+			struct = filespecs[filename].structs[st]
 			functions_list = functions_list + \
 				"% " + st + " " + struct.name + " " + struct.repl + "\n"
-		for _bare in filesspec[filename].typedefs:
+		for _bare in filespecs[filename].typedefs:
 			functions_list = functions_list + "- " + _bare.orig + ":\n"
-			for fn in filesspec[filename].typedefs[_bare]:
+			for fn in filespecs[filename].typedefs[_bare]:
 				if fn.no_dlsym: continue
 				functions_list = functions_list + "  - " + fn.name + "\n"
-		for funtype in filesspec[filename].structsuses:
+		for funtype in filespecs[filename].structsuses:
 			assert(funtype.redirected is not None)
 			functions_list = functions_list + "% " + funtype.orig + " -> " + funtype.redirected.orig + "\n"
 	
@@ -992,7 +994,7 @@ def checkRun(root: str, gbls: SortedGlobals, redirects: SortedRedirects, filessp
 	return functions_list

 def generate_files(root: str, files: Iterable[str], ver: str, gbls: SortedGlobals, redirects: SortedRedirects, \
- filespecs: FilesSpecific) -> None:
+ filespecs: FileSpecifics) -> None:
 	# Files header and guard
 	files_header = {
 		"wrapper.c": """
@@ -1148,6 +1150,13 @@ def generate_files(root: str, files: Iterable[str], ver: str, gbls: SortedGlobal
 	for fhdr in files_header:
 		files_guard[fhdr] = trim(files_guard[fhdr])
 	
+	# TODO: put the informations below in a structure (like in box64, where
+	# typedefs & co. are stored in an ABI-specific structure)
+	
+	return_x87: str = "DKdf"
+	if any(c not in FileSpec.values for c in return_x87):
+		raise NotImplementedError("Invalid character")
+	
 	# Typedefs
 	#           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      s
 	tdtypes = ["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*", "...", "...", "void*"]
@@ -1309,6 +1318,22 @@ def generate_files(root: str, files: Iterable[str], ver: str, gbls: SortedGlobal
 			if not clauses.empty():
 				file.write("#endif\n")
 		
+		# Finally, write predicate functions
+		
+		# isRetX87Wrapper
+		file.write("\nint isRetX87Wrapper(wrapper_t fun) {\n")
+		for clauses in gbls:
+			empty = True
+			for funtype in gbls[clauses]:
+				if funtype.orig[0] in return_x87: # TODO: put this in a function (functions would request the ABI for more info)
+					if empty and (not clauses.empty()):
+						file.write("#if " + str(clauses) + "\n")
+						empty = False
+					file.write("\tif (fun == &" + funtype.orig + ") return 1;\n")
+			if not empty:
+				file.write("#endif\n")
+		file.write("\treturn 0;\n}\n")
+		
 		file.write(files_guard["wrapper.c"].format(lbr="{", rbr="}", version=ver))
 	
 	# Rewrite the wrapper.h file:
@@ -1398,6 +1423,6 @@ if __name__ == '__main__':
 		if v == "--":
 			limit.append(i)
 	Define.defines = list(map(DefineType, sys.argv[2:limit[0]]))
-	if main(sys.argv[1], sys.argv[limit[0]+1:], "2.0.0.10") != 0:
+	if main(sys.argv[1], sys.argv[limit[0]+1:], "2.0.0.11") != 0:
 		exit(2)
 	exit(0)
@@ -23,6 +23,7 @@
 #include "custommem.h"
 #include "dictionnary.h"
 #include "rcfile.h"
+#include "gltools.h"
 #ifdef DYNAREC
 #include "dynablock.h"
 #include "dynarec/arm_lock_helper.h"
@@ -78,13 +79,15 @@ void free_tlsdatasize(void* p)
    tlsdatasize_t *data = (tlsdatasize_t*)p;
    box_free(data->ptr);
    box_free(p);
+    if(my_context)
+        pthread_setspecific(my_context->tlskey, NULL);
 }

 int unlockMutex()
 {
    int ret = unlockCustommemMutex();
    int i;
-    #ifdef DYNAREC
+    #ifdef USE_CUSTOM_MUTEX
    void* tid = (void*)GetTID();
    #define GO(A, B)                    \
        i = (arm_lock_storeifref2(&A, NULL, tid)==tid);  \
@@ -139,6 +142,7 @@ static void init_mutexes(box86context_t* context)
 {

 #ifdef DYNAREC
+#ifdef USE_CUSTOM_MUTEX
    arm_lock_stored(&context->mutex_dyndump, 0);
    arm_lock_stored(&context->mutex_once, 0);
    arm_lock_stored(&context->mutex_once2, 0);
@@ -146,6 +150,21 @@ static void init_mutexes(box86context_t* context)
    arm_lock_stored(&context->mutex_tls, 0);
    arm_lock_stored(&context->mutex_thread, 0);
    arm_lock_stored(&context->mutex_bridge, 0);
+#else
+    pthread_mutexattr_t attr;
+    pthread_mutexattr_init(&attr);
+    pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_ERRORCHECK);
+    pthread_mutex_init(&context->mutex_dyndump, &attr);
+    pthread_mutex_init(&context->mutex_once, &attr);
+    pthread_mutex_init(&context->mutex_once2, &attr);
+    pthread_mutex_init(&context->mutex_trace, &attr);
+    pthread_mutex_init(&context->mutex_tls, &attr);
+    pthread_mutex_init(&context->mutex_thread, &attr);
+    pthread_mutex_init(&context->mutex_bridge, &attr);
+
+    pthread_mutexattr_destroy(&attr);
+#endif
+    pthread_mutex_init(&context->mutex_lock, NULL);
 #else
    pthread_mutexattr_t attr;
    pthread_mutexattr_init(&attr);
@@ -210,7 +229,7 @@ box86context_t *NewBox86Context(int argc)
 #ifdef BUILD_LIB
    context->deferedInit = 0;
 #else
-    context->deferedInit = 1;
+    context->deferredInit = 1;
 #endif
    context->sel_serial = 1;

@@ -220,11 +239,12 @@ box86context_t *NewBox86Context(int argc)
    context->local_maplib = NewLibrarian(context, 1);
    context->versym = NewDictionnary();
    context->system = NewBridge();
-    context->globaldefver = NewDefaultVersion();
-    context->weakdefver = NewDefaultVersion();
+    // Cannot use Bridge name as the map is not initialized yet
    // create vsyscall
    context->vsyscall = AddBridge(context->system, iFEv, x86Syscall, 0, NULL);
    addAlternate((void*)0xffffe400, (void*)context->vsyscall);
+    // create exit bridge
+    context->exit_bridge = AddBridge(context->system, NULL, NULL, 0, NULL);
 #ifdef BUILD_LIB
    context->box86lib = RTLD_DEFAULT;   // not ideal
 #else
@@ -238,7 +258,7 @@ box86context_t *NewBox86Context(int argc)
    init_mutexes(context);
    pthread_atfork(NULL, NULL, atfork_child_box64context);

-    pthread_key_create(&context->tlskey, free_tlsdatasize);
+    pthread_key_create(&context->tlskey, NULL/*free_tlsdatasize*/);

    InitFTSMap(context);

@@ -246,11 +266,13 @@ box86context_t *NewBox86Context(int argc)
    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);

+    context->globdata = NewMapSymbols();
+
    initAllHelpers(context);

    return context;
 }
-
+void freeALProcWrapper(box86context_t* context);
 EXPORTDYN
 void FreeBox86Context(box86context_t** context)
 {
@@ -266,6 +288,7 @@ void FreeBox86Context(box86context_t** context)

    box86context_t* ctx = *context;   // local copy to do the cleanning

+    //clean_current_emuthread();    // cleaning main thread seems a bad idea
    if(ctx->local_maplib)
        FreeLibrarian(&ctx->local_maplib, NULL);
    if(ctx->maplib)
@@ -288,8 +311,8 @@ void FreeBox86Context(box86context_t** context)
    if(ctx->zydis)
        DeleteX86Trace(ctx);

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

    /*box_free(ctx->argv);*/
    
@@ -336,7 +359,6 @@ void FreeBox86Context(box86context_t** context)
    void* ptr;
    if ((ptr = pthread_getspecific(ctx->tlskey)) != NULL) {
        free_tlsdatasize(ptr);
-        pthread_setspecific(ctx->tlskey, NULL);
    }
    pthread_key_delete(ctx->tlskey);

@@ -350,24 +372,25 @@ void FreeBox86Context(box86context_t** context)
        }
    }

-    free_neededlib(&ctx->neededlibs);
+    free_neededlib(ctx->neededlibs);
+    ctx->neededlibs = NULL;

    if(ctx->emu_sig)
        FreeX86Emu(&ctx->emu_sig);
-    FreeDefaultVersion(&ctx->globaldefver);
-    FreeDefaultVersion(&ctx->weakdefver);
+
+    FreeMapSymbols(&ctx->globdata);

    finiAllHelpers(ctx);

-#ifndef DYNAREC
+#ifndef USE_CUSTOM_MUTEX
    pthread_mutex_destroy(&ctx->mutex_once);
    pthread_mutex_destroy(&ctx->mutex_once2);
    pthread_mutex_destroy(&ctx->mutex_trace);
    pthread_mutex_destroy(&ctx->mutex_tls);
    pthread_mutex_destroy(&ctx->mutex_thread);
    pthread_mutex_destroy(&ctx->mutex_bridge);
-    pthread_mutex_destroy(&ctx->mutex_lock);
 #endif
+    pthread_mutex_destroy(&ctx->mutex_lock);

    freeCycleLog(ctx);

@@ -375,20 +398,45 @@ void FreeBox86Context(box86context_t** context)
 }

 int AddElfHeader(box86context_t* ctx, elfheader_t* head) {
-    int idx = ctx->elfsize;
-    if(idx==ctx->elfcap) {
-        // resize...
-        ctx->elfcap += 16;
-        ctx->elfs = (elfheader_t**)box_realloc(ctx->elfs, sizeof(elfheader_t*) * ctx->elfcap);
+    int idx = 0;
+    while(idx<ctx->elfsize && ctx->elfs[idx]) idx++;
+    if(idx == ctx->elfsize) {
+        if(idx==ctx->elfcap) {
+            // resize...
+            ctx->elfcap += 16;
+            ctx->elfs = (elfheader_t**)box_realloc(ctx->elfs, sizeof(elfheader_t*) * ctx->elfcap);
+        }
+        ctx->elfs[idx] = head;
+        ctx->elfsize++;
+    } else {
+        ctx->elfs[idx] = head;
    }
-    ctx->elfs[idx] = head;
-    ctx->elfsize++;
    printf_log(LOG_DEBUG, "Adding \"%s\" as #%d in elf collection\n", ElfName(head), idx);
    return idx;
 }

+void RemoveElfHeader(box86context_t* ctx, elfheader_t* head) {
+    if(GetTLSBase(head)) {
+        // should remove the tls info
+        int tlsbase = GetTLSBase(head);
+        /*if(tlsbase == -ctx->tlssize) {
+            // not really correct, but will do for now
+            ctx->tlssize -= GetTLSSize(head);
+            if(!(++ctx->sel_serial))
+                ++ctx->sel_serial;
+        }*/
+    }
+    for(int i=0; i<ctx->elfsize; ++i)
+        if(ctx->elfs[i] == head) {
+            ctx->elfs[i] = NULL;
+            return;
+        }
+}
+
+
 int AddTLSPartition(box86context_t* context, int tlssize) {
    int oldsize = context->tlssize;
+    // should in fact first try to map a hole, but rewinding all elfs and checking filled space, like with the mapmem utilities
    context->tlssize += tlssize;
    context->tlsdata = box_realloc(context->tlsdata, context->tlssize);
    memmove(context->tlsdata+tlssize, context->tlsdata, oldsize);   // move to the top, using memmove as regions will probably overlap
@@ -3,6 +3,6 @@

 #define BOX86_MAJOR 0
 #define BOX86_MINOR 3
-#define BOX86_REVISION 0
+#define BOX86_REVISION 4

 #endif //__BOX86_VERSION_H_
@@ -222,6 +222,8 @@ Op is 20-27
 // 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 ADD_REG_LSR_IMM5_COND(cond, dst, src1, src2, imm5) \
+    EMIT((cond) | 0x00800000 | ((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)
@@ -390,6 +392,8 @@ Op is 20-27
 #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) )
+// ldr reg, [pc+imm]
+#define LDR_literal(reg, imm12) EMIT(c__ | 0b010<<25 | 1<<24 | (((imm12)>=0)?1:0)<<23 | 0<<21 | 1<<20 | 0b1111<<16 | (reg)<<12 | (abs(imm12)))

 // str reg, [addr, #+/-imm9]
 #define STR_IMM9(reg, addr, imm9) EMIT(0xe5000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
@@ -589,8 +593,12 @@ Op is 20-27

 // 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))
+// Count leading 0 bit of Rm, store result in Rd with cond
+#define CLZ_COND(cond, Rd, Rm)  EMIT(cond | 0b00010110<<20 | 0b1111<<16 | (Rd)<<12 | 0b1111<<8 | 0b0001<<4 | (Rm))
 // Reverse bits of Rm, store result in Rd
 #define RBIT(Rd, Rm) EMIT(c__ | 0b01101111<<20 | 0b1111<<16 | (Rd)<<12 | 0b1111<<8 | 0b0011<<4 | (Rm))
+// Reverse bits of Rm, store result in Rd with cond
+#define RBIT_COND(cond, Rd, Rm) EMIT(cond | 0b01101111<<20 | 0b1111<<16 | (Rd)<<12 | 0b1111<<8 | 0b0011<<4 | (Rm))

 #define PLD_gen(U, R, Rn, Imm5, type, Rm) (0b1111<<28 | 0b0111<<24 | (U)<<23 | (R)<<22 | 0b01<<20 | (Rn)<<16 | 0b1111<<12 | (Imm5)<<7 | (type)<<5 | (Rm))
 // Preload Cache Rn+Rm
@@ -600,7 +608,21 @@ Op is 20-27

 #define DMB_gen(opt)    (0b1111<<28 | 0b01010111<<20 | 0b1111<<16 | 0b1111<<12 | 0b0000<<8 | 0b0101<<4 | (opt))
 // Data memory barrier Inner Sharable
+#ifdef PANDORA
+// The Pandora is single core and ordered, DMB is not usefull here so just emit nothing
+#define DMB_ISH()
+#else
 #define DMB_ISH()   EMIT(DMB_gen(0b1011))
+#endif
+
+#define DSB_gen(opt)    (0b1111<<28 | 0b01010111<<20 | 0b1111<<16 | 0b1111<<12 | 0b0000<<8 | 0b0100<<4 | (opt))
+// Data memory barrier Inner Sharable
+#ifdef PANDORA
+// The Pandora is single core and ordered, DSB is not usefull here so just emit nothing
+#define DSB_ISH()
+#else
+#define DSB_ISH()   EMIT(DSB_gen(0b1011))
+#endif

 #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
@@ -626,11 +648,15 @@ Op is 20-27
 #define VMRS(Rt)    EMIT(c__ | (0b1110<<24) | (0b1111<<20) | (0b0001<<16) | ((Rt)<<12) | (0b1010<<8) | (0b0001<<4) | (0b0000))
 // Move to FPSCR from Arm register
 #define VMSR(Rt)    EMIT(c__ | (0b1110<<24) | (0b1110<<20) | (0b0001<<16) | ((Rt)<<12) | (0b1010<<8) | (0b0001<<4) | (0b0000))
+// Move to FPSCR from Arm register with cond
+#define VMSR_cond(cond, Rt)    EMIT(cond | (0b1110<<24) | (0b1110<<20) | (0b0001<<16) | ((Rt)<<12) | (0b1010<<8) | (0b0001<<4) | (0b0000))
 // Move to FPSCR from Arm flags APSR
 #define VMRS_APSR()    VMRS(15)

 // Move between Rt to Sm
 #define VMOVtoV(Sm, Rt) EMIT(c__ | (0b1110<<24) | (0b000<<21) | (0<<20) | ((((Sm)&0b11110)>>1)<<16) | ((Rt)<<12) | (0b1010<<8) | (((Sm)&1)<<7) |(0b00<<6) | (1<<4))
+// Move between Rt to Sm with condition
+#define VMOVtoVcond(cond, Sm, Rt) EMIT(cond | (0b1110<<24) | (0b000<<21) | (0<<20) | ((((Sm)&0b11110)>>1)<<16) | ((Rt)<<12) | (0b1010<<8) | (((Sm)&1)<<7) |(0b00<<6) | (1<<4))
 // Move between Sm to Rt
 #define VMOVfrV(Rt, Sm) EMIT(c__ | (0b1110<<24) | (0b000<<21) | (1<<20) | ((((Sm)&0b11110)>>1)<<16) | ((Rt)<<12) | (0b1010<<8) | (((Sm)&1)<<7) |(0b00<<6) | (1<<4))

@@ -1093,6 +1119,7 @@ Op is 20-27
 #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))
 #define VMULL_U64_U32(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(1, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
+#define VMULL_P64(Dd, Dn, Dm)       EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, 1, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))

 #define VMUL_NEON_gen(op, D, size, Vn, Vd, N, Q, M, Vm)    (0b1111<<28 | 0b001<<25 | (op)<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
 #define VMULQ_32(Dd, Dn, Dm)     EMIT(VMUL_NEON_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
@@ -1390,4 +1417,17 @@ Op is 20-27
 #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))

+// AES
+#define AESD_gen(D, size, Vd, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0110<<7 | 1<<6 | (M)<<5 | (Vm))
+#define AESD(Dd, Dm)            EMIT(AESD_gen(((Dd)>>4)&1, 0, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
+#define AESE_gen(D, size, Vd, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0110<<7 | 0<<6 | (M)<<5 | (Vm))
+#define AESE(Dd, Dm)            EMIT(AESE_gen(((Dd)>>4)&1, 0, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
+#define AESIMC_gen(D, size, Vd, M, Vm)  (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0111<<7 | 1<<6 | (M)<<5 | (Vm))
+#define AESIMC(Dd, Dm)          EMIT(AESIMC_gen(((Dd)>>4)&1, 0, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
+#define AESMC_gen(D, size, Vd, M, Vm)   (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b0111<<7 | 0<<6 | (M)<<5 | (Vm))
+#define AESMC(Dd, Dm)           EMIT(AESMC_gen(((Dd)>>4)&1, 0, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
 #endif  //__ARM_EMITTER_H__
@@ -335,9 +335,9 @@ NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 0 size 1 @<1> Rm @ param=1 char align[4
 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> 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> 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 1 1 0 1 size @<1> @<1> 1 1 0 1 @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]!
-NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 1 1 @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]
-NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> Rm @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 0 1 @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 1 1 @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> Rm @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\], <Rm>
 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> 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> 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>
@@ -131,6 +131,7 @@ arm_lock_storeifref_loop:
    beq     arm_lock_storeifref_loop
    mov     r0, r1
 arm_lock_storeifref_exit:
+    dmb     ish
    bx      lr

 arm_lock_decifnot0b:
@@ -174,13 +175,11 @@ arm_lock_incif0b_0:
    bx      lr

 arm_lock_stored:
-    dmb     ish
    str     r1, [r0]
    dmb     ish
    bx      lr

 arm_lock_storeb:
-    dmb     ish
    strb    r1, [r0]
    dmb     ish
    bx      lr
@@ -9,6 +9,7 @@
 .extern LinkNext

 .global arm_next
+    .4byte  0   // NULL pointer before arm64_next, for getDB
 arm_next:
    // emu is r0
    //stm     r0, {r4-r12,r14}   // don't put put back reg value in emu, faster but more tricky to debug
@@ -159,7 +159,7 @@ const char* print_register_list_fpu(uint8_t start, uint8_t size, uint8_t double_
 		start = (start >> 4) + ((start & 0xF) << 1);
 	}
 	
-	char tmp[6];
+	char tmp[7];
 	for (int cur = start; cur < start + size - 1; ++cur) {
 		sprintf(tmp, "%c%d, ", regChr, cur);
 		strcat(ret, tmp);
@@ -3049,7 +3049,7 @@ const char* arm_print(uint32_t opcode) {
 		int param1_1 = (opcode >> 5) & 0x1;
 		int param2_1 = (opcode >> 4) & 0x1;
 		
-		char align[4];
+		char align[5];
 		if (param2_1 == 1) {
 			sprintf(align, ":%d", 16 << size);
 		} else {
@@ -3064,7 +3064,7 @@ const char* arm_print(uint32_t opcode) {
 		int param1_1 = (opcode >> 5) & 0x1;
 		int param2_1 = (opcode >> 4) & 0x1;
 		
-		char align[4];
+		char align[5];
 		if (param2_1 == 1) {
 			sprintf(align, ":%d", 16 << size);
 		} else {
@@ -3080,7 +3080,7 @@ const char* arm_print(uint32_t opcode) {
 		int param1_1 = (opcode >> 5) & 0x1;
 		int param2_1 = (opcode >> 4) & 0x1;
 		
-		char align[4];
+		char align[5];
 		if (param2_1 == 1) {
 			sprintf(align, ":%d", 16 << size);
 		} else {
@@ -41,21 +41,37 @@ uint32_t X31_hash_code(void* addr, int len)
 	return (uint32_t)h;
 }

-dynablocklist_t* NewDynablockList(uintptr_t text, int textsz, int direct)
+dynablock_t* InvalidDynablock(dynablock_t* db, int need_lock)
 {
-    if(!textsz) {
-        printf_log(LOG_NONE, "Error, creating a NULL sized Dynablock\n");
-        return NULL;
+    if(db) {
+        if(db->gone)
+            return NULL; // already in the process of deletion!
+        dynarec_log(LOG_DEBUG, "InvalidDynablock(%p), db->block=%p x86=%p:%p already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size-1, db->gone);
+        if(need_lock)
+            mutex_lock(&my_context->mutex_dyndump);
+        // remove jumptable
+        setJumpTableDefault(db->x86_addr);
+        db->done = 0;
+        db->gone = 1;
+        if(need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    }
-    dynablocklist_t* ret = (dynablocklist_t*)box_calloc(1, sizeof(dynablocklist_t));
-    ret->text = text;
-    ret->textsz = textsz;
-    if(direct && textsz) {
-        ret->direct = (dynablock_t**)box_calloc(textsz, sizeof(dynablock_t*));
-        if(!ret->direct) {printf_log(LOG_NONE, "Warning, fail to create direct block for dynablock @%p\n", (void*)text);}
+    return db;
+}
+
+void FreeInvalidDynablock(dynablock_t* db, int need_lock)
+{
+    if(db) {
+        if(!db->gone)
+            return; // already in the process of deletion!
+        dynarec_log(LOG_DEBUG, "FreeInvalidDynablock(%p), db->block=%p x86=%p:%p already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size-1, db->gone);
+        if(need_lock)
+            mutex_lock(&my_context->mutex_dyndump);
+        FreeDynarecMap((uintptr_t)db->actual_block);
+        customFree(db);
+        if(need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    }
-    dynarec_log(LOG_DEBUG, "New Dynablocklist %p, from %p->%p\n", ret, (void*)text, (void*)(text+textsz));
-    return ret;
 }

 void FreeDynablock(dynablock_t* db, int need_lock)
@@ -63,367 +79,239 @@ void FreeDynablock(dynablock_t* db, int need_lock)
    if(db) {
        if(db->gone)
            return; // already in the process of deletion!
-        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);
+        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x86=%p:%p already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size-1, db->gone);
        if(need_lock)
            mutex_lock(&my_context->mutex_dyndump);
-        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, 0);
-        }
-        // 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);
-            box_free(db->sons);
-            box_free(db->instsize);
-        }
-        box_free(db);
+        dynarec_log(LOG_DEBUG, " -- FreeDyrecMap(%p, %d)\n", db->actual_block, db->size);
+        db->done = 0;
+        db->gone = 1;
+        if(db->previous)
+            FreeInvalidDynablock(db->previous, 0);
+        FreeDynarecMap((uintptr_t)db->actual_block);
+        customFree(db);
        if(need_lock)
            mutex_unlock(&my_context->mutex_dyndump);
    }
 }

-void FreeDynablockList(dynablocklist_t** dynablocks)
-{
-    if(!dynablocks)
-        return;
-    if(!*dynablocks)
-        return;
-    dynarec_log(LOG_DEBUG, "Free Dynablocklist %p, with Direct Blocks %p\n", *dynablocks, (*dynablocks)->direct);
-    if((*dynablocks)->direct) {
-        for (int i=0; i<(*dynablocks)->textsz; ++i) {
-            if((*dynablocks)->direct[i] && !(*dynablocks)->direct[i]->father) 
-                FreeDynablock((*dynablocks)->direct[i], my_context?1:0);
-        }
-        box_free((*dynablocks)->direct);
-    }
-    (*dynablocks)->direct = NULL;

-    box_free(*dynablocks);
-    *dynablocks = NULL;
-}

 void MarkDynablock(dynablock_t* db)
 {
    if(db) {
-        if(db->father)
-            db = db->father;    // mark only father
-        if(db->need_test)
-            return; // already done
-        db->need_test = 1;
-        setJumpTableDefault(db->x86_addr);
-        for(int i=0; i<db->sons_size; ++i)
-            setJumpTableDefault(db->sons[i]->x86_addr);
+        dynarec_log(LOG_DEBUG, "MarkDynablock %p %p-%p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1);
+        if(!setJumpTableIfRef(db->x86_addr, db->jmpnext, db->block)) {
+            dynablock_t* old = db;
+            db = getDB((uintptr_t)old->x86_addr);
+            if(!old->gone && db!=old) {
+                printf_log(LOG_INFO, "Warning, couldn't mark block as dirty for %p, block=%p, current_block=%p\n", old->x86_addr, old, db);
+                // the block is lost, need to invalidate it...
+                old->gone = 1;
+                old->done = 0;
+                if(!db || db->previous)
+                    FreeInvalidDynablock(old, 1);
+                else
+                    db->previous = old;
+            }
+        }
    }
 }

-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)
+static 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)
+static int MarkedDynablock(dynablock_t* db)
 {
-    if(!dynablocks)
-        return;
-    if(!dynablocks->direct)
-        return;
-    uintptr_t startdb = dynablocks->text;
-    uintptr_t enddb = startdb + dynablocks->textsz -1;
-    uintptr_t start = addr;
-    uintptr_t end = addr+size-1;
-    if(start<startdb)
-        start = startdb;
-    if(end>enddb)
-        end = enddb;
-    dynablock_t *db;
-    if(end>startdb && start<enddb)
-        for(uintptr_t i = start; i<end; ++i)
-            if((db=dynablocks->direct[i-startdb]))
-                if(IntervalIntersects((uintptr_t)db->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) {
-        int need_lock = my_context?1:0;
-        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, need_lock);
-        );
-        kh_destroy(dynablocks, blocks);
+    if(db) {
+        if(getNeedTest((uintptr_t)db->x86_addr))
+            return 1; // already done
    }
+    return 0;
 }
-void MarkRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+
+void MarkRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
 {
-    if(!dynablocks)
+    // Mark will try to find *any* blocks that intersect the range to mark
+    if(!db)
        return;
-    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);
+    dynarec_log(LOG_DEBUG, "MarkRangeDynablock %p-%p .. startdb=%p, sizedb=%p\n", (void*)addr, (void*)addr+size-1, (void*)db->x86_addr, (void*)db->x86_size);
+    if(IntervalIntersects((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1))
+        MarkDynablock(db);
+}
+
+int FreeRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
+{
+    if(!db)
+        return 1;
+
+    int need_lock = my_context?1:0;
+    if(IntervalIntersects((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1)) {
+        FreeDynablock(db, need_lock);
+        return 0;
    }
+    return 1;
 }

-dynablock_t* FindDynablockDynablocklist(void* addr, kh_dynablocks_t* dynablocks)
+dynablock_t *AddNewDynablock(uintptr_t addr)
 {
-    if(!dynablocks)
-        return NULL;
-    dynablock_t* 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;
+    dynablock_t* block;
+    #if 0
+    // check if memory as the correct flags
+    int prot = getProtection(addr);
+    if(!(prot&(PROT_EXEC|PROT_DYNAREC|PROT_DYNAREC_R))) {
+        dynarec_log(LOG_VERBOSE, "Block asked on a memory with no execution flags 0x%02X\n", prot);
        return NULL;
    }
-    if((addr<dynablocks->text) || (addr>=(dynablocks->text+dynablocks->textsz))) {
-        // 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_VERBOSE, "Block already exist in Direct Map\n");
-            *created = 0;
-            return block;
-        }
-    }
    
-    if (!*created)
-        return block;
-    
-    if(!dynablocks->direct) {
-        dynablock_t** p = (dynablock_t**)box_calloc(dynablocks->textsz, sizeof(dynablock_t*));
-        if(arm_lock_storeifnull(&dynablocks->direct, p)!=NULL)
-            box_free(p);    // someone already create the direct array, too late...
-    }
-
+    #endif
    // create and add new block
    dynarec_log(LOG_VERBOSE, "Ask for DynaRec Block creation @%p\n", (void*)addr);
-
-    block = (dynablock_t*)box_calloc(1, sizeof(dynablock_t));
-    block->parent = dynablocks; 
-    dynablock_t* tmp = (dynablock_t*)arm_lock_storeifref(&dynablocks->direct[addr-dynablocks->text], block, NULL);
-    if(tmp !=  block) {
-        // a block appeard!
-        box_free(block);
-        *created = 0;
-        return tmp;
-    }
-    *created = 1;
+    block = (dynablock_t*)customCalloc(1, sizeof(dynablock_t));
    return block;
 }

 //TODO: move this to dynrec_arm.c and track allocated structure to avoid memory leak
-static __thread struct __jmp_buf_tag dynarec_jmpbuf;
+#ifdef ANDROID
+#define JUMPBUFF sigjmp_buf
+#else
+#define JUMPBUFF struct __jmp_buf_tag
+#endif
+static __thread JUMPBUFF dynarec_jmpbuf;
+#ifdef ANDROID
+#define DYN_JMPBUF dynarec_jmpbuf
+#else
+#define DYN_JMPBUF &dynarec_jmpbuf
+#endif

 void cancelFillBlock()
 {
-    longjmp(&dynarec_jmpbuf, 1);
+    longjmp(DYN_JMPBUF, 1);
 }

 /* 
    return NULL if block is not found / cannot be created. 
    Don't create if create==0
 */
-static dynablock_t* internalDBGetBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, dynablock_t* current, int need_lock)
+static dynablock_t* internalDBGetBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, int need_lock)
 {
    if(hasAlternate((void*)addr))
        return NULL;
-    // try the quickest way first: get parent of current and check if ok!
-    dynablocklist_t *dynablocks = NULL;
-    dynablock_t* block = NULL;
-    if(current && current->done) {
-        dynablocks = current->parent;
-        if(dynablocks && !(addr>=dynablocks->text && addr<(dynablocks->text+dynablocks->textsz)))
-            dynablocks = NULL;
-    }
-    // nope, lets do the long way
-    if(!dynablocks) {
-        dynablocks = getDBFromAddress(addr);
-        if(!dynablocks) {
-            dynablocks = GetDynablocksFromAddress(emu->context, addr);
-            if(!dynablocks)
+    dynablock_t* block = getDB(addr);
+    if(block || !create)
+        return block;
+
+    if(need_lock) {
+        if(box86_dynarec_wait) {
+            mutex_lock(&my_context->mutex_dyndump);
+        } else {
+            if(mutex_trylock(&my_context->mutex_dyndump))   // FillBlock not available for now
                return NULL;
        }
-    }
-    // check direct first, without lock
-    if(dynablocks->direct/* && (addr>=dynablocks->text) && (addr<(dynablocks->text+dynablocks->textsz))*/)
-        if((block = dynablocks->direct[addr-dynablocks->text]))
-            return block;
-
-    int created = create;
-    if(need_lock)
-        mutex_lock(&my_context->mutex_dyndump);
-    block = AddNewDynablock(dynablocks, addr, &created);
-    if(!created) {
-        if(need_lock)
+        block = getDB(addr);    // just in case
+        if(block) {
            mutex_unlock(&my_context->mutex_dyndump);
-        return block;   // existing block...
+            return block;
+        }
    }
+    
+    block = AddNewDynablock(addr);

    // fill the block
    block->x86_addr = (void*)addr;
-    if(sigsetjmp(&dynarec_jmpbuf, 1)) {
+    if(sigsetjmp(DYN_JMPBUF, 1)) {
        printf_log(LOG_INFO, "FillBlock at %p triggered a segfault, cancelling\n", (void*)addr);
+        FreeDynablock(block, 0);
        if(need_lock)
            mutex_unlock(&my_context->mutex_dyndump);
        return NULL;
    }
    void* ret = FillBlock(block, filladdr);
-    if(need_lock)
-        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
-        }
-        box_free(block);
+        customFree(block);
        block = NULL;
    }
    // check size
-    if(block && block->x86_size) {
+    if(block) {
        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);   // already protected in FillBlock
-        // but check if it already need test
-        if(block->need_test || (block->father && block->father->need_test))
-            return NULL;
+        if(blocksz>my_context->max_db_size)
+            my_context->max_db_size = blocksz;
        // 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);
+        if(!addJumpTableIfDefault(block->x86_addr, block->dirty?block->jmpnext:block->block)) {
+            FreeDynablock(block, 0);
+            block = getDB(addr);
+            MarkDynablock(block);   // just in case...
+        } else {
+            if(block->x86_size)
+                block->done = 1;    // don't validate the block if the size is null, but keep the block
+        }
    }
+    if(need_lock)
+        mutex_unlock(&my_context->mutex_dyndump);

-    dynarec_log(LOG_DEBUG, " --- DynaRec Block %s @%p:%p (%p, 0x%x bytes, with %d son(s))\n", created?"created":"recycled", (void*)addr, (void*)(addr+((block)?block->x86_size:0)), (block)?block->block:0, (block)?block->size:0, (block)?block->sons_size:0);
+    dynarec_log(LOG_DEBUG, "%04d| --- DynaRec Block created @%p:%p (%p, 0x%x bytes)\n", GetTID(), (void*)addr, (void*)(addr+((block)?block->x86_size:1)-1), (block)?block->block:0, (block)?block->size:0);

    return block;
 }

-dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t** current)
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create)
 {
-    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, *current, 1);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
+    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, 1);
+    if(db && db->done && db->block && getNeedTest(addr)) {
+        if(db->always_test)
+            sched_yield();  // just calm down...
        if(AreaInHotPage((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr + db->x86_size - 1)) {
-            dynarec_log(LOG_DEBUG, "Not running block %p from %p:%p with for %p because it's in a hotpage\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, (void*)addr);
-            return NULL;
+            emu->test.test = 0;
+            if(box86_dynarec_fastpage) {
+                uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+                if(hash==db->hash) { // seems ok, run it without reprotecting it
+                    setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+                    return db;
+                }
+                db->done = 0;   // invalidating the block, it's already not good
+                dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, hash, db->hash, (void*)addr);
+                // Free db, it's now invalid!
+                FreeDynablock(db, 1);
+                return NULL;    // not building a new one, it's still a hotpage
+            } else {
+                dynarec_log(LOG_INFO, "Not running block %p from %p:%p with for %p because it's in a hotpage\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, (void*)addr);
+                return NULL;
+            }
        }
-        if(mutex_trylock(&my_context->mutex_dyndump))
-            return NULL;
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = (getProtection((uintptr_t)father->x86_addr)&PROT_READ)?X31_hash_code(father->x86_addr, father->x86_size):0;
-        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, 0);
+        uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+        int need_lock = mutex_trylock(&my_context->mutex_dyndump);
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            dynablock_t* old = InvalidDynablock(db, need_lock);
            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, addr, create, *current, 0);
+            db = internalDBGetBlock(emu, addr, addr, create, need_lock);
+            if(db) {
+                if(db->previous)
+                    FreeInvalidDynablock(db->previous, need_lock);
+                db->previous = old;
+            } else
+                FreeInvalidDynablock(old, need_lock);
        } else {
-            father->need_test = 0;
-            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            dynarec_log(LOG_DEBUG, "Validating block %p from %p:%p (hash:%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, db->hash, (void*)addr);
+            protectDB((uintptr_t)db->x86_addr, db->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);
+            if(isprotectedDB((uintptr_t)db->x86_addr, db->x86_size) && !db->always_test) {
+                setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+            }
        }
-        mutex_unlock(&my_context->mutex_dyndump);
+        if(!need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    } 
+    if(!db || !db->block || !db->done)
+        emu->test.test = 0;
    return db;
 }

@@ -431,27 +319,36 @@ 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, 1);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
-        if(mutex_trylock(&my_context->mutex_dyndump))
-            return NULL;
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = (getProtection((uintptr_t)father->x86_addr)&PROT_READ)?X31_hash_code(father->x86_addr, father->x86_size):0;
-        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, 0);
+    dynablock_t *db = internalDBGetBlock(emu, addr, filladdr, create, 1);
+    if(db && db->done && db->block && getNeedTest(filladdr)) {
+        if(db->always_test)
+            sched_yield();  // just calm down...
+        int need_lock = mutex_trylock(&my_context->mutex_dyndump);
+        uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating alt block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            dynablock_t* old = InvalidDynablock(db, need_lock);
            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, filladdr, create, NULL, 0);
+            db = internalDBGetBlock(emu, addr, filladdr, create, need_lock);
+            if(db) {
+                if(db->previous)
+                    FreeInvalidDynablock(db->previous, need_lock);
+                db->previous = old;
+            } else
+                FreeInvalidDynablock(old, need_lock);
        } else {
-            father->need_test = 0;
-            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            protectDB((uintptr_t)db->x86_addr, db->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);
+            if(isprotectedDB((uintptr_t)db->x86_addr, db->x86_size) && !db->always_test) {
+                setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+            }
        }
-        mutex_unlock(&my_context->mutex_dyndump);
+        if(!need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    } 
+    if(!db || !db->block || !db->done)
+        emu->test.test = 0;
    return db;
 }
@@ -1,36 +1,26 @@
 #ifndef __DYNABLOCK_PRIVATE_H_
 #define __DYNABLOCK_PRIVATE_H_

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

 typedef struct dynablock_s {
-    dynablocklist_t* parent;
-    void*           block;
+    void*           block;  // block-sizeof(void*) == self
+    void*           actual_block;   // the actual start of the block (so block-sizeof(void*))
+    struct dynablock_s*    previous;   // a previous block that might need to be freed
    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;
+    uint8_t         always_test;
+    uint8_t         dirty;      // if need to be tested as soon as it's created
    int             isize;
-    dynablock_t**   sons;   // sons (kind-of dummy dynablock...)
-    int             sons_size;
-    dynablock_t*    father; // set only in the case of a son
    instsize_t*     instsize;
+    void*           jmpnext;    // a branch jmpnext code when block is marked
 } dynablock_t;

-typedef struct dynablocklist_s {
-    uintptr_t           text;
-    int                 textsz;
-    int                 maxsz;     // maxblock size (for this block or previous block)
-    dynablock_t**       direct;    // direct mapping (waste of space, so the array is created at first write)
-} dynablocklist_t;
-
 #endif //__DYNABLOCK_PRIVATE_H_
@@ -20,6 +20,7 @@
 #include "dynablock.h"
 #include "dynablock_private.h"
 #include "bridge.h"
+void x86test_check(x86emu_t* ref, uintptr_t ip);
 #endif

 #ifdef ARM
@@ -44,26 +45,27 @@ void* LinkNext(x86emu_t* emu, uintptr_t addr, void* x2)
        printf_log(LOG_NONE, "Warning, jumping to NULL address from %p (db=%p, x86addr=%p/%s)\n", x2, db, x86addr, pcname);
    }
    #endif
-    dynablock_t* current = NULL;
    void * jblock;
-    dynablock_t* block = DBGetBlock(emu, addr, 1, &current);
+    dynablock_t* block = NULL;
+    if(hasAlternate((void*)addr)) {
+        printf_log(LOG_INFO, "Jmp address has alternate: %p", (void*)addr);
+        addr = (uintptr_t)getAlternate((void*)addr);
+        R_EIP = addr;
+        printf_log(LOG_INFO, " -> %p\n", (void*)addr);
+        block = DBGetBlock(emu, addr, 1);
+    } else
+        block = DBGetBlock(emu, addr, 1);
    if(!block) {
        // no block, let link table as is...
-        if(hasAlternate((void*)addr)) {
-            printf_log(LOG_INFO, "Jmp address has alternate: %p", (void*)addr);
-            addr = (uintptr_t)getAlternate((void*)addr);
-            R_EIP = addr;
-            printf_log(LOG_INFO, " -> %p\n", (void*)addr);
-            block = DBGetBlock(emu, addr, 1, &current);
-        }
-        if(!block) {
-            #ifdef HAVE_TRACE
+        #ifdef HAVE_TRACE
+        if(LOG_INFO<=box86_dynarec_log) {
            dynablock_t* db = FindDynablockFromNativeAddress(x2);
-            dynarec_log(LOG_INFO, "Warning, jumping to a no-block address %p from %p (db=%p, x86addr=%p)\n", (void*)addr, x2, db, db?(void*)getX86Address(db, (uintptr_t)x2):NULL);
-            #endif
-            //tableupdate(arm_epilog, addr, table);
-            return arm_epilog;
+            elfheader_t* h = FindElfAddress(my_context, (uintptr_t)x2);
+            dynarec_log(LOG_INFO, "Warning, jumping to a no-block address %p from %p (db=%p, x64addr=%p(elf=%s))\n", (void*)addr, x2, db, db?(void*)getX86Address(db, (uintptr_t)x2-4):NULL, h?ElfName(h):"(none)");
        }
+        #endif
+        //tableupdate(arm_epilog, addr, table);
+        return arm_epilog;
    }
    if(!block->done) {
        // not finished yet... leave linker
@@ -79,166 +81,122 @@ void* LinkNext(x86emu_t* emu, uintptr_t addr, void* x2)
 }
 #endif

+#ifdef ANDROID
+#define JUMPBUFF sigjmp_buf
+#else
+#define JUMPBUFF struct __jmp_buf_tag
+#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
-            }
-        }
+    uint32_t old_esp = R_ESP;
+    uint32_t old_ebx = R_EBX;
+    uint32_t old_edi = R_EDI;
+    uint32_t old_esi = R_ESI;
+    uint32_t old_ebp = R_EBP;
+    uint32_t old_eip = R_EIP;
+    // save defered flags
+    defered_flags_t old_df = emu->df;
+    uint32_t old_op1 = emu->op1;
+    uint32_t old_op2 = emu->op2;
+    uint32_t old_res = emu->res;
+    // uc_link
+    i386_ucontext_t* old_uc_link = emu->uc_link;
+    emu->uc_link = NULL;
+
+    PushExit(emu);
+    R_EIP = addr;
+    emu->df = d_none;
+    DynaRun(emu);
+    emu->quit = 0;  // reset Quit flags...
+    emu->df = d_none;
+    emu->uc_link = old_uc_link;
+    if(emu->flags.quitonlongjmp && emu->flags.longjmp) {
+        if(emu->flags.quitonlongjmp==1)
+            emu->flags.longjmp = 0;   // don't change anything because of the longjmp
+    } else {
+        // restore defered flags
+        emu->df = old_df;
+        emu->op1 = old_op1;
+        emu->op2 = old_op2;
+        emu->res = old_res;
+        // and the old registers
+        R_EBX = old_ebx;
+        R_EDI = old_edi;
+        R_ESI = old_esi;
+        R_EBP = old_ebp;
+        R_ESP = old_esp;
+        R_EIP = old_eip;  // and set back instruction pointer
    }
-#ifdef DYNAREC
-    if(!box86_dynarec)
-#endif
-        EmuCall(emu, addr);
-#ifdef DYNAREC
-    else {
-        uint32_t old_esp = R_ESP;
-        uint32_t old_ebx = R_EBX;
-        uint32_t old_edi = R_EDI;
-        uint32_t old_esi = R_ESI;
-        uint32_t old_ebp = R_EBP;
-        uint32_t old_eip = R_EIP;
-        PushExit(emu);
-        R_EIP = addr;
-        emu->df = d_none;
-        dynablock_t* block = NULL;
-        dynablock_t* current = NULL;
-        while(!emu->quit) {
-            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, "%04d|Calling Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
-                Run(emu, 1);
-            } else {
-                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
-                arm_prolog(emu, block->block);
-                #endif
-            }
-            if(emu->fork) {
-                int forktype = emu->fork;
-                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...
-        emu->df = d_none;
-        if(emu->quitonlongjmp && emu->longjmp) {
-            emu->longjmp = 0;   // don't change anything because of the longjmp
-        } else {
-            R_EBX = old_ebx;
-            R_EDI = old_edi;
-            R_ESI = old_esi;
-            R_EBP = old_ebp;
-            R_ESP = old_esp;
-            R_EIP = old_eip;  // and set back instruction pointer
-        }
-    }
-#endif
-    // clear the setjmp
-    if(ejb && jmpbuf_reset)
-        ejb->jmpbuf_ok = 0;
 }

 int my_setcontext(x86emu_t* emu, void* ucp);
 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
+    JUMPBUFF jmpbuf[1] = {0};
+    int skip = 0;
+    JUMPBUFF *old_jmpbuf = emu->jmpbuf;
+    emu->flags.jmpbuf_ready = 0;

+    while(!(emu->quit)) {
+        if(!emu->jmpbuf || (emu->flags.need_jmpbuf && emu->jmpbuf!=jmpbuf)) {
+            emu->jmpbuf = jmpbuf;
+            emu->flags.jmpbuf_ready = 1;
+            #ifdef ANDROID
+            if((skip=sigsetjmp(*(JUMPBUFF*)emu->jmpbuf, 1))) 
+            #else
+            if((skip=sigsetjmp(emu->jmpbuf, 1))) 
+            #endif
+            {
+                printf_log(LOG_DEBUG, "Setjmp DynaRun\n");
+                #ifdef DYNAREC
+                if(box86_dynarec_test) {
+                    if(emu->test.clean)
+                        x86test_check(emu, R_EIP);
+                    emu->test.clean = 0;
+                }
+                #endif
            }
        }
-    }
-#ifdef DYNAREC
-    if(!box86_dynarec)
-#endif
-        return Run(emu, 0);
-#ifdef DYNAREC
-    else {
-        dynablock_t* block = NULL;
-        dynablock_t* current = NULL;
-        while(!emu->quit) {
-            block = DBGetBlock(emu, R_EIP, 1, &current);
-            current = block;
+        if(emu->flags.need_jmpbuf)
+            emu->flags.need_jmpbuf = 0;
+
+#ifndef DYNAREC
+        Run(emu, 0);
+#else
+        if(!box86_dynarec)
+            Run(emu, 0);
+        else {
+            dynablock_t* block = (skip || ACCESS_FLAG(F_TF))?NULL:DBGetBlock(emu, R_EIP, 1);
            if(!block || !block->block || !block->done) {
+                skip = 0;
                // no block, of block doesn't have DynaRec content (yet, temp is not null)
                // Use interpreter (should use single instruction step...)
-                dynarec_log(LOG_DEBUG, "%04d|Running Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Running Interpreter @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
+                if(box86_dynarec_test)
+                    emu->test.clean = 0;
                Run(emu, 1);
            } else {
-                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);
+                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x86 insts (hash=0x%x) emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize, block->hash, emu);
                // block is here, let's run it!
-                #ifdef ARM
                arm_prolog(emu, block->block);
-                #endif
            }
            if(emu->fork) {
                int forktype = emu->fork;
                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]);
-                }
            }
-            else if(emu->quit && emu->uc_link) {
+            if(emu->quit && emu->uc_link) {
                emu->quit = 0;
                my_setcontext(emu, emu->uc_link);
            }
        }
+#endif
+        if(emu->flags.need_jmpbuf)
+            emu->quit = 0;
    }
    // clear the setjmp
-    if(ejb && jmpbuf_reset)
-        ejb->jmpbuf_ok = 0;
-    return 0;
-#endif
+    emu->jmpbuf = old_jmpbuf;
 }
@@ -73,8 +73,8 @@ void add_next(dynarec_arm_t *dyn, uintptr_t addr) {
        }
    // add slots
    if(dyn->next_sz == dyn->next_cap) {
-        dyn->next_cap += 16;
-        dyn->next = (uintptr_t*)box_realloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
+        dyn->next_cap += 64;
+        dyn->next = (uintptr_t*)dynaRealloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
    }
    dyn->next[dyn->next_sz++] = addr;
 }
@@ -83,16 +83,33 @@ uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr) {
    uintptr_t best = 0;
    int i = 0;
    while((i<dyn->next_sz) && (best!=addr)) {
-        if(dyn->next[i]<addr) { // remove the address, it's before current address
-            dyn->next[i] = 0;
-        } else {
-            if((dyn->next[i]<best) || !best)
-                best = dyn->next[i];
+        if(dyn->next[i]) {
+            if(dyn->next[i]<addr) { // remove the address, it's before current address
+                dyn->next[i] = 0;
+            } else {
+                if((dyn->next[i]<best) || !best)
+                    best = dyn->next[i];
+            }
        }
        ++i;
    }
    return best;
 }
+void add_jump(dynarec_arm_t *dyn, int ninst) {
+    // add slots
+    if(dyn->jmp_sz == dyn->jmp_cap) {
+        dyn->jmp_cap += 64;
+        dyn->jmps = (int*)dynaRealloc(dyn->jmps, dyn->jmp_cap*sizeof(int));
+    }
+    dyn->jmps[dyn->jmp_sz++] = ninst;
+}
+int get_first_jump(dynarec_arm_t *dyn, int next) {
+    for(int i=0; i<dyn->jmp_sz; ++i)
+        if(dyn->insts[dyn->jmps[i]].x86.jmp == next)
+            return i;
+    return -2;
+}
+
 #define PK(A) (*((uint8_t*)(addr+(A))))
 int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n)
 {
@@ -241,35 +258,31 @@ int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n)
    return (i==n)?1:0;
 }

-static instsize_t* addInst(instsize_t* insts, size_t* size, size_t* cap, int x86_size, int arm_size)
+void addInst(instsize_t* insts, size_t* size, int x86_size, int native_size)
 {
    // x86 instruction is <16 bytes
    int toadd;
-    if(x86_size>arm_size)
+    if(x86_size>native_size)
        toadd = 1 + x86_size/15;
    else
-        toadd = 1 + arm_size/15;
-    if((*size)+toadd>(*cap)) {
-        *cap = (*size)+toadd;
-        insts = (instsize_t*)box_realloc(insts, (*cap)*sizeof(instsize_t));
-    }
+        toadd = 1 + native_size/15;
    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)
+        if(native_size>15)
            insts[*size].nat = 15;
        else
-            insts[*size].nat = arm_size;
-        arm_size -= insts[*size].nat;
+            insts[*size].nat = native_size;
+        native_size -= insts[*size].nat;
        ++(*size);
        --toadd;
    }
-    return insts;
 }

+
 static void fillPredecessors(dynarec_arm_t* dyn)
 {
    int pred_sz = 1;    // to be safe
@@ -281,6 +294,13 @@ static void fillPredecessors(dynarec_arm_t* dyn)
            ++dyn->insts[dyn->insts[i].x86.jmp_insts].pred_sz;
        }
    }
+    // remove "has_next" from orphean branch
+    for(int i=0; i<dyn->size-1; ++i) {
+        if(!dyn->insts[i].x86.has_next) {
+            if(dyn->insts[i+1].x86.has_next && !dyn->insts[i+1].pred_sz)
+                dyn->insts[i+1].x86.has_next = 0;
+        }
+    }
    // second the "has_next"
    for(int i=0; i<dyn->size-1; ++i) {
        if(dyn->insts[i].x86.has_next) {
@@ -288,7 +308,7 @@ static void fillPredecessors(dynarec_arm_t* dyn)
            ++dyn->insts[i+1].pred_sz;
        }
    }
-    dyn->predecessor = (int*)box_malloc(pred_sz*sizeof(int));
+    dyn->predecessor = (int*)dynaMalloc(pred_sz*sizeof(int));
    // fill pred pointer
    int* p = dyn->predecessor;
    for(int i=0; i<dyn->size; ++i) {
@@ -305,8 +325,6 @@ static void fillPredecessors(dynarec_arm_t* dyn)
            dyn->insts[j].pred[dyn->insts[j].pred_sz++] = i;
        }
    }
-
-
 }

 // updateNeed goes backward, from last intruction to top
@@ -332,9 +350,8 @@ static int updateNeed(dynarec_arm_t* dyn, int ninst, uint8_t need) {
        need = dyn->insts[ninst].x86.need_after&~dyn->insts[ninst].x86.gen_flags;
        if(dyn->insts[ninst].x86.may_set)
            need |= dyn->insts[ninst].x86.gen_flags;    // forward the flags
-        // Consume X_PEND if relevant
-        if((need&X_PEND) && (dyn->insts[ninst].x86.set_flags&SF_PENDING))
-            need &=~X_PEND;
+        else if((need&X_PEND) && (dyn->insts[ninst].x86.set_flags&SF_PENDING))
+            need &=~X_PEND;     // Consume X_PEND if relevant
        need |= dyn->insts[ninst].x86.use_flags;
        if(dyn->insts[ninst].x86.need_before == need)
            return ninst - 1;
@@ -349,43 +366,77 @@ static int updateNeed(dynarec_arm_t* dyn, int ninst, uint8_t need) {
            else
                updateNeed(dyn, dyn->insts[ninst].pred[i], need);
        }
-        if(!ok)
-            return ninst - 1;
        --ninst;
+        if(!ok)
+            return ninst;
    }
    return ninst;
 }

+void* current_helper = NULL;
+
+void CancelBlock(int need_lock)
+{
+    if(need_lock)
+        mutex_lock(&my_context->mutex_dyndump);
+    dynarec_arm_t* helper = (dynarec_arm_t*)current_helper;
+    current_helper = NULL;
+    if(helper) {
+        dynaFree(helper->next);
+        dynaFree(helper->insts);
+        dynaFree(helper->predecessor);
+        if(helper->dynablock && helper->dynablock->actual_block) {
+            FreeDynarecMap((uintptr_t)helper->dynablock->actual_block);
+            helper->dynablock->actual_block = NULL;
+        }
+    }
+    if(need_lock)
+        mutex_unlock(&my_context->mutex_dyndump);
+}
+
 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* current_helper = NULL;
-
-void CancelBlock()
-{
-    dynarec_arm_t* helper = (dynarec_arm_t*)current_helper;
-    current_helper = NULL;
-    if(!helper)
-        return;
-    box_free(helper->next);
-    box_free(helper->insts);
-    box_free(helper->predecessor);
-    if(helper->dynablock && helper->dynablock->block)
-        FreeDynarecMap(helper->dynablock, (uintptr_t)helper->dynablock->block, helper->dynablock->size);
-}
+void arm_epilog();
+void arm_next();

 void* CreateEmptyBlock(dynablock_t* block, uintptr_t addr) {
    block->isize = 0;
    block->done = 0;
-    block->size = 0;
-    block->block = NULL;
-    block->need_test = 0;
+    size_t sz = 8*sizeof(void*);
+    void* actual_p = (void*)AllocDynarecMap(sz);
+    void* p = actual_p + sizeof(void*);
+    if(actual_p==NULL) {
+        dynarec_log(LOG_INFO, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
+        CancelBlock(0);
+        return NULL;
+    }
+    block->size = sz;
+    block->actual_block = actual_p;
+    block->block = p;
+    block->jmpnext = p;
+    *(dynablock_t**)actual_p = block;
+    *(void**)(p+4*sizeof(void*)) = arm_epilog;
+    CreateJmpNext(block->jmpnext, p+4*sizeof(void*));
+    // all done...
+    __clear_cache(actual_p, actual_p+sz);   // need to clear the cache before execution...
    return block;
 }

 void* FillBlock(dynablock_t* block, uintptr_t addr) {
+    /*
+        A Block must have this layout:
+
+        0x0000..0x0003  : dynablock_t* : self
+        0x0004..4+4*n   : actual Native instructions, (n is the total number)
+        B ..    B+3     : dynablock_t* : self (as part of JmpNext, that simulate another block)
+        B+4 ..  B+15    : 3 Native code for jmpnext (or jmp epilog in case of empty block)
+        B+16 .. B+19    : jmpnext (or jmp_epilog) address
+        B+20 .. B+31    : empty (in case an architecture needs more than 3 opcodes)
+        B+32 .. B+32+sz : instsize (compressed array with each instruction lenght on x86 and native side)
+
+    */
 dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    if(IsInHotPage(addr)) {
        dynarec_log(LOG_DEBUG, "Cancelling dynarec FillBlock on hotpage for %p\n", (void*)addr);
@@ -395,10 +446,6 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
        dynarec_log(LOG_DEBUG, "Asked to fill a block in fobidden zone\n");
        return CreateEmptyBlock(block, addr);
    }
-    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;
-    }
    if(current_helper) {
        dynarec_log(LOG_DEBUG, "Cancelling dynarec FillBlock at %p as anothor one is going on\n", (void*)addr);
        return NULL;
@@ -412,23 +459,19 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    helper.start = addr;
    uintptr_t start = addr;
    helper.cap = 64; // needs epilog handling
-    helper.insts = (instruction_arm_t*)box_calloc(helper.cap, sizeof(instruction_arm_t));
+    helper.insts = (instruction_arm_t*)dynaCalloc(helper.cap, sizeof(instruction_arm_t));
    // pass 0, addresses, x86 jump addresses, overall size of the block
    uintptr_t end = arm_pass0(&helper, addr);
-    // no need for next anymore
-    box_free(helper.next);
-    helper.next_sz = helper.next_cap = 0;
-    helper.next = NULL;
    // basic checks
    if(!helper.size) {
        dynarec_log(LOG_DEBUG, "Warning, null-sized dynarec block (%p)\n", (void*)addr);
-        CancelBlock();
-        return (void*)block;
+        CancelBlock(0);
+        return CreateEmptyBlock(block, addr);
    }
    if(!isprotectedDB(addr, 1)) {
        dynarec_log(LOG_INFO, "Warning, write on current page on pass0, aborting dynablock creation (%p)\n", (void*)addr);
        AddHotPage(addr);
-        CancelBlock();
+        CancelBlock(0);
        return NULL;
    }
    // already protect the block and compute hash signature
@@ -437,24 +480,57 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
        protectDB(addr, end-addr);  //end is 1byte after actual end
    uint32_t hash = X31_hash_code((void*)addr, end-addr);
    // calculate barriers
-    for(int i=0; i<helper.size; ++i)
-        if(helper.insts[i].x86.jmp) {
-            uintptr_t j = helper.insts[i].x86.jmp;
-            if(j<start || j>=end || j==helper.insts[i].x86.addr) {
-                helper.insts[i].x86.jmp_insts = -1;
-                helper.insts[i].x86.need_after |= X_PEND;
-            } else {
-                // find jump address instruction
-                int k=-1;
-                for(int i2=0; i2<helper.size && k==-1; ++i2) {
-                    if(helper.insts[i2].x86.addr==j)
-                        k=i2;
+    for(int ii=0; ii<helper.jmp_sz; ++ii) {
+        int i = helper.jmps[ii];
+        uintptr_t j = helper.insts[i].x86.jmp;
+        if(j<start || j>=end || j==helper.insts[i].x86.addr) {
+            if(j==helper.insts[i].x86.addr) // if there is a loop on some opcode, make the block "always to tested"
+                helper.always_test = 1;
+            helper.insts[i].x86.jmp_insts = -1;
+            helper.insts[i].x86.need_after |= X_PEND;
+        } else {
+            // find jump address instruction
+            int k=-1;
+            int search = ((j>=helper.insts[0].x86.addr) && j<helper.insts[0].x86.addr+helper.isize)?1:0;
+            int imin = 0;
+            int imax = helper.size-1;
+            int i2 = helper.size/2;
+            // dichotomy search
+            while(search) {
+                if(helper.insts[i2].x86.addr == j) {
+                    k = i2;
+                    search = 0;
+                } else if(helper.insts[i2].x86.addr>j) {
+                    imax = i2;
+                    i2 = (imax+imin)/2;
+                } else {
+                    imin = i2;
+                    i2 = (imax+imin)/2;
+                }
+                if(search && (imax-imin)<2) {
+                    search = 0;
+                    if(helper.insts[imin].x86.addr==j)
+                        k = imin;
+                    else if(helper.insts[imax].x86.addr==j)
+                        k = imax;
                }
-                if(k!=-1 && !helper.insts[i].barrier_maybe)
-                    helper.insts[k].x86.barrier |= BARRIER_FULL;
-                helper.insts[i].x86.jmp_insts = k;
            }
+            /*for(int i2=0; i2<helper.size && k==-1; ++i2) {
+                if(helper.insts[i2].x86.addr==j)
+                    k=i2;
+            }*/
+            if(k!=-1 && !helper.insts[i].barrier_maybe)
+                helper.insts[k].x86.barrier |= BARRIER_FULL;
+            helper.insts[i].x86.jmp_insts = k;
        }
+    }
+    // no need for next and jmps anymore
+    dynaFree(helper.next);
+    helper.next_sz = helper.next_cap = 0;
+    helper.next = NULL;
+    dynaFree(helper.jmps);
+    helper.jmp_sz = helper.jmp_cap = 0;
+    helper.jmps = NULL;
    // fill predecessors with the jump address
    fillPredecessors(&helper);

@@ -468,29 +544,69 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    // 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(block, sz);
-    if(p==NULL) {
-        dynarec_log(LOG_DEBUG, "AllocDynarecMap(%p, %d) failed, cancelling block\n", block, sz);
-        CancelBlock();
+    size_t insts_rsize = (helper.insts_size+2)*sizeof(instsize_t);
+    insts_rsize = (insts_rsize+7)&~7;   // round the size...
+    size_t arm_size = (helper.arm_size+7)&~7;   // round the size...
+    // ok, now allocate mapped memory, with executable flag on
+    size_t sz = sizeof(void*) + arm_size + 8*sizeof(void*) + insts_rsize;
+    //           dynablock_t* block (arm insts) jmpnext code    instsize
+    void* actual_p = (void*)AllocDynarecMap(sz);
+    void* p = (void*)(((uintptr_t)actual_p) + sizeof(void*));
+    void* next = p + arm_size;
+    void* instsize = next + 8*sizeof(void*);
+    if(actual_p==NULL) {
+        dynarec_log(LOG_INFO, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
+        CancelBlock(0);
        return NULL;
    }
    helper.block = p;
+    block->actual_block = actual_p;
    helper.arm_start = (uintptr_t)p;
-    if(helper.sons_size) {
-        helper.sons_x86 = (uintptr_t*)alloca(helper.sons_size*sizeof(uintptr_t));
-        helper.sons_arm = (void**)alloca(helper.sons_size*sizeof(void*));
-    }
-    // pass 3, emit (log emit arm opcode)
+    helper.jmp_next = (uintptr_t)next+sizeof(void*);
+    helper.instsize = (instsize_t*)instsize;
+    *(dynablock_t**)actual_p = block;
+    // pass 3, emit (log emit native opcode)
    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":"");
    }
+    size_t oldarmsize = helper.arm_size;
+    size_t oldinstsize = helper.insts_size;
    helper.arm_size = 0;
+    helper.insts_size = 0;  // reset
    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);
+    // keep size of instructions for signal handling
+    block->instsize = instsize;
+    // ok, free the helper now
+    dynaFree(helper.insts);
+    helper.insts = NULL;
+    helper.instsize = NULL;
+    dynaFree(helper.predecessor);
+    helper.predecessor = NULL;
+    block->size = sz;
+    block->isize = helper.size;
+    block->block = p;
+    block->jmpnext = next+sizeof(void*);
+    block->always_test = helper.always_test;
+    block->dirty = block->always_test;
+    *(dynablock_t**)next = block;
+    *(void**)(next+5*sizeof(void*)) = arm_next;
+    CreateJmpNext(block->jmpnext, next+5*sizeof(void*));
+    //block->x86_addr = (void*)start;
+    block->x86_size = end-start;
+    // all done...
+    __clear_cache(actual_p, actual_p+sz);   // need to clear the cache before execution...
+    block->hash = X31_hash_code(block->x86_addr, block->x86_size);
+    // Check if something changed, to abbort if it as
+    if((block->hash != hash)) {
+        dynarec_log(LOG_DEBUG, "Warning, a block changed while being processed hash(%p:%zu)=%x/%x\n", block->x86_addr, block->x86_size, block->hash, hash);
+        AddHotPage(addr);
+        CancelBlock(0);
+        return NULL;
+    }
+    if((oldarmsize!=helper.arm_size)) {
+        printf_log(LOG_NONE, "BOx86: Warning, size difference in block between pass2 (%zu) & pass3 (%zu)!\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) {
@@ -500,76 +616,19 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
            printf_log(LOG_NONE, "\t%d -> %d\n", helper.insts[i].size2, helper.insts[i].size);
        }
        printf_log(LOG_NONE, " ------------\n");
-    }
-    // all done...
-    __clear_cache(p, p+sz);   // need to clear the cache before execution...
-    // keep size of instructions for signal handling
-    {
-        size_t cap = 1;
-        for(int i=0; i<helper.size; ++i)
-            cap += 1 + ((helper.insts[i].x86.size>helper.insts[i].size)?helper.insts[i].x86.size:helper.insts[i].size)/15;
-        size_t size = 0;
-        block->instsize = (instsize_t*)box_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
-    box_free(helper.insts);
-    helper.insts = NULL;
-    box_free(helper.next);
-    helper.next = NULL;
-    block->size = sz;
-    block->isize = helper.size;
-    block->block = p;
-    block->need_test = 0;
-    //block->x86_addr = (void*)start;
-    block->x86_size = end-start;
-    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);
-        CancelBlock();
-        AddHotPage(addr);
+        CancelBlock(0);
        return NULL;
    }
+    if(insts_rsize/sizeof(instsize_t)<helper.insts_size) {
+        printf_log(LOG_NONE, "BOX86: Warning, ists_size difference in block between pass2 (%zu) and pass3 (%zu), allocated: %zu\n", oldinstsize, helper.insts_size, insts_rsize/sizeof(instsize_t));
+    }
    if(!isprotectedDB(addr, end-addr)) {
-        dynarec_log(LOG_DEBUG, "Warning, block unprotected while beeing processed %p:%zd, cancelling\n", block->x86_addr, block->x86_size);
+        dynarec_log(LOG_DEBUG, "Warning, block unprotected while being processed %p:%zu, marking as need_test\n", block->x86_addr, block->x86_size);
        AddHotPage(addr);
-        block->need_test = 1;
+        block->dirty = 1;
        //protectDB(addr, end-addr);
    }
-    // fill sons if any
-    dynablock_t** sons = NULL;
-    int sons_size = 0;
-    if(helper.sons_size) {
-        sons = (dynablock_t**)box_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];
-                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
-            box_free(sons);
-    }
-    box_free(helper.predecessor);
-    helper.predecessor = NULL;
    current_helper = NULL;
-    block->done = 1;
+    //block->done = 1;
    return (void*)block;
 }
@@ -27,6 +27,10 @@
 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"

+int isRetX87Wrapper(wrapper_t fun);
+void emit_signal(x86emu_t* emu, int sig, void* addr, int code);
+void emit_div0(x86emu_t* emu, void* addr, int code);
+
 uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
 {
    uint8_t nextop, opcode;
@@ -1000,7 +1004,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("(LOCK)XCHG Eb, Gb");
            // Do the swap
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGB(x14);
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);
@@ -1027,7 +1031,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x87:
            INST_NAME("(LOCK)XCHG Ed, Gd");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGD;
                GETED;
                if(gd!=ed) {
@@ -1077,7 +1081,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            } else {
                gd = gb1;   // no need to extract
            }
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
                eb2 = ((ed&4)>>2);    // L or H
@@ -1092,7 +1096,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Ed, Gd");
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(xEAX+(nextop&7), gd);
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 4095, 0, 0, &lock);
@@ -1106,7 +1110,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            gd = (nextop&0x38)>>3;
            gb1 = xEAX+(gd&3);
            gb2 = ((gd&4)>>2);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                    wback = (nextop&7);
                    wb2 = (wback>>2);
                    wback = xEAX+(wback&3);
@@ -1117,10 +1121,6 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        ed = wback;
                    }
                } else {
-                    if(box86_dynarec_strongmem && 
-                     (dyn->insts[ninst].x86.barrier || !ninst || (box86_dynarec_strongmem>1) || (ninst && dyn->insts[ninst-1].x86.barrier))) {
-                        SMDMB();
-                    }
                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 4095, 0, 0, &lock);
                    SMREADLOCK(lock);
                    LDRB_IMM9(x14, wback, fixedaddress);
@@ -1132,7 +1132,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Gd, Ed");
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(gd, xEAX+(nextop&7));
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 4095, 0, 0, &lock);
@@ -1144,7 +1144,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Ed, Seg");
            nextop=F8;
            MOV32(x3, offsetof(x86emu_t, segs[(nextop&0x38)>>3]));
-            if((nextop&0xC0)==0xC0) {   // reg <= seg
+            if(MODREG) {   // reg <= seg
                LDRH_REG(xEAX+(nextop&7), xEmu, x3);
            } else {                    // mem <= seg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
@@ -1157,7 +1157,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("LEA Gd, Ed");
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg? that's an invalid operation
+            if(MODREG) {   // reg <= reg? that's an invalid operation
                DEFAULT;
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, gd, &fixedaddress, 0, 0, 0, NULL);
@@ -1169,7 +1169,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x8E:
            INST_NAME("MOV Seg,Ew");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
            } else {
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 255, 0, 0, NULL);
@@ -1189,7 +1189,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x8F:
            INST_NAME("POP Ed");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                POP1(xEAX+(nextop&7));
            } else {
                POP1(x2); // so this can handle POP [ESP] and maybe some variant too
@@ -1245,6 +1245,12 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            AND_REG_LSL_IMM5(xFlags, xFlags, x1, 0);
            ORR_IMM8(xFlags, xFlags, 2, 0);
            SET_DFNONE(x1);
+            if(box86_wine || 1) {    // should this be done all the time?
+                TSTS_IMM8_ROR(xFlags, 0x1, 12); // 0x100 == F_TF
+                B_NEXT(cEQ);
+                // go to epilog, TF should trigger at end of next opcode, so using Interpretor only
+                jump_to_epilog(dyn, addr, 0, ninst);
+            }
            break;
        case 0x9E:
            INST_NAME("SAHF");
@@ -1264,6 +1270,8 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV AL, Ob");
            u32 = F32;
            MOV32(x2, u32);
+            if(isLockAddress(u32)) lock=1; else lock = 0;
+            SMREADLOCK(lock);
            LDRB_IMM9(x2, x2, 0);
            BFI(xEAX, x2, 0, 8);
            break;
@@ -1271,33 +1279,41 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV EAX, Od");
            u32 = F32;
            MOV32(x2, u32);
+            if(isLockAddress(u32)) lock=1; else lock = 0;
+            SMREADLOCK(lock);            
            LDR_IMM9(xEAX, x2, 0);
            break;
        case 0xA2:
            INST_NAME("MOV Ob, AL");
            u32 = F32;
            MOV32(x2, u32);
+            if(isLockAddress(u32)) lock=1; else lock = 0;
            STRB_IMM9(xEAX, x2, 0);
-            SMWRITE();
+            SMWRITELOCK(lock);
            break;
        case 0xA3:
            INST_NAME("MOV Od, EAX");
            u32 = F32;
            MOV32(x2, u32);
+            if(isLockAddress(u32)) lock=1; else lock = 0;
            STR_IMM9(xEAX, x2, 0);
-            SMWRITE();
+            SMWRITELOCK(lock);
            break;
        case 0xA4:
            INST_NAME("MOVSB");
            GETDIR(x3, 1);
+            SMREAD();
            LDRBAI_REG_LSL_IMM5(x1, xESI, x3, 0);
            STRBAI_REG_LSL_IMM5(x1, xEDI, x3, 0);
+            SMWRITE();
            break;
        case 0xA5:
            INST_NAME("MOVSD");
            GETDIR(x3, 4);
+            SMREAD();
            LDRAI_REG_LSL_IMM5(x1, xESI, x3, 0);
            STRAI_REG_LSL_IMM5(x1, xEDI, x3, 0);
+            SMWRITE();
            break;
        case 0xA6:
            INST_NAME("CMPSB");
@@ -1334,11 +1350,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("STOSB");
            GETDIR(x3, 1);
            STRBAI_REG_LSL_IMM5(xEAX, xEDI, x3, 0);
+            SMWRITE();
            break;
        case 0xAB:
            INST_NAME("STOSD");
            GETDIR(x3, 4);
            STRAI_REG_LSL_IMM5(xEAX, xEDI, x3, 0);
+            SMWRITE();
            break;
        case 0xAC:
            INST_NAME("LODSB");
@@ -1444,49 +1462,42 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 4:
                case 6:
                    INST_NAME("SHL Eb, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETEB(x1);
-                    u8 = (F8)&0x1f;
-                    UFLAG_IF{
-                        MOV32(x14, u8); UFLAG_OP2(x14)
-                    };
-                    UFLAG_OP1(ed);
-                    MOV_REG_LSL_IMM5(ed, ed, u8);
-                    EBBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shl8);
+                    if(geted_ib(dyn, addr, ninst, nextop)&0x1f) {
+                        SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                        GETEB(x1);
+                        u8 = (F8)&0x1f;
+                        emit_shl8c(dyn, ninst, ed, u8, x2, x14);
+                        EBBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
                case 5:
                    INST_NAME("SHR Eb, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETEB(x1);
-                    u8 = (F8)&0x1f;
-                    UFLAG_IF{
-                        MOV32(x14, u8); UFLAG_OP2(x14)
-                    };
-                    UFLAG_OP1(ed);
-                    if(u8) {
-                        MOV_REG_LSR_IMM5(ed, ed, u8);
+                    if(geted_ib(dyn, addr, ninst, nextop)&0x1f) {
+                        SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                        GETEB(x1);
+                        u8 = (F8)&0x1f;
+                        emit_shr8c(dyn, ninst, ed, u8, x2, x14);
                        EBBACK;
+                    } else {
+                        FAKEED;
+                        F8;
                    }
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shr8);
                    break;
                case 7:
                    INST_NAME("SAR Eb, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETSEB(x1);
-                    u8 = (F8)&0x1f;
-                    UFLAG_IF{
-                        MOV32(x14, u8); UFLAG_OP2(x14)
-                    };
-                    UFLAG_OP1(ed);
-                    if(u8) {
-                        MOV_REG_ASR_IMM5(ed, ed, u8);
+                    if(geted_ib(dyn, addr, ninst, nextop)&0x1f) {
+                        SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                        GETSEB(x1);
+                        u8 = (F8)&0x1f;
+                        emit_sar8c(dyn, ninst, ed, u8, x2, x14);
                        EBBACK;
+                    } else {
+                        FAKEED;
+                        F8;
                    }
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_sar8);
                    break;
            }
            break;
@@ -1617,7 +1628,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC6:
            INST_NAME("MOV Eb, Ib");
            nextop=F8;
-            if((nextop&0xC0)==0xC0) {   // reg <= u8
+            if(MODREG) {   // reg <= u8
                u8 = F8;
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
@@ -1635,7 +1646,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC7:
            INST_NAME("MOV Ed, Id");
            nextop=F8;
-            if((nextop&0xC0)==0xC0) {   // reg <= i32
+            if(MODREG) {   // reg <= i32
                i32 = F32S;
                ed = xEAX+(nextop&7);
                MOV32(ed, i32);
@@ -1663,8 +1674,10 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                if((PK(0)==0) && (PK(1)==0) && (PK(2)==0) && (PK(3)==0))
                {
                    addr+=4;
+                    SKIPTEST(x14);
                    MESSAGE(LOG_DEBUG, "Exit x86 Emu\n");
                    MOV32(x14, ip+1+2);
+                    SMEND();
                    STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
                    MOVW(x1, 1);
                    STR_IMM9(x1, xEmu, offsetof(x86emu_t, quit));
@@ -1672,6 +1685,8 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    *need_epilog = 1;
                } else {
                    MESSAGE(LOG_DUMP, "Native Call to %s\n", GetNativeName(GetNativeFnc(ip)));
+                    SKIPTEST(x14);
+                    SMEND();
                    x87_forget(dyn, ninst, x3, x14, 0);
                    if((box86_log<2) && !cycle_log) {   // call the wrapper directly
                        uintptr_t ncall[2]; // to avoid BUSERROR!!!
@@ -1707,18 +1722,24 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                MOV32(x2, offsetof(box86context_t, signals[SIGTRAP]));
                LDR_REG_LSL_IMM5(x3, x1, x2, 0);
                CMPS_IMM8(x3, 0);
-                B_NEXT(cNE);
+                B_NEXT(cEQ);
+                STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
+                MOV32(xEIP, addr);
+                STR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
                MOV32(x1, SIGTRAP);
-                CALL_(raise, -1, 0);
+                MOV32(x2, addr);
+                MOV32(x3, 128);
+                CALL(emit_signal, -1, 0);
                break;
            }
            break;
        case 0xCD:
            SETFLAGS(X_ALL, SF_SET);    // Hack, set all flags (to an unknown state...)
+            SKIPTEST(x14);
            SMEND();
-            if(PK(0)==0x80) {
+            u8 = F8;
+            if(u8==0x80) {
                INST_NAME("Syscall");
-                u8 = F8;
                BARRIER(BARRIER_FLOAT);
                MOV32(xEIP, ip+2);
                STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
@@ -1734,7 +1755,11 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                jump_to_epilog(dyn, 0, xEIP, ninst);
            } else {
                INST_NAME("INT Ib");
-                DEFAULT;
+                if(box86_wine && u8==0x2D && 0) {
+                    MESSAGE(LOG_DEBUG, "Hack for wine/int 2d\n");
+                } else {
+                    DEFAULT;
+                }
            }
            break;

@@ -1747,115 +1772,68 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            *ok = 0;
            break;
        case 0xD0:
-        case 0xD2:
            nextop = F8;
            switch((nextop>>3)&7) {
                case 0:
-                    if(opcode==0xD0) {
-                        INST_NAME("ROL Eb, 1");
-                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
-                        GETEB(x1);
-                        emit_rol8c(dyn, ninst, ed, 1, x2, x14);
-                        EBBACK;
-                    } else {
-                        INST_NAME("ROL Eb, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                        MESSAGE(LOG_DUMP, "Need Optimization ROL 8b\n");
-                        SETFLAGS(X_OF|X_CF, SF_SET);
-                        GETEB(x1);
-                        CALL_(rol8, x1, (1<<x3));
-                        EBBACK;
-                    }
+                    INST_NAME("ROL Eb, 1");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                    GETEB(x1);
+                    emit_rol8c(dyn, ninst, ed, 1, x2, x14);
+                    EBBACK;
                    break;
                case 1:
-                    if(opcode==0xD0) {
-                        INST_NAME("ROR Eb, 1");
-                        MOVW(x2, 1);
-                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
-                        GETEB(x1);
-                        emit_ror8c(dyn, ninst, ed, 1, x2, x14);
-                        EBBACK;
-                    } else {
-                        INST_NAME("ROR Eb, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                        SETFLAGS(X_OF|X_CF, SF_SET);
-                        MESSAGE(LOG_DUMP, "Need Optimization ROR 8b\n");
-                        GETEB(x1);
-                        CALL_(ror8, x1, (1<<x3));
-                        EBBACK;
-                    }
+                    INST_NAME("ROR Eb, 1");
+                    MOVW(x2, 1);
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                    GETEB(x1);
+                    emit_ror8c(dyn, ninst, ed, 1, x2, x14);
+                    EBBACK;
                    break;
                case 2:
-                    if(opcode==0xD0) {INST_NAME("RCL Eb, 1");} else {INST_NAME("RCL Eb, CL");}
+                    INST_NAME("RCL Eb, 1");
                    MESSAGE(LOG_DUMP, "Need Optimization RCL 8b\n");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD0) {MOVW(x2, 1);} else {AND_IMM8(x2, xECX, 0x1f);}
+                    MOVW(x2, 1);
                    GETEB(x1);
                    CALL_(rcl8, x1, (1<<x3));
                    EBBACK;
                    break;
                case 3:
-                    if(opcode==0xD0) {INST_NAME("RCR Eb, 1");} else {INST_NAME("RCR Eb, CL");}
+                    INST_NAME("RCR Eb, 1");
                    MESSAGE(LOG_DUMP, "Need Optimization RCR 8b\n");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD0) {MOVW(x2, 1);} else {AND_IMM8(x2, xECX, 0x1f);}
+                    MOVW(x2, 1);
                    GETEB(x1);
                    CALL_(rcr8, x1, (1<<x3));
                    EBBACK;
                    break;
                case 4:
                case 6:
-                    if(opcode==0xD0) {
-                        INST_NAME("SHL Eb, 1");
-                        MOVW(x2, 1);
-                    } else {
-                        INST_NAME("SHL Eb, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SHL Eb, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETEB(x1);
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_LSL_REG(ed, ed, x2);
+                    emit_shl8c(dyn, ninst, ed, 1, x2, x14);
                    EBBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shl8);
                    break;
                case 5:
-                    if(opcode==0xD0) {
-                        INST_NAME("SHR Eb, 1");
-                        MOVW(x2, 1);
-                    } else {
-                        INST_NAME("SHR Eb, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SHR Eb, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETEB(x1);
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_LSR_REG(ed, ed, x2);
+                    emit_shr8c(dyn, ninst, ed, 1, x2, x14);
                    EBBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shr8);
                    break;
                case 7:
-                    if(opcode==0xD0) {
-                        INST_NAME("SAR Eb, 1");
-                        MOVW(x2, 1);
-                    } else {
-                        INST_NAME("SAR Eb, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SAR Eb, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETSEB(x1);
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_ASR_REG(ed, ed, x2);
+                    emit_sar8c(dyn, ninst, ed, 1, x2, x14);
                    EBBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_sar8);
                    break;
            }
            break;
+
        case 0xD1:
            nextop = F8;
            switch((nextop>>3)&7) {
@@ -1936,6 +1914,125 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
            }
            break;
+        case 0xD2:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("ROL Eb, CL");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        TSTS_IMM8(xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    }
+                    AND_IMM8(x2, xECX, 0x7);
+                    RSB_IMM8(x2, x2, 8);
+                    GETEB(x1);
+                    BFI(ed, ed, 8, 8);
+                    MOV_REG_LSR_REG(ed, ed, x2);
+                    EBBACK;
+                    UFLAG_IF {  // calculate flags directly
+                        CMPS_IMM8(x2, 7);
+                            MOV_REG_LSR_IMM5_COND(cEQ, x3, ed, 7);
+                            ADD_REG_LSL_IMM5_COND(cEQ, x3, x3, ed, 0);
+                            BFI_COND(cEQ, xFlags, x3, F_OF, 1);
+                        BFI(xFlags, ed, F_CF, 1);
+                        UFLAG_DF(x2, d_none);
+                    }
+                    break;
+                case 1:
+                    INST_NAME("ROR Eb, CL");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        TSTS_IMM8(xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    }
+                    AND_IMM8(x2, xECX, 0x7);
+                    GETEB(x1);
+                    BFI(ed, ed, 8, 8);
+                    MOV_REG_LSR_REG(ed, ed, x2);
+                    EBBACK;
+                    UFLAG_IF {  // calculate flags directly
+                        CMPS_IMM8(x2, 1);
+                            MOV_REG_LSR_IMM5_COND(cEQ, x2, ed, 6); // x2 = d>>30
+                            XOR_REG_LSR_IMM8_COND(cEQ, x2, x2, x2, 1); // x2 = ((d>>30) ^ ((d>>30)>>1))
+                            BFI_COND(cEQ, xFlags, x2, F_OF, 1);
+                        MOV_REG_LSR_IMM5(x2, ed, 7);
+                        BFI(xFlags, x2, F_CF, 1);
+                        UFLAG_DF(x2, d_none);
+                    }
+                    break;
+                case 2:
+                    INST_NAME("RCL Eb, CL");
+                    MESSAGE(LOG_DUMP, "Need Optimization RCL 8b\n");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    AND_IMM8(x2, xECX, 0x1f);
+                    GETEB(x1);
+                    CALL_(rcl8, x1, (1<<x3));
+                    EBBACK;
+                    break;
+                case 3:
+                    INST_NAME("RCR Eb, CL");
+                    MESSAGE(LOG_DUMP, "Need Optimization RCR 8b\n");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    AND_IMM8(x2, xECX, 0x1f);
+                    GETEB(x1);
+                    CALL_(rcr8, x1, (1<<x3));
+                    EBBACK;
+                    break;
+                case 4:
+                case 6:
+                    INST_NAME("SHL Eb, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETEB(x1);
+                    emit_shl8(dyn, ninst, ed, x2, x2, x14);
+                    EBBACK;
+                    break;
+                case 5:
+                    INST_NAME("SHR Eb, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETEB(x1);
+                    emit_shr8(dyn, ninst, ed, x2, x2, x14);
+                    EBBACK;
+                    break;
+                case 7:
+                    INST_NAME("SAR Eb, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETSEB(x1);
+                    emit_sar8(dyn, ninst, ed, x2, x2, x14);
+                    EBBACK;
+                    break;
+            }
+            break;
        case 0xD3:
            nextop = F8;
            switch((nextop>>3)&7) {
@@ -2158,8 +2255,8 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                tmp = dyn->insts[ninst].pass2choice = 1;
            else if ((getProtection(u32)&PROT_READ) && (PKa(u32+0)==0x8B) && (((PKa(u32+1))&0xC7)==0x04) && (PKa(u32+2)==0x24) && (PKa(u32+3)==0xC3))
                tmp = dyn->insts[ninst].pass2choice = 2;
-            /*else if(isNativeCall(dyn, u32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
-                tmp = dyn->insts[ninst].pass2choice = 3;*/
+            else if(isNativeCall(dyn, u32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
+                tmp = dyn->insts[ninst].pass2choice = 3;
            else 
                tmp = dyn->insts[ninst].pass2choice = 0;
            #else
@@ -2169,26 +2266,31 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 1:
                    //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    MESSAGE(LOG_DUMP, "Hack for Call 0\n");
+                    SKIPTEST(x14);
                    MOV32(xEIP, addr);
                    PUSH1(xEIP);
                    break;
                case 2:
                    //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    MESSAGE(LOG_DUMP, "Hack for Call x86.get_pc_thunk.reg\n");
+                    SKIPTEST(x14);
                    u8 = PK(i32+1);
                    gd = xEAX+((u8&0x38)>>3);
                    MOV32(gd, addr);
                    break;
-                // disabling this to avoid fetching data outside current block (in case this part changed, this block will not been marck as dirty)
-                /*case 3:
+                case 3:
                    SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    BARRIER(BARRIER_FULL);
                    BARRIER_NEXT(BARRIER_FULL);
                    MOV32(x2, addr);
                    PUSH1(x2);
                    MESSAGE(LOG_DUMP, "Native Call to %s (retn=%d)\n", GetNativeName(GetNativeFnc(dyn->insts[ninst].natcall-1)), dyn->insts[ninst].retn);
+                    SKIPTEST(x14);
                    // calling a native function
-                    x87_forget(dyn, ninst, x3, x14, 0);
+                    if(isRetX87Wrapper(*(wrapper_t*)(dyn->insts[ninst].natcall+2))) {
+                        // return value will be on the stack, so the stack depth needs to be updated
+                        x87_purgecache(dyn, ninst, 0, x3, x14, x1);
+                    }
                    if((box86_log<2) && !cycle_log && dyn->insts[ninst].natcall) {   // call the wrapper directly
                        uintptr_t ncall[2];
                        memcpy(ncall,  (void*)(dyn->insts[ninst].natcall+2), 2*sizeof(void*));   // the wrapper + function
@@ -2219,16 +2321,22 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    B_NEXT(cNE);    // not quitting, so lets continue
                    MARK;
                    jump_to_epilog(dyn, 0, xEIP, ninst);
-                    break;*/
+                    break;
                default:
                    if((box86_dynarec_safeflags>1) || (ninst && dyn->insts[ninst-1].x86.set_flags)) {
                        READFLAGS(X_PEND);  // that's suspicious
                    } else {
-                        //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
+                        SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    }
                    // regular call
-                    BARRIER(BARRIER_FLOAT);
                    //BARRIER_NEXT(BARRIER_FULL);
+                    if(/*box86_dynarec_callret &&*/ box86_dynarec_bigblock>1) {
+                        BARRIER(BARRIER_FULL);
+                    } else {
+                        BARRIER(BARRIER_FLOAT);
+                        *need_epilog = 0;
+                        *ok = 0;
+                    }
                    *need_epilog = 0;
                    *ok = 0;
                    MOV32(x2, addr);
@@ -2252,11 +2360,11 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                INST_NAME("JMP Ib");
                i32 = F8S;
            }
-            JUMP(addr+i32, 0);
+            JUMP((uintptr_t)getAlternate((void*)(addr+i32)), 0);
            if(dyn->insts[ninst].x86.jmp_insts==-1) {
                // out of the block
                fpu_purgecache(dyn, ninst, 1, x1, x2, x3);
-                jump_to_next(dyn, addr+i32, 0, ninst);
+                jump_to_next(dyn, (uintptr_t)getAlternate((void*)(addr+i32)), 0, ninst);
            } else {
                // inside the block
                CacheTransform(dyn, ninst, CHECK_CACHE(), x1, x2, x3);
@@ -2271,9 +2379,20 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            *ok = 0;
            break;
        
-        case 0xEE:
-            INST_NAME("OUT dx, al");
-            //ignored!
+        case 0xEC:                      /* IN AL, DX */
+        case 0xED:                      /* IN EAX, DX */
+        case 0xEE:                      /* OUT DX, AL */
+        case 0xEF:                      /* OUT DX, EAX */
+            INST_NAME(opcode==0xEC?"IN AL, DX":(opcode==0xED?"IN EAX, DX":(opcode==0xEE?"OUT DX? AL":"OUT DX, EAX")));
+            SETFLAGS(X_ALL, SF_SET);    // Hack to set flags in "don't care" state
+            STM(xEmu, (1<<xEAX)|(1<<xECX)|(1<<xEDX)|(1<<xEBX)|(1<<xESP)|(1<<xEBP)|(1<<xESI)|(1<<xEDI)|(1<<xFlags));
+            MOV32(xEIP, ip);
+            STR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+            CALL(arm_priv, -1, 0);
+            LDR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+            jump_to_epilog(dyn, 0, xEIP, ninst);
+            *need_epilog = 0;
+            *ok = 0;
            break;

        case 0xF0:
@@ -2297,11 +2416,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        TSTS_REG_LSL_IMM5(xECX, xECX, 0);
                        B_NEXT(cEQ);    // end of loop
                        GETDIR(x3, 2);
+                        SMREAD();
                        MARK;
                        LDRHAI_REG_LSL_IMM5(x1, xESI, x3);
                        STRHAI_REG_LSL_IMM5(x1, xEDI, x3);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        break;
                    case 0xAB:
                        INST_NAME("REP STOSW");
@@ -2312,6 +2433,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        STRHAI_REG_LSL_IMM5(xEAX, xEDI, x3);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        break;
                    case 0xAF:
                        if(opcode==0xF2) {INST_NAME("REPNZ SCASW");} else {INST_NAME("REPZ SCASW");}
@@ -2390,11 +2512,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        TSTS_REG_LSL_IMM5(xECX, xECX, 0);
                        B_NEXT(cEQ);    // end of loop
                        GETDIR(x3,1);
+                        SMREAD();
                        MARK;
                        LDRBAI_REG_LSL_IMM5(x1, xESI, x3, 0);
                        STRBAI_REG_LSL_IMM5(x1, xEDI, x3, 0);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        // done
                        break;
                    case 0xA5:
@@ -2402,11 +2526,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        TSTS_REG_LSL_IMM5(xECX, xECX, 0);
                        B_NEXT(cEQ);    // end of loop
                        GETDIR(x3,4);
+                        SMREAD();
                        MARK;
                        LDRAI_REG_LSL_IMM5(x1, xESI, x3, 0);
                        STRAI_REG_LSL_IMM5(x1, xEDI, x3, 0);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        // done
                        break;
                    case 0xA6:
@@ -2468,6 +2594,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        STRBAI_REG_LSL_IMM5(xEAX, xEDI, x3, 0);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        break;
                    case 0xAB:
                        INST_NAME("REP STOSD");
@@ -2478,6 +2605,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        STRAI_REG_LSL_IMM5(xEAX, xEDI, x3, 0);
                        SUBS_IMM8(xECX, xECX, 1);
                        B_MARK(cNE);
+                        SMWRITE();
                        break;
                    case 0xAC:
                        INST_NAME("REP LODSB");
@@ -2576,7 +2704,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2:
                    INST_NAME("NOT Eb");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        wback = (nextop&7);
                        wb2 = (wback>>2);
                        wback = xEAX+(wback&3);
@@ -2597,12 +2725,12 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 4:
                    INST_NAME("MUL AL, Ed");
                    SETFLAGS(X_ALL, SF_PENDING);
-                    UFLAG_DF(x2, d_mul8);
                    GETEB(x1);
                    UBFX(x2, xEAX, 0, 8);
                    MUL(x1, x1, x2);
                    BFI(xEAX, x1, 0, 16);
                    UFLAG_RES(x1);
+                    UFLAG_DF(x2, d_mul8);
                    break;
                case 5:
                    INST_NAME("IMUL AL, Eb");
@@ -2661,52 +2789,66 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 4:
                    INST_NAME("MUL EAX, Ed");
                    SETFLAGS(X_ALL, SF_PENDING);
-                    UFLAG_DF(x2, d_mul32);
                    GETED;
                    UMULL(xEDX, xEAX, xEAX, ed);
                    UFLAG_RES(xEAX);
                    UFLAG_OP1(xEDX);
+                    UFLAG_DF(x2, d_mul32);
                    break;
                case 5:
                    INST_NAME("IMUL EAX, Ed");
                    SETFLAGS(X_ALL, SF_PENDING);
-                    UFLAG_DF(x2, d_imul32);
                    GETED;
                    SMULL(xEDX, xEAX, xEAX, ed);
                    UFLAG_RES(xEAX);
                    UFLAG_OP1(xEDX);
+                    UFLAG_DF(x2, d_imul32);
                    break;
                case 6:
                    INST_NAME("DIV Ed");
                    SETFLAGS(X_ALL, SF_SET);
-                    if(arm_div && ninst && dyn->insts 
+                    if(ninst && dyn->insts && (nextop==0xF0)
                       && dyn->insts[ninst-1].x86.addr 
-                       && (*(uint8_t*)(dyn->insts[ninst-1].x86.addr)==0x31 
-                        || *(uint8_t*)(dyn->insts[ninst-1].x86.addr)==0x33)
-                       && *(uint8_t*)(dyn->insts[ninst-1].x86.addr+1)==0xD2) {
-                        // previous instruction is XOR EDX, EDX, so its 32bits / 32bits
-                        SET_DFNONE(x2); // flags are undefined
-                        GETED;
-                        UDIV(x2, xEAX, ed); // x1 = xEAX / ed
-                        MLS(xEDX, x2, ed, xEAX);  // x14 = xEAX mod ed (i.e. xEAX - x1*ed)
-                        MOV_REG(xEAX, x2);
+                       && *(uint8_t*)(dyn->insts[ninst-1].x86.addr)==0xB8 
+                       && *(uint32_t*)(dyn->insts[ninst-1].x86.addr+1)==0) {
+                        // hack for some protection that check a divide by zero actualy trigger a div0 exception
+                        MESSAGE(LOG_INFO, "Divide by 0 hack\n");
+                        STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
+                        MOV32(xEIP, addr);
+                        STR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+                        MOV_REG(x1, xEIP);
+                        MOV32(x2, 0);
+                        CALL(emit_div0, -1, 0);
                    } else {
-                        GETEDH(x1);
-                        if(arm_div) {
-                            CMPS_IMM8(xEDX, 0);
-                            B_MARK(cEQ);
-                        }
-                        if(ed!=x1) {MOV_REG(x1, ed);}
-                        STM(xEmu, (1<<xEAX) | (1<<xECX) | (1<<xEDX));
-                        CALL(div32, -1, 0);
-                        LDM(xEmu, (1<<xEAX) | (1<<xECX) | (1<<xEDX));
-                        if(arm_div) {
-                            B_NEXT(c__);
-                            MARK;
-                            UDIV(x2, xEAX, ed); // x2 = xEAX / ed
-                            MLS(xEDX, x2, ed, xEAX);  // x14 = xEAX mod ed (i.e. xEAX - x2*ed)
-                            MOV_REG(xEAX, x2);
+                        if(arm_div && ninst && dyn->insts 
+                        && dyn->insts[ninst-1].x86.addr 
+                        && (*(uint8_t*)(dyn->insts[ninst-1].x86.addr)==0x31 
+                            || *(uint8_t*)(dyn->insts[ninst-1].x86.addr)==0x33)
+                        && *(uint8_t*)(dyn->insts[ninst-1].x86.addr+1)==0xD2) {
+                            // previous instruction is XOR EDX, EDX, so its 32bits / 32bits
                            SET_DFNONE(x2); // flags are undefined
+                            GETED;
+                            UDIV(x2, xEAX, ed); // x1 = xEAX / ed
+                            MLS(xEDX, x2, ed, xEAX);  // x14 = xEAX mod ed (i.e. xEAX - x1*ed)
+                            MOV_REG(xEAX, x2);
+                        } else {
+                            GETEDH(x1);
+                            if(arm_div) {
+                                CMPS_IMM8(xEDX, 0);
+                                B_MARK(cEQ);
+                            }
+                            if(ed!=x1) {MOV_REG(x1, ed);}
+                            STM(xEmu, (1<<xEAX) | (1<<xECX) | (1<<xEDX));
+                            CALL(div32, -1, 0);
+                            LDM(xEmu, (1<<xEAX) | (1<<xECX) | (1<<xEDX));
+                            if(arm_div) {
+                                B_NEXT(c__);
+                                MARK;
+                                UDIV(x2, xEAX, ed); // x2 = xEAX / ed
+                                MLS(xEDX, x2, ed, xEAX);  // x14 = xEAX mod ed (i.e. xEAX - x2*ed)
+                                MOV_REG(xEAX, x2);
+                                SET_DFNONE(x2); // flags are undefined
+                            }
                        }
                    }
                    break;
@@ -2764,13 +2906,17 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            ORRS_IMM8(xFlags, xFlags, 1, 0);
            break;
        case 0xFA:
-            INST_NAME("CLI");
-            BFC(xFlags, F_IF, 1);
-            break;
        case 0xFB:
-            INST_NAME("STI");
-            MOVW(x1, 1);
-            BFI(xFlags, x1, F_IF, 1);
+            INST_NAME(opcode==0xFA?"STI":"CLI");
+            SETFLAGS(X_ALL, SF_SET);    // Hack to set flags in "don't care" state
+            STM(xEmu, (1<<xEAX)|(1<<xECX)|(1<<xEDX)|(1<<xEBX)|(1<<xESP)|(1<<xEBP)|(1<<xESI)|(1<<xEDI)|(1<<xFlags));
+            MOV32(xEIP, ip);
+            STR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+            CALL(arm_priv, -1, 0);
+            LDR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+            jump_to_epilog(dyn, 0, xEIP, ninst);
+            *need_epilog = 0;
+            *ok = 0;
            break;
        case 0xFC:
            INST_NAME("CLD");
@@ -2823,10 +2969,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2: // CALL Ed
                    INST_NAME("CALL Ed");
-                    PASS2IF((box86_dynarec_safeflags>1) || (ninst && dyn->insts[ninst-1].x86.set_flags), 1) {
+                    PASS2IF((box86_dynarec_safeflags>1) ||
+                        ((ninst && dyn->insts[ninst-1].x86.set_flags)
+                        || ((ninst>1) && dyn->insts[ninst-2].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
+                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
                    }
                    GETEDH(xEIP);
                    BARRIER(BARRIER_FLOAT);
@@ -28,7 +28,7 @@
    gd = (nextop&0x38)>>3;  \
    a = sse_get_reg(dyn, ninst, x1, gd, w)
 #define GETEX(a, w)    \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        a = sse_get_reg(dyn, ninst, x1, nextop&7, w); \
    } else {    \
        SMREAD(); \
@@ -40,7 +40,7 @@
    gd = (nextop&0x38)>>3;  \
    a = mmx_get_reg(dyn, ninst, x1, x2, x3, gd)
 #define GETEM(a)    \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        a = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7); \
    } else {    \
        SMREAD(); \
@@ -90,6 +90,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("UD2");
            SETFLAGS(X_ALL, SF_SET);    // Hack to set flags in "don't care" state
            //CALL(arm_ud, -1, 0);
+            SKIPTEST(x14);
            UDF(0);
            break;

@@ -110,7 +111,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVUPS Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -132,7 +133,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -149,7 +150,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x12:
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                INST_NAME("MOVHLPS Gx,Ex");
                GETGX(v0, 1);
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
@@ -168,7 +169,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            INST_NAME("MOVLPS Ex,Gx");
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                VMOVD(v1, v0);
            } else {
@@ -191,7 +192,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETEX(q0, 0);
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = fpu_get_scratch_quad(dyn);
                VMOVQ(q1, q0);
            } else q1 = q0;
@@ -202,7 +203,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETEX(q0, 0);
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = fpu_get_scratch_quad(dyn);
                VMOVQ(q1, q0);
            } else q1 = q0;
@@ -211,7 +212,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x16:
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                INST_NAME("MOVLHPS Gx,Ex");
                GETGX(v0, 1);
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
@@ -228,7 +229,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            INST_NAME("MOVHPS Ex,Gx");
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                VMOVD(v1, v0+1);
            } else {
@@ -248,7 +249,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x18:
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                INST_NAME("NOP (multibyte)");
            } else
            switch((nextop>>3)&7) {
@@ -281,7 +282,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVAPS Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -297,7 +298,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -311,7 +312,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
            } else {
                SMREAD();
@@ -326,7 +327,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -339,7 +340,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, gd);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -354,7 +355,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, gd);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -387,7 +388,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET);
            nextop = F8;
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                if(v1<16)
                    d0 = v1;
@@ -421,6 +422,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x31:
            INST_NAME("RDTSC");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
+            SKIPTEST(x14);
            CALL(ReadTSC, xEAX, 0);   // will return the u64 in x1:xEAX
            MOV_REG(xEDX, x1);
            break;
@@ -464,7 +466,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("MOVBE Gd, Ed");
                    nextop=F8;
                    GETGD;
-                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                    if(MODREG) {   // reg <= reg
                        REV(gd, xEAX+(nextop&7));
                    } else {                    // mem <= reg
                        SMREAD();
@@ -477,12 +479,12 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("MOVBE Ed, Gd");
                    nextop=F8;
                    GETGD;
-                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                    if(MODREG) {   // reg <= reg
                        REV(xEAX+(nextop&7), gd);
                    } else {                    // mem <= reg
                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 4095, 0, 0, NULL);
                        REV(x1, gd);
-                        STR_IMM9(gd, ed, fixedaddress);
+                        STR_IMM9(x1, ed, fixedaddress);
                        SMWRITE();
                    }
                    break;
@@ -521,7 +523,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            GETFLAGS;   \
            nextop=F8;  \
            GETGD;      \
-            if((nextop&0xC0)==0xC0) {   \
+            if(MODREG) {   \
                ed = xEAX+(nextop&7);   \
                MOV_REG_COND(YES, gd, ed); \
            } else { \
@@ -623,7 +625,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGD;
            XOR_REG_LSL_IMM5(gd, gd, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                // EX is an xmm reg
                GETEX(q0, 0);
                VMOVfrV_D(x1, x2, q0);
@@ -656,10 +658,13 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x51:
            INST_NAME("SQRTPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            #if 0
+            // the approximate is not precise enough and and some incorrect behavior
            v2 = fpu_get_scratch_quad(dyn);
            v1 = fpu_get_scratch_quad(dyn);
            // Newton-Raphson: Xn+1 = Xn*(3-d*Xn²)/2 converge to 1/sqrt(d) with X0 = result of VSQRTE
@@ -692,22 +697,72 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            } else {
                VMULQ_F32(v0, v2, q0);  // v0 = X2*d ~ SQRT(d)
            }
+            #else
+            if((q0<15) && (v0<15)) {
+                d0 = q0;
+                d1 = v0;
+            } else {
+                s0 = fpu_get_scratch_quad(dyn);
+                // source ok or not
+                if(q0<15) {
+                    d0 = q0;
+                } else {
+                    VMOVQ(s0, q0);
+                    d0 = s0;
+                }
+                if(v0<15)
+                    d1 = v0;
+                else
+                    d1 = s0;
+            }
+            for(int i=0; i<4; ++i) {
+                VSQRT_F32(d1*2+i, d0*2+i);
+            }
+            if(v0<15) {
+                //done
+            } else {
+                VMOVQ(v0, d1);
+            }
+            #endif
            break;
        case 0x52:
            INST_NAME("RSQRTPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            q1 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            v0 = fpu_get_scratch_quad(dyn);
+            if(q1==q0 || !box86_dynarec_fastnan)
+                v1 = fpu_get_scratch_quad(dyn);
+            else
+                v1 = q1;
            VRSQRTEQ_F32(v0, q0);
+            VMULQ_F32(v1, v0, q0);
+            VRSQRTSQ_F32(v1, v1, v0);
+            VMULQ_F32(q1, v1, v0);
+            if(!box86_dynarec_fastnan) {
+                VCEQQ_F32(v1, q1, q1);  // 00 is nan
+                VANDQ(q1, q1, v1);  // if nan, remove
+                VBICQ(v0, v0, v1);  // if nan, keep
+                VORRQ(q1, q1, v0);
+            }
            break;
        case 0x53:
            INST_NAME("RCPPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
-            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            q1 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            if(q0 == q1)
+                v1 = fpu_get_scratch_quad(dyn);
+            else
+                v1 = q1;
+            v0 = fpu_get_scratch_quad(dyn);
            VRECPEQ_F32(v0, q0);
+            VRECPSQ_F32(v1, v0, q0);
+            VMULQ_F32(q1, v0, v1);
            break;
        case 0x54:
            INST_NAME("ANDPS Gx, Ex");
@@ -791,7 +846,23 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGX(v0, 1);
            GETEX(v1, 0);
+            // VMIN/VMAX wll put default NaN if any input is NaN
+            // but x86 will copy if either v0[x] or v1[x] is NaN, so lets force a copy if source is NaN
+            if(!box86_dynarec_fastnan) {
+                if(v0==v1) {
+                    q1 = fpu_get_scratch_quad(dyn);
+                    VMOVQ(q1, v1);
+                } else
+                    q1 = v1;
+            }
            VMINQ_F32(v0, v0, v1);
+            if(!box86_dynarec_fastnan) {
+                q0 = fpu_get_scratch_quad(dyn);
+                VCEQQ_F32(q0, v0, v0);   // 0 is NaN, -1 is not NaN
+                VANDQ(v0, v0, q0);
+                VBICQ(q0, q1, q0);
+                VORRQ(v0, v0, q0);
+            }
            break;
        case 0x5E:
            INST_NAME("DIVPS Gx, Ex");
@@ -815,14 +886,30 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGX(v0, 1);
            GETEX(v1, 0);
+            // VMIN/VMAX wll put default NaN if any input is NaN
+            // but x86 will copy if either v0[x] or v1[x] is NaN, so lets force a copy if source is NaN
+            if(!box86_dynarec_fastnan) {
+                if(v0==v1) {
+                    q1 = fpu_get_scratch_quad(dyn);
+                    VMOVQ(q1, v1);
+                } else
+                    q1 = v1;
+            }
            VMAXQ_F32(v0, v0, v1);
+            if(!box86_dynarec_fastnan) {
+                q0 = fpu_get_scratch_quad(dyn);
+                VCEQQ_F32(q0, v0, v0);   // 0 is NaN, -1 is not NaN
+                VANDQ(v0, v0, q0);
+                VBICQ(q0, q1, q0);
+                VORRQ(v0, v0, q0);
+            }
            break;
        case 0x60:
            INST_NAME("PUNPCKLBW Gm,Em");
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -833,7 +920,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -844,7 +931,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -886,7 +973,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            q0 = fpu_get_scratch_quad(dyn);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                d1 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                VMOVD(q0+1, d1);
            } else {
@@ -901,7 +988,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -913,7 +1000,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -925,7 +1012,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGM(d0);
            GETEM(d1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = fpu_get_scratch_double(dyn);
                VMOVD(v0, d1);
            } else v0 = d1;
@@ -960,7 +1047,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVQ Gm, Em");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                v0 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, gd);
                VMOVD(v0, v1);
@@ -976,7 +1063,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            gd = (nextop&0x38)>>3;
            i32 = -1;
            v0 = mmx_get_reg(dyn, ninst, x1, x2, x3, gd);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                u8 = F8;
                v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                // use stack as temporary storage
@@ -1021,7 +1108,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 2:
                    INST_NAME("PSRLW Em, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        d0 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                    } else {
                        SMREAD();
@@ -1044,7 +1131,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 4:
                    INST_NAME("PSRAW Em, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        d0 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                    } else {
                        SMREAD();
@@ -1063,7 +1150,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 6:
                    INST_NAME("PSLLW Em, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        d0 = mmx_get_reg(dyn, ninst, x1, x2, x3, nextop&7);
                    } else {
                        SMREAD();
@@ -1191,7 +1278,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVD Ed, Gm");
            nextop = F8;
            GETGM(v0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX + (nextop&7);
                VMOVfrDx_32(ed, v0, 0);
            } else {
@@ -1205,7 +1292,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVQ Em, Gm");
            nextop = F8;
            GETGM(v0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = mmx_get_reg_empty(dyn, ninst, x1, x2, x3, nextop&7);
                VMOVD(v1, v0);
            } else {
@@ -1344,7 +1431,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop=F8;  \
            MOVW_COND(NO, x3, 0); \
            MOVW_COND(YES, x3, 1);  \
-            if((nextop&0xC0)==0xC0) { \
+            if(MODREG) { \
                ed = (nextop&7);    \
                eb1 = xEAX+(ed&3);  \
                eb2 = ((ed&4)>>2);  \
@@ -1471,7 +1558,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
            } else {
                SMREAD();
@@ -1527,7 +1614,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;
            } else {
@@ -1594,7 +1681,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        MESSAGE(LOG_DUMP, "Need Optimization\n");
                        // A purge is probably not needed, just a full refresh
                        fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
-                        if((nextop&0xC0)==0xC0) {
+                        if(MODREG) {
                            DEFAULT;
                        } else {
                            addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
@@ -1606,7 +1693,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        INST_NAME("FXRSTOR Ed");
                        MESSAGE(LOG_DUMP, "Need Optimization\n");
                        fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
-                        if((nextop&0xC0)==0xC0) {
+                        if(MODREG) {
                            DEFAULT;
                        } else {
                            addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
@@ -1627,7 +1714,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        break;
                    case 3:
                        INST_NAME("STMXCSR Md");
-                        if((nextop&0xC0)==0xC0) {
+                        if(MODREG) {
                            ed = xEAX+(nextop&7);
                            LDR_IMM9(ed, xEmu, offsetof(x86emu_t, mxcsr));
                        } else {
@@ -1640,7 +1727,20 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        INST_NAME("XRSTOR Ed (not implemented");
                        FAKEED;
                        SETFLAGS(X_ALL, SF_SET);    // Hack to set flags in "don't care" state
+                        STM(xEmu, (1<<xEAX)|(1<<xECX)|(1<<xEDX)|(1<<xEBX)|(1<<xESP)|(1<<xEBP)|(1<<xESI)|(1<<xEDI)|(1<<xFlags));
+                        STR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
                        CALL(arm_ud, -1, 0);
+                        LDR_IMM9(xEIP, xEmu, offsetof(x86emu_t, ip));
+                        jump_to_epilog(dyn, 0, xEIP, ninst);
+                        *need_epilog = 0;
+                        *ok = 0;
+                        break;
+                    case 7:
+                        INST_NAME("CLFLUSH Ed");
+                        MESSAGE(LOG_DUMP, "Need Optimization?\n");
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
+                        if(ed!=x1) {MOV_REG(x1, ed);}
+                        CALL_(arm_clflush, -1, 0);
                        break;
                    default:
                        DEFAULT;
@@ -1710,7 +1810,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;
            } else {
@@ -1742,7 +1842,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVZX Gd, Eb");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
                eb2 = (ed&4)>>2;    // L or H
@@ -1757,7 +1857,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVZX Gd, Ew");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                UXTH(gd, ed, 0);
            } else {
@@ -1775,7 +1875,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
                    gd = x2;
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                    } else {
                        SMREAD();
@@ -1795,7 +1895,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("BTS Ed, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1827,7 +1927,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("BTR Ed, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1859,7 +1959,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("BTC Ed, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1899,7 +1999,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;
            } else {
@@ -1971,7 +2071,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVSX Gd, Eb");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
                eb2 = (ed&4)>>2;    // L or H
@@ -1986,7 +2086,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVSX Gd, Ew");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                SXTH(gd, ed, 0);
            } else {
@@ -2011,11 +2111,12 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGD;
            GETED;
-            MOV_REG(x1, gd);
            if(gd!=ed) {
-                MOV_REG(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, x1, x3, x14);
+            emit_add32(dyn, ninst, ed, gd, x3, x14);
            WBACK;
            break;
        case 0xC2:
@@ -2054,7 +2155,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            // ignoring "NTI"
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(xEAX+(nextop&7), gd);   // doesn't exist
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 4095, 0, 0, NULL);
@@ -2065,7 +2166,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("PINSRW Gm,Ed,Ib");
            nextop = F8;
            GETGM(d0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                u8 = (F8)&3;
                ed = xEAX+(nextop&7);
                VMOVtoDx_16(d0, u8, ed);
@@ -2080,7 +2181,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("PEXTRW Gd,Em,Ib");
            nextop = F8;
            gd = xEAX+((nextop&0x38)>>3);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETEM(d0);
                u8 = (F8)&3;
                VMOVfrDx_U16(gd, d0, u8);
@@ -2343,7 +2444,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,

        case 0xD7:
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                INST_NAME("PMOVMSKB Gd, Em");
                GETEM(d0);
                gd = xEAX+((nextop&0x38)>>3);
@@ -2487,7 +2588,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOVNTQ Em, Gm");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                DEFAULT;
            } else {
                v0 = mmx_get_reg(dyn, ninst, x1, x2, x3, gd);
@@ -29,7 +29,8 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
    int32_t i32, j32;
    uint32_t u32;
    uint8_t gd, ed;
-    uint8_t wback, u8;
+    uint8_t wback, u8, wb1, wb2;
+    uint8_t gb1, gb2, eb1, eb2;
    int fixedaddress;

    MAYUSE(j32);
@@ -46,6 +47,17 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            emit_add32(dyn, ninst, gd, ed, x3, x14);
            break;

+        case 0x20:
+            INST_NAME("AND Eb, Gb");
+            SETFLAGS(X_ALL, SF_SET_PENDING);
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            GETGB(x2);
+            GETEBO(x14);
+            emit_and8(dyn, ninst, x1, x2, x14, x2);
+            EBBACK;
+            break;
+
        case 0x2B:
            INST_NAME("SUB Gd, FS:Ed");
            SETFLAGS(X_ALL, SF_SET_PENDING);
@@ -86,6 +98,13 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            emit_cmp32(dyn, ninst, gd, ed, x3, x14);
            break;

+        case 0x64:
+            addr = dynarecFS(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+        case 0x65:
+            addr = dynarecGS(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+
        case 0x67:
            // reduced EA size...
            opcode = F8;
@@ -95,7 +114,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    nextop = F8;
                    grab_fsdata(dyn, addr, ninst, x14);
                    GETGD;  // don't need GETGW here
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        if(ed!=gd) {
                            BFI(ed, gd, 0, 16);
@@ -112,7 +131,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    nextop=F8;
                    grab_fsdata(dyn, addr, ninst, x14);
                    GETGD;
-                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                    if(MODREG) {   // reg <= reg
                        MOV_REG(gd, xEAX+(nextop&7));
                    } else {                    // mem <= reg
                        SMREAD();
@@ -125,7 +144,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("POP FS:Ew16");
                    nextop=F8;
                    grab_fsdata(dyn, addr, ninst, x14);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        POP1(xEAX+(nextop&7));  // 67 ignored
                    } else {
                        addr = geted16(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
@@ -165,7 +184,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    switch((nextop>>3)&7) {
                        case 6: // Push Ed
                            INST_NAME("PUSH FS:Ew");
-                            if((nextop&0xC0)==0xC0) {   // reg
+                            if(MODREG) {   // reg
                                DEFAULT;
                            } else {                    // mem <= i32
                                SMREAD();
@@ -182,13 +201,36 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    DEFAULT;
            }
            break;
-
+        case 0x88:
+            INST_NAME("MOV Eb, Gb");
+            grab_fsdata(dyn, addr, ninst, x14);
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            gb2 = ((gd&4)>>2);
+            gb1 = xEAX+(gd&3);
+            if(gb2) {
+                gd = x3;
+                UXTB(gd, gb1, gb2);
+            } else {
+                gd = gb1;   // no need to extract
+            }
+            if(MODREG) {
+                ed = (nextop&7);
+                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
+                eb2 = ((ed&4)<<1);    // L or H
+                BFI(eb1, gd, eb2, 8);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
+                STRB_REG_LSL_IMM5(gd, ed, x14, 0);
+                SMWRITE2();
+            }
+            break;
        case 0x89:
            INST_NAME("MOV FS:Ed, Gd");
            grab_fsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(xEAX+(nextop&7), gd);
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
@@ -202,7 +244,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            grab_fsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(gd, xEAX+(nextop&7));
            } else {                    // mem <= reg
                SMREAD();
@@ -215,7 +257,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("POP FS:Ed");
            grab_fsdata(dyn, addr, ninst, x14);
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                POP1(xEAX+(nextop&7));
            } else {
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
@@ -224,6 +266,9 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                SMWRITE2();
            }
            break;
+        case 0x90:
+            INST_NAME("NOP");
+            break;

        case 0xA0:
            INST_NAME("MOV AL, FS:Id");
@@ -288,7 +333,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV FS:Ed, Id");
            grab_fsdata(dyn, addr, ninst, x14);
            nextop=F8;
-            if((nextop&0xC0)==0xC0) {   // reg <= i32
+            if(MODREG) {   // reg <= i32
                i32 = F32S;
                ed = xEAX+(nextop&7);
                MOV32(ed, i32);
@@ -313,12 +358,15 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 2: // CALL Ed
                    INST_NAME("CALL FS:Ed");
-                    PASS2IF(ninst && dyn->insts[ninst-1].x86.set_flags, 1) {
+                    PASS2IF((box86_dynarec_safeflags>1) ||
+                        ((ninst && dyn->insts[ninst-1].x86.set_flags)
+                        || ((ninst>1) && dyn->insts[ninst-2].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
+                    if(MODREG) {   // reg
                        MOV_REG(xEIP, xEAX+(nextop&7));
                    } else {                    // mem
                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
@@ -333,7 +381,7 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 6: // Push Ed
                    INST_NAME("PUSH FS:Ed");
-                    if((nextop&0xC0)==0xC0) {   // reg
+                    if(MODREG) {   // reg
                        DEFAULT;
                    } else {                    // mem <= i32
                        SMREAD();
@@ -172,6 +172,13 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            emit_cmp32(dyn, ninst, gd, ed, x3, x14);
            break;

+        case 0x64:
+            addr = dynarecFS(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+        case 0x65:
+            addr = dynarecGS(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+
        case 0x69:
            INST_NAME("IMUL Gd, GS:Ed, Id");
            SETFLAGS(X_ALL, SF_PENDING);
@@ -369,7 +376,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(xEAX+(nextop&7), gd);
            } else {                    // mem <= reg
                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
@@ -383,7 +390,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {   // reg <= reg
+            if(MODREG) {   // reg <= reg
                MOV_REG(gd, xEAX+(nextop&7));
            } else {                    // mem <= reg
                SMREAD();
@@ -396,7 +403,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("POP GS:Ed");
            grab_tlsdata(dyn, addr, ninst, x14);
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                POP1((xEAX+(nextop&7)));
            } else {
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
@@ -405,6 +412,9 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                SMWRITE2();
            }
            break;
+        case 0x90:
+            INST_NAME("NOP");
+            break;

        case 0xA1:
            INST_NAME("MOV EAX, GS:Id");
@@ -444,7 +454,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV GS:Eb, Ib");
            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
-            if((nextop&0xC0)==0xC0) {   // reg <= u8
+            if(MODREG) {   // reg <= u8
                u8 = F8;
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
@@ -463,7 +473,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV GS:Ed, Id");
            grab_tlsdata(dyn, addr, ninst, x14);
            nextop=F8;
-            if((nextop&0xC0)==0xC0) {   // reg <= i32
+            if(MODREG) {   // reg <= i32
                i32 = F32S;
                ed = xEAX+(nextop&7);
                MOV32(ed, i32);
@@ -650,7 +660,7 @@ uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 6: // Push Ed
                    INST_NAME("PUSH GS:Ed");
-                    if((nextop&0xC0)==0xC0) {   // reg
+                    if(MODREG) {   // reg
                        DEFAULT;
                    } else {                    // mem <= i32
                        SMREAD();
@@ -18,6 +18,7 @@
 #include "dynarec_arm.h"
 #include "dynarec_arm_private.h"
 #include "arm_printer.h"
+#include "custommem.h"

 #include "dynarec_arm_helper.h"
 #include "dynarec_arm_functions.h"
@@ -47,8 +48,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("ADD Ew, Gw");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x2);
-            GETEW(x1);
+            GETSGW(x2);
+            GETSEW(x1);
            emit_add16(dyn, ninst, x1, x2, x14, x3, (wb1 && (wback==x3))?1:0);
            EWBACK;
            break;
@@ -56,16 +57,16 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("ADD Gw, Ew");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x1);
-            GETEW(x2);
+            GETSGW(x1);
+            GETSEW(x2);
            emit_add16(dyn, ninst, x1, x2, x3, x14, 0);
            GWBACK;
            break;
        case 0x05:
            INST_NAME("ADD AX, Iw");
            SETFLAGS(X_ALL, SF_SET_PENDING);
-            i32 = F16;
-            UXTH(x1, xEAX, 0);
+            i32 = F16S;
+            SXTH(x1, xEAX, 0);
            emit_add16c(dyn, ninst, x1, i32, x3, x14);
            BFI(xEAX, x1, 0, 16);
            break;
@@ -123,8 +124,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x2);
-            GETEW(x1);
+            GETSGW(x2);
+            GETSEW(x1);
            emit_adc16(dyn, ninst, x1, x2, x14, x3, (wb1 && (wback==x3))?1:0);
            EWBACK;
            break;
@@ -133,8 +134,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x1);
-            GETEW(x2);
+            GETSGW(x1);
+            GETSEW(x2);
            emit_adc16(dyn, ninst, x1, x2, x14, x3, 0);
            GWBACK;
            break;
@@ -142,8 +143,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("ADC AX, Iw");
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
-            i32 = F16;
-            UXTH(x1, xEAX, 0);
+            i32 = F16S;
+            SXTH(x1, xEAX, 0);
            emit_adc16c(dyn, ninst, x1, i32, x3, x14);
            BFI(xEAX, x1, 0, 16);
            break;
@@ -153,8 +154,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x2);
-            GETEW(x1);
+            GETSGW(x2);
+            GETSEW(x1);
            emit_sbb16(dyn, ninst, x1, x2, x14, x3, (wb1 && (wback==x3))?1:0);
            EWBACK;
            break;
@@ -164,7 +165,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGW(x1);
-            GETEW(x2);
+            GETSEW(x2);
            emit_sbb16(dyn, ninst, x1, x2, x14, x3, 0);
            GWBACK;
            break;
@@ -172,9 +173,9 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("SBB AX, Iw");
            READFLAGS(X_CF);
            SETFLAGS(X_ALL, SF_SET_PENDING);
-            i16 = F16;
-            UXTH(x1, xEAX, 0);
-            emit_sbb16c(dyn, ninst, x1, i16, x3, x14);
+            i32 = F16S;
+            SXTH(x1, xEAX, 0);
+            emit_sbb16c(dyn, ninst, x1, i32, x3, x14);
            BFI(xEAX, x1, 0, 16);
            break;
        case 0x1E:
@@ -219,8 +220,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("SUB Ew, Gw");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x2);
-            GETEW(x1);
+            GETSGW(x2);
+            GETSEW(x1);
            emit_sub16(dyn, ninst, x1, x2, x14, x3, (wb1 && (wback==x3))?1:0);
            EWBACK;
            break;
@@ -228,16 +229,16 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("SUB Gw, Ew");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            GETGW(x1);
-            GETEW(x2);
+            GETSGW(x1);
+            GETSEW(x2);
            emit_sub16(dyn, ninst, x1, x2, x3, x14, 0);
            GWBACK;
            break;
        case 0x2D:
            INST_NAME("SUB AX, Iw");
            SETFLAGS(X_ALL, SF_SET_PENDING);
-            i32 = F16;
-            UXTH(x1, xEAX, 0);
+            i32 = F16S;
+            SXTH(x1, xEAX, 0);
            emit_sub16c(dyn, ninst, x1, i32, x3, x14);
            BFI(xEAX, x1, 0, 16);
            break;
@@ -388,13 +389,13 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("IMUL Gw,Ew,Iw");
            SETFLAGS(X_ALL, SF_PENDING);
            nextop = F8;
-            UFLAG_DF(x1, d_imul16);
            GETEW(x1);
            i16 = F16S;
            MOVW(x2, i16);
            SMULBB(x2, ed, x2);
            UFLAG_RES(x2);
            BFI(xEAX+((nextop&0x38)>>3), x2, 0, 16);
+            UFLAG_DF(x1, d_imul16);
            break;
        case 0x6A:
            INST_NAME("PUSH Ib");
@@ -406,13 +407,13 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("IMUL Gw,Ew,Ib");
            SETFLAGS(X_ALL, SF_PENDING);
            nextop = F8;
-            UFLAG_DF(x1, d_imul16);
            GETEW(x1);
            i16 = F8S;
            MOVW(x2, i16);
            SMULBB(x2, ed, x2);
            UFLAG_RES(x2);
            BFI(xEAX+((nextop&0x38)>>3), x2, 0, 16);
+            UFLAG_DF(x1, d_imul16);
            break;

        case 0x81:
@@ -507,7 +508,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x87:
            INST_NAME("(LOCK) XCHG Ew, Gw");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGD;
                UXTH(x14, gd, 0);
                ed = xEAX + (nextop&7);
@@ -541,7 +542,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Ew, Gw");
            nextop = F8;
            GETGD;  // don't need GETGW here
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                if(ed!=gd) {
                    BFI(ed, gd, 0, 16);
@@ -556,7 +557,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Gw, Ew");
            nextop = F8;
            GETGD;  // don't need GETGW neither
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                if(ed!=gd) {
                    BFI(gd, ed, 0, 16);
@@ -573,7 +574,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            MOV32(x2, offsetof(x86emu_t, segs[(nextop&0x38)>>3]));
            LDRH_REG(x1, xEmu, x2);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                BFI(ed, x1, 0, 16);
            } else {
@@ -586,7 +587,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x8E:
            INST_NAME("MOV Seg,Ew");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                MOV32(x2, offsetof(x86emu_t, segs[(nextop&0x38)>>3]));
                STRH_REG(ed, xEmu, x2);
@@ -649,7 +650,8 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV, AX, Od");
            u32 = F32;
            MOV32(x2, u32);
-            SMREAD();
+            if(isLockAddress(u32)) lock=1; else lock = 0;
+            SMREADLOCK(lock);
            LDRH_IMM8(x2, x2, 0);
            BFI(xEAX, x2, 0, 16);
            break;
@@ -657,8 +659,9 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("MOV Od, AX");
            u32 = F32;
            MOV32(x2, u32);
+            if(isLockAddress(u32)) lock=1; else lock = 0;
            STRH_IMM8(xEAX, x2, 0);
-            SMWRITE();
+            SMWRITELOCK(lock);
            break;
        case 0xA5:
            INST_NAME("MOVSW");
@@ -723,21 +726,29 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("ROL Ew, Ib");
-                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    GETEW(x1);
-                    u8 = F8;
-                    MOVW(x2, u8);
-                    CALL_(rol16, x1, (1<<x3));
-                    EWBACK;
+                    if(geted_ib(dyn, addr, ninst, nextop)&31) {
+                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                        GETEW(x1);
+                        u8 = F8;
+                        emit_rol16c(dyn, ninst, x1, u8&15, x2, x14);
+                        EWBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
                case 1:
                    INST_NAME("ROR Ew, Ib");
-                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    GETEW(x1);
-                    u8 = F8;
-                    MOVW(x2, u8);
-                    CALL_(ror16, x1, (1<<x3));
-                    EWBACK;
+                    if(geted_ib(dyn, addr, ninst, nextop)&31) {
+                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                        GETEW(x1);
+                        u8 = F8;
+                        emit_ror16c(dyn, ninst, x1, u8&15, x2, x14);
+                        EWBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
                case 2:
                    INST_NAME("RCL Ew, Ib");
@@ -762,39 +773,42 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 4:
                case 6:
                    INST_NAME("SHL Ew, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETEW(x1);
-                    u8 = F8;
-                    MOVW(x2, (u8&0x1f));
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_LSL_REG(ed, ed, x2);
-                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shl16);
+                    if(geted_ib(dyn, addr, ninst, nextop)&31) {
+                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                        GETEW(x1);
+                        u8 = F8;
+                        emit_shl16c(dyn, ninst, x1, u8&31, x2, x14);
+                        EWBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
                case 5:
                    INST_NAME("SHR Ed, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETEW(x1);
-                    u8 = F8;
-                    MOVW(x2, (u8&0x1f));
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_LSR_REG(ed, ed, x2);
-                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shr16);
+                    if(geted_ib(dyn, addr, ninst, nextop)&31) {
+                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                        GETEW(x1);
+                        u8 = F8;
+                        emit_shr16c(dyn, ninst, x1, u8&31, x2, x14);
+                        EWBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
                case 7:
                    INST_NAME("SAR Ed, Ib");
-                    SETFLAGS(X_ALL, SF_PENDING);
-                    GETSEW(x1);
-                    u8 = F8;
-                    MOVW(x2, (u8&0x1f));
-                    UFLAG_OP12(ed, x2)
-                    MOV_REG_ASR_REG(ed, ed, x2);
-                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_sar16);
+                    if(geted_ib(dyn, addr, ninst, nextop)&31) {
+                        SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
+                        GETSEW(x1);
+                        u8 = F8;
+                        emit_sar16c(dyn, ninst, x1, u8&31, x2, x14);
+                        EWBACK;
+                    } else {
+                        FAKEED;
+                        F8;
+                    }
                    break;
            }
            break;
@@ -802,7 +816,7 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0xC7:
            INST_NAME("MOV Ew, Iw");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                u16 = F16;
                MOVW(x1, u16);
@@ -816,101 +830,178 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;

        case 0xD1:
-        case 0xD3:
            nextop = F8;
            switch((nextop>>3)&7) {
                case 0:
-                    if(opcode==0xD1) {
-                        INST_NAME("ROL Ew, 1");
-                        MOVW(x2, 1);
-                    } else {
-                        INST_NAME("ROL Ew, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    INST_NAME("ROL Ew, 1");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETEW(x1);
-                    CALL_(rol16, x1, (1<<x3));
+                    emit_rol16c(dyn, ninst, x1, 1, x2, x14);
                    EWBACK;
                    break;
                case 1:
-                    if(opcode==0xD1) {
-                        INST_NAME("ROR Ew, 1");
-                        MOVW(x2, 1);
-                    } else {
-                        INST_NAME("ROR Ew, CL");
-                        AND_IMM8(x2, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    INST_NAME("ROR Ew, 1");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET_PENDING);
                    GETEW(x1);
-                    CALL_(ror16, x1, (1<<x3));
+                    emit_ror16c(dyn, ninst, x1, 1, x2, x14);
                    EWBACK;
                    break;
                case 2:
-                    if(opcode==0xD1) {INST_NAME("RCL Ew, 1"); } else { INST_NAME("RCL Ew, CL");}
+                    INST_NAME("RCL Ew, 1");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD1) {MOVW(x2, 1);} else {AND_IMM8(x2, xECX, 0x1f);}
+                    MOVW(x2, 1);
                    GETEW(x1);
                    CALL_(rcl16, x1, (1<<x3));
                    EWBACK;
                    break;
                case 3:
-                    if(opcode==0xD1) {INST_NAME("RCR Ew, 1");} else {INST_NAME("RCR Ew, CL");}
+                    INST_NAME("RCR Ew, 1");
                    READFLAGS(X_CF);
                    SETFLAGS(X_OF|X_CF, SF_SET);
-                    if(opcode==0xD1) {MOVW(x2, 1);} else {AND_IMM8(x2, xECX, 0x1f);}
+                    MOVW(x2, 1);
                    GETEW(x1);
                    CALL_(rcr16, x1, (1<<x3));
                    EWBACK;
                    break;
                case 4:
                case 6:
-                    if(opcode==0xD1) {
-                        INST_NAME("SHL Ew, 1");
-                        MOVW(x14, 1);
-                    } else {
-                        INST_NAME("SHL Ew, CL");
-                        AND_IMM8(x14, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SHL Ew, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETEW(x1);
-                    UFLAG_OP12(ed, x14)
-                    MOV_REG_LSL_REG(ed, ed, x14);
+                    emit_shl16c(dyn, ninst, x1, 1, x2, x14);
                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shl16);
                    break;
                case 5:
-                    if(opcode==0xD1) {
-                        INST_NAME("SHR Ew, 1");
-                        MOVW(x14, 1);
-                    } else {
-                        INST_NAME("SHR Ew, CL");
-                        AND_IMM8(x14, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SHR Ew, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETEW(x1);
-                    UFLAG_OP12(ed, x14)
-                    MOV_REG_LSR_REG(ed, ed, x14);
+                    emit_shr16c(dyn, ninst, x1, 1, x2, x14);
                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_shr16);
                    break;
                case 7:
-                    if(opcode==0xD1) {
-                        INST_NAME("SAR Ew, 1");
-                        MOVW(x14, 1);
-                    } else {
-                        INST_NAME("SAR Ew, CL");
-                        AND_IMM8(x14, xECX, 0x1f);
-                    }
-                    SETFLAGS(X_ALL, SF_PENDING);
+                    INST_NAME("SAR Ew, 1");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
                    GETSEW(x1);
-                    UFLAG_OP12(ed, x14)
-                    MOV_REG_ASR_REG(ed, ed, x14);
+                    emit_sar16c(dyn, ninst, x1, 1, x2, x14);
+                    EWBACK;
+                    break;
+            }
+            break;
+            
+        case 0xD3:
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("ROL Ew, CL");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        TSTS_IMM8(xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    }
+                    AND_IMM8(x2, xECX, 0x0f);
+                    RSB_IMM8(x2, x2, 16);
+                    GETEW(x1);
+                    ORR_REG_LSL_IMM5(ed, ed, ed, 16);
+                    MOV_REG_LSR_REG(ed, ed, x2);
+                    EWBACK;
+                    UFLAG_IF {  // calculate flags directly
+                        CMPS_IMM8(x2, 15);
+                            ADD_REG_LSR_IMM5_COND(cEQ, x3, ed, ed, 15);
+                            BFI_COND(cEQ, xFlags, x3, F_OF, 1);
+                        BFI(xFlags, ed, F_CF, 1);
+                        UFLAG_DF(x2, d_none);
+                    }
+                    break;
+                case 1:
+                    INST_NAME("ROR Ew, CL");
+                    SETFLAGS(X_OF|X_CF, SF_SUBSET);
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        TSTS_IMM8(xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    }
+                    AND_IMM8(x2, xECX, 0x0f);
+                    GETEW(x1);
+                    ORR_REG_LSL_IMM5(ed, ed, ed, 16);
+                    MOV_REG_LSR_REG(ed, ed, x2);
+                    EWBACK;
+                    UFLAG_IF {  // calculate flags directly
+                        CMPS_IMM8(x2, 1);
+                            MOV_REG_LSR_IMM5_COND(cEQ, x2, ed, 14); // x2 = d>>14
+                            XOR_REG_LSR_IMM8_COND(cEQ, x2, x2, x2, 1); // x2 = ((d>>14) ^ ((d>>14)>>1))
+                            BFI_COND(cEQ, xFlags, x2, F_OF, 1);
+                        MOV_REG_LSR_IMM5(x2, ed, 15);
+                        BFI(xFlags, x2, F_CF, 1);
+                        UFLAG_DF(x2, d_none);
+                    }
+                    break;
+                case 2:
+                    INST_NAME("RCL Ew, CL");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    AND_IMM8(x2, xECX, 0x1f);
+                    GETEW(x1);
+                    CALL_(rcl16, x1, (1<<x3));
+                    EWBACK;
+                    break;
+                case 3:
+                    INST_NAME("RCR Ew, CL");
+                    READFLAGS(X_CF);
+                    SETFLAGS(X_OF|X_CF, SF_SET);
+                    AND_IMM8(x2, xECX, 0x1f);
+                    GETEW(x1);
+                    CALL_(rcr16, x1, (1<<x3));
+                    EWBACK;
+                    break;
+                case 4:
+                case 6:
+                    INST_NAME("SHL Ew, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETEW(x1);
+                    emit_shl16(dyn, ninst, x1, x2, x2, x14);
+                    EWBACK;
+                    break;
+                case 5:
+                    INST_NAME("SHR Ew, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETEW(x1);
+                    emit_shr16(dyn, ninst, x1, x2, x2, x14);
+                    EWBACK;
+                    break;
+                case 7:
+                    INST_NAME("SAR Ew, CL");
+                    SETFLAGS(X_ALL, SF_SET_PENDING);    // some flags are left undefined
+                    if(box86_dynarec_safeflags>1)
+                        MAYSETFLAGS();
+                    UFLAG_IF {
+                        ANDS_IMM8(x2, xECX, 0x1f);
+                        B_NEXT(cEQ);
+                    } else {
+                        AND_IMM8(x2, xECX, 0x1f);
+                    }
+                    GETSEW(x1);
+                    emit_sar16(dyn, ninst, x1, x2, x2, x14);
                    EWBACK;
-                    UFLAG_RES(ed);
-                    UFLAG_DF(x3, d_sar16);
                    break;
            }
            break;
@@ -1058,13 +1149,13 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 5:
                    INST_NAME("IMUL AX, Ew");
                    SETFLAGS(X_ALL, SF_PENDING);
-                    UFLAG_DF(x1, d_imul16);
                    GETEW(x1);
                    SMULBB(x2, xEAX, ed);
                    UFLAG_RES(x2);
                    BFI(xEAX, x2, 0, 16);
                    MOV_REG_LSR_IMM5(x2, x2, 16);
                    BFI(xEDX, x2, 0, 16);
+                    UFLAG_DF(x1, d_imul16);
                    break;
                case 6:
                    INST_NAME("DIV Ew");
@@ -1128,6 +1219,11 @@ uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    emit_dec16(dyn, ninst, x1, x2, x14);
                    EWBACK;
                    break;
+                case 6:
+                    INST_NAME("PUSH Ew");
+                    GETEW(x1);
+                    PUSH16(x1, xESP);
+                    break;
                default:
                    DEFAULT;
            }
@@ -21,10 +21,11 @@

 #include "dynarec_arm_functions.h"
 #include "dynarec_arm_helper.h"
+#include "emu/x86compstrings.h"

 // Get EX as a quad
 #define GETEX(a, w)             \
-    if((nextop&0xC0)==0xC0) {   \
+    if(MODREG) {   \
        a = sse_get_reg(dyn, ninst, x1, nextop&7, w);  \
    } else {                    \
        SMREAD();               \
@@ -35,12 +36,18 @@
 #define GETGX(a, w)         \
    gd = (nextop&0x38)>>3;  \
    a = sse_get_reg(dyn, ninst, x1, gd, w)
+#define GETGX_empty(a)      \
+    gd = (nextop&0x38)>>3;  \
+    a = sse_get_reg_empty(dyn, ninst, x1, gd)
+
+#define GETG    gd = (nextop&0x38)>>3

 uintptr_t dynarec660F(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, j32;
+    uint32_t u32;
    uint8_t gd, ed;
    uint8_t wback, wb1;
    uint8_t eb1, eb2;
@@ -66,7 +73,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVUPD Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -87,7 +94,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -105,7 +112,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVLPD Gx, Eq");
            nextop = F8;
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                // access register instead of memory is bad opcode!
                DEFAULT;
                return addr;
@@ -126,7 +133,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVLPD Eq, Gx");
            nextop = F8;
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                // access register instead of memory is bad opcode!
                DEFAULT;
                return addr;
@@ -147,7 +154,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("UNPCKLPD Gx, Ex");
            nextop = F8;
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                VMOVD(v0+1, v1);
            } else {
@@ -161,7 +168,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGX(v0, 1);
            VMOVD(v0, v0+1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                VMOVD(v0+1, v1+1);
            } else {
@@ -174,7 +181,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVHPD Gx, Ed");
            nextop = F8;
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                // access register instead of memory is bad opcode!
                DEFAULT;
                return addr;
@@ -195,7 +202,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVHPD Ed, Gx");
            nextop = F8;
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                // access register instead of memory is bad opcode!
                DEFAULT;
                return addr;
@@ -223,14 +230,14 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVAPD Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
-                VLD1Q_32(v0, ed);
+                VLD1Q_64(v0, ed);
                SMWRITE2();
            }
            break;
@@ -239,12 +246,28 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
-                VST1Q_32(v0, ed);
+                VST1Q_64(v0, ed);
+                SMWRITE2();
+            }
+            break;
+
+
+        case 0x2B:
+            INST_NAME("MOVNTPD Ex, Gx");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
+            if(MODREG) {
+                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
+                VMOVQ(v1, v0);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
+                VST1Q_64(v0, ed);
                SMWRITE2();
            }
            break;
@@ -270,7 +293,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    nextop = F8;
                    GETGX(q0, 1);
                    q1 = fpu_get_scratch_quad(dyn);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        v0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                    } else {
                        SMREAD();
@@ -444,6 +467,108 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    q0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                    VABSQ_S32(q0, q1);
                    break;
+
+                case 0xDB:
+                    INST_NAME("AESIMC Gx, Ex");  // AES-NI
+                    nextop = F8;
+                    if(arm_aes) {
+                        GETEX(q1, 0);
+                        GETGX_empty(q0);
+                        AESIMC(q0, q1);
+                    } else {
+                        GETEX(q1, 0);
+                        GETGX_empty(q0);
+                        if(q0!=q1) {
+                            VMOVQ(q0, q1);
+                        }
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        MOV32(x1, gd);
+                        CALL(arm_aesimc, -1, 0);
+                    }
+                    break;
+                case 0xDC:
+                    INST_NAME("AESENC Gx, Ex");  // AES-NI
+                    nextop = F8;
+                    if(arm_aes) {
+                        GETEX(q1, 0);
+                        GETGX(q0, 1);
+                        v0 = fpu_get_scratch_quad(dyn);  // ARM64 internal operation differs a bit from x86_64
+                        VEORQ(v0, q0, q1);
+                        AESE(v0, q1);
+                        AESMC(v0, v0);
+                        VEORQ(q0, v0, q1);
+                    } else {
+                        GETG;
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        MOV32(x1, gd);
+                        CALL(arm_aese, -1, 0);
+                        GETGX(q0, 1);
+                        GETEX(q1, 0);
+                        VEORQ(q0, q0, q1);
+                    }
+                    break;
+                case 0xDD:
+                    INST_NAME("AESENCLAST Gx, Ex");  // AES-NI
+                    nextop = F8;
+                    if(arm_aes) {
+                        GETEX(q1, 0);
+                        GETGX(q0, 1);
+                        v0 = fpu_get_scratch_quad(dyn);  // ARM64 internal operation differs a bit from x86_64
+                        VEORQ(v0, q0, q1);
+                        AESE(v0, q1);
+                        VEORQ(q0, v0, q1);
+                    } else {
+                        GETG;
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        MOV32(x1, gd);
+                        CALL(arm_aeselast, -1, 0);
+                        GETGX(q0, 1);
+                        GETEX(q1, 0);
+                        VEORQ(q0, q0, q1);
+                    }
+                    break;
+                case 0xDE:
+                    INST_NAME("AESDEC Gx, Ex");  // AES-NI
+                    nextop = F8;
+                    if(arm_aes) {
+                        GETEX(q1, 0);
+                        GETGX(q0, 1);
+                        v0 = fpu_get_scratch_quad(dyn);  // ARM64 internal operation differs a bit from x86_64
+                        VEORQ(v0, q0, q1);
+                        AESD(v0, q1);
+                        AESIMC(v0, v0);
+                        VEORQ(q0, v0, q1);
+                    } else {
+                        GETG;
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        MOV32(x1, gd);
+                        CALL(arm_aesd, -1, 0);
+                        GETGX(q0, 1);
+                        GETEX(q1, 0);
+                        VEORQ(q0, q0, q1);
+                    }
+                    break;
+                case 0xDF:
+                    INST_NAME("AESDECLAST Gx, Ex");  // AES-NI
+                    nextop = F8;
+                    if(arm_aes) {
+                        GETEX(q1, 0);
+                        GETGX(q0, 1);
+                        v0 = fpu_get_scratch_quad(dyn);  // ARM64 internal operation differs a bit from x86_64
+                        VEORQ(v0, q0, q1);
+                        AESD(v0, q1);
+                        VEORQ(q0, v0, q1);
+                    } else {
+                        GETG;
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        MOV32(x1, gd);
+                        CALL(arm_aesdlast, -1, 0);
+                        GETGX(q0, 1);
+                        GETEX(q1, 0);
+                        VEORQ(q0, q0, q1);
+                    }
+                    break;
+
                default:
                    DEFAULT;
            }
@@ -468,6 +593,163 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                        VEXTQ_8(q0, q1, q0, u8);
                    }
                    break;
+
+                case 0x16:
+                    INST_NAME("PEXTRD Ed, Gx, Ib");
+                    nextop = F8;
+                    GETGX(q0, 0);
+                    if(MODREG) {
+                        ed = xEAX+(nextop&7);
+                        u8 = F8;
+                        VMOVfrDx_32(ed, q0+((u8&2)>>1), u8&1);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, 0, NULL);
+                        u8 = F8;
+                        VST1LANE_32(q0+((u8&2)>>1), wback, u8&1);
+                        SMWRITE2();
+                    }
+                    break;
+
+                case 0x20:
+                    INST_NAME("PINSRB Gx, ED, Ib");
+                    nextop = F8;
+                    GETGX(q0, 1);
+                    if(MODREG) {
+                        ed = xEAX+(nextop&7);
+                        u8 = F8;
+                        VMOVtoDx_8(q0+((u8&8)>>3), u8&7, ed);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, 0, NULL);
+                        u8 = F8;
+                        VLD1LANE_8(q0+((u8&8)>>3), wback, u8&7);
+                        SMWRITE2();
+                    }
+                    break;
+
+                case 0x22:
+                    INST_NAME("PINSRD Gx, ED, Ib");
+                    nextop = F8;
+                    GETGX(q0, 1);
+                    if(MODREG) {
+                        ed = xEAX+(nextop&7);
+                        u8 = F8;
+                        VMOVtoDx_32(q0+((u8&2)>>1), u8&1, ed);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, 0, NULL);
+                        u8 = F8;
+                        VLD1LANE_32(q0+((u8&2)>>1), wback, u8&1);
+                        SMWRITE2();
+                    }
+                    break;
+
+                case 0x44:
+                    INST_NAME("PCLMULQDQ Gx, Ex, Ib");
+                    nextop = F8;
+                    if(arm_pmull) {
+                        GETGX(q0, 1);
+                        GETEX(q1, 0);
+                        u8 = F8;
+                        switch (u8&0b00010001) {
+                            case 0b00000000:
+                                VMULL_P64(q0, q0, q1);
+                                break;
+                            case 0b00010001:
+                                VMULL_P64(q0, q0+1, q1+1);
+                                break;
+                            case 0b00000001:
+                                VMULL_P64(q0, q0+1, q1);
+                                break;
+                            case 0b00010000:
+                                VMULL_P64(q0, q0, q1+1);
+                                break;
+                        }
+                    } else {
+                        GETG;
+                        sse_forget_reg(dyn, ninst, gd, x1);
+                        if(MODREG) {
+                            ed = nextop&7;
+                            sse_forget_reg(dyn, ninst, ed, x1);
+                            MOV32(x2, ed);
+                        } else {
+                            MOV32(x2, 0);
+                            addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
+                            if(ed!=x2) {
+                                MOV_REG(x2, ed);
+                            }
+                        }
+                        MOV32(x1, gd); // gx
+                        u8 = F8;
+                        MOV32(x3, u8);
+                        CALL(arm_pclmul, -1, 0);
+                    }
+                    break;
+
+                case 0x63:
+                    INST_NAME("PCMPISTRI Gx, Ex, Ib");
+                    SETFLAGS(X_OF|X_CF|X_AF|X_ZF|X_SF|X_PF, SF_SET);
+                    nextop = F8;
+                    GETG;
+                    if(!sse_reflect_reg(dyn, ninst, gd, x2)) {
+                        u32 = offsetof(x86emu_t, xmm[gd]);
+                        if(!(u32&3) && (u32>>2)<256) {
+                            ADD_IMM8_ROR(x2, xEmu, u32>>2, 15);
+                        } else {
+                            MOV32(x2, u32);
+                            ADD_REG_LSL_IMM5(x2, xEmu, x2, 0);
+                        }
+                    }
+                    if(MODREG) {
+                        ed = (nextop&7);
+                        if(!sse_reflect_reg(dyn, ninst, ed, x1)) {
+                            u32 = offsetof(x86emu_t, xmm[ed]);
+                            if(!(u32&3) && (u32>>2)<256) {
+                                ADD_IMM8_ROR(x1, xEmu, u32>>2, 15);
+                            } else {
+                                MOV32(x1, u32);
+                                ADD_REG_LSL_IMM5(x1, xEmu, x1, 0);
+                            }
+                        }
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
+                        if(ed!=x1) {
+                            MOV_REG(x1, ed);
+                        }
+                    }
+                    u8 = F8;
+                    MOV32(x3, u8);
+                    CALL(sse42_compare_string_implicit_len, x1, 0);
+                    CMPS_IMM8(x1, 0);
+                    MOVW_COND(cEQ, xECX, (u8&1)?8:16);
+                    if(u8&0b1000000) {
+                        CLZ_COND(cNE, xECX, x1);
+                        RSB_COND_IMM8(cNE, xECX, xECX, 31);
+                    } else {
+                        RBIT_COND(cNE, xECX, x1);
+                        CLZ_COND(cNE, xECX, xECX);
+                    }
+                    break;
+                    
+                case 0xDF:
+                    INST_NAME("AESKEYGENASSIST Gx, Ex, Ib");  // AES-NI
+                    nextop = F8;
+                    GETG;
+                    sse_forget_reg(dyn, ninst, gd, x1);
+                    MOV32(x1, gd); // gx
+                    if(MODREG) {
+                        ed = nextop&7;
+                        sse_forget_reg(dyn, ninst, ed, x2);
+                        MOV32(x2, ed);
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0, 0, NULL);
+                        if(ed!=x2) {
+                            MOV_REG(x2, ed);
+                        }
+                    }
+                    u8 = F8;
+                    MOV32(x3, u8);
+                    CALL(arm_aeskeygenassist, -1, 0);
+                    break;
+
                default:
                    DEFAULT;
            }
@@ -478,7 +760,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            GETFLAGS;   \
            nextop=F8;  \
            GETGD;      \
-            if((nextop&0xC0)==0xC0) {   \
+            if(MODREG) {   \
                ed = xEAX+(nextop&7);   \
            } else { \
                SMREAD(); \
@@ -586,7 +868,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETEX(q0, 0);
            gd = xEAX+((nextop&0x38)>>3);
-            if((nextop&0xC0)==0xC0)
+            if(MODREG)
                q1 = fpu_get_scratch_quad(dyn);
            else 
                q1 = q0;
@@ -711,11 +993,21 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                q1 = v1;
            // need rounding!
            //VCVTQ_S32_F32(v0, v1);
-            u8 = sse_setround(dyn, ninst, x1, x2, x14);
-            VCVTR_S32_F32(q0*2+0, q1*2+0);
-            VCVTR_S32_F32(q0*2+1, q1*2+1);
-            VCVTR_S32_F32(q0*2+2, q1*2+2);
-            VCVTR_S32_F32(q0*2+3, q1*2+3);
+            if(!box86_dynarec_fastround)
+                u8 = sse_setround_reset(dyn, ninst, x1, x2, x14);
+            else
+                u8 = sse_setround(dyn, ninst, x1, x2, x14);
+            for(int i=0; i<4; ++i) {
+                VCVTR_S32_F32(q0*2+i, q1*2+i);
+                if(!box86_dynarec_fastround) {
+                    VMRS(x3);   // get the FPCSR reg and test FPU exception (IO only)
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    MOV_IMM_COND(cNE, x1, 0b10, 1);   // 0x80000000
+                    VMOVtoVcond(cNE, q0*2+i, x1);
+                    BIC_IMM8_COND(cNE, x3, x3, 1, 0);    // reset FPU
+                    VMSR_cond(cNE, x3);
+                }
+        }
            if(q0!=v0) {
                VMOVQ(v0, q0);
            }
@@ -920,7 +1212,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                VMOVD(v0+1, v1);
            } else {
@@ -934,7 +1226,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGX(q0, 1);
            VMOVD(q0, q0+1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                VMOVD(q0+1, q1+1);
            } else {
@@ -956,7 +1248,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVDQA Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -972,7 +1264,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            i32 = -1;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                u8 = F8;
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
@@ -1103,7 +1395,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            switch((nextop>>3)&7) {
                case 2:
                    INST_NAME("PSRLW Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1126,7 +1418,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 4:
                    INST_NAME("PSRAW Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1145,7 +1437,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 6:
                    INST_NAME("PSLLW Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1176,7 +1468,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            switch((nextop>>3)&7) {
                case 2:
                    INST_NAME("PSRLD Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1199,7 +1491,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 4:
                    INST_NAME("PSRAD Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1218,7 +1510,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 6:
                    INST_NAME("PSLLD Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1248,7 +1540,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            switch((nextop>>3)&7) {
                case 2:
                    INST_NAME("PSRLQ Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1271,7 +1563,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 3:
                    INST_NAME("PSRLDQ Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1296,7 +1588,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 6:
                    INST_NAME("PSLLQ Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1319,7 +1611,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    break;
                case 7:
                    INST_NAME("PSLLDQ Ex, Ib");
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                    } else {
                        SMREAD();
@@ -1373,7 +1665,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX + (nextop&7);
                VMOVfrDx_32(ed, v0, 0);
            } else {
@@ -1387,7 +1679,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVDQA Ex,Gx");
            nextop = F8;
            GETGX(v0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -1403,7 +1695,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;  // there is an AND below, so 32bits is the same (no need for GETGW)
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);   // no need for extract
            } else {
                SMREAD();
@@ -1445,7 +1737,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;  // there is an AND below, to 32bits is the same
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;   // no need for extract
            } else {
@@ -1495,8 +1787,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("IMUL Gw,Ew");
            SETFLAGS(X_ALL, SF_PENDING);
            nextop = F8;
-            UFLAG_DF(x1, d_imul16);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
            } else {
                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0, 0, NULL);
@@ -1507,6 +1798,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            SMULBB(x2, gd, ed);
            UFLAG_RES(x2);
            BFI((xEAX+((nextop&0x38)>>3)), x2, 0, 16);
+            UFLAG_DF(x1, d_imul16);
            break;

        case 0xB3:
@@ -1515,7 +1807,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;  // there is an AND below, to 32bits is the same
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;   // no need for extract
            } else {
@@ -1544,7 +1836,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVZX Gw, Eb");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
                eb2 = (ed&4)>>2;    // L or H
@@ -1560,7 +1852,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVZX Gw, Ew");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX + (nextop&7);
                UXTH(x1, ed, 0);
            } else {
@@ -1578,7 +1870,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    INST_NAME("BT Ew, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                    } else {
                        SMREAD();
@@ -1598,7 +1890,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    INST_NAME("BTS Ew, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1629,7 +1921,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    INST_NAME("BTR Ew, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1660,7 +1952,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                    INST_NAME("BTC Ew, Ib");
                    SETFLAGS(X_CF, SF_SUBSET);
                    SET_DFNONE(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -1696,7 +1988,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            SET_DFNONE(x1);
            nextop = F8;
            GETGD;  // there is an AND below, to 32bits is the same
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                wback = 0;   // no need for extract
            } else {
@@ -1755,7 +2047,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVSX Gw, Eb");
            nextop = F8;
            GETGD;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = (nextop&7);
                eb1 = xEAX+(ed&3);  // Ax, Cx, Dx or Bx
                eb2 = (ed&4)>>2;    // L or H
@@ -1828,7 +2120,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("PINSRW Gx,Ed,Ib");
            nextop = F8;
            GETGX(v0, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                u8 = (F8)&7;
                ed = xEAX+(nextop&7);
                VMOVtoDx_16(v0+(u8/4), u8&3, ed);
@@ -1843,7 +2135,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("PEXTRW Gd,Ex,Ib");
            nextop = F8;
            gd = xEAX+((nextop&0x38)>>3);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETEX(v0, 0);
                u8 = (F8)&7;
                VMOVfrDx_U16(gd, v0+(u8/4), (u8&3));
@@ -1957,7 +2249,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVQ Ex,Gx");
            nextop = F8;
            GETGX(q0, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVD(q1, q0);
                VEOR(q1+1, q1+1, q1+1);
@@ -1978,7 +2270,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            break;
        case 0xD7:
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                INST_NAME("PMOVMSKB Gd, Ex");
                GETEX(q0, 0);
                gd = xEAX+((nextop&0x38)>>3);
@@ -2149,7 +2441,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -2264,7 +2556,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            q0 = fpu_get_scratch_quad(dyn);
            VMOVQ(q0, v0);
            VTRN_32(q0, q0+1);  // transpose GX
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = fpu_get_scratch_quad(dyn);
                VMOVQ(q1, v1);
            } else {
@@ -2295,13 +2587,13 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            VPADDLQ_U32(q0, q0);
            break;
        case 0xF7:
-            INST_NAME("MASKMOVDQU Gx, Ex")
+            INST_NAME("MASKMOVDQU Gx, Ex");
            nextop = F8;
            GETGX(q0, 1);
            GETEX(q1, 0);
            v0 = fpu_get_scratch_quad(dyn);
            VLD1Q_8(v0, xEDI);  // hopefully it's always aligned?
-            if((nextop&0xC0)==0xC0)
+            if(MODREG)
                v1 = fpu_get_scratch_quad(dyn); // need to preserve the register
            else
                v1 = q1;
@@ -41,7 +41,6 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
        case 0x81:
        case 0x83:
            nextop = F8;
-            SMDMB();
            switch((nextop>>3)&7) {
                case 0: //ADD
                    if(opcode==0x81) {
@@ -50,8 +49,7 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                        INST_NAME("LOCK ADD Ew, Ib");
                    }
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i16 = F16S; else i16 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_add16c(dyn, ninst, ed, i16, x3, x14);
@@ -69,9 +67,9 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            STREXH(x3, x1, wback);
                            CMPS_IMM8(x3, 0);
                            B_MARKLOCK(cNE);
+                            SMDMB();
                        }
                        if(!fixedaddress) {
-                            SMDMB();
                            B_NEXT(c__);
                        }
                        if(!fixedaddress || (fixedaddress && (fixedaddress&1))) {
@@ -84,8 +82,8 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            CMPS_IMM8(x3, 0);
                            B_MARK(cNE);
                            STRH_IMM8(x1, wback, 0); // put the whole value
+                            SMDMB();
                        }
-                        SMDMB();
                    }
                    break;
                default:
@@ -101,7 +99,7 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                case 0: // INC Ew
                    INST_NAME("INC Ew");
                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        GETEW(x1);
                        emit_inc16(dyn, ninst, ed, x3, x14);
                        EWBACK(x1);
@@ -118,9 +116,9 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            STREXH(x3, x1, wback);
                            CMPS_IMM8(x3, 0);
                            B_MARKLOCK(cNE);
+                            SMDMB();
                        }
                        if(!fixedaddress) {
-                            SMDMB();
                            B_NEXT(c__);
                        }
                        if(!fixedaddress || (fixedaddress && (fixedaddress&1))) {
@@ -133,13 +131,14 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            CMPS_IMM8(x3, 0);
                            B_MARK(cNE);
                            STRH_IMM8(x1, wback, 0);
+                            SMDMB();
                        }
                    }
                    break;
                case 1: //DEC Ew
                    INST_NAME("DEC Ew");
                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        GETEW(x1);
                        emit_dec32(dyn, ninst, ed, x3, x14);
                        EWBACK(x1);
@@ -156,9 +155,9 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            STREXH(x3, x1, wback);
                            CMPS_IMM8(x3, 0);
                            B_MARKLOCK(cNE);
+                            SMDMB();
                        }
                        if(!fixedaddress) {
-                            SMDMB();
                            B_NEXT(c__);
                        }
                        if(!fixedaddress || (fixedaddress && (fixedaddress&3))) {
@@ -171,6 +170,7 @@ uintptr_t dynarec66F0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                            CMPS_IMM8(x3, 0);
                            B_MARK(cNE);
                            STRH_IMM8(x1, wback, 0);
+                            SMDMB();
                        }
                    }
                    break;
@@ -50,7 +50,7 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    nextop = F8;
                    grab_fsdata(dyn, addr, ninst, x14);
                    GETGD;  // don't need GETGW here
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        if(ed!=gd) {
                            BFI(ed, gd, 0, 16);
@@ -67,7 +67,7 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    nextop=F8;
                    grab_fsdata(dyn, addr, ninst, x14);
                    GETGD;
-                    if((nextop&0xC0)==0xC0) {   // reg <= reg
+                    if(MODREG) {   // reg <= reg
                        ed = xEAX+(nextop&7);
                        if(ed!=gd) {
                            BFI(gd, ed, 0, 16);
@@ -83,7 +83,7 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    INST_NAME("POP FS:Ew16");
                    nextop=F8;
                    grab_fsdata(dyn, addr, ninst, x14);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        POP1(xEAX+(nextop&7));  // 67 ignored
                    } else {
                        addr = geted16(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
@@ -123,7 +123,7 @@ uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    switch((nextop>>3)&7) {
                        case 6: // Push Ed
                            INST_NAME("PUSH FS:Ew");
-                            if((nextop&0xC0)==0xC0) {   // reg
+                            if(MODREG) {   // reg
                                DEFAULT;
                            } else {                    // mem <= i32
                                SMREAD();
@@ -194,7 +194,7 @@ 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, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                VMOV_i_32(v1, 0b01110000);
            } else {
@@ -204,7 +204,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xE9:
            INST_NAME("FLDL2T");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_l2t));
                VLDR_32(v1, x2, 0);
@@ -216,7 +216,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEA:     
            INST_NAME("FLDL2E");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_l2e));
                VLDR_32(v1, x2, 0);
@@ -228,7 +228,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEB:
            INST_NAME("FLDPI");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_pi));
                VLDR_32(v1, x2, 0);
@@ -240,7 +240,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEC:
            INST_NAME("FLDLG2");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_lg2));
                VLDR_32(v1, x2, 0);
@@ -252,7 +252,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xED:
            INST_NAME("FLDLN2");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_ln2));
                VLDR_32(v1, x2, 0);
@@ -264,7 +264,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEE:
            INST_NAME("FLDZ");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                VMOV_8(v1/2, 0);  // float is *2...
            } else {
@@ -309,7 +309,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xF2:
            INST_NAME("FPTAN");
-            v2 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v2 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            v1 = x87_get_st(dyn, ninst, x1, x2, 1, NEON_CACHE_ST_D);
            VMOV_64(0, v1);    // prepare call to tan
            CALL_1D(tan, 0);
@@ -319,15 +319,21 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            LDRH_IMM8(x1, xEmu, offsetof(x86emu_t, sw));
            BFC(x1, 9, 2); //C1 C2 = 0 0
            STRH_IMM8(x1, xEmu, offsetof(x86emu_t, sw));
-            // so here: F64: Imm8 = abcd efgh that gives => aBbbbbbb bbcdefgh 0000000 00000000 00000000...
-            // and want 1.0 = 0x3ff0000000000000
-            // so 00111111 11110000 00000000 00000000....
-            // a = 0, b = 1, c = 1, d = 1, efgh=0
-            // 0b01110000
-            if(ST_IS_F(0)) {
-                VMOV_i_32(v2, 0b01110000);
+            if(PK(0)==0xdd && PK(1)==0xd8) {
+                MESSAGE(LOG_DUMP, "Optimized next DD D8 fstp st0, st0, not emiting 1\n");
+                u8 = F8;
+                u8 = F8;
            } else {
-                VMOV_i_64(v2, 0b01110000);
+                // so here: F64: Imm8 = abcd efgh that gives => aBbbbbbb bbcdefgh 0000000 00000000 00000000...
+                // and want 1.0 = 0x3ff0000000000000
+                // so 00111111 11110000 00000000 00000000....
+                // a = 0, b = 1, c = 1, d = 1, efgh=0
+                // 0b01110000
+                if(ST_IS_F(0)) {
+                    VMOV_i_32(v2, 0b01110000);
+                } else {
+                    VMOV_i_64(v2, 0b01110000);
+                }
            }
            break;
        case 0xF3:
@@ -589,7 +595,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2:
                    INST_NAME("FST float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    if(ST_IS_F(0))
                        s0 = v1;
                    else {
@@ -608,7 +614,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 3:
                    INST_NAME("FSTP float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    if(ST_IS_F(0))
                        s0 = v1;
                    else {
@@ -161,7 +161,7 @@ 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, x1, NEON_CACHE_ST_F);
+                    v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0, 0, NULL);
                    LDRSH_IMM8(x1, wback, fixedaddress);
                    if(ST_IS_F(0)) {
@@ -175,7 +175,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 1:
                    INST_NAME("FISTTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
                    s0 = fpu_get_scratch_single(dyn);
@@ -202,7 +202,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2:
                    INST_NAME("FIST Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    u8 = x87_setround_reset(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
@@ -224,7 +224,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 3:
                    INST_NAME("FISTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    u8 = x87_setround_reset(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
@@ -923,7 +923,7 @@ void emit_dec32(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)
 {
    IFX(X_PEND) {
-        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
        SET_DF(s3, d_dec8);
    } else IFX(X_ZF|X_OF|X_AF|X_SF|X_PF) {
        SET_DFNONE(s3);
@@ -154,7 +154,17 @@ void emit_shr32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
            }
            B_NEXT(cEQ);
    }
-    IFX(X_ZF|X_CF) {
+    IFX(X_CF) {
+        SUB_IMM8(s4, s2, 1);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, F_CF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s2==1
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 31);
+            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
+    }
+    IFX(X_ZF) {
        MOVS_REG_LSR_REG(s1, s1, s2);
    } else {
        MOV_REG_LSR_REG(s1, s1, s2);
@@ -162,25 +172,14 @@ void emit_shr32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
    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) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF), 0);
        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);
    }
@@ -203,7 +202,21 @@ void emit_shr32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s
        }
        return;
    }
-    IFX(X_ZF|X_CF) {
+    IFX(X_CF) {
+        if(c==1) {
+            BFI(xFlags, s1, F_CF, 1);
+        } else {
+            MOV_REG_LSR_IMM5(s3, s1, c-1);
+            BFI(xFlags, s3, F_CF, 1);
+        }
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 31);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+    IFX(X_ZF) {
        MOVS_REG_LSR_IMM5(s1, s1, c);
    } else {
        MOV_REG_LSR_IMM5(s1, s1, c);
@@ -211,26 +224,14 @@ void emit_shr32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s
    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) {
+        BIC_IMM8(xFlags, xFlags, (1<<F_ZF), 0);
        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);
    }
@@ -274,6 +275,565 @@ void emit_sar32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s
        MOV_REG_LSR_IMM5(s3, s1, 31);
        BFI(xFlags, s3, F_SF, 1);
    }
+    IFX(X_OF)
+        if(c==1) {
+            BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHL8 instruction, from s1 , shift s2, store result in s1 using s3 and s4 as scratch. s3 can be same as s2
+void emit_shl8(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_shl8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF | X_OF) {
+        RSB_IMM8(s4, s2, 8);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, F_CF, 1);
+    }
+    MOV_REG_LSL_REG(s1, s1, s2);
+    IFX(X_ZF) {
+        ANDS_IMM8(s1, s1, 0xff);
+    } else IFX(X_PEND) {
+        AND_IMM8(s1, s1, 0xff);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        MOV32(s4, 1);
+        BFI_COND(cEQ, xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s2==1
+            IFX(X_SF) {} else {MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 7);}
+            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);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHL8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shl8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_shl8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c<8) {
+        IFX(X_CF|X_OF) {
+            MOV_REG_LSR_IMM5(s3, s1, 8-c);
+            BFI(xFlags, s3, F_CF, 1);
+        }
+        MOV_REG_LSL_IMM5(s1, s1, c);
+
+        IFX(X_PEND) {
+            AND_IMM8(s1, s1, 0xff);
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        IFX(X_ZF) {
+            TSTS_IMM8(s1, 0xff);
+            MOVW_COND(cNE, s4, 0);
+            MOVW_COND(cEQ, s4, 1);
+            BFI(xFlags, s4, F_ZF, 1);
+        }
+        IFX(X_SF) {
+            MOV_REG_LSR_IMM5(s4, s1, 7);
+            BFI(xFlags, s4, F_SF, 1);
+        }
+        IFX(X_OF) {
+            if(c==1) {
+                IFX(X_SF) {} else {MOV_REG_LSR_IMM5(s4, s1, 7);}
+                XOR_REG_LSL_IMM5(s4, s4, xFlags, 0);  // CF is set if OF is asked
+                BFI(xFlags, s4, F_OF, 1);
+            } else {
+                BFC(xFlags, F_OF, 1);
+            }
+        }
+        IFX(X_PF) {
+            emit_pf(dyn, ninst, s1, s3, s4);
+        }
+    } else {
+        IFX(X_CF) {
+            MOV_REG_LSL_IMM5(s3, s1, c-1);
+            MOV_REG_LSR_IMM5(s4, s3, 7);
+            BFI(xFlags, s4, F_CF, 1);
+        }
+        MOVW(s1, 0);
+        IFX(X_OF) {
+            BFC(xFlags, F_OF, 1);
+        }
+        IFX(X_SF) {
+            BFC(xFlags, F_SF, 1);
+        }
+        IFX(X_PF | X_ZF) {
+            MOVW(s3, 1);
+            IFX(X_ZF) {
+                BFI(xFlags, s3, F_ZF, 1);
+            }
+            IFX(X_PF) {
+                BFI(xFlags, s3, F_PF, 1);
+            }
+        }
+    }
+}
+
+// emit SHR8 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s2 can be same as s3
+void emit_shr8(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_shr8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        SUB_IMM8(s4, s2, 1);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, 0, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s2==1
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 7);
+            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
+    }
+    MOV_REG_LSR_REG(s1, s1, s2);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s1, 0xff);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    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 SHR8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shr8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_shr8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        if(c==1) {
+            BFI(xFlags, s1, 0, 1);
+        } else {
+            MOV_REG_LSR_IMM5(s3, s1, c-1);
+            BFI(xFlags, s3, 0, 1);
+        }
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 7);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+    MOV_REG_LSR_IMM5(s1, s1, c);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s1, 0xff);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    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 SAR8 instruction, from s1 , shift s2, store result in s1 using s3 and s4 as scratch, s2 can be same as s3
+void emit_sar8(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_sar8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        SUB_IMM8(s4, s2, 1);
+        MOV_REG_ASR_REG(s4, s1, s4);
+        BFI(xFlags, s4, 0, 1);
+    }
+    MOV_REG_ASR_REG(s1, s1, s2);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s1, 0xff);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);
+            BFC_COND(cEQ, xFlags, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SAR8 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_sar8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_sar8);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        MOV_REG_ASR_IMM5(s3, s1, c-1);
+        BFI(xFlags, s3, 0, 1);
+    }
+    MOV_REG_ASR_IMM5(s1, s1, c);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_IMM8(s1, 0xff);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 7);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF)
+        if(c==1) {
+            BFC(xFlags, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHL16 instruction, from s1 , shift s2, store result in s1 using s3 and s4 as scratch. s3 can be same as s2
+void emit_shl16(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_shl16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF | X_OF) {
+        RSB_IMM8(s4, s2, 16);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, F_CF, 1);
+    }
+    MOV_REG_LSL_REG(s1, s1, s2);
+    UBFX(s1, s1, 0, 16);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_REG_LSL_IMM5(s1, s1, 0);
+        MOV32(s4, 1);
+        BFI_COND(cEQ, xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s2==1
+            IFX(X_SF) {} else {MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 15);}
+            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);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SHL16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shl16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_shl16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(c<16) {
+        IFX(X_CF|X_OF) {
+            MOV_REG_LSR_IMM5(s3, s1, 16-c);
+            BFI(xFlags, s3, F_CF, 1);
+        }
+        MOV_REG_LSL_IMM5(s1, s1, c);
+
+        UBFX(s1, s1, 0, 16);
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        IFX(X_ZF) {
+            TSTS_REG_LSL_IMM5(s1, s1, 0);
+            MOVW_COND(cNE, s4, 0);
+            MOVW_COND(cEQ, s4, 1);
+            BFI(xFlags, s4, F_ZF, 1);
+        }
+        IFX(X_SF) {
+            MOV_REG_LSR_IMM5(s4, s1, 15);
+            BFI(xFlags, s4, F_SF, 1);
+        }
+        IFX(X_OF) {
+            if(c==1) {
+                IFX(X_SF) {} else {MOV_REG_LSR_IMM5(s4, s1, 15);}
+                XOR_REG_LSL_IMM5(s4, s4, xFlags, 0);  // CF is set if OF is asked
+                BFI(xFlags, s4, F_OF, 1);
+            } else {
+                BFC(xFlags, F_OF, 1);
+            }
+        }
+        IFX(X_PF) {
+            emit_pf(dyn, ninst, s1, s3, s4);
+        }
+    } else {
+        IFX(X_CF) {
+            MOV_REG_LSL_IMM5(s3, s1, c-1);
+            MOV_REG_LSR_IMM5(s4, s3, 15);
+            BFI(xFlags, s4, F_CF, 1);
+        }
+        MOVW(s1, 0);
+        IFX(X_OF) {
+            BFC(xFlags, F_OF, 1);
+        }
+        IFX(X_SF) {
+            BFC(xFlags, F_SF, 1);
+        }
+        IFX(X_PF | X_ZF) {
+            MOVW(s3, 1);
+            IFX(X_ZF) {
+                BFI(xFlags, s3, F_ZF, 1);
+            }
+            IFX(X_PF) {
+                BFI(xFlags, s3, F_PF, 1);
+            }
+        }
+    }
+}
+
+// emit SHR16 instruction, from s1 , s2, store result in s1 using s3 and s4 as scratch, s2 can be same as s3
+void emit_shr16(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_shr16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        SUB_IMM8(s4, s2, 1);
+        MOV_REG_LSR_REG(s4, s1, s4);
+        BFI(xFlags, s4, 0, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);   // if s2==1
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 16);
+            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
+    }
+    MOV_REG_LSR_REG(s1, s1, s2);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_REG_LSL_IMM5(s1, s1, 0);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    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 SHR16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_shr16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_shr16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        if(c==1) {
+            BFI(xFlags, s1, 0, 1);
+        } else {
+            MOV_REG_LSR_IMM5(s3, s1, c-1);
+            BFI(xFlags, s3, 0, 1);
+        }
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 15);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+    MOV_REG_LSR_IMM5(s1, s1, c);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_REG_LSL_IMM5(s1, s1, 0);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    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 SAR16 instruction, from s1 , shift s2, store result in s1 using s3 and s4 as scratch, s2 can be same as s3
+void emit_sar16(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_sar16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        SUB_IMM8(s4, s2, 1);
+        MOV_REG_ASR_REG(s4, s1, s4);
+        BFI(xFlags, s4, 0, 1);
+    }
+    MOV_REG_ASR_REG(s1, s1, s2);
+    UBFX(s1, s1, 0, 16);
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_REG_LSL_IMM5(s1, s1, 0);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF) {
+        CMPS_IMM8(s2, 1);
+            BFC_COND(cEQ, xFlags, F_OF, 1);
+    }
+    IFX(X_PF) {
+        emit_pf(dyn, ninst, s1, s3, s4);
+    }
+}
+
+// emit SAR16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_sar16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4)
+{
+    if(!c)
+        return;
+    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_sar16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    IFX(X_CF) {
+        MOV_REG_ASR_IMM5(s3, s1, c-1);
+        BFI(xFlags, s3, 0, 1);
+    }
+    MOV_REG_ASR_IMM5(s1, s1, c);
+    IFX(X_PEND | X_ZF) {
+        UXTH(s1, s1, 0);
+    }
+    IFX(X_PEND) {
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_ZF) {
+        TSTS_REG_LSL_IMM5(s1, s1, 0);
+        MOVW_COND(cNE, s4, 0);
+        MOVW_COND(cEQ, s4, 1);
+        BFI(xFlags, s4, F_ZF, 1);
+    }
+    IFX(X_SF) {
+        MOV_REG_LSR_IMM5(s4, s1, 15);
+        BFI(xFlags, s4, F_SF, 1);
+    }
+    IFX(X_OF)
+        if(c==1) {
+            BFC(xFlags, F_OF, 1);
+    }
    IFX(X_PF) {
        emit_pf(dyn, ninst, s1, s3, s4);
    }
@@ -366,6 +926,12 @@ void emit_shrd32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int
        }
        return;
    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 31);
+            BFI(xFlags, s4, F_OF, 1);   // store sign for later use
+        }
+    }
    IFX(X_CF) {
        MOVS_REG_LSR_IMM5(s1, s1, c);
    } else {
@@ -394,9 +960,11 @@ void emit_shrd32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int
    }
    IFX(X_OF) {
        if(c==1) {
-            MOV_REG_LSR_IMM5(s4, s1, 30);
-            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            MOV_REG_LSR_IMM5(s4, s1, 31);
+            XOR_REG_LSR_IMM8(s4, s4, xFlags, F_OF); // set if sign changed
            BFI(xFlags, s4, F_OF, 1);
+        } else {
+            BFC(xFlags, F_OF, 1);
        }
    }
    IFX(X_PF) {
@@ -425,6 +993,12 @@ void emit_shld32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int
        }
        return;
    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 31);
+            BFI(xFlags, s4, F_OF, 1);   // store sign for later use
+        }
+    }
    IFX(X_CF) {
        MOV_REG_LSR_IMM5(s3, s1, 32-c);
        BFI(xFlags, s3, F_CF, 1);
@@ -436,9 +1010,9 @@ void emit_shld32c(dynarec_arm_t* dyn, int ninst, int s1, int s2, int32_t c, int
    }
    IFX(X_OF) {
        if(c==1) {
-            UBFX(s3, s1, (31-c), 1);
-            XOR_IMM8_COND(cCS, s3, s3, 1);
-            BFI(xFlags, s3, F_OF, 1);
+            MOV_REG_LSR_IMM5(s4, s1, 31);
+            XOR_REG_LSR_IMM8(s4, s4, xFlags, F_OF); // set if sign changed
+            BFI(xFlags, s4, F_OF, 1);
        } else {
            BFC(xFlags, F_OF, 1);
        }
@@ -474,7 +1048,7 @@ void emit_shrd32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
        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);
+        SET_DF(s4, d_shrd32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
    }
@@ -484,6 +1058,10 @@ void emit_shrd32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
        }
        B_NEXT(cEQ);

+    IFX(X_OF) {
+        MOV_REG_LSR_IMM5(s4, s1, 31);
+        BFI(xFlags, s4, F_OF, 1);   // store sign for later use
+    }
    IFX(X_CF) {
        MOVS_REG_LSR_REG(s1, s1, s3);
    } else {
@@ -513,8 +1091,8 @@ void emit_shrd32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
    }
    IFX(X_OF) {
        CMPS_IMM8(s3, 31);  // 32-c
-            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 30);
-            XOR_REG_LSR_IMM8_COND(cEQ, s4, s4, s4, 1);
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 31);
+            XOR_REG_LSR_IMM8_COND(cEQ, s4, s4, xFlags, F_OF);
            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
    }
    IFX(X_PF) {
@@ -531,7 +1109,7 @@ void emit_shld32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
        STR_IMM9(s1, xEmu, offsetof(x86emu_t, op1));
        STR_IMM9(s3, xEmu, offsetof(x86emu_t, op2));
        // same flags calc as shl32
-        SET_DF(s4, d_shl32);
+        SET_DF(s4, d_shld32);
    } else IFX(X_ALL) {
        SET_DFNONE(s4);
    }
@@ -544,6 +1122,10 @@ void emit_shld32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
        }
        B_NEXT(cEQ);

+    IFX(X_OF) {
+        MOV_REG_LSR_IMM5(s4, s1, 31);
+        BFI(xFlags, s4, F_OF, 1);   // store sign for later use
+    }
    IFX(X_CF) {
        RSB_IMM8(s4, s3, 32);
        MOV_REG_LSR_REG(s4, s1, s4);
@@ -555,13 +1137,10 @@ void emit_shld32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4)
        MOV_REG_LSL_REG(s1, s1, s3);
    }
    IFX(X_OF) {
-        MOVW_COND(cCS, s4, 1);
-        MOVW_COND(cCC, s4, 0);
-        CMPS_IMM8(s3, 1);
-            ORR_REG_LSR_IMM5_COND(cEQ, s4, s4, s1, 30);
+        CMPS_IMM8(s3, 31);  // 32-c
+            MOV_REG_LSR_IMM5_COND(cEQ, s4, s1, 31);
+            XOR_REG_LSR_IMM8_COND(cEQ, s4, s4, xFlags, F_OF);
            BFI_COND(cEQ, xFlags, s4, F_OF, 1);
-        //else
-            BFC_COND(cNE, xFlags, F_OF, 1);
    }
    RSB_IMM8(s3, s3, 32);
    IFX(X_ZF) {
@@ -602,8 +1181,8 @@ void emit_rol8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
        return;
    }
    if(c&7) {
-        BFI(s1, s1, 24, 8);  // duplicate the value for rotate left
-        MOV_REG_ROR_IMM5(s1, s1, 32-(c&7));
+        BFI(s1, s1, 8, 8);
+        MOV_REG_LSR_IMM5(s1, s1, 8-(c&7));
    }
    //AND_IMM8(s1, s1, 0xff);
    IFX(X_PEND) {
@@ -614,7 +1193,7 @@ void emit_rol8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
    }
    IFX(X_OF) {
        if(c==1) {
-            ADD_REG_LSR_IMM5(s3, s1, s1, 7);
+            XOR_REG_LSR_IMM8(s3, s1, s1, 7);
            BFI(xFlags, s3, F_OF, 1);
        }
    }
@@ -638,7 +1217,7 @@ void emit_ror8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
    }
    if(c&7) {
        BFI(s1, s1, 8, 8);  // duplicate the value for Rotate right
-        MOV_REG_ROR_IMM5(s1, s1, c&7);
+        MOV_REG_LSR_IMM5(s1, s1, c&7);
    }
    IFX(X_PEND) {
        AND_IMM8(s1, s1, 0xff);
@@ -656,3 +1235,75 @@ void emit_ror8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4
        }
    }
 }
+
+// emit ROL16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_rol16c(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_rol16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    if(c&15) {
+        BFI(s1, s1, 16, 16);
+        MOV_REG_LSR_IMM5(s1, s1, 16-(c&15));
+    }
+    IFX(X_PEND) {
+        UBFX(s1, s1, 0, 16);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF) {
+        BFI(xFlags, s1, F_CF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            XOR_REG_LSR_IMM8(s3, s1, s1, 15);
+            BFI(xFlags, s3, F_OF, 1);
+        }
+    }
+}
+
+// emit ROR16 instruction, from s1 , constant c, store result in s1 using s3 and s4 as scratch
+void emit_ror16c(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_ror16);
+    } else IFX(X_ALL) {
+        SET_DFNONE(s4);
+    }
+    if(!c) {
+        IFX(X_PEND) {
+            STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+        }
+        return;
+    }
+    if(c&15) {
+        BFI(s1, s1, 16, 16);  // duplicate the value for Rotate right
+        MOV_REG_LSR_IMM5(s1, s1, c&15);
+    }
+    IFX(X_PEND) {
+        UBFX(s1, s1, 0, 16);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, res));
+    }
+    IFX(X_CF) {
+        MOV_REG_LSR_IMM5(s3, s1, 15);
+        BFI(xFlags, s3, F_CF, 1);
+    }
+    IFX(X_OF) {
+        if(c==1) {
+            MOV_REG_LSR_IMM5(s4, s1, 14);
+            XOR_REG_LSR_IMM8(s4, s4, s4, 1);
+            BFI(xFlags, s4, F_OF, 1);
+        }
+    }
+}
@@ -42,8 +42,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGB(x2);
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                wback = (nextop&7);    
                wb2 = (wback>>2);      
                wback = xEAX+(wback&3);
@@ -66,8 +65,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_add32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -85,8 +83,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("LOCK OR Eb, Gb");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGB(x2);
                wback = (nextop&7);    
                wb2 = (wback>>2);      
@@ -111,8 +108,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_or32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -132,8 +128,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGB(x2);
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                wback = (nextop&7);    
                wb2 = (wback>>2);      
                wback = xEAX+(wback&3);
@@ -157,8 +152,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_adc32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -178,8 +172,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGB(x2);
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                wback = (nextop&7);    
                wb2 = (wback>>2);      
                wback = xEAX+(wback&3);
@@ -203,8 +196,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_sbb32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -222,8 +214,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("LOCK AND Eb, Gb");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGB(x2);
                wback = (nextop&7);    
                wb2 = (wback>>2);      
@@ -248,8 +239,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_and32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -268,8 +258,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGB(x2);
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                wback = (nextop&7);    
                wb2 = (wback>>2);      
                wback = xEAX+(wback&3);
@@ -292,8 +281,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_sub32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -311,8 +299,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("LOCK XOR Eb, Gb");
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                GETGB(x2);
                wback = (nextop&7);    
                wb2 = (wback>>2);      
@@ -337,8 +324,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SETFLAGS(X_ALL, SF_SET_PENDING);
            nextop = F8;
            GETGD;
-            SMDMB();
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                ed = xEAX+(nextop&7);
                emit_xor32(dyn, ninst, ed, gd, x3, x14);
            } else {
@@ -368,8 +354,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        INST_NAME("LOCK ADD Ed, Ib");
                    }
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_add32c(dyn, ninst, ed, i32, x3, x14);
@@ -409,8 +394,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 1: //OR
                    if(opcode==0x81) {INST_NAME("LOCK OR Ed, Id");} else {INST_NAME("LOCK OR Ed, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_or32c(dyn, ninst, ed, i32, x3, x14);
@@ -430,8 +414,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    if(opcode==0x81) {INST_NAME("LOCK ADC Ed, Id");} else {INST_NAME("LOCK ADC Ed, Ib");}
                    READFLAGS(X_CF);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_adc32c(dyn, ninst, ed, i32, x3, x14);
@@ -451,8 +434,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    if(opcode==0x81) {INST_NAME("LOCK SBB Ed, Id");} else {INST_NAME("LOCK SBB Ed, Ib");}
                    READFLAGS(X_CF);
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_sbb32c(dyn, ninst, ed, i32, x3, x14);
@@ -471,8 +453,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 4: //AND
                    if(opcode==0x81) {INST_NAME("LOCK AND Ed, Id");} else {INST_NAME("LOCK AND Ed, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_and32c(dyn, ninst, ed, i32, x3, x14);
@@ -491,12 +472,11 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 5: //SUB
                    if(opcode==0x81) {INST_NAME("LOCK SUB Ed, Id");} else {INST_NAME("LOCK SUB Ed, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_sub32c(dyn, ninst, ed, i32, x3, x14);
                    } else {
-                        SMDMB();
                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, 1, LOCK_LOCK);
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        if(!fixedaddress) {
@@ -532,12 +512,11 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 6: //XOR
                    if(opcode==0x81) {INST_NAME("LOCK XOR Ed, Id");} else {INST_NAME("LOCK XOR Ed, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        ed = xEAX+(nextop&7);
                        emit_xor32c(dyn, ninst, ed, i32, x3, x14);
                    } else {
-                        SMDMB();
                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0, 0, LOCK_LOCK);
                        if(opcode==0x81) i32 = F32S; else i32 = F8S;
                        MARKLOCK;
@@ -552,7 +531,6 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 7: //CMP
                    if(opcode==0x81) {INST_NAME("(LOCK) CMP Ed, Id");} else {INST_NAME("(LOCK) CMP Ed, Ib");}
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
                    GETEDH(x1);
                    // No need to LOCK, this is readonly
                    if(opcode==0x81) i32 = F32S; else i32 = F8S;
@@ -566,7 +544,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            }
            break;

-        // generic case, no lock needed, the value is only read (note that on x86 locked read is always followed wy a locked write)
+        // generic case, no lock needed, the value is only read (note that on x86 locked read is always followed by a locked write)
        #define GO(A)   \
        case A+0x02:    \
        case A+0x03:    \
@@ -585,17 +563,6 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            SMDMB();
            --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:
-            SMDMB();
-            --addr;
-            break;
    
        case 0x0F:
            nextop = F8;
@@ -603,9 +570,8 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 0xB0:
                    INST_NAME("LOCK CMPXCHG Eb, Gb");
                    SETFLAGS(X_ALL, SF_SET_PENDING);
-                    SMDMB();
                    nextop = F8;
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        wback = (nextop&7);    
                        wb2 = (wback>>2);      
                        wback = xEAX+(wback&3);
@@ -649,8 +615,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    nextop = F8;
                    GETGD;
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        // same as no lock, all in regs...
                        CMPS_REG_LSL_IMM5(xEAX, ed, 0);
@@ -721,8 +686,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    SET_DFNONE(x1);
                    nextop = F8;
                    GETGD;
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -736,30 +700,26 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    }
                    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);
+                    BIC_REG_LSL_REG(ed, ed, x14, x2);
                    if(wback) {
                        STREX(x14, ed, wback);
                        CMPS_IMM8(x14, 0);
                        B_MARKLOCK(cNE);
                    }
-                    MARK3;
                    SMDMB();
                    break;

                case 0xBA:
                    nextop = F8;
-                    SMDMB();
                    switch((nextop>>3)&7) {
                        case 4:
                            INST_NAME("(LOCK) BT Ed, Ib");
                            SETFLAGS(X_CF, SF_SUBSET);
                            SET_DFNONE(x1);
                            gd = x2;
-                            if((nextop&0xC0)==0xC0) {
+                            if(MODREG) {
                                ed = xEAX+(nextop&7);
                            } else {
                                addr = geted(dyn, addr, ninst, nextop, &ed, x3, &fixedaddress, 4095-32, 0, 0, LOCK_LOCK);
@@ -778,26 +738,21 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            INST_NAME("LOCK BTS Ed, Ib");
                            SETFLAGS(X_CF, SF_SUBSET);
                            SET_DFNONE(x1);
-                            gd = x2;
-                            if((nextop&0xC0)==0xC0) {
+                            if(MODREG) {
                                ed = xEAX+(nextop&7);
-                                u8 = F8;
-                                MOVW(gd, u8&0x1f);
+                                u8 = F8&0x1f;
                                wback = 0;
                            } else {
                                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, 0, LOCK_LOCK);
-                                u8 = F8;
+                                u8 = F8&0x1f;
                                ed = x1;
-                                MOVW(gd, u8&0x1f);
                                MARKLOCK;
                                LDREX(ed, wback);
                            }
-                            MOV_REG_LSR_REG(x14, ed, x2);
-                            ANDS_IMM8(x14, x14, 1);
+                            MOV_REG_LSR_IMM5(x14, ed, u8);
                            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);
+                            BFI(ed, x14, u8, 1);
                            if(wback) {
                                STREX(x14, ed, wback);
                                CMPS_IMM8(x14, 0);
@@ -809,26 +764,46 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            INST_NAME("LOCK BTR Ed, Ib");
                            SETFLAGS(X_CF, SF_SUBSET);
                            SET_DFNONE(x1);
-                            gd = x2;
-                            if((nextop&0xC0)==0xC0) {
+                            if(MODREG) {
                                ed = xEAX+(nextop&7);
-                                u8 = F8;
-                                MOVW(gd, u8&0x1f);
+                                u8 = F8&0x1f;
                                wback = 0;
                            } else {
                                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, 0, LOCK_LOCK);
-                                u8 = F8;
+                                u8 = F8&0x1f;
                                ed = x1;
-                                MOVW(gd, u8&0x1f);
                                MARKLOCK;
                                LDREX(ed, wback);
                            }
-                            MOV_REG_LSR_REG(x14, ed, x2);
-                            ANDS_IMM8(x14, x14, 1);
+                            MOV_REG_LSR_IMM5(x14, ed, u8);
                            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);
+                            BFC(ed, u8, 1);
+                            if(wback) {
+                                STREX(x14, ed, wback);
+                                CMPS_IMM8(x14, 0);
+                                B_MARKLOCK(cNE);
+                            }
+                            MARK3;
+                            break;
+                        case 7:
+                            INST_NAME("LOCK BTC Ed, Ib");
+                            SETFLAGS(X_CF, SF_SUBSET);
+                            SET_DFNONE(x1);
+                            if(MODREG) {
+                                ed = xEAX+(nextop&7);
+                                u8 = F8&0x1f;
+                                wback = 0;
+                            } else {
+                                addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 0, 0, 0, LOCK_LOCK);
+                                u8 = F8&0x1f;
+                                ed = x1;
+                                MARKLOCK;
+                                LDREX(ed, wback);
+                            }
+                            MOV_REG_LSR_IMM5(x14, ed, u8);
+                            BFI(xFlags, x14, F_CF, 1);
+                            MOV32(x14, 1);
+                            XOR_REG_LSL_IMM5(ed, ed, x14, u8);
                            if(wback) {
                                STREX(x14, ed, wback);
                                CMPS_IMM8(x14, 0);
@@ -847,8 +822,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    SET_DFNONE(x1);
                    nextop = F8;
                    GETGD;
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        wback = 0;
                    } else {
@@ -879,7 +853,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    nextop = F8;
                    SMDMB();
                    GETGB(x1);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        wback = (nextop&7); 
                        wb2 = (wback>>2); 
                        wback = xEAX+(wback&3); 
@@ -909,8 +883,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    SETFLAGS(X_ALL, SF_SET_PENDING);
                    nextop = F8;
                    GETGD;
-                    SMDMB();
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        if(gd!=ed) {
                            XOR_REG_LSL_IMM5(gd, gd, ed, 0);    // swap gd, ed
@@ -964,7 +937,6 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                        case 1:
                            INST_NAME("LOCK CMPXCHG8B Gq, Eq");
                            SETFLAGS(X_ZF, SF_SUBSET);
-                            SMDMB();
                            addr = geted(dyn, addr, ninst, nextop, &wback, x1, &fixedaddress, 0, 0, 1, LOCK_LOCK);
                            MARKLOCK;
                            if(!fixedaddress || (fixedaddress && (fixedaddress&7))) {
@@ -1010,8 +982,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;

                default:
-                    // dafault to NO LOCK
-                    addr-=2;
+                    DEFAULT;
            }
            break;

@@ -1023,7 +994,7 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 0: // INC Ed
                    INST_NAME("INC Ed");
                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        emit_inc32(dyn, ninst, ed, x3, x14);
                    } else {
@@ -1039,9 +1010,9 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            STREX(x3, x1, wback);
                            CMPS_IMM8(x3, 0);
                            B_MARKLOCK(cNE);
+                            SMDMB();
                        }
                        if(!fixedaddress) {
-                            SMDMB();
                            B_NEXT(c__);
                        }
                        if(!fixedaddress || (fixedaddress && (fixedaddress&3))) {
@@ -1054,13 +1025,14 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            CMPS_IMM8(x3, 0);
                            B_MARK(cNE);
                            STR_IMM9(x1, wback, 0);
+                            SMDMB();
                        }
                    }
                    break;
                case 1: //DEC Ed
                    INST_NAME("DEC Ed");
                    SETFLAGS(X_ALL&~X_CF, SF_SUBSET_PENDING);
-                    if((nextop&0xC0)==0xC0) {
+                    if(MODREG) {
                        ed = xEAX+(nextop&7);
                        emit_dec32(dyn, ninst, ed, x3, x14);
                    } else {
@@ -1076,9 +1048,9 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            STREX(x3, x1, wback);
                            CMPS_IMM8(x3, 0);
                            B_MARKLOCK(cNE);
+                            SMDMB();
                        }
                        if(!fixedaddress) {
-                            SMDMB();
                            B_NEXT(c__);
                        }
                        if(!fixedaddress || (fixedaddress && (fixedaddress&3))) {
@@ -1091,19 +1063,18 @@ uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                            CMPS_IMM8(x3, 0);
                            B_MARK(cNE);
                            STR_IMM9(x1, wback, 0);
+                            SMDMB();
                        }
                    }
                    break;
                default:
-                    // default to NO LOCK
-                    addr-=2;
+                    DEFAULT;
            }
            SMDMB();
            break;
       
        default:
-            // default to NO LOCK
-            --addr;
+            DEFAULT;
    }

    return addr;
@@ -24,7 +24,7 @@

 // Get Ex as a double, not a quad (warning, x2 and x3 may get used)
 #define GETEX(a, w)                                 \
-    if((nextop&0xC0)==0xC0) {                       \
+    if(MODREG) {                       \
        a = sse_get_reg(dyn, ninst, x1, nextop&7,w);\
    } else {    \
        parity = getedparity(dyn, ninst, addr, nextop, 3);  \
@@ -68,7 +68,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVSD Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                VMOVD(v0, d0);
@@ -87,7 +87,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                VMOVD(d0, v0);
            } else {
@@ -102,7 +102,7 @@ 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;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVD(v0, d0);
@@ -133,7 +133,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -165,7 +165,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                d0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -328,7 +328,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -364,7 +364,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -434,7 +434,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -460,7 +460,7 @@ uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("LDDQ Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -24,7 +24,7 @@

 // Get Ex as a single, not a quad, with b a intermediary double
 #define GETEX(a, b)                                 \
-    if((nextop&0xC0)==0xC0) {                       \
+    if(MODREG) {                       \
        b = sse_get_reg(dyn, ninst, x1, nextop&7,0);\
        if(b<16)                                    \
            a = b;                                  \
@@ -81,7 +81,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVSS Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = sse_get_reg(dyn, ninst, x1, gd, 1);
                q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                if(v0<16)
@@ -121,7 +121,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q0 = sse_get_reg(dyn, ninst, x1, nextop&7, 1);
                if(v0<16)
                    d0 = v0;
@@ -156,7 +156,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVSLDUP Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -173,7 +173,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVSHDUP Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                q1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -223,7 +223,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                if(v0<16)
                    d0 = v0;
@@ -268,7 +268,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            GETGD;
            s0 = fpu_get_scratch_single(dyn);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v0 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                if(v0<16)
                    d0 = v0;
@@ -420,7 +420,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
        case 0x5B:
            INST_NAME("CVTTPS2DQ Gx, Ex");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -523,7 +523,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVDQU Gx,Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVQ(v0, v1);
@@ -543,7 +543,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("PSHUFHW Gx, Ex, Ib");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
            } else {
                SMREAD();
@@ -579,7 +579,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVQ Gx, Ex");
            nextop = F8;
            gd = (nextop&0x38)>>3;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                VMOVD(v0, v1);
@@ -596,7 +596,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            nextop = F8;
            gd = (nextop&0x38)>>3;
            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
                VMOVQ(v1, v0);
            } else {
@@ -666,7 +666,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            INST_NAME("MOVQ2DQ Gx, Em");
            nextop = F8;
            GETGX_empty(v0);
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = mmx_get_reg(dyn, ninst, x1, x2, x3, (nextop&7));
                VEORQ(v0, v0, v0);  // usefull?
                VMOV_64(v0, v1);
@@ -688,7 +688,7 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
        case 0xE6:
            INST_NAME("CVTDQ2PD Gx, Ex");
            nextop = F8;
-            if((nextop&0xC0)==0xC0) {
+            if(MODREG) {
                v1 = sse_get_reg(dyn, ninst, x1, nextop&7, 0);
                if(v1<16)
                    v0 = v1;
@@ -27,6 +27,8 @@
 #include "dynarec_arm_private.h"
 #include "dynarec_arm_functions.h"
 #include "bridge.h"
+#include "arm_printer.h"
+#include "custommem.h"

 void arm_fstp(x86emu_t* emu, void* p)
 {
@@ -67,12 +69,24 @@ void arm_fxtract(x86emu_t* emu)
 }
 void arm_fprem(x86emu_t* emu)
 {
-    int64_t tmp64s = ST0.d / ST1.d;
-    ST0.d -= ST1.d * tmp64s;
-    emu->sw.f.F87_C2 = 0;
-    emu->sw.f.F87_C1 = (tmp64s&1);
-    emu->sw.f.F87_C3 = ((tmp64s>>1)&1);
-    emu->sw.f.F87_C0 = ((tmp64s>>2)&1);
+    int e0, e1;
+    int64_t ll;
+    frexp(ST0.d, &e0);
+    frexp(ST1.d, &e1);
+    int32_t 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;
+    }
 }
 void arm_frndint(x86emu_t* emu)
 {
@@ -109,7 +123,7 @@ void arm_fistp64(x86emu_t* emu, int64_t* ed)
        memcpy(ed, &STll(0).sq, sizeof(int64_t));
    } else {
        int64_t tmp;
-        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL) || !isfinite(ST0.d))
+        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(ST0.d))
            tmp = 0x8000000000000000LL;
        else
            tmp = fpu_round(emu, ST0.d);
@@ -120,7 +134,7 @@ void arm_fistp64(x86emu_t* emu, int64_t* ed)
 int64_t arm_fist64_0(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = nearbyint(d);
@@ -129,7 +143,7 @@ int64_t arm_fist64_0(double d)
 int64_t arm_fist64_1(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = floor(d);
@@ -138,7 +152,7 @@ int64_t arm_fist64_1(double d)
 int64_t arm_fist64_2(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = ceil(d);
@@ -147,7 +161,7 @@ int64_t arm_fist64_2(double d)
 int64_t arm_fist64_3(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = trunc(d);
@@ -170,9 +184,22 @@ void arm_fld(x86emu_t* emu, uint8_t* ed)

 void arm_ud(x86emu_t* emu)
 {
+    emu->test.test = 0;
    emit_signal(emu, SIGILL, (void*)R_EIP, 0);
 }

+void arm_priv(x86emu_t* emu)
+{
+    emu->test.test = 0;
+    emit_signal(emu, SIGSEGV, (void*)R_EIP, 128);
+}
+
+void arm_singlestep(x86emu_t* emu)
+{
+    emu->test.test = 0;
+    emit_signal(emu, SIGTRAP, (void*)R_EIP, 2);
+}
+
 void arm_fsave(x86emu_t* emu, uint8_t* ed)
 {
    fpu_savenv(emu, (char*)ed, 0);
@@ -199,15 +226,185 @@ void arm_frstor(x86emu_t* emu, uint8_t* ed)

 void arm_fprem1(x86emu_t* emu)
 {
-    // simplified version
-    int64_t tmp64s = lrint(ST0.d / ST1.d);
-    ST0.d -= ST1.d*tmp64s;
-    emu->sw.f.F87_C2 = 0;
-    emu->sw.f.F87_C1 = (tmp64s&1);
-    emu->sw.f.F87_C3 = ((tmp64s>>1)&1);
-    emu->sw.f.F87_C0 = ((tmp64s>>2)&1);
+    int e0, e1;
+    int64_t ll;
+    frexp(ST0.d, &e0);
+    frexp(ST1.d, &e1);
+    int32_t 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;
+    }
 }

+static uint8_t ff_mult(uint8_t a, uint8_t b)
+{
+	int retval = 0;
+
+	for(int i = 0; i < 8; i++) {
+		if((b & 1) == 1)
+			retval ^= a;
+
+		if((a & 0x80)) {
+			a <<= 1;
+			a  ^= 0x1b;
+		} else {
+			a <<= 1;
+		}
+
+		b >>= 1;
+	}
+
+	return retval;
+}
+
+void arm_aesimc(x86emu_t* emu, int xmm)
+{
+    sse_regs_t eax1 = emu->xmm[xmm];
+
+    for(int j=0; j<4; ++j) {
+        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x0E, eax1.ub[0+j*4]) ^ ff_mult(0x0B, eax1.ub[1+j*4]) ^ ff_mult(0x0D, eax1.ub[2+j*4]) ^ ff_mult(0x09, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[1+j*4] = ff_mult(0x09, eax1.ub[0+j*4]) ^ ff_mult(0x0E, eax1.ub[1+j*4]) ^ ff_mult(0x0B, eax1.ub[2+j*4]) ^ ff_mult(0x0D, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[2+j*4] = ff_mult(0x0D, eax1.ub[0+j*4]) ^ ff_mult(0x09, eax1.ub[1+j*4]) ^ ff_mult(0x0E, eax1.ub[2+j*4]) ^ ff_mult(0x0B, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x0B, eax1.ub[0+j*4]) ^ ff_mult(0x0D, eax1.ub[1+j*4]) ^ ff_mult(0x09, eax1.ub[2+j*4]) ^ ff_mult(0x0E, eax1.ub[3+j*4]);
+    }
+}
+void arm_aesmc(x86emu_t* emu, int xmm)
+{
+    sse_regs_t eax1 = emu->xmm[xmm];
+
+    for(int j=0; j<4; ++j) {
+        emu->xmm[xmm].ub[0+j*4] = ff_mult(0x02, eax1.ub[0+j*4]) ^ ff_mult(0x03, eax1.ub[1+j*4]) ^               eax1.ub[2+j*4]  ^               eax1.ub[3+j*4] ;
+        emu->xmm[xmm].ub[1+j*4] =               eax1.ub[0+j*4]  ^ ff_mult(0x02, eax1.ub[1+j*4]) ^ ff_mult(0x03, eax1.ub[2+j*4]) ^               eax1.ub[3+j*4] ;
+        emu->xmm[xmm].ub[2+j*4] =               eax1.ub[0+j*4]  ^               eax1.ub[1+j*4]  ^ ff_mult(0x02, eax1.ub[2+j*4]) ^ ff_mult(0x03, eax1.ub[3+j*4]);
+        emu->xmm[xmm].ub[3+j*4] = ff_mult(0x03, eax1.ub[0+j*4]) ^               eax1.ub[1+j*4]  ^               eax1.ub[2+j*4]  ^ ff_mult(0x02, eax1.ub[3+j*4]);
+    }
+}
+void arm_aesdlast(x86emu_t* emu, int xmm)
+{
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  N  K  H  E  B  O  L  I  F  C  P  M  J  G  D
+    const uint8_t invshiftrows[] = {0,13,10, 7, 4, 1,14,11, 8, 5, 2,15,12, 9, 6, 3};
+    const uint8_t invsubbytes[256] = {
+        0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38, 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
+        0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87, 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
+        0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d, 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
+        0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2, 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
+        0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
+        0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda, 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
+        0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a, 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
+        0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02, 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
+        0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea, 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
+        0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85, 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
+        0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89, 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
+        0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20, 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
+        0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31, 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
+        0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d, 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
+        0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0, 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
+        0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26, 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
+    };
+
+    sse_regs_t eax1;
+    for(int i=0; i<16; ++i)
+        eax1.ub[i] = emu->xmm[xmm].ub[invshiftrows[i]];
+    //STATE ← InvSubBytes( STATE );
+    for(int i=0; i<16; ++i)
+        emu->xmm[xmm].ub[i] = invsubbytes[eax1.ub[i]];
+
+}
+static const uint8_t shiftrows[] = {0, 5,10,15, 4, 9,14, 3, 8,13, 2, 7,12, 1, 6,11};
+static const uint8_t subbytes[256] = {
+    0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
+    0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
+    0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
+    0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
+    0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
+    0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
+    0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
+    0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
+    0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
+    0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
+    0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
+    0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
+    0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
+    0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
+    0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
+    0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
+};
+void arm_aeselast(x86emu_t* emu, int xmm)
+{
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  F  K  P  E  J  O  D  I  N  C  H  M  B  G  L
+    sse_regs_t eax1;
+    for(int i=0; i<16; ++i)
+        eax1.ub[i] = emu->xmm[xmm].ub[shiftrows[i]];
+    //STATE ← SubBytes( STATE );
+    for(int i=0; i<16; ++i)
+        emu->xmm[xmm].ub[i] = subbytes[eax1.ub[i]];
+}
+void arm_aesd(x86emu_t* emu, int xmm)
+{
+    arm_aesdlast(emu, xmm);
+    arm_aesimc(emu, xmm);
+}
+void arm_aese(x86emu_t* emu, int xmm)
+{
+    arm_aeselast(emu, xmm);
+    arm_aesmc(emu, xmm);
+}
+void arm_aeskeygenassist(x86emu_t* emu, int gx, void* p, uint32_t u8)
+{
+    sse_regs_t *EX = ((uintptr_t)p>8)?((sse_regs_t*)p):&emu->xmm[(uintptr_t)p];
+    sse_regs_t *GX = &emu->xmm[gx];
+    for (int i = 4; i < 8; ++i)
+        GX->ub[i] = subbytes[EX->ub[i]];
+    for (int i = 12; i < 16; ++i)
+        GX->ub[i] = subbytes[EX->ub[i]];
+    GX->ud[0] = GX->ud[1];
+    uint8_t tmp8u = GX->ub[4];
+    GX->ud[1] = GX->ud[1] >> 8;
+    GX->ub[7] = tmp8u;
+    GX->ud[1] ^= u8;
+    GX->ud[2] = GX->ud[3];
+    tmp8u = GX->ub[12];
+    GX->ud[3] = GX->ud[3] >> 8;
+    GX->ub[15] = tmp8u;
+    GX->ud[3] ^= u8;
+}
+
+void arm_pclmul(x86emu_t* emu, int gx, void* p, uint32_t u8)
+{
+    int ex = (((uintptr_t)p)<8)?((uintptr_t)p):-1;
+    sse_regs_t *EX = (ex==-1)?((sse_regs_t*)p):&emu->xmm[ex];
+    sse_regs_t *GX = &emu->xmm[gx];
+    int g = (u8&1)?1:0;
+    int e = (u8&0b10000)?1:0;
+    uint64_t result[2] = {0};
+    uint64_t op2 = EX->q[e];
+    for (int i=0; i<64; ++i)
+        if(GX->q[g]&(1LL<<i)) {
+            result[0] ^= (EX->q[e]<<i);
+            if(i)
+                result[1] ^= (EX->q[e]>>(64-i));
+
+        }
+    GX->q[0] = result[0];
+    GX->q[1] = result[1];
+}
+
+void arm_clflush(x86emu_t* emu, void* p)
+{
+    cleanDBFromAddressRange((uintptr_t)p, 8, 0);
+}

 // Get a FPU single scratch reg
 int fpu_get_scratch_single(dynarec_arm_t* dyn)
@@ -608,6 +805,7 @@ void neoncacheUnwind(neoncache_t* cache)
                case NEON_CACHE_XMMW:
                    cache->ssecache[cache->neoncache[i].n].reg = i+FPUFIRST;
                    cache->ssecache[cache->neoncache[i].n].write = (cache->neoncache[i].t==NEON_CACHE_XMMW)?1:0;
+                    cache->neoncache[i] = cache->neoncache[i+1];
                    ++cache->fpu_reg;
                    ++i;    // next next
                    break;
@@ -694,6 +892,12 @@ uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop)
    }
    return addr;
 }
+// return Ib on a mod/rm opcode without emiting anything
+uint8_t geted_ib(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop)
+{
+    addr = fakeed(dyn, addr, ninst, nextop);
+    return F8;
+}
 #undef F8
 #undef F32

@@ -702,13 +906,13 @@ int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int
 #define PK(a)       *(uint8_t*)(addr+a)
 #define PK32(a)     *(uint32_t*)(addr+a)

-    if(!addr)
+    if(!addr || !getProtection(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 = (uintptr_t)getAlternate(*(void**)a1); 
    }
-    if(addr<0x10000)    // too low, that is suspicious
+    if(addr<0x10000 || !getProtection(addr))    // 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) {
@@ -737,62 +941,90 @@ const char* getCacheName(int t, int n)
    return buff;
 }

-// is inst clean for a son branch?
-int isInstClean(dynarec_arm_t* dyn, int ninst)
-{
-    // check flags cache
-    if(dyn->insts[ninst].f_entry.dfnone || dyn->insts[ninst].f_entry.pending)
-        return 0;
-    if(dyn->insts[ninst].x86.state_flags)
-        return 0;
-    // check neoncache
-    neoncache_t* n = &dyn->insts[ninst].n;
-    if(n->news || n->stack || n->stack_next)
-        return 0;
-    for(int i=0; i<24; ++i)
-        if(n->neoncache[i].v)
-            return 0;
-    return 1;
-}
-
 static int flagsCacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
    int jmp = dyn->insts[ninst].x86.jmp_insts;
    if(jmp<0)
        return 0;
    if(dyn->insts[ninst].f_exit.dfnone)  // flags are fully known, nothing we can do more
        return 0;
-/*    if((dyn->f.pending!=SF_SET)
-    && (dyn->f.pending!=SF_SET_PENDING)) {
-        if(dyn->f.pending!=SF_PENDING) {*/
-    switch (dyn->insts[jmp].f_entry.pending) {
-        case SF_UNKNOWN: return 0;
-        case SF_SET: 
-            if(dyn->insts[ninst].f_exit.pending!=SF_SET && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING) 
-                return 1; 
-            else 
-                return 0;
-        case SF_SET_PENDING:
-            if(dyn->insts[ninst].f_exit.pending!=SF_SET 
-            && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING
-            && dyn->insts[ninst].f_exit.pending!=SF_PENDING) 
-                return 1; 
-            else 
-                return 0;
-        case SF_PENDING:
-            if(dyn->insts[ninst].f_exit.pending!=SF_SET 
-            && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING
-            && dyn->insts[ninst].f_exit.pending!=SF_PENDING)
-                return 1;
-            else
-                return (dyn->insts[jmp].f_entry.dfnone  == dyn->insts[ninst].f_exit.dfnone)?0:1;
-    }
    if(dyn->insts[jmp].f_entry.dfnone && !dyn->insts[ninst].f_exit.dfnone)
        return 1;
+    switch (dyn->insts[jmp].f_entry.pending) {
+        case SF_UNKNOWN: return 0;
+        case SF_SET:
+            if(dyn->insts[ninst].f_exit.pending!=SF_SET && dyn->insts[ninst].f_exit.pending!=SF_SET_PENDING)
+                return 1;
+            else
+                return 0;
+        case SF_SET_PENDING:
+            if(dyn->insts[ninst].f_exit.pending==SF_SET_PENDING)
+                return 0;
+            return 1;
+        case SF_PENDING:
+            if(dyn->insts[ninst].f_exit.pending==SF_PENDING || dyn->insts[ninst].f_exit.pending==SF_SET_PENDING)
+                return 0;
+            return (dyn->insts[jmp].f_entry.dfnone  == dyn->insts[ninst].f_exit.dfnone)?0:1;
+    }
    return 0;
 }
+
 int CacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
    int ret = 0;
    if (fpuCacheNeedsTransform(dyn, ninst)) ret|=1;
    if (flagsCacheNeedsTransform(dyn, ninst)) ret|=2;
    return ret;
-}
+}
+
+void inst_name_pass3(dynarec_arm_t* dyn, int ninst, const char* 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, inst=%d, barrier=%d state=%d/%d(%d), %s=%X/%X, use=%X, need=%X/%X sm=%d/%d",
+            (box86_dynarec_dump>1)?"\e[32m":"",
+            (void*)(dyn->arm_start+dyn->insts[ninst].address),
+            dyn->insts[ninst].size/4,
+            ninst,
+            dyn->insts[ninst].x86.barrier,
+            dyn->insts[ninst].x86.state_flags,
+            dyn->f.pending,
+            dyn->f.dfnone,
+            dyn->insts[ninst].x86.may_set?"may":"set",
+            dyn->insts[ninst].x86.set_flags,
+            dyn->insts[ninst].x86.gen_flags,
+            dyn->insts[ninst].x86.use_flags,
+            dyn->insts[ninst].x86.need_before,
+            dyn->insts[ninst].x86.need_after,
+            dyn->smread, dyn->smwrite);
+        if(dyn->insts[ninst].pred_sz) {
+            dynarec_log(LOG_NONE, ", pred=");
+            for(int ii=0; ii<dyn->insts[ninst].pred_sz; ++ii)
+                dynarec_log(LOG_NONE, "%s%d", ii?"/":"", dyn->insts[ninst].pred[ii]);
+        }
+        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts>=0)
+            dynarec_log(LOG_NONE, ", jmp=%d", dyn->insts[ninst].x86.jmp_insts);
+        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts==-1)
+            dynarec_log(LOG_NONE, ", jmp=out");
+        for(int ii=0; ii<24; ++ii) {
+            switch(dyn->insts[ninst].n.neoncache[ii].t) {
+                case NEON_CACHE_ST_D: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_ST_F: dynarec_log(LOG_NONE, " S%d:%s", (ii+8)*2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_MM: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_XMMW: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;
+                case NEON_CACHE_XMMR: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;
+                case NEON_CACHE_SCR: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_NONE:
+                default:    break;
+            }
+        }
+        if(dyn->n.stack || dyn->insts[ninst].n.stack_next || dyn->insts[ninst].n.x87stack)
+            dynarec_log(LOG_NONE, " X87:%d/%d(+%d/-%d)%d", dyn->n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, dyn->insts[ninst].n.stack_pop, dyn->insts[ninst].n.x87stack);
+        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
+            dynarec_log(LOG_NONE, " %s:%d/%d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
+        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");
+    }
+}
+
+void print_opcode(dynarec_arm_t* dyn, int ninst, uint32_t opcode)
+{
+    dynarec_log(LOG_NONE, "\t%08x\t%s\n", opcode, arm_print(opcode));
+}
@@ -23,12 +23,24 @@ 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_aesd(x86emu_t* emu, int xmm);
+void arm_aese(x86emu_t* emu, int xmm);
+void arm_aesdlast(x86emu_t* emu, int xmm);
+void arm_aeselast(x86emu_t* emu, int xmm);
+void arm_aesimc(x86emu_t* emu, int xmm);
+void arm_aeskeygenassist(x86emu_t* emu, int gx, void* p, uint32_t u8);
+void arm_pclmul(x86emu_t* emu, int gx, void* p, uint32_t u8);
+
+void arm_clflush(x86emu_t* emu, void* p);
+
 int64_t arm_fist64_0(double d);
 int64_t arm_fist64_1(double d);
 int64_t arm_fist64_2(double d);
 int64_t arm_fist64_3(double d);

 void arm_ud(x86emu_t* emu);
+void arm_priv(x86emu_t* emu);
+void arm_singlestep(x86emu_t* emu);

 #define FPUFIRST    8

@@ -71,9 +83,6 @@ int CacheNeedsTransform(dynarec_arm_t* dyn, int i1);
 // Undo the changes of a neoncache to get the status before the instruction
 void neoncacheUnwind(neoncache_t* cache);

-// is inst clean for a son branch?
-int isInstClean(dynarec_arm_t* dyn, int ninst);
-
 // predecessor access
 int isPred(dynarec_arm_t* dyn, int ninst, int pred);
 int getNominalPred(dynarec_arm_t* dyn, int ninst);
@@ -82,10 +91,14 @@ int getNominalPred(dynarec_arm_t* dyn, int ninst);
 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);
+uint8_t geted_ib(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop);

 // Is what pointed at addr a native call? And if yes, to what function?
 int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn);

 const char* getCacheName(int t, int n);

+void inst_name_pass3(dynarec_arm_t* dyn, int ninst, const char* name);
+void print_opcode(dynarec_arm_t* dyn, int ninst, uint32_t opcode);
+
 #endif //__DYNAREC_ARM_FUNCTIONS_H__
@@ -4,6 +4,7 @@
 #include <pthread.h>
 #include <errno.h>
 #include <assert.h>
+#include <string.h>

 #include "debug.h"
 #include "box86context.h"
@@ -39,6 +40,9 @@ uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, u
    if(hint>0 && hint<xEAX) scratch = hint;
    if(hint==xEIP) scratch = hint;  // allow this one as a scratch and return register
    MAYUSE(scratch);
+
+    if(l==LOCK_LOCK) { /*SMDMB();*/DMB_ISH(); }
+
    if(!(nextop&0xC0)) {
        if((nextop&7)==4) {
            uint8_t sib = F8;
@@ -176,8 +180,8 @@ uintptr_t geted16(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop,
    uint32_t m = nextop&0xC7;
    uint32_t n = (m>>6)&3;
    int32_t offset = 0;
-    if(!n && m==6) {
-        offset = F16;
+    if(!n && (m&7)==6) {
+        offset = F16S;
        MOVW(ret, offset);
    } else {
        switch(n) {
@@ -185,7 +189,7 @@ uintptr_t geted16(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop,
            case 1: offset = F8S; break;
            case 2: offset = F16S; break;
        }
-        if(offset && (abs(offset)>absmax || (offset&mask))) {
+        if(offset && (abs(offset)<=absmax && !(offset&mask))) {
            *fixaddress = offset;
            offset = 0;
        }
@@ -861,10 +865,29 @@ void x87_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int
    MESSAGE(LOG_DUMP, "\t---Purge x87 Cache and Synch Stackcount\n");
 }

-#ifdef HAVE_TRACE
 static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
 {
-    x87_stackcount(dyn, ninst, s1);
+    // Synch top & stack counter
+    int a = dyn->n.x87stack;
+    if(a) {
+    // Add x87stack to emu fpu_stack
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, fpu_stack));
+        if(a>0) {
+            ADD_IMM8(s2, s2, a);
+        } else {
+            SUB_IMM8(s2, s2, -a);
+        }
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, fpu_stack));
+        // Sub x87stack to top, with and 7
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+        if(a>0) {
+            SUB_IMM8(s2, s2, a);
+        } else {
+            ADD_IMM8(s2, s2, -a);
+        }
+        AND_IMM8(s2, s2, 7);
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    }
    int ret = 0;
    for (int i=0; (i<8) && (!ret); ++i)
        if(dyn->n.x87cache[i] != -1)
@@ -872,13 +895,20 @@ static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
    if(!ret)    // nothing to do
        return;
    // Get top
-    LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    if(!a) {
+        // already there
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    }
    // loop all cache entries
    for (int i=0; i<8; ++i)
        if(dyn->n.x87cache[i]!=-1) {
-            ADD_IMM8(s3, s2, dyn->n.x87cache[i]);
-            AND_IMM8(s3, s3, 7);    // (emu->top + i)&7
-            ADD_REG_LSL_IMM5(s3, xEmu, s3, 3);    // fpu[(emu->top+i)&7] lsl 3 because fpu are double, so 8 bytes
+            if(dyn->n.x87cache[i]) {
+                ADD_IMM8(s3, s2, dyn->n.x87cache[i]);
+                AND_IMM8(s3, s3, 7);    // (emu->top + i)&7
+                ADD_REG_LSL_IMM5(s3, xEmu, s3, 3);    // fpu[(emu->top+i)&7] lsl 3 because fpu are double, so 8 bytes
+            } else {
+                ADD_REG_LSL_IMM5(s3, xEmu, s2, 3);
+            }
            if(ST_IS_F(dyn->n.x87cache[i])) {
                VCVT_F64_F32(0, dyn->n.x87reg[i]*2);    // use D0 as scratch...
                VSTR_64(0, s3, offsetof(x86emu_t, x87));
@@ -887,7 +917,31 @@ static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
            }
        }
 }
-#endif
+
+static void x87_unreflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
+{
+    // go back with the top & stack counter
+    int a = dyn->n.x87stack;
+    if(a) {
+    // Add x87stack to emu fpu_stack
+        LDR_IMM9(s1, xEmu, offsetof(x86emu_t, fpu_stack));
+        if(a>0) {
+            SUB_IMM8(s1, s1, a);
+        } else {
+            ADD_IMM8(s1, s1, -a);
+        }
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, fpu_stack));
+        // Sub x87stack to top, with and 7
+        LDR_IMM9(s1, xEmu, offsetof(x86emu_t, top));
+        if(a>0) {
+            ADD_IMM8(s1, s1, a);
+        } else {
+            SUB_IMM8(s1, s1, -a);
+        }
+        AND_IMM8(s1, s1, 7);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, top));
+    }
+}

 int x87_get_current_cache(dynarec_arm_t* dyn, int ninst, int st, int t)
 {
@@ -998,8 +1052,10 @@ void x87_forget(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st)
    if(ret==-1)    // nothing to do
        return;
    MESSAGE(LOG_DUMP, "\tForget x87 Cache for ST%d\n", st);
+    int reg = dyn->n.x87reg[ret];
    #if STEP == 1
-    neoncache_promote_double(dyn, ninst, st);
+    if(dyn->n.neoncache[reg].t==NEON_CACHE_ST_F)
+        neoncache_promote_double(dyn, ninst, st);
    #endif
    // prepare offset to fpu => s1
    ADD_IMM8(s1, xEmu, offsetof(x86emu_t, x87));
@@ -1011,11 +1067,17 @@ void x87_forget(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st)
        AND_IMM8(s2, s2, 7);    // (emu->top + i)&7
    }
    ADD_REG_LSL_IMM5(s2, s1, s2, 3);    // fpu[(emu->top+i)&7] lsl 3 because fpu are double, so 8 bytes
-    VSTR_64(dyn->n.x87reg[ret], s2, 0);    // save the value
+    if(dyn->n.neoncache[reg].t==NEON_CACHE_ST_F) {
+        // this should not happens
+        VCVT_F64_F32(0, reg*2);    // use D0 as scratch...
+        VSTR_64(0, s2, 0);
+    } else {
+        VSTR_64(reg, s2, 0);    // save the value
+    }
    MESSAGE(LOG_DUMP, "\t--------Forget x87 Cache for ST%d\n", st);
    // and forget that cache
-    fpu_free_reg_double(dyn, dyn->n.x87reg[ret]);
-    dyn->n.neoncache[dyn->n.x87reg[ret]].v = 0;
+    fpu_free_reg_double(dyn, reg);
+    dyn->n.neoncache[reg].v = 0;
    dyn->n.x87cache[ret] = -1;
    dyn->n.x87reg[ret] = -1;
 }
@@ -1223,7 +1285,6 @@ void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1)
        MESSAGE(LOG_DUMP, "\t------ Purge MMX Cache\n");
    }
 }
-#ifdef HAVE_TRACE
 static void mmx_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
    int old = -1;
@@ -1241,7 +1302,6 @@ static void mmx_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
            VST1_32_W(dyn->n.mmxcache[i], s1);
        }
 }
-#endif


 // SSE / SSE2 helpers
@@ -1287,6 +1347,26 @@ int sse_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a)
    dyn->n.ssecache[a].write = 1; // it will be write...
    return dyn->n.ssecache[a].reg;
 }
+// forget neon register for a SSE reg
+void sse_forget_reg(dynarec_arm_t* dyn, int ninst, int a, int s1)
+{
+    if(dyn->n.ssecache[a].v==-1)
+        return;
+    if(dyn->n.neoncache[dyn->n.ssecache[a].reg+0-FPUFIRST].t == NEON_CACHE_XMMW) {
+        int offs = offsetof(x86emu_t, xmm[a]);
+        if(!(offs&3) && (offs>>2)<256) {
+            ADD_IMM8_ROR(s1, xEmu, offs>>2, 15);
+        } else {
+            MOV32(s1, offs);
+            ADD_REG_LSL_IMM5(s1, xEmu, s1, 0);
+        }
+        VST1Q_64(dyn->n.ssecache[a].reg, s1);
+    }
+    fpu_free_reg_quad(dyn, dyn->n.ssecache[a].reg);
+    dyn->n.ssecache[a].v = -1;
+    return;
+}
+
 // purge the SSE cache only(needs 1 scratch registers)
 static void sse_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1)
 {
@@ -1321,7 +1401,6 @@ static void sse_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1)
        MESSAGE(LOG_DUMP, "\t------ Purge SSE Cache\n");
    }
 }
-#ifdef HAVE_TRACE
 static void sse_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
    int old = -1;
@@ -1345,7 +1424,24 @@ static void sse_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
            VST1Q_32_W(dyn->n.ssecache[i].reg, s1);
        }
 }
-#endif
+
+int sse_reflect_reg(dynarec_arm_t* dyn, int ninst, int a, int s1)
+{
+    if(dyn->n.ssecache[a].v==-1)
+        return 0;
+    if(dyn->n.ssecache[a].write) {
+        int offs = offsetof(x86emu_t, xmm[a]);
+        if(!(offs&3) && (offs>>2)<256) {
+            ADD_IMM8_ROR(s1, xEmu, offs>>2, 15);
+        } else {
+            MOV32(s1, offs);
+            ADD_REG_LSL_IMM5(s1, xEmu, s1, 0);
+        }
+        VST1Q_32(dyn->n.ssecache[a].reg, s1);
+        return 1;
+    }
+    return 0;
+}

 void fpu_pushcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
@@ -1446,8 +1542,10 @@ static void swapCache(dynarec_arm_t* dyn, int ninst, int i, int j, neoncache_t *
        quad =1;
    if(cache->neoncache[j].t==NEON_CACHE_XMMR || cache->neoncache[j].t==NEON_CACHE_XMMW)
        quad =1;
-    if(quad)    // if quad, we need to swap the whole quad!
+    if(quad) {    // if quad, we need to swap the whole quad!
        i&=~1;
+        j&=~1;
+    }
    if(!cache->neoncache[i].v && ((quad && !cache->neoncache[i+1].v) || !quad)) {
        // a mov is enough, no need to swap
        MESSAGE(LOG_DUMP, "\t  - Moving %d <- %d\n", i, j);
@@ -1717,7 +1815,7 @@ static void fpuCacheTransform(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
        stack_cnt = cache_i2.stack;
    }
    neoncache_t cache = dyn->n;
-    // sanitize XMM regs alocation
+    // sanitize XMM regs alocation, cache_i2 has been sanitized in neoncacheUnwind already
    for(int i=0; i<24; i+=2) {
        if(cache.neoncache[i].t == NEON_CACHE_XMMR || cache.neoncache[i].t == NEON_CACHE_XMMW)
            cache.neoncache[i+1] = cache.neoncache[i];
@@ -1797,7 +1895,7 @@ static void flagsCacheTransform(dynarec_arm_t* dyn, int ninst, int s1)
    if(dyn->f.dfnone)  // flags are fully known, nothing we can do more
        return;
    MESSAGE(LOG_DUMP, "\tFlags fetch ---- ninst=%d -> %d\n", ninst, jmp);
-    int go = 0;
+    int go = (dyn->insts[jmp].f_entry.dfnone && !dyn->f.dfnone)?1:0;
    switch (dyn->insts[jmp].f_entry.pending) {
        case SF_UNKNOWN: break;
        case SF_SET: 
@@ -1815,21 +1913,19 @@ static void flagsCacheTransform(dynarec_arm_t* dyn, int ninst, int s1)
            && dyn->f.pending!=SF_SET_PENDING
            && dyn->f.pending!=SF_PENDING)
                go = 1;
-            else
-                go = (dyn->insts[jmp].f_entry.dfnone  == dyn->f.dfnone)?0:1;
+            else if(dyn->insts[jmp].f_entry.dfnone!=dyn->f.dfnone)
+                go = 1;
            break;
    }
-    if(dyn->insts[jmp].f_entry.dfnone && !dyn->f.dfnone)
-        go = 1;
    if(go) {
        if(dyn->f.pending!=SF_PENDING) {
            LDR_IMM9(s1, xEmu, offsetof(x86emu_t, df));
            TSTS_REG_LSL_IMM5(s1, s1, 0);
-            j32 = (GETMARK3)-(dyn->arm_size+8);
+            j32 = (GETMARKF2)-(dyn->arm_size+8);
            Bcond(cEQ, j32);
        }
        CALL_(UpdateFlags, -1, 0);
-        MARK3;
+        MARKF2;
    }
 #endif
 }
@@ -1841,16 +1937,17 @@ void CacheTransform(dynarec_arm_t* dyn, int ninst, int cacheupd, int s1, int s2,
        flagsCacheTransform(dyn, ninst, s1);
 }

-#ifdef HAVE_TRACE
 void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
 {
    x87_reflectcache(dyn, ninst, s1, s2, s3);
-    if(trace_emm)
-       mmx_reflectcache(dyn, ninst, s1);
-    if(trace_xmm)
-       sse_reflectcache(dyn, ninst, s1);
+    mmx_reflectcache(dyn, ninst, s1);
+    sse_reflectcache(dyn, ninst, s1);
+}
+
+void fpu_unreflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
+{
+    x87_unreflectcache(dyn, ninst, s1, s2, s3);
 }
-#endif

 void fpu_reset(dynarec_arm_t* dyn)
 {
@@ -1887,3 +1984,64 @@ void emit_pf(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4)
    MVN_REG_LSR_REG(s4, s4, s3);
    BFI(xFlags, s4, F_PF, 1);
 }
+
+void fpu_reset_cache(dynarec_arm_t* dyn, int ninst, int reset_n)
+{
+    MESSAGE(LOG_DEBUG, "Reset Caches with %d\n",reset_n);
+    #if STEP > 1
+    // for STEP 2 & 3, just need to refrest with current, and undo the changes (push & swap)
+    dyn->n = dyn->insts[ninst].n;
+    neoncacheUnwind(&dyn->n);
+    #ifdef HAVE_TRACE
+    if(box86_dynarec_dump)
+        if(memcmp(&dyn->n, &dyn->insts[reset_n].n, sizeof(neon_cache_t))) {
+            MESSAGE(LOG_DEBUG, "Warning, difference in neoncache: reset=");
+            for(int i=0; i<24; ++i)
+                if(dyn->insts[reset_n].n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[reset_n].n.neoncache[i].t, dyn->insts[reset_n].n.neoncache[i].n));
+            if(dyn->insts[reset_n].n.combined1 || dyn->insts[reset_n].n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[reset_n].n.swapped?"SWP":"CMB", dyn->insts[reset_n].n.combined1, dyn->insts[reset_n].n.combined2);
+            if(dyn->insts[reset_n].n.stack_push || dyn->insts[reset_n].n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[reset_n].n.stack_push, -dyn->insts[reset_n].n.stack_pop);
+            MESSAGE(LOG_DEBUG, " ==> ");
+            for(int i=0; i<24; ++i)
+                if(dyn->insts[ninst].n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[ninst].n.neoncache[i].t, dyn->insts[ninst].n.neoncache[i].n));
+            if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
+            if(dyn->insts[ninst].n.stack_push || dyn->insts[ninst].n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
+            MESSAGE(LOG_DEBUG, " -> ");
+            for(int i=0; i<24; ++i)
+                if(dyn->n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->n.neoncache[i].t, dyn->n.neoncache[i].n));
+            if(dyn->n.combined1 || dyn->n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->n.swapped?"SWP":"CMB", dyn->n.combined1, dyn->n.combined2);
+            if(dyn->n.stack_push || dyn->n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->n.stack_push, -dyn->n.stack_pop);
+            MESSAGE(LOG_DEBUG, "\n");
+        }
+    #endif //HAVE_TRACE
+    #else
+    dyn->n = dyn->insts[reset_n].n;
+    #endif
+}
+
+// propagate ST stack state, especial stack pop that are deferred
+void fpu_propagate_stack(dynarec_arm_t* dyn, int ninst)
+{
+    if(dyn->n.stack_pop) {
+        for(int j=0; j<24; ++j)
+            if((dyn->n.neoncache[j].t == NEON_CACHE_ST_D || dyn->n.neoncache[j].t == NEON_CACHE_ST_F)) {
+                if(dyn->n.neoncache[j].n<dyn->n.stack_pop)
+                    dyn->n.neoncache[j].v = 0;
+                else
+                    dyn->n.neoncache[j].n-=dyn->n.stack_pop;
+            }
+        dyn->n.stack_pop = 0;
+    }
+    dyn->n.stack = dyn->n.stack_next;
+    dyn->n.news = 0;
+    dyn->n.stack_push = 0;
+    dyn->n.swapped = 0;
+}
@@ -27,29 +27,34 @@
 #define PKip(a) *(uint8_t*)(ip+a)
 #define PKa(a)  *(uint8_t*)(a)

+#define MODREG  ((nextop&0xC0)==0xC0)
+
 // Strong mem emulation helpers
-// Sequence of Read will trigger a DMB on "first" read if strongmem is 2
-// Squence of Write will trigger a DMB on "last" write if strongmem is 1
+#define SMREAD_MIN  2
+#define SMWRITE_MIN 1
+// Sequence of Read will trigger a DMB on "first" read if strongmem is >= SMREAD_MIN
+// Sequence of Write will trigger a DMB on "last" write if strongmem is >= 1
+// All Write operation that might use a lock all have a memory barrier if strongmem is >= SMWRITE_MIN
 // Opcode will read
-#define SMREAD()    if(!dyn->smread && box86_dynarec_strongmem>1) {SMDMB();}
+#define SMREAD()    if((dyn->smread==0) && (box86_dynarec_strongmem>SMREAD_MIN)) {SMDMB();} else dyn->smread=1
 // Opcode will read with option forced lock
-#define SMREADLOCK(lock)    if(lock || (!dyn->smread && box86_dynarec_strongmem>1)) {SMDMB();}
-// Opcode migh read (depend on nextop)
+#define SMREADLOCK(lock)    if((lock) || ((dyn->smread==0) && (box86_dynarec_strongmem>SMREAD_MIN))) {SMDMB();}
+// Opcode might read (depend on nextop)
 #define SMMIGHTREAD()   if(!MODREG) {SMREAD();}
 // Opcode has wrote
 #define SMWRITE()   dyn->smwrite=1
 // Opcode has wrote (strongmem>1 only)
-#define SMWRITE2()   if(box86_dynarec_strongmem>1) dyn->smwrite=1
+#define SMWRITE2()   if(box86_dynarec_strongmem>SMREAD_MIN) dyn->smwrite=1
 // Opcode has wrote with option forced lock
-#define SMWRITELOCK(lock)   if(lock) {SMDMB();} else dyn->smwrite=1
-// Opcode migh have wrote (depend on nextop)
+#define SMWRITELOCK(lock)   if(lock || (box86_dynarec_strongmem>SMWRITE_MIN)) {SMDMB();} else dyn->smwrite=1
+// Opcode might have wrote (depend on nextop)
 #define SMMIGHTWRITE()   if(!MODREG) {SMWRITE();}
 // Start of sequence
 #define SMSTART()   SMEND()
 // End of sequence
-#define SMEND()     if(dyn->smwrite && box86_dynarec_strongmem) {DMB_ISH();} dyn->smwrite=0; dyn->smread=0;
+#define SMEND()     if(dyn->smwrite && box86_dynarec_strongmem) {if(box86_dynarec_strongmem){DSB_ISH();}else{DMB_ISH();}} dyn->smwrite=0; dyn->smread=0;
 // Force a Data memory barrier (for LOCK: prefix)
-#define SMDMB()     DMB_ISH(); dyn->smwrite=0; dyn->smread=1
+#define SMDMB()     if(box86_dynarec_strongmem){DSB_ISH();}else{DMB_ISH();} dyn->smwrite=0; dyn->smread=1

 //LOCK_* define
 #define LOCK_LOCK   (int*)1
@@ -120,6 +125,8 @@
                }
 // GETGW extract x86 register in gd, that is i
 #define GETGW(i) gd = xEAX+((nextop&0x38)>>3); UXTH(i, gd, 0); gd = i;
+// GETSGW extract x86 signed register in gd, that is i
+#define GETSGW(i) gd = xEAX+((nextop&0x38)>>3); SXTH(i, gd, 0); gd = i;
 //GETEWW will use i for ed, and can use w for wback.
 #define GETEWW(w, i) if((nextop&0xC0)==0xC0) {  \
                    wback = xEAX+(nextop&7);\
@@ -260,17 +267,19 @@
 // CALL_1RD will use S as scratch. Return value in D0, 1 ARG in R
 #define CALL_1RD(F, R, S, M) call_rd(dyn, ninst, F, R, S, 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;}
-#define GETMARK2 ((dyn->insts)?dyn->insts[ninst].mark2:(dyn->arm_size+4))
-#define MARK3   if(dyn->insts) {dyn->insts[ninst].mark3 = (uintptr_t)dyn->arm_size;}
-#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 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;}
+#define GETMARK2    ((dyn->insts)?dyn->insts[ninst].mark2:(dyn->arm_size+4))
+#define MARK3       if(dyn->insts) {dyn->insts[ninst].mark3 = (uintptr_t)dyn->arm_size;}
+#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 MARKF2      if(dyn->insts) {dyn->insts[ninst].markf2 = (uintptr_t)dyn->arm_size;}
+#define GETMARKF2   ((dyn->insts)?dyn->insts[ninst].markf2:(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 j32)
@@ -333,7 +342,17 @@


 #define SET_DFNONE(S)    if(!dyn->f.dfnone) {MOVW(S, d_none); STR_IMM9(S, xEmu, offsetof(x86emu_t, df)); dyn->f.dfnone=1;}
-#define SET_DF(S, N)     if(N) {MOVW(S, N); STR_IMM9(S, xEmu, offsetof(x86emu_t, df)); dyn->f.dfnone=0;} else SET_DFNONE(S)
+#define SET_DF(S, N)     \
+    if(N) {             \
+        MOVW(S, N);\
+        STR_IMM9(S, xEmu, offsetof(x86emu_t, df));  \
+        if(dyn->f.pending==SF_PENDING && dyn->insts[ninst].x86.need_after && !(dyn->insts[ninst].x86.need_after&X_PEND)) {  \
+            CALL_(UpdateFlags, -1, 0);              \
+            dyn->f.pending = SF_SET;                \
+            SET_NODF();     \
+        }                   \
+        dyn->f.dfnone=0;\
+    } else SET_DFNONE(S)
 #define SET_NODF()          dyn->f.dfnone = 0
 #define SET_DFOK()          dyn->f.dfnone = 1

@@ -375,7 +394,7 @@
    } else dyn->f.pending = SF_SET
 #endif
 #ifndef JUMP
-#define JUMP(A, C) 
+#define JUMP(A, C) SMEND()
 #endif
 #ifndef BARRIER
 #define BARRIER(A) 
@@ -510,10 +529,24 @@ void* arm_next(x86emu_t* emu, uintptr_t addr);
 #define emit_shr32      STEPNAME(emit_shr32)
 #define emit_shr32c     STEPNAME(emit_shr32c)
 #define emit_sar32c     STEPNAME(emit_sar32c)
+#define emit_shl8       STEPNAME(emit_shl8)
+#define emit_shl8c      STEPNAME(emit_shl8c)
+#define emit_shr8       STEPNAME(emit_shr8)
+#define emit_shr8c      STEPNAME(emit_shr8c)
+#define emit_sar8       STEPNAME(emit_sar8)
+#define emit_sar8c      STEPNAME(emit_sar8c)
+#define emit_shl16      STEPNAME(emit_shl16)
+#define emit_shl16c     STEPNAME(emit_shl16c)
+#define emit_shr16      STEPNAME(emit_shr16)
+#define emit_shr16c     STEPNAME(emit_shr16c)
+#define emit_sar16      STEPNAME(emit_sar16)
+#define emit_sar16c     STEPNAME(emit_sar16c)
 #define emit_rol32c     STEPNAME(emit_rol32c)
 #define emit_rol8c      STEPNAME(emit_rol8c)
 #define emit_ror32c     STEPNAME(emit_ror32c)
 #define emit_ror8c      STEPNAME(emit_ror8c)
+#define emit_rol16c     STEPNAME(emit_rol16c)
+#define emit_ror16c     STEPNAME(emit_ror16c)
 #define emit_shrd32c    STEPNAME(emit_shrd32c)
 #define emit_shld32c    STEPNAME(emit_shld32c)
 #define emit_shrd32     STEPNAME(emit_shrd32)
@@ -543,16 +576,19 @@ void* arm_next(x86emu_t* emu, uintptr_t addr);
 #define mmx_get_reg_empty STEPNAME(mmx_get_reg_empty)
 #define sse_get_reg     STEPNAME(sse_get_reg)
 #define sse_get_reg_empty STEPNAME(sse_get_reg_empty)
+#define sse_forget_reg  STEPNAME(sse_forget_reg)
+#define sse_reflect_reg STEPNAME(sse_reflect_reg)

 #define fpu_pushcache   STEPNAME(fpu_pushcache)
 #define fpu_popcache    STEPNAME(fpu_popcache)
 #define fpu_reset       STEPNAME(fpu_reset)
+#define fpu_reset_cache STEPNAME(fpu_reset_cache)
+#define fpu_propagate_stack STEPNAME(fpu_propagate_stack)
 #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
+#define fpu_unreflectcache STEPNAME(fpu_unreflectcache)

 // get the single reg that from the double "reg" (so Dx[idx])
 #define fpu_get_single_reg      STEPNAME(fpu_get_single_reg)
@@ -645,10 +681,24 @@ void emit_shl32c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s
 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_shl8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shl8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_shr8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shr8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sar8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_sar8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_shl16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shl16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_shr16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_shr16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_sar16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_sar16c(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_rol8c(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_ror8c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_rol16c(dynarec_arm_t* dyn, int ninst, int s1, int32_t c, int s3, int s4);
+void emit_ror16c(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_shrd32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
@@ -737,17 +787,24 @@ int mmx_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int
 int sse_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a, int forwrite);
 // get neon register for a SSE reg, but don't try to synch it if it needed to be created
 int sse_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a);
+// forget an SSE reg
+void sse_forget_reg(dynarec_arm_t* dyn, int ninst, int a, int s1);
+// sync reg with emu for sse reg (return 1 if reflect was done, s1 = &xmm[a] then)
+int sse_reflect_reg(dynarec_arm_t* dyn, int ninst, int a, int s1);

 // common coproc helpers
 // reset the cache
 void fpu_reset(dynarec_arm_t* dyn);
+// reset the cache with n
+void fpu_reset_cache(dynarec_arm_t* dyn, int ninst, int reset_n);
+// propagate stack state
+void fpu_propagate_stack(dynarec_arm_t* dyn, int ninst);
 // purge the FPU cache (needs 3 scratch registers) next=1 if for a conditionnal branch jumping out of block (no tracking updated)
 void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 void x87_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1);
-#ifdef HAVE_TRACE
 void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
-#endif
+void fpu_unreflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 void fpu_pushcache(dynarec_arm_t* dyn, int ninst, int s1);
 void fpu_popcache(dynarec_arm_t* dyn, int ninst, int s1);

@@ -794,4 +851,21 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
 #define MAYUSE(A)   
 #endif

+#define NOTEST(s1)                                          \
+    if(box86_dynarec_test) {                                \
+        MOVW(s1, 0);                                        \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.test));  \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.clean)); \
+    }
+#define SKIPTEST(s1)                                        \
+    if(box86_dynarec_test) {                                \
+        MOVW(s1, 0);                                        \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.clean)); \
+    }
+#define GOTEST(s1, s2)                                      \
+    if(box86_dynarec_test) {                                \
+        MOVW(s2, 1);                                        \
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, test.test));  \
+    }
+
 #endif //__DYNAREC_ARM_HELPER_H__
@@ -0,0 +1,12 @@
+#include <stdint.h>
+#include <stdlib.h>
+
+#include "arm_emitter.h"
+
+#define EMIT(A) *block = (A); ++block
+void CreateJmpNext(void* addr, void* next)
+{
+    uint32_t* block = (uint32_t*)addr;
+    LDR_literal(x2, ((intptr_t)next - (intptr_t)addr)-8);
+    BX(x2);
+}
@@ -36,7 +36,6 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
    uintptr_t init_addr = addr;
    MAYUSE(init_addr);
    int need_epilog = 1;
-    dyn->sons_size = 0;
    // Clean up (because there are multiple passes)
    dyn->f.pending = 0;
    dyn->f.dfnone = 0;
@@ -57,44 +56,7 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->f.pending = 0;
                fpu_reset(dyn);
            } else {
-                MESSAGE(LOG_DEBUG, "Reset Caches with %d\n",reset_n);
-                #if STEP > 1
-                // for STEP 2 & 3, just need to refrest with current, and undo the changes (push & swap)
-                dyn->n = dyn->insts[ninst].n;
-                neoncacheUnwind(&dyn->n);
-                #ifdef HAVE_TRACE
-                if(box86_dynarec_dump)
-                    if(memcmp(&dyn->n, &dyn->insts[reset_n].n, sizeof(neon_cache_t))) {
-                        MESSAGE(LOG_DEBUG, "Warning, difference in neoncache: reset=");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->insts[reset_n].n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[reset_n].n.neoncache[i].t, dyn->insts[reset_n].n.neoncache[i].n));
-                        if(dyn->insts[reset_n].n.combined1 || dyn->insts[reset_n].n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[reset_n].n.swapped?"SWP":"CMB", dyn->insts[reset_n].n.combined1, dyn->insts[reset_n].n.combined2);
-                        if(dyn->insts[reset_n].n.stack_push || dyn->insts[reset_n].n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[reset_n].n.stack_push, -dyn->insts[reset_n].n.stack_pop);
-                        MESSAGE(LOG_DEBUG, " ==> ");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->insts[ninst].n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[ninst].n.neoncache[i].t, dyn->insts[ninst].n.neoncache[i].n));
-                        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
-                        if(dyn->insts[ninst].n.stack_push || dyn->insts[ninst].n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
-                        MESSAGE(LOG_DEBUG, " -> ");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->n.neoncache[i].t, dyn->n.neoncache[i].n));
-                        if(dyn->n.combined1 || dyn->n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->n.swapped?"SWP":"CMB", dyn->n.combined1, dyn->n.combined2);
-                        if(dyn->n.stack_push || dyn->n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->n.stack_push, -dyn->n.stack_pop);
-                        MESSAGE(LOG_DEBUG, "\n");
-                    }
-                #endif //HAVE_TRACE
-                #else
-                dyn->n = dyn->insts[reset_n].n;
-                #endif
+                fpu_reset_cache(dyn, ninst, reset_n);
                dyn->f = dyn->insts[reset_n].f_exit;
                if(dyn->insts[ninst].x86.barrier&BARRIER_FLOAT) {
                    MESSAGE(LOG_DEBUG, "Apply Barrier Float\n");
@@ -108,29 +70,29 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            }
            reset_n = -1;
        }
-        // propagate ST stack state, especial stack pop that are defered
-        if(dyn->n.stack_pop) {
-            for(int j=0; j<24; ++j)
-                if((dyn->n.neoncache[j].t == NEON_CACHE_ST_D || dyn->n.neoncache[j].t == NEON_CACHE_ST_F)) {
-                    if(dyn->n.neoncache[j].n<dyn->n.stack_pop)
-                        dyn->n.neoncache[j].v = 0;
-                    else
-                        dyn->n.neoncache[j].n-=dyn->n.stack_pop;
-                }
-            dyn->n.stack_pop = 0;
-        }
-        dyn->n.stack = dyn->n.stack_next;
-        dyn->n.news = 0;
-        dyn->n.stack_push = 0;
-        dyn->n.swapped = 0;
+        fpu_propagate_stack(dyn, ninst);
        NEW_INST;
+        if(!ninst) {
+            GOTEST(x1, x2);
+        }
        if(dyn->insts[ninst].pred_sz>1) {SMSTART();}
        fpu_reset_scratch(dyn);
        if((dyn->insts[ninst].x86.need_before&~X_PEND) && !dyn->insts[ninst].pred_sz) {
            READFLAGS(dyn->insts[ninst].x86.need_before&~X_PEND);
        }
+        if(box86_dynarec_test)  {
+            MESSAGE(LOG_DUMP, "TEST INIT  ----\n");
+            fpu_reflectcache(dyn, ninst, x1, x2, x3);
+            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(x86test_init, -1, 0);
+            fpu_unreflectcache(dyn, ninst, x1, x2, x3);
+            MESSAGE(LOG_DUMP, "----------\n");
+        }
 #ifdef HAVE_TRACE
-        if(my_context->dec && box86_dynarec_trace) {
+        else if(my_context->dec && box86_dynarec_trace) {
        if((trace_end == 0) 
            || ((ip >= trace_start) && (ip < trace_end)))  {
                MESSAGE(LOG_DUMP, "TRACE ----\n");
@@ -140,6 +102,7 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                STR_IMM9(x1, xEmu, offsetof(x86emu_t, ip));
                MOVW(x2, 1);
                CALL(PrintTrace, -1, 0);
+                fpu_unreflectcache(dyn, ninst, x1, x2, x3);
                MESSAGE(LOG_DUMP, "----------\n");
            }
        }
@@ -162,13 +125,26 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->f.dfnone = 0;
            }
        }
+        #ifndef PROT_READ
+        #define PROT_READ 1
+        #endif
        #if STEP != 0
        if(!ok && !need_epilog && (addr < (dyn->start+dyn->isize))) {
            ok = 1;
            // we use the 1st predecessor here
            int ii = ninst+1;
-            while(ii<dyn->size && !dyn->insts[ii].pred_sz) 
-                ++ii;
+            if(ii<dyn->size && !dyn->insts[ii].pred_sz) {
+                while(ii<dyn->size && (!dyn->insts[ii].pred_sz || (dyn->insts[ii].pred_sz==1 && dyn->insts[ii].pred[0]==ii-1))) {
+                    // may need to skip opcodes to advance
+                    ++ninst;
+                    NEW_INST;
+                    MESSAGE(LOG_DEBUG, "Skipping unused opcode\n");
+                    INST_NAME("Skipped opcode");
+                    INST_EPILOG;
+                    addr += dyn->insts[ii].x86.size;
+                    ++ii;
+                }
+            }
            if((dyn->insts[ii].x86.barrier&BARRIER_FULL)==BARRIER_FULL)
                reset_n = -2;    // hack to say Barrier!
            else {
@@ -180,20 +156,18 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            }
        }
        #else
-        #ifndef PROT_READ
-        #define PROT_READ 1
-        #endif
        if(dyn->forward) {
-            if(dyn->forward_to == addr) {
+            if(dyn->forward_to == addr && !need_epilog) {
                // we made it!
-                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Forward extend block %p -> %p\n", (void*)dyn->forward, (void*)dyn->forward_to);
+                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Forward extend block for %d bytes %p -> %p\n", dyn->forward_to-dyn->forward, (void*)dyn->forward, (void*)dyn->forward_to);
                dyn->forward = 0;
                dyn->forward_to = 0;
                dyn->forward_size = 0;
                dyn->forward_ninst = 0;
+                ok = 1; // in case it was 0
            } else if ((dyn->forward_to < addr) || !ok) {
                // something when wrong! rollback
-                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Could not forward extend block %p -> %p\n", (void*)dyn->forward, (void*)dyn->forward_to);
+                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Could not forward extend block for %d bytes %p -> %p\n", dyn->forward_to-dyn->forward, (void*)dyn->forward, (void*)dyn->forward_to);
                ok = 0;
                dyn->size = dyn->forward_size;
                ninst = dyn->forward_ninst;
@@ -204,46 +178,85 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->forward_ninst = 0;
            }
            // else just continue
-        } else  if(!ok && !need_epilog && box86_dynarec_bigblock && (getProtection(addr+3)&PROT_READ))
+        } else if(!ok && !need_epilog && box86_dynarec_bigblock && (getProtection(addr+3)&~PROT_READ))
            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)) 
-                    /*&& box86_dynarec_bigblock*/))
+                    /*||(((next-addr)<30) && is_instructions(dyn, addr, next-addr))*/ ))
                {
                    ok = 1;
                    // need to find back that instruction to copy the caches, as previous version cannot be used anymore
-                    reset_n = -2;
-                    for(int ii=0; ii<ninst; ++ii)
-                        if(dyn->insts[ii].x86.jmp == next) {
-                            reset_n = ii;
-                            ii=ninst;
-                        }
+                    // and pred table is not ready yet
+                    reset_n = get_first_jump(dyn, next);
                    if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Extend block %p, %p -> %p (ninst=%d, jump from %d)\n", dyn, (void*)addr, (void*)next, ninst, reset_n);
-                } else if(next && (next-addr<box86_dynarec_forward) 
-                && (getProtection(next)&PROT_READ)/*box86_dynarec_bigblock>=forwardblock*/) {
-                    dyn->forward = addr;
-                    dyn->forward_to = next;
-                    dyn->forward_size = dyn->size;
-                    dyn->forward_ninst = ninst;
-                    reset_n = -2;
-                    ok = 1;
+                } else if(next && (next-addr)<box86_dynarec_forward && (getProtection(next)&PROT_READ)/*box86_dynarec_bigblock>=stopblock*/) {
+                    if(!((box86_dynarec_bigblock<stopblock) && !isJumpTableDefault((void*)next))) {
+                        if(dyn->forward) {
+                            if(next<dyn->forward_to)
+                                dyn->forward_to = next;
+                            reset_n = -2;
+                            ok = 1;
+                        } else {
+                            dyn->forward = addr;
+                            dyn->forward_to = next;
+                            dyn->forward_size = dyn->size;
+                            dyn->forward_ninst = ninst;
+                            reset_n = -2;
+                            ok = 1;
+                        }
+                    }
                }
            }
        #endif
-        if(ok<0)  {ok = 0; need_epilog=1;}
+        if(ok<0)  {
+            ok = 0; need_epilog=1; 
+            #if STEP == 0
+            if(ninst) {
+                --ninst;
+                if(!dyn->insts[ninst].x86.barrier) {
+                    BARRIER(BARRIER_FLOAT);
+                }
+                dyn->insts[ninst].x86.need_after |= X_PEND;
+                ++ninst;
+            }
+            if(dyn->forward) {
+                // stopping too soon
+                dyn->size = dyn->forward_size;
+                ninst = dyn->forward_ninst+1;
+                addr = dyn->forward;
+                dyn->forward = 0;
+                dyn->forward_to = 0;
+                dyn->forward_size = 0;
+                dyn->forward_ninst = 0;
+            }
+            #endif
+        }
        ++ninst;
        #if STEP == 0
        if(ok && (((box86_dynarec_bigblock<stopblock) && !isJumpTableDefault((void*)addr)) 
-            || (addr>=box86_nodynarec_start && addr<box86_nodynarec_end)
-            /*|| (dyn->size >= 2000)*/))    // is this still needed? maybe it should be a parameter?
+            || (addr>=box86_nodynarec_start && addr<box86_nodynarec_end)))
        #else
        if(ok && (ninst==dyn->size))
        #endif
        {
+            #if STEP == 0
+            if(dyn->forward) {
+                // stopping too soon
+                dyn->size = dyn->forward_size;
+                ninst = dyn->forward_ninst+1;
+                addr = dyn->forward;
+                dyn->forward = 0;
+                dyn->forward_to = 0;
+                dyn->forward_size = 0;
+                dyn->forward_ninst = 0;
+            }
+            #endif
            int j32;
            MAYUSE(j32);
-            MESSAGE(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size); 
+            MESSAGE(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size);
+            if(!box86_dynarec_dump && addr>=box86_nodynarec_start && addr<box86_nodynarec_end)
+                dynarec_log(LOG_INFO, "Stopping block in no-dynarec zone\n");
            --ninst;
            if(!dyn->insts[ninst].x86.barrier) {
                BARRIER(BARRIER_FLOAT);
@@ -252,16 +265,18 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            dyn->insts[ninst].x86.need_after |= X_PEND;
            #endif
            ++ninst;
+            NOTEST(x3);
            fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
            jump_to_next(dyn, addr, 0, ninst);
            ok=0; need_epilog=0;
        }
    }
    if(need_epilog) {
+        NOTEST(x3);
        fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
        jump_to_epilog(dyn, ip, 0, ninst);  // no linker here, it's an unknow instruction
    }
    FINI;
-    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d, %d sons)\n", dyn->size, dyn->sons_size);
+    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d)\n", dyn->size);
    return addr;
 }
@@ -4,6 +4,7 @@
    dyn->isize = addr-sav_addr;\
    dyn->insts[ninst].x86.addr = addr;\
    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr
+
 #define MESSAGE(A, ...)  
 #define MAYSETFLAGS()   dyn->insts[ninst].x86.may_set = 1
 #define READFLAGS(A)    \
@@ -15,13 +16,13 @@
        dyn->f.pending=(B)&SF_SET_PENDING;      \
        dyn->f.dfnone=((B)&SF_SET)?1:0;
 #define EMIT(A)     
-#define JUMP(A, C)      if((A)>addr) add_next(dyn, (uintptr_t)(A)); dyn->insts[ninst].x86.jmp = A; dyn->insts[ninst].x86.jmp_cond = C
+#define JUMP(A, C)      add_jump(dyn, ninst); add_next(dyn, (uintptr_t)A); SMEND(); dyn->insts[ninst].x86.jmp = A; dyn->insts[ninst].x86.jmp_cond = C
 #define BARRIER(A)      if(A!=BARRIER_MAYBE) {fpu_purgecache(dyn, ninst, 0, x1, x2, x3); dyn->insts[ninst].x86.barrier = A;} else dyn->insts[ninst].barrier_maybe = 1
 #define BARRIER_NEXT(A) dyn->insts[ninst+1].x86.barrier = A
 #define NEW_INST \
        ++dyn->size;                            \
        if(dyn->size+3>=dyn->cap) {             \
-                dyn->insts = (instruction_arm_t*)box_realloc(dyn->insts, sizeof(instruction_arm_t)*dyn->cap*2); \
+                dyn->insts = (instruction_arm_t*)dynaRealloc(dyn->insts, sizeof(instruction_arm_t)*dyn->cap*2);\
                memset(&dyn->insts[dyn->cap], 0, sizeof(instruction_arm_t)*dyn->cap);   \
                dyn->cap *= 2;                  \
        }                                       \
@@ -39,7 +40,7 @@
 #define DEFAULT                         \
        --dyn->size;                    \
        *ok = -1;                       \
-        if(box86_dynarec_log>=LOG_INFO || box86_dynarec_dump) {\
+        if(box86_dynarec_log>=LOG_INFO || box86_dynarec_dump || box86_dynarec_missing) {\
        dynarec_log(LOG_NONE, "%p: Dynarec stopped @ %s (%s:%u) because of Opcode %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X %02X", \
        (void*)ip, __FUNCTION__, __FILE__, __LINE__, PKip(0), \
        PKip(1), PKip(2), PKip(3),      \
@@ -2,7 +2,6 @@
 #define FINI
 #define MESSAGE(A, ...)  
 #define EMIT(A)     
-
 #define NEW_INST                                \
        dyn->insts[ninst].f_entry = dyn->f;     \
        dyn->n.combined1 = dyn->n.combined2 = 0;\
@@ -1,13 +1,16 @@
 #define INIT        dyn->arm_size = 0
-#define FINI        if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
+#define FINI                                                                                            \
+        if(ninst) {                                                                                     \
+                dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);     \
+                dyn->insts_size += 1+((dyn->insts[ninst].x86.size>dyn->insts[ninst].size)?dyn->insts[ninst].x86.size:dyn->insts[ninst].size)/15; \
+        }

 #define MESSAGE(A, ...)  
-#define EMIT(A)     dyn->insts[ninst].size+=4; dyn->arm_size+=4
+#define EMIT(A)     do{dyn->insts[ninst].size+=4; dyn->arm_size+=4;}while(0)
 #define NEW_INST                                                                                        \
        if(ninst) {                                                                                     \
                dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);     \
-                if(ninst && isInstClean(dyn, ninst))                                                    \
-                        ++dyn->sons_size;                                                               \
+                dyn->insts_size += 1+((dyn->insts[ninst-1].x86.size>dyn->insts[ninst-1].size)?dyn->insts[ninst-1].x86.size:dyn->insts[ninst-1].size)/15; \
        }
 #define INST_EPILOG dyn->insts[ninst].epilog = dyn->arm_size; 
 #define INST_NAME(name) 
@@ -1,65 +1,20 @@
 #define INIT    
-#define FINI
-#define EMIT(A)     \
-    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 FINI        \
+    if(ninst)       \
+        addInst(dyn->instsize, &dyn->insts_size, dyn->insts[ninst].x86.size, dyn->insts[ninst].size/4); \
+    addInst(dyn->instsize, &dyn->insts_size, 0, 0);
+#define EMIT(A)                                         \
+    do{                                                 \
+        if(box86_dynarec_dump) print_opcode(dyn, ninst, (uint32_t)(A)); \
+        if((uintptr_t)dyn->block<(uintptr_t)dyn->next-sizeof(void*))\
+            *(uint32_t*)(dyn->block) = (uint32_t)(A);   \
+        dyn->block += 4; dyn->arm_size += 4;            \
+        dyn->insts[ninst].size2 += 4;                   \
+    }while(0)

 #define MESSAGE(A, ...)  if(box86_dynarec_dump) dynarec_log(LOG_NONE, __VA_ARGS__)
 #define NEW_INST        \
-    if(ninst && isInstClean(dyn, ninst)) {                      \
-        dyn->sons_x86[dyn->sons_size] = (uintptr_t)ip;          \
-        dyn->sons_arm[dyn->sons_size] = dyn->block;             \
-        MESSAGE(LOG_DUMP, "----> potential Son here %p/%p\n", (void*)ip, dyn->block);  \
-        ++dyn->sons_size;                                       \
-    }
-
+    if(ninst)                                                   \
+        addInst(dyn->instsize, &dyn->insts_size, dyn->insts[ninst-1].x86.size, dyn->insts[ninst-1].size/4);
 #define INST_EPILOG     
-#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, inst=%d, barrier=%d state=%d/%d(%d), %s=%X/%X, use=%X, need=%X/%X sm=%d/%d", \
-            (box86_dynarec_dump>1)?"\e[32m":"", \
-            (void*)(dyn->arm_start+dyn->insts[ninst].address),  \
-            dyn->insts[ninst].size/4,           \
-            ninst,                              \
-            dyn->insts[ninst].x86.barrier,      \
-            dyn->insts[ninst].x86.state_flags,  \
-            dyn->f.pending,                     \
-            dyn->f.dfnone,                      \
-            dyn->insts[ninst].x86.may_set?"may":"set",          \
-            dyn->insts[ninst].x86.set_flags,    \
-            dyn->insts[ninst].x86.gen_flags,    \
-            dyn->insts[ninst].x86.use_flags,    \
-            dyn->insts[ninst].x86.need_before,  \
-            dyn->insts[ninst].x86.need_after,   \
-            dyn->smread, dyn->smwrite);         \
-        if(dyn->insts[ninst].pred_sz) {         \
-            dynarec_log(LOG_NONE, ", pred=");     \
-            for(int ii=0; ii<dyn->insts[ninst].pred_sz; ++ii)\
-                dynarec_log(LOG_NONE, "%s%d", ii?"/":"", dyn->insts[ninst].pred[ii]);\
-        }                                       \
-        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts>=0)\
-            dynarec_log(LOG_NONE, ", jmp=%d", dyn->insts[ninst].x86.jmp_insts);\
-        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts==-1)\
-            dynarec_log(LOG_NONE, ", jmp=out");\
-        for(int ii=0; ii<24; ++ii) {            \
-            switch(dyn->insts[ninst].n.neoncache[ii].t) {    \
-                case NEON_CACHE_ST_D: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;            \
-                case NEON_CACHE_ST_F: dynarec_log(LOG_NONE, " S%d:%s", (ii+8)*2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;        \
-                case NEON_CACHE_MM: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;              \
-                case NEON_CACHE_XMMW: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;  \
-                case NEON_CACHE_XMMR: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;  \
-                case NEON_CACHE_SCR: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;             \
-                case NEON_CACHE_NONE:           \
-                default:    break;              \
-            }                                   \
-        }                                       \
-        if(dyn->n.stack || dyn->insts[ninst].n.stack_next || dyn->insts[ninst].n.x87stack)     \
-            dynarec_log(LOG_NONE, " X87:%d/%d(+%d/-%d)%d", dyn->n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, dyn->insts[ninst].n.stack_pop, dyn->insts[ninst].n.x87stack); \
-        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)                     \
-            dynarec_log(LOG_NONE, " %s:%d/%d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);   \
-        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");                       \
-    }
-
+#define INST_NAME(name) inst_name_pass3(dyn, ninst, name)
@@ -5,6 +5,7 @@

 typedef struct x86emu_s x86emu_t;
 typedef struct dynablock_s dynablock_t;
+typedef struct instsize_s instsize_t;

 #define BARRIER_MAYBE   8

@@ -68,7 +69,7 @@ typedef struct instruction_arm_s {
    int                 pred_sz;    // size of predecessor list
    int                 *pred;      // predecessor array
    uintptr_t           mark, mark2, mark3;
-    uintptr_t           markf;
+    uintptr_t           markf, markf2;
    uintptr_t           markseg;
    uintptr_t           marklock;
    int                 pass2choice;// value for choices that are fixed on pass2 for pass3
@@ -90,27 +91,35 @@ typedef struct dynarec_arm_s {
    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
+    uintptr_t           jmp_next;   // address of the jump_next address
    flagcache_t         f;
    neoncache_t         n;          // cache for the 8..31 double reg from fpu, plus x87 stack delta
    uintptr_t*          next;       // variable array of "next" jump address
    int                 next_sz;
    int                 next_cap;
-    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
+    int*                jmps;       // variable array of jump instructions
+    int                 jmp_sz;
+    int                 jmp_cap;
    int*                predecessor;// single array of all predecessor
    dynablock_t*        dynablock;
+    instsize_t*         instsize;
+    size_t              insts_size; // size of the instruction size array (calculated)
    uint8_t             smread;    // for strongmem model emulation
    uint8_t             smwrite;    // for strongmem model emulation
    uintptr_t           forward;    // address of the last end of code while testing forward
    uintptr_t           forward_to; // address of the next jump to (to check if everything is ok)
    int32_t             forward_size;   // size at the forward point
    int                 forward_ninst;  // ninst at the forward point
+    uint8_t             always_test;
 } 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);
+void add_jump(dynarec_arm_t *dyn, int ninst);
+int get_first_jump(dynarec_arm_t *dyn, int next);
 int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n);
 int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n);

+void CreateJmpNext(void* addr, void* next);
+
 #endif //__DYNAREC_ARM_PRIVATE_H_
@@ -44,4 +44,17 @@ typedef struct instruction_x86_s {

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

+//#define USE_CUSTOM_MEM
+#ifdef USE_CUSTOM_MEM
+#define dynaMalloc customMalloc
+#define dynaCalloc customCalloc
+#define dynaRealloc customRealloc
+#define dynaFree customFree
+#else
+#define dynaMalloc box_malloc
+#define dynaCalloc box_calloc
+#define dynaRealloc box_realloc
+#define dynaFree box_free
+#endif
+
 #endif //__DYNAREC_PRIVATE_H_
@@ -233,9 +233,9 @@ NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 0 size 1 @<1> Rm @ param=1 char align[4
 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> 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> 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 1 1 0 1 size @<1> @<1> 1 1 0 1 @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]!
-NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 1 1 @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]
-NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> Rm @ param2_1=1 char align[4] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\], <Rm>
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 0 1 @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]!
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> 1 1 1 1 @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\]
+NEON 1 1 1 1 0 1 0 0 1 D 1 0 Rn Vd 1 1 0 1 size @<1> @<1> Rm @ param2_1=1 char align[5] !@ %@ sprintf(align, ":%d", 16 << size)@ align[0] = '\0'@ @ VLD2<c>.\%d8 &l&l size\ {\vecname[0x20 + d]\[], \vecname[0x21 + param1_1 + d]\[]}, [<Rn>\align\], <Rm>
 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> 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> 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>
@@ -45,7 +45,9 @@ const char* DumpSection(Elf32_Shdr *s, char* SST) {
        GO(SHT_GNU_ATTRIBUTES);
        GO(SHT_GNU_HASH);
        GO(SHT_GNU_LIBLIST);
-        GO(SHT_CHECKSUM);
+        #ifndef TERMUX
+          GO(SHT_CHECKSUM);
+        #endif
        GO(SHT_LOSUNW);
        //GO(SHT_SUNW_move);
        GO(SHT_SUNW_COMDAT);
@@ -1,10 +1,6 @@
 #ifndef __ELFLOADER_PRIVATE_H_
 #define __ELFLOADER_PRIVATE_H_

-#ifdef DYNAREC
-typedef struct dynablocklist_s dynablocklist_t;
-#endif
-
 typedef struct library_s library_t;
 typedef struct needed_libs_s needed_libs_t;
 typedef struct kh_mapsymbols_s kh_mapsymbols_t;
@@ -12,9 +8,21 @@ typedef struct kh_mapsymbols_s kh_mapsymbols_t;
 #include <elf.h>
 #include "elfloader.h"

-struct elfheader_s {
+
+typedef struct multiblock_s {
+    void*       p;
+    uintptr_t   offs;
+    uintptr_t   paddr;
+    uintptr_t   align;
+    size_t      size;
+    size_t      asize;
+    uint8_t     flags;
+} multiblock_t;
+
+typedef struct elfheader_s {
    char*       name;
    char*       path;   // Resolved path to file
+    char*       soname; // soname of the elf
    uint16_t    numPHEntries;
    Elf32_Phdr  *PHEntries;
    uint16_t    numSHEntries;
@@ -38,8 +46,10 @@ struct elfheader_s {
    Elf32_Verdef*   VerDef;
    int         szVerDef;
    int         e_type;
+    uint32_t    flags;

    intptr_t    delta;  // should be 0
+    void*       image;

    uintptr_t   entrypoint;
    uintptr_t   initentry;
@@ -70,6 +80,9 @@ struct elfheader_s {
    int         textsz;
    uintptr_t   bss;
    int         bsssz;
+    uintptr_t   ehframe;
+    uintptr_t   ehframe_end;
+    uintptr_t   ehframehdr;

    uintptr_t   paddr;
    uintptr_t   vaddr;
@@ -87,20 +100,25 @@ struct elfheader_s {

    int         init_done;
    int         fini_done;
+    int         refcnt;     // ref count for the elf
+    int         malloc_hook_2;  // this elf hook malloc, hacking it

    char*       memory; // char* and not void* to allow math on memory pointer
-    void**      multiblock;
-    uintptr_t*  multiblock_offs;
-    uint32_t*   multiblock_size;
+    multiblock_t*  multiblocks;
    int         multiblock_n;

    library_t   *lib;
-    needed_libs_t *neededlibs;
+    needed_libs_t *needed;
+
+    FILE*       file;
+    int         fileno;

    kh_mapsymbols_t   *mapsymbols;
    kh_mapsymbols_t   *weaksymbols;
    kh_mapsymbols_t   *localsymbols;
-};
+    kh_defaultversion_t *globaldefver;  // the global default version for symbols (the XXX@@vvvv of symbols)
+    kh_defaultversion_t *weakdefver;    // the weak default version for symbols (the XXX@@vvvv of symbols)
+} elfheader_t;

 #define R_386_NONE	0
 #define R_386_32	1
@@ -242,8 +242,10 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
                    printf_dump(LOG_DEBUG, "The DT_INIT_ARRAYSZ is %d\n", h->initarray_sz);
                    break;
                case DT_PREINIT_ARRAYSZ:
-                    if(val)
-                        printf_dump(LOG_NONE, "Warning, PreInit Array (size=%d) present and ignored!\n", val);
+                    #ifndef ANDROID
+                        if(val)
+                            printf_dump(LOG_NONE, "Warning, PreInit Array (size=%d) present and ignored!\n", val);
+                    #endif
                    break;
                case DT_FINI: // Exit hook
                    h->finientry = ptr;
@@ -273,6 +275,10 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
                    h->VerDef = (Elf32_Verdef*)ptr;
                    printf_dump(LOG_DEBUG, "The DT_VERDEF is at address %p\n", h->VerDef);
                    break;
+                case DT_FLAGS:
+                    h->flags = val;
+                    printf_dump(LOG_DEBUG, "The DT_FLAGS is 0x%x\n", h->flags);
+                    break;
                }
            }
            if(h->rel) {
@@ -351,6 +357,17 @@ elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
            h->bsssz = h->SHEntries[ii].sh_size;
            printf_dump(LOG_DEBUG, "The .bss is at address %p, and is %d big\n", (void*)h->bss, h->bsssz);
        }
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".eh_frame");
+        if(ii) {
+            h->ehframe = (uintptr_t)(h->SHEntries[ii].sh_addr);
+            h->ehframe_end = h->ehframe + h->SHEntries[ii].sh_size;
+            printf_dump(LOG_DEBUG, "The .eh_frame section is at address %p..%p\n", (void*)h->ehframe, (void*)h->ehframe_end);
+        }
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".eh_frame_hdr");
+        if(ii) {
+            h->ehframehdr = (uintptr_t)(h->SHEntries[ii].sh_addr);
+            printf_dump(LOG_DEBUG, "The .eh_frame_hdr section is at address %p\n", (void*)h->ehframehdr);
+        }

        LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".dynstr", "DynSym Strings", SHT_STRTAB, (void**)&h->DynStr, NULL);
        LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".dynsym", "DynSym", SHT_DYNSYM, (void**)&h->DynSym, &h->numDynSym);
@@ -399,3 +416,74 @@ const char* GetParentSymbolVersion(elfheader_t* h, int index)
    }
    return NULL;
 }
+
+int GetVersionIndice(elfheader_t* h, const char* vername)
+{
+    if(!vername)
+        return 0;
+    if(h->VerDef) {
+        Elf32_Verdef *def = (Elf32_Verdef*)((uintptr_t)h->VerDef + h->delta);
+        while(def) {
+            Elf32_Verdaux *aux = (Elf32_Verdaux*)((uintptr_t)def + def->vd_aux);
+            if(!strcmp(h->DynStr+aux->vda_name, vername))
+                return def->vd_ndx;
+            def = def->vd_next?((Elf32_Verdef*)((uintptr_t)def + def->vd_next)):NULL;
+        }
+    }
+    return 0;
+}
+
+int GetNeededVersionCnt(elfheader_t* h, const char* libname)
+{
+    if(!libname)
+        return 0;
+    if(h->VerNeed) {
+        Elf32_Verneed *ver = (Elf32_Verneed*)((uintptr_t)h->VerNeed + h->delta);
+        while(ver) {
+            char *filename = h->DynStr + ver->vn_file;
+            Elf32_Vernaux *aux = (Elf32_Vernaux*)((uintptr_t)ver + ver->vn_aux);
+            if(!strcmp(filename, libname))
+                return ver->vn_cnt;
+            ver = ver->vn_next?((Elf32_Verneed*)((uintptr_t)ver + ver->vn_next)):NULL;
+        }
+    }
+    return 0;
+}
+
+const char* GetNeededVersionString(elfheader_t* h, const char* libname, int idx)
+{
+    if(!libname)
+        return 0;
+    if(h->VerNeed) {
+        Elf32_Verneed *ver = (Elf32_Verneed*)((uintptr_t)h->VerNeed + h->delta);
+        while(ver) {
+            char *filename = h->DynStr + ver->vn_file;
+            Elf32_Vernaux *aux = (Elf32_Vernaux*)((uintptr_t)ver + ver->vn_aux);
+            if(!strcmp(filename, libname)) {
+                for(int j=0; j<ver->vn_cnt; ++j) {
+                    if(j==idx) 
+                        return h->DynStr+aux->vna_name;
+                    aux = (Elf32_Vernaux*)((uintptr_t)aux + aux->vna_next);
+                }
+                return NULL;    // idx out of bound, return NULL...
+           }
+            ver = ver->vn_next?((Elf32_Verneed*)((uintptr_t)ver + ver->vn_next)):NULL;
+        }
+    }
+    return NULL;
+}
+
+int GetNeededVersionForLib(elfheader_t* h, const char* libname, const char* ver)
+{
+    if(!libname || !ver)
+        return 0;
+    int n = GetNeededVersionCnt(h, libname);
+    if(!n)
+        return 0;
+    for(int i=0; i<n; ++i) {
+        const char* vername = GetNeededVersionString(h, libname, i);
+        if(vername && !strcmp(ver, vername))
+            return 1;
+    }
+    return 0;
+}
@@ -1,4 +1,27 @@
+#include <string.h>
+
+#define F8      *(uint8_t*)(addr++)
+#define F8S     *(int8_t*)(addr++)
+#define F16     *(uint16_t*)(addr+=2, addr-2)
+#define F16S    *(int16_t*)(addr+=2, addr-2)
+#define F32     *(uint32_t*)(addr+=4, addr-4)
+#define F32S    *(int32_t*)(addr+=4, addr-4)
+#define F32S64  (uint64_t)(int64_t)F32S
+#define F64     *(uint64_t*)(addr+=8, addr-8)
+#define F64S    *(int64_t*)(addr+=8, addr-8)
+#define PK(a)   *(uint8_t*)(addr+a)
+#ifdef DYNAREC
+#define STEP if(step) return 0;
+#define STEP2 if(step) {R_EIP = addr; return 0;}
+#define STEP3 if(*step) (*step)++;
+#else
+#define STEP
+#define STEP2
+#define STEP3
+#endif
+
 // ModRM utilities macros
+#define MODREG ((nextop&0xC0)==0xC0)
 #define getecommon(A, T) \
    if(!(nextop&0xC0)) { \
        if((nextop&7)==4) { \
@@ -85,7 +108,7 @@
            case 7: base =      R_BX; break;    \
        }                                       \
        switch((m>>6)&3) {                      \
-            case 0: if(m==6) base = F16; break; \
+            case 0: if(m&7==6) base = F16S; break; \
            case 1: base += F8S; break;         \
            case 2: base += F16S; break;        \
        }                                       \
@@ -94,47 +117,119 @@


 #define geteb(A) \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
    } else getecommon(A, reg32_t)
+#define testeb(A, O) \
+    if(MODREG) { \
+        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
+    } else { \
+        reg32_t* ret; \
+        getecommono(ret, reg32_t, O);\
+        test->memsize = 1; \
+        test->memaddr = (uintptr_t)ret;\
+        test->mem[0] = ret->byte[0];\
+        A = (reg32_t*)test->mem;\
+    }
 #define getebo(A, O)          \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
    } else getecommono(A, reg32_t, O)
 #define geted(A) \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = &emu->regs[(nextop&7)]; \
    } else getecommon(A, reg32_t)
 #define getedo(A, O) \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = &emu->regs[(nextop&7)]; \
    } else getecommono(A, reg32_t, O)
+#define tested(A, SZ, O) \
+    if(MODREG) { \
+        A = &emu->regs[(nextop&7)]; \
+    } else {\
+        reg32_t* ret; \
+        getecommono(ret, reg32_t, O);\
+        test->memsize = SZ; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->dword, SZ);\
+        A = (reg32_t*)test->mem;\
+    }
 #define getem(A) \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = &emu->mmx[(nextop&7)]; \
    } else getecommon(A, mmx87_regs_t)
+#define testem(A) \
+    if(MODREG) { \
+        A = &emu->mmx[(nextop&7)]; \
+    } else {\
+        mmx87_regs_t* ret; \
+        getecommon(ret, mmx87_regs_t) \
+        test->memsize = 8; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->ub, 8);\
+        A = (mmx87_regs_t*)test->mem;\
+    }
 #define getex(A) \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = &emu->xmm[(nextop&7)]; \
    } else getecommon(A, sse_regs_t)
+#define testex(A) \
+    if(MODREG) { \
+        A = &emu->xmm[(nextop&7)]; \
+    } else {\
+        sse_regs_t* ret; \
+        getecommon(ret, sse_regs_t);\
+        test->memsize = 16; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->ub, 16);\
+        A = (sse_regs_t*)test->mem;\
+    }
 #define getew16(A)  \
-    if((nextop&0xC0)==0xC0) { \
+    if(MODREG) { \
        A = &emu->regs[(nextop&7)]; \
    } else getecommon16(A, reg32_t)
 #define getew16o(A, O)              \
-    if((nextop&0xC0)==0xC0) {       \
+    if(MODREG) {       \
        A = &emu->regs[(nextop&7)]; \
    } else getecommon16o(A, reg32_t, O)
+#define testtew16(A, O)                 \
+    if(MODREG) {           \
+        A = &emu->regs[(nextop&7)];     \
+    } else {                            \
+        reg32_t* ret;                   \
+        getecommon16o(A, reg32_t, O);   \
+        test->memsize = 2;              \
+        test->memaddr = (uintptr_t)ret; \
+        *(uint16_t*)test->mem = ret->word[0];\
+        A = (reg32_t*)test->mem;        \
+    }

 // Macros for ModR/M gets
+#ifdef TEST_INTERPRETER
+#define GET_EB      testeb(oped, 0)
+#define GET_ED      tested(oped, 4, 0)
+#define GET_ED8     tested(oped, 8, 0)
+#define GET_EDT     tested(oped, 10, 0)
+#define GET_ED_OFFS(o) tested(oped, 4, o)
+#define GET_EB_OFFS(o) testeb(oped, o)
+#define GET_EM      testem(opem)
+#define GET_EX      testex(opex)
+#define GET_EW16    testew16(oped, 0)
+#define GET_EW16_OFFS(o)    testtew16(oped, o)
+#else
 #define GET_EB      geteb(oped)
 #define GET_ED      geted(oped)
+#define GET_ED8     geted(oped)
+#define GET_EDT     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)
+#endif
+#define GET_ED_     geted(oped)
+#define GET_EW16_   getew16(oped)
 #define EB          oped
 #define ED          oped
 #define EM          opem
@@ -1,555 +0,0 @@
-    opcode = F8;
-    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;
-
-    GO(0x00, add)                   /* ADD 0x01 ~> 0x05 */
-    GO(0x08, or)                    /*  OR 0x09 ~> 0x0D */
-    GO(0x10, adc)                   /* ADC 0x11 ~> 0x15 */
-    GO(0x18, sbb)                   /* SBB 0x19 ~> 0x1D */
-    GO(0x20, and)                   /* AND 0x21 ~> 0x25 */
-    GO(0x28, sub)                   /* SUB 0x29 ~> 0x2D */
-    GO(0x30, xor)                   /* XOR 0x31 ~> 0x35 */
-    //GO(0x38, cmp)                   /* CMP 0x39 ~> 0x3D */
-    #undef GO
-    _66_0x06:                      /* PUSH ES */
-        Push16(emu, emu->segs[_ES]);
-        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 */
-        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_0x1E:                      /* PUSH DS */
-        Push16(emu, emu->segs[_DS]);
-        NEXT;
-    _66_0x1F:                      /* POP DS */
-        emu->segs[_DS] = Pop16(emu);    // no check, no use....
-        emu->segs_serial[_DS] = 0;
-        NEXT;
-        
-    _66_0x26:                      /* ES: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    _66_0x2E:                      /* CS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    _66_0x36:                      /* SS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-
-    _66_0x39:
-        nextop = F8;
-        GET_EW;
-        cmp16(emu, EW->word[0], GW.word[0]);
-        NEXT;
-    _66_0x3B:
-        nextop = F8;
-        GET_EW;
-        cmp16(emu, GW.word[0], EW->word[0]);
-        NEXT;
-    _66_0x3D:
-        cmp16(emu, R_AX, F16);
-        NEXT;
-    _66_0x3E:                      /* DS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    
-    _66_0x40:
-    _66_0x41:
-    _66_0x42:
-    _66_0x43:
-    _66_0x44:
-    _66_0x45:
-    _66_0x46:
-    _66_0x47:                              /* INC Reg */
-        tmp8u = opcode&7;
-        emu->regs[tmp8u].word[0] = inc16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x48:
-    _66_0x49:
-    _66_0x4A:
-    _66_0x4B:
-    _66_0x4C:
-    _66_0x4D:
-    _66_0x4E:
-    _66_0x4F:                              /* DEC Reg */
-        tmp8u = opcode&7;
-        emu->regs[tmp8u].word[0] = dec16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x50:
-    _66_0x51:
-    _66_0x52:
-    _66_0x53:
-    _66_0x54:
-    _66_0x55:
-    _66_0x56:
-    _66_0x57:                              /* PUSH Reg */
-        tmp8u = opcode&7;
-        Push16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x58:
-    _66_0x59:
-    _66_0x5A:
-    _66_0x5B:
-    _66_0x5C:                      
-    _66_0x5D:
-    _66_0x5E:
-    _66_0x5F:                              /* POP Reg */
-        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
-
-    _66_0x68:                       /* PUSH u16 */
-        tmp16u = F16;
-        Push16(emu, tmp16u);
-        NEXT;
-    _66_0x69:                      /* IMUL Gw,Ew,Iw */
-        nextop = F8;
-        GET_EW;
-        tmp16u = F16;
-        GW.word[0] = imul16(emu, EW->word[0], tmp16u);
-        NEXT;
-    _66_0x6A:                      /* PUSH Ib (as signed word) */
-        tmp16s = F8S;
-        Push16(emu, (uint16_t)tmp16s);
-        NEXT;
-    _66_0x6B:                      /* IMUL Gw,Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        tmp16s = F8S;
-        GW.word[0] = imul16(emu, EW->word[0], (uint16_t)tmp16s);
-        NEXT;
-
-    _66_0x81:                              /* GRP3 Ew,Iw */
-    _66_0x83:                              /* GRP3 Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        if(opcode==0x81) 
-            tmp16u = F16;
-        else {
-            tmp16s = F8S;
-            tmp16u = (uint16_t)tmp16s;
-        }
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = add16(emu, EW->word[0], tmp16u); break;
-            case 1: EW->word[0] =  or16(emu, EW->word[0], tmp16u); break;
-            case 2: EW->word[0] = adc16(emu, EW->word[0], tmp16u); break;
-            case 3: EW->word[0] = sbb16(emu, EW->word[0], tmp16u); break;
-            case 4: EW->word[0] = and16(emu, EW->word[0], tmp16u); break;
-            case 5: EW->word[0] = sub16(emu, EW->word[0], tmp16u); break;
-            case 6: EW->word[0] = xor16(emu, EW->word[0], tmp16u); break;
-            case 7:               cmp16(emu, EW->word[0], tmp16u); break;
-        }
-        NEXT;
-
-    _66_0x85:                              /* TEST Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        test16(emu, EW->word[0], GW.word[0]);
-        NEXT;
-
-    _66_0x87:                              /* XCHG Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        tmp16u = GW.word[0];
-        GW.word[0] = EW->word[0];
-        EW->word[0] = tmp16u;
-        NEXT;
-
-    _66_0x89:                              /* MOV Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        EW->word[0] = GW.word[0];
-        NEXT;
-
-    _66_0x8B:                              /* MOV Gw,Ew */
-        nextop = F8;
-        GET_EW;
-        GW.word[0] = EW->word[0];
-        NEXT;
-    _66_0x8C:                              /* MOV Ew,Seg */
-        nextop = F8;
-        GET_EW;
-        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;
-        GET_EW;
-        EW->word[0] = Pop16(emu);
-        NEXT;
-    _66_0x90:                              /* NOP */
-        NEXT;
-
-    _66_0x91:
-    _66_0x92:
-    _66_0x93:
-    _66_0x94:
-    _66_0x95:
-    _66_0x96:
-    _66_0x97:                      /* XCHG reg,EAX */
-        tmp16u = R_AX;
-        R_AX = emu->regs[opcode&7].word[0];
-        emu->regs[opcode&7].word[0] = tmp16u;
-        NEXT;
-
-    _66_0x98:                               /* CBW */
-        emu->regs[_AX].sword[0] = emu->regs[_AX].sbyte[0];
-        NEXT;
-    _66_0x99:                              /* CWD */
-        R_DX=((R_AX & 0x8000)?0xFFFF:0x0000);
-        NEXT;
-
-    _66_0x9C:                              /* PUSHFW */
-        CHECK_FLAGS(emu);
-        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 */
-        R_AX = *(uint16_t*)F32;
-        NEXT;
-    _66_0xA3:                              /* MOV Ow,AX */
-        *(uint16_t*)F32 = R_AX;
-        NEXT;
-    _66_0xA5:                              /* MOVSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        *(uint16_t*)R_EDI = *(uint16_t*)R_ESI;
-        R_EDI += tmp8s;
-        R_ESI += tmp8s;
-        NEXT;
-
-    _66_0xA7:                              /* CMPSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        tmp16u  = *(uint16_t*)R_EDI;
-        tmp16u2 = *(uint16_t*)R_ESI;
-        R_EDI += tmp8s;
-        R_ESI += tmp8s;
-        cmp16(emu, tmp16u2, tmp16u);
-        NEXT;
-
-    _66_0xA9:                             /* TEST AX,Iw */
-        test16(emu, R_AX, F16);
-        NEXT;
-
-    _66_0xAB:                              /* STOSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        *(uint16_t*)R_EDI = R_AX;
-        R_EDI += tmp8s;
-        NEXT;
-    _66_0xAD:                              /* LODSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        R_AX = *(uint16_t*)R_ESI;
-        R_ESI += tmp8s;
-        NEXT;
-    _66_0xAF:                              /* SCASW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        cmp16(emu, R_AX, *(uint16_t*)R_EDI);
-        R_EDI += tmp8s;
-        NEXT;
-
-    _66_0xB8:                              /* MOV AX,Iw */
-    _66_0xB9:                              /* MOV CX,Iw */
-    _66_0xBA:                              /* MOV DX,Iw */
-    _66_0xBB:                              /* MOV BX,Iw */
-    _66_0xBC:                              /*    ...     */
-    _66_0xBD:
-    _66_0xBE:
-    _66_0xBF:
-        emu->regs[opcode&7].word[0] = F16;
-        NEXT;
-
-    _66_0xC1:                              /* GRP2 Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        tmp8u = F8 /*& 0x1f*/;
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = rol16(emu, EW->word[0], tmp8u); break;
-            case 1: EW->word[0] = ror16(emu, EW->word[0], tmp8u); break;
-            case 2: EW->word[0] = rcl16(emu, EW->word[0], tmp8u); break;
-            case 3: EW->word[0] = rcr16(emu, EW->word[0], tmp8u); break;
-            case 4:
-            case 6: EW->word[0] = shl16(emu, EW->word[0], tmp8u); break;
-            case 5: EW->word[0] = shr16(emu, EW->word[0], tmp8u); break;
-            case 7: EW->word[0] = sar16(emu, EW->word[0], tmp8u); break;
-        }
-        NEXT;
-
-    _66_0xC7:                              /* MOV Ew,Iw */
-        nextop = F8;
-        GET_EW;
-        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;
-        GET_EW;
-        tmp8u=(opcode==0xD3)?R_CL:1;
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = rol16(emu, EW->word[0], tmp8u); break;
-            case 1: EW->word[0] = ror16(emu, EW->word[0], tmp8u); break;
-            case 2: EW->word[0] = rcl16(emu, EW->word[0], tmp8u); break;
-            case 3: EW->word[0] = rcr16(emu, EW->word[0], tmp8u); break;
-            case 4: 
-            case 6: EW->word[0] = shl16(emu, EW->word[0], tmp8u); break;
-            case 5: EW->word[0] = shr16(emu, EW->word[0], tmp8u); break;
-            case 7: EW->word[0] = sar16(emu, EW->word[0], tmp8u); break;
-        }
-        NEXT;
-    
-    _66_0xD9:
-        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 = 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 */
-        nextop = F8;
-        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;
-                    *(uint16_t*)R_EDI = *(uint16_t*)R_ESI;
-                    R_EDI += tmp8s;
-                    R_ESI += tmp8s;
-                }
-                break;
-            case 0xA7:              /* REP(N)Z CMPSW */
-                tmp16u = 0;
-                tmp16u2 = 0;
-                while(tmp32u) {
-                    --tmp32u;
-                    tmp16u  = *(uint16_t*)R_EDI;
-                    tmp16u2 = *(uint16_t*)R_ESI;
-                    R_EDI += tmp8s;
-                    R_ESI += tmp8s;
-                    if((tmp16u==tmp16u2)==(opcode==0xF2))
-                        break;
-                }
-                if(R_ECX) cmp16(emu, tmp16u2, tmp16u);
-                break;
-            case 0xAB:              /* REP STOSW */
-                while(tmp32u) {
-                    --tmp32u;
-                    *(uint16_t*)R_EDI = R_AX;
-                    R_EDI += tmp8s;
-                }
-                break;
-            case 0xAD:              /* REP LODSW */
-                while(tmp32u) {
-                    --tmp32u;
-                    R_AX = *(uint16_t*)R_ESI;
-                    R_ESI += tmp8s;
-                }
-                break;
-            case 0xAF:              /* REP(N)Z SCASW */
-                tmp16u = 0;
-                while(tmp32u) {
-                    --tmp32u;
-                    tmp16u = *(uint16_t*)R_EDI;
-                    R_EDI += tmp8s;
-                    if((R_AX==tmp16u)==(opcode==0xF2))
-                        break;
-                }
-                if(R_ECX) cmp16(emu, R_AX, tmp16u);
-                break;
-            default:
-                goto _default;
-        }
-        R_ECX = tmp32u;
-        NEXT;
-
-    _66_0xF5:                      /* CMC */
-        CHECK_FLAGS(emu);
-        CONDITIONAL_SET_FLAG(!ACCESS_FLAG(F_CF), F_CF);
-        NEXT;
-
-    _66_0xF7:                      /* GRP3 Ew(,Iw) */
-        nextop = F8;
-        GET_EW;
-        switch((nextop>>3)&7) {
-            case 0: 
-            case 1:                 /* TEST Ew,Iw */
-                test16(emu, EW->word[0], F16);
-                break;
-            case 2:                 /* NOT Ew */
-                EW->word[0] = not16(emu, EW->word[0]);
-                break;
-            case 3:                 /* NEG Ew */
-                EW->word[0] = neg16(emu, EW->word[0]);
-                break;
-            case 4:                 /* MUL AX,Ew */
-                mul16(emu, EW->word[0]);
-                break;
-            case 5:                 /* IMUL AX,Ew */
-                imul16_eax(emu, EW->word[0]);
-                break;
-            case 6:                 /* DIV Ew */
-                div16(emu, EW->word[0]);
-                break;
-            case 7:                 /* IDIV Ew */
-                idiv16(emu, EW->word[0]);
-                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;
-        GET_EW;
-        switch((nextop>>3)&7) {
-            case 0:                 /* INC Ed */
-                EW->word[0] = inc16(emu, EW->word[0]);
-                break;
-            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 = R_EIP;
-                R_EIP = ip;
-                printf_log(LOG_NONE, "Illegal Opcode 66 %02X %02X\n", opcode, nextop);
-                emu->quit=1;
-                emu->error |= ERR_ILLEGAL;
-                goto fini;
-        }
-        NEXT;
@@ -0,0 +1,120 @@
+#include <stdint.h>
+
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run_private.h"
+#include "x86emu_private.h"
+#include "x86compstrings.h"
+
+static int overrideIfDataInvalid(sse_regs_t* mem, int lmem, sse_regs_t* reg, int lreg, int j, int i, int imm8)
+{
+    int valid1 = (i<lreg);
+    int valid2 = (j<lmem);
+    if(!valid1 && !valid2)
+        switch((imm8>>2)&3) {
+            case 0b00:
+            case 0b01:  return 0;
+            case 0b10:
+            case 0b11:  return 1;
+        }
+    if(!valid1 && valid2)
+        switch((imm8>>2)&3) {
+            case 0b00:
+            case 0b01:
+            case 0b10:  return 0;
+            case 0b11:  return 1;
+        }
+    if(valid1 && !valid2)
+        return 0;
+    switch((imm8>>2)&3) {
+        case 0b01:  // range
+            switch (imm8&3) {
+                case 0b00:  // ub
+                    return (i&1)?((reg->ub[i]>=mem->ub[j])):((reg->ub[i]<=mem->ub[j]));
+                case 0b01:  // uw
+                    return (i&1)?((reg->uw[i]>=mem->uw[j])):((reg->uw[i]<=mem->uw[j]));
+                case 0b10:  // sb
+                    return (i&1)?((reg->sb[i]>=mem->sb[j])):((reg->sb[i]<=mem->sb[j]));
+                case 0b11:  // sw
+                    return (i&1)?((reg->sw[i]>=mem->sw[j])):((reg->sw[i]<=mem->sw[j]));
+            }
+            break;
+        default:    // the others
+            switch (imm8&1) {
+                case 0: // byte
+                    return (reg->ub[i] == mem->ub[j]);
+                case 1: // word
+                    return (reg->uw[i] == mem->uw[j]);
+            }
+    }
+}
+
+uint32_t sse42_compare_string_explicit_len(x86emu_t* emu, sse_regs_t* mem, int lmem, sse_regs_t* reg, int lreg, uint8_t imm8)
+{
+    // get number of packed byte/word
+    int n_packed = (imm8&1)?8:16;
+    if(lreg<0) lreg = -lreg;
+    if(lmem<0) lmem = -lmem;
+    if(lreg>n_packed) lreg = n_packed;
+    if(lmem>n_packed) lmem = n_packed;
+    // aggregate to intres1
+    uint32_t intres1 = 0;
+    switch((imm8>>2)&3) {
+        case 0b00:  //Equal any
+            for(int j=0; j<n_packed; ++j)
+                for(int i=0; i<n_packed; ++i) {
+                    intres1 |= overrideIfDataInvalid(mem, lmem, reg, lreg, j, i, imm8)<<j;
+                }
+            break;
+        case 0b01:  // Range
+            for(int j=0; j<n_packed; ++j)
+                for(int i=0; i<n_packed; i+=2) {
+                    intres1 |= (overrideIfDataInvalid(mem, lmem, reg, lreg, j, i, imm8) & overrideIfDataInvalid(mem, lmem, reg, lreg, j, i+1, imm8))<<j;
+                }
+            break;
+        case 0b10:  // Equal each
+            for(int i=0; i<n_packed; ++i) {
+                intres1 |= overrideIfDataInvalid(mem, lmem, reg, lreg, i, i, imm8)<<i;
+            }
+            break;
+        case 0b11:  // Equal ordered
+            intres1 = (1<<n_packed)-1;
+            for(int j=0; j<n_packed; ++j)
+                for(int i=0; i<n_packed-j; ++i) {
+                    int k = i+j;
+                    intres1 &= (((1<<n_packed)-1)^(1<<j)) | overrideIfDataInvalid(mem, lmem, reg, lreg, k, i, imm8)<<j;
+                }
+            break;
+    }
+    // build intres2
+    uint32_t intres2 = intres1;
+    switch((imm8>>4)&3) {
+        case 0b01: intres2 ^= ((1<<n_packed)-1); break;
+        case 0b11: intres2 ^= ((1<<lmem)-1); break;
+    }
+    // and now set the flags
+    RESET_FLAGS(emu);
+    CONDITIONAL_SET_FLAG(intres2, F_CF);
+    CONDITIONAL_SET_FLAG(lmem<n_packed, F_ZF);
+    CONDITIONAL_SET_FLAG(lreg<n_packed, F_SF);
+    CONDITIONAL_SET_FLAG(intres2&1, F_OF);
+    CLEAR_FLAG(F_AF);
+    CLEAR_FLAG(F_PF);
+    
+    return intres2;
+}
+
+uint32_t sse42_compare_string_implicit_len(x86emu_t* emu, sse_regs_t* mem, sse_regs_t* reg, uint8_t imm8)
+{
+    int lmem = 0;
+    int lreg = 0;
+    // get lmem and lreg
+    if(imm8&1) {
+        while(lmem<8 && mem->uw[lmem]) ++lmem;
+        while(lreg<8 && reg->uw[lreg]) ++lreg;
+    } else {
+        while(lmem<16 && mem->ub[lmem]) ++lmem;
+        while(lreg<16 && reg->ub[lreg]) ++lreg;
+    }
+    return sse42_compare_string_explicit_len(emu, mem, lmem, reg, lreg, imm8);
+}
@@ -0,0 +1,13 @@
+#ifndef __X86_CMPSTRINGS_H__
+#define __X86_CMPSTRINGS_H__
+
+#include <stdint.h>
+
+#include "regs.h"
+
+typedef struct x86emu_s x86emu_t;
+
+uint32_t sse42_compare_string_explicit_len(x86emu_t* emu, sse_regs_t* a, int la, sse_regs_t* b, int lb, uint8_t imm8);
+uint32_t sse42_compare_string_implicit_len(x86emu_t* emu, sse_regs_t* a, sse_regs_t* b, uint8_t imm8);
+
+#endif //__X86_CMPSTRINGS_H__
@@ -54,17 +54,6 @@ uint32_t* GetParityTab()
    return x86emu_parity_tab;
 }

-void PushExit(x86emu_t* emu)
-{
-    uintptr_t endMarker = AddCheckBridge(my_context->system, NULL, NULL, 0, "(ExitEmulation)");
-    Push(emu, endMarker);
-}
-
-void* GetExit()
-{
-    return (void*)AddCheckBridge(my_context->system, NULL, NULL, 0, "(ExitEmulation)");
-}
-
 static void internalX86Setup(x86emu_t* emu, box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
 {
    emu->context = context;
@@ -198,6 +187,11 @@ void FreeX86Emu(x86emu_t **emu)
        return;
    printf_log(LOG_DEBUG, "%04d|Free a X86 Emu (%p)\n", GetTID(), *emu);

+    if((*emu)->test.emu) {
+        internalFreeX86((*emu)->test.emu);
+        box_free((*emu)->test.emu);
+        (*emu)->test.emu = NULL;
+    }
    internalFreeX86(*emu);

    box_free(*emu);
@@ -231,6 +225,10 @@ void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
 	newemu->top = emu->top;
    newemu->fpu_stack = emu->fpu_stack;
    memcpy(newemu->xmm, emu->xmm, sizeof(emu->xmm));
+    newemu->df = emu->df;
+    newemu->op1 = emu->op1;
+    newemu->op2 = emu->op2;
+    newemu->res = emu->res;
    newemu->mxcsr = emu->mxcsr;
    newemu->quit = emu->quit;
    newemu->error = emu->error;
@@ -240,6 +238,34 @@ void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
    newemu->regs[_SP].dword[0] = emu->regs[_SP].dword[0] + (intptr_t)(newst - oldst);
 }

+void CopyEmu(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->eflags, &emu->eflags, sizeof(emu->eflags));
+    newemu->old_ip = emu->old_ip;
+    memcpy(newemu->segs, emu->segs, sizeof(emu->segs));
+    memcpy(newemu->segs_serial, emu->segs_serial, sizeof(emu->segs_serial));
+    memcpy(newemu->segs_offs, emu->segs_offs, sizeof(emu->segs_offs));
+	memcpy(newemu->x87, emu->x87, sizeof(emu->x87));
+	memcpy(newemu->mmx, emu->mmx, sizeof(emu->mmx));
+    memcpy(newemu->xmm, emu->xmm, sizeof(emu->xmm));
+    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));
+	newemu->cw = emu->cw;
+	newemu->sw = emu->sw;
+	newemu->top = emu->top;
+    newemu->fpu_stack = emu->fpu_stack;
+    newemu->df = emu->df;
+    newemu->op1 = emu->op1;
+    newemu->op2 = emu->op2;
+    newemu->res = emu->res;
+    newemu->mxcsr = emu->mxcsr;
+    newemu->quit = emu->quit;
+    newemu->error = emu->error;
+}
+
 box86context_t* GetEmuContext(x86emu_t* emu)
 {
    return emu->context;
@@ -311,7 +337,7 @@ void ResetFlags(x86emu_t *emu)
 const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
 {
    static char buff[800];
-    char* regname[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
+    static const char* regname[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
    char tmp[80];
    buff[0] = '\0';
    if(trace_emm) {
@@ -396,8 +422,6 @@ void UnimpOpcode(x86emu_t* emu)
        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;
-    emu->error |= ERR_UNIMPL;
 }

 void EmuCall(x86emu_t* emu, uintptr_t addr)
@@ -414,9 +438,9 @@ void EmuCall(x86emu_t* emu, uintptr_t addr)
    Run(emu, 0);
    emu->quit = 0;  // reset Quit flags...
    emu->df = d_none;
-    if(emu->quitonlongjmp && emu->longjmp) {
-        if(emu->quitonlongjmp==1)
-            emu->longjmp = 0;   // don't change anything because of the longjmp
+    if(emu->flags.quitonlongjmp && emu->flags.longjmp) {
+        if(emu->flags.quitonlongjmp==1)
+            emu->flags.longjmp = 0;   // don't change anything because of the longjmp
    } else {
        R_EBX = old_ebx;
        R_EDI = old_edi;
@@ -475,8 +499,8 @@ uint64_t ReadTSC(x86emu_t* emu)
    uint64_t ret;
    __asm__ volatile("rdtsc" : "=A"(ret));
    return ret;
-#if 0
 #elif defined(__ARM_ARCH)
+#if 0
 #if (__ARM_ARCH >= 6)
    static int init = 1;
    if(init) {
@@ -23,6 +23,32 @@ typedef struct forkpty_s {
    void*   f;  // forkpty function
 } forkpty_t;

+typedef struct x86emu_s x86emu_t;
+
+typedef struct x86test_s {
+    x86emu_t*   emu;
+    uintptr_t   memaddr;
+    int         memsize;
+    int         test;
+    int         clean;
+    uint8_t     mem[16];
+} x86test_t;
+
+typedef struct emu_flags_s {
+    uint32_t    need_jmpbuf:1;    // need a new jmpbuff for signal handling
+    uint32_t    quitonlongjmp:2;  // quit if longjmp is called
+    uint32_t    quitonexit:2;     // quit if exit/_exit is called
+    uint32_t    longjmp:1;        // if quit because of longjmp
+    uint32_t    jmpbuf_ready:1;   // the jmpbuf in the emu is ok and don't need refresh
+} emu_flags_t;
+
+#ifdef ANDROID
+#include <setjmp.h>
+#define JUMPBUFF sigjmp_buf
+#else
+#define JUMPBUFF struct __jmp_buf_tag
+#endif
+
 typedef struct x86emu_s {
    // cpu
 	reg32_t     regs[8];
@@ -52,17 +78,18 @@ typedef struct x86emu_s {
    uintptr_t   prev2_ip, prev_ip;
    #endif
    // segments
-    uint32_t    segs[6];        // only 32bits value?
+    uint16_t    segs[6];
+    uint16_t    dummy_seg6, dummy_seg7; // to stay aligned
    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;
    int         fork;   // quit because need to fork
-    forkpty_t*  forkpty_info;
    int         exit;
-    int         quitonlongjmp;  // quit if longjmp is called
-    int         longjmp;        // if quit because of longjmp
+    forkpty_t*  forkpty_info;
+    emu_flags_t flags;
+    x86test_t   test;       // used for dynarec testing
    // parent context
    box86context_t *context;
    // cpu helpers
@@ -74,6 +101,7 @@ typedef struct x86emu_s {
    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)
+    JUMPBUFF*   jmpbuf;

    i386_ucontext_t *uc_link; // to handle setcontext

@@ -67,7 +67,7 @@ x86emu_t* x86emu_fork(x86emu_t* emu, int forktype)
    R_EAX = v;
    return emu;
 }
-
+void emit_signal(x86emu_t* emu, int sig, void* addr, int code); // in signals.c
 extern int errno;
 void x86Int3(x86emu_t* emu)
 {
@@ -350,8 +350,8 @@ void x86Int3(x86emu_t* emu)
        }
        return;
    }
-    if(0 && my_context->signals[SIGTRAP])
-        raise(SIGTRAP);
+    if(1 && my_context->signals[SIGTRAP])
+        emit_signal(emu, SIGTRAP, (void*)R_EIP, 128);
    else
        printf_log(LOG_INFO, "%04d|Warning, ignoring unsupported Int 3 call @%p\n", GetTID(), (void*)R_EIP);
    //emu->quit = 1;
@@ -484,10 +484,8 @@ uint8_t rcl8(x86emu_t *emu, uint8_t d, uint8_t s)
 		CONDITIONAL_SET_FLAG(cf, F_CF);
        /* OVERFLOW is set *IFF* cnt==1, then it is the 
           xor of CF and the most significant bit.  Blecck. */
-        /* parenthesized this expression since it appears to
-           be causing OF to be misset */
-        CONDITIONAL_SET_FLAG(cnt == 1 && XOR2(cf + ((res >> 6) & 0x2)),
-							 F_OF);
+		if(cnt == 1)
+        	CONDITIONAL_SET_FLAG((cf ^ (res >> 7)) & 0x1, F_OF);

    }
 	return (uint8_t)res;
@@ -513,8 +511,8 @@ uint16_t rcl16(x86emu_t *emu, uint16_t d, uint8_t s)
 			res |= 1 << (cnt - 1);
 		}
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		CONDITIONAL_SET_FLAG(cnt == 1 && XOR2(cf + ((res >> 14) & 0x2)),
-							 F_OF);
+		if(cnt == 1)
+        	CONDITIONAL_SET_FLAG((cf ^ (res >> 15)) & 0x1, F_OF);
 	}
 	return (uint16_t)res;
 }
@@ -539,8 +537,8 @@ uint32_t rcl32(x86emu_t *emu, uint32_t d, uint8_t s)
 			res |= 1 << (cnt - 1);
 		}
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		CONDITIONAL_SET_FLAG(cnt == 1 && XOR2(cf + ((res >> 30) & 0x2)),
-							 F_OF);
+		if(cnt == 1)
+        	CONDITIONAL_SET_FLAG((cf ^ (res >> 31)) & 0x1, F_OF);
 	}
 	return res;
 }
@@ -592,6 +590,11 @@ uint8_t rcr8(x86emu_t *emu, uint8_t d, uint8_t s)
               (i.e. packed bit array or unpacked.)
             */
 			ocf = ACCESS_FLAG(F_CF) != 0;
+			/* OVERFLOW is set *IFF* cnt==1, then it is the 
+			xor of CF and the most significant bit.  Blecck. */
+			/* parenthesized... */
+			CONDITIONAL_SET_FLAG((ocf ^ (d >> 7)) & 0x1,
+								 F_OF);
        } else
            cf = (d >> (cnt - 1)) & 0x1;

@@ -620,13 +623,6 @@ uint8_t rcr8(x86emu_t *emu, uint8_t d, uint8_t s)
        }
        /* set the new carry flag, based on the variable "cf" */
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-        /* OVERFLOW is set *IFF* cnt==1, then it is the 
-           xor of CF and the most significant bit.  Blecck. */
-        /* parenthesized... */
-		if (cnt == 1) {
-			CONDITIONAL_SET_FLAG(XOR2(ocf + ((d >> 6) & 0x2)),
-								 F_OF);
-		}
 	}
 	return (uint8_t)res;
 }
@@ -648,6 +644,8 @@ uint16_t rcr16(x86emu_t *emu, uint16_t d, uint8_t s)
 		if (cnt == 1) {
 			cf = d & 0x1;
 			ocf = ACCESS_FLAG(F_CF) != 0;
+			CONDITIONAL_SET_FLAG((ocf ^ (d >> 15)) & 0x1,
+								 F_OF);
 		} else
 			cf = (d >> (cnt - 1)) & 0x1;
 		mask = (1 << (16 - cnt)) - 1;
@@ -657,10 +655,6 @@ uint16_t rcr16(x86emu_t *emu, uint16_t d, uint8_t s)
 			res |= 1 << (16 - cnt);
 		}
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		if (cnt == 1) {
-			CONDITIONAL_SET_FLAG(XOR2(ocf + ((d >> 14) & 0x2)),
-								 F_OF);
-		}
 	}
 	return (uint16_t)res;
 }
@@ -682,6 +676,8 @@ uint32_t rcr32(x86emu_t *emu, uint32_t d, uint8_t s)
 		if (cnt == 1) {
 			cf = d & 0x1;
 			ocf = ACCESS_FLAG(F_CF) != 0;
+			CONDITIONAL_SET_FLAG((ocf ^ (d >> 31)) & 0x1,
+								 F_OF);
 		} else
 			cf = (d >> (cnt - 1)) & 0x1;
 		mask = (1 << (32 - cnt)) - 1;
@@ -692,10 +688,6 @@ uint32_t rcr32(x86emu_t *emu, uint32_t d, uint8_t s)
 			res |= 1 << (32 - cnt);
 		}
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		if (cnt == 1) {
-			CONDITIONAL_SET_FLAG(XOR2(ocf + ((d >> 30) & 0x2)),
-								 F_OF);
-		}
 	}
 	return res;
 }
@@ -885,8 +877,7 @@ uint16_t shld16 (x86emu_t *emu, uint16_t d, uint16_t fill, uint8_t s)
 			res = d;
 		}
 		if (cnt == 1) {
-			CONDITIONAL_SET_FLAG((((res & 0x8000) == 0x8000) ^
-								  (ACCESS_FLAG(F_CF) != 0)), F_OF);
+			CONDITIONAL_SET_FLAG(((res ^ d) >> 15)&1, F_OF);
 		} else {
 			CLEAR_FLAG(F_OF);
 		}
@@ -927,8 +918,7 @@ uint32_t shld32 (x86emu_t *emu, uint32_t d, uint32_t fill, uint8_t s)
 		res = d;
 	}
 	if (cnt == 1) {
-		CONDITIONAL_SET_FLAG((((res & 0x80000000) == 0x80000000) ^
-								(ACCESS_FLAG(F_CF) != 0)), F_OF);
+		CONDITIONAL_SET_FLAG(((res ^ d) >> 31)&1, F_OF);
 	} else {
 		CLEAR_FLAG(F_OF);
 	}
@@ -959,7 +949,7 @@ uint16_t shrd16 (x86emu_t *emu, uint16_t d, uint16_t fill, uint8_t s)
 		}

 		if (cnt == 1) {
-			CONDITIONAL_SET_FLAG(XOR2(res >> 14), F_OF);
+			CONDITIONAL_SET_FLAG(((res ^ d) >> 15)&1, F_OF);
        } else {
 			CLEAR_FLAG(F_OF);
        }
@@ -1005,7 +995,7 @@ uint32_t shrd32 (x86emu_t *emu, uint32_t d, uint32_t fill, uint8_t s)
 		res = d;
 	}
 	if (cnt == 1) {
-		CONDITIONAL_SET_FLAG(XOR2(res >> 30), F_OF);
+		CONDITIONAL_SET_FLAG(((res ^ d) >> 31)&1, F_OF);
 	} else {
 		CLEAR_FLAG(F_OF);
 	}
@@ -0,0 +1,475 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test64(x86test_t *test, int seg, uintptr_t addr)
+#else
+uintptr_t Run64(x86emu_t *emu, int seg, uintptr_t addr)
+#endif
+{
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uintptr_t tlsdata = (seg==_FS)?GetFSBaseEmu(emu):GetGSBaseEmu(emu);
+    switch(opcode) {
+        case 0x01:              /* ADD GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = add32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x03:              /* ADD Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = add32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x06:              /* PUSH ES */
+            Push(emu, emu->segs[_ES]);
+            break;
+        case 0x07:             /* POP ES */
+            emu->segs[_ES] = Pop(emu);
+            emu->segs_serial[_ES] = 0;
+            break;
+
+        case 0x0f:
+            #ifdef TEST_INTERPRETER
+            return Test640F(test, tlsdata, addr);
+            #else
+            return Run640F(emu, tlsdata, addr);
+            #endif
+
+        case 0x11:              /* ADC GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = adc32(emu, ED->dword[0], GD.dword[0]);
+            break;
+        
+        case 0x13:              /* ADC Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = adc32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x20:              /* AND Seg:Eb,Gb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            EB->byte[0] = and8(emu, EB->byte[0], GB);
+            break;
+        case 0x21:              /* AND Seg:Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = and32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x23:              /* AND Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = and32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x29:              /* SUB GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = sub32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x2B:              /* SUB Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = sub32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x31:              /* XOR Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = xor32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x33:              /* XOR Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = xor32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x39:              /* CMP Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            cmp32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x3B:              /* CMP Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            cmp32(emu, GD.dword[0], ED->dword[0]);
+            break;
+        
+        case 0x40:
+        case 0x41:
+        case 0x42:
+        case 0x43:
+        case 0x44:
+        case 0x45:
+        case 0x46:
+        case 0x47:             /* INC Reg */
+            tmp8u = opcode&7;
+            emu->regs[tmp8u].dword[0] = inc32(emu, emu->regs[tmp8u].dword[0]);
+            break;
+        case 0x48:
+        case 0x49:
+        case 0x4A:
+        case 0x4B:
+        case 0x4C:
+        case 0x4D:
+        case 0x4E:
+        case 0x4F:            /* DEC Reg */
+            tmp8u = opcode&7;
+            emu->regs[tmp8u].dword[0] = dec32(emu, emu->regs[tmp8u].dword[0]);
+            break;
+        case 0x50:              /* PUSH Reg */
+        case 0x51:
+        case 0x52:
+        case 0x53:
+        case 0x54:               /* PUSH ESP */
+        case 0x55:
+        case 0x56:
+        case 0x57:              
+        case 0x58:              /* POP Reg */
+        case 0x59:
+        case 0x5A:
+        case 0x5B:
+        case 0x5C:              /* POP ESP */
+        case 0x5D:
+        case 0x5E:
+        case 0x5F:
+            // Segment override if for memory loc, no stack segment 
+            return addr-1;
+
+        case 0x64:              /* FS: */
+            #ifdef TEST_INTERPRETER
+            return Test64(test, _FS, addr-1); // put FS back
+            #else
+            return Run64(emu, _FS, addr-1); // put FS back
+            #endif
+        case 0x65:                      /* GS: */
+            #ifdef TEST_INTERPRETER
+            return Test64(test, _GS, addr);
+            #else
+            return Run64(emu, _GS, addr);
+            #endif
+            break;
+        case 0x66:
+            #ifdef TEST_INTERPRETER
+            return Test6466(test, tlsdata, addr);
+            #else
+            return Run6466(emu, tlsdata, addr);
+            #endif
+        case 0x67:
+            #ifdef TEST_INTERPRETER
+            return Test6467(test, tlsdata, addr);
+            #else
+            return Run6467(emu, tlsdata, addr);
+            #endif
+
+        case 0x69:              /* IMUL Gd,Ed,Id */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            tmp32u = F32;
+            GD.dword[0] = imul32(emu, ED->dword[0], tmp32u);
+            break;
+        
+        case 0x70:
+        case 0x71:
+        case 0x72:
+        case 0x73:
+        case 0x74:
+        case 0x75:
+        case 0x76:
+        case 0x77:
+        case 0x78:
+        case 0x79:
+        case 0x7A:
+        case 0x7B:
+        case 0x7C:
+        case 0x7D:
+        case 0x7E:
+        case 0x7F:
+            // just ignore the prefix
+            return addr-1;
+        case 0x80:             /* GRP Eb,Ib */
+        case 0x82:             // 0x82 and 0x80 are the same opcodes it seems?
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = F8;
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = add8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] =  or8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = adc8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = sbb8(emu, EB->byte[0], tmp8u); break;
+                case 4: EB->byte[0] = and8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = sub8(emu, EB->byte[0], tmp8u); break;
+                case 6: EB->byte[0] = xor8(emu, EB->byte[0], tmp8u); break;
+                case 7:               cmp8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+        case 0x81:              /* GRP Ed,Id */
+        case 0x83:              /* GRP Ed,Ib */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            if(opcode==0x83) {
+                tmp32s = F8S;
+                tmp32u = (uint32_t)tmp32s;
+            } else
+                tmp32u = F32;
+            switch((nextop>>3)&7) {
+                case 0: ED->dword[0] = add32(emu, ED->dword[0], tmp32u); break;
+                case 1: ED->dword[0] =  or32(emu, ED->dword[0], tmp32u); break;
+                case 2: ED->dword[0] = adc32(emu, ED->dword[0], tmp32u); break;
+                case 3: ED->dword[0] = sbb32(emu, ED->dword[0], tmp32u); break;
+                case 4: ED->dword[0] = and32(emu, ED->dword[0], tmp32u); break;
+                case 5: ED->dword[0] = sub32(emu, ED->dword[0], tmp32u); break;
+                case 6: ED->dword[0] = xor32(emu, ED->dword[0], tmp32u); break;
+                case 7:                cmp32(emu, ED->dword[0], tmp32u); break;
+            }
+            break;
+        case 0x87:
+            nextop = F8;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+            GET_ED_OFFS(tlsdata);
+            if(MODREG) {
+                tmp32u = GD.dword[0];
+                GD.dword[0] = ED->dword[0];
+                ED->dword[0] = tmp32u;
+            } else {
+                if(((uintptr_t)ED)&3)
+                {
+                    // not aligned, dont't try to "LOCK"
+                    tmp32u = ED->dword[0];
+                    ED->dword[0] = GD.dword[0];
+                    GD.dword[0] = tmp32u;
+                } else {
+                    // XCHG is supposed to automaticaly LOCK memory bus
+                    GD.dword[0] = arm_lock_xchg(ED, GD.dword[0]);
+                }
+            }
+#else
+            GET_ED_OFFS(tlsdata);
+            if(!MODREG)
+                pthread_mutex_lock(&emu->context->mutex_lock); // XCHG always LOCK (but when accessing memory only)
+            tmp32u = GD.dword[0];
+            GD.dword[0] = ED->dword[0];
+            ED->dword[0] = tmp32u;
+            if(!MODREG)
+                pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+            break;
+        case 0x88:              /* MOV Seg:Eb,Gb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            EB->byte[0] = GB;
+            break;
+        case 0x89:              /* MOV Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = GD.dword[0];
+            break;
+        case 0x8A:              /* MOV Gb,GS:Eb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            GB = EB->byte[0];
+            break;
+        case 0x8B:              /* MOV Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = ED->dword[0];
+            break;
+        case 0x8C:              /* MOV Seg:Ed, Seg */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = emu->segs[((nextop&0x38)>>3)];
+            break;
+        case 0x8E:              /* MOV Seg,Seg:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            emu->segs[((nextop&0x38)>>3)] = ED->word[0];
+            emu->segs_serial[((nextop&0x38)>>3)] = 0;
+            break;
+        case 0x8F:              /* POP Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = Pop(emu);
+            break;
+        case 0x90:              /* NOP */
+            break;
+
+        case 0x9C:              /* PUSHFD */
+            // Segment override if for memory loc, no stack segment 
+            return addr-1;   // so ignore prefix and continue
+
+        case 0xA0:              /* MOV AL,Ob */
+            tmp32s = F32S;
+            R_AL = *(uint8_t*)((tlsdata) + tmp32s);
+            break;
+        case 0xA1:              /* MOV EAX,Ov */
+            tmp32s = F32S;
+            R_EAX = *(uint32_t*)((tlsdata) + tmp32s);
+            break;
+        case 0xA2:              /* MOV Ob,AL */
+            tmp32s = F32S;
+            *(uint8_t*)((tlsdata) + tmp32s) = R_AL;
+            break;
+        case 0xA3:             /* MOV Od,EAX */
+            tmp32s = F32S;
+            *(uint32_t*)((tlsdata) + tmp32s) = R_EAX;
+            break;
+
+        case 0xA8:            /* TEST AL, Ib */
+            test8(emu, R_AL, F8);
+            break;
+
+        case 0xC0:             /* GRP2 Eb,Ib */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = F8/* & 0x1f*/; // masking done in each functions
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = rol8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] = ror8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = rcl8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = rcr8(emu, EB->byte[0], tmp8u); break;
+                case 4:
+                case 6: EB->byte[0] = shl8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = shr8(emu, EB->byte[0], tmp8u); break;
+                case 7: EB->byte[0] = sar8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+
+        case 0xC6:              /* MOV Eb,Ib */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            EB->byte[0] = F8;
+            break;
+        case 0xC7:              /* MOV Ed,Id */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = F32;
+            break;
+
+        case 0xD0:                      /* GRP2 Eb,1 */
+        case 0xD2:                      /* GRP2 Eb,CL */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = (opcode==0xD0)?1:R_CL;
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = rol8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] = ror8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = rcl8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = rcr8(emu, EB->byte[0], tmp8u); break;
+                case 4: 
+                case 6: EB->byte[0] = shl8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = shr8(emu, EB->byte[0], tmp8u); break;
+                case 7: EB->byte[0] = sar8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+        case 0xD1:                      /* GRP2 Ed,1 */
+        case 0xD3:                      /* GRP2 Ed,CL */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            tmp8u = (opcode==0xD1)?1:R_CL;
+            switch((nextop>>3)&7) {
+                case 0: ED->dword[0] = rol32(emu, ED->dword[0], tmp8u); break;
+                case 1: ED->dword[0] = ror32(emu, ED->dword[0], tmp8u); break;
+                case 2: ED->dword[0] = rcl32(emu, ED->dword[0], tmp8u); break;
+                case 3: ED->dword[0] = rcr32(emu, ED->dword[0], tmp8u); break;
+                case 4: 
+                case 6: ED->dword[0] = shl32(emu, ED->dword[0], tmp8u); break;
+                case 5: ED->dword[0] = shr32(emu, ED->dword[0], tmp8u); break;
+                case 7: ED->dword[0] = sar32(emu, ED->dword[0], tmp8u); break;
+            }
+            break;
+
+        case 0xE9:
+        case 0xEB:
+            return addr-1;       // ignore FS: to execute regular opcode
+            break;
+
+        case 0xFF:              /* GRP 5 Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            switch((nextop>>3)&7) {
+                case 0:                 /* INC Ed */
+                    ED->dword[0] = inc32(emu, ED->dword[0]);
+                    break;
+                case 1:                 /* DEC Ed */
+                    ED->dword[0] = dec32(emu, ED->dword[0]);
+                    break;
+                case 2:                 /* CALL NEAR Ed */
+                    R_EIP = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                    Push(emu, addr);
+                    addr = R_EIP;
+                    break;
+                case 3:                 /* CALL FAR Ed */
+                    if(nextop>0xc0) {
+                        printf_log(LOG_NONE, "Illegal Opcode %02X %02X\n", opcode, nextop);
+                        emu->quit=1;
+                        emu->error |= ERR_ILLEGAL;
+                        return 0;
+                    } else {
+                        Push16(emu, R_CS);
+                        Push(emu, addr);
+                        addr = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                        R_CS = (ED+1)->word[0];
+                    }
+                    break;
+                case 4:                 /* JMP NEAR Ed */
+                    addr = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                    break;
+                case 5:                 /* JMP FAR Ed */
+                    if(nextop>0xc0) {
+                        printf_log(LOG_NONE, "Illegal Opcode 0x%02X 0x%02X\n", opcode, nextop);
+                        emu->quit=1;
+                        emu->error |= ERR_ILLEGAL;
+                        return 0;
+                    } else {
+                        addr = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                        R_CS = (ED+1)->word[0];
+                    }
+                    break;
+                case 6:                 /* Push Ed */
+                    Push(emu, ED->dword[0]);
+                    break;
+                default:
+                    printf_log(LOG_NONE, "Illegal Opcode 0x%02X 0x%02X\n", opcode, nextop);
+                    emu->quit=1;
+                    emu->error |= ERR_ILLEGAL;
+                    return 0;
+            }
+            break;
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,55 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test640F(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run640F(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0xAF:                      /* IMUL Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = imul32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0xB6:                              /* MOVZX Gd,GS:Eb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            GD.dword[0] = EB->byte[0];
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,55 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6466(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run6466(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0x03:                              /* ADD Gw, FS:Ew */
+            nextop = F8;
+            GET_EW_OFFS(tlsdata);
+            GW.word[0] = add16(emu, GW.word[0], EW->word[0]);
+            break;
+
+        case 0x8B:                              /* MOV Gw,FS:Ew */
+            nextop = F8;
+            GET_EW_OFFS(tlsdata);
+            GW.word[0] = EW->word[0];
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,92 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6467(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run6467(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0x3B:                              /* CMP GD, FS:Ed16 */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            cmp32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x89:                              /* MOV ED16,Gd */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            ED->dword[0] = GD.dword[0];
+            break;
+
+        case 0x8B:                              /* MOV Gd,Ed16 */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            GD.dword[0] = ED->dword[0];
+            break;
+
+        case 0x8F:                              /* POP FS:Ed */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            ED->dword[0] = Pop(emu);
+            break;
+
+        case 0xA1:                              /* MOV EAX,Ov16 */
+            tmp32u = F16;
+            R_EAX = *(uint32_t*)(tlsdata + tmp32u);
+            break;
+        case 0xA3:                              /* MOV Ov16,EAX */
+            tmp32u = F16;
+            #ifdef TEST_INTERPRETER
+            test->memaddr = tlsdata + tmp32u;
+            test->memsize = 4;
+            *(uint32_t*)(test->mem) = R_EAX;
+            #else
+            *(uint32_t*)(tlsdata + tmp32u) = R_EAX;
+            #endif
+            break;
+        case 0xFF:                              /* GRP 5 Ed */
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 6:                         /* Push Ed */
+                    GET_EW16_OFFS(tlsdata);
+                    Push(emu, ED->dword[0]);
+                    break;
+                default:
+                    return 0;
+            }
+            break;
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -14,19 +14,36 @@
 #include "x86trace.h"
 #include "box86context.h"

-#define F8      *(uint8_t*)(ip++)
-#define F8S     *(int8_t*)(ip++)
-#define F16     *(uint16_t*)(ip+=2, ip-2)
-#define F16S    *(int16_t*)(ip+=2, ip-2)
-#define F32     *(uint32_t*)(ip+=4, ip-4)
-#define F32S    *(int32_t*)(ip+=4, ip-4)
-#define PK(a)   *(uint8_t*)(ip+a)
-
 #include "modrm.h"
+#include "x86compstrings.h"

-void Run660F(x86emu_t *emu)
+static uint8_t ff_mult(uint8_t a, uint8_t b)
+{
+	int retval = 0;
+
+	for(int i = 0; i < 8; i++) {
+		if((b & 1) == 1)
+			retval ^= a;
+
+		if((a & 0x80)) {
+			a <<= 1;
+			a  ^= 0x1b;
+		} else {
+			a <<= 1;
+		}
+
+		b >>= 1;
+	}
+
+	return retval;
+}
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test660F(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run660F(x86emu_t *emu, uintptr_t addr)
+#endif
 {
-    uintptr_t ip = R_EIP+2; //skip 66 0F
    uint8_t opcode;
    uint8_t nextop;
    reg32_t *oped;
@@ -38,11 +55,63 @@ void Run660F(x86emu_t *emu)
    int32_t tmp32s;
    sse_regs_t *opex, eax1, *opx2;
    mmx87_regs_t *opem;
+    float tmpf;
+    double tmpd;
    #ifndef NOALIGN
    int is_nan;
    #endif
+    // AES opcodes constants
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  F  K  P  E  J  O  D  I  N  C  H  M  B  G  L
+    const uint8_t shiftrows[] = {0, 5,10,15, 4, 9,14, 3, 8,13, 2, 7,12, 1, 6,11};
+    const uint8_t subbytes[256] = {
+        0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
+        0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
+        0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
+        0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
+        0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
+        0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
+        0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
+        0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
+        0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
+        0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
+        0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
+        0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
+        0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
+        0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
+        0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
+        0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
+    };
+                            //   A0 B1 C2 D3 E4 F5 G6 H7 I8 J9 Ka Lb Mc Nd Oe Pf
+                            //   A  N  K  H  E  B  O  L  I  F  C  P  M  J  G  D
+    const uint8_t invshiftrows[] = {0,13,10, 7, 4, 1,14,11, 8, 5, 2,15,12, 9, 6, 3};
+    const uint8_t invsubbytes[256] = {
+        0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38, 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
+        0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87, 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
+        0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d, 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
+        0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2, 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
+        0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
+        0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda, 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
+        0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a, 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
+        0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02, 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
+        0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea, 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
+        0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85, 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
+        0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89, 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
+        0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20, 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
+        0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31, 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
+        0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d, 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
+        0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0, 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
+        0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26, 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
+    };


+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    #ifdef TERMUX
+    extern int isinff(float);
+    extern int isnanf(float);
+    #endif
    opcode = F8;
    switch(opcode) {

@@ -144,8 +213,7 @@ void Run660F(x86emu_t *emu)
                EW->word[0] = 0;    // dummy operation. Should be CR0 -> Ew
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x10:                      /* MOVUPD Gx, Ex */
@@ -160,7 +228,7 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x12:                      /* MOVLPD Gx, Eq */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        if((uintptr_t)ED & 7)
            memcpy(&GX.q[0], ED, 8);
        else
@@ -168,7 +236,7 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x13:                      /* MOVLPD Eq, Gx */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        if((uintptr_t)ED & 7)
            memcpy(ED, &GX.q[0], 8);
        else
@@ -187,18 +255,18 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x16:                      /* MOVHPD Gx, Ed */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        GX.q[1] = *(uint64_t*)ED;
        break;
    case 0x17:                      /* MOVHPD Ed, Gx */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        *(uint64_t*)ED = GX.q[1];
        break;

    case 0x1F:                      /* NOP (multi-byte) */
        nextop = F8;
-        GET_ED;
+        GET_ED_;
        break;
    
    case 0x28:                      /* MOVAPD Gx, Ex */
@@ -219,16 +287,20 @@ void Run660F(x86emu_t *emu)
        GX.d[0] = EM->sd[0];
        GX.d[1] = EM->sd[1];
        break;
-
-
+    case 0x2B:                      /* MOVNTPD Ex, Gx */
+        nextop = F8;
+        GET_EX;
+        EX->q[0] = GX.q[0]; // Address needs to ba aligned for MOVNTPD
+        EX->q[1] = GX.q[1];
+        break;
    case 0x2C:                      /* CVTTPD2PI Gm, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(!isfinite(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GM.ud[0] = 0x80000000;
        else
            GM.sd[0] = EX->d[0];
-        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>0x7fffffff)
+        if(!isfinite(EX->d[1]) || EX->d[1]>(double)0x7fffffff || EX->d[1]<-(double)0x80000000U)
            GM.ud[1] = 0x80000000;
        else
            GM.sd[1] = EX->d[1];
@@ -237,7 +309,7 @@ void Run660F(x86emu_t *emu)
        nextop = F8;
        GET_EX;
        for(int i=0; i<2; ++i)
-            if(isnan(EX->d[i]) || isinf(EX->d[i]) || EX->d[i]>0x7fffffff)
+            if(!isfinite(EX->d[i]) || EX->d[i]>(double)0x7fffffff || EX->d[i]<-(double)0x80000000U)
                GM.ud[i] = 0x80000000;
            else
                switch(emu->mxcsr.f.MXCSR_RC) {
@@ -347,7 +419,34 @@ void Run660F(x86emu_t *emu)
                        GX.sw[4+i] = EX->sw[i*2+0] - EX->sw[i*2+1];
                }
                break;
-                
+            case 0x06:  /* PHSUBD Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<2; ++i)
+                    GX.sd[i] = GX.sd[i*2+0] - GX.sd[i*2+1];
+                if(&GX == EX) {
+                    GX.q[1] = GX.q[0];
+                } else {
+                    for (int i=0; i<2; ++i)
+                        GX.sd[2+i] = EX->sd[i*2+0] - EX->sd[i*2+1];
+                }
+                break;
+            case 0x07:  /* PHSUBSW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<4; ++i) {
+                    tmp32s = GX.sw[i*2+0]-GX.sw[i*2+1];
+                    GX.sw[i] = (tmp32s<-32768)?-32768:((tmp32s>32767)?32767:tmp32s);
+                }
+                if(&GX == EX) {
+                    GX.q[1] = GX.q[0];
+                } else {
+                    for (int i=0; i<4; ++i) {
+                        tmp32s = EX->sw[i*2+0] - EX->sw[i*2+1];
+                        GX.sw[4+i] = (tmp32s<-32768)?-32768:((tmp32s>32767)?32767:tmp32s);
+                    }
+                }
+                break;
            case 0x08:  /* PSIGNB Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -375,6 +474,15 @@ void Run660F(x86emu_t *emu)
                }
            break;

+            case 0x10:  /* PBLENDVB Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<16; ++i) {
+                    if(emu->xmm[0].ub[i]&0x80)
+                        GX.ub[i] = EX->ub[i];
+                }
+                break;
+
            case 0x14:  /* BLENDVPS Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -383,6 +491,14 @@ void Run660F(x86emu_t *emu)
                        GX.ud[i] = EX->ud[i];
                }
                break;
+            case 0x15:  /* BLENDVPD Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<2; ++i) {
+                    if(emu->xmm[0].q[i]&0x8000000000000000LL)
+                        GX.q[i] = EX->q[i];
+                }
+                break;

            case 0x17:      // PTEST GX, EX
                nextop = F8;
@@ -390,6 +506,10 @@ void Run660F(x86emu_t *emu)
                CHECK_FLAGS(emu);
                CONDITIONAL_SET_FLAG(!((GX.q[0]&EX->q[0])|(GX.q[1]&EX->q[1])), F_ZF);
                CONDITIONAL_SET_FLAG(!(((~GX.q[0])&EX->q[0])|((~GX.q[1])&EX->q[1])), F_CF);
+                CLEAR_FLAG(F_AF);
+                CLEAR_FLAG(F_OF);
+                CLEAR_FLAG(F_SF);
+                CLEAR_FLAG(F_PF);
                break;

            case 0x1C:  /* PABSB Gx, Ex */
@@ -451,6 +571,36 @@ void Run660F(x86emu_t *emu)
                    GX.sq[i] = EX->sd[i];
                break;

+            case 0x28:  /* PMULDQ Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                GX.sq[1] = ((int64_t)GX.sd[2])*(int64_t)EX->sd[2];
+                GX.sq[0] = ((int64_t)GX.sd[0])*(int64_t)EX->sd[0];
+                break;
+            case 0x29:  /* PCMPEQQ Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=1; i>=0; --i)
+                    GX.sq[i] = (GX.sq[i]==EX->sq[i])?-1LL:0LL;
+                break;
+            case 0x2A:  /* MOVNTDQA Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                GX.q[0] = EX->q[0];
+                GX.q[1] = EX->q[1];
+                break;
+            case 0x2B:  /* PACKUSDW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<4; ++i)
+                    GX.uw[i] = (GX.sd[i]<0)?0:((GX.sd[i]>65535)?65535:GX.sd[i]);
+                if(&GX==EX)
+                    GX.q[1] = GX.q[0];
+                else
+                    for(int i=0; i<4; ++i)
+                        GX.uw[i+4] = (EX->sd[i]<0)?0:((EX->sd[i]>65535)?65535:EX->sd[i]);
+                break;
+
            case 0x30: /* PMOVZXBW Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -488,6 +638,19 @@ void Run660F(x86emu_t *emu)
                    GX.q[i] = EX->ud[i];
                break;

+            case 0x37: /* PCMPGTQ Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=1; i>=0; --i)
+                    GX.sq[i] = (GX.sq[i]>EX->sq[i])?-1LL:0LL;
+                break;
+            case 0x38:  /* PMINSB Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<16; ++i)
+                    if(GX.sb[i]>EX->sb[i])
+                        GX.sb[i] = EX->sb[i];
+                break;
            case 0x39:  /* PMINSD Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -495,7 +658,27 @@ void Run660F(x86emu_t *emu)
                    if(GX.sd[i]>EX->sd[i])
                        GX.sd[i] = EX->sd[i];
                break;
-
+            case 0x3A:  /* PMINUW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<8; ++i)
+                    if(GX.uw[i]>EX->uw[i])
+                        GX.uw[i] = EX->uw[i];
+                break;
+            case 0x3B:  /* PMINUD Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<4; ++i)
+                    if(GX.ud[i]>EX->ud[i])
+                        GX.ud[i] = EX->ud[i];
+                break;
+            case 0x3C:  /* PMAXSB Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<16; ++i)
+                    if(GX.sb[i]<EX->sb[i])
+                        GX.sb[i] = EX->sb[i];
+                break;
            case 0x3D:  /* PMAXSD Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -503,7 +686,20 @@ void Run660F(x86emu_t *emu)
                    if(GX.sd[i]<EX->sd[i])
                        GX.sd[i] = EX->sd[i];
                break;
-
+            case 0x3E:  /* PMAXUW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<8; ++i)
+                    if(GX.uw[i]<EX->uw[i])
+                        GX.uw[i] = EX->uw[i];
+                break;
+            case 0x3F:  /* PMAXUD Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for(int i=0; i<4; ++i)
+                    if(GX.ud[i]<EX->ud[i])
+                        GX.ud[i] = EX->ud[i];
+                break;
            case 0x40:  /* PMULLD Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -511,14 +707,136 @@ void Run660F(x86emu_t *emu)
                    GX.sd[i] *= EX->sd[i];
                }
                break;
+            case 0x41:  /* PHMINPOSUW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                tmp16u = EX->uw[0];
+                tmp16s = 0;
+                for(int i=1; i<8; ++i) {
+                    if(EX->uw[i]<tmp16u) {
+                        tmp16u = EX->uw[i];
+                        tmp16s = i;
+                    }
+                }
+                GX.q[1] = 0;
+                GX.uw[0] = tmp16u;
+                GX.uw[1] = tmp16s;
+                GX.ud[1] = 0;
+                break;

+            case 0xDB:  /* AESIMC Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                //STATE ← InvMixColumns( STATE );
+                if (EX == &GX) {
+                    for(int i=0; i<16; ++i)
+                        eax1.ub[i] = EX->ub[i];
+                    for(int j=0; j<4; ++j) {
+                        GX.ub[0+j*4] = ff_mult(0x0E, eax1.ub[0+j*4]) ^ ff_mult(0x0B, eax1.ub[1+j*4]) ^ ff_mult(0x0D, eax1.ub[2+j*4]) ^ ff_mult(0x09, eax1.ub[3+j*4]);
+                        GX.ub[1+j*4] = ff_mult(0x09, eax1.ub[0+j*4]) ^ ff_mult(0x0E, eax1.ub[1+j*4]) ^ ff_mult(0x0B, eax1.ub[2+j*4]) ^ ff_mult(0x0D, eax1.ub[3+j*4]);
+                        GX.ub[2+j*4] = ff_mult(0x0D, eax1.ub[0+j*4]) ^ ff_mult(0x09, eax1.ub[1+j*4]) ^ ff_mult(0x0E, eax1.ub[2+j*4]) ^ ff_mult(0x0B, eax1.ub[3+j*4]);
+                        GX.ub[3+j*4] = ff_mult(0x0B, eax1.ub[0+j*4]) ^ ff_mult(0x0D, eax1.ub[1+j*4]) ^ ff_mult(0x09, eax1.ub[2+j*4]) ^ ff_mult(0x0E, eax1.ub[3+j*4]);
+                    }
+                } else {
+                    for(int j=0; j<4; ++j) {
+                        GX.ub[0+j*4] = ff_mult(0x0E, EX->ub[0+j*4]) ^ ff_mult(0x0B, EX->ub[1+j*4]) ^ ff_mult(0x0D, EX->ub[2+j*4]) ^ ff_mult(0x09, EX->ub[3+j*4]);
+                        GX.ub[1+j*4] = ff_mult(0x09, EX->ub[0+j*4]) ^ ff_mult(0x0E, EX->ub[1+j*4]) ^ ff_mult(0x0B, EX->ub[2+j*4]) ^ ff_mult(0x0D, EX->ub[3+j*4]);
+                        GX.ub[2+j*4] = ff_mult(0x0D, EX->ub[0+j*4]) ^ ff_mult(0x09, EX->ub[1+j*4]) ^ ff_mult(0x0E, EX->ub[2+j*4]) ^ ff_mult(0x0B, EX->ub[3+j*4]);
+                        GX.ub[3+j*4] = ff_mult(0x0B, EX->ub[0+j*4]) ^ ff_mult(0x0D, EX->ub[1+j*4]) ^ ff_mult(0x09, EX->ub[2+j*4]) ^ ff_mult(0x0E, EX->ub[3+j*4]);
+                    }
+                }
+                break;
+            case 0xDC:  /* AESENC Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                //STATE ← SRC1;
+                //RoundKey ← SRC2;
+                //STATE ← ShiftRows( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = GX.ub[shiftrows[i]];
+                //STATE ← SubBytes( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = subbytes[eax1.ub[i]];
+                //STATE ← MixColumns( STATE );
+                for(int j=0; j<4; ++j) {
+                    GX.ub[0+j*4] = ff_mult(0x02, eax1.ub[0+j*4]) ^ ff_mult(0x03, eax1.ub[1+j*4]) ^               eax1.ub[2+j*4]  ^               eax1.ub[3+j*4] ;
+                    GX.ub[1+j*4] =               eax1.ub[0+j*4]  ^ ff_mult(0x02, eax1.ub[1+j*4]) ^ ff_mult(0x03, eax1.ub[2+j*4]) ^               eax1.ub[3+j*4] ;
+                    GX.ub[2+j*4] =               eax1.ub[0+j*4]  ^               eax1.ub[1+j*4]  ^ ff_mult(0x02, eax1.ub[2+j*4]) ^ ff_mult(0x03, eax1.ub[3+j*4]);
+                    GX.ub[3+j*4] = ff_mult(0x03, eax1.ub[0+j*4]) ^               eax1.ub[1+j*4]  ^               eax1.ub[2+j*4]  ^ ff_mult(0x02, eax1.ub[3+j*4]);
+                }
+                //DEST[127:0] ← STATE XOR RoundKey;
+                GX.q[0] ^= EX->q[0];
+                GX.q[1] ^= EX->q[1];
+                break;
+            case 0xDD:  /* AESENCLAST Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                //STATE ← SRC1;
+                //RoundKey ← SRC2;
+                //STATE ← ShiftRows( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = GX.ub[shiftrows[i]];
+                //STATE ← SubBytes( STATE );
+                for(int i=0; i<16; ++i)
+                    GX.ub[i] = subbytes[eax1.ub[i]];
+                //DEST[127:0] ← STATE XOR RoundKey;
+                GX.q[0] ^= EX->q[0];
+                GX.q[1] ^= EX->q[1];
+                break;
+            case 0xDE:  /* AESDEC Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                //STATE ← SRC1;
+                //RoundKey ← SRC2;
+                //STATE ← InvShiftRows( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = GX.ub[invshiftrows[i]];
+                //STATE ← InvSubBytes( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = invsubbytes[eax1.ub[i]];
+                //STATE ← InvMixColumns( STATE );
+                for(int j=0; j<4; ++j) {
+                    GX.ub[0+j*4] = ff_mult(0x0E, eax1.ub[0+j*4]) ^ ff_mult(0x0B, eax1.ub[1+j*4]) ^ ff_mult(0x0D, eax1.ub[2+j*4]) ^ ff_mult(0x09, eax1.ub[3+j*4]);
+                    GX.ub[1+j*4] = ff_mult(0x09, eax1.ub[0+j*4]) ^ ff_mult(0x0E, eax1.ub[1+j*4]) ^ ff_mult(0x0B, eax1.ub[2+j*4]) ^ ff_mult(0x0D, eax1.ub[3+j*4]);
+                    GX.ub[2+j*4] = ff_mult(0x0D, eax1.ub[0+j*4]) ^ ff_mult(0x09, eax1.ub[1+j*4]) ^ ff_mult(0x0E, eax1.ub[2+j*4]) ^ ff_mult(0x0B, eax1.ub[3+j*4]);
+                    GX.ub[3+j*4] = ff_mult(0x0B, eax1.ub[0+j*4]) ^ ff_mult(0x0D, eax1.ub[1+j*4]) ^ ff_mult(0x09, eax1.ub[2+j*4]) ^ ff_mult(0x0E, eax1.ub[3+j*4]);
+                }
+                //DEST[127:0] ← STATE XOR RoundKey;
+                GX.q[0] ^= EX->q[0];
+                GX.q[1] ^= EX->q[1];
+                break;
+            case 0xDF:  /* AESDECLAST Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                //STATE ← SRC1;
+                //RoundKey ← SRC2;
+                //STATE ← InvShiftRows( STATE );
+                for(int i=0; i<16; ++i)
+                    eax1.ub[i] = GX.ub[invshiftrows[i]];
+                //STATE ← InvSubBytes( STATE );
+                for(int i=0; i<16; ++i)
+                    GX.ub[i] = invsubbytes[eax1.ub[i]];
+                //DEST[127:0] ← STATE XOR RoundKey;
+                GX.q[0] ^= EX->q[0];
+                GX.q[1] ^= EX->q[1];
+                break;
+                
+            case 0xF0: /* MOVBE Gw, Ew */
+                nextop = F8;
+                GET_ED;
+                GD.word[0] = __builtin_bswap16(ED->word[0]);
+                break;
+            case 0xF1: /* MOVBE Ew, Gw */
+                nextop = F8;
+                GET_ED;
+                ED->word[0] = __builtin_bswap16(GD.word[0]);
+                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;

-    case 0x3A:  // these are some SSE3 opcodes
+    case 0x3A:  // these are some SSE3 & SSE4.x opcodes
        opcode = F8;
        switch(opcode) {
            case 0x08:          // roundps GX, EX, u8
@@ -528,28 +846,28 @@ void Run660F(x86emu_t *emu)
                switch((tmp8u & 4) ? (emu->mxcsr.f.MXCSR_RC) : (tmp8u & 3))
                {
                    case ROUND_Nearest:
-                        GX.f[0] = nearbyint(EX->f[0]);
-                        GX.f[1] = nearbyint(EX->f[1]);
-                        GX.f[2] = nearbyint(EX->f[2]);
-                        GX.f[3] = nearbyint(EX->f[3]);
+                        GX.f[0] = nearbyintf(EX->f[0]);
+                        GX.f[1] = nearbyintf(EX->f[1]);
+                        GX.f[2] = nearbyintf(EX->f[2]);
+                        GX.f[3] = nearbyintf(EX->f[3]);
                        break;
                    case ROUND_Down:
-                        GX.f[0] = floor(EX->f[0]);
-                        GX.f[1] = floor(EX->f[1]);
-                        GX.f[2] = floor(EX->f[2]);
-                        GX.f[3] = floor(EX->f[3]);
+                        GX.f[0] = floorf(EX->f[0]);
+                        GX.f[1] = floorf(EX->f[1]);
+                        GX.f[2] = floorf(EX->f[2]);
+                        GX.f[3] = floorf(EX->f[3]);
                        break;
                    case ROUND_Up:
-                        GX.f[0] = ceil(EX->f[0]);
-                        GX.f[1] = ceil(EX->f[1]);
-                        GX.f[2] = ceil(EX->f[2]);
-                        GX.f[3] = ceil(EX->f[3]);
+                        GX.f[0] = ceilf(EX->f[0]);
+                        GX.f[1] = ceilf(EX->f[1]);
+                        GX.f[2] = ceilf(EX->f[2]);
+                        GX.f[3] = ceilf(EX->f[3]);
                        break;
                    case ROUND_Chop:
-                        GX.f[0] = trunc(EX->f[0]);
-                        GX.f[1] = trunc(EX->f[1]);
-                        GX.f[2] = trunc(EX->f[2]);
-                        GX.f[3] = trunc(EX->f[3]);
+                        GX.f[0] = truncf(EX->f[0]);
+                        GX.f[1] = truncf(EX->f[1]);
+                        GX.f[2] = truncf(EX->f[2]);
+                        GX.f[3] = truncf(EX->f[3]);
                        break;
                }
                break;
@@ -584,16 +902,16 @@ void Run660F(x86emu_t *emu)
                switch((tmp8u & 4) ? (emu->mxcsr.f.MXCSR_RC) : (tmp8u & 3))
                {
                    case ROUND_Nearest:
-                        GX.f[0] = nearbyint(EX->f[0]);
+                        GX.f[0] = nearbyintf(EX->f[0]);
                        break;
                    case ROUND_Down:
-                        GX.f[0] = floor(EX->f[0]);
+                        GX.f[0] = floorf(EX->f[0]);
                        break;
                    case ROUND_Up:
-                        GX.f[0] = ceil(EX->f[0]);
+                        GX.f[0] = ceilf(EX->f[0]);
                        break;
                    case ROUND_Chop:
-                        GX.f[0] = trunc(EX->f[0]);
+                        GX.f[0] = truncf(EX->f[0]);
                        break;
                }
                break;
@@ -617,7 +935,24 @@ void Run660F(x86emu_t *emu)
                        break;
                }
                break;
-
+            case 0x0C:  /* PBLENDPS Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                for (int i=0; i<4; ++i) {
+                    if(tmp8u&(1<<i))
+                        GX.ud[i] = EX->ud[i];
+                }
+                break;
+            case 0x0D:  /* PBLENDPD Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                for (int i=0; i<2; ++i) {
+                    if(tmp8u&(1<<i))
+                        GX.q[i] = EX->q[i];
+                }
+                break;
            case 0x0E:  /* PBLENDW Gx, Ex, Ib */
                nextop = F8;
                GET_EX;
@@ -634,43 +969,44 @@ void Run660F(x86emu_t *emu)
                if(tmp8u>31)
                    {GX.q[0] = GX.q[1] = 0;}
                else
-                #if 0
-                    if(tmp8u>15) {
-                    tmp8u=(tmp8u-16)*8;
-                    if (tmp8u < 64) {
-                        GX.q[0] = (GX.q[0] >> tmp8u) | (GX.q[1] << (64 - tmp8u));
-                        GX.q[1] = (GX.q[1] >> tmp8u);
-                    } else {
-                        GX.q[0] = GX.q[1] >> (tmp8u - 64);
-                        GX.q[1] = 0;
-                    }                    
-                } else {
-                    tmp8u*=8;
-                    if (tmp8u < 64) {
-                        GX.q[0] = (EX->q[0] >> tmp8u) | (EX->q[1] << (64 - tmp8u));
-                        GX.q[1] = (EX->q[1] >> tmp8u) | (GX.q[0] << (64-tmp8u));
-                    } else {
-                        tmp8u -= 64;
-                        GX.q[0] = (EX->q[1] >> tmp8u) | (GX.q[0] << (64 - tmp8u));
-                        GX.q[1] = (GX.q[0] >> tmp8u) | (GX.q[1] >> (64 - tmp8u));
-                    }                    
-                }
-                #else
                {
                    for (int i=0; i<16; ++i, ++tmp8u)
                        eax1.ub[i] = (tmp8u>15)?((tmp8u>31)?0:GX.ub[tmp8u-16]):EX->ub[tmp8u];
                    GX.q[0] = eax1.q[0];
                    GX.q[1] = eax1.q[1];
                }
-                #endif
                break;

+            case 0x14:      // PEXTRB ED, GX, u8
+                nextop = F8;
+                GET_ED;
+                tmp8u = F8;
+                if(MODREG)
+                    ED->dword[0] = GX.ub[tmp8u&0x0f];
+                else
+                    ED->byte[0] = GX.ub[tmp8u&0x0f];
+                break;
+            case 0x15:      // PEXTRW Ew,Gx,Ib
+                nextop = F8;
+                GET_ED;
+                tmp8u = F8;
+                if(MODREG)
+                    ED->dword[0] = GX.uw[tmp8u&7];  // 16bits extract, 0 extended
+                else
+                    ED->word[0] = GX.uw[tmp8u&7];
+                break;
            case 0x16:      // PEXTRD ED, GX, u8
                nextop = F8;
                GET_ED;
                tmp8u = F8;
                ED->dword[0] = GX.ud[tmp8u&3];
                break;
+            case 0x17:      // EXTRACTPS ED, GX, u8
+                nextop = F8;
+                GET_ED;
+                tmp8u = F8;
+                ED->dword[0] = GX.ud[tmp8u&3];
+                break;

            case 0x20:            // PINSRB GX, Ed, Ib
                nextop = F8;
@@ -682,7 +1018,7 @@ void Run660F(x86emu_t *emu)
                nextop = F8;
                GET_EX;
                tmp8u = F8;
-                if((nextop&0xC0)==0xC0) tmp8s = (tmp8u>>6)&3; else tmp8s = 0;
+                if(MODREG) tmp8s = (tmp8u>>6)&3; else tmp8s = 0;
                GX.ud[(tmp8u>>4)&3] = EX->ud[tmp8s];
                for(int i=0; i<4; ++i)
                    if(tmp8u&(1<<i))
@@ -695,9 +1031,146 @@ void Run660F(x86emu_t *emu)
                GX.ud[tmp8u&0x3] = ED->dword[0];
                break;

+            case 0x40:  /* DPPS Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmpf = 0.0f;
+                for(int i=0; i<4; ++i)
+                    if(tmp8u&(1<<(i+4)))
+                        tmpf += GX.f[i]*EX->f[i];
+                for(int i=0; i<4; ++i)
+                    GX.f[i] = (tmp8u&(1<<i))?tmpf:0.0f;
+                break;
+            case 0x41:  /* DPPD Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmpd = 0.0;
+                if(tmp8u&(1<<(4+0)))
+                    tmpd += GX.d[0]*EX->d[0];
+                if(tmp8u&(1<<(4+1)))
+                    tmpd += GX.d[1]*EX->d[1];
+                GX.d[0] = (tmp8u&(1<<(0)))?tmpd:0.0;
+                GX.d[1] = (tmp8u&(1<<(1)))?tmpd:0.0;
+                break;
+            case 0x42:  /* MPSADBW Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                {
+                    int src = tmp8u&3;
+                    int dst = (tmp8u>>2)&1;
+                    int b[11];
+                    for (int i=0; i<11; ++i)
+                        b[i] = GX.ub[dst*4+i];
+                    for(int i=0; i<8; ++i) {
+                        int tmp = abs(b[i+0]-EX->ub[src*4+0]);
+                        tmp += abs(b[i+1]-EX->ub[src*4+1]);
+                        tmp += abs(b[i+2]-EX->ub[src*4+2]);
+                        tmp += abs(b[i+3]-EX->ub[src*4+3]);
+                        GX.uw[i] = tmp;
+                    }
+                }
+                break;
+                
+            case 0x44:  /* PCLMULQDQ Gx, Ex, Ib */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                {
+                    int g = (tmp8u&1)?1:0;
+                    int e = (tmp8u&0b10000)?1:0;
+                    uint64_t result[2] = { 0 };
+                    for (int i=0; i<64; ++i)
+                        if(GX.q[g]&(1LL<<i)) {
+                            result[0] ^= (EX->q[e]<<i);
+                            if(i)
+                                result[1] ^= (EX->q[e]>>(64-i));
+                        }
+
+                    GX.q[0] = result[0];
+                    GX.q[1] = result[1];
+                }
+                break;
+
+            case 0x60:  /* PCMPESTRM */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmp32u = sse42_compare_string_explicit_len(emu, EX, R_EDX, &GX, R_EAX, tmp8u);
+                if(tmp8u&0b1000000) {
+                    switch(tmp8u&1) {
+                        case 0: for(int i=0; i<16; ++i) GX.ub[i] = ((tmp32u>>i)&1)?0xff:0x00; break;
+                        case 1: for(int i=0; i<8; ++i) GX.uw[i] = ((tmp32u>>i)&1)?0xffff:0x0000; break;
+                    }
+                } else {
+                    GX.q[1] = GX.q[0] = 0;
+                    GX.uw[0] = tmp32u;
+                }
+                break;
+            case 0x61:  /* PCMPESTRI */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmp32u = sse42_compare_string_explicit_len(emu, EX, R_EDX, &GX, R_EAX, tmp8u);
+                if(!tmp32u)
+                    R_ECX = (tmp8u&1)?8:16;
+                else if(tmp8u&0b1000000)
+                    R_ECX = 31-__builtin_clz(tmp32u);
+                else
+                    R_ECX = __builtin_ffs(tmp32u) - 1;
+                break;
+            case 0x62:  /* PCMPISTRM */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmp32u = sse42_compare_string_implicit_len(emu, EX, &GX, tmp8u);
+                if(tmp8u&0b1000000) {
+                    switch(tmp8u&1) {
+                        case 0: for(int i=0; i<16; ++i) GX.ub[i] = ((tmp32u>>i)&1)?0xff:0x00; break;
+                        case 1: for(int i=0; i<8; ++i) GX.uw[i] = ((tmp32u>>i)&1)?0xffff:0x0000; break;
+                    }
+                } else {
+                    GX.q[1] = GX.q[0] = 0;
+                    GX.uw[0] = tmp32u;
+                }
+                break;
+            case 0x63:  /* PCMPISTRI */
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                tmp32u = sse42_compare_string_implicit_len(emu, EX, &GX, tmp8u);
+                if(!tmp32u)
+                    R_ECX = (tmp8u&1)?8:16;
+                else if(tmp8u&0b1000000)
+                    R_ECX = 31-__builtin_clz(tmp32u);
+                else
+                    R_ECX = __builtin_ffs(tmp32u) - 1;
+                break;
+                
+            case 0xDF:      // AESKEYGENASSIST Gx, Ex, u8
+                nextop = F8;
+                GET_EX;
+                tmp32u = F8;
+                for (int i = 4; i < 8; ++i)
+                    GX.ub[i] = subbytes[EX->ub[i]];
+                for (int i = 12; i < 16; ++i)
+                    GX.ub[i] = subbytes[EX->ub[i]];
+                GX.ud[0] = GX.ud[1];
+                tmp8u = GX.ub[4];
+                GX.ud[1] = GX.ud[1] >> 8;
+                GX.ub[7] = tmp8u;
+                GX.ud[1] ^= tmp32u;
+                GX.ud[2] = GX.ud[3];
+                tmp8u = GX.ub[12];
+                GX.ud[3] = GX.ud[3] >> 8;
+                GX.ub[15] = tmp8u;
+                GX.ud[3] ^= tmp32u;
+                break;
+
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;

@@ -782,23 +1255,24 @@ void Run660F(x86emu_t *emu)
        nextop = F8;
        GET_EX;
        for(int i=0; i<4; ++i)
-            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>0x7fffffff)
+            // note, converting 0x7fffffff to float gives 0x80000000 basicaly, because of the loss or precision in the mantice. So uing >= here...
+            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>=(float)0x80000000U || EX->f[i]<-(float)0x80000000U)
                GX.sd[i] = 0x80000000;
-        else
-            switch(emu->mxcsr.f.MXCSR_RC) {
-                case ROUND_Nearest:
-                    GX.sd[i] = nearbyintf(EX->f[i]);
-                    break;
-                case ROUND_Down:
-                    GX.sd[i] = floorf(EX->f[i]);
-                    break;
-                case ROUND_Up:
-                    GX.sd[i] = ceilf(EX->f[i]);
-                    break;
-                case ROUND_Chop:
-                    GX.sd[i] = EX->f[i];
-                    break;
-            }
+            else
+                switch(emu->mxcsr.f.MXCSR_RC) {
+                    case ROUND_Nearest:
+                        GX.sd[i] = nearbyintf(EX->f[i]);
+                        break;
+                    case ROUND_Down:
+                        GX.sd[i] = floorf(EX->f[i]);
+                        break;
+                    case ROUND_Up:
+                        GX.sd[i] = ceilf(EX->f[i]);
+                        break;
+                    case ROUND_Chop:
+                        GX.sd[i] = EX->f[i];
+                        break;
+                }
        break;
    case 0x5C:                      /* SUBPD Gx, Ex */
        nextop = F8;
@@ -1016,8 +1490,7 @@ void Run660F(x86emu_t *emu)
                    for (int i=0; i<8; ++i) EX->uw[i] <<= tmp8u;
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x72:  /* GRP */
@@ -1043,8 +1516,7 @@ void Run660F(x86emu_t *emu)
                    for (int i=0; i<4; ++i) EX->ud[i] <<= tmp8u;
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x73:  /* GRP */
@@ -1096,8 +1568,7 @@ void Run660F(x86emu_t *emu)
                }
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x74:  /* PCMPEQB Gx,Ex */
@@ -1190,7 +1661,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1217,7 +1693,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1263,7 +1744,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u)) {
            SET_FLAG(F_CF);
@@ -1332,8 +1818,7 @@ void Run660F(x86emu_t *emu)
                break;

            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0xBB:                      /* BTC Ew,Gw */
@@ -1345,7 +1830,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1500,20 +1990,19 @@ void Run660F(x86emu_t *emu)
        nextop = F8;
        GET_EX;
        EX->q[0] = GX.q[0];
-        if((nextop&0xC0)==0xC0)
+        if(MODREG)
            EX->q[1] = 0;
        break;
    case 0xD7:  /* PMOVMSKB Gd,Ex */
        nextop = F8;
-        if((nextop&0xC0)==0xC0) {
+        if(MODREG) {
            GET_EX;
            GD.dword[0] = 0;
            for (int i=0; i<16; ++i)
                if(EX->ub[i]&0x80)
                    GD.dword[0] |= (1<<i);
        } else {
-            ip = R_EIP;
-            UnimpOpcode(emu);
+            return 0;
        }
        break;
    case 0xD8:  /* PSUBUSB Gx,Ex */
@@ -1617,11 +2106,11 @@ void Run660F(x86emu_t *emu)
    case 0xE6:  /* CVTTPD2DQ Gx, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GX.sd[0] = 0x80000000;
        else
            GX.sd[0] = EX->d[0];
-        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>0x7fffffff)
+        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>(double)0x7fffffff || EX->d[1]<-(double)0x80000000U)
            GX.sd[1] = 0x80000000;
        else
            GX.sd[1] = EX->d[1];
@@ -1799,10 +2288,8 @@ void Run660F(x86emu_t *emu)
        break;
 
    default:
-        ip = R_EIP;
-        UnimpOpcode(emu);
+        return 0;
    }

-    R_EIP = ip;
-    return;
+    return addr;
 }
@@ -0,0 +1,103 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test66D9(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run66D9(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+
+    nextop = F8;
+    switch(nextop) {
+        
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+        case 0xD0:
+        case 0xE0:
+        case 0xE5:
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xFC:
+        case 0xE1:
+        case 0xE4:
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            return 0;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 4:     /* FLDENV m */
+                    // warning, incomplete
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_loadenv(emu, (char*)ED, 1);
+                    #endif
+                    break;
+                case 6:     /* FNSTENV m */
+                    // warning, incomplete
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_savenv(emu, (char*)ED, 1);
+                    #endif
+                    break;
+                default:
+                    return 0;
+            }
+    }
+
+    return addr;
+}
@@ -0,0 +1,122 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test66DD(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run66DD(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+
+    nextop = F8;
+    switch(nextop) {
+
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+        case 0xD0:
+        case 0xE0:
+        case 0xE5:
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xFC:
+        case 0xE1:
+        case 0xE4:
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            return 0;
+        break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 4: /* FRSTOR m94byte */
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_loadenv(emu, (char*)ED, 1);
+                    // get the STx
+                    {
+                        char* p =(char*)ED;
+                        p += 14;
+                        for (int i=0; i<8; ++i) {
+                            LD2D(p, &ST(i).d);
+                            p+=10;
+                        }
+                    }
+                    #endif
+                    break;
+                case 6: /* FNSAVE m94byte */
+                    GET_EW;
+                    // ENV first...
+                    #ifndef TEST_INTERPRETER
+                    fpu_savenv(emu, (char*)ED, 1);
+                    // save the STx
+                    {
+                        char* p =(char*)ED;
+                        p += 14;
+                        for (int i=0; i<8; ++i) {
+                            D2LD(&ST(i).d, p);
+                            p+=10;
+                        }
+                    }
+                    #endif
+                    reset_fpu(emu);
+                    break;
+                default:
+                    return 0;
+            }
+    }
+
+    return addr;
+}
@@ -0,0 +1,74 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+#include "x86compstrings.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test66F20F(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run66F20F(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    int8_t tmp8s;
+    uint16_t tmp16u;
+    int16_t tmp16s;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    sse_regs_t *opex, eax1, *opx2;
+    mmx87_regs_t *opem;
+    float tmpf;
+    double tmpd;
+
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    opcode = F8;
+    switch(opcode) {
+
+         case 0x38:  // SSE 4.x
+            opcode = F8;
+            switch(opcode) {
+            
+                case 0xF1:  // CRC32 Gd, Ew
+                    nextop = F8;
+                    GET_EW;
+                    for(int j=0; j<2; ++j) {
+                        GD.dword[0] ^=  EW->byte[j];
+                        for (int i = 0; i < 8; i++) {
+                            if (GD.dword[0] & 1)
+                                GD.dword[0] = (GD.dword[0] >> 1) ^ 0x82f63b78;
+                            else
+                                GD.dword[0] = (GD.dword[0] >> 1);
+                        }
+                    }
+                    break;
+
+                default: 
+                    return 0;
+            }
+        break;
+
+    default:
+        return 0;
+    }
+
+    return addr;
+}
@@ -0,0 +1,147 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "my_cpuid.h"
+#include "bridge.h"
+#include "signals.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test67(x86test_t *test, int rep, uintptr_t addr)
+#else
+uintptr_t Run67(x86emu_t *emu, int rep, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u, tmp8u2;
+    int8_t tmp8s;
+    uint16_t tmp16u, tmp16u2;
+    int16_t tmp16s;
+    uint32_t tmp32u, tmp32u2, tmp32u3;
+    int32_t tmp32s, tmp32s2;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    double d;
+    float f;
+    int64_t ll;
+    sse_regs_t *opex, eax1;
+    mmx87_regs_t *opem, eam1;
+    uintptr_t tlsdata;
+
+#ifdef TEST_INTERPRETER
+    x86emu_t *emu = test->emu;
+#endif
+    opcode = F8;
+
+    while(opcode==0x67)
+        opcode = F8;
+
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+
+        case 0x64:
+            tlsdata = GetFSBaseEmu(emu);
+            #ifdef TEST_INTERPRETER
+            return Test6467(test, tlsdata, addr);
+            #else
+            return Run6467(emu, tlsdata, addr);
+            #endif
+
+        case 0x66:                      /* MoooRE opcodes */
+            #ifdef TEST_INTERPRETER
+            return Test6766(test, rep, addr);
+            #else
+            return Run6766(emu, rep, addr);
+            #endif
+
+        case 0x6C:                      /* INSB */
+            tmp32u = rep?R_CX:1;
+            while(tmp32u--) {
+                #ifdef TEST_INTERPRETER
+                *(int8_t*)(R_DI+GetESBaseEmu(emu)) = 0;         // faking port read, using actual segment ES, just in case
+                #endif
+                if(ACCESS_FLAG(F_DF))
+                    R_DI-=1;
+                else
+                    R_DI+=1;
+            }
+            if(rep) R_CX = 0;
+            break;
+
+        case 0xAC:                      /* LODSB */
+            tmp32u = rep?R_CX:1;
+            while(tmp32u--) {
+                R_AL = *(int8_t*)(R_SI+GetDSBaseEmu(emu));
+                if(ACCESS_FLAG(F_DF))
+                    R_SI-=1;
+                else
+                    R_SI+=1;
+            }
+            if(rep) R_CX = 0;
+            break;
+
+        case 0xE0:                      /* LOOPNZ */
+            CHECK_FLAGS(emu);
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX && !ACCESS_FLAG(F_ZF))
+                addr += tmp8s;
+            break;
+        case 0xE1:                      /* LOOPZ */
+            CHECK_FLAGS(emu);
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX && ACCESS_FLAG(F_ZF))
+                addr += tmp8s;
+            break;
+        case 0xE2:                      /* LOOP */
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX)
+                addr += tmp8s;
+            break;
+        case 0xE3:                      /* JCXZ */
+            tmp8s = F8S;
+            if(!R_CX)
+                addr += tmp8s;
+            break;
+
+        case 0xE8:                      /* CALL Id */
+            tmp32s = F32S; // call is relative
+            Push(emu, addr);
+            addr += tmp32s;
+            break;
+
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,58 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6766(x86test_t *test, int rep, uintptr_t addr)
+#else
+uintptr_t Run6766(x86emu_t *emu, int rep, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+
+    while((opcode==0x2E) || (opcode==0x66))   // ignoring CS: or multiple 0x66
+        opcode = F8;
+
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+
+        case 0x8D:                              /* LEA Gw,Ew */
+            nextop = F8;
+            GET_EW16_;
+            GW.word[0] = (uint16_t)(uintptr_t)ED;
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -24,6 +24,7 @@
 #include "x86trace.h"
 #endif
 #include "x86tls.h"
+#include "x86emu.h"

 #define PARITY(x)   (((emu->x86emu_parity_tab[(x) / 32] >> ((x) % 32)) & 1) == 0)
 #define XOR2(x) 	(((x) ^ ((x)>>1)) & 0x1)
@@ -456,6 +457,8 @@ void UpdateFlags(x86emu_t *emu)
                    CONDITIONAL_SET_FLAG((emu->res & 0xff) == 0, F_ZF);
                    CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
                    CONDITIONAL_SET_FLAG(emu->res & 0x80, F_SF);
+                    if(emu->op2==1)
+                        CLEAR_FLAG(F_OF);
                }
            } else {
                if (emu->op1&0x80) {
@@ -479,6 +482,8 @@ void UpdateFlags(x86emu_t *emu)
                    CONDITIONAL_SET_FLAG((emu->res & 0xffff) == 0, F_ZF);
                    CONDITIONAL_SET_FLAG(emu->res & 0x8000, F_SF);
                    CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+                    if(emu->op2==1)
+                        CLEAR_FLAG(F_OF);
                }
            } else {
                if (emu->op1&0x8000) {
@@ -501,6 +506,8 @@ void UpdateFlags(x86emu_t *emu)
                CONDITIONAL_SET_FLAG(emu->res == 0, F_ZF);
                CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
                CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+                if(emu->op2==1)
+                    CLEAR_FLAG(F_OF);
            }
            break;
        case d_shr8:
@@ -514,7 +521,7 @@ void UpdateFlags(x86emu_t *emu)
                    CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
                }
                if (cnt == 1) {
-                    CONDITIONAL_SET_FLAG(XOR2(emu->res >> 6), F_OF);
+                    CONDITIONAL_SET_FLAG(emu->op1 & 0x80, F_OF);
                }
            } else {
                CONDITIONAL_SET_FLAG((emu->op1 >> (emu->op2-1)) & 0x1, F_CF);
@@ -534,7 +541,7 @@ void UpdateFlags(x86emu_t *emu)
                    CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
                }
                if (cnt == 1) {
-                    CONDITIONAL_SET_FLAG(XOR2(emu->res >> 14), F_OF);
+                    CONDITIONAL_SET_FLAG(emu->op1 & 0x8000, F_OF);
                }
            } else {
                CLEAR_FLAG(F_CF);
@@ -553,7 +560,35 @@ void UpdateFlags(x86emu_t *emu)
                CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
            }
            if (cnt == 1) {
-                CONDITIONAL_SET_FLAG(XOR2(emu->res >> 30), F_OF);
+                CONDITIONAL_SET_FLAG(emu->op1 & 0x80000000, F_OF);
+            }
+            break;
+        case d_shrd32:
+            cnt = emu->op2;
+            if (cnt > 0) {
+                cc = emu->op1 & (1 << (cnt - 1));
+                CONDITIONAL_SET_FLAG(cc, F_CF);
+                CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
+                CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
+                CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+                if (cnt == 1) {
+                    CONDITIONAL_SET_FLAG((emu->op1 ^ emu->res) & 0x80000000, F_OF);
+                }
+            }
+            break;
+        case d_shld32:
+            cnt = emu->op2;
+            if (cnt > 0) {
+                cc = emu->op1 & (1 << (32 - cnt));
+                CONDITIONAL_SET_FLAG(cc, F_CF);
+                CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
+                CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
+                CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+                if (cnt == 1) {
+                    CONDITIONAL_SET_FLAG((emu->op1 ^ emu->res) & 0x80000000, F_OF);
+                } else {
+                    CLEAR_FLAG(F_OF);
+                }
            }
            break;
        case d_sub8:
@@ -882,6 +917,7 @@ void PrintTrace(x86emu_t* emu, uintptr_t ip, int dynarec)
                printFunctionAddr(*(uintptr_t*)(R_ESP), "=> ");
            } else if(peek==0x55) {
                printf_log(LOG_NONE, " => STACK_TOP: %p", *(void**)(R_ESP));
+                printFunctionAddr(*(uintptr_t*)(R_ESP), "top:");
                printFunctionAddr(ip, "here: ");
            } else if(peek==0xE8) { // Call
                uintptr_t nextaddr = ip + 5 + PK32(1);
@@ -890,7 +926,7 @@ void PrintTrace(x86emu_t* emu, uintptr_t ip, int dynarec)
                const uint8_t pk1 = PK(1);
                if(((pk1>>3)&7)==2 || ((pk1>>3)&7)==4) { // jmp/call near
                    uintptr_t nextaddr = (uintptr_t)getAlternate((void*)evalED(emu, ip+1));
-                    printf_log(LOG_NONE, " => %p", nextaddr);
+                    printf_log(LOG_NONE, " => %p", (void*)nextaddr);
                    printFunctionAddr(nextaddr, " / ");
                }
            }
@@ -4,6 +4,7 @@
 #include <stdint.h>
 #include "regs.h"
 #include "x86emu_private.h"
+#include "box86context.h"
 typedef struct x86emu_s x86emu_t;

 static inline uint8_t Fetch8(x86emu_t *emu) {return *(uint8_t*)(R_EIP++);}
@@ -41,15 +42,25 @@ static inline uint32_t Pop(x86emu_t *emu)
    return *st;
 }

+#ifdef TEST_INTERPRETER
+#define Push(E, V)  do{E->regs[_SP].dword[0] -=4; test->memsize = 4; *(uint64_t*)test->mem = (V); test->memaddr = E->regs[_SP].dword[0];}while(0)
+#define Push16(E, V)  do{E->regs[_SP].dword[0] -=2; test->memsize = 2; *(uint16_t*)test->mem = (V); test->memaddr = E->regs[_SP].dword[0];}while(0)
+#else
 static inline void Push(x86emu_t *emu, uint32_t v)
 {
    R_ESP -= 4;
    *((uint32_t*)R_ESP) = v;
 }
+#endif

+static inline void PushExit(x86emu_t* emu)
+{
+    R_ESP -= 4;
+    *((uint32_t*)R_ESP) = my_context->exit_bridge;
+}

+#if 0
 // the op code definition can be found here: http://ref.x86asm.net/geek32.html
-
 static inline reg32_t* GetECommon(x86emu_t* emu, uint32_t m)
 {
    if (m<=7) {
@@ -188,23 +199,71 @@ static inline sse_regs_t* GetGx(x86emu_t *emu, uint32_t v)
    uint8_t m = (v&0x38)>>3;
    return &emu->xmm[m&7];
 }
-
+#endif
 void UpdateFlags(x86emu_t *emu);

 #define CHECK_FLAGS(emu) if(emu->df) UpdateFlags(emu)
 #define RESET_FLAGS(emu) emu->df = d_none

-void Run67(x86emu_t *emu);
-void Run0F(x86emu_t *emu);
-void Run660F(x86emu_t *emu);
-void Run66D9(x86emu_t *emu);    // x87
-void Run6766(x86emu_t *emu);
-void RunGS(x86emu_t *emu);
-void RunGS0F(x86emu_t *emu, uintptr_t tlsdata);
-void RunFS(x86emu_t *emu);
-void RunFS66(x86emu_t *emu, uintptr_t tlsdata);
-void RunLock(x86emu_t *emu);
-void RunLock66(x86emu_t *emu);
+uintptr_t Run0F(x86emu_t *emu, uintptr_t addr, int *step);
+uintptr_t Run64(x86emu_t *emu, int seg, uintptr_t addr);
+uintptr_t Run640F(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Run6466(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Run6467(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Run66(x86emu_t *emu, int rep, uintptr_t addr);
+uintptr_t Run660F(x86emu_t *emu, uintptr_t addr);
+uintptr_t Run66F20F(x86emu_t *emu, uintptr_t addr);
+uintptr_t Run6664(x86emu_t *emu, int seg, uintptr_t addr);
+uintptr_t Run66D9(x86emu_t *emu, uintptr_t addr);
+uintptr_t Run66DD(x86emu_t *emu, uintptr_t addr);
+uintptr_t Run66F0(x86emu_t *emu, uintptr_t addr);
+uintptr_t Run67(x86emu_t *emu, int rep, uintptr_t addr);
+uintptr_t Run670F(x86emu_t *emu, int rep, uintptr_t addr);
+uintptr_t Run6766(x86emu_t *emu, int rep, uintptr_t addr);
+uintptr_t Run67660F(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunD8(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunD9(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDA(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDB(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDC(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDD(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDE(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunDF(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunF0(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunF066(x86emu_t *emu, uintptr_t addr);
+uintptr_t RunF20F(x86emu_t *emu, uintptr_t addr, int *step);
+uintptr_t RunF30F(x86emu_t *emu, uintptr_t addr);
+
+#ifdef DYNAREC
+uintptr_t Test0F(x86test_t *test, uintptr_t addr, int *step);
+uintptr_t Test64(x86test_t *test, int seg, uintptr_t addr);
+uintptr_t Test640F(x86test_t *test, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Test6466(x86test_t *test, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Test6467(x86test_t *test, uintptr_t tlsdata, uintptr_t addr);
+uintptr_t Test66(x86test_t *test, int rep, uintptr_t addr);
+uintptr_t Test660F(x86test_t *test, uintptr_t addr);
+uintptr_t Test66F20F(x86test_t *test, uintptr_t addr);
+uintptr_t Test6664(x86test_t *test, int seg, uintptr_t addr);
+uintptr_t Test66D9(x86test_t *test, uintptr_t addr);
+uintptr_t Test66DD(x86test_t *test, uintptr_t addr);
+uintptr_t Test66F0(x86test_t *test, uintptr_t addr);
+uintptr_t Test67(x86test_t *test, int rep, uintptr_t addr);
+uintptr_t Test670F(x86test_t *test, int rep, uintptr_t addr);
+uintptr_t Test6766(x86test_t *test, int rep, uintptr_t addr);
+uintptr_t Test67660F(x86test_t *test, uintptr_t addr);
+uintptr_t TestD8(x86test_t *test, uintptr_t addr);
+uintptr_t TestD9(x86test_t *test, uintptr_t addr);
+uintptr_t TestDA(x86test_t *test, uintptr_t addr);
+uintptr_t TestDB(x86test_t *test, uintptr_t addr);
+uintptr_t TestDC(x86test_t *test, uintptr_t addr);
+uintptr_t TestDD(x86test_t *test, uintptr_t addr);
+uintptr_t TestDE(x86test_t *test, uintptr_t addr);
+uintptr_t TestDF(x86test_t *test, uintptr_t addr);
+uintptr_t TestF0(x86test_t *test, uintptr_t addr);
+uintptr_t TestF066(x86test_t *test, uintptr_t addr);
+uintptr_t TestF20F(x86test_t *test, uintptr_t addr, int *step);
+uintptr_t TestF30F(x86test_t *test, uintptr_t addr);
+#endif

 void x86Syscall(x86emu_t *emu);
 void x86Int3(x86emu_t* emu);
@@ -1,7 +1,40 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wstrict-aliasing"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestD8(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunD8(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    float f;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
-    switch(nextop) {
+    switch (nextop) {
        case 0xC0:
        case 0xC1:
        case 0xC2:
@@ -159,7 +192,9 @@
                }
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
+   return addr;
+}
 #pragma GCC diagnostic pop