Bump version to v0.1.0

Clean up Gameshell section

.gitignore

@@ -1,3 +1,4 @@
 build
-build_llvm
+build*
 .vscode
+.gdb_history

CHANGELOG.md

@@ -1,3 +1,28 @@
+v0.1.0
+=======
+* Dynarec!!! Only for ARM (note that dynarec doesn't support JITed code for now)
+* Added real support for getcontext/set/context/makecontext/swapcontext
+* Preliminary signal handling
+* Fixes to SDL(1/2) Image and SDL1 Mixer, and to SDL1.2 RWops usage
+* Fixed numerous issues in opcodes (both interpretor and dynarec). FTL works fine now, among many others
+* Added wrapped gtk support (still ongoing, many libs involved)
+* Make loading of libs more configurable
+* If a wrapped native libs is not found, try to use emulated one
+* Add en env. var. to force use of emulated lib for certain libs
+* Add an env. var. to define wich libGL to use
+* Added Install / Uninstall target (using systemd binfmt)
+* Added more hardware target (RPis, GameShell...)
+* Wrapped more libs (including FreeType, smpeg, ncurses, sndfile...)
+
+v0.0.4
+=======
+* Improved the Init of dependant Libs. More things work now.
+* Added a lot of wrapped functions.
+* Added a few wrapped libs (like libz or some other x11 related libs)
+* For trace enabled build, Trace can be enabled only after a certain amount of opcodes (still, a debugger would be better)
+* Some fixes in a few opcodes, and implemented x87 rounding (SuperMeatBoy behaves better now)
+* FTL 1.6.9 still have corrupted music, but older 1.5.13 seems fine (different set of libs)
+
 v0.0.2
 =======
 * A full commercial games runs fine on the Pandora: Airline Tycoon Deluxe

CMakeLists.txt

@@ -3,18 +3,80 @@ cmake_minimum_required(VERSION 2.6)
 project(box86)
 
 enable_testing()
+enable_language(C)
 
 option(PANDORA "Set to ON if targeting an OpenPandora device" ${PANDORA})
+option(PYRA "Set to ON if targeting an OpenPandora device" ${PYRA})
+option(RPI2 "Set to ON if targeting an RaspberryPI2 device" ${RPI2})
+option(RPI3 "Set to ON if targeting an RaspberryPI3 device" ${RPI3})
+option(RPI4 "Set to ON if targeting an RaspberryPI4 device" ${RPI4})
+option(GAMESHELL "Set to ON if targeting a Gameshell device" ${GAMESHELL})
 option(USE_CCACHE "Set to ON to use ccache if present in the system" ${USE_CCACHE})
 option(LD80BITS "Set to ON if host device have 80bits long double (i.e. i386)" ${LD80BITS})
 option(NOALIGN "Set to ON if host device doesn't need re-align (i.e. i386)" ${NOALIGN})
 option(HAVE_TRACE "Set to ON to have Trace ability (needs ZydisInfo library)" ${HAVE_TRACE})
+option(USE_FLOAT "Set to ON to use only float, no double, in all x87 Emulation" ${USE_FLOAT})
+option(NOLOADADDR "Set to ON to avoid fixing the load address of Box86" ${NO_LOADAADR})
+option(ARM_DYNAREC "Set to ON to use ARM Dynamic Recompilation (WIP, don't use yet)" ${ARM_DYNAREC})
+option(NOGETCLOCK "Set to ON to avoid using clock_gettime with CLOCK_MONOTONIC_COARSE for RDTSC opcode (use gltimeofday instead)" ${NOGETCLOCK})
 
+find_package(PythonInterp)
+if(NOT PYTHONINTERP_FOUND)
+    message( FATAL_ERROR "You need a Python interpretor, CMake will exit." )
+endif()
 
 # Pandora
 if(PANDORA)
     add_definitions(-DPANDORA)
     #add_definitions(-mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp -ftree-vectorize -fsingle-precision-constant -ffast-math)
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
+endif()
+
+# Gameshell
+if(GAMESHELL)
+    add_definitions(-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard")
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
+endif()
+
+if(PYRA)
+    add_definitions(-mcpu=cortex-a15 -mfpu=neon -mfloat-abi=hard -marm)
+    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a15 -mfpu=neon -mfloat-abi=hard")
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
+endif()
+
+if(RPI2)
+    add_definitions(-DRPI)
+    add_definitions(-DRPI2)
+    add_definitions(-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard")
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
+endif()
+if(RPI3)
+    add_definitions(-DRPI)
+    add_definitions(-DRPI3)
+    add_definitions(-marm -march=armv8-a+crc -mtune=cortex-a53 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -march=armv8-a+crc -mtune=cortex-a53 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
+endif()
+if(RPI4)
+    add_definitions(-DRPI)
+    add_definitions(-DRPI3)
+    add_definitions(-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard)
+    set(CMAKE_ASM_FLAGS  "-marm -march=armv8-a+crc -mtune=cortex-a72 -mfpu=neon-fp-armv8 -mfloat-abi=hard")
+    set(ARM_DYNAREC ON)
+    set(LD80BITS OFF)
+    set(NOALIGN OFF)
 endif()
 
 if(LD80BITS)
@@ -29,7 +91,33 @@ if(HAVE_TRACE)
     add_definitions(-DHAVE_TRACE)
 endif()
 
-add_definitions(-g -std=gnu99 -funwind-tables -O3 -fvisibility=hidden -no-pie -fjump-tables)
+if(USE_FLOAT)
+    add_definitions(-DUSE_FLOAT)
+endif()
+
+if(ARM_DYNAREC)
+    add_definitions(-DDYNAREC)
+    add_definitions(-DARM)
+    add_definitions(-marm)
+    enable_language(ASM)
+endif()
+
+if(NOGETCLOCK)
+    add_definitions(-DNOGETCLOCK)
+endif()
+
+if(BOX86LIB)
+    add_definitions(-DBUILD_LIB)
+    add_definitions(-fPIC)
+    if(NOT BOX86_ROOT)
+        message(FATAL_ERROR "BOX86_ROOT not set when build as lib")
+    endif()
+else()
+    set(BOX86_ROOT ${CMAKE_HOME_DIRECTORY})
+endif()
+
+
+add_definitions(-g -std=gnu99 -funwind-tables -O3 -fvisibility=hidden)
 
 if(USE_CCACHE)
     find_program(CCACHE_FOUND ccache)
@@ -39,114 +127,318 @@ if(USE_CCACHE)
     endif()
 endif()
 
-#include_directories(include)
+include_directories(
+    "${BOX86_ROOT}/src/include"
+    "${BOX86_ROOT}/src"
+    "${BOX86_ROOT}/src/wrapped/generated"
+)
 
 set(ELFLOADER_SRC
-    "${CMAKE_HOME_DIRECTORY}/src/bridge.c"
-    "${CMAKE_HOME_DIRECTORY}/src/main.c"
-    "${CMAKE_HOME_DIRECTORY}/src/pathcoll.c"
-    "${CMAKE_HOME_DIRECTORY}/src/box86context.c"
-    "${CMAKE_HOME_DIRECTORY}/src/fileutils.c"
-    "${CMAKE_HOME_DIRECTORY}/src/elfloader.c"
-    "${CMAKE_HOME_DIRECTORY}/src/elfparser.c"
-    "${CMAKE_HOME_DIRECTORY}/src/elfload_dump.c"
-    "${CMAKE_HOME_DIRECTORY}/src/library.c"
-    "${CMAKE_HOME_DIRECTORY}/src/librarian.c"
-    "${CMAKE_HOME_DIRECTORY}/src/stack.c"
-    "${CMAKE_HOME_DIRECTORY}/src/threads.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrapper.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86emu.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86run.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86run66.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86run_private.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86syscall.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86primop.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86trace.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x86int3.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x87emu_private.c"
-    "${CMAKE_HOME_DIRECTORY}/src/x87run.c"
-    "${CMAKE_HOME_DIRECTORY}/src/myalign.c"
-    "${CMAKE_HOME_DIRECTORY}/src/callback.c"
-    "${CMAKE_HOME_DIRECTORY}/src/sdl1rwops.c"
-    "${CMAKE_HOME_DIRECTORY}/src/sdl2rwops.c"
+    "${BOX86_ROOT}/src/main.c"
+    "${BOX86_ROOT}/src/box86context.c"
+    "${BOX86_ROOT}/src/tools/bridge.c"
+    "${BOX86_ROOT}/src/tools/pathcoll.c"
+    "${BOX86_ROOT}/src/tools/fileutils.c"
+    "${BOX86_ROOT}/src/tools/callback.c"
+    "${BOX86_ROOT}/src/tools/box86stack.c"
+    "${BOX86_ROOT}/src/tools/gtkclass.c"
+    "${BOX86_ROOT}/src/elfs/elfloader.c"
+    "${BOX86_ROOT}/src/elfs/elfparser.c"
+    "${BOX86_ROOT}/src/elfs/elfload_dump.c"
+    "${BOX86_ROOT}/src/librarian/library.c"
+    "${BOX86_ROOT}/src/librarian/librarian.c"
+    "${BOX86_ROOT}/src/emu/x86emu.c"
+    "${BOX86_ROOT}/src/emu/x86run.c"
+    "${BOX86_ROOT}/src/emu/x86run66.c"
+    "${BOX86_ROOT}/src/emu/x86run_private.c"
+    "${BOX86_ROOT}/src/emu/x86syscall.c"
+    "${BOX86_ROOT}/src/emu/x86primop.c"
+    "${BOX86_ROOT}/src/emu/x86trace.c"
+    "${BOX86_ROOT}/src/emu/x86int3.c"
+    "${BOX86_ROOT}/src/emu/x87emu_private.c"
+    "${BOX86_ROOT}/src/emu/x87run.c"
+    "${BOX86_ROOT}/src/libtools/myalign.c"
+    "${BOX86_ROOT}/src/libtools/myfts.c"
+    "${BOX86_ROOT}/src/libtools/sdl1rwops.c"
+    "${BOX86_ROOT}/src/libtools/sdl2rwops.c"
+    "${BOX86_ROOT}/src/libtools/signals.c"
+    "${BOX86_ROOT}/src/libtools/threads.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec.c"
+    "${BOX86_ROOT}/src/wrapped/generated/wrapper.c"
 )
 
 set(WRAPPEDS
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibc.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibpthread.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibrt.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibgl.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibglu.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibx11.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibasound.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibdl.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibm.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl1.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl1mixer.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl1image.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl2.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl2mixer.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsdl2image.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedsmpeg2.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedvorbisfile.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedpng16.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibxxf86vm.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibxrandr.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedlibz.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedopenal.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedalut.c"
-    "${CMAKE_HOME_DIRECTORY}/src/wrappedldlinux.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibc.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibpthread.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibrt.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibgl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibglu.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibx11.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibasound.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibdl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1mixer.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1image.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl1ttf.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl2mixer.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsdl2image.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsmpeg.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedsmpeg2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedvorbisfile.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibvorbis.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibogg.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpng12.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpng16.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxxf86vm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxrandr.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxext.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedxinerama.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxfixes.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcursor.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxrender.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxi.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxss.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxpm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxau.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxdmcp.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibx11xcb.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxcb.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxtst.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibxt.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibz.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedopenal.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedalut.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibjpeg.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedcurl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedudev0.c"
+    "${BOX86_ROOT}/src/wrapped/wrappeddbus.c"
+    "${BOX86_ROOT}/src/wrapped/wrappediconv.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedcrypto.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibssl.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedutil.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibuuid.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibresolv.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpulsesimple.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpulse.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibsndfile.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedfontconfig.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedfreetype.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedbz2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibsm.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibice.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibusb1.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibncursesw.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibformw.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibtinfo.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgtkx112.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgdkx112.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgdkpixbuf2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgio2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgmodule2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgobject2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedglib2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgstreamer010.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgstinterfaces010.c"
+    "${BOX86_ROOT}/src/wrapped/wrappeddbusmenugtk.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedatk.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpangoft2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpangocairo.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgthread2.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedcairo.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedpango.c"
+    "${BOX86_ROOT}/src/wrapped/wrappeddbusglib1.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedgudev1.c"
+#    "${BOX86_ROOT}/src/wrapped/wrappedmonounity.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedldlinux.c"
 )
-set(WRAPPER "${CMAKE_HOME_DIRECTORY}/src/wrapper.c" "${CMAKE_HOME_DIRECTORY}/src/wrapper.h")
-add_custom_command(OUTPUT ${WRAPPER} COMMAND "${CMAKE_HOME_DIRECTORY}/rebuild_wrappers.py" "${CMAKE_HOME_DIRECTORY}" MAIN_DEPENDENCY "${CMAKE_HOME_DIRECTORY}/rebuild_wrappers.py" DEPENDS ${WRAPPEDS})
-set_source_files_properties(${WRAPPER} PROPERTIES GENERATED TRUE)
 
-add_executable(box86 ${ELFLOADER_SRC} ${WRAPPEDS})
-target_link_libraries(box86 m dl rt pthread)
+set(WRAPPEDS_HEAD "")
+foreach(A ${WRAPPEDS})
+    string(REPLACE ".c" "_private.h" B ${A})
+    set(WRAPPEDS_HEAD ${WRAPPEDS_HEAD} ${B})
+endforeach()
 
-target_link_libraries(box86 -no-pie -Wl,-Ttext-segment,0xA8000000)
+set(WRAPPER "${BOX86_ROOT}/src/wrapped/generated/wrapper.c" "${BOX86_ROOT}/src/wrapped/generated/wrapper.h")
 
-add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_custom_command(OUTPUT ${WRAPPER} COMMAND "${PYTHON_EXECUTABLE}" "${BOX86_ROOT}/rebuild_wrappers.py" "${BOX86_ROOT}" "PANDORA" "HAVE_LD80BITS" "NOALIGN" "HAVE_TRACE" "USE_FLOAT" MAIN_DEPENDENCY "${BOX86_ROOT}/rebuild_wrappers.py" DEPENDS ${WRAPPEDS} ${WRAPPEDS_HEAD})
+
+if(ARM_DYNAREC)
+    set(DYNAREC_SRC
+        "${BOX86_ROOT}/src/dynarec/dynablock.c"
+
+        "${BOX86_ROOT}/src/dynarec/dynarec_arm.c"
+        "${BOX86_ROOT}/src/dynarec/dynarec_arm_functions.c"
+        "${BOX86_ROOT}/src/dynarec/arm_printer.c"
+
+        "${BOX86_ROOT}/src/dynarec/arm_prolog.S"
+        "${BOX86_ROOT}/src/dynarec/arm_epilog.S"
+        "${BOX86_ROOT}/src/dynarec/arm_linker.S"
+        "${BOX86_ROOT}/src/dynarec/arm_table.S"
+    )
+
+    set(DYNAREC_PASS
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_helper.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_pass.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_00.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_0f.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_65.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_66.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_67.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_d8.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_d9.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_da.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_db.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_dc.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_dd.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_de.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_df.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_f0.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_660f.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_f20f.c"
+    "${BOX86_ROOT}/src/dynarec/dynarec_arm_f30f.c"
+    )
+
+    set(DYNAREC_PASS "${BOX86_ROOT}/src/wrapped/generated/wrapper.h" ${DYNAREC_PASS})
+
+    add_library(dynarec_arm OBJECT ${DYNAREC_SRC})
+
+    add_library(arm_pass0 OBJECT ${DYNAREC_PASS})
+    set_target_properties(arm_pass0 PROPERTIES COMPILE_FLAGS "-DSTEP=0")
+    add_library(arm_pass1 OBJECT ${DYNAREC_PASS})
+    set_target_properties(arm_pass1 PROPERTIES COMPILE_FLAGS "-DSTEP=1")
+    add_library(arm_pass2 OBJECT ${DYNAREC_PASS})
+    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(dynarec STATIC 
+        $<TARGET_OBJECTS:dynarec_arm> 
+        $<TARGET_OBJECTS:arm_pass0> 
+        $<TARGET_OBJECTS:arm_pass1>
+        $<TARGET_OBJECTS:arm_pass2>
+        $<TARGET_OBJECTS:arm_pass3>
+    )
+endif()
+
+if(BOX86LIB)
+#build a static lib...
+add_library(box86 OBJECT ${ELFLOADER_SRC} ${WRAPPEDS})
+
+else(BOX86LIB)
+
+if(USE_FLOAT)
+    set(BOX86 box86f)
+elseif(HAVE_TRACE)
+    set(BOX86 box86t)
+else()
+    set(BOX86 box86)
+endif()
+
+add_executable(${BOX86} ${ELFLOADER_SRC} ${WRAPPEDS})
+target_link_libraries(${BOX86} m dl rt pthread)
+if(ARM_DYNAREC)
+    target_link_libraries(${BOX86} dynarec)
+endif()
+
+if(NOT NOLOADADDR)
+    target_link_libraries(${BOX86} -Wl,-Ttext-segment,0x03800000)
+endif()
+
+string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "i686"  _x86)
+string(COMPARE EQUAL "${CMAKE_HOST_SYSTEM_PROCESSOR}" "x86_64"  _x86_64)
+
+if(NOT _x86 AND NOT _x86_64)
+  install(TARGETS ${BOX86}
+    RUNTIME DESTINATION bin)
+  install(FILES ${CMAKE_SOURCE_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
+  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.6 DESTINATION /usr/lib/i386-linux-gnu/)
+  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libgcc_s.so.1 DESTINATION /usr/lib/i386-linux-gnu/)
+endif()
+
+if(NOT TARGET uninstall)
+    configure_file(
+        "${CMAKE_CURRENT_SOURCE_DIR}/cmake_uninstall.cmake.in"
+        "${CMAKE_CURRENT_BINARY_DIR}/cmake_uninstall.cmake"
+        IMMEDIATE @ONLY)
+
+    add_custom_target(uninstall
+        COMMAND ${CMAKE_COMMAND} -P ${CMAKE_CURRENT_BINARY_DIR}/cmake_uninstall.cmake)
+endif()
+
+add_test(test01 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test01 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref01.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test02 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test02 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test02 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref02.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test03 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test03 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test03 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref03.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test04 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test04 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test04 -D TEST_ARGS2=yeah -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref04.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test05 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test05 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test05 -D TEST_ARGS2=7 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref05.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test06 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test06 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test06 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref06.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test07 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test07 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test07 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref07.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test08 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test08 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test08 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref08.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
 
-add_test(test09 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/box86 
+add_test(test09 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
     -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test09 -D TEST_OUTPUT=tmpfile.txt 
     -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref09.txt
     -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(test10 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test10 -D TEST_OUTPUT=tmpfile.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref10.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(test11 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test11 -D TEST_OUTPUT=tmpfile.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref11.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(test12 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test12 -D TEST_OUTPUT=tmpfile.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref12.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+add_test(test13 ${CMAKE_COMMAND} -D TEST_PROGRAM=${CMAKE_BINARY_DIR}/${BOX86} 
+    -D TEST_ARGS=${CMAKE_SOURCE_DIR}/tests/test13 -D TEST_OUTPUT=tmpfile.txt 
+    -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/ref13.txt
+    -P ${CMAKE_SOURCE_DIR}/runTest.cmake )
+
+    file(GLOB extension_tests "${CMAKE_SOURCE_DIR}/tests/extensions/*.c")
+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/${testname} -D TEST_OUTPUT=tmpfile.txt
+        -D TEST_REFERENCE=${CMAKE_SOURCE_DIR}/tests/extensions/${testname}.txt
+        -P ${CMAKE_SOURCE_DIR}/runTest.cmake)
+endforeach()
+
+endif(BOX86LIB)

COMPILE.md

@@ -0,0 +1,54 @@
+Compiling
+----
+#### for Pandora
+
+ `mkdir build; cd build; cmake .. -DPANDORA=1; make`
+
+#### for Pyra
+
+ `mkdir build; cd build; cmake .. -DPYRA=1; make`
+
+#### for Gameshell
+
+`mkdir build; cd build; cmake .. -DGAMESHELL=1; make`
+
+#### for RaspberryPI
+
+  _a build for model 2, 3 and 4 can be done. Model 1 and 0 cannot (at least not with Dynarec, as they lack NEON support)_
+ 
+`mkdir build; cd build; cmake .. -DRPI4=1; make`
+ 
+  _For Pi4. Change to RPI2 or RPI3 for other models._
+
+#### for Other ARM Linux
+
+ `mkdir build; cd build; cmake .. ; make`
+
+#### for x86 Linux
+
+ `mkdir build; cd build; cmake .. -DLD80BITS=1 -DNOALIGN=1; make`
+
+#### use ccmake
+
+Alternatively, you can use the curses-bases ccmake (or any other gui frontend for cmake) to select wich platform to use interactively.
+
+#### Customize your build
+
+*use ccache if you have it*
+Add `-DUSE_CCACHE=1` if you have ccache (it's better if you plan to touch the sources)
+
+
+*to have a Trace Enabled build*
+To have a trace enabled build (warning, it will be slower), add `-DHAVE_TRACE=1` but you will need, at runtime, to have the [Zydis library](https://github.com/zyantific/zydis) library in your `LD_LIBRARY_PATH` or in the system lib folders.
+
+*to have ARM Dynarec*
+The Dynarec is only avaiable on ARM Cpu. Notes also that VFPv3 and NEON are required for the Dynarec. Activate it by using `-DARM_DYNAREC=1`. Also, be sure to use `-marm` in compilation flags (because many compileur use Thumb as default, and the dynarec will not work in this mode).
+
+----
+
+Testing
+----
+A few tests are included.
+They can be launched with `ctest`
+They are very basic and don't test much for now.
+

LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2018-2019 Sebastien Chevalier ("ptitSeb")
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

README.md

@@ -2,17 +2,71 @@
 
 Linux Userspace x86 Emulator with a twist
 
-Box86 will let run x86 Linux program (games) on non-x86 Linux, like ARM (needs to be 32bits little-endian).
+[![Donate](https://img.shields.io/badge/Donate-PayPal-green.svg)](https://www.paypal.com/cgi-bin/webscr?cmd=_donations&business=LU8Y2T62ZWFHU)
 
-Also, Box86 use native version for some "system" libraries, like libc, libm, or SDL and OpenGL, leading to more performance and easier integration with host system.
+Box86 will let you run x86 Linux programs (games) on non-x86 Linux, like ARM (host system needs to be 32bit little-endian).
 
-Most x86 Games needs OpenGL, so on ARM platform, a solution like gl4es is probably needed.
+Because Box86 uses the native version for some "system" libraries, like libc, libm, or SDL and OpenGL, it's easy to integrate and use, and performances can be surprinsigly high in some cases.
 
-Current version is higly experimental and early, and most stuff wont run and run correctly. For now, WorldOfGoo does run correctly. FTL runs but sound is distorted...
+Most x86 Games need OpenGL, so on ARM platforms, a solution like [gl4es](https://github.com/ptitSeb/gl4es) is probably needed.
 
-If you are serious about developping Box86, you should install ccache and activate it's support in the cmake project (use ccmake for example)
-To have the TRACE enabled (i.e. dumping to stdout all individual x86 instruction execute, with dump of registers), you'll also need [Zydis library](https://github.com/zyantific/zydis) accessible on your system.
+Box86 now integrate a DynaRec for ARM platform, providing a speed boost between 5 to 10 times compared to only the interpretor.
 
-Some x86 internal opcode use parts of "Realmode X86 Emulator Library", see [x86primop.c](src/x86primop.c) for copyright details
+Many games already work, like for example: WorldOfGoo, Airline Tycoon Deluxe or FTL. Many of the GameMaker linux games also run fine (there a long list, among them are UNDERTALE, A Risk of Rain, and Cook Server Delicious)
 
-Change log is accessible [here](CHANGELOG.md)
+If you are serious about developing Box86, you should install ccache and activate it's support in the cmake project (use ccmake for example)
+To have TRACE enabled (i.e. dumping to stdout all individual x86 instructions executed, with dump of registers), you'll also need [Zydis library](https://github.com/zyantific/zydis) accessible on your system.
+
+Some x86 internal opcodes use parts of "Realmode X86 Emulator Library", see [x86primop.c](src/x86primop.c) for copyright details
+
+Here are 6 videos, the first 2 of "Airline Tycoon Deluxe" and "Heretic 2" running on a gigahertz OpenPandora (the second one use the dynarec), and the next 2 of "Bit.Trip.Runner" and "Neverwinter Night" running on an ODroid XU4 (without dynarec), and the last 2 on a Pi4: Shovel Knight (video from @ITotalJustice) and Freedom Planet (video from @djazz), also without dynarec.
+
+[![Play on Youtube](https://img.youtube.com/vi/bLt0hMoFDLk/3.jpg)](https://www.youtube.com/watch?v=bLt0hMoFDLk) [![Play on Youtube](https://img.youtube.com/vi/MM7kWYts7IA/3.jpg)](https://www.youtube.com/watch?v=MM7kWYts7IA) [![Play on Youtube](https://img.youtube.com/vi/8hr71S029Hg/1.jpg)](https://www.youtube.com/watch?v=8hr71S029Hg) [![Play on Youtube](https://img.youtube.com/vi/B4YN37z3-ws/1.jpg)](https://www.youtube.com/watch?v=B4YN37z3-ws) [![Play on Youtube](https://img.youtube.com/vi/xk8Q30mxqPg/1.jpg)](https://www.youtube.com/watch?v=xk8Q30mxqPg) [![Play on Youtube](https://img.youtube.com/vi/_QMRMVvYrqU/1.jpg)](https://www.youtube.com/watch?v=_QMRMVvYrqU)
+
+----
+
+Compiling
+----
+How to compile can be found [here](COMPILE.md)
+
+----
+
+Usage
+----
+
+There are a few environment variables to control Box86 behaviour.
+
+See [here](USAGE.md) for all variables and what they do.
+
+----
+
+Version history
+----
+
+The change log is [here](CHANGELOG.md)
+
+----
+
+A note about 64bits platform
+----
+
+Because box86 works by directly translating function calls from x86 to host system, the host system (the one box86 is running on) needs to have 32bits library. Box86 doesn't include any 32bits <-> 64bits translation. So basically, to run box86 on, for example, an ARM64 platform, you will need to build box86 for arm 32bits, and also need to have a chroot with 32bits library.
+
+Also note that, even if, on day, there is a box86_64, this one will only be able to run x86_64 binary on 64bits platform. You will still need box86 (and see 32bits chroot) to run x86 binary (in fact, like it is the case on actual x86_64 linux)
+
+----
+
+A note about Unity game emulation
+----
+
+Running Unity games is not possible for now. Mono itself uses signals that are not well emulated enough. So the solution is to use a native version of the libmono used by Unity. It can be found here: https://github.com/Unity-Technologies/mono and it needs to be built from source. When built copy `libmonosgen-2.0.so` to `libmonounity.so` and put it somewhere it can be dlopen'd (so in `usr/lib` or friend or somewhere in your `LD_LIBRARY_PATH`).
+Note that libmonounity is not completely wrapped yet, and the mechanism to call x86 library from libmonounity is not done yet, so the use of libmonounity is not enable for now. 
+
+TL;DR: mono games are not working for now anyway.
+
+----
+
+Final words
+----
+
+(If you use Box86 in your project, please don't forget to mention Box86)

USAGE.md

@@ -0,0 +1,112 @@
+Usage
+----
+
+There are many environment variable to control Box86 behaviour. 
+
+##### BOX86_LOG
+Controls the Verbose level of the log
+ * 0
+ * NONE : No message (except some fatal error)
+ * 1
+ * INFO : Current default. Show some log
+ * 2
+ * DEBUG : Verbose a lot of stuffs (like relocations or function called)
+ * 3
+ * DUMP : All DEBUG plus DUMP of all ELF Info
+
+#### BOX86_LD_LIBRARY_PATH
+Controls the folder scanned to look for x86 libs. Default is current folder and `lib` in current folder.
+Also `/usr/lib/i386-linux-gnu` and `/lib/i386-linux-gnu` are added if they exists
+
+#### BOX86_PATH
+Controls the folder scanned to look for x86 executable. Default is current folder and `bin` in current folder.
+
+#### BOX86_DLSYM_ERROR
+* 0 : default. Don't log `dlsym` error
+* 1 : Log dlsym error
+
+#### BOX86_TRACE_FILE
+Send all log and trace to a file instead of `stdout`
+
+#### BOX86_TRACE
+Only on build with trace enabled. Trace allow the logging of all instruction execute, along with register dump
+* 0 : No trace
+* 1 : Trace enabled. The trace start after the init of all depending libs is done
+* symbolname : Trace only `symbolname` (trace is disable if the symbol is not found)
+* 0xXXXXXXX-0xYYYYYYY : trace only between the 2 addresses
+
+#### BOX86_TRACE_INIT
+Use BOX86_TRACE_INIT instead of BOX_TRACE to start trace before init of Libs and main program
+* 0 : No trace
+* 1 : Trace enabled. The trace start with the init of all depending libs is done
+
+#### BOX86_TRACE_START
+Only on build with trace enabled.
+* NNNNNNN : Start trace only after NNNNNNNN opcode execute (number is an `uint64_t`)
+
+#### BOX86_TRACE_XMM
+Only on build with trace enabled.
+* 0 : Default, the XMM (i.e. SSE/SSE2) register will not be logged with the general and x87 registers
+* 1 : Dump the XMM registers
+
+#### BOX86_LOAD_ADDR
+Try to load at 0xXXXXXX main binaray (if binary is a PIE)
+* 0xXXXXXXXX the load address (only active on PIE programs)
+
+#### BOX86_NOSIGSEGV
+To disable handling of SigSEGV (to ease debugging mainly)
+* 0 : default, let x86 program set sighandler for SEGV
+* 1 : disable handling of SigSEGV
+
+#### BOX86_X11COLOR16
+PANDORA only: to try convert X11 color from 32 bits to 16 bits (to avoid light green on light cyan windows
+* 0 : default, don't touch X11 colors
+* 1 : Change colors arguments in XSetForeground, XSetBackground and XCreateGC
+
+#### BOX86_X11THREADS
+To call XInitThreads when loading X11 (for old Loki games with Loki_Compat lib)
+* 0 : default, don't force call XInitThreads
+* 1 : Call XInitThreads as soon as libX11 is loaded
+
+#### BOX86_DYNAREC_LOG
+Set level of DynaRec log
+ * 0 :
+ * NONE : No Log for DynaRec
+ * 1 :
+ * INFO : Minimum Dynarec Log (only unimplemented OpCode)
+ * 2 :
+ * DEBUG : Debug Log for Dynarec (with detail on block created / executed)
+ * 3 :
+ * DUMP : All above plus dump of the x86 -> native opcode conversions
+
+#### BOX86_DYNAREC
+ * 0 : Disable Dynarec
+ * 1 : Enable Dynarec (default)
+
+#### BOX86_DYNAREC_LINKER
+ * 0 : Disable Dynarec Linker (use that on debug, with dynarec log >= 2, to have detail on wich block get executed)
+ * 1 : Enable Dynarec Linker (default)
+
+#### BOX86_DYNAREC_TRACE
+ * 0 : Disable trace for generated code (default)
+ * 1 : Enable trace for generated code (like regular Trace, this will slow down a lot and generate huge logs)
+
+#### BOX86_LIBGL
+ * libXXXX set the name for libGL (default to libGL.so.1)
+ * /PATH/TO/libGLXXX set the name and path for libGL
+ You can also use SDL_VIDEO_GL_DRIVER
+
+#### BOX86_LD_PRELOAD
+ * XXXX[:YYYYY] force loading XXXX (and YYYY...) libraries with the binary
+ PreLoaded libs can be emulated or native, and are treated the same way as if they were comming from the binary
+ 
+#### BOX86_EMULATED_LIBS
+ * XXXX[:YYYYY] force lib XXXX (and YYYY...) to be emulated (and not wrapped)
+Some games use old version of some lib, with an ABI incompatible with native version.
+Note that LittleInforno for example is auto detected, and libvorbis.so.0 is automatical added to emulated libs
+
+#### BOX86_ALLOWMISSINGLIBS
+Allow box86 to continue even if a lib is missing
+ * 0 : default, stop if a lib cannot be loaded
+ * 1 : continue even if a needed lib cannot be loaded. Unadvised, this will, in most cases, crash later on.
+

cmake_uninstall.cmake.in

@@ -0,0 +1,21 @@
+if(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
+  message(FATAL_ERROR "Cannot find install manifest: @CMAKE_BINARY_DIR@/install_manifest.txt")
+endif(NOT EXISTS "@CMAKE_BINARY_DIR@/install_manifest.txt")
+
+file(READ "@CMAKE_BINARY_DIR@/install_manifest.txt" files)
+string(REGEX REPLACE "\n" ";" files "${files}")
+foreach(file ${files})
+  message(STATUS "Uninstalling $ENV{DESTDIR}${file}")
+  if(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
+    exec_program(
+      "@CMAKE_COMMAND@" ARGS "-E remove \"$ENV{DESTDIR}${file}\""
+      OUTPUT_VARIABLE rm_out
+      RETURN_VALUE rm_retval
+      )
+    if(NOT "${rm_retval}" STREQUAL 0)
+      message(FATAL_ERROR "Problem when removing $ENV{DESTDIR}${file}")
+    endif(NOT "${rm_retval}" STREQUAL 0)
+  else(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
+    message(STATUS "File $ENV{DESTDIR}${file} does not exist.")
+  endif(IS_SYMLINK "$ENV{DESTDIR}${file}" OR EXISTS "$ENV{DESTDIR}${file}")
+endforeach(file)

rebuild_wrappers.py

@@ -4,7 +4,7 @@ import os
 import glob
 import sys
 
-values = ['E', 'e', 'v', 'c', 'w', 'i', 'I', 'C', 'W', 'u', 'U', 'f', 'd', 'D', 'L', 'p', 'V']
+values = ['E', 'e', 'v', 'c', 'w', 'i', 'I', 'C', 'W', 'u', 'U', 'f', 'd', 'D', 'K', 'l', 'L', 'p', 'V', "O", "S"]
 def splitchar(s):
 	ret = [len(s)]
 	i = 0
@@ -15,48 +15,181 @@ def splitchar(s):
 		ret.append(values.index(c))
 	return ret
 
-def main(root):
+def value(define):
+	return define[9:-1] if define.startswith("!") else define[8:-1]
+
+def splitdef(dnf, defines):
+	cunjs = dnf.split(" || ")
+	clauses = [c.split(" && ") for c in cunjs]
+	
+	ret = [len(cunjs)]
+	
+	for cunj in clauses:
+		for c in cunj:
+			ret.append(len(c))
+	for cunj in clauses:
+		for c in cunj:
+			ret.append(defines.index(value(c)) * 2 + (1 if c.startswith("!") else 0))
+	ret.append(0)
+	return ret
+
+def invert(define):
+	return define[1:] if define.startswith("!") else ("!" + define)
+
+def main(root, defines, ver):
 	# Initialize variables: gbl for all values, vals for file-per-file values, redirects for redirections
-	gbl = []
+	gbl = {}
 	vals = {}
 	redirects = {}
 	
 	# First read the files inside the headers
-	for filepath in glob.glob(os.path.join(root, "src", "wrapped*_private.h")):
+	for filepath in glob.glob(os.path.join(root, "src/wrapped/", "wrapped*_private.h")):
 		filename = filepath.split("/")[-1]
-		locval = []
+		locval = {}
+		dependants = []
 		with open(filepath, 'r') as file:
 			for line in file:
 				ln = line.strip()
+				# If the line is a `#' line (#ifdef LD80BITS/#ifndef LD80BITS/header)
+				if ln.startswith("#"):
+					preproc_cmd = ln[1:].strip()
+					try:
+						if preproc_cmd.startswith("if defined(GO)"):
+							continue #if defined(GO) && defined(GOM)...
+						elif preproc_cmd.startswith("if !(defined(GO)"):
+							continue #if !(defined(GO) && defined(GOM)...)
+						elif preproc_cmd.startswith("error"):
+							continue #error meh!
+						elif preproc_cmd.startswith("endif"):
+							if dependants != []:
+								dependants.pop()
+						elif preproc_cmd.startswith("ifdef"):
+							if preproc_cmd[5:].strip() not in defines:
+								raise KeyError(preproc_cmd[5:].strip())
+							dependants.append("defined(" + preproc_cmd[5:].strip() + ")")
+						elif preproc_cmd.startswith("ifndef"):
+							if preproc_cmd[5:].strip() not in defines:
+								raise KeyError(preproc_cmd[5:].strip())
+							dependants.append("!defined(" + preproc_cmd[6:].strip() + ")")
+						elif preproc_cmd.startswith("else"):
+							dependants[-1] = invert(dependants[-1])
+						else:
+							raise NotImplementedError("Unknown preprocessor directive: {0} ({1}:{2})".format(
+								preproc_cmd.split(" ")[0], filename, line[:-1]
+							))
+					except KeyError as k:
+						raise NotImplementedError("Unknown key: {0} ({1}:{2})".format(
+							k.args[0], filename, line[:-1]
+						), k)
 				# If the line is a `GO...' line (GO/GOM/GO2/...)...
-				if ln.startswith("GO"):
+				elif ln.startswith("GO"):
 					# ... then look at the second parameter of the line
 					ln = ln.split(",")[1].split(")")[0].strip()
+					
+					if ln[1] not in ["F"]:
+						raise NotImplementedError("Bad middle letter {0} ({1}:{2})".format(ln[1], filename, line[:-1]))
 					if any(c not in values for c in ln[2:]) or (('v' in ln[2:]) and (len(ln) > 3)):
 						old = ln
 						# This needs more work
 						acceptables = ['v', 'o', '0', '1'] + values
 						if any(c not in acceptables for c in ln[2:]):
-							raise NotImplementedError("{0} ({1})".format(ln[2:], line))
+							raise NotImplementedError("{0} ({1}:{2})".format(ln[2:], filename, line[:-1]))
 						# Ok, this is acceptable: there is 0, 1, stdout and void
 						ln = (ln
 							.replace("v", "")   # void   -> nothing
 							.replace("o", "p")  # stdout -> pointer
 							.replace("0", "p")  # 0      -> pointer
 							.replace("1", "i")) # 1      -> integer
-						redirects[old] = ln
+						redirects.setdefault(" && ".join(dependants), {})
+						redirects[" && ".join(dependants)][old] = ln
 					# Simply append the function name if it's not yet existing
-					if ln not in locval:
-						locval.append(ln)
-					if ln not in gbl:
-						gbl.append(ln)
+					locval.setdefault(" && ".join(dependants), [])
+					gbl.setdefault(" && ".join(dependants), [])
+					if ln not in locval[" && ".join(dependants)]:
+						locval[" && ".join(dependants)].append(ln)
+					if ln not in gbl[" && ".join(dependants)]:
+						gbl[" && ".join(dependants)].append(ln)
 		
 		# Sort the file local values and add it to the dictionary
-		locval.sort(key=lambda v: splitchar(v))
+		for k in locval:
+			locval[k].sort(key=splitchar)
 		vals[filename] = locval
 	
+	gbl_vals = {}
+	for k in gbl:
+		ks = k.split(" && ")
+		for v in gbl[k]:
+			if k == "":
+				gbl_vals[v] = []
+				continue
+			if v in gbl_vals:
+				if gbl_vals[v] == []:
+					continue
+				for other_key in gbl_vals[v]:
+					other_key_vals = other_key.split(" && ")
+					for other_key_val in other_key_vals:
+						if other_key_val not in ks:
+							break
+					else:
+						break
+				else:
+					gbl_vals[v].append(k)
+			else:
+				gbl_vals[v] = [k]
+	for v in gbl_vals:
+		for k in gbl_vals[v]:
+			if " && ".join([invert(v2) for v2 in k.split(" && ")]) in gbl_vals[v]:
+				gbl_vals[v] = []
+				break
+	gbl = {}
+	gbl_idxs = []
+	for k in gbl_vals:
+		if len(gbl_vals[k]) == 1:
+			key = gbl_vals[k][0]
+		else:
+			key = "(" + (") || (".join(gbl_vals[k])) + ")"
+		gbl[key] = gbl.get(key, []) + [k]
+		if (key not in gbl_idxs) and (key != "()"):
+			gbl_idxs.append(key)
+	gbl_idxs.sort(key=lambda v: splitdef(v, defines))
+	
+	redirects_vals = {}
+	for k in redirects:
+		ks = k.split(" && ")
+		for v in redirects[k]:
+			if k == "":
+				redirects_vals[(v, redirects[k][v])] = []
+				continue
+			if (v, redirects[k][v]) in redirects_vals:
+				if redirects_vals[(v, redirects[k][v])] == []:
+					continue
+				for other_key in redirects_vals[(v, redirects[k][v])]:
+					if other_key == "()":
+						break
+					other_key_vals = other_key.split(" && ")
+					for other_key_val in other_key_vals:
+						if other_key_val not in ks:
+							break
+					else:
+						break
+				else:
+					redirects_vals[(v, redirects[k][v])].append(k)
+			else:
+				redirects_vals[(v, redirects[k][v])] = [k]
+	redirects = {}
+	redirects_idxs = []
+	for k, v in redirects_vals:
+		key = "(" + (") || (".join(redirects_vals[(k, v)])) + ")"
+		val = redirects.get(key, {})
+		val[k] = v
+		redirects[key] = val
+		if (key not in redirects_idxs) and (key != "()"):
+			redirects_idxs.append(key)
+	redirects_idxs.sort(key=lambda v: splitdef(v, defines))
+	
 	# Sort the table
-	gbl.sort(key=lambda v: splitchar(v))
+	for k in gbl:
+		gbl[k].sort(key=lambda v: splitchar(v))
 	
 	def length(s):
 		l = len(s)
@@ -70,15 +203,15 @@ def main(root):
 	# File headers and guards
 	files_headers = {
 		"wrapper.c": """/*****************************************************************
- * File automatically generated by rebuild_wrappers.py (v1.0.0.02)
+ * File automatically generated by rebuild_wrappers.py (v%s)
  *****************************************************************/
 #include <stdio.h>
 #include <stdlib.h>
 #include <stdint.h>
 
 #include "wrapper.h"
-#include "x86emu_private.h"
-#include "x87emu_private.h"
+#include "emu/x86emu_private.h"
+#include "emu/x87emu_private.h"
 #include "regs.h"
 #include "x86emu.h"
 
@@ -88,9 +221,36 @@ typedef union ui64_s {
     uint32_t    d[2];
 } ui64_t;
 
+extern void* my__IO_2_1_stderr_;
+extern void* my__IO_2_1_stdin_ ;
+extern void* my__IO_2_1_stdout_;
+
+static void* io_convert(void* v)
+{
+	if(!v)
+		return v;
+	if(v==my__IO_2_1_stderr_)
+		return stderr;
+	if(v==my__IO_2_1_stdin_)
+		return stdin;
+	if(v==my__IO_2_1_stdout_)
+		return stdout;
+	return v;
+}
+
+#ifdef USE_FLOAT
+#define ST0val ST0.f
+#else
+#define ST0val ST0.d
+#endif
+#ifndef NOALIGN
+int of_convert(int flag) { return (flag&~0740000) | ((flag&0140000)<<2) | ((flag&0600000)>>2);}
+#else
+int of_convert(int flag) { return flag; }
+#endif
 """,
 		"wrapper.h": """/*****************************************************************
- * File automatically generated by rebuild_wrappers.py (v1.0.0.02)
+ * File automatically generated by rebuild_wrappers.py (v%s)
  *****************************************************************/
 #ifndef __WRAPPER_H_
 #define __WRAPPER_H_
@@ -103,13 +263,16 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 
 // list of defined wrapper
 // v = void, i = int32, u = uint32, U/I= (u)int64
+// l = signed long, L = unsigned long (long is an int with the size of a pointer)
 // p = pointer, P = callback
-// f = float, d = double, D = long double, L = fake long double
+// f = float, d = double, D = long double, K = fake long double
 // V = vaargs, E = current x86emu struct, e = ref to current x86emu struct
 // 0 = constant 0, 1 = constant 1
 // o = stdout
 // C = unsigned byte c = char
 // W = unsigned short w = short
+// O = libc O_ flags bitfield
+// S = _IO_2_1_stdXXX_ pointer (or FILE*)
 // Q = ...
 // S8 = struct, 8 bytes
 
@@ -122,31 +285,58 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 	}
 	
 	# Transform strings into arrays
-	gbl = [[c for c in v] for v in gbl]
+	for k in gbl:
+		gbl[k] = [[c for c in v] for v in gbl[k]]
 	
 	# Rewrite the wrapper.h file:
-	with open(os.path.join(root, "src", "wrapper.h"), 'w') as file:
-		file.write(files_headers["wrapper.h"])
-		for v in gbl:
+	with open(os.path.join(root, "src/wrapped/generated/", "wrapper.h"), 'w') as file:
+		file.write(files_headers["wrapper.h"] % ver)
+		for v in gbl["()"]:
 			file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
-		for v in redirects:
+		for k in gbl_idxs:
+			if k != "()":
+				file.write("\n#if " + k + "\n")
+				for v in gbl[k]:
+					file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
+				file.write("#endif\n")
+		file.write("\n")
+		for v in redirects["()"]:
 			file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
+		for k in redirects_idxs:
+			if k != "()":
+				file.write("\n#if " + k + "\n")
+				for v in redirects[k]:
+					file.write("void " + ''.join(v) + "(x86emu_t *emu, uintptr_t fnc);\n")
+				file.write("#endif\n")
 		file.write(files_guards["wrapper.h"])
 	
 	# Rewrite the wrapper.c file:
-	with open(os.path.join(root, "src", "wrapper.c"), 'w') as file:
-		file.write(files_headers["wrapper.c"])
+	with open(os.path.join(root, "src/wrapped/generated/", "wrapper.c"), 'w') as file:
+		file.write(files_headers["wrapper.c"] % ver)
 		
 		# First part: typedefs
-		for v in gbl:
-    		#		  E  		    e	 		   v  		c  			w  		i  			I  			C  			W  			u  			U  			f  		d  			D   		L			p  			V 
-			types = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "void*", "void*"]
+		for v in gbl["()"]:
+			#         E            e             v       c         w          i          I          C          W           u           U           f        d         D              K         l           L            p        V			O        S
+			types = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "intptr_t", "uintptr_t", "void*", "void*", "int32_t", "void*"]
 			if len(values) != len(types):
-    				raise NotImplementedError("len(values) = {lenval} != len(types) = {lentypes}".format(lenval=len(values), lentypes=len(types)))
+					raise NotImplementedError("len(values) = {lenval} != len(types) = {lentypes}".format(lenval=len(values), lentypes=len(types)))
 			
 			file.write("typedef " + types[values.index(v[0])] + " (*" + ''.join(v) + "_t)"
 				+ "(" + ', '.join(types[values.index(t)] for t in v[2:]) + ");\n")
+		for k in gbl_idxs:
+			if k != "()":
+				file.write("\n#if " + k + "\n")
+				for v in gbl[k]:
+					#         E            e             v       c         w          i          I          C          W           u           U           f        d         D              K         l           L            p        V			O        S
+					types = ["x86emu_t*", "x86emu_t**", "void", "int8_t", "int16_t", "int32_t", "int64_t", "uint8_t", "uint16_t", "uint32_t", "uint64_t", "float", "double", "long double", "double", "intptr_t", "uintptr_t", "void*", "void*", "int32_t", "void*"]
+					if len(values) != len(types):
+							raise NotImplementedError("len(values) = {lenval} != len(types) = {lentypes}".format(lenval=len(values), lentypes=len(types)))
+					
+					file.write("typedef " + types[values.index(v[0])] + " (*" + ''.join(v) + "_t)"
+						+ "(" + ', '.join(types[values.index(t)] for t in v[2:]) + ");\n")
+				file.write("#endif\n")
 		
+		file.write("\n")
 		# Next part: function definitions
 		
 		# Helper functions to write the function definitions
@@ -154,16 +344,16 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 			if len(args) == 0:
 				return ""
 			if d % 4 != 0:
-				raise ValueError("{d} is not a multiple of 4. Did you try passing a V and something else?")
+				raise ValueError("{d} is not a multiple of 4. Did you try passing a V and something else?".format(d=d))
 			
 			if args[0] == "0":
 				return "(void*)(R_ESP + {p}), ".format(p=d) + function_args(args[1:], d + 4)
-			elif args[0] == 1:
+			elif args[0] == "1":
 				return "1, " + function_args(args[1:], d)
 			
 			arg = [
 				"emu, ",                          # E
-				"&emu, ",						  # e
+				"&emu, ",                         # e
 				"",                               # v
 				"*(int8_t*)(R_ESP + {p}), ",      # c
 				"*(int16_t*)(R_ESP + {p}), ",     # w
@@ -176,12 +366,16 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 				"*(float*)(R_ESP + {p}), ",       # f
 				"*(double*)(R_ESP + {p}), ",      # d
 				"*(long double*)(R_ESP + {p}), ", # D
-				"FromLD((void*)(R_ESP + {p})), ", # L
+				"FromLD((void*)(R_ESP + {p})), ", # K
+				"*(intptr_t*)(R_ESP + {p}), ",    # l
+				"*(uintptr_t*)(R_ESP + {p}), ",   # L
 				"*(void**)(R_ESP + {p}), ",       # p
-				"(void*)(R_ESP + {p}), "      	  # V
+				"(void*)(R_ESP + {p}), ",         # V
+				"of_convert(*(int32_t*)(R_ESP + {p})), ",         # O
+				"io_convert(*(void**)(R_ESP + {p})), ",         # S
 			]
-			#		  E  e  v  c  w  i  I  C  W  u  U  f  d  D   L   p  V  
-			deltas = [0, 0, 4, 4, 4, 4, 8, 4, 4, 4, 8, 4, 8, 12, 12, 4, 0]
+			#         E  e  v  c  w  i  I  C  W  u  U  f  d  D   K   l  L  p  V  O  S
+			deltas = [0, 0, 4, 4, 4, 4, 8, 4, 4, 4, 8, 4, 8, 12, 12, 4, 4, 4, 0, 4, 4]
 			if len(values) != len(arg):
 				raise NotImplementedError("len(values) = {lenval} != len(arg) = {lenarg}".format(lenval=len(values), lenarg=len(arg)))
 			if len(values) != len(deltas):
@@ -202,27 +396,44 @@ typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 				"R_EAX=(unsigned short)fn({0});",                               # W
 				"R_EAX=(uint32_t)fn({0});",                                     # u
 				"ui64_t r; r.u=(uint64_t)fn({0}); R_EAX=r.d[0]; R_EDX=r.d[1];", # U
-				"float fl=fn({0}); fpu_do_push(emu); ST0.d = fl;",              # f
-				"double db=fn({0}); fpu_do_push(emu); ST0.d = db;",             # d
-				"long double ld=fn({0}); fpu_do_push(emu); ST0.d = ld;",        # D
-				"double db=fn({0}); fpu_do_push(emu); ST0.d = db;",             # L
+				"float fl=fn({0}); fpu_do_push(emu); ST0val = fl;",             # f
+				"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # d
+				"long double ld=fn({0}); fpu_do_push(emu); ST0val = ld;",       # D
+				"double db=fn({0}); fpu_do_push(emu); ST0val = db;",            # K
+				"R_EAX=(intptr_t)fn({0});",                                     # l
+				"R_EAX=(uintptr_t)fn({0});",                                    # L
 				"R_EAX=(uintptr_t)fn({0});",                                    # p
 				"\n#error Invalid return type: va_list\n",                      # V
+				"\n#error Invalid return type: at_flags\n",                     # O
+				"\n#error Invalid return type: _io_file*\n",                    # S
 			]
 			if len(values) != len(vals):
 				raise NotImplementedError("len(values) = {lenval} != len(vals) = {lenvals}".format(lenval=len(values), lenvals=len(vals)))
 			f.write(vals[values.index(rettype)].format(function_args(args)[:-2]) + " }\n")
 		
-		for v in gbl:
+		for v in gbl["()"]:
 			function_writer(file, ''.join(v), ''.join(v) + "_t", v[0], v[2:])
-		for v in redirects:
-			function_writer(file, v, redirects[v] + "_t", v[0], v[2:])
+		for k in gbl_idxs:
+			if k != "()":
+				file.write("\n#if " + k + "\n")
+				for v in gbl[k]:
+					function_writer(file, ''.join(v), ''.join(v) + "_t", v[0], v[2:])
+				file.write("#endif\n")
+		file.write("\n")
+		for v in redirects["()"]:
+			function_writer(file, v, redirects["()"][v] + "_t", v[0], v[2:])
+		for k in redirects_idxs:
+			if k != "()":
+				file.write("\n#if " + k + "\n")
+				for v in redirects[k]:
+					function_writer(file, v, redirects[k][v] + "_t", v[0], v[2:])
+				file.write("#endif\n")
 		
 		file.write(files_guards["wrapper.c"])
 	
 	return 0
 
 if __name__ == '__main__':
-	if main(sys.argv[1]) != 0:
+	if main(sys.argv[1], sys.argv[2:], "1.1.1.06") != 0:
 		exit(2)
 	exit(0)

runTest.cmake

@@ -13,6 +13,7 @@ if( NOT TEST_REFERENCE )
 endif( NOT TEST_REFERENCE )
 
 set(ENV{BOX86_LOG} 0)
+set(ENV{LD_LIBRARY_PATH} ${CMAKE_SOURCE_DIR}/x86lib)
 # run the test program, capture the stdout/stderr and the result var
 execute_process(
   COMMAND ${TEST_PROGRAM} ${TEST_ARGS} ${TEST_ARGS2}

src/box86context.c

@@ -1,8 +1,9 @@
+#define _GNU_SOURCE         /* See feature_test_macros(7) */
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#define _GNU_SOURCE         /* See feature_test_macros(7) */
 #include <dlfcn.h>
+#include <signal.h>
 
 #include "box86context.h"
 #include "elfloader.h"
@@ -13,9 +14,74 @@
 #include "bridge.h"
 #include "library.h"
 #include "callback.h"
+#include "wrapper.h"
+#include "myfts.h"
+#ifdef DYNAREC
+#include <sys/mman.h>
+#include "dynablock.h"
+
+typedef struct mmaplist_s {
+    void*         block;
+    uintptr_t     offset;           // offset in the block
+} mmaplist_t;
+
+#define MMAPSIZE (4*1024*1024)       // allocate 4Mo sized blocks
+
+uintptr_t AllocDynarecMap(box86context_t *context, int size)
+{
+    // make size 0x10 bytes aligned
+    size = (size+0x0f)&~0x0f;
+    pthread_mutex_lock(&context->mutex_mmap);
+    // look for free space
+    for(int i=0; i<context->mmapsize; ++i) {
+        if(context->mmaplist[i].offset+size < MMAPSIZE) {
+            uintptr_t ret = context->mmaplist[i].offset + (uintptr_t)context->mmaplist[i].block;
+            context->mmaplist[i].offset+=size;
+            pthread_mutex_unlock(&context->mutex_mmap);
+            return ret;
+        }
+    }
+    // no luck, add a new one !
+    int i = context->mmapsize++;    // yeah, usefull post incrementation
+    dynarec_log(LOG_DEBUG, "Ask for DynaRec Block Alloc #%d\n", context->mmapsize);
+    context->mmaplist = (mmaplist_t*)realloc(context->mmaplist, context->mmapsize*sizeof(mmaplist_t));
+    void* p = mmap(NULL, MMAPSIZE, PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+    if(p==MAP_FAILED) {
+        dynarec_log(LOG_INFO, "Cannot create memory map of %d byte for dynarec block #%d\n", MMAPSIZE, i);
+        --context->mmapsize;
+        pthread_mutex_unlock(&context->mutex_mmap);
+        return 0;
+    }
+    context->mmaplist[i].block = p;
+    context->mmaplist[i].offset=size;
+    pthread_mutex_unlock(&context->mutex_mmap);
+    return (uintptr_t)p;
+}
+#endif
 
 void x86Syscall(x86emu_t *emu);
 
+/// maxval not inclusive
+int getrand(int maxval)
+{
+    if(maxval<1024) {
+        return ((random()&0x7fff)*maxval)/0x7fff;
+    } else {
+        uint64_t r = random();
+        r = (r*maxval) / RAND_MAX;
+        return r;
+    }
+}
+
+void free_tlsdatasize(void* p)
+{
+    if(!p)
+        return;
+    tlsdatasize_t *data = (tlsdatasize_t*)p;
+    free(data->tlsdata);
+    free(p);
+}
+
 box86context_t *NewBox86Context(int argc)
 {
     // init and put default values
@@ -23,20 +89,39 @@ box86context_t *NewBox86Context(int argc)
 
     context->deferedInit = 1;
 
-    context->maplib = NewLibrarian();
+    context->maplib = NewLibrarian(context);
     context->system = NewBridge();
     // create vsyscall
-    context->vsyscall = AddBridge(context->system, vFv, x86Syscall);
-
+    context->vsyscall = AddBridge(context->system, vFv, x86Syscall, 0);
+#ifdef BUILD_LIB
+    context->box86lib = RTLD_DEFAULT;   // not ideal
+#else
     context->box86lib = dlopen(NULL, RTLD_NOW|RTLD_GLOBAL);
+#endif
     context->dlprivate = NewDLPrivate();
 
     context->callbacks = NewCallbackList();
 
     context->argc = argc;
-    context->argv = (char**)calloc(context->argc, sizeof(char*));
+    context->argv = (char**)calloc(context->argc+1, sizeof(char*));
 
     pthread_mutex_init(&context->mutex_once, NULL);
+    pthread_mutex_init(&context->mutex_once2, NULL);
+    pthread_mutex_init(&context->mutex_trace, NULL);
+    pthread_mutex_init(&context->mutex_lock, NULL);
+
+    pthread_key_create(&context->tlskey, free_tlsdatasize);
+
+#ifdef DYNAREC
+    pthread_mutex_init(&context->mutex_blocks, NULL);
+    pthread_mutex_init(&context->mutex_mmap, NULL);
+    context->dynablocks = NewDynablockList(0, 0, 0);
+#endif
+    InitFTSMap(context);
+
+    for (int i=0; i<4; ++i) context->canary[i] = 1 +  getrand(255);
+    context->canary[getrand(4)] = 0;
+    printf_log(LOG_DEBUG, "Setting up canary (for Stack protector) at GS:0x14, value:%08X\n", *(uint32_t*)context->canary);
 
     return context;
 }
@@ -49,25 +134,42 @@ void FreeBox86Context(box86context_t** context)
     if(--(*context)->forked >= 0)
         return;
 
+    FreeFTSMap(*context);
+
+    if((*context)->maplib)
+        FreeLibrarian(&(*context)->maplib);
+
+#ifdef DYNAREC
+    dynarec_log(LOG_INFO, "Free global Dynarecblocks\n");
+    if((*context)->dynablocks)
+        FreeDynablockList(&(*context)->dynablocks);
+    for (int i=0; i<(*context)->mmapsize; ++i)
+        if((*context)->mmaplist[i].block)
+            munmap((*context)->mmaplist[i].block, MMAPSIZE);
+    free((*context)->mmaplist);
+    pthread_mutex_destroy(&(*context)->mutex_blocks);
+    pthread_mutex_destroy(&(*context)->mutex_mmap);
+#endif
+    
     if((*context)->emu)
         FreeX86Emu(&(*context)->emu);
 
     FreeCollection(&(*context)->box86_path);
     FreeCollection(&(*context)->box86_ld_lib);
+    FreeCollection(&(*context)->box86_emulated_libs);
     if((*context)->zydis)
         DeleteX86Trace(*context);
 
     if((*context)->deferedInitList)
         free((*context)->deferedInitList);
 
+#ifndef BUILD_LIB
     if((*context)->box86lib)
         dlclose((*context)->box86lib);
+#endif
 
     FreeDLPrivate(&(*context)->dlprivate);
 
-    if((*context)->maplib)
-        FreeLibrarian(&(*context)->maplib);
-    
     for(int i=0; i<(*context)->argc; ++i)
         free((*context)->argv[i]);
     free((*context)->argv);
@@ -83,6 +185,9 @@ void FreeBox86Context(box86context_t** context)
 
     free((*context)->stack);
 
+    free((*context)->fullpath);
+    free((*context)->box86path);
+
     FreeBridge(&(*context)->system);
 
     if((*context)->glwrappers)
@@ -92,7 +197,32 @@ void FreeBox86Context(box86context_t** context)
 
     FreeCallbackList(&(*context)->callbacks);
 
+    void* ptr;
+    if ((ptr = pthread_getspecific((*context)->tlskey)) != NULL) {
+        free_tlsdatasize(ptr);
+        pthread_setspecific((*context)->tlskey, NULL);
+    }
+    pthread_key_delete((*context)->tlskey);
+
+    if((*context)->tlsdata)
+        free((*context)->tlsdata);
+
     pthread_mutex_destroy(&(*context)->mutex_once);
+    pthread_mutex_destroy(&(*context)->mutex_once2);
+    pthread_mutex_destroy(&(*context)->mutex_trace);
+    pthread_mutex_destroy(&(*context)->mutex_lock);
+
+    if((*context)->atfork_sz) {
+        free((*context)->atforks);
+        (*context)->atforks = NULL;
+        (*context)->atfork_sz = (*context)->atfork_cap = 0;
+    }
+
+    for(int i=0; i<MAX_SIGNAL; ++i)
+        if((*context)->signals[i]!=0 && (*context)->signals[i]!=1) {
+            signal(i, SIG_DFL);
+        }
+
 
     free(*context);
     *context = NULL;
@@ -110,3 +240,9 @@ int AddElfHeader(box86context_t* ctx, elfheader_t* head) {
     printf_log(LOG_DEBUG, "Adding \"%s\" as #%d in elf collection\n", ElfName(head), idx);
     return idx;
 }
+
+int AddTLSPartition(box86context_t* context, int tlssize) {
+    context->tlssize += tlssize;
+    context->tlsdata = realloc(context->tlsdata, context->tlssize);
+    return -context->tlssize;   // negative offset
+}

src/box86version.h

@@ -2,7 +2,7 @@
 #define __BOX86_VERSION_H_
 
 #define BOX86_MAJOR 0
-#define BOX86_MINOR 0
-#define BOX86_REVISION 2
+#define BOX86_MINOR 1
+#define BOX86_REVISION 0
 
 #endif //__BOX86_VERSION_H_

src/callback.c

@@ -1,134 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdint.h>
-#include <string.h>
-
-#include "debug.h"
-#include "x86emu.h"
-#include "x86run.h"
-#include "x86emu_private.h"
-#include "x86run_private.h"
-#include "box86context.h"
-#include "stack.h"
-#include "khash.h"
-
-typedef struct onecallback_s {
-    x86emu_t    *emu;
-    uintptr_t   fnc;
-    int         nb_args;
-    void*       arg[8];
-    void*       stack;
-} onecallback_t;
-
-KHASH_MAP_INIT_INT(callbacks, onecallback_t*)
-
-typedef struct callbacklist_s {
-    kh_callbacks_t      *list;
-    int                 cap;
-    int                 sz;
-} callbacklist_t;
-
-x86emu_t* AddCallback(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4)
-{
-    callbacklist_t *callbacks = emu->context->callbacks;
-    int stsize = 1024*1024;
-    void* stack = malloc(stsize);
-    x86emu_t * newemu = NewX86Emu(emu->context, fnc, (uintptr_t)stack, stsize);
-	SetupX86Emu(newemu, emu->shared_global, emu->globals);
-    newemu->trace_start = emu->trace_start;
-    newemu->trace_end = emu->trace_end;
-
-    onecallback_t * cb;
-    int ret;
-    khint_t k = kh_put(callbacks, callbacks->list, (uintptr_t)newemu, &ret);
-    cb = kh_value(callbacks->list, k) = (onecallback_t*)calloc(1, sizeof(onecallback_t));
-
-    cb->stack = stack;
-    cb->emu = newemu;
-    cb->fnc = fnc;
-    cb->nb_args = nb_args;
-    cb->arg[0] = arg1;
-    cb->arg[1] = arg2;
-    cb->arg[2] = arg3;
-    cb->arg[3] = arg4;
-
-    return newemu;
-}
-
-onecallback_t* FindCallback(x86emu_t* emu)
-{
-    // find the callback first
-    callbacklist_t *callbacks = emu->context->callbacks;
-    khint_t k = kh_get(callbacks, callbacks->list, (uintptr_t)emu);
-    if(k==kh_end(callbacks->list))
-        return NULL;
-    return kh_value(callbacks->list, k);
-}
-
-void FreeCallback(x86emu_t* emu)
-{
-    // find the callback first
-    callbacklist_t *callbacks = emu->context->callbacks;
-    khint_t k = kh_get(callbacks, callbacks->list, (uintptr_t)emu);
-    if(k==kh_end(callbacks->list))
-        return;
-    onecallback_t* cb = kh_value(callbacks->list, k);
-    FreeX86Emu(&cb->emu);
-    free(cb->stack);
-    free(cb);
-    kh_del(callbacks, callbacks->list, k);
-}
-
-uint32_t RunCallback(x86emu_t* emu)
-{
-    onecallback_t *cb = FindCallback(emu);
-    if(cb) {
-        for (int i=cb->nb_args-1; i>=0; --i)    // reverse order
-            Push(emu, (uint32_t)cb->arg[i]);
-        EmuCall(emu, cb->fnc);
-        return R_EAX;
-    }
-    printf_log(LOG_INFO, "Warning, Callback not found?!\n");
-    return 0;
-}
-
-void SetCallbackArg(x86emu_t* emu, int arg, void* val)
-{
-    onecallback_t *cb = FindCallback(emu);
-    if(cb) {
-        cb->arg[arg] = val;
-    }
-}
-
-void SetCallbackNArg(x86emu_t* emu, int narg)
-{
-    onecallback_t *cb = FindCallback(emu);
-    if(cb) {
-        cb->nb_args = narg;
-    }
-}
-
-
-callbacklist_t* NewCallbackList()
-{
-    callbacklist_t* callbacks = (callbacklist_t*)calloc(1, sizeof(callbacklist_t));
-    callbacks->list = kh_init(callbacks);
-    return callbacks;
-}
-
-void FreeCallbackList(callbacklist_t** callbacks)
-{
-    if(!callbacks)
-        return;
-    if(!*callbacks)
-        return;
-    onecallback_t* cb;
-    kh_foreach_value((*callbacks)->list, cb,
-        FreeX86Emu(&cb->emu);
-        free(cb->stack);
-        free(cb);
-    );
-    kh_destroy(callbacks, (*callbacks)->list);
-    free(*callbacks);
-    *callbacks = NULL;
-}

src/callback.h

@@ -1,18 +0,0 @@
-#ifndef __CALLBACK_H__
-#define __CALLBACK_H__
-
-#include <stdint.h>
-
-typedef struct callbacklist_s callbacklist_t;
-typedef struct x86emu_s x86emu_t;
-
-callbacklist_t* NewCallbackList();
-void FreeCallbackList(callbacklist_t** callbacks);
-
-x86emu_t* AddCallback(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4);
-void FreeCallback(x86emu_t* emu);
-uint32_t RunCallback(x86emu_t* emu);
-void SetCallbackArg(x86emu_t* emu, int arg, void* val);
-void SetCallbackNArg(x86emu_t* emu, int narg);
-
-#endif //__CALLBACK_H__

src/debug.h

@@ -1,17 +0,0 @@
-#ifndef __DEBUG_H_
-#define __DEBUG_H_
-
-extern int box86_log;    // log level
-extern int trace_xmm;    // include XMM reg in trace?
-#define LOG_NONE 0
-#define LOG_INFO 1
-#define LOG_DEBUG 2
-#define LOG_DUMP 3
-
-extern FILE* ftrace;
-
-#define printf_log(L, ...) if(L<=box86_log) fprintf(ftrace, __VA_ARGS__)
-
-#define EXPORT __attribute__((visibility("default")))
-
-#endif //__DEBUG_H_

src/dynarec/arm_emitter.h

@@ -0,0 +1,906 @@
+#ifndef __ARM_EMITTER_H__
+#define __ARM_EMITTER_H__
+/*
+    ARM Emitter
+
+Most ARM instruction are encoded using the same pattern
+
+Condition is stored on bits 28-31
+Barel roll is stored on bits 5-6 for op, 7-11 for value, 0-4 for reg
+Source register is 12-15
+Dest register is 16-19
+Op is 20-27
+*/
+
+// x86 Register mapping
+#define xEAX    4
+#define xECX    5
+#define xEDX    6
+#define xEBX    7
+#define xESP    8
+#define xEBP    9
+#define xESI    10
+#define xEDI    11
+#define xEIP    12
+// scratch registers
+#define x1      1
+#define x2      2
+#define x3      3
+#define x12     12
+// emu is r0
+#define xEmu    0
+// ARM SP is r13
+#define xSP     13      
+
+// barel roll operations (4 possibles)
+#define brLSL(i, r) (0<<4 | 0<<5 | ((i&31)<<7) | r)
+#define brLSR(i, r) (0<<4 | 1<<5 | ((i&31)<<7) | r)
+#define brASR(i, r) (0<<4 | 2<<5 | ((i&31)<<7) | r)
+#define brROR(i, r) (0<<4 | 3<<5 | ((i&31)<<7) | r)
+#define brIMM(r)    (abs(r))
+// barel roll with a register
+#define brRLSL(i, r) (1<<4 | 0<<5 | ((i&15)<<8) | r)
+#define brRLSR(i, r) (1<<4 | 1<<5 | ((i&15)<<8) | r)
+#define brRASR(i, r) (1<<4 | 2<<5 | ((i&15)<<8) | r)
+#define brRROR(i, r) (1<<4 | 3<<5 | ((i&15)<<8) | r)
+
+
+// conditions
+#define cEQ (0b0000<<28)
+#define cNE (0b0001<<28)
+#define cCS (0b0010<<28)
+#define cHS cCS
+#define cCC (0b0011<<28)
+#define cLO cCC
+#define cMI (0b0100<<28)
+#define cPL (0b0101<<28)
+#define cVS (0b0110<<28)
+#define cVC (0b0111<<28)
+#define cHI (0b1000<<28)
+#define cLS (0b1001<<28)
+#define cGE (0b1010<<28)
+#define cLT (0b1011<<28)
+#define cGT (0b1100<<28)
+#define cLE (0b1101<<28)
+#define c__ (0b1110<<28)    // means all
+
+//  nop
+#define NOP     EMIT(0xe1a00000)
+
+// mov dst, src
+#define MOV_REG(dst, src) EMIT(0xe1a00000 | ((dst) << 12) | (src) )
+// movxx dst, src
+#define MOV_REG_COND(cond, dst, src) EMIT( cond | 0x01a00000 | ((dst) << 12) | (src) )
+
+// movw dst, #imm16
+#define MOVW(dst, imm16) EMIT(0xe3000000 | ((dst) << 12) | (((imm16) & 0xf000) << 4) | brIMM((imm16) & 0x0fff) )
+// movt dst, #imm16
+#define MOVT(dst, imm16) EMIT(0xe3400000 | ((dst) << 12) | (((imm16) & 0xf000) << 4) | brIMM((imm16) & 0x0fff) )
+// pseudo insruction: mov reg, #imm with imm a 32bits value
+#define MOV32(dst, imm32)                   \
+    MOVW(dst, ((uint32_t)imm32)&0xffff);    \
+    if (((uint32_t)imm32)>>16) {            \
+        MOVT(dst, (((uint32_t)imm32)>>16)); }
+// pseudo insruction: mov reg, #imm with imm a 32bits value, fixed size
+#define MOV32_(dst, imm32)                   \
+    MOVW(dst, ((uint32_t)imm32)&0xffff);    \
+    MOVT(dst, (((uint32_t)imm32)>>16))
+// movw.cond dst, #imm16
+#define MOVW_COND(cond, dst, imm16) EMIT(cond | 0x03000000 | ((dst) << 12) | (((imm16) & 0xf000) << 4) | brIMM((imm16) & 0x0fff) )
+// mov dst #imm8 ror imm4*2
+#define MOV_IMM(dst, imm8, rimm4) EMIT(0xe3a00000 | ((dst) << 12) | (imm8) | ((rimm4) << 8) )
+// mov.cond dst #imm8 ror imm4*2
+#define MOV_IMM_COND(cond, dst, imm8, rimm4) EMIT(cond | 0x03a00000 | ((dst) << 12) | (imm8) | ((rimm4) << 8) )
+
+// mov dst, src lsl imm5
+#define MOV_REG_LSL_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (0<<5) | (imm5<<7))
+// mov dst, src lsr imm5
+#define MOV_REG_LSR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (1<<5) | (imm5<<7))
+// mov dst, src asr imm5
+#define MOV_REG_ASR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (2<<5) | (imm5<<7))
+// mov dst, src ror imm5
+#define MOV_REG_ROR_IMM5(dst, src, imm5) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (0<<4) | (3<<5) | (imm5<<7))
+
+// mov dst, src lsl rs
+#define MOV_REG_LSL_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (0<<5) | (rs<<8))
+// mov dst, src lsr rs
+#define MOV_REG_LSR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (1<<5) | (rs<<8))
+// mov dst, src asr rs
+#define MOV_REG_ASR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (2<<5) | (rs<<8))
+// mov dst, src ror rs
+#define MOV_REG_ROR_REG(dst, src, rs) EMIT(0xe1a00000 | ((dst) << 12) | (src) | (1<<4) | (3<<5) | (rs<<8))
+
+// sub dst, src, #(imm8)
+#define SUB_IMM8(dst, src, imm8) \
+    EMIT(0xe2400000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// sub cond dst, src, #(imm8)
+#define SUB_COND_IMM8(dst, src, imm8) \
+    EMIT((cond) | 0x02400000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// sub.s dst, src, #(imm8)
+#define SUBS_IMM8(dst, src, imm8) \
+    EMIT(0xe2500000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add dst, src1, src2, lsl #imm
+#define SUB_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe0400000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// add.s dst, src1, src2, lsl #imm
+#define SUBS_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe0500000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+
+// sbc dst, src, #(imm8)
+#define SBC_IMM8(dst, src, imm8) \
+    EMIT(0xe2c00000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+
+// rsb dst, src, #(imm8)
+#define RSB_IMM8(dst, src, imm8) \
+    EMIT(0xe2600000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// rsb cond dst, src, #(imm8)
+#define RSB_COND_IMM8(cond, dst, src, imm8) \
+    EMIT((cond) | 0x02600000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+
+// and dst, src1, src2, lsl #imm
+#define AND_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0000000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// and.s dst, src1, src2, lsl #imm
+#define ANDS_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0100000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// and dst, src1, src2, lsr #imm
+#define AND_REG_LSR_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0000000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm5, src2) )
+// and dst, src, #(imm8)
+#define AND_IMM8(dst, src, imm8) \
+    EMIT(0xe2000000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// and.s dst, src, #(imm8)
+#define ANDS_IMM8(dst, src, imm8) \
+    EMIT(0xe2100000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add dst, src, #(imm8)
+#define ADD_IMM8(dst, src, imm8) \
+    EMIT(0xe2800000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add.s dst, src, #(imm8)
+#define ADDS_IMM8(dst, src, imm8) \
+    EMIT(0xe2900000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// add dst, src1, src2, lsl #imm
+#define ADD_REG_LSL_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0800000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// add.s dst, src1, src2, lsl #imm
+#define ADDS_REG_LSL_IMM5(dst, src1, src2, imm8) \
+    EMIT(0xe0900000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// add dst, src1, src2, lsr #imm
+#define ADD_REG_LSR_IMM5(dst, src1, src2, imm5) \
+    EMIT(0xe0800000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm5, src2) )
+// cmp.s dst, src1, src2, lsl #imm
+#define CMPS_REG_LSL_IMM5(src1, src2, imm5) \
+    EMIT(0xe1500000 | ((0) << 12) | ((src1) << 16) | brLSL(imm5, src2) )
+// cmp.s dst, src, #imm
+#define CMPS_IMM8(src, imm8) \
+    EMIT(0xe3500000 | ((0) << 12) | ((src) << 16) | brIMM(imm8) )
+// tst.s dst, src1, src2, lsl #imm
+#define TSTS_REG_LSL_IMM8(src1, src2, imm8) \
+    EMIT(0xe1100000 | ((0) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// tst.s dst, src1, #imm
+#define TSTS_IMM8(src, imm8) \
+    EMIT(0xe3100000 | ((0) << 12) | ((src) << 16) | brIMM(imm8) )
+// tst.s dst, src1, #imm ror rot*2
+#define TSTS_IMM8_ROR(src, imm8, rot) \
+    EMIT(0xe3100000 | ((0) << 12) | ((src) << 16) | ((rot)<<8) | brIMM(imm8) )
+// orr dst, src1, src2, lsl #imm
+#define ORR_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// orr.s dst, src1, src2, lsl #imm
+#define ORRS_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe1900000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// orr dst, src1, #imm8
+#define ORR_IMM8(dst, src, imm8, rot) \
+    EMIT(0xe3800000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// orr dst, src1, src2, lsl #rs
+#define ORR_REG_LSL_REG(dst, src1, src2, rs) \
+    EMIT(0xe1800000 | ((dst) << 12) | ((src1) << 16) | brRLSL(rs, src2) )
+// xor dst, src1, src2, lsl #imm
+#define XOR_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// xor.s dst, src1, src2, lsl #imm
+#define XORS_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe0300000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// xor dst, src, #(imm8)
+#define XOR_IMM8(dst, src, imm8) \
+    EMIT(0xe2200000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// xor.cond dst, src, #(imm8)
+#define XOR_IMM8_COND(cond, dst, src, imm8) \
+    EMIT(cond | 0x02200000 | ((dst) << 12) | ((src) << 16) | brIMM(imm8) )
+// xor dst, src1, src2, lsl #rs
+#define XOR_REG_LSL_REG(dst, src1, src2, rs) \
+    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brRLSL(rs, src2) )
+// xor dst, src1, src2, lsl #imm
+#define XOR_REG_LSR_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe0200000 | ((dst) << 12) | ((src1) << 16) | brLSR(imm8, src2) )
+// bic dst, src1, src2, lsl #imm
+#define BIC_REG_LSL_IMM8(dst, src1, src2, imm8) \
+    EMIT(0xe1c00000 | ((dst) << 12) | ((src1) << 16) | brLSL(imm8, src2) )
+// bic dst, src, IMM8
+#define BIC_IMM8(dst, src, imm8, rot) \
+    EMIT(0xe3c00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// bic dst, src1, src2, lsl #imm
+#define BIC_REG_LSL_REG(dst, src1, src2, rs) \
+    EMIT(0xe1c00000 | ((dst) << 12) | ((src1) << 16) | brRLSL(rs, src2) )
+// mvn dst, src1, src2, lsl #imm
+#define MVN_REG_LSL_IMM8(dst, rm, imm8) \
+    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm8, rm) )
+// mvn dst, src, IMM8
+#define MVN_IMM8(dst, src, imm8, rot) \
+    EMIT(0xe3e00000 | ((dst) << 12) | ((src) << 16) | ((rot)<<8) | imm8 )
+// mvn dst, src1, src2, lsr #imm
+#define MVN_REG_LSR_IMM8(dst, rm, imm8) \
+    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | brLSR(imm8, rm) )
+// mvn dst, rm lsr rs
+#define MVN_REG_LSR_REG(dst, rm, rs) \
+    EMIT(0xe1e00000 | ((dst) << 12) | (0 << 16) | (1<<4) | (1<<5) | (rs<<8) | (rm) )
+// mvn with condition dst, src1, src2, lsl #imm
+#define MVN_COND_REG_LSL_IMM8(cond, dst, rm, imm8) \
+    EMIT(cond | 0x01e00000 | ((dst) << 12) | (0 << 16) | brLSL(imm8, rm) )
+
+// Single data transfert construction
+#define SDT_REG(Cond, P, U, B, W, L, Rn, Rd, ShiftRm) (Cond | (0b00<<26) | (1<<25) | (P<<24) | (U<<23) | (B<<22) | (U<<23) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | ShiftRm)
+#define SDT_OFF(Cond, P, U, B, W, L, Rn, Rd, Imm12)   (Cond | (0b00<<26) | (0<<25) | (P<<24) | (U<<23) | (B<<22) | (U<<23) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | Imm12)
+// ldr reg, [addr, #+/-imm9]
+#define LDR_IMM9(reg, addr, imm9) EMIT(0xe5100000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldrxx reg, [addr, #+/-imm9]
+#define LDR_IMM9_COND(cond, reg, addr, imm9) EMIT(cond | 0x05100000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldrb reg, [addr, #+/-imm9]
+#define LDRB_IMM9(reg, addr, imm9) EMIT(0xe5500000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldr reg, [addr, #+/-imm9]!
+#define LDR_IMM9_W(reg, addr, imm9) EMIT(0xe5300000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldr reg, [addr, rm lsl imm5]
+#define LDR_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(c__ | (0b011<<25) | (1<<24) | (1<<23) | (0<<21) | (1<<20) | ((reg) << 12) | ((addr) << 16) | brLSL(imm5, rm) )
+// ldr reg, [addr, rm lsr imm5]
+#define LDR_REG_LSR_IMM5(reg, addr, rm, imm5) EMIT(c__ | (0b011<<25) | (1<<24) | (1<<23) | (0<<21) | (1<<20) | ((reg) << 12) | ((addr) << 16) | brLSR(imm5, rm) )
+// ldrb reg, [addr, rm lsl imm5]
+#define LDRB_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe5d00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// ldr reg, [addr], #+/-imm9
+#define LDRAI_IMM9_W(reg, addr, imm9)   EMIT(0xe4100000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldrb reg, [addr], #+/-imm9
+#define LDRBAI_IMM9_W(reg, addr, imm9)  EMIT(0xe4100000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// ldr reg, [addr], rm lsl imm5
+#define LDRAI_REG_LSL_IMM5(reg, addr, rm, imm5)  EMIT(0xe6900000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// ldrb reg, [addr], rm lsl imm5
+#define LDRBAI_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe6d00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// ldrd reg, reg+1, [addr, #+/-imm9], reg must be even, reg+1 is implicit
+#define LDRD_IMM8(reg, addr, imm8) EMIT(c__ | 0b000<<25 | 1<<24  | (((imm8)<0)?0:1)<<23 | 1<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | ((abs(imm8))&0xf0)<<(8-4) | (0b1101<<4) | ((abs(imm8))&0x0f) )
+
+// str reg, [addr, #+/-imm9]
+#define STR_IMM9(reg, addr, imm9) EMIT(0xe5000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// strb reg, [addr, #+/-imm9]
+#define STRB_IMM9(reg, addr, imm9) EMIT(0xe5400000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// str reg, [addr], #+/-imm9
+#define STRAI_IMM9_W(reg, addr, imm9)  EMIT(0xe4000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// str reg, [addr, rm lsl imm5]
+#define STR_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe5800000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// strb reg, [addr, rm lsl imm5]
+#define STRB_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe5c00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// strb reg, [addr], #+/-imm9
+#define STRBAI_IMM9_W(reg, addr, imm9)  EMIT(0xe4400000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
+// str reg, [addr], rm lsl imm5
+#define STRAI_REG_LSL_IMM5(reg, addr, rm, imm5)  EMIT(0xe6800000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// strb reg, [addr], rm lsl imm5
+#define STRBAI_REG_LSL_IMM5(reg, addr, rm, imm5) EMIT(0xe6c00000 | ((reg) << 12) | ((addr) << 16) | (1<<25) | brLSL(imm5, rm) )
+// strd reg, reg+1, [addr, #+/-imm8], reg must be even, reg+1 is implicit
+#define STRD_IMM8(reg, addr, imm8) EMIT(c__ | 0b000<<25 | 1<<24 | (((imm8)<0)?0:1)<<23 | 1<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | ((abs(imm8))&0xf0)<<(8-4) | (0b1111<<4) | ((abs(imm8))&0x0f) )
+
+// bx reg
+#define BX(reg) EMIT(0xe12fff10 | (reg) )
+// bx cond reg
+#define BXcond(C, reg) EMIT(C | 0x012fff10 | (reg) )
+
+// blx reg
+#define BLX(reg) EMIT(0xe12fff30 | (reg) )
+
+// b cond offset
+#define Bcond(C, O) EMIT(C | (0b101<<25) | (0<<24) | ((O)>>2)&0xffffff)
+
+// bl cond offset
+#define BLcond(C, O) EMIT(C | (0b101<<25) | (1<<24) | ((O)>>2)&0xffffff)
+
+// push reg!, {list}
+//                           all |    const    |pre-index| subs    | no PSR  |writeback| store   |   base    |reg list
+#define PUSH(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (1<<21) | (0<<20) | (reg<<16) | (list))
+
+// pop reg!, {list}
+//                           all |    const    |postindex|  add    | no PSR  |writeback|  load   |   base    |reg list
+#define POP(reg, list)  EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (1<<21) | (1<<20) | (reg<<16) | (list))
+
+// STMDB reg, {list}
+//                            all |    const    |pre-index| subs    | no PSR  |  no wb  | store   |   base    |reg list
+#define STMDB(reg, list) EMIT(c__ | (0b100<<25) | (1<<24) | (0<<23) | (0<<22) | (0<<21) | (0<<20) | (reg<<16) | (list))
+// STMia reg, {list}
+//                          all |    const    |postindex|   add   | no PSR  |  no wb  |  store  |   base    |reg list
+#define STM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (0<<20) | (reg<<16) | (list))
+// LDMia reg, {list}
+//                          all |    const    |postindex|   add   | no PSR  |  no wb  |  load   |   base    |reg list
+#define LDM(reg, list) EMIT(c__ | (0b100<<25) | (0<<24) | (1<<23) | (0<<22) | (0<<21) | (1<<20) | (reg<<16) | (list))
+
+
+// Half Word and signed data transfert construction
+#define HWS_REG(Cond, P, U, W, L, Rn, Rd, S, H, Rm)     (Cond | (0b000<<25) | (P<<24) | (U<<23) | (0<<22) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | (1<<7) | (S<<6) | (H<<5) | (1<<4) | Rm)
+#define HWS_OFF(Cond, P, U, W, L, Rn, Rd, S, H, Imm8)   (Cond | (0b000<<25) | (P<<24) | (U<<23) | (1<<22) | (W<<21) | (L<<20) | (Rn<<16) | (Rd<<12) | ((Imm8&0xf0)<<(8-4)) | (1<<7) | (S<<6) | (H<<5) | (1<<4) | (Imm8&0x0f))
+
+#define LDRSB_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 1, addr, reg, 1, 0, abs(imm8)))
+#define LDRSH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 1, addr, reg, 1, 1, abs(imm8)))
+#define LDRH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1,  (((imm8)<0)?0:1), 0, 1, addr, reg, 0, 1, abs(imm8)))
+#define LDRH_IMM8_COND(cond, reg, addr, imm8) EMIT(HWS_OFF(cond, 1, (((imm8)<0)?0:1), 0, 1, addr, reg, 0, 1, abs(imm8)))
+#define STRH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 0, addr, reg, 0, 1, abs(imm8)))
+#define STRSH_IMM8(reg, addr, imm8) EMIT(HWS_OFF(c__, 1, (((imm8)<0)?0:1), 0, 0, addr, reg, 1, 1, abs(imm8)))
+
+#define LDRHAI_REG_LSL_IMM5(reg, addr, rm) EMIT(HWS_REG(c__, 0, 1, 0, 1, addr, reg, 0, 1, rm))
+#define STRHAI_REG_LSL_IMM5(reg, addr, rm) EMIT(HWS_REG(c__, 0, 1, 0, 0, addr, reg, 0, 1, rm))
+
+// Mul Long construction
+#define MULLONG(Cond, U, A, S, RdHi, RdLo, Rs, Rm)     (Cond | (0b00001<<23) | (U<<22) | (A<<21) | (S<<20) | (RdHi<<16) | (RdLo<<12) | (Rs<<8) | (0b1001<<4) | (Rm))
+
+#define UMULL(RdHi, RdLo, Rs, Rm)   EMIT(MULLONG(c__, 0, 0, 0, RdHi, RdLo, Rs, Rm))
+#define SMULL(RdHi, RdLo, Rs, Rm)   EMIT(MULLONG(c__, 1, 0, 0, RdHi, RdLo, Rs, Rm))
+
+// Mul and MulA
+#define MULMULA(Cond, A, S, Rd, Rn, Rs, Rm)     (Cond | (0b000000<<22) | (A<<21) | (S<<20) | (Rd<<16) | (Rn<<12) | (Rs<<8) | (0b1001<<4) | (Rm))
+#define MUL(Rd, Rm, Rn)     EMIT(MULMULA(c__, 0, 0, (Rd), 0, (Rm), (Rn)))
+
+#define SMUL_16_gen(cond, Rd, Rm, M, N, Rn) (cond | 0b00010110<<20 | (Rd)<<16 | (Rm)<<8 | 1<<7 | (M)<<6 | (N)<<5 | (Rn))
+// Signed Mul between Rn[0..15] * Rm[0..15] => Rd
+#define SMULBB(Rd, Rn, Rm)  EMIT(SMUL_16_gen(c__, Rd, Rm, 0, 0, Rn))
+
+// SXTB rd, rm ror rot
+#define SXTB(rd, rm, rot)   EMIT(c__ | (0b01101010<<20) | (0x0f<<16) | ((rd)<<12) | (rot)<<10 | (0b0111<<4) | (rm))
+// UXTB rd, rm ror rot
+#define UXTB(rd, rm, rot)   EMIT(c__ | (0b01101110<<20) | (0x0f<<16) | ((rd)<<12) | (rot)<<10 | (0b0111<<4) | (rm))
+// SXTH rd, rm ror rot
+#define SXTH(rd, rm, rot)   EMIT(c__ | (0b01101011<<20) | (0x0f<<16) | ((rd)<<12) | (rot)<<10 | (0b0111<<4) | (rm))
+// UXTH rd, rm ror rot
+#define UXTH(rd, rm, rot)   EMIT(c__ | (0b01101111<<20) | (0x0f<<16) | ((rd)<<12) | (rot)<<10 | (0b0111<<4) | (rm))
+
+// UBFX: Unsigned Bit Field Extract: extract any number of bits from Rn, zero extend and put in Rd
+#define UBFX(rd, rn, lsb, width)    EMIT(c__ | (0b0111111<<21) | (((width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b101<<4) | (rn))
+
+// BFI: Bit Field Insert: copy any number of low order bit from Rn to any position of Rd
+#define BFI(rd, rn, lsb, width) EMIT(c__ | (0b0111110<<21) | (((lsb)+(width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b001<<4) | (rn))
+// BFI_COND: Bit Field Insert with condition: copy any number of low order bit from Rn to any position of Rd
+#define BFI_COND(cond, rd, rn, lsb, width) EMIT(cond | (0b0111110<<21) | (((lsb)+(width)-1)<<16) | ((rd)<<12) | ((lsb)<<7) | (0b001<<4) | (rn))
+
+// REV: Reverse byte of a 32bits word
+#define REV(rd, rm) EMIT(c__ | (0b01101<<23) | (0<<22) | (0b11<<20) | (0b1111<<16) | ((rd)<<12) | (0b1111<<8) | (0b0011<<4) | (rm))
+
+#define SSAT_gen(cond, sat_imm, Rd, imm5, sh, Rn) (cond | 0b0110101<<21 | (sat_imm)<<16 | (Rd)<<12 | (imm5)<<7 | (sh)<<6 | 0b01<<4 | (Rn))
+// Signed Staturate Rn to 2^(staturate_to-1) into Rd. Optionnaly shift left Rn before saturate
+#define SSAT_REG_LSL_IMM5(Rd, saturate_to, Rn, shift)   EMIT(SSAT_gen(c__, (saturate_to)-1, Rd, shift, 0, Rn))
+#define USAT_gen(cond, sat_imm, Rd, imm5, sh, Rn) (cond | 0b0110111<<21 | (sat_imm)<<16 | (Rd)<<12 | (imm5)<<7 | (sh)<<6 | 0b01<<4 | (Rn))
+// Unsigned Staturate Rn to 2^(staturate_to-1) into Rd. Optionnaly shift left Rn before saturate
+#define USAT_REG_LSL_IMM5(Rd, saturate_to, Rn, shift)   EMIT(SSAT_gen(c__, (saturate_to), Rd, shift, 0, Rn))
+
+#define MRS_gen(cond, Rd)   (cond | 0b00010<<23 | 0b1111<<16 | (Rd)<<12)
+// Load MRS ASPR flags
+#define MRS_aspr(Rd)    EMIT(MRS_gen(c__, Rd))
+
+#define MSR_gen(cond, mask, Rn) (cond | 0b00010<<23 | 0b10<<20 | (mask)<<18 | 0b1111<<12 | (Rn))
+// Store MRS ASPR nzcvq flags from Rn
+#define MSR_nzcvq(Rn)   EMIT(MSR_gen(c__, 0b10, Rn))
+#define MSR_imm_gen(cond, mask, imm12) (cond | 0b00110<<23 | 0b10<<20 | (mask)<<18 | 0b1111<<12 | (imm12))
+#define MSR_nzcvq_0()   EMIT(MSR_imm_gen(c__, 0b10, 0))
+
+#define LDREXD_gen(cond, Rn, Rt) (cond | 0b000<<25 | 0b11011<<20 | (Rn)<<16 | (Rt)<<12 | 0b1111<<8 | 0b1001<<4 | 0b1111)
+// Load Exclusive Rt/Rt+1 from Rn (tagging the memory)
+#define LDREXD(Rn, Rt)  EMIT(LDREXD_gen(c__, Rn, Rt))
+
+#define STREXD_gen(cond, Rd, Rn, Rt)   (cond | 0b000<<25 | 0b11010<<20 | (Rn)<<16 | (Rd)<<12 | 0b1111<<8 | 0b1001<<4 | (Rt))
+// Store Exclusive Rt/Rt+1 to Rn, with result in Rd if tag is ok (Rd!=Rn && Rd!=Rt && Rd!=Rt+1), Rd==1 if store failed
+#define STREXD(Rd, Rn, Rt)  EMIT(STREXD_gen(c__, Rd, Rn, Rt))
+
+// 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))
+// 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))
+
+#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
+#define PLD(Rn, Rm) EMIT(PLD_gen(1, 1, Rn, 0, 0, Rm))
+// Preload Cache Rn-Rm
+#define PLDn(Rn, Rm) EMIT(PLD_gen(0, 1, Rn, 0, 0, Rm))
+
+// VFPU
+#define TRANSFERT64(C, op) ((0b1100<<24) | (0b010<<21) | (0b101<<9) | ((C)<<8) | ((op)<<4))
+
+// Move from FPSCR to Arm register
+#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 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 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))
+
+// Move between Rt to Sm and Rt2 to Sm+1 (Sm cannot be 31!, rt and Rt2 can be the same)
+#define VMOVtoV_64(Sm, Rt, Rt2) EMIT(c__ | (0b1100<<24) | (0b010<<21) | (0<<20) | ((Rt2)<<16) | ((Rt)<<12) | (0b1010<<8) | (0b00<<6) | (1<<4) | ((Sm)&1)>>1 | ((Sm)&0b10000)<<1)
+// Move between Sm to Rt and Sm+1 to Rt2 (Sm cannot be 31!, Rt and Rt2 must be different)
+#define VMOVfrV_64(Rt, Rt2, Sm) EMIT(c__ | (0b1100<<24) | (0b010<<21) | (1<<20) | ((Rt2)<<16) | ((Rt)<<12) | (0b1010<<8) | (0b00<<6) | (1<<4) | ((Sm)&1)>>1 | ((Sm)&0b10000)<<1)
+
+// Move between Rt/Rt2 to Dm
+#define VMOVtoV_D(Dm, Rt, Rt2) EMIT(c__ | (0b1100<<24) | (0b010<<21) | (0<<20) | ((Rt2)<<16) | ((Rt)<<12) | (0b1011<<8) | (0b00<<6) | (1<<4) | ((Dm)&0b1111) | (((Dm)>>4)&1)<<5)
+// Move between Dm and Rt/Rt2 (Rt and Rt2 must be different)
+#define VMOVfrV_D(Rt, Rt2, Dm) EMIT(c__ | (0b1100<<24) | (0b010<<21) | (1<<20) | ((Rt2)<<16) | ((Rt)<<12) | (0b1011<<8) | (0b00<<6) | (1<<4) | ((Dm)&0b1111) | (((Dm)>>4)&1)<<5)
+
+// Move between Dd and Dm
+#define VMOV_64(Dd, Dm)     EMIT(c__ | (0b11101<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b01<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15))
+// Move between Sd and Sm
+#define VMOV_32(Sd, Sm)     EMIT(c__ | (0b11101<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0b01<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15))
+// Move with condition between Dd and Dm
+#define VMOVcond_64(cond, Dd, Dm)     EMIT(cond | (0b11101<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b01<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15))
+// Move with condition between Sd and Sm
+#define VMOVcond_32(cond, Sd, Sm)     EMIT(cond | (0b11101<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0b01<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15))
+// Move between Sd and Imm
+#define VMOV_i_32(Sd, Imm8)     EMIT(c__ | (0b11101<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Imm8)>>4)&0xf)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0b00<<6) | ((Imm8)&15))
+
+#define VMOVtoDx_gen(Vd, D, Rt, opc1, opc2)   (0b1110<<24 | 0<<23 | (opc1)<<21 | (Vd)<<16 | (Rt)<<12 | 0b1011<<8 | (D)<<7 | (opc2)<<5 | 1<<4)
+// Move between Rt and Dm[x]
+#define VMOVtoDx_32(Dm, x, Rt) EMIT(c__ | VMOVtoDx_gen((Dm)&15, ((Dm)>>4)&1, Rt, x, 0))
+// Move between Rt and Dm[x]
+#define VMOVtoDx_16(Dm, x, Rt) EMIT(c__ | VMOVtoDx_gen((Dm)&15, ((Dm)>>4)&1, Rt, (x)>>1, ((x)&1)<<1 | 1))
+// Move between Rt and Dm[x]
+#define VMOVtoDx_8(Dm, x, Rt)  EMIT(c__ | VMOVtoDx_gen((Dm)&15, ((Dm)>>4)&1, Rt, 2| (x)>>2, (x)&3))
+
+#define VMOVfrDx_gen(Vn, N, Rt, U, opc1, opc2) (0b1110<<24 | (U)<<23 | (opc1)<<21 | 1<<20 | (Vn)<<16 | (Rt)<<12 | 0b1011<<8 | (N)<<7 | (opc2)<<5 | 1<<4)
+// Move between Dn[x] and Rt
+#define VMOVfrDx_32(Rt, Dn, x) EMIT(c__ | VMOVfrDx_gen((Dn)&15, ((Dn)>>4)&1, Rt, 0, x, 0))
+// Move between Dn[x] and Rt
+#define VMOVfrDx_U16(Rt, Dn, x) EMIT(c__ | VMOVfrDx_gen((Dn)&15, ((Dn)>>4)&1, Rt, 1, (x)>>1, ((x)&1)<<1 | 1))
+// Move between Dn[x] and Rt
+#define VMOVfrDx_U8(Rt, Dn, x) EMIT(c__ | VMOVfrDx_gen((Dn)&15, ((Dn)>>4)&1, Rt, 1, 2| (x)>>2, (x)&3))
+// Move between Dn[x] and Rt
+#define VMOVfrDx_S16(Rt, Dn, x) EMIT(c__ | VMOVfrDx_gen((Dn)&15, ((Dn)>>4)&1, Rt, 0, (x)>>1, ((x)&1)<<1 | 1))
+// Move between Dn[x] and Rt
+#define VMOVfrDx_S8(Rt, Dn, x) EMIT(c__ | VMOVfrDx_gen((Dn)&15, ((Dn)>>4)&1, Rt, 0, 2| (x)>>2, (x)&3))
+
+// Load from memory to double  VLDR Dd, [Rn, #+/-imm8], imm8&3 ignored!
+#define VLDR_64(Dd, Rn, Imm8)    EMIT(c__ | (0b1101<<24) | (((Imm8)<0)?0:1)<<23 | ((((Dd)>>4)&1)<<22) | (1<<20) | ((Rn)<<16) | (((Dd)&15)<<12) | (0b1011<<8) | ((abs(Imm8)>>2)&255))
+// Load from memory to single  VLDR Sd, [Rn, #+/-imm8], imm8&3 ignored!
+#define VLDR_32(Sd, Rn, Imm8)    EMIT(c__ | (0b1101<<24) | (((Imm8)<0)?0:1)<<23 | (((Sd)&1)<<22) | (1<<20) | ((Rn)<<16) | ((((Sd)>>1)&15)<<12) | (0b1010<<8) | ((abs(Imm8)>>2)&255))
+
+// Store to memory to double  VSTR Dd, [Rn, #+/-imm8], imm8&3 ignored!
+#define VSTR_64(Dd, Rn, Imm8)    EMIT(c__ | (0b1101<<24) | (((Imm8)<0)?0:1)<<23 | ((((Dd)>>4)&1)<<22) | (0<<20) | ((Rn)<<16) | (((Dd)&15)<<12) | (0b1011<<8) | ((abs(Imm8)>>2)&255))
+// Store to memory to single  VSTR Sd, [Rn, #+/-imm8], imm8&3 ignored!
+#define VSTR_32(Sd, Rn, Imm8)    EMIT(c__ | (0b1101<<24) | (((Imm8)<0)?0:1)<<23 | (((Sd)&1)<<22) | (0<<20) | ((Rn)<<16) | ((((Sd)>>1)&15)<<12) | (0b1010<<8) | ((abs(Imm8)>>2)&255))
+
+// Convert from single Sm to double Dd
+#define VCVT_F64_F32(Dd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) |  ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0111<<16) | (((Dd)&15)<<12) | (0b101<<9) | (0<<8) | (0b11<<6) | (((Sm)&1)<<5) | (0<<4) | (((Sm)>>1)&15))
+// Convert from double Dm to single Sd
+#define VCVT_F32_F64(Sd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) |  (((Sd)&1)<<22) | (0b11<<20) | (0b0111<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (1<<8) | (0b11<<6) | ((((Dm)>>4)&1)<<5) | (0<<4) | ((Dm)&15))
+
+// Convert from double Dm to int32 Sd, with Round toward Zero mode
+#define VCVT_S32_F64(Sd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (1<<8) | (1<<7) | (1<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )
+// Convert from double Dm to int32 Sd, with Round selection from FPSCR
+#define VCVTR_S32_F64(Sd, Dm)   EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15) )
+// Convert from int32 Sm to double Dd
+#define VCVT_F64_S32(Dd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b111<<19) | (0b000<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (1<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+// Convert from single Sm to int32 Sd, with Round toward Zero mode
+#define VCVT_S32_F32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (1<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+// Convert from single Sm to int32 Sd, with Round selection from FPSCR
+#define VCVTR_S32_F32(Sd, Sm)   EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+// Convert from int32 Sm to single Dd
+#define VCVT_F32_S32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b111<<19) | (0b000<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (1<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+#define VCVT_16_gen(cond, D, op, U, Vd, sf, sx, i, imm4)   (cond | 0b1110<<24 | 1<<23 | (D)<<22 | 0b11<<20 | 1<<19 | (op)<<18 | 1<<17 | (U)<<16 | (Vd)<<12 | 0b101<<9 | (sf)<<8 | (sx)<<7 | 1<<6 | (i)<<5 | (imm4))
+// Inplace convert from F64 to S16. Rounding is not selectable
+#define VCVT_S16_F64(Dd)    EMIT(VCVT_16_gen(c__, ((Dd)>>4)&1, 1, 0, (Dd)&15, 1, 0, 16&1, (16>>1)&15))
+
+// Mutiply F64 Dd = Dn*Dm
+#define VMUL_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b10<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+
+// Divide F64 Dd = Dn/Dm
+#define VDIV_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b00<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+
+// Add F64 Dd = Dn + Dm
+#define VADD_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (0<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+
+// Sub F64 Dd = Dn + Dm
+#define VSUB_F64(Dd, Dn, Dm)    EMIT(c__ | (0b1110<<24) | (0<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dn)&15)<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (((Dn>>4)&1)<<7) | (1<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+
+// Mutiply F32 Sd = Sn*Sm
+#define VMUL_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (0<<23) | (((Sd)&1)<<22) | (0b10<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+// Divide F32 Sd = Sn/Sm
+#define VDIV_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b00<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+// ADD F32 Sd = Sn + Sm
+#define VADD_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (0<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (0<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+// Sub F32 Sd = Sn + Sm
+#define VSUB_F32(Sd, Sn, Sm)    EMIT(c__ | (0b1110<<24) | (0<<23) | (((Sd)&1)<<22) | (0b11<<20) | ((((Sn)>>1)&15)<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (((Sn)&1)<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+// Cmp between 2 double Dd and Dm
+#define VCMP_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0100<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+// Cmp between 1 double Dd and 0.0
+#define VCMP_F64_0(Dd)      EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0101<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0<<7) | (1<<6) | (0<<5) | (0) )
+// Cmp between 2 single Sd and Sm
+#define VCMP_F32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b11<<20) | (0b0100<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0<<7) | (1<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+// Cmp between 1 single Sd and 0.0
+#define VCMP_F32_0(Sd)      EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b11<<20) | (0b0101<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0<<7) | (1<<6) | (0<<5) | (0) )
+
+// Neg F64 Dd = - Dm
+#define VNEG_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b01<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+
+// Sqrt F64 Dd = - Dm
+#define VSQRT_F64(Dd, Dm)    EMIT(c__ | (0b1110<<24) | (1<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (0b0001<<16) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b11<<6) | (((Dm>>4)&1)<<5) | ((Dm)&15) )
+// Sqrt F32 Dd = - Dm
+#define VSQRT_F32(Sd, Sm)    EMIT(c__ | (0b1110<<24) | (1<<23) | (((Sd)&1)<<22) | (0b11<<20) | (0b0001<<16) | ((((Sd)>>1)&15)<<12) | (0b101<<9) | (0<<8) | (0b11<<6) | (((Sm)&1)<<5) | (((Sm)>>1)&15) )
+
+// Abs Dd = |Dm|
+#define VABS_F64(Dd, Dm)     EMIT(c__ | (0b11101<<23) | ((((Dd)>>4)&1)<<22) | (0b11<<20) | (((Dd)&15)<<12) | (0b101<<9) | (1<<8) | (0b11<<6) | ((((Dm)>>4)&1)<<5) | ((Dm)&15))
+
+// NEON
+
+// L is 1 for VLD1, 0 for VST1 Dd is V:Vd, type:0b0111=64, 0b1010=128, 0b0110=192, 0b0010=256, size:0=8,1=16,2=32,3=64, align:"4<<align", wback=rm!=15, reg_index:rm!=13&&rm!=15
+#define Vxx1gen(L, D, Rn, Vd, type, size, align, Rm) (0b1111<<28 | 0b0100<<24 | 0<<23 | (D)<<22 | (L)<<21 | 0<<20 | (Rn)<<16 | (Vd)<<12 | (type)<<8 | (size)<<6 | (align)<<4 | (Rm))
+// Load [Rn] => Dd/Dd+1. Align is 4
+#define VLD1Q_32(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, 15))
+// Load [Rn] => Dd. Align is 4
+#define VLD1_32(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 2, 0, 15))
+// Load [Rn]! => Dd. Align is 4
+#define VLD1_32_W(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 2, 0, 13))
+// Load [Rn]! => Dd/Dd+1. Align is 4
+#define VLD1Q_32_W(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, 13))
+// Load [Rn, Rm]! => Dd/Dd+1. If Rm==15, no writeback, Rm ignored, else writeback Rn <- Rn+Rm. Align is 4
+#define VLD1Q_32_REG(Dd, Rn, Rm) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, Rm))
+// Load [Rn] => Dd/Dd+1. Align is 4
+#define VLD1Q_8(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 0, 0, 15))
+// Load [Rn] => Dd/Dd+1. Align is 4
+#define VLD1Q_16(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 1, 0, 15))
+// Load [Rn] => Dd/Dd+1. Align is 4
+#define VLD1Q_64(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 3, 0, 15))
+// Load [Rn] => Dd. Align is 4
+#define VLD1_64(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 3, 0, 15))
+// Load [Rn] => Dd. Align is 4
+#define VLD1_32(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 2, 0, 15))
+// Load [Rn] => Dd. Align is 4
+#define VLD1_16(Dd, Rn) EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 1, 0, 15))
+// Load [Rn] => Dd. Align is 4
+#define VLD1_8(Dd, Rn)  EMIT(Vxx1gen(1, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 0, 0, 15))
+
+// Store [Rn]! => Dd. Align is 4
+#define VST1_32_W(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 2, 0, 13))
+// Store [Rn] => Dd/Dd+1. Align is 4
+#define VST1Q_32(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, 15))
+// Store [Rn]! => Dd/Dd+1. Align is 4
+#define VST1Q_32_W(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, 13))
+// Store [Rn, Rm]! => Dd/Dd+1. If Rm==15, no writeback, Rm ignored, else writeback Rn <- Rn+Rm. Align is 4
+#define VST1Q_32_REG(Dd, Rn, Rm) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 2, 0, Rm))
+// Store [Rn] => Dd/Dd+1. Align is 4
+#define VST1Q_8(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 0, 0, 15))
+// Store [Rn] => Dd/Dd+1. Align is 4
+#define VST1Q_16(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 1, 0, 15))
+// Store [Rn] => Dd/Dd+1. Align is 4
+#define VST1Q_64(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b1010, 3, 0, 15))
+// Store [Rn] => Dd. Align is 4
+#define VST1_64(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 3, 0, 15))
+// Store [Rn] => Dd. Align is 4
+#define VST1_32(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 2, 0, 15))
+// Store [Rn] => Dd. Align is 4
+#define VST1_16(Dd, Rn) EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 1, 0, 15))
+// Store [Rn] => Dd. Align is 4
+#define VST1_8(Dd, Rn)  EMIT(Vxx1gen(0, ((Dd)>>4)&1, Rn, ((Dd)&0x0f), 0b0111, 0, 0, 15))
+
+#define VEOR_gen(D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | 0b00<<20 | (Vn)<<16 | (Vd)<<12 | 0b0001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VEOR(Dd, Dn, Dm)    EMIT(VEOR_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VEORQ(Dd, Dn, Dm)   EMIT(VEOR_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMOVL_gen(U, D, imm3, Vd, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (imm3)<<19 | (Vd)<<12 | 0b1010<<8 | (M)<<5 | 1<<4 | (Vm))
+#define VMOVL_S32(Dd, Dm)   EMIT(VMOVL_gen(0, ((Dd)>>4)&1, 0b100, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+#define VMOVL_U32(Dd, Dm)   EMIT(VMOVL_gen(1, ((Dd)>>4)&1, 0b100, (Dd)&15, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMOV_gen(D,Vd, M, Vm, Q) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 0b10<<20 | (Vm)<<16 | (Vd)<<12 | 0b0001<<8 | (M)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VMOVD(Dd, Dm) EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 0))
+#define VMOVQ(Dd, Dm) EMIT(VMOV_gen(((Dd)>>4)&1, (Dd)&15, ((Dm)>>4)&1, (Dm)&15, 1))
+
+#define VLD1LANE_gen(D, Rn, Vd, size, index_align, Rm) (0b1111<<28 | 0b0100<<24 | 1<<23 | (D)<<22 | 0b10<<20 | (Rn)<<16 | (Vd)<<12 | (size)<<10 | (index_align)<<4 | (Rm))
+#define VLD1LANE_8(Dd, Rn, index)    EMIT(VLD1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 0, (index)<<1, 15))
+#define VLD1LANE_16(Dd, Rn, index)   EMIT(VLD1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 1, (index)<<2, 15))
+#define VLD1LANE_32(Dd, Rn, index)   EMIT(VLD1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 2, (index)<<3, 15))
+
+#define VLD1ALL_gen(D, Rn, Vd, size, T, a, Rm) (0b1111<<28 | 0b0100<<24 | 1<<23 | (D)<<22 | 0b10<<20 | (Rn)<<16 | (Vd)<<12 | 0b11<<10| (size)<<6 | (T)<<5 | (a)<<4 | (Rm))
+#define VLD1ALL_16(Dd, Rn)      EMIT(VLD1ALL_gen(((Dd)>>4)&1, Rn, (Dd)&15, 1, 0, 0, 15))
+#define VLD1ALL_32(Dd, Rn)      EMIT(VLD1ALL_gen(((Dd)>>4)&1, Rn, (Dd)&15, 2, 0, 0, 15))
+#define VLD1QALL_32(Dd, Rn)     EMIT(VLD1ALL_gen(((Dd)>>4)&1, Rn, (Dd)&15, 2, 1, 0, 15))
+
+#define VST1LANE_gen(D, Rn, Vd, size, index_align, Rm) (0b1111<<28 | 0b0100<<24 | 1<<23 | (D)<<22 | 0b00<<20 | (Rn)<<16 | (Vd)<<12 | (size)<<10 | (index_align)<<4 | (Rm))
+#define VST1LANE_8(Dd, Rn, index)    EMIT(VST1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 0, (index)<<1, 15))
+#define VST1LANE_16(Dd, Rn, index)   EMIT(VST1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 1, (index)<<2, 15))
+#define VST1LANE_32(Dd, Rn, index)   EMIT(VST1LANE_gen(((Dd)>>4)&1, Rn, (Dd)&15, 2, (index)<<3, 15))
+
+#define VADD_gen(size, D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VADD_8(Dd, Dn, Dm)     EMIT(VADD_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VADDQ_8(Dd, Dn, Dm)    EMIT(VADD_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VADD_16(Dd, Dn, Dm)    EMIT(VADD_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VADDQ_16(Dd, Dn, Dm)   EMIT(VADD_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VADD_32(Dd, Dn, Dm)    EMIT(VADD_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VADDQ_32(Dd, Dn, Dm)   EMIT(VADD_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VADD_64(Dd, Dn, Dm)    EMIT(VADD_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VADDQ_64(Dd, Dn, Dm)   EMIT(VADD_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSUB_gen(size, D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VSUB_8(Dd, Dn, Dm)     EMIT(VSUB_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSUBQ_8(Dd, Dn, Dm)    EMIT(VSUB_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSUB_16(Dd, Dn, Dm)    EMIT(VSUB_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSUBQ_16(Dd, Dn, Dm)   EMIT(VSUB_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSUB_32(Dd, Dn, Dm)    EMIT(VSUB_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSUBQ_32(Dd, Dn, Dm)   EMIT(VSUB_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSUB_64(Dd, Dn, Dm)    EMIT(VSUB_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSUBQ_64(Dd, Dn, Dm)   EMIT(VSUB_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VORR_gen(D,Vd, N, Vn, M, Vm, Q) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 0b10<<20 | (Vn)<<16 | (Vd)<<12 | 0b0001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VORRD(Dd, Dn, Dm) EMIT(VORR_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 0))
+#define VORRQ(Dd, Dn, Dm) EMIT(VORR_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 1))
+
+#define VAND_gen(D,Vd, N, Vn, M, Vm, Q) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 0b00<<20 | (Vn)<<16 | (Vd)<<12 | 0b0001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VANDD(Dd, Dn, Dm) EMIT(VAND_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 0))
+#define VANDQ(Dd, Dn, Dm) EMIT(VAND_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 1))
+
+#define VBIC_gen(D,Vd, N, Vn, M, Vm, Q) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | 0b01<<20 | (Vn)<<16 | (Vd)<<12 | 0b0001<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VBICD(Dd, Dn, Dm) EMIT(VBIC_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 0))
+#define VBICQ(Dd, Dn, Dm) EMIT(VBIC_gen(((Dd)>>4)&1, (Dd)&15, ((Dn)>>4)&1, (Dn)&15, ((Dm)>>4)&1, (Dm)&15, 1))
+
+#define VCEQ_I_gen(size, D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VCEQ_8(Dd, Dn, Dm)     EMIT(VCEQ_I_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_8(Dd, Dn, Dm)    EMIT(VCEQ_I_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_16(Dd, Dn, Dm)    EMIT(VCEQ_I_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_16(Dd, Dn, Dm)   EMIT(VCEQ_I_gen(1, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_32(Dd, Dn, Dm)    EMIT(VCEQ_I_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_32(Dd, Dn, Dm)   EMIT(VCEQ_I_gen(2, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_64(Dd, Dn, Dm)    EMIT(VCEQ_I_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_64(Dd, Dn, Dm)   EMIT(VCEQ_I_gen(3, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQ_F_gen(sz, D, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | (sz)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1110<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VCEQ_F32(Dd, Dn, Dm)    EMIT(VCEQ_F_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCEQQ_F32(Dd, Dn, Dm)   EMIT(VCEQ_F_gen(0, ((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VCGT_I_gen(U, D, size, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0011<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VCGT_U8(Dd, Dn, Dm)     EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_U8(Dd, Dn, Dm)    EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_U16(Dd, Dn, Dm)    EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_U16(Dd, Dn, Dm)   EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_U32(Dd, Dn, Dm)    EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_U32(Dd, Dn, Dm)   EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_U64(Dd, Dn, Dm)    EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 3, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_U64(Dd, Dn, Dm)   EMIT(VCGT_I_gen(1, ((Dd)>>4)&1, 3, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_S8(Dd, Dn, Dm)     EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_S8(Dd, Dn, Dm)    EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_S16(Dd, Dn, Dm)    EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_S16(Dd, Dn, Dm)   EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_S32(Dd, Dn, Dm)    EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_S32(Dd, Dn, Dm)   EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_S64(Dd, Dn, Dm)    EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 3, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_S64(Dd, Dn, Dm)   EMIT(VCGT_I_gen(0, ((Dd)>>4)&1, 3, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCGT_F_gen(D, sz, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | 1<<21 | (sz)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1110<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VCGT_F32(Dd, Dn, Dm)    EMIT(VCGT_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGTQ_F32(Dd, Dn, Dm)   EMIT(VCGT_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VCGE_F_gen(D, sz, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b0011<<24 | 0<<23 | (D)<<22 | 0<<21 | (sz)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1110<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VCGE_F32(Dd, Dn, Dm)    EMIT(VCGE_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCGEQ_F32(Dd, Dn, Dm)   EMIT(VCGE_F_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSHR_gen(U, D, imm6, Vd, L, Q, M, Vm) (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (imm6)<<16 | (Vd)<<12 | (L)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VSHR_U8(Dd, Dm, imm3)    EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_S8(Dd, Dm, imm3)    EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_U16(Dd, Dm, imm4)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b01<<4 | (16-(imm4)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_S16(Dd, Dm, imm4)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b01<<4 | (16-(imm4)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_U32(Dd, Dm, imm5)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b1<<5 | (32-(imm5)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_S32(Dd, Dm, imm5)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b1<<5 | (32-(imm5)), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_U64(Dd, Dm, imm6)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, (64-(imm6)), (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHR_S64(Dd, Dm, imm6)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, (64-(imm6)), (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_U8(Dd, Dm, imm3)    EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_S8(Dd, Dm, imm3)    EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b001<<3 | (8-(imm3)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_U16(Dd, Dm, imm4)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b01<<4 | (16-(imm4)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_S16(Dd, Dm, imm4)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b01<<4 | (16-(imm4)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_U32(Dd, Dm, imm5)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, 0b1<<5 | (32-(imm5)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_S32(Dd, Dm, imm5)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, 0b1<<5 | (32-(imm5)), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_U64(Dd, Dm, imm6)   EMIT(VSHR_gen(1, ((Dd)>>4)&1, (64-(imm6)), (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHRQ_S64(Dd, Dm, imm6)   EMIT(VSHR_gen(0, ((Dd)>>4)&1, (64-(imm6)), (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSHL_gen(D, imm6, Vd, L, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | 1<<23 | (D)<<22 | (imm6)<<16 | (Vd)<<12 | 0b0101<<8 | (L)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VSHL_8(Dd, Dm, imm3)    EMIT(VSHL_gen(((Dd)>>4)&1, 0b001<<3 | ((imm3)+8), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_16(Dd, Dm, imm4)   EMIT(VSHL_gen(((Dd)>>4)&1, 0b01<<4 | ((imm4)+16), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_32(Dd, Dm, imm5)   EMIT(VSHL_gen(((Dd)>>4)&1, 0b1<<5 | ((imm5)+32), (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_64(Dd, Dm, imm6)   EMIT(VSHL_gen(((Dd)>>4)&1, (imm6), (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_8(Dd, Dm, imm3)   EMIT(VSHL_gen(((Dd)>>4)&1, 0b001<<3 | ((imm3)+8), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_16(Dd, Dm, imm4)  EMIT(VSHL_gen(((Dd)>>4)&1, 0b01<<4 | ((imm4)+16), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_32(Dd, Dm, imm5)  EMIT(VSHL_gen(((Dd)>>4)&1, 0b1<<5 | ((imm5)+32), (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_64(Dd, Dm, imm6)  EMIT(VSHL_gen(((Dd)>>4)&1, (imm6), (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSHLR_gen(U, D, size, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b001<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0100<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VSHL_U8(Dd, Dm, Dn)     EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_U16(Dd, Dm, Dn)    EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_U32(Dd, Dm, Dn)    EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_U64(Dd, Dm, Dn)    EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_U8(Dd, Dm, Dn)    EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_U16(Dd, Dm, Dn)   EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_U32(Dd, Dm, Dn)   EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_U64(Dd, Dm, Dn)   EMIT(VSHLR_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_S8(Dd, Dm, Dn)     EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_S16(Dd, Dm, Dn)    EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_S32(Dd, Dm, Dn)    EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHL_S64(Dd, Dm, Dn)    EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_S8(Dd, Dm, Dn)    EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_S16(Dd, Dm, Dn)   EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_S32(Dd, Dm, Dn)   EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VSHLQ_S64(Dd, Dm, Dn)   EMIT(VSHLR_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VTRN_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0000<<8 | 1<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VTRN_32(Dd, Dm)     EMIT(VTRN_gen(((Dd)>>4)&1, 2, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VZIP_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0001<<8 | 1<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VZIP_8(Dd, Dm)     EMIT(VZIP_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VZIP_16(Dd, Dm)    EMIT(VZIP_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VZIP_32(Dd, Dm)    VTRN_32(Dd, Dm)
+#define VZIPQ_8(Dd, Dm)    EMIT(VZIP_gen(((Dd)>>4)&1, 0, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VZIPQ_16(Dd, Dm)   EMIT(VZIP_gen(((Dd)>>4)&1, 1, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VZIPQ_32(Dd, Dm)   EMIT(VZIP_gen(((Dd)>>4)&1, 2, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VUZP_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0001<<8 | 0<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VUZP_8(Dd, Dm)     EMIT(VUZP_gen(((Dd)>>4)&1, 0, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VUZP_16(Dd, Dm)    EMIT(VUZP_gen(((Dd)>>4)&1, 1, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VUZP_32(Dd, Dm)    VTRN_32(Dd, Dm)
+#define VUZPQ_8(Dd, Dm)    EMIT(VUZP_gen(((Dd)>>4)&1, 0, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VUZPQ_16(Dd, Dm)   EMIT(VUZP_gen(((Dd)>>4)&1, 1, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VUZPQ_32(Dd, Dm)   EMIT(VUZP_gen(((Dd)>>4)&1, 2, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VADD_F_gen(D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | (D)<<22 | 0<<20 | (Vn)<<16 | (Vd)<<12 | 0b1101<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VADD__F32(Dd, Dn, Dm)   EMIT(VADD_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VADDQ_F32(Dd, Dn, Dm)   EMIT(VADD_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSUB_F_gen(D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | (D)<<22 | 1<<21 | 0<<20 | (Vn)<<16 | (Vd)<<12 | 0b1101<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VSUB__F32(Dd, Dn, Dm)   EMIT(VSUB_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VSUBQ_F32(Dd, Dn, Dm)   EMIT(VSUB_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMUL_F_gen(D, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | (D)<<22 | 0<<20 | (Vn)<<16 | (Vd)<<12 | 0b1101<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+#define VMUL__F32(Dd, Dn, Dm)   EMIT(VMUL_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VMULQ_F32(Dd, Dn, Dm)   EMIT(VMUL_F_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VCVT_NEON_gen(D, size, Vd, op, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b11<<16 | (Vd)<<12 | 0b11<<9 | (op)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VCVTQ_F32_S32(Dd, Dm)   EMIT(VCVT_NEON_gen(((Dd)>>4)&1, 2, (Dd)&15, 0b00, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCVTQ_S32_F32(Dd, Dm)   EMIT(VCVT_NEON_gen(((Dd)>>4)&1, 2, (Dd)&15, 0b10, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VCVTn_F32_S32(Dd, Dm)   EMIT(VCVT_NEON_gen(((Dd)>>4)&1, 2, (Dd)&15, 0b00, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VCVTn_S32_F32(Dd, Dm)   EMIT(VCVT_NEON_gen(((Dd)>>4)&1, 2, (Dd)&15, 0b10, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMULL_NEON_gen(U, D, size, Vn, Vd, op, N, M, Vm)    (0b1111<<28 | 0b001<<25 | (U)<<24 | 1<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b11<<10 | (op)<<9 | (N)<<7 | (M)<<5 | (Vm))
+#define VMULL_S32_S16(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
+#define VMULL_U32_U16(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(1, ((Dd)>>4)&1, 1, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
+#define VMULL_S64_S32(Dd, Dn, Dm)   EMIT(VMULL_NEON_gen(0, ((Dd)>>4)&1, 2, (Dn)&15, (Dd)&15, 0, ((Dn)>>4)&1, ((Dm)>>4)&1, (Dm)&15))
+#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 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))
+#define VMULQ_16(Dd, Dn, Dm)     EMIT(VMUL_NEON_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VMUL_32(Dd, Dn, Dm)      EMIT(VMUL_NEON_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VMUL_16(Dd, Dn, Dm)      EMIT(VMUL_NEON_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VEXT_gen(D, Vn, Vd, imm4, N, Q, M, Vm)  (0b1111<<28 | 0b0010<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (Vn)<<16 | (Vd)<<12 | (imm4)<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VEXT_8(Dd, Dn, Dm, imm4)    EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, imm4, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VEXTQ_8(Dd, Dn, Dm, imm4)   EMIT(VEXT_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, imm4, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VMVN_gen(D, size, Vd, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | (Vd)<<12 | 0b1011<<7 | (Q)<<6 | (M)<<5 | 0<<4 | (Vm))
+#define VMVNQ(Dd, Dm)       EMIT(VMVN_gen(((Dd)>>4)&1, 0, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VQMOVN_gen(D, size, Vd, op, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0010<<8 | (op)<<6 | (M)<<5 | (Vm))
+// Vector Saturating Move and Narrow, for 16 bits signed integer (Dm is Qm in fact)
+#define VQMOVN_S16(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0b10, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 32 bits signed integer (Dm is Qm in fact)
+#define VQMOVN_S32(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0b10, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 64 bits signed integer (Dm is Qm in fact)
+#define VQMOVN_S64(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0b10, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 16 bits signed integer (Dm is Qm in fact)
+#define VQMOVUN_S16(Dd, Dm) EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0b01, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 32 bits signed integer (Dm is Qm in fact)
+#define VQMOVUN_S32(Dd, Dm) EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0b01, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 64 bits signed integer (Dm is Qm in fact)
+#define VQMOVUN_S64(Dd, Dm) EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0b01, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 16 bits unsigned integer (Dm is Qm in fact)
+#define VQMOVN_U16(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0b11, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 32 bits signed integer (Dm is Qm in fact)
+#define VQMOVN_U32(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0b11, ((Dm)>>4)&1, (Dm)&15))
+// Vector Saturating Move and Narrow, for 64 bits signed integer (Dm is Qm in fact)
+#define VQMOVN_U64(Dd, Dm)  EMIT(VQMOVN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0b11, ((Dm)>>4)&1, (Dm)&15))
+
+#define VTBL_gen(D, Vn, Vd, len, N, op, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (Vn)<<16 | (Vd)<<12 | 0b10<<10 | (len)<<8 | (N)<<7 | (op)<<6 | (M)<<5 | (Vm))
+// Vector Table lookup, take elements from Dn, and using lookup table Dm, put elements in Dd
+#define VTBL1_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b00, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Vector Table lookup, take elements from Dn/Dn+1, and using lookup table Dm, put elements in Dd
+#define VTBL2_8(Dd, Dn, Dm) EMIT(VTBL_gen(((Dd)>>4)&1, (Dn)&15, (Dd)&15, 0b01, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VRECPE_gen(D, size, Vd, F, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b11<<16 | (Vd)<<12 | 0b010<<9 | (F)<<8 | (Q)<<6 | (M)<<5 | (Vm))
+// Vector Reciprocal Estimate of Dm to Dd
+#define VRECPEQ_F32(Dd, Dm) EMIT(VRECPE_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VRECPS_gen(D, size, Vn, Vd, N, Q, M, Vm) (0b1111<<28 | 0b0010<<24 | 0<<23 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b1111<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+// Vector Reciprocal Step of Dn and Dm to Dd
+#define VRECPSQ_F32(Dd, Dn, Dm) EMIT(VRECPS_gen(((Dd)>>4)&1, 0, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#define VSWAP_gen(D, size, Vd, Q, M, Vm)    (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<20 | (size)<<18 | 0b10<<16 | (Vd)<<12 | 0b0000<<7 | (Q)<<6 | (M)<<5 | (Vm))
+// Swap Double vectors content
+#define VSWAP(Dd, Dm)   EMIT(VSWAP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, ((Dm)>>4)&1, (Dm)&15)))
+// Swap Quad vectors content
+#define VSWAPQ(Dd, Dm)  EMIT(VSWAP_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 1, ((Dm)>>4)&1, (Dm)&15)))
+
+#define VQADD_gen(U, D, size, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b001<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12  | 0b0000<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+// Add with saturation unsigned 8bits
+#define VQADDQ_U8(Dd, Dn, Dm)   EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 16bits
+#define VQADDQ_U16(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 32bits
+#define VQADDQ_U32(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 64bits
+#define VQADDQ_U64(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 8bits
+#define VQADDQ_S8(Dd, Dn, Dm)   EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 16bits
+#define VQADDQ_S16(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 32bits
+#define VQADDQ_S32(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 64bits
+#define VQADDQ_S64(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 8bits
+#define VQADD_U8(Dd, Dn, Dm)   EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 16bits
+#define VQADD_U16(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 32bits
+#define VQADD_U32(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation unsigned 64bits
+#define VQADD_U64(Dd, Dn, Dm)  EMIT(VQADD_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 8bits
+#define VQADD_S8(Dd, Dn, Dm)   EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 16bits
+#define VQADD_S16(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 32bits
+#define VQADD_S32(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Add with saturation signed 64bits
+#define VQADD_S64(Dd, Dn, Dm)  EMIT(VQADD_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VQSUB_gen(U, D, size, Vn, Vd, N, Q, M, Vm)  (0b1111<<28 | 0b001<<25 | (U)<<24 | (D)<<22 | (size)<<20 | (Vn)<<16 | (Vd)<<12 | 0b0010<<8 | (N)<<7 | (Q)<<6 | (M)<<5 | 1<<4 | (Vm))
+// Substract with saturation unsigned 8bits
+#define VQSUBQ_U8(Dd, Dn, Dm)   EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 16bits
+#define VQSUBQ_U16(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 32bits
+#define VQSUBQ_U32(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 64bits
+#define VQSUBQ_U64(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 8bits
+#define VQSUBQ_S8(Dd, Dn, Dm)   EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 16bits
+#define VQSUBQ_S16(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 32bits
+#define VQSUBQ_S32(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 64bits
+#define VQSUBQ_S64(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 1, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 8bits
+#define VQSUB_U8(Dd, Dn, Dm)   EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 16bits
+#define VQSUB_U16(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 32bits
+#define VQSUB_U32(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation unsigned 64bits
+#define VQSUB_U64(Dd, Dn, Dm)  EMIT(VQSUB_gen(1, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 8bits
+#define VQSUB_S8(Dd, Dn, Dm)   EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b00, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 16bits
+#define VQSUB_S16(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b01, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 32bits
+#define VQSUB_S32(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b10, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+// Substract with saturation signed 64bits
+#define VQSUB_S64(Dd, Dn, Dm)  EMIT(VQSUB_gen(0, ((Dd)>>4)&1, 0b11, (Dn)&15, (Dd)&15, ((Dn)>>4)&1, 0, ((Dm)>>4)&1, (Dm)&15))
+
+#define VNEGN_gen(D, size, Vd, F, Q, M, Vm) (0b1111<<28 | 0b0011<<24 | 1<<23 | (D)<<22 | 0b11<<20 | (size)<<18 | 0b01<<16 | (Vd)<<12 | (F)<<10 | 0b111<<7 | (Q)<<6 | (M)<<5 | (Vm))
+#define VNEGN_8(Dd, Dm)     EMIT(VNEGN_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGN_16(Dd, Dm)    EMIT(VNEGN_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGN_32(Dd, Dm)    EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 0, ((Dm)>>4)&1, (Dm)&15))
+// no VNEGN_64...
+#define VNEGNQ_8(Dd, Dm)    EMIT(VNEGN_gen(((Dd)>>4)&1, 0b00, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGNQ_16(Dd, Dm)   EMIT(VNEGN_gen(((Dd)>>4)&1, 0b01, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGNQ_32(Dd, Dm)   EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 0, 1, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGN_F32(Dd, Dm)   EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 0, ((Dm)>>4)&1, (Dm)&15))
+#define VNEGNQ_F32(Dd, Dm)  EMIT(VNEGN_gen(((Dd)>>4)&1, 0b10, (Dd)&15, 1, 1, ((Dm)>>4)&1, (Dm)&15))
+
+#endif  //__ARM_EMITTER_H__

src/dynarec/arm_epilog.S

@@ -0,0 +1,18 @@
+//arm epilog for dynarec
+//Save stuff, prepare stack and register
+//called with pointer to emu as 1st parameter
+//and address to jump to as 2nd parameter
+
+.text
+.align 4
+
+.global arm_epilog
+arm_epilog:
+    //update register -> emu
+    //pop     {r0}
+    stm     r0, {r4-r12}   // put back reg value in emu, including EIP (so r12 must be EIP now)
+    //restore all used register
+    //vpop     {d8-d15}
+    pop     {r4-r11, pc}
+    //end, return...
+    //bx      lr

src/dynarec/arm_linker.S

@@ -0,0 +1,24 @@
+//arm update linker table for dynarec
+//called with pointer to emu as 1st parameter
+//and address of table to as 2nd parameter
+//ip is at r12
+
+.text
+.align 4
+
+.extern UpdateLinkTable
+
+.global arm_linker
+arm_linker:
+    // emu is r0
+    // table offset is r1
+    // IP address is r12
+    push    {r0, r1, r12}
+    mov     r2, r12
+    // call the function
+    bl      UpdateLinkTable
+    // return offset is jump address
+    mov     r3, r0
+    pop     {r0, r1, r12}
+    bx      r3
+

src/dynarec/arm_printer.c

@@ -0,0 +1,923 @@
+#include <stdint.h>
+#include <stddef.h>
+#include <string.h>
+#include <stdio.h>
+
+#include "arm_printer.h"
+
+// conditions
+#define cEQ (0b0000<<28)
+#define cNE (0b0001<<28)
+#define cCS (0b0010<<28)
+#define cHS cCS
+#define cCC (0b0011<<28)
+#define cLO cCC
+#define cMI (0b0100<<28)
+#define cPL (0b0101<<28)
+#define cVS (0b0110<<28)
+#define cVC (0b0111<<28)
+#define cHI (0b1000<<28)
+#define cLS (0b1001<<28)
+#define cGE (0b1010<<28)
+#define cLT (0b1011<<28)
+#define cGT (0b1100<<28)
+#define cLE (0b1101<<28)
+#define c__ (0b1110<<28)    // means all
+
+static const char* conds[] = {
+    "EQ", "NE", "CS", "CC", "MI", "PL",
+    "VS", "VC", "HI", "LS", "GE", "LT",
+    "GT", "LE", ""
+};
+
+static const char* opcodes[] = {
+    "AND", "EOR", "SUB", "RSB", "ADD", "ADC",
+    "SBC", "RSC", "TST", "TEQ", "CMP", "CMN",
+    "ORR", "MOV", "BIC", "MVN"
+};
+
+static const char* blocktransfert_other[] = {
+    "STMDA", "STMIA", "STMDB", "STMIB",
+    "LDMDA", "LDMIA", "LDMDB", "LDMIB"
+};
+static const char* blocktransfert_stack[] = {
+    "STMED", "STMEA", "STMFD", "STMFA",
+    "LDMFA", "LDMFD", "LDMEA", "LDMED"
+};
+static const char* regname[] = {
+    "r0", "r1", "r2", "r3", "r4",
+    "r5", "r6", "r7", "r8", "r9",
+    "r10", "r11", "r12", "sp", "lr", "pc"
+};
+
+static const char* shift_type[] = {
+    "lsl", "lsr", "asr", "ror"
+};
+
+const char* print_shift(int shift, int comma) {
+    static char ret[20];
+    memset(ret, 0, sizeof(ret));
+    if (shift) {
+        int sh_op = (shift>>1)&3;
+        if(shift&1) {
+            int rs = shift>>4;
+            sprintf(ret, "%s%s %s", comma?",":"", shift_type[sh_op], regname[rs]);
+        } else {
+            uint8_t amount = shift>>3;
+            sprintf(ret, "%s%s #%u", comma?",":"", shift_type[sh_op], amount);
+        }
+    }
+    return ret;    
+}
+
+//  nop
+#define NOP     (0xe1a00000)
+
+const char* arm_print(uint32_t opcode)
+{
+    static __thread char ret[100];
+    memset(ret, 0, sizeof(ret));
+    if((opcode & (0b1111<<28))==(0b1111<<28)) {
+        // NEON?
+        if(((opcode>>24)&0b1111)==0b0100 && ((opcode>>20)&0b1001)==0b0000)
+        {
+            // VLD1 / VST1
+            int D = (opcode>>22)&1;
+            int Rn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int type = (opcode>>8)&15;
+            int size = (opcode>>6)&3;
+            int align = (opcode>>4)&3;
+            int l = (opcode>>21)&1;
+            int Rm = opcode&15;
+            int regs = 0;
+            int Dd = (D<<4) | Vd;
+            char reglist[50] = {0};
+            switch (type) {
+                case 0b0111: regs = 1; sprintf(reglist, "D%d", Dd); break;
+                case 0b1010: regs = 2; sprintf(reglist, "D%d, D%d", Dd, Dd+1); break;
+                case 0b0110: regs = 3; sprintf(reglist, "D%d, D%d, D%d", Dd, Dd+1, Dd+2); break;
+                case 0b0010: regs = 4; sprintf(reglist, "D%d, D%d, D%d, D%d", Dd, Dd+1, Dd+2, Dd+3); break;
+            }
+            char salign[10] = {0};
+            if (align) sprintf(salign, ":%d", 4<<align);
+            char wback[10] = {0};
+            if(Rm!=15) {
+                if(Rm==13) strcpy(wback, "!");
+                else sprintf(wback, ", %s", regname[Rm]);
+            }
+            if (!regs)
+                strcpy(ret, "VLD1?????");
+            else
+                sprintf(ret, "V%s1%s.%d {%s}, [%s%s]%s", l?"LD":"ST", (regs>2)?"q":"", 8<<size, reglist, regname[Rn], salign, wback);
+        } else
+        if(((opcode>>24)&0b1111)==0b0011 && ((opcode>>20)&0b1011)==0b0000 && ((opcode>>8)&0b1111)==0b0001 && ((opcode>>4)&1)==1)
+        {
+            // VEOR
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VEORQ Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VEOR D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1000)==0b0000 && ((opcode>>8)&0b1111)==0b1000 && ((opcode>>4)&1)==0)
+        {
+            // VADD
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = (opcode>>20)&3;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VADDQ.%d Q%d, Q%d, Q%d", 8<<size, Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VADD.%d D%d, D%d, D%d", 8<<size, Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0011 && ((opcode>>20)&0b1000)==0b0000 && ((opcode>>8)&0b1111)==0b1000 && ((opcode>>4)&1)==1)
+        {
+            // VCEQ
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = (opcode>>20)&3;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VCEQQ.%d Q%d, Q%d, Q%d", 8<<size, Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VCEQ.%d D%d, D%d, D%d", 8<<size, Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1011)==0b0010 && ((opcode>>8)&0b1111)==0b0001 && ((opcode>>4)&1)==1)
+        {
+            // VORR / VMOV (simd)
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if(Vn!=Vm) {
+                if (Q)
+                    sprintf(ret, "VORRQ Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+                else
+                    sprintf(ret, "VORR D%d, D%d, D%d", Dd, Dn, Dm);
+            } else {
+                if (Q)
+                    sprintf(ret, "VMOVQ Q%d, Q%d", Dd/2, Dm/2);
+                else
+                    sprintf(ret, "VMOV D%d, D%d", Dd, Dm);
+            }
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1011)==0b0000 && ((opcode>>8)&0b1111)==0b0001 && ((opcode>>4)&1)==1)
+        {
+            // VAND
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VANDQ Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VAND D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1011)==0b0001 && ((opcode>>8)&0b1111)==0b0001 && ((opcode>>4)&1)==1)
+        {
+            // VBIC
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VBICQ Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VBIC D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0100 && ((opcode>>20)&0b1011)==0b1000 && ((opcode>>8)&0b0011)==0b0000)
+        {
+            // VST1LANE
+            int D = (opcode>>22)&1;
+            int Rn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int size = (opcode>>10)&3;
+            int index_align = (opcode>>4)&15;
+            int Rm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int rsize = 32;
+            char ralign[30] = {0};
+            int index = 0;
+            int alignment = 1;
+            switch(size) {
+                case 0:
+                    rsize = 8;
+                    index = (index_align>>1)&7;
+                    break;
+                case 1:
+                    rsize = 16;
+                    index = (index_align>>2)&3;
+                    alignment += (index_align&1);
+                    break;
+                case 2:
+                    rsize = 32;
+                    index = (index_align>>3)&1;
+                    if((index_align&3)!=0)
+                        alignment = 4;
+                    break;
+            }
+            if(alignment!=1)
+                sprintf(ret, ":%d", 4<<alignment);
+            char post[40] = {0};
+            if(Rm!=15) {
+                if (Rm==13)
+                    strcpy(post, "!");
+                else
+                    sprintf(post, ", %s", regname[Rm]);
+            }
+            sprintf(ret, "VST1.%d {D%d[%d]}, [%s%s]%s", rsize, Dd, index, regname[Rn], ralign, post);
+        } else
+        if(((opcode>>24)&0b1111)==0b0100 && ((opcode>>20)&0b1011)==0b1010 && ((opcode>>8)&0b0011)==0b0000)
+        {
+            // VLD1LANE
+            int D = (opcode>>22)&1;
+            int Rn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int size = (opcode>>10)&3;
+            int index_align = (opcode>>4)&15;
+            int Rm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int rsize = 32;
+            char ralign[30] = {0};
+            int index = 0;
+            int T = 0;
+            int alignment = 1;
+            char list[40] = {0};
+            switch(size) {
+                case 0:
+                    rsize = 8;
+                    index = (index_align>>1)&7;
+                    sprintf(list, "{D%d[%d]}", Dd, index);
+                    break;
+                case 1:
+                    rsize = 16;
+                    index = (index_align>>2)&3;
+                    alignment += (index_align&1);
+                    sprintf(list, "{D%d[%d]}", Dd, index);
+                    break;
+                case 2:
+                    rsize = 32;
+                    index = (index_align>>3)&1;
+                    if((index_align&3)!=0)
+                        alignment = 4;
+                    sprintf(list, "{D%d[%d]}", Dd, index);
+                    break;
+                case 3:
+                    rsize = 8 <<((index_align>>2)&3);
+                    index = -1;
+                    T = (index_align>>1)&1;
+                    if(index_align&1)
+                        alignment = (index_align>>2)&3;
+                    if(T)
+                        sprintf(list, "{D%d[], D%d[]}", Dd, Dd+1);
+                    else
+                        sprintf(list, "{D%d[]}", Dd);
+                    break;
+            }
+            if(alignment!=1)
+                sprintf(ret, ":%d", 4<<alignment);
+            char post[40] = {0};
+            if(Rm!=15) {
+                if (Rm==13)
+                    strcpy(post, "!");
+                else
+                    sprintf(post, ", %s", regname[Rm]);
+            }
+            sprintf(ret, "VLD1.%d %s, [%s%s]%s", rsize, list, regname[Rn], ralign, post);
+        } else
+        if(((opcode>>24)&0b1110)==0b0010 && ((opcode>>20)&0b1000)==0b1000 && ((opcode>>16)&0b0111)==0b0000 && ((opcode>>8)&0b1111)==0b1010 && ((opcode>>4)&1)==1)
+        {
+            // VMOVL
+            int D = (opcode>>22)&1;
+            int U = (opcode>>24)&1;
+            int imm3 = (opcode>>19)&7;
+            int Vd = (opcode>>12)&15;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int Dd = (D<<4) | Vd;
+            int Dm = (M<<4) | Vm;
+            sprintf(ret, "VMOVL.%s%d Q%d, D%d", U?"U":"S", 4<<imm3,Dd/2, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1011)==0b0000 && ((opcode>>8)&0b1111)==0b1101 && ((opcode>>4)&1)==0)
+        {
+            // VADD_F
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = 0;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VADD.F32 Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VADD.F32 D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0010 && ((opcode>>20)&0b1011)==0b0010 && ((opcode>>8)&0b1111)==0b1101 && ((opcode>>4)&1)==0)
+        {
+            // VSUB_F
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = 0;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VSUB.F32 Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VSUB.F32 D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0011 && ((opcode>>20)&0b1011)==0b0000 && ((opcode>>8)&0b1111)==0b1101 && ((opcode>>4)&1)==1)
+        {
+            // VMUL_F
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = 0;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VMUL.F32 Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VMUL.F32 D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        if(((opcode>>24)&0b1111)==0b0011 && ((opcode>>20)&0b1011)==0b0010 && ((opcode>>8)&0b1111)==0b1110 && ((opcode>>4)&1)==0)
+        {
+            // VCGT.F32
+            int D = (opcode>>22)&1;
+            int Vn = (opcode>>16)&15;
+            int Vd = (opcode>>12)&15;
+            int N = (opcode>>7)&1;
+            int Q = (opcode>>6)&1;
+            int M = (opcode>>5)&1;
+            int Vm = (opcode)&15;
+            int size = 0;
+            int Dd = (D<<4) | Vd;
+            int Dn = (N<<4) | Vn;
+            int Dm = (M<<4) | Vm;
+            if (Q)
+                sprintf(ret, "VCGT.F32 Q%d, Q%d, Q%d", Dd/2, Dn/2, Dm/2);
+            else
+                sprintf(ret, "VCGT.F32 D%d, D%d, D%d", Dd, Dn, Dm);
+        } else
+        strcpy(ret, "?????");
+    } else {
+        const char* cond = conds[(opcode>>28)&15];
+        if((opcode&0b00001111111111111111111111010000)==0b00000001001011111111111100010000) {
+            int l = (opcode>>5)&1;
+            sprintf(ret, "B%sX%s r%d", l?"L":"", cond, opcode&0b1111);
+        } else if (((opcode>>25)&0b111)==0b101) {
+            int32_t offset = opcode&0xffffff;
+            if(opcode&0x800000) offset |= 0xff000000;
+            sprintf(ret, "B%s%s %+d", cond, (opcode&(1<<24))?"L":"", offset+2);
+        } else if((opcode&0b00001111110000000000000010010000)==0b00000000000000000000000010010000) {
+            int a = (opcode>>21)&1;
+            int s = (opcode>>20)&1;
+            int rd = (opcode>>16)&15;
+            int rn = (opcode>>12)&15;
+            int rs = (opcode>>8)&15;
+            int rm = (opcode)&15;
+            if(a) {
+                sprintf(ret, "MLA%s%s %s, %s, %s, %s", cond, s?"S":"", regname[rd], regname[rm], regname[rs], regname[rn]);
+            } else {
+                sprintf(ret, "MUL%s%s %s, %s, %s", cond, s?"S":"", regname[rd], regname[rm], regname[rs]);
+            }
+        } else if((opcode&0b00001111100000000000000010010000)==0b00000000100000000000000010010000) {
+            int u = (opcode>>22)&1;
+            int a = (opcode>>21)&1;
+            int s = (opcode>>20)&1;
+            int rdhi = (opcode>>16)&15;
+            int rdlo = (opcode>>12)&15;
+            int rs = (opcode>>8)&15;
+            int rm = (opcode)&15;
+            if(a) {
+                sprintf(ret, "%sMLAL%s%s %s, %s, %s, %s", u?"S":"U", cond, s?"S":"", regname[rdlo], regname[rdhi], regname[rm], regname[rs]);
+            } else {
+                sprintf(ret, "%sMULL%s%s %s, %s, %s, %s", u?"S":"U", cond, s?"S":"", regname[rdlo], regname[rdhi], regname[rm], regname[rs]);
+            }
+        } else if((opcode&0b00001111101100000000111111110000)==0b00000001000000000000000010010000) {
+            int b = (opcode>>22)&1;
+            int rn = (opcode>>16)&15;
+            int rd = (opcode>>12)&15;
+            int rm = (opcode)&15;
+            sprintf(ret, "SWP%s%s r%d, r%d, [r%d]", cond, b?"B":"", rd, rm, rn);
+        } else if((opcode&0b000011111111111110000111111110000)==0b00000110101111110000111100110000) {
+            int rd = (opcode>>12)&15;
+            int rm = (opcode)&15;
+            sprintf(ret, "BSWAP%s %s, %s", cond, regname[rd], regname[rm]);
+        } else if((opcode&0b00001111111100000000000000000000)==0b00000011000000000000000000000000) {
+            uint16_t imm16 = opcode&0x0fff | ((opcode>>4)&0xf000);
+            int rn = (opcode>>12)&15;
+            sprintf(ret, "MOVW%s %s, #0x%x", cond, regname[rn], imm16);
+        } else if((opcode&0b00001111111100000000000000000000)==0b00000011010000000000000000000000) {
+            uint16_t imm16 = opcode&0x0fff | ((opcode>>4)&0xf000);
+            int rn = (opcode>>12)&15;
+            sprintf(ret, "MOVT%s %s, #0x%x", cond, regname[rn], imm16);
+        } else
+        {
+        uint32_t cat = (opcode>>25)&0b111;
+            switch (cat) {
+                case 0b000:
+                    // many things are in here, but Branches are already printed, so only exchange and hlf data tranfert are left
+                    if(((opcode>>5)&3)!=0 && ((opcode>>4)&0b1001)==0b1001) {
+                        int p = (opcode>>24)&1;
+                        int u = (opcode>>23)&1;
+                        int o = (opcode>>22)&1;
+                        int w = (opcode>>21)&1;
+                        int l = (opcode>>20)&1;
+                        int rn = (opcode>>16)&15;
+                        int rd = (opcode>>12)&15;
+                        int offset = ((opcode>>4)&0xf0) | (opcode&0xf);
+                        int rm = opcode&15;
+                        int sh = (opcode>>5)&3;
+                        const char* shs[] = {"swp", "H", "SB", "SH"};
+                        char op2[40];
+                        if(o) {
+                            sprintf(op2, "#%d", u?offset:-offset);
+                        } else {
+                            sprintf(op2, "%s%s", u?"":"-", regname[rm]);
+                        }
+                        char addr[50];
+                        if(p) { // pre-index
+                            if(o && !offset) {
+                                sprintf(addr, "[%s]", regname[rn]);
+                            } else {
+                                sprintf(addr, "[%s, %s]%s", regname[rn], op2, w?"!":"");
+                            }
+                        } else {
+                            sprintf(addr, "[%s], %s", regname[rn], op2);
+                        }
+                        sprintf(ret, "%s%s%s%s %s, %s", l?"LDR":"STR", cond, shs[sh], (w && p)?"T":"", regname[rd], addr);
+                        break;
+                    } else
+                    if((((opcode>>25)&0b111)==0b000) && (((opcode>>20)&1)==0) && ((opcode>>4)&0b1101)==0b1101)
+                    {
+                        //STRD / LDRD
+                        int p = (opcode>>24)&1;
+                        int u = (opcode>>23)&1;
+                        int o = (opcode>>22)&1;
+                        int w = (opcode>>21)&1;
+                        int rn = (opcode>>16)&15;
+                        int rt = (opcode>>12)&15;
+                        int imm4H = (opcode>>8)&15;
+                        int imm4L = opcode&15;
+                        int rm = opcode&15;
+                        int s = (opcode>>5)&1;
+                        int imm8 = (imm4H<<4) | imm4L;
+                        char op2[40];
+                        if(o) {
+                            sprintf(op2, "#%d", u?imm8:-imm8);
+                        } else {
+                            sprintf(op2, "%s%s", u?"":"-", regname[rm]);
+                        }
+                        char addr[50];
+                        if(p) { // pre-index
+                            if(o && !imm8) {
+                                sprintf(addr, "[%s]", regname[rn]);
+                            } else {
+                                sprintf(addr, "[%s, %s]%s", regname[rn], op2, w?"!":"");
+                            }
+                        } else {
+                            sprintf(addr, "[%s], %s", regname[rn], op2);
+                        }
+                        sprintf("%s%sD%s %s, %s, %s", (rt&1)?"!!":"", s?"STR":"LDR", cond, regname[rt], regname[rt+1], addr);
+                    }
+
+                case 0b001:
+                     // data operation
+                    {
+                        int i = (opcode>>25)&1;
+                        int s = (opcode>>20)&1;
+                        int op = (opcode>>21)&15;
+                        int rn = (opcode>>16)&15;
+                        int rd = (opcode>>12)&15;
+                        int op2 = opcode&0xfff;
+                        char tmp[40] = {0};
+                        if(i) { // op2 is immediate
+                            uint8_t imm = op2&255;
+                            int rot = (op2>>8)&0xf;
+                            sprintf(tmp, "#%d", imm<<(rot*2));
+                        } else {
+                            int rm = op2&15;
+                            int shift = op2>>4;
+                            sprintf(tmp, "%s%s", regname[rm], print_shift(shift, 1));
+                        }
+                        if(op==0b1101 || op==0b1111) {  // MOV, MVN
+                            sprintf(ret, "%s%s%s %s, %s", opcodes[op], cond, s?"S":"", regname[rd], tmp);
+                        } else if(op==0b1000 || op==0b1001 || op==0b1010 || op==0b1011) {
+                            // TST, TEQ, CMP, CMN
+                            sprintf(ret, "%s%s %s, %s", opcodes[op], cond, regname[rn], tmp);
+                        } else {
+                            sprintf(ret, "%s%s%s %s, %s, %s", opcodes[op], cond, s?"S":"", regname[rd], regname[rn], tmp);
+                        }
+                    }
+                    break;
+                case 0b100:
+                    // block data transfert
+                    {
+                        int p = (opcode>>24)&1;
+                        int u = (opcode>>23)&1;
+                        int s = (opcode>>22)&1;
+                        int w = (opcode>>21)&1;
+                        int l = (opcode>>20)&1;
+                        int rn = (opcode>>16)&15;
+                        int list = opcode&0b111111111111111;
+                        int op = (l<<2)|(p<<1)|(u);
+                        sprintf(ret, "%s %s%s,{", (rn==13)?blocktransfert_stack[op]:blocktransfert_other[op], regname[rn], w?"!":"");
+                        int last=-2;
+                        int cnt = 0;
+                        char tmp[5];
+                        for (int i=0; i<16; ++i) {
+                            if(list&(1<<i)) { // present
+                                if(last==i-1) {
+                                    last = i;
+                                    ++cnt;
+                                } else {
+                                    sprintf(tmp, "r%d", i);
+                                    if(last>=0) strcat(ret, ",");
+                                    strcat(ret, tmp);
+                                    last=i;
+                                    cnt = 0;
+                                }
+                            } else {
+                                if(last==i-1) {
+                                    if(cnt) {
+                                        sprintf(tmp, "-r%d", last);
+                                        strcat(ret, tmp);
+                                    }
+                                    last=-2;
+                                }
+                            }
+                        }
+                        if(last==15) {
+                            sprintf(tmp, "-r%d", last);
+                            strcat(ret, tmp);
+                        }
+                        strcat(ret, "}");
+                        if(s)
+                            strcat(ret, "^");
+                    }
+                    break;
+                case 0b010:
+                case 0b011:
+                    if(cat==0b011 && ((opcode>>16)&15)==15 && ((opcode>>4)&15)==0b0111) {
+                        //Sign extention
+                        int b = ((opcode>>20)&1);
+                        int s = ((opcode>>22)&1);
+                        int rd = ((opcode>>12)&15);
+                        int rm = ((opcode)&15);
+                        int rot = ((opcode>>10)&3);
+                        char tmp[20] = {0};
+                        if (rot)
+                            sprintf(tmp, " ror %d", rot*8);
+                        sprintf(ret, "%sXT%s %s, %s%s", s?"U":"S", b?"H":"B", regname[rd], regname[rm], tmp);
+                    } else if(((opcode>>21)&0b1111111)==0b0111111 && ((opcode>>4)&7)==0b101) {
+                        int widthm1 = ((opcode>>16)&31);
+                        int rd = ((opcode>>12)&15);
+                        int lsb = ((opcode>>7)&31);
+                        int rn = ((opcode)&15);
+                        sprintf(ret, "UBFX %s, %s, #%d, #%d", regname[rd], regname[rn], lsb, widthm1+1);
+                    } else if(((opcode>>21)&0b1111111)==0b0111110 && ((opcode>>4)&7)==0b001) {
+                        int msb = ((opcode>>16)&31);
+                        int rd = ((opcode>>12)&15);
+                        int lsb = ((opcode>>7)&31);
+                        int rn = ((opcode)&15);
+                        sprintf(ret, "BFI %s, %s, #%d, #%d", regname[rd], regname[rn], lsb, (msb-lsb+1));
+                    }
+                    else
+                    {
+                        // single data tranfert
+                        int i = ((opcode>>25)&1);
+                        int p = ((opcode>>24)&1);
+                        int u = ((opcode>>23)&1);
+                        int b = ((opcode>>22)&1);
+                        int w = ((opcode>>21)&1);
+                        int l = ((opcode>>20)&1);
+                        int rn = ((opcode>>16)&15);
+                        int rd = ((opcode>>12)&15);
+                        int offset = (opcode&0xfff);
+                        char op2[40];
+                        if(i) {
+                            int shift = offset>>4;
+                            int rm = offset&15;
+                            sprintf(op2, "%s%s%s", u?"":"-", regname[rm], print_shift(shift, 1));
+                        } else {
+                            sprintf(op2, "#%d", u?offset:-offset);
+                        }
+                        char addr[50];
+                        if(p) { // pre-index
+                            if(!offset) {
+                                sprintf(addr, "[%s]", regname[rn]);
+                            } else {
+                                if(i)
+                                    sprintf(addr, "[%s, %s]%s", regname[rn], op2, w?"!":"");
+                                else
+                                    sprintf(addr, "[%s, %s]%s", regname[rn], op2, w?"!":"");
+                            }
+                        } else {
+                            sprintf(addr, "[%s], %s", regname[rn], op2);
+                        }
+                        sprintf(ret, "%s%s%s%s %s, %s", l?"LDR":"STR", cond, b?"B":"", (w && p)?"T":"", regname[rd], addr);
+                    }
+                    break;
+                case 0b111:
+                    if(((opcode>>21)&0b1111111)==0b1110000 && (((opcode>>8)&0b1111)==0b1010) && ((opcode&0b1111111)==0b0010000)) {
+                        // VMOV ARM to/from Sm
+                        int rt = ((opcode>>12)&15);
+                        int vn = ((opcode>>16)&15)<<1 | ((opcode>>7)&1);
+                        int op = ((opcode>>20)&1);
+                        if(op==0)
+                            sprintf(ret, "VMOV%s S%d, %s", cond, vn, regname[rt]);
+                        else
+                            sprintf(ret, "VMOV%s %s, S%d", cond, regname[rt], vn);
+                    } else
+                    if(((opcode>>21)&0b1111111)==0b1110111 && (((opcode>>16)&0b1111)==0b0001) && ((opcode&0b111111111111)==0b101000010000)) {
+                        // VMOV ARM to/from FPCSR
+                        int rt = ((opcode>>12)&15);
+                        int op = ((opcode>>20)&1);
+                        if(op==0)
+                            sprintf(ret, "VMSR%s FPSCR, %s", cond, regname[rt]);
+                        else
+                            sprintf(ret, "VMRS%s %s, FPCSR", cond, regname[rt]);
+                    } else
+                    if(((opcode>>19)&0b111110111)==0b111010111 && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b101)==0b100)) {
+                        // VCVT float / integer
+                        int sz = ((opcode>>8)&1);
+                        int op = ((opcode>>7)&1);
+                        int opc2 = ((opcode>>16)&7);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int d, m;
+                        int to_integer = (opc2&4)?1:0;
+                        if(to_integer) {
+                            d = (Vd<<1) | D;
+                            m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        } else {
+                            m = ((Vm<<1)|M);
+                            d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        }
+                        switch(opc2) {
+                            case 0:
+                                sprintf(ret, "VCVT%s.F%d.%c32 %c%d, S%d", cond, sz?64:32, op?'S':'U', sz?'D':'S', d, m);
+                                break;
+                            case 0b100:
+                                sprintf(ret, "VCVT%s%s.U32.F%d S%d, %c%d", op?"":"R", cond, sz?64:32, d, sz?'D':'S', m);
+                                break;
+                            case 0b101:
+                                sprintf(ret, "VCVT%s%s.S32.F%d S%d, %c%d", op?"":"R", cond, sz?64:32, d, sz?'D':'S', m);
+                                break;
+                            default:
+                                sprintf(ret, "VCVT????%s ???", cond);
+                        }
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11100010 && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b101)==0b000)) {
+                        // VMUL
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vn = (opcode>>16)&15;
+                        int Vd = (opcode>>12)&15;
+                        int N = (opcode>>7)&1;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int n = (sz)?((N<<4)|Vn):((Vn<<1)|N);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VMUL%s.F%d %c%d, %c%d, %c%d", cond, sz?64:32, r, d, r, n, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101000 && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b101)==0b000)) {
+                        // VDIV
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vn = (opcode>>16)&15;
+                        int Vd = (opcode>>12)&15;
+                        int N = (opcode>>7)&1;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int n = (sz)?((N<<4)|Vn):((Vn<<1)|N);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VDIV%s.F%d %c%d, %c%d, %c%d", cond, sz?64:32, r, d, r, n, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11100011 && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b101)==0b000)) {
+                        // VADD
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vn = (opcode>>16)&15;
+                        int Vd = (opcode>>12)&15;
+                        int N = (opcode>>7)&1;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int n = (sz)?((N<<4)|Vn):((Vn<<1)|N);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VADD%s.F%d %c%d, %c%d, %c%d", cond, sz?64:32, r, d, r, n, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11100011 && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b101)==0b100)) {
+                        // VSUB
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vn = (opcode>>16)&15;
+                        int Vd = (opcode>>12)&15;
+                        int N = (opcode>>7)&1;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int n = (sz)?((N<<4)|Vn):((Vn<<1)|N);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VSUB%s.F%d %c%d, %c%d, %c%d", cond, sz?64:32, r, d, r, n, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0001) && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b1101)==0b0100)) {
+                        // VNEG
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        //int opc2 = (opcode>>16)&15;
+                        //int opc3 = (opcode>>6)&3;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VNEG%s.F%d %c%d, %c%d", cond, sz?64:32, r, d, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0001) && (((opcode>>9)&0b111)==0b101) && (((opcode>>4)&0b1101)==0b1100)) {
+                        // VSQRT
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        //int opc2 = (opcode>>16)&15;
+                        //int opc3 = (opcode>>6)&3;
+                        int d = (sz)?((D<<4)|Vd):((Vd<<1)|D);
+                        int m = (sz)?((M<<4)|Vm):((Vm<<1)|M);
+                        char r = sz?'D':'S';
+                        sprintf(ret, "VSQRT%s.F%d %c%d, %c%d", cond, sz?64:32, r, d, r, m);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0100) && (((opcode>>8)&0b1110)==0b1010) && (((opcode>>4)&0b0101)==0b0100)) {
+                        // VCMP single/double reg
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int vd = (sz)?((D<<4) | Vd):(D | (Vd<<1));
+                        int vm = (sz)?((M<<4) | Vm):(M | (Vm<<1));
+                        int E = (opcode>>7)&1;
+                        sprintf(ret, "VCMP%s%s.F%d %s%d, %s%d", E?"E":"", cond, sz?64:32, sz?"D":"S", vd, sz?"D":"S", vm);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0101) && (((opcode>>8)&0b1110)==0b1010) && (((opcode>>4)&0b0101)==0b0100)) {
+                        // VCMP single/double reg to 0
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int vd = (sz)?((D<<4) | Vd):(D | (Vd<<1));
+                        int E = (opcode>>7)&1;
+                        sprintf(ret, "VCMP%s%s.F%d %s%d, #0", E?"E":"", cond, sz?64:32, sz?"D":"S", vd);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>8)&0b1110)==0b1010) && (((opcode>>4)&0b1101)==0b0100)) {
+                        // VMOV single/double reg
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int vd = (sz)?((D<<4) | Vd):(D | (Vd<<1));
+                        int vm = (sz)?((M<<4) | Vm):(M | (Vm<<1));
+                        sprintf(ret, "VMOV%s.F%d %s%d, %s%d", cond, sz?64:32, sz?"D":"S", vd, sz?"D":"S", vm);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0000) && (((opcode>>8)&0b1110)==0b1010) && (((opcode>>4)&0b1101)==0b1100)) {
+                        // VABS single/double reg
+                        int sz = ((opcode>>8)&1);
+                        int D = (opcode>>22)&1;
+                        int Vd = (opcode>>12)&15;
+                        int M = (opcode>>5)&1;
+                        int Vm = (opcode)&15;
+                        int vd = (sz)?((D<<4) | Vd):(D | (Vd<<1));
+                        int vm = (sz)?((M<<4) | Vm):(M | (Vm<<1));
+                        sprintf(ret, "VABS%s.F%d %s%d, %s%d", cond, sz?64:32, sz?"D":"S", vd, sz?"D":"S", vm);
+                    } else
+                    if(((opcode>>20)&0b11111011)==0b11101011 && (((opcode>>16)&0b1111)==0b0111) && (((opcode>>4)&0b11101101)==0b10101100)) {
+                        // VCVT float / float
+                        int sz = ((opcode>>8)&1);
+                        if(sz) {    // double to single
+                            int vd = ((opcode>>12)&15)<<1 | ((opcode>>22)&1);
+                            int vm = ((opcode>>5)&1)<<4 | (opcode&15);
+                            sprintf(ret, "VCVT%s.F32.F64 S%d, D%d", cond, vd, vm);
+                        } else {
+                            int vd = (opcode>>12)&15 | ((opcode>>22)&1)<<4;
+                            int vm = ((opcode>>5)&1) | (opcode&15)<<1;
+                            sprintf(ret, "VCVT%s.F64.F32 D%d, S%d", cond, vd, vm);
+                        }
+                    }
+                    break;
+                case 0b110:
+                    if(((opcode>>21)&0b1111111)==0b1100010 && (((opcode>>8)&0b1111)==0b1010) && (((opcode>>4)&0b1101)==0b0001)) {
+                        // VMOV ARM to/from 2*SM
+                        int rt = ((opcode>>12)&15);
+                        int rt2 = ((opcode>>16)&15);
+                        int vm = (opcode&15)<<1 | ((opcode>>5)&1);
+                        int op = ((opcode>>20)&1);
+                        if(op==0)
+                            sprintf(ret, "VMOV%s S%d, S%d, %s, %s", cond, vm, vm+1, regname[rt], regname[rt2]);
+                        else
+                            sprintf(ret, "VMOV%s %s, %s, S%d, S%d", cond, regname[rt], regname[rt2], vm, vm+1);
+                    } else
+                    if(((opcode>>21)&0b1111111)==0b1100010 && (((opcode>>8)&0b1111)==0b1011) && (((opcode>>4)&0b1101)==0b0001)) {
+                        // VMOV ARM to/from 2*SM
+                        int rt = ((opcode>>12)&15);
+                        int rt2 = ((opcode>>16)&15);
+                        int vm = (opcode&15) | ((opcode>>5)&1)<<4;
+                        int op = ((opcode>>20)&1);
+                        if(op==0)
+                            sprintf(ret, "VMOV%s D%d, %s, %s", cond, vm, regname[rt], regname[rt2]);
+                        else
+                            sprintf(ret, "VMOV%s %s, %s, D%d", cond, regname[rt], regname[rt2], vm);
+                    } else
+                    if(((opcode>>21)&0b1111001)==0b1101000 && (((opcode>>8)&0b1110)==0b1010)) {
+                        // VLDR/VSTR to/from Single/Double
+                        uint32_t ldr = ((opcode>>20)&1);
+                        uint32_t u = ((opcode>>23)&1);
+                        uint32_t rn = ((opcode>>16)&15);
+                        int32_t imm8 = (opcode&255)<<2;
+                        uint32_t D = (opcode>>22)&1;
+                        uint32_t Vd = (opcode>>12)&15;
+                        uint32_t notsingle = ((opcode>>8)&1);
+                        uint32_t vd = (notsingle)?(D<<4 | Vd):(Vd<<1 | D);
+                        char offset[50] = {0};
+                        if(imm8)
+                            sprintf(offset, ", #%d", u?imm8:-imm8);
+                        sprintf(ret, "V%s%s %s%d, [%s%s]", ldr?"LDR":"STR", cond, notsingle?"D":"S", vd, regname[rn], offset);
+                    }
+                    break;
+                default:
+                    sprintf(ret, "???%s", cond);
+            }
+        }
+    }
+
+    return ret;
+}

src/dynarec/arm_printer.h

@@ -0,0 +1,6 @@
+#ifndef _ARM_PRINTER_H_
+#define _ARM_PRINTER_H_
+
+const char* arm_print(uint32_t opcode);
+
+#endif //_ARM_PRINTER_H_

src/dynarec/arm_prolog.S

@@ -0,0 +1,19 @@
+//arm prologue for dynarec
+//Save stuff, prepare stack and register
+//called with pointer to emu as 1st parameter
+//and address to jump to as 2nd parameter
+
+.text
+.align 4
+
+.global arm_prolog
+arm_prolog:
+    //save all used register
+    push     {r4-r11, lr}
+    //vpush     {d8-d15}
+    //setup emu -> register
+    //push     {r0}
+    ldm      r0, {r4-r11}    // all 8 register in direct access, no EIP (so r12 can be used as scratch)
+    //TODO
+    //jump to function
+    bx       r1

src/dynarec/arm_table.S

@@ -0,0 +1,19 @@
+//arm try to update linker table for dynarec
+//and address of table to as 1st parameter
+//jump addres as second and ip as third
+// don't retry if failed to write...
+
+.text
+.align 4
+
+.global arm_tableupdate
+arm_tableupdate:
+    // jump address is r0 and IP address is r1
+    // table offset is r2
+    push    {r4, r5}
+    ldrexd  r4, r5, [r2]
+    strexd  r3, r0, r1, [r2]
+    pop     {r4, r5}
+    mov     r0, r3
+    bx      lr
+

src/dynarec/dynablock.c

@@ -0,0 +1,179 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "tools/bridge_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynablock.h"
+#include "dynablock_private.h"
+#include "elfloader.h"
+#ifdef ARM
+#include "dynarec_arm.h"
+#else
+#error Unsupported architecture!
+#endif
+
+#include "khash.h"
+
+KHASH_MAP_INIT_INT(dynablocks, dynablock_t*)
+
+
+dynablocklist_t* NewDynablockList(uintptr_t base, uintptr_t text, int textsz)
+{
+    dynablocklist_t* ret = (dynablocklist_t*)calloc(1, sizeof(dynablocklist_t));
+    ret->blocks = kh_init(dynablocks);
+    ret->base = base;
+    ret->text = text;
+    ret->textsz = textsz;
+    pthread_rwlock_init(&ret->rwlock_blocks, NULL);
+
+    return ret;
+}
+
+void FreeDynablockList(dynablocklist_t** dynablocks)
+{
+    if(!dynablocks)
+        return;
+    if(!*dynablocks)
+        return;
+    dynarec_log(LOG_INFO, "Free %d Blocks from Dynablocklist (with %d buckets)%s\n", kh_size((*dynablocks)->blocks), kh_n_buckets((*dynablocks)->blocks), ((*dynablocks)->direct)?" With Direct mapping enabled":"");
+    dynablock_t* db;
+    kh_foreach_value((*dynablocks)->blocks, db,
+        free(db->table);
+        free(db);
+    );
+    kh_destroy(dynablocks, (*dynablocks)->blocks);
+    if((*dynablocks)->direct)
+        free((*dynablocks)->direct);
+    (*dynablocks)->direct = 0;
+
+    pthread_rwlock_destroy(&(*dynablocks)->rwlock_blocks);
+
+    free(*dynablocks);
+    *dynablocks = NULL;
+}
+
+void ConvertHash2Direct(dynablocklist_t* dynablocks)
+{
+    if(dynablocks->textsz==0 || dynablocks->text==0)
+        return; // nothing to do
+    // create the new set
+    dynablock_t **direct = (dynablock_t**)calloc(dynablocks->textsz, sizeof(dynablock_t*));
+    kh_dynablocks_t *blocks = kh_init(dynablocks);
+    // transfert
+    int ret;
+    khint_t k;
+    dynablock_t* db;
+    uintptr_t key;
+    uintptr_t start = dynablocks->text-dynablocks->base;
+    uintptr_t end = dynablocks->text + dynablocks->textsz-dynablocks->base;
+    kh_foreach(dynablocks->blocks, key, db,
+        if(key>=start && key<end)
+            direct[key-start] = db;
+        else {
+            k = kh_put(dynablocks, blocks, key, &ret);
+            if(ret) {   // don't try to insert if already done...
+                kh_value(blocks, k) = db;
+            }
+        }
+    );
+    // destroy old and do the swap (should swap before destroy, but hash access is always behind mutex)
+    kh_destroy(dynablocks, dynablocks->blocks);
+    dynablocks->blocks = blocks;
+    dynablocks->direct = direct;
+}
+
+#define MAGIC_SIZE 256
+/* 
+    return NULL if block is not found / cannot be created. 
+    Don't create if create==0
+*/
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t* current)
+{
+    // try the quickest way first: get parent of current and check if ok!
+    dynablocklist_t *dynablocks = NULL;
+    dynablock_t* block = NULL;
+    if(current) {
+        dynablocks = current->parent;    
+        if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz) {
+            block = dynablocks->direct[addr-dynablocks->text];
+            if(block)
+                return block;
+        }
+        if(!(addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz))
+            dynablocks = NULL;
+    }
+    // nope, lets do the long way
+    if(!dynablocks)
+        dynablocks = GetDynablocksFromAddress(emu->context, addr);
+    if(!dynablocks)
+        return NULL;
+    // check direct first, without lock
+    if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz)
+        block = dynablocks->direct[addr-dynablocks->text];
+    if(block)
+        return block;
+    // nope, put rwlock in read mode and check hash
+    pthread_rwlock_rdlock(&dynablocks->rwlock_blocks);
+    // but first, check again just in case it has been created while waiting for mutex
+    if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz)
+        block = dynablocks->direct[addr-dynablocks->text];
+    if(block) {
+        pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+        return block;
+    }
+    // check if the block exist
+    int ret;
+    khint_t k = kh_get(dynablocks, dynablocks->blocks, addr-dynablocks->base);
+    if(k!=kh_end(dynablocks->blocks)) {
+        /*atomic_store(&dynalock, 0);*/
+        block = kh_value(dynablocks->blocks, k);
+        pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+        return block;
+    }
+    // unlock now
+    pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+    // Blocks doesn't exist. If creation is not allow, just return NULL
+    if(!create)
+        return block;
+    // Lock as write now!
+    pthread_rwlock_wrlock(&dynablocks->rwlock_blocks);
+    // create and add new block
+    dynarec_log(LOG_DEBUG, "Ask for DynaRec Block creation @%p\n", addr);
+    if(dynablocks->direct && addr>=dynablocks->text && addr<=dynablocks->text+dynablocks->textsz)
+    {
+        block = dynablocks->direct[addr-dynablocks->text] = (dynablock_t*)calloc(1, sizeof(dynablock_t));
+    } else {
+        k = kh_put(dynablocks, dynablocks->blocks, addr-dynablocks->base, &ret);
+        if(!ret) {
+            // Ooops, block is already there? retreive it and bail out
+            block = kh_value(dynablocks->blocks, k);    
+            pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+            return block;
+        }
+        block = kh_value(dynablocks->blocks, k) = (dynablock_t*)calloc(1, sizeof(dynablock_t));
+        // check if size of hash map == magic size, if yes, convert to direct before unlocking
+        if(!dynablocks->direct && kh_size(dynablocks->blocks)==MAGIC_SIZE)
+            ConvertHash2Direct(dynablocks);
+    }
+    block->parent = dynablocks;
+    // create an empty block first, so if other thread want to execute the same block, they can, but using interpretor path
+    pthread_rwlock_unlock(&dynablocks->rwlock_blocks);
+
+    // fill the block
+    FillBlock(emu, block, addr);
+    dynarec_log(LOG_DEBUG, " --- DynaRec Block created @%p (%p, 0x%x bytes)\n", addr, block->block, block->size);
+
+    return block;
+}

src/dynarec/dynablock_private.h

@@ -0,0 +1,26 @@
+#ifndef __DYNABLOCK_PRIVATE_H_
+#define __DYNABLOCK_PRIVATE_H_
+
+typedef struct dynablocklist_s dynablocklist_t;
+
+typedef struct dynablock_s {
+    dynablocklist_t *parent;
+    void*       block;
+    int         size;
+    uintptr_t*  table;
+    int         tablesz;
+    int         done;
+} dynablock_t;
+
+typedef struct kh_dynablocks_s kh_dynablocks_t;
+
+typedef struct dynablocklist_s {
+    kh_dynablocks_t     *blocks;
+    pthread_rwlock_t    rwlock_blocks;
+    uintptr_t           base;
+    uintptr_t           text;
+    int                 textsz;
+    dynablock_t         **direct;    // direct mapping (waste of space, so not always there)
+} dynablocklist_t;
+
+#endif //__DYNABLOCK_PRIVATE_H_

src/dynarec/dynarec.c

@@ -0,0 +1,167 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "tools/bridge_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#ifdef DYNAREC
+#include "dynablock.h"
+#include "dynablock_private.h"
+#endif
+
+#ifdef ARM
+void arm_prolog(x86emu_t* emu, void* addr);
+void arm_epilog();
+void arm_linker();
+int arm_tableupdate(void* jump, uintptr_t addr, void** table);
+#endif
+
+#ifdef DYNAREC
+void* UpdateLinkTable(x86emu_t* emu, void** table, uintptr_t addr)
+{
+    dynablock_t* block = DBGetBlock(emu, addr, 1, NULL);    // keep a copy of parent block?
+    int r;
+    if(block==0) {
+        // no block, don't try again, ever
+        #ifdef ARM
+        r = arm_tableupdate(arm_epilog, addr, table);
+        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
+        #else
+        table[0] = arm_epilog;
+        table[1] = addr;
+        #endif
+        return arm_epilog;
+    }
+    if(!block->done) {
+        // not finished yet... leave linker
+        #ifdef ARM
+        r = arm_tableupdate(arm_linker, addr, table);
+        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
+        #else
+        table[0] = arm_linker;
+        table[1] = addr;
+        #endif
+        return arm_epilog;
+    }
+    if(!block->block) {
+        // null block, but done, go to epilog
+        #ifdef ARM
+        r = arm_tableupdate(arm_epilog, addr, table);
+        if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
+        #else
+        table[0] = arm_epilog;
+        table[1] = addr;
+        #endif
+        return arm_epilog;
+    }
+    #ifdef ARM
+    r = arm_tableupdate(block->block, addr, table);
+    if(r) dynarec_log(LOG_DEBUG, "Linker: failed to set table data @%p, for emu=%p\n", table, emu);
+    #else
+    table[0] = block->block;
+    table[1] = addr;
+    #endif
+    return block->block;
+}
+#endif
+
+void DynaCall(x86emu_t* emu, uintptr_t addr)
+{
+#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;
+        while(!emu->quit) {
+            block = DBGetBlock(emu, R_EIP, 1, block);
+            if(!block || !block->block || !block->done) {
+                // no block, of block doesn't have DynaRec content (yet, temp is not null)
+                // Use interpreter (should use single instruction step...)
+                dynarec_log(LOG_DEBUG, "Calling Interpretor @%p, emu=%p\n", R_EIP, emu);
+                Run(emu, 1);
+            } else {
+                dynarec_log(LOG_DEBUG, "Calling DynaRec Block @%p (%p) emu=%p\n", R_EIP, block->block, 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);
+            }
+        }
+        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
+}
+
+int DynaRun(x86emu_t* emu)
+{
+#ifdef DYNAREC
+    if(!box86_dynarec)
+#endif
+        return Run(emu, 0);
+#ifdef DYNAREC
+    else {
+        dynablock_t* block = NULL;
+        while(!emu->quit) {
+            block = DBGetBlock(emu, R_EIP, 1, block);
+            if(!block || !block->block || !block->done) {
+                // no block, of block doesn't have DynaRec content (yet, temp is not null)
+                // Use interpreter (should use single instruction step...)
+                dynarec_log(LOG_DEBUG, "Running Interpretor @%p, emu=%p\n", R_EIP, emu);
+                Run(emu, 1);
+            } else {
+                dynarec_log(LOG_DEBUG, "Running DynaRec Block @%p (%p) emu=%p\n", R_EIP, block->block, 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);
+            }
+        }
+    }
+    return 0;
+#endif
+}

src/dynarec/dynarec_arm.c

@@ -0,0 +1,123 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "tools/bridge_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynablock.h"
+#include "dynablock_private.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+
+void printf_x86_instruction(zydis_dec_t* dec, instruction_x86_t* inst, const char* name) {
+    uint8_t *ip = (uint8_t*)inst->addr;
+    if(ip[0]==0xcc && ip[1]=='S' && ip[2]=='C') {
+        uint32_t a = *(uint32_t*)(ip+3);
+        if(a==0) {
+            dynarec_log(LOG_NONE, "%p: Exit x86emu\n", (void*)ip);
+        } else {
+            dynarec_log(LOG_NONE, "%p: Native call to %p\n", (void*)ip, (void*)a);
+        }
+    } else {
+        if(dec) {
+            dynarec_log(LOG_NONE, "%p: %s\n", inst->addr, DecodeX86Trace(dec, inst->addr));
+        } else {
+            dynarec_log(LOG_NONE, "%p: ", inst->addr);
+            for(int i=0; i<inst->size; ++i) {
+                dynarec_log(LOG_NONE, "%02X ", ip[i]);
+            }
+            dynarec_log(LOG_NONE, " %s\n", name);
+        }
+    }
+}
+
+void arm_pass0(dynarec_arm_t* dyn, uintptr_t addr);
+void arm_pass1(dynarec_arm_t* dyn, uintptr_t addr);
+void arm_pass2(dynarec_arm_t* dyn, uintptr_t addr);
+void arm_pass3(dynarec_arm_t* dyn, uintptr_t addr);
+
+void FillBlock(x86emu_t* emu, dynablock_t* block, uintptr_t addr) {
+    // init the helper
+    dynarec_arm_t helper = {0};
+    helper.emu = emu;
+    arm_pass0(&helper, addr);
+    if(!helper.size) {
+        dynarec_log(LOG_DEBUG, "Warning, null-sized dynarec block (%p)\n", (void*)addr);
+        block->done = 1;
+        return;
+    }
+    helper.cap = helper.size+3; // needs epilog handling
+    helper.insts = (instruction_arm_t*)calloc(helper.cap, sizeof(instruction_arm_t));
+    // pass 1, addresses, x86 jump addresses, flags
+    arm_pass1(&helper, addr);
+    // calculate barriers
+    uintptr_t start = helper.insts[0].x86.addr;
+    uintptr_t end = helper.insts[helper.size].x86.addr+helper.insts[helper.size].x86.size;
+    for(int i=0; i<helper.size; ++i)
+        if(helper.insts[i].x86.jmp) {
+            uintptr_t j = helper.insts[i].x86.jmp;
+            if(j<start || j>=end)
+                helper.insts[i].x86.jmp_insts = -1;
+            else {
+                // find jump address instruction
+                int k=-1;
+                for(int i2=0; i2<helper.size && k==-1; ++i2) {
+                    if(helper.insts[i2].x86.addr==j)
+                        k=i2;
+                }
+                if(k!=-1)   // -1 if not found, mmm, probably wrong, exit anyway
+                    helper.insts[k].x86.barrier = 1;
+                helper.insts[i].x86.jmp_insts = k;
+            }
+        }
+    // remove useless flags calulation
+    for(int i=0; i<helper.size; ++i)
+        if(helper.insts[i].x86.flags==X86_FLAGS_CHANGE) {
+            int done = 0;
+            for(int i2=i+1; i2<helper.size+1 && done==0; ++i2) {
+                if(helper.insts[i2].x86.barrier || helper.insts[i2].x86.jmp)
+                    done = 1;
+                else if(helper.insts[i2].x86.flags==X86_FLAGS_USE)
+                    done = 1;
+                else if(helper.insts[i2].x86.flags==X86_FLAGS_CHANGE) {
+                    done = 1;
+                    helper.insts[i].x86.flags=X86_FLAGS_NONE;
+                }
+            }
+        }
+    // pass 2, instruction size
+    arm_pass2(&helper, addr);
+    // ok, now allocate mapped memory, with executable flag on
+    int sz = helper.arm_size;
+    void* p = (void*)AllocDynarecMap(emu->context, sz);
+    if(p==NULL) {
+        free(helper.insts);
+        return;
+    }
+    helper.block = p;
+    helper.tablesz = helper.tablei;
+    if(helper.tablesz)
+        helper.table = (uintptr_t*)calloc(helper.tablesz, sizeof(uintptr_t));
+    // pass 3, emit (log emit arm opcode)
+    dynarec_log(LOG_DEBUG, "Emitting %d bytes for %d x86 bytes\n", helper.arm_size, helper.isize);
+    helper.arm_size = 0;
+    arm_pass3(&helper, addr);
+    // all done...
+    __builtin___clear_cache(p, p+helper.arm_size);   // need to clear the cache before execution...
+    free(helper.insts);
+    block->table = helper.table;
+    block->tablesz = helper.tablesz;
+    block->size = sz;
+    block->block = p;
+    block->done = 1;
+}

src/dynarec/dynarec_arm_65.c

@@ -0,0 +1,116 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_helper.h"
+
+uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    int32_t i32;
+    uint32_t u32;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    switch(opcode) {
+        case 0x33:
+            grab_tlsdata(dyn, addr, ninst, x12);
+            INST_NAME("GS:XOR Gd, Ed");
+            nextop = F8;
+            GETGD;
+            GETEDO(x12);
+            XOR_REG_LSL_IMM8(gd, gd, ed, 0);
+            UFLAG_RES(gd);
+            UFLAG_DF(1, d_xor32);
+            UFLAGS(0);
+            break;
+
+        case 0x89:
+            grab_tlsdata(dyn, addr, ninst, x12);
+            INST_NAME("MOV GS:Ed, Gd");
+            nextop=F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {   // reg <= reg
+                MOV_REG(xEAX+(nextop&7), gd);
+            } else {                    // mem <= reg
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                STR_REG_LSL_IMM5(gd, ed, x12, 0);
+            }
+            break;
+
+        case 0x8B:
+            grab_tlsdata(dyn, addr, ninst, x12);
+            INST_NAME("MOV Gd, GS:Ed");
+            nextop=F8;
+            GETGD;
+            if((nextop&0xC0)==0xC0) {   // reg <= reg
+                MOV_REG(gd, xEAX+(nextop&7));
+            } else {                    // mem <= reg
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                LDR_REG_LSL_IMM5(gd, ed, x12, 0);
+            }
+            break;
+
+        case 0xA1:
+            grab_tlsdata(dyn, addr, ninst, x1);
+            INST_NAME("MOV EAX, GS:Id");
+            i32 = F32S;
+            if(i32>-4096 && i32<4096) {
+                LDR_IMM9(xEAX, x1, i32);
+            } else {
+                MOV32(x2, i32);
+                ADD_REG_LSL_IMM5(x1, x1, x2, 0);
+                LDR_IMM9(xEAX, x1, 0);
+            }
+            break;
+
+        case 0xA3:
+            grab_tlsdata(dyn, addr, ninst, x1);
+            INST_NAME("MOV GS:Od, EAX");
+            u32 = F32;
+            MOV32(x2, u32);
+            ADD_REG_LSL_IMM5(x2, x1, x2, 0);
+            STR_IMM9(xEAX, x2, 0);
+            break;
+
+        case 0xC7:
+            grab_tlsdata(dyn, addr, ninst, x12);
+            INST_NAME("MOV GS:Ed, Id");
+            nextop=F8;
+            if((nextop&0xC0)==0xC0) {   // reg <= i32
+                i32 = F32S;
+                ed = xEAX+(nextop&7);
+                MOV32(ed, i32);
+            } else {                    // mem <= i32
+                addr = geted(dyn, addr, ninst, nextop, &ed, x2, &fixedaddress, 0, 0);
+                i32 = F32S;
+                MOV32(x3, i32);
+                STR_REG_LSL_IMM5(x3, ed, x12, 0);
+            }
+            break;
+
+        default:
+            *ok = 0;
+            DEFAULT;
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_67.c

@@ -0,0 +1,104 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t opcode = F8;
+    uint8_t nextop, u8;
+    int8_t  i8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    switch(opcode) {
+       
+
+        #define GO(NO, YES)   \
+            BARRIER(2); \
+            JUMP(addr+i8);\
+            if(dyn->insts) {    \
+                if(dyn->insts[ninst].x86.jmp_insts==-1) {   \
+                    /* out of the block */                  \
+                    i32 = dyn->insts[ninst+1].address-(dyn->arm_size+8); \
+                    Bcond(NO, i32);     \
+                    jump_to_linker(dyn, addr+i8, 0, ninst); \
+                } else {    \
+                    /* inside the block */  \
+                    i32 = dyn->insts[dyn->insts[ninst].x86.jmp_insts].address-(dyn->arm_size+8);    \
+                    Bcond(YES, i32);    \
+                }   \
+            }
+        case 0xE0:
+            INST_NAME("LOOPNZ (16bits)");
+            USEFLAG(1);
+            i8 = F8S;
+            UXTH(x1, xECX, 0);
+            SUBS_IMM8(x1, x1, 1);
+            BFI(xECX, x1, 0, 16);
+            B_NEXT(cEQ);    // CX is 0, no LOOP
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            GO(cNE, cEQ);
+            break;
+        case 0xE1:
+            INST_NAME("LOOPZ (16bits)");
+            USEFLAG(1);
+            i8 = F8S;
+            UXTH(x1, xECX, 0);
+            SUBS_IMM8(x1, x1, 1);
+            BFI(xECX, x1, 0, 16);
+            B_NEXT(cEQ);    // CX is 0, no LOOP
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            GO(cEQ, cNE);
+            break;
+        case 0xE2:
+            INST_NAME("LOOP (16bits)");
+            USEFLAG(1);
+            i8 = F8S;
+            UXTH(x1, xECX, 0);
+            SUBS_IMM8(x1, x1, 1);
+            BFI(xECX, x1, 0, 16);
+            GO(cEQ, cNE);
+            break;
+        case 0xE3:
+            INST_NAME("JCXZ");
+            USEFLAG(1);
+            i8 = F8S;
+            UXTH(x1, xECX, 0);
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            GO(cNE, cEQ);
+            break;
+        #undef GO
+
+        default:
+            *ok = 0;
+            DEFAULT;
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_d8.c

@@ -0,0 +1,242 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2, v3;
+    int s0, s1, s2;
+    int d0, d1;
+    int fixedaddress;
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FADD ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VADD_F64(v1, v1, v2);
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FMUL ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VMUL_F64(v1, v1, v2);
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCOM ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            break;
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FCOMP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FSUB ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v1, v1, v2);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FSUBR ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v1, v2, v1);
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FDIV ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v1, v1, v2);
+            break;
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            INST_NAME("FDIVR ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v1, v2, v1);
+            break;
+      
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FADD ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VADD_F64(v1, v1, d1);
+                    break;
+                case 1:
+                    INST_NAME("FMUL ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VMUL_F64(v1, v1, d1);
+                    break;
+                case 2:
+                    INST_NAME("FCOM ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VCMP_F64(v1, d1);
+                    FCOM(x1, x2);
+                    break;
+                case 3:
+                    INST_NAME("FCOMP ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VCMP_F64(v1, d1);
+                    FCOM(x1, x2);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 4:
+                    INST_NAME("FSUB ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VSUB_F64(v1, v1, d1);
+                    break;
+                case 5:
+                    INST_NAME("FSUBR ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VSUB_F64(v1, d1, v1);
+                    break;
+                case 6:
+                    INST_NAME("FDIV ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VDIV_F64(v1, v1, d1);
+                    break;
+                case 7:
+                    INST_NAME("FDIVR ST0, float[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_F32(d1, s0);
+                    VDIV_F64(v1, d1, v1);
+                    break;
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_d9.c

@@ -0,0 +1,327 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+#include "emu/x87emu_private.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+static double d_1   = 1.0;
+static double d_l2t = L2T;
+static double d_l2e = L2E;
+static double d_pi  = PI;
+static double d_lg2 = LG2;
+static double d_ln2 = LN2;
+static double d_0   = 0.0;
+
+uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    int v1, v2, v3;
+    int s0, s1, s2;
+    int i1, i2, i3;
+    switch(nextop) {
+
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FLD STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            v2 = x87_do_push(dyn, ninst);
+            VMOV_64(v2, v1);
+            break;
+
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FXCH STx");
+            // swap the cache value, not the double value itself :p
+            i1 = x87_get_cache(dyn, ninst, x1, x2, nextop&7);
+            i2 = x87_get_cache(dyn, ninst, x1, x2, 0);
+            i3 = dyn->x87cache[i1];
+            dyn->x87cache[i1] = dyn->x87cache[i2];
+            dyn->x87cache[i2] = i3;
+            break;
+
+        case 0xD0:
+            INST_NAME("FNOP");
+            break;
+
+        case 0xE0:
+            INST_NAME("FCHS");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            VNEG_F64(v1, v1);
+            break;
+        case 0xE1:
+            INST_NAME("FABS");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            VABS_F64(v1, v1);
+            break;
+
+        case 0xE4:
+            INST_NAME("FTST");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            VCMP_F64_0(v1);
+            FCOM(x1, x2);   // same flags...
+            break;
+        case 0xE5:
+            INST_NAME("FXAM");
+            x87_refresh(dyn, ninst, x1, x2, 0);
+            CALL(fpu_fxam, -1, 0);  // should be possible inline, but is it worth it?
+            break;
+
+        case 0xE8:
+            INST_NAME("FLD1");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_1));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xE9:
+            INST_NAME("FLDL2T");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_l2t));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xEA:     
+            INST_NAME("FLDL2E");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_l2e));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xEB:
+            INST_NAME("FLDPI");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_pi));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xEC:
+            INST_NAME("FLDLG2");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_lg2));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xED:
+            INST_NAME("FLDLN2");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_ln2));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xEE:
+            INST_NAME("FLDZ");
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_0));
+            VLDR_64(v1, x2, 0);
+            break;
+
+        case 0xFA:
+            INST_NAME("FSQRT");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            VSQRT_F64(v1, v1);
+            break;
+
+        case 0xFC:
+            INST_NAME("FRNDINT");
+            // use C helper for now, nothing staightforward is available
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_frndint, -1, 0);
+            /*
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            VCMP_F64_0(v1);
+            VMRS_APSR();
+            B_NEXT(cVS);    // Unordered, skip
+            B_NEXT(cEQ);    // Zero, skip
+            u8 = x87_setround(dyn, ninst, x1, x2, x3);
+            VCVT_S32_F64(x1, v1);   // limit to 32bits....
+            VCVT_F64_S32(v1, x1);
+            x87_restoreround(dyn, ninst, u8);
+            */
+            break;
+        case 0xF0:
+            INST_NAME("F2XM1");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_f2xm1, -1, 0);
+            break;
+        case 0xF1:
+            INST_NAME("FYL2X");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fyl2x, -1, 0);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xF2:
+            INST_NAME("FTAN");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_ftan, -1, 0);
+            v1 = x87_do_push(dyn, ninst);
+            MOV32(x2, (&d_1));
+            VLDR_64(v1, x2, 0);
+            break;
+        case 0xF3:
+            INST_NAME("FPATAN");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fpatan, -1, 0);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xF4:
+            INST_NAME("FXTRACT");
+            x87_do_push_empty(dyn, ninst, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fxtract, -1, 0);
+            break;
+        case 0xF8:
+            INST_NAME("FPREM");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fprem, -1, 0);
+            break;
+        case 0xF9:
+            INST_NAME("FYL2XP1");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fyl2xp1, -1, 0);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xFB:
+            INST_NAME("FSINCOS");
+            x87_do_push_empty(dyn, ninst, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fsincos, -1, 0);
+            break;
+        case 0xFD:
+            INST_NAME("FSCALE");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            x87_forget(dyn, ninst, x1, x2, 1);
+            CALL(arm_fscale, -1, 0);
+            break;
+        case 0xFE:
+            INST_NAME("FSIN");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_fsin, -1, 0);
+            break;
+        case 0xFF:
+            INST_NAME("FCOS");
+            x87_forget(dyn, ninst, x1, x2, 0);
+            CALL(arm_fcos, -1, 0);
+            break;
+
+
+        case 0xD1:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+        case 0xE2:
+        case 0xE3:
+        case 0xE6:
+        case 0xE7:
+        case 0xEF:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            *ok = 0;
+            DEFAULT;
+            break;
+             
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FLD ST0, float[ED]");
+                    v1 = x87_do_push(dyn, ninst);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    s0 = fpu_get_scratch_single(dyn);
+                    // to avoid bus error
+                    //VLDR_32(s0, ed, 0);
+                    LDR_IMM9(x2, ed, fixedaddress);
+                    VMOVtoV(s0, x2);
+                    VCVT_F64_F32(v1, s0);
+                    break;
+                case 2:
+                    INST_NAME("FST float[ED], ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VCVT_F32_F64(s0, v1);
+                    // to avoid bus error...
+                    //VSTR_32(s0, ed, 0);
+                    VMOVfrV(x2, s0);
+                    STR_IMM9(x2, ed, fixedaddress);
+                    break;
+                case 3:
+                    INST_NAME("FSTP float[ED], ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VCVT_F32_F64(s0, v1);
+                    // to avoid bus error...
+                    //VSTR_32(s0, ed, 0);
+                    VMOVfrV(x2, s0);
+                    STR_IMM9(x2, ed, fixedaddress);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 5:
+                    INST_NAME("FLDCW Ew");
+                    GETEW(x1);
+                    STRH_IMM8(x1, xEmu, offsetof(x86emu_t, cw));    // hopefully cw is not too far for an imm8
+                    UBFX(x1, x1, 10, 2);    // extract round
+                    STR_IMM9(x1, xEmu, offsetof(x86emu_t, round));
+                    break;
+                case 7:
+                    INST_NAME("FNSTCW Ew");
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0);
+                    ed = x1;
+                    wb1 = 1;
+                    LDRH_IMM8(x1, xEmu, offsetof(x86emu_t, cw));
+                    EWBACK;
+                    break;
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_da.c

@@ -0,0 +1,219 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2;
+    int d0;
+    int s0;
+    int fixedaddress;
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FCMOVB ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_CF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cNE, v1, v2);   // F_CF==1
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FCMOVE ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cNE, v1, v2);   // F_ZF==1
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCMOVBE ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, flags[F_CF]));
+            ORRS_REG_LSL_IMM8(x1, x1, x2, 0);
+            VMOVcond_64(cNE, v1, v2);   // F_CF==1 | F_ZF==1
+            break;
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FCMOVU ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_PF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cNE, v1, v2);   // F_PF==1
+            break;       
+        case 0xE9:
+            INST_NAME("FUCOMPP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst);
+            break;
+
+        case 0xE4:
+        case 0xF0:
+        case 0xF1:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFD:
+            *ok = 0;
+            DEFAULT;
+            break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FIADD ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VADD_F64(v1, v1, d0);
+                    break;
+                case 1:
+                    INST_NAME("FIMUL ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VMUL_F64(v1, v1, d0);
+                    break;
+                case 2:
+                    INST_NAME("FICOM ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VCMP_F64(v1, d0);
+                    FCOM(x1, x2);
+                    break;
+                case 3:
+                    INST_NAME("FICOMP ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VCMP_F64(v1, d0);
+                    FCOM(x1, x2);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 4:
+                    INST_NAME("FISUB ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VSUB_F64(v1, v1, d0);
+                    break;
+                case 5:
+                    INST_NAME("FISUBR ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VSUB_F64(v1, d0, v1);
+                    break;
+                case 6:
+                    INST_NAME("FIDIV ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VDIV_F64(v1, v1, d0);
+                    break;
+                case 7:
+                    INST_NAME("FIDIVR ST0, Ed");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    GETED;
+                    d0 = fpu_get_scratch_double(dyn);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(d0, s0);
+                    VDIV_F64(v1, d0, v1);
+                    break;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_db.c

@@ -0,0 +1,276 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "emu/x87emu_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+static double maxint32 = (double)0x7fffffff;
+
+uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2, v3;
+    int s0, s1, s2;
+    int d0;
+    int fixedaddress;
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FCMOVNB ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_CF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cEQ, v1, v2);   // F_CF==0
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FCMOVNE ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cEQ, v1, v2);   // F_ZF==0
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCMOVNBE ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_ZF]));
+            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, flags[F_CF]));
+            ORRS_REG_LSL_IMM8(x1, x1, x2, 0);
+            VMOVcond_64(cEQ, v1, v2);   // F_CF==0 | F_ZF==0
+            break;
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FCMOVNU ST0, STx");
+            USEFLAG(1);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            LDR_IMM9(x1, xEmu, offsetof(x86emu_t, flags[F_PF]));
+            TSTS_REG_LSL_IMM8(x1, x1, 0);
+            VMOVcond_64(cEQ, v1, v2);   // F_PF==0
+            break;
+        case 0xE2:
+            INST_NAME("FNCLEX");
+            MOVW(x1, 0);
+            LDRH_IMM8(x2, xEmu, offsetof(x86emu_t, sw));
+            BFI(x2, x1, 0, 8);  // IE .. PE, SF, ES
+            BFI(x2, x1, 15, 1); // B
+            STRH_IMM8(x2, xEmu, offsetof(x86emu_t, sw));
+            break;
+        case 0xE3:
+            INST_NAME("FNINIT");
+            fpu_purgecache(dyn, ninst, x1, x2, x3); // maybe only x87, not SSE?
+            CALL(reset_fpu, -1, 0);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FUCOMI ST0, STx");
+            UFLAGS(0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOMI(x1, x2);
+            UFLAGS(1);
+            break;
+        case 0xF0:  
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FCOMI ST0, STx");
+            UFLAGS(0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOMI(x1, x2);
+            UFLAGS(1);
+            break;
+
+        case 0xE0:
+        case 0xE1:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            *ok = 0;
+            DEFAULT;
+            break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FILD ST0, Ed");
+                    v1 = x87_do_push(dyn, ninst);
+                    GETED;
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, ed);
+                    VCVT_F64_S32(v1, s0);
+                    break;
+                case 1:
+                    INST_NAME("FISTTP Ed, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 4095, 0);
+                        ed = x1;
+                    }
+                    s0 = fpu_get_scratch_single(dyn);
+                    MSR_nzcvq_0();
+                    VMRS(x12);   // Get FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x12, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
+                    VCVT_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU exception (IO only)
+                    VMOVfrV(ed, s0);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    MOV_IMM_COND(cNE, ed, 0b10, 1);   // 0x80000000
+                    WBACK;
+                    VMSR(x12);  // put back values
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 2:
+                    INST_NAME("FIST Ed, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x12); // x1 have the modified RPSCR reg
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 4095, 0);
+                        ed = x1;
+                    }
+                    s0 = fpu_get_scratch_single(dyn);
+                    MSR_nzcvq_0();
+                    // x1 already have FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x1, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
+                    VCVTR_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU execption (invalid operation only)
+                    VMOVfrV(ed, s0);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    MOV_IMM_COND(cNE, ed, 0b10, 1);   // 0x80000000
+                    WBACK;
+                    x87_restoreround(dyn, ninst, u8);
+                    break;
+                case 3:
+                    INST_NAME("FISTP Ed, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x12); // x1 have the modified RPSCR reg
+                    if((nextop&0xC0)==0xC0) {
+                        ed = xEAX+(nextop&7);
+                        wback = 0;
+                    } else {
+                        addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 4095, 0);
+                        ed = x1;
+                    }
+                    s0 = fpu_get_scratch_single(dyn);
+                    MSR_nzcvq_0();
+                    // The FPCSR reg to clear exceptions flags
+                    ORR_IMM8(x3, x1, 0b001, 6); // enable exceptions
+                    BIC_IMM8(x3, x3, 0b10011111, 0);
+                    VMSR(x3);
+                    VCVTR_S32_F64(s0, v1);
+                    VMRS(x3);   // get the FPCSR reg and test FPU execption (denormal and overflow only)
+                    VMOVfrV(ed, s0);
+                    TSTS_IMM8_ROR(x3, 0b00000001, 0);
+                    MOV_IMM_COND(cNE, ed, 0b10, 1);   // 0x80000000
+                    WBACK;
+                    x87_do_pop(dyn, ninst);
+                    x87_restoreround(dyn, ninst, u8);
+                    break;
+                case 5:
+                    INST_NAME("FLD tbyte");
+                    x87_do_push_empty(dyn, ninst, x1);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {
+                        MOV_REG(x1, ed);
+                    }
+                    CALL(arm_fld, -1, 0);
+                    break;
+                case 7:
+                    INST_NAME("FSTP tbyte");
+                    x87_forget(dyn, ninst, x1, x3, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {
+                        MOV_REG(x1, ed);
+                    }
+                    CALL(arm_fstp, -1, 0);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_dc.c

@@ -0,0 +1,221 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2, v3;
+    int s0, s1;
+    int d0, d1;
+    int fixedaddress;
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FADD STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VADD_F64(v1, v1, v2);
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FMUL STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VMUL_F64(v1, v1, v2);
+            break;
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FCOM ST0, STx"); //yep
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            break;
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FCOMP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FSUBR STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v1, v2, v1);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FSUB STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v1, v1, v2);
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FDIVR STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v1, v2, v1);
+            break;       
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            INST_NAME("FDIV STx, ST0");
+            v2 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v1, v1, v2);
+            break;
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FADD ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VADD_F64(v1, v1, d1);
+                    break;
+                case 1:
+                    INST_NAME("FMUL ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VMUL_F64(v1, v1, d1);
+                    break;
+                case 2:
+                    INST_NAME("FCOM ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VCMP_F64(v1, d1);
+                    FCOM(x1, x2);
+                    break;
+                case 3:
+                    INST_NAME("FCOMP ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VCMP_F64(v1, d1);
+                    FCOM(x1, x2);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 4:
+                    INST_NAME("FSUB ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VSUB_F64(v1, v1, d1);
+                    break;
+                case 5:
+                    INST_NAME("FSUBR ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VSUB_F64(v1, d1, v1);
+                    break;
+                case 6:
+                    INST_NAME("FDIV ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VDIV_F64(v1, v1, d1);
+                    break;
+                case 7:
+                    INST_NAME("FDIVR ST0, double[ED]");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 1023, 3);
+                    d1 = fpu_get_scratch_double(dyn);
+                    VLDR_64(d1, wback, fixedaddress);
+                    VDIV_F64(v1, d1, v1);
+                    break;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_dd.c

@@ -0,0 +1,183 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "emu/x87emu_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    int v1, v2, v3;
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FFREE STx");
+            break;  // not handling Tag...
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+            INST_NAME("FST ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VMOV_64(v2, v1);
+            break;
+        case 0xD8:
+            INST_NAME("FSTP ST0, ST0");
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            INST_NAME("FSTP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VMOV_64(v2, v1);
+            x87_do_pop(dyn, ninst);
+            break;
+
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FUCOM ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FUCOMP ST0, STx");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst);
+            break;
+
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+        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 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            *ok = 0;
+            DEFAULT;
+            break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FLD double");
+                    v1 = x87_do_push(dyn, ninst);
+                    #if 0
+                    // can bus error...
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VLDR_64(v1, ed, fixedaddress);
+                    #else
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 255, 0);
+                    LDRD_IMM8(x2, ed, fixedaddress);
+                    VMOVtoV_D(v1, x2, x3);
+                    #endif
+                    break;
+                case 1:
+                    INST_NAME("FISTTP i64, ST0");
+                    x87_forget(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(arm_fistp64, -1, 0);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 2:
+                    INST_NAME("FST double");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VSTR_64(v1, ed, fixedaddress); // check alignment?
+                    break;
+                case 3:
+                    INST_NAME("FSTP double");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                    VSTR_64(v1, ed, fixedaddress); // check alignment?
+                    x87_do_pop(dyn, ninst);
+                    break;
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_de.c

@@ -0,0 +1,163 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2, v3;
+    int s0, s1, s2;
+    int fixedaddress;
+
+    switch(nextop) {
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+            INST_NAME("FADDP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VADD_F64(v2, v2, v1);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+            INST_NAME("FMULP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VMUL_F64(v2, v2, v1);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xD9:  /* FCOMPP */
+            INST_NAME("FCOMPP ST0, ST1");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, 1);
+            VCMP_F64(v1, v2);
+            FCOM(x1, x2);
+            x87_do_pop(dyn, ninst);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xE0:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+            INST_NAME("FSUBRP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v2, v1, v2);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FSUBP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VSUB_F64(v2, v2, v1);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FDIVRP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v2, v1, v2);
+            x87_do_pop(dyn, ninst);
+            break;
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            INST_NAME("FDIVP STx, ST0");
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VDIV_F64(v2, v2, v1);
+            x87_do_pop(dyn, ninst);
+            break;
+
+        case 0xD0:
+        case 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+        case 0xD8:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+            *ok = 0;
+            DEFAULT;
+            break;
+
+       
+        default:
+            switch((nextop>>3)&7) {
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_df.c

@@ -0,0 +1,244 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "emu/x87emu_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int v1, v2, v3;
+    int s0, s1;
+    int d0;
+    int fixedaddress;
+
+    switch(nextop) {
+        case 0xE0:
+            INST_NAME("FNSTSW AX");
+            LDR_IMM9(x2, xEmu, offsetof(x86emu_t, top));
+            LDRH_IMM8(x1, xEmu, offsetof(x86emu_t, sw));
+            AND_IMM8(x2, x2, 7);
+            BFI(x1, x2, 11, 3); // inject top
+            BFI(xEAX, x1, 0, 16);
+            break;
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xEF:
+            INST_NAME("FUCOMIP ST0, STx");
+            UFLAGS(0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOMI(x1, x2);
+            x87_do_pop(dyn, ninst);
+            UFLAGS(1);
+            break;
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+            INST_NAME("FCOMIP ST0, STx");
+            UFLAGS(0);
+            v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+            v2 = x87_get_st(dyn, ninst, x1, x2, nextop&7);
+            VCMP_F64(v1, v2);
+            FCOMI(x1, x2);
+            x87_do_pop(dyn, ninst);
+            UFLAGS(1);
+            break;
+
+        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 0xD1:
+        case 0xD2:
+        case 0xD3:
+        case 0xD4:
+        case 0xD5:
+        case 0xD6:
+        case 0xD7:
+        case 0xD8:
+        case 0xD9:
+        case 0xDA:
+        case 0xDB:
+        case 0xDC:
+        case 0xDD:
+        case 0xDE:
+        case 0xDF:
+        case 0xE1:
+        case 0xE2:
+        case 0xE3:
+        case 0xE4:
+        case 0xE5:
+        case 0xE6:
+        case 0xE7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFC:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            *ok = 0;
+            DEFAULT;
+            break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 0:
+                    INST_NAME("FILD ST0, Ew");
+                    v1 = x87_do_push(dyn, ninst);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0);
+                    LDRSH_IMM8(x1, wback, fixedaddress);
+                    s0 = fpu_get_scratch_single(dyn);
+                    VMOVtoV(s0, x1);
+                    VCVT_F64_S32(v1, s0);
+                    break;
+                case 1:
+                    INST_NAME("FISTTP Ew, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
+                    ed = x1;
+                    s0 = fpu_get_scratch_single(dyn);
+                    VCVT_S32_F64(s0, v1);
+                    VMOVfrV(ed, s0);
+                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
+                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
+                    MRS_aspr(x3);   // check Q flag
+                    TSTS_IMM8_ROR(x3, 0b10, 3);
+                    B_MARK(cEQ);    // not saturated, continue
+                    MOV32(ed, 0xffff8000);
+                    MARK;
+                    STRH_IMM8(ed, wback, fixedaddress);
+                    x87_do_pop(dyn, ninst);
+                    x87_restoreround(dyn, ninst, u8);
+                    break;
+                case 2:
+                    INST_NAME("FIST Ew, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
+                    ed = x1;
+                    s0 = fpu_get_scratch_single(dyn);
+                    VCVTR_S32_F64(s0, v1);
+                    VMOVfrV(ed, s0);
+                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
+                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
+                    MRS_aspr(x3);   // check Q flag
+                    TSTS_IMM8_ROR(x3, 0b10, 3);
+                    B_MARK(cEQ);    // not saturated, continue
+                    MOV32(ed, 0xffff8000);
+                    MARK;
+                    STRH_IMM8(ed, wback, fixedaddress);
+                    x87_restoreround(dyn, ninst, u8);
+                    break;
+                case 3:
+                    INST_NAME("FISTP Ew, ST0");
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0);
+                    u8 = x87_setround(dyn, ninst, x1, x2, x12);
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0);
+                    ed = x1;
+                    s0 = fpu_get_scratch_single(dyn);
+                    VCVTR_S32_F64(s0, v1);
+                    VMOVfrV(ed, s0);
+                    MSR_nzcvq_0();    // reset Q flag (and all other flags too)
+                    SSAT_REG_LSL_IMM5(ed, 16, ed, 0);
+                    MRS_aspr(x3);   // check Q flag
+                    TSTS_IMM8_ROR(x3, 0b10, 3);
+                    B_MARK(cEQ);    // not saturated, continue
+                    MOV32(ed, 0xffff8000);
+                    MARK;
+                    STRH_IMM8(ed, wback, fixedaddress);
+                    x87_do_pop(dyn, ninst);
+                    x87_restoreround(dyn, ninst, u8);
+                    break;
+                case 4:
+                    INST_NAME("FBLD ST0, tbytes");
+                    x87_do_push_empty(dyn, ninst, x1);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(fpu_fbld, -1, 0);
+                    break;
+                case 5: // could be inlined for most thing, but is it usefull?
+                    INST_NAME("FILD ST0, i64");
+                    x87_do_push_empty(dyn, ninst, x1);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(arm_fild64, -1, 0);
+                    break;
+                case 6:
+                    INST_NAME("FBSTP tbytes, ST0");
+                    x87_forget(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(fpu_fbst, -1, 0);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                case 7: // could be inlined for most thing, but is it usefull?
+                    INST_NAME("FISTP i64, ST0");
+                    x87_forget(dyn, ninst, x1, x2, 0);
+                    addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                    if(ed!=x1) {MOV_REG(x1, ed);}
+                    CALL(arm_fistp64, -1, 0);
+                    x87_do_pop(dyn, ninst);
+                    break;
+                default:
+                    *ok = 0;
+                    DEFAULT;
+            }
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_f0.c

@@ -0,0 +1,113 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_helper.h"
+
+
+uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = PK(0); // don't increment addr
+    uint8_t u8;
+    uint32_t u32;
+    int32_t i32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    int locked = 0;
+    switch(nextop) {
+        // generic case
+        #define GO(A)   \
+        case A+0x00:    \
+        case A+0x01:    \
+        case A+0x02:    \
+        case A+0x03:    \
+        case A+0x04:    \
+        case A+0x05:
+        GO(0x00);
+        GO(0x08);
+        GO(0x10);
+        GO(0x18);
+        GO(0x20);
+        GO(0x28);
+        GO(0x30);
+        #undef GO
+
+        case 0x81:
+        case 0x83:
+            MESSAGE(LOG_DUMP, "LOCK\n");
+            locked = 1;
+            LOCK;
+            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+
+        case 0x86:  // for this two, the lock is already done by the opcode, so just ignoring it
+        case 0x87:
+            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
+            break;
+
+        case 0x0F:
+            nextop = PK(1);
+            switch(nextop) {
+                case 0xB0:
+                case 0xB1:
+                case 0xB3:
+                case 0xBA:
+                case 0xBB:
+                case 0xC0:
+                case 0xC1:
+                case 0xC7:
+                    MESSAGE(LOG_DUMP, "LOCK\n");
+                    locked = 1;
+                    LOCK;
+                    addr = dynarec0F(dyn, addr+1, ip, ninst, ok, need_epilog);
+                    break;
+
+                default:
+                    addr = dynarec0F(dyn, addr+1, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+            }
+            break;
+
+        case 0xFF:
+            nextop = PK(1);
+            switch((nextop>>3)&7)
+            {
+                case 0:
+                case 1:
+                    MESSAGE(LOG_DUMP, "LOCK\n");
+                    locked = 1;
+                    LOCK;
+                    addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);
+                    break;
+                default:
+                addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+            }
+            break;
+       
+        default:
+            addr = dynarec00(dyn, addr, ip, ninst, ok, need_epilog);    // no lock, regular instruction...
+    }
+    if(locked) {UNLOCK;}
+
+    return addr;
+}
+

src/dynarec/dynarec_arm_f20f.c

@@ -0,0 +1,288 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+// Get Ex as a double, not a quad
+#define GETEX(a) \
+    if((nextop&0xC0)==0xC0) { \
+        a = sse_get_reg(dyn, ninst, x1, nextop&7); \
+    } else {    \
+        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3); \
+        a = fpu_get_scratch_double(dyn);            \
+        VLDR_64(a, ed, fixedaddress);               \
+    }
+
+uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t nextop = F8;
+    uint8_t u8;
+    int32_t i32, i32_;
+    int16_t i16;
+    uint32_t u32;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    uint8_t eb1, eb2;
+    int v0, v1;
+    int q0, q1;
+    int d0, d1;
+    int s0, s1;
+    int fixedaddress;
+    switch(nextop) {
+        
+        case 0x10:
+            INST_NAME("MOVSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                VMOVD(v0, d0);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                LDR_IMM9(x2, ed, fixedaddress+0);
+                LDR_IMM9(x3, ed, fixedaddress+4);
+                VMOVtoV_D(v0, x2, x3);
+                VEOR(v0+1, v0+1, v0+1); // upper 64bits set to 0
+            }
+            break;
+        case 0x11:
+            INST_NAME("MOVSD Ex, Gx");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                VMOVD(d0, v0);
+            } else {
+                VMOVfrV_D(x2, x3, v0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                STR_IMM9(x2, ed, fixedaddress+0);
+                STR_IMM9(x3, ed, fixedaddress+4);
+            }
+            break;
+        case 0x12:
+            INST_NAME("MOVDDUP Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                VMOVD(v0, d0);
+                VMOVD(v0+1, d0);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-4, 0);
+                LDR_IMM9(x2, ed, fixedaddress+0);
+                LDR_IMM9(x3, ed, fixedaddress+4);
+                VMOVtoV_D(v0, x2, x3);
+                VMOVD(v0+1, v0);
+            }
+            break;
+
+        case 0x2A:
+            INST_NAME("CVTSI2SD Gx, Ed");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETED;
+            s0 = fpu_get_scratch_single(dyn);
+            VMOVtoV(s0, ed);
+            VCVT_F64_S32(v0, s0);
+            break;
+
+        case 0x2C:
+            INST_NAME("CVTTSD2SI Gd, Ex");
+            nextop = F8;
+            GETGD;
+            s0 = fpu_get_scratch_single(dyn);
+            if((nextop&0xC0)==0xC0) {
+                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                d0 = fpu_get_scratch_double(dyn);
+                VLDR_64(d0, ed, fixedaddress);
+            }
+            VCVT_S32_F64(s0, d0);
+            VMOVfrV(gd, s0);
+            break;
+        case 0x2D:
+            INST_NAME("CVTSD2SI Gd, Ex");
+            u8 = x87_setround(dyn, ninst, x1, x2, x12);
+            nextop = F8;
+            GETGD;
+            s0 = fpu_get_scratch_single(dyn);
+            if((nextop&0xC0)==0xC0) {
+                d0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                d0 = fpu_get_scratch_double(dyn);
+                VLDR_64(d0, ed, fixedaddress);
+            }
+            VCVTR_S32_F64(s0, d0);
+            VMOVfrV(gd, s0);
+            x87_restoreround(dyn, ninst, u8);
+            break;
+
+        case 0x51:
+            INST_NAME("SQRTSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VSQRT_F64(v0, d0);
+            break;
+
+        case 0x58:
+            INST_NAME("ADDSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VADD_F64(v0, v0, d0);
+            break;
+        case 0x59:
+            INST_NAME("MULSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VMUL_F64(v0, v0, d0);
+            break;
+        case 0x5A:
+            INST_NAME("CVTSD2SS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            d1 = fpu_get_scratch_double(dyn);
+            s0 = fpu_get_scratch_single(dyn);
+            VCVT_F32_F64(s0, d0);
+            VMOV_64(d1, v0);
+            VMOV_32(d1*2, s0);
+            VMOV_64(v0, d1);
+            break;
+
+        case 0x5C:
+            INST_NAME("SUBSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VSUB_F64(v0, v0, d0);
+            break;
+        case 0x5D:
+            INST_NAME("MINSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            // MINSD: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
+            VCMP_F64(v0, d0);
+            VMRS_APSR();
+            VMOVcond_64(cPL, v0, d0);
+            break;
+        case 0x5E:
+            INST_NAME("DIVSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VDIV_F64(v0, v0, d0);
+            break;
+        case 0x5F:
+            INST_NAME("MAXSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            // MAXSD: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
+            VCMP_F64(d0, v0);
+            VMRS_APSR();
+            VMOVcond_64(cPL, v0, d0);
+            break;
+
+        case 0x70:
+            INST_NAME("PSHUFLW Gx, Ex, Ib");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                v1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(v1, ed);
+            }
+
+            u8 = F8;
+            // only low part need to be suffled. VTBL only handle 8bits value, so the 16bits suffles need to be changed in 8bits
+            u32 = 0;
+            for (int i=0; i<2; ++i) {
+                u32 |= (((u8>>(i*2))&3)*2+0)<<(i*16+0);
+                u32 |= (((u8>>(i*2))&3)*2+1)<<(i*16+8);
+            }
+            MOV32(x2, u32);
+            u32 = 0;
+            for (int i=2; i<4; ++i) {
+                u32 |= (((u8>>(i*2))&3)*2+0)<<((i-2)*16+0);
+                u32 |= (((u8>>(i*2))&3)*2+1)<<((i-2)*16+8);
+            }
+            MOV32(x3, u32);
+            d0 = fpu_get_scratch_double(dyn);
+            VMOVtoV_D(d0, x2, x3);
+            VTBL1_8(v0, v1, d0);
+            if(v0!=v1) {
+                VMOVD(v0+1, v1+1);
+            }
+            break;
+
+        case 0xC2:
+            INST_NAME("CMPSD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(d0);
+            VCMP_F64(v0, d0);
+            VMRS_APSR();
+            MOVW(x2, 0);
+            u8 = F8;
+            switch(u8&7) {
+                case 0: MVN_COND_REG_LSL_IMM8(cEQ, x2, x2, 0); break;   // Equal
+                case 1: MVN_COND_REG_LSL_IMM8(cCC, x2, x2, 0); break;   // Less than
+                case 2: MVN_COND_REG_LSL_IMM8(cLS, x2, x2, 0); break;   // Less or equal
+                case 3: MVN_COND_REG_LSL_IMM8(cVS, x2, x2, 0); break;   // NaN
+                case 4: MVN_COND_REG_LSL_IMM8(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
+                case 5: MVN_COND_REG_LSL_IMM8(cPL, x2, x2, 0); break;   // Greater or equal or unordered
+                case 6: MVN_COND_REG_LSL_IMM8(cHI, x2, x2, 0); break;   // Greater or unordered
+                case 7: MVN_COND_REG_LSL_IMM8(cVC, x2, x2, 0); break;   // not NaN
+            }
+            VMOVtoV_D(v0, x2, x2);
+            break;
+
+        default:
+            *ok = 0;
+            DEFAULT;
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_f30f.c

@@ -0,0 +1,503 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+// Get Ex as a single, not a quad, with b a intermediary double
+#define GETEX(a, b) \
+    if((nextop&0xC0)==0xC0) { \
+        b = sse_get_reg(dyn, ninst, x1, nextop&7);  \
+        if(b<16)                                    \
+            a = b;                                  \
+        else {                                      \
+            a = fpu_get_scratch_double(dyn);        \
+            VMOV_64(a, b);                          \
+        }                                           \
+        a *= 2;                                     \
+    } else {    \
+        addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3); \
+        a = fpu_get_scratch_single(dyn);            \
+        VLDR_32(a, ed, fixedaddress);               \
+    }
+
+uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog)
+{
+    uint8_t opcode = F8;
+    uint8_t nextop, u8;
+    int32_t i32, i32_;
+    uint32_t u32;
+    int16_t i16;
+    uint16_t u16;
+    uint8_t gd, ed;
+    uint8_t wback, wb1, wb2;
+    uint8_t eb1, eb2;
+    int fixedaddress;
+    int v0, v1;
+    int s0, s1;
+    int d0, d1;
+    int q0, q1;
+    switch(opcode) {
+
+        case 0x10:
+            INST_NAME("MOVSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                v0 = sse_get_reg(dyn, ninst, x1, gd);
+                q0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOVD(d0, v0);
+                }
+                if(q0<16)
+                    d1 = q0;
+                else {
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVD(d1, q0);
+                }
+                VMOV_32(d0*2, d1*2);
+                if(v0!=d0) {
+                    VMOVD(v0, d0);
+                }
+            } else {
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                LDR_IMM9(x2, ed, fixedaddress);   // to avoid bus errors
+                VEORQ(v0, v0, v0);
+                VMOVtoDx_32(v0, 0, x2);
+            }
+            break;
+        case 0x11:
+            INST_NAME("MOVSS Ex, Gx");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                q0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOVD(d0, v0);
+                }
+                if(q0<16)
+                    d1 = q0;
+                else {
+                    d1 = fpu_get_scratch_double(dyn);
+                    VMOVD(d1, q0);
+                }
+                VMOV_32(d1*2, d0*2);
+                if(q0!=d1) {
+                    VMOVD(q0, d1);
+                }
+            } else {
+                VMOVfrDx_32(x2, v0, 0);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095, 0);
+                STR_IMM9(x2, ed, fixedaddress);
+            }
+            break;
+
+        case 0x2A:
+            INST_NAME("CVTSI2SS Gx, Ed");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETED;
+            if(v0<16)
+                d0 = v0;
+            else {
+                d0 = fpu_get_scratch_double(dyn);
+                VMOV_64(d0, v0);
+            }
+            s0 = d0*2;
+            VMOVtoV(s0, ed);
+            VCVT_F32_S32(s0, s0);
+            if(d0!=v0) {
+                VMOV_64(v0, d0);
+            }
+            break;
+
+        case 0x2C:
+            INST_NAME("CVTTSS2SI Gd, Ex");
+            nextop = F8;
+            GETGD;
+            s0 = fpu_get_scratch_single(dyn);
+            if((nextop&0xC0)==0xC0) {
+                v0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+                VMOV_32(s0, d0*2);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                VLDR_32(s0, ed, fixedaddress);
+            }
+            VCVT_S32_F32(s0, s0);
+            VMOVfrV(gd, s0);
+            break;
+        case 0x2D:
+            INST_NAME("CVTSS2SI Gd, Ex");
+            u8 = x87_setround(dyn, ninst, x1, x2, x12);
+            nextop = F8;
+            GETGD;
+            s0 = fpu_get_scratch_single(dyn);
+            if((nextop&0xC0)==0xC0) {
+                v0 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                if(v0<16)
+                    d0 = v0;
+                else {
+                    d0 = fpu_get_scratch_double(dyn);
+                    VMOV_64(d0, v0);
+                }
+                VMOV_32(s0, d0*2);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 1023, 3);
+                VLDR_32(s0, ed, fixedaddress);
+            }
+            VCVTR_S32_F32(s0, s0);
+            VMOVfrV(gd, s0);
+            x87_restoreround(dyn, ninst, u8);
+            break;
+
+        case 0x51:
+            INST_NAME("SQRTSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VSQRT_F32(d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x52:
+            INST_NAME("RSQRTSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            s1 = fpu_get_scratch_single(dyn);
+            // so here: F32: Imm8 = abcd efgh that gives => aBbbbbbc defgh000 00000000 00000000
+            // and want 1.0f = 0x3f800000
+            // so 00111111 10000000 00000000 00000000
+            // a = 0, b = 1, c = 1, d = 1, efgh=0
+            // 0b01110000
+            VMOV_i_32(s1, 0b01110000);
+            VDIV_F32(s0, s1, s0);
+            VSQRT_F32(d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x53:
+            INST_NAME("RCPSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            s1 = fpu_get_scratch_single(dyn);
+            VMOV_i_32(s1, 0b01110000);
+            VDIV_F32(d1*2, s1, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+
+        case 0x58:
+            INST_NAME("ADDSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VADD_F32(d1*2, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x59:
+            INST_NAME("MULSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VMUL_F32(d1*2, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x5A:
+            INST_NAME("CVTSS2SD Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            VCVT_F64_F32(v0, s0);
+            break;
+        case 0x5B:
+            INST_NAME("CVTTPS2DQ Gx, Ex");
+            nextop = F8;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                v1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(v1, ed);
+            }
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+            VCVTQ_S32_F32(v0, v1);
+            break;
+        case 0x5C:
+            INST_NAME("SUBSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VSUB_F32(d1*2, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x5D:
+            INST_NAME("MINSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            // MINSS: if any input is NaN, or Ex[0]<Gx[0], copy Ex[0] -> Gx[0]
+            VCMP_F32(d1*2, s0);
+            VMRS_APSR();
+            VMOVcond_32(cPL, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x5E:
+            INST_NAME("DIVSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VDIV_F32(d1*2, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+        case 0x5F:
+            INST_NAME("MAXSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            // MAXSS: if any input is NaN, or Ex[0]>Gx[0], copy Ex[0] -> Gx[0]
+            VCMP_F32(s0, d1*2);
+            VMRS_APSR();
+            VMOVcond_32(cPL, d1*2, s0);
+            if(v0!=d1) {
+                VMOV_64(v0, d1);
+            }
+            break;
+
+        case 0x6F:
+            INST_NAME("MOVDQU Gx,Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                VMOVQ(v0, v1);
+            } else {
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-12, 0);
+                LDR_IMM9(x2, ed, fixedaddress+0);
+                LDR_IMM9(x3, ed, fixedaddress+4);
+                VMOVtoV_D(v0, x2, x3);
+                LDR_IMM9(x2, ed, fixedaddress+8);
+                LDR_IMM9(x3, ed, fixedaddress+12);
+                VMOVtoV_D(v0+1, x2, x3);
+            }
+            break;
+        case 0x70:
+            INST_NAME("PSHUFHW Gx, Ex, Ib");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                v1 = fpu_get_scratch_quad(dyn);
+                VLD1Q_64(v1, ed);
+            }
+            u8 = F8;
+            // only high part need to be suffled. VTBL only handle 8bits value, so the 16bits suffles need to be changed in 8bits
+            u32 = 0;
+            for (int i=0; i<2; ++i) {
+                u32 |= (((u8>>(i*2))&3)*2+0)<<(i*16+0);
+                u32 |= (((u8>>(i*2))&3)*2+1)<<(i*16+8);
+            }
+            MOV32(x2, u32);
+            u32 = 0;
+            for (int i=2; i<4; ++i) {
+                u32 |= (((u8>>(i*2))&3)*2+0)<<((i-2)*16+0);
+                u32 |= (((u8>>(i*2))&3)*2+1)<<((i-2)*16+8);
+            }
+            MOV32(x3, u32);
+            d0 = fpu_get_scratch_double(dyn);
+            VMOVtoV_D(d0, x2, x3);
+            VTBL1_8(v0+1, v1+1, d0);
+            if(v0!=v1) {
+                VMOVD(v0, v1);
+            }
+            break;
+
+
+        case 0x7E:
+            INST_NAME("MOVQ Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg(dyn, ninst, x1, nextop&7);
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                VMOVD(v0, v1);
+            } else {
+                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0);
+                VLD1_64(v0, ed);
+            }
+            VEOR(v0+1, v0+1, v0+1);
+            break;
+        case 0x7F:
+            INST_NAME("MOVDQU Ex,Gx");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
+                v0 = sse_get_reg(dyn, ninst, x1, gd);
+                VMOVQ(v1, v0);
+            } else {
+                v0 = sse_get_reg(dyn, ninst, x1, gd);
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 4095-12, 0);
+                VMOVfrV_D(x2, x3, v0);
+                STR_IMM9(x2, ed, fixedaddress+0);
+                STR_IMM9(x3, ed, fixedaddress+4);
+                VMOVfrV_D(x2, x3, v0+1);
+                STR_IMM9(x2, ed, fixedaddress+8);
+                STR_IMM9(x3, ed, fixedaddress+12);
+            }
+            break;
+
+        case 0xC2:
+            INST_NAME("CMPSS Gx, Ex");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd);
+            GETEX(s0, d0);
+            if(v0<16)
+                d1 = v0;
+            else {
+                d1 = fpu_get_scratch_double(dyn);
+                VMOV_64(d1, v0);
+            }
+            VCMP_F32(d1*2, s0);
+            VMRS_APSR();
+            MOVW(x2, 0);
+            u8 = F8;
+            switch(u8&7) {
+                case 0: MVN_COND_REG_LSL_IMM8(cEQ, x2, x2, 0); break;   // Equal
+                case 1: MVN_COND_REG_LSL_IMM8(cCC, x2, x2, 0); break;   // Less than
+                case 2: MVN_COND_REG_LSL_IMM8(cLS, x2, x2, 0); break;   // Less or equal
+                case 3: MVN_COND_REG_LSL_IMM8(cVS, x2, x2, 0); break;   // NaN
+                case 4: MVN_COND_REG_LSL_IMM8(cNE, x2, x2, 0); break;   // Not Equal (or unordered on ARM, not on X86...)
+                case 5: MVN_COND_REG_LSL_IMM8(cPL, x2, x2, 0); break;   // Greater or equal or unordered
+                case 6: MVN_COND_REG_LSL_IMM8(cHI, x2, x2, 0); break;   // Greater or unordered
+                case 7: MVN_COND_REG_LSL_IMM8(cVC, x2, x2, 0); break;   // not NaN
+            }
+            VMOVtoDx_32(v0, 0, x2);
+            break;
+
+        default:
+            *ok = 0;
+            DEFAULT;
+    }
+    return addr;
+}
+

src/dynarec/dynarec_arm_functions.c

@@ -0,0 +1,290 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <errno.h>
+#include <string.h>
+#include <math.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "tools/bridge_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "emu/x86run_private.h"
+#include "emu/x87emu_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "dynarec_arm_functions.h"
+
+void arm_popf(x86emu_t* emu, uint32_t f)
+{
+    emu->packed_eflags.x32 = (f & 0x3F7FD7) | 0x2; // mask off res2 and res3 and on res1
+    UnpackFlags(emu);
+    RESET_FLAGS(emu);
+}
+
+void arm_fstp(x86emu_t* emu, void* p)
+{
+    if(ST0.ll!=STld(0).ref)
+        D2LD(&ST0.d, p);
+    else
+        memcpy(p, &STld(0).ld, 10);
+}
+
+void arm_print_armreg(x86emu_t* emu, uintptr_t reg, uintptr_t n)
+{
+    dynarec_log(LOG_DEBUG, "R%d=0x%x (%d)\n", n, reg, reg);
+}
+
+void arm_f2xm1(x86emu_t* emu)
+{
+    ST0.d = exp2(ST0.d) - 1.0;
+}
+void arm_fyl2x(x86emu_t* emu)
+{
+    ST(1).d = log2(ST0.d)*ST(1).d;
+}
+void arm_ftan(x86emu_t* emu)
+{
+    ST0.d = tan(ST0.d);
+}
+void arm_fpatan(x86emu_t* emu)
+{
+    ST1.d = atan2(ST1.d, ST0.d);
+}
+void arm_fxtract(x86emu_t* emu)
+{
+    int32_t tmp32s = (ST1.ll&0x7ff0000000000000LL)>>52;
+    tmp32s -= 1023;
+    ST1.d /= exp2(tmp32s);
+    ST0.d = tmp32s;
+}
+void arm_fprem(x86emu_t* emu)
+{
+    int32_t tmp32s = ST0.d / ST1.d;
+    ST0.d -= ST1.d * tmp32s;
+    emu->sw.f.F87_C2 = 0;
+    emu->sw.f.F87_C0 = (tmp32s&1);
+    emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
+    emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
+}
+void arm_fyl2xp1(x86emu_t* emu)
+{
+    ST(1).d = log2(ST0.d + 1.0)*ST(1).d;
+}
+void arm_fsincos(x86emu_t* emu)
+{
+    sincos(ST1.d, &ST1.d, &ST0.d);
+}
+void arm_frndint(x86emu_t* emu)
+{
+    ST0.d = fpu_round(emu, ST0.d);
+}
+void arm_fscale(x86emu_t* emu)
+{
+    ST0.d *= exp2(trunc(ST1.d));
+}
+void arm_fsin(x86emu_t* emu)
+{
+    ST0.d = sin(ST0.d);
+}
+void arm_fcos(x86emu_t* emu)
+{
+    ST0.d = cos(ST0.d);
+}
+
+void arm_fbld(x86emu_t* emu, uint8_t* ed)
+{
+    fpu_fbld(emu, ed);
+}
+
+void arm_fild64(x86emu_t* emu, int64_t* ed)
+{
+    ST0.d = *ed;
+    STll(0).ll = *ed;
+    STll(0).ref = ST0.ll;
+}
+
+void arm_fbstp(x86emu_t* emu, uint8_t* ed)
+{
+    fpu_fbst(emu, ed);
+}
+
+void arm_fistp64(x86emu_t* emu, int64_t* ed)
+{
+    if(STll(0).ref==ST(0).ll) {
+        *ed = STll(0).ll;
+    } else {
+        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL))
+            *ed = 0x8000000000000000LL;
+        else
+            *ed = (int64_t)ST0.d;
+    }
+}
+
+void arm_fld(x86emu_t* emu, uint8_t* ed)
+{
+    memcpy(&STld(0).ld, ed, 10);
+    LD2D(&STld(0), &ST(0).d);
+    STld(0).ref = ST0.ll;
+}
+
+void arm_cpuid(x86emu_t* emu, uint32_t tmp32u)
+{
+    switch(tmp32u) {
+        case 0x0:
+            // emulate a P4
+            R_EAX = 0x80000004;
+            // return GenuineIntel
+            R_EBX = 0x756E6547;
+            R_EDX = 0x49656E69;
+            R_ECX = 0x6C65746E;
+            break;
+        case 0x1:
+            R_EAX = 0x00000101; // familly and all
+            R_EBX = 0;          // Brand indexe, CLFlush, Max APIC ID, Local APIC ID
+            R_EDX =   1         // fpu 
+                    | 1<<8      // cmpxchg8
+                    | 1<<11     // sep (sysenter & sysexit)
+                    | 1<<15     // cmov
+                    | 1<<19     // clflush (seems to be with SSE2)
+                    | 1<<23     // mmx
+                    //| 1<<24     // fxsr (fxsave, fxrestore)
+                    | 1<<25     // SSE
+                    | 1<<26     // SSE2
+                    ;
+            R_ECX =   1<<12     // fma
+                    | 1<<13     // cx16 (cmpxchg16)
+                    ;           // also 1<<0 is SSE3 and 1<<9 is SSSE3
+            break;
+        case 0x2:   // TLB and Cache info. Sending 1st gen P4 info...
+            R_EAX = 0x665B5001;
+            R_EBX = 0x00000000;
+            R_ECX = 0x00000000;
+            R_EDX = 0x007A7000;
+            break;
+        
+        case 0x4:   // Cache info
+            switch (R_ECX) {
+                case 0: // L1 data cache
+                    R_EAX = (1 | (1<<5) | (1<<8));   //type
+                    R_EBX = (63 | (7<<22)); // size
+                    R_ECX = 63;
+                    R_EDX = 1;
+                    break;
+                case 1: // L1 inst cache
+                    R_EAX = (2 | (1<<5) | (1<<8)); //type
+                    R_EBX = (63 | (7<<22)); // size
+                    R_ECX = 63;
+                    R_EDX = 1;
+                    break;
+                case 2: // L2 cache
+                    R_EAX = (3 | (2<<5) | (1<<8)); //type
+                    R_EBX = (63 | (15<<22));    // size
+                    R_ECX = 4095;
+                    R_EDX = 1;
+                    break;
+
+                default:
+                    R_EAX = 0x00000000;
+                    R_EBX = 0x00000000;
+                    R_ECX = 0x00000000;
+                    R_EDX = 0x00000000;
+                    break;
+            }
+            break;
+        case 0x5:   //mwait info
+            R_EAX = 0;
+            R_EBX = 0;
+            R_ECX = 1 | 2;
+            R_EDX = 0;
+            break;
+        case 0x6:   // thermal
+        case 0x9:   // direct cache access
+        case 0xA:   // Architecture performance monitor
+            R_EAX = 0;
+            R_EBX = 0;
+            R_ECX = 0;
+            R_EDX = 0;
+            break;
+        case 0x7:   // extended bits...
+            /*if(R_ECX==0)    R_EAX = 0;
+            else*/ R_EAX = R_ECX = R_EBX = R_EDX = 0;
+            break;
+
+        case 0x80000000:        // max extended
+            break;              // no extended, so return same value as beeing the max value!
+        default:
+            printf_log(LOG_INFO, "Warning, CPUID command %X unsupported (ECX=%08x)\n", tmp32u, R_ECX);
+            R_EAX = 0;
+    }   
+}
+
+// Get a FPU single scratch reg
+int fpu_get_scratch_single(dynarec_arm_t* dyn)
+{
+    return dyn->fpu_scratch++;  // return an Sx
+}
+// Get a FPU double scratch reg
+int fpu_get_scratch_double(dynarec_arm_t* dyn)
+{
+    int i = (dyn->fpu_scratch+1)&(~1);
+    dyn->fpu_scratch = i+2;
+    return i/2; // return a Dx
+}
+// Get a FPU quad scratch reg
+int fpu_get_scratch_quad(dynarec_arm_t* dyn)
+{
+    int i = (dyn->fpu_scratch+3)&(~3);
+    dyn->fpu_scratch = i+4;
+    return i/2; // Return a Dx, not a Qx
+}
+// Reset scratch regs counter
+int fpu_reset_scratch(dynarec_arm_t* dyn)
+{
+    dyn->fpu_scratch = 0;
+}
+#define FPUFIRST    8
+// Get a FPU double reg
+int fpu_get_reg_double(dynarec_arm_t* dyn)
+{
+    // TODO: check upper limit?
+    int i=0;
+    while (dyn->fpuused[i]) ++i;
+    dyn->fpuused[i] = 1;
+    return i+FPUFIRST; // return a Dx
+}
+// Free a FPU double reg
+void fpu_free_reg_double(dynarec_arm_t* dyn, int reg)
+{
+    // TODO: check upper limit?
+    int i=reg-FPUFIRST;
+    dyn->fpuused[i] = 0;
+}
+// Get a FPU quad reg
+int fpu_get_reg_quad(dynarec_arm_t* dyn)
+{
+    int i=0;
+    while (dyn->fpuused[i] || dyn->fpuused[i+1]) i+=2;
+    dyn->fpuused[i] = dyn->fpuused[i+1] = 1;
+    return i+FPUFIRST; // Return a Dx, not a Qx
+}
+// Free a FPU quad reg
+void fpu_free_reg_quad(dynarec_arm_t* dyn, int reg)
+{
+    int i=reg-FPUFIRST;
+    dyn->fpuused[i] = dyn->fpuused[i+1] = 0;
+}
+// Reset fpu regs counter
+int fpu_reset_reg(dynarec_arm_t* dyn)
+{
+    dyn->fpu_reg = 0;
+    for (int i=0; i<24; ++i)
+        dyn->fpuused[i]=0;
+}

src/dynarec/dynarec_arm_functions.h

@@ -0,0 +1,50 @@
+#ifndef __DYNAREC_ARM_FUNCTIONS_H__
+#define __DYNAREC_ARM_FUNCTIONS_H__
+
+typedef struct x86emu_s x86emu_t;
+
+void arm_popf(x86emu_t* emu, uint32_t f);
+void arm_fstp(x86emu_t* emu, void* p);
+
+void arm_print_armreg(x86emu_t* emu, uintptr_t reg, uintptr_t n);
+
+void arm_f2xm1(x86emu_t* emu);
+void arm_fyl2x(x86emu_t* emu);
+void arm_ftan(x86emu_t* emu);
+void arm_fpatan(x86emu_t* emu);
+void arm_fxtract(x86emu_t* emu);
+void arm_fprem(x86emu_t* emu);
+void arm_fyl2xp1(x86emu_t* emu);
+void arm_fsincos(x86emu_t* emu);
+void arm_frndint(x86emu_t* emu);
+void arm_fscale(x86emu_t* emu);
+void arm_fsin(x86emu_t* emu);
+void arm_fcos(x86emu_t* emu);
+void arm_fbld(x86emu_t* emu, uint8_t* ed);
+void arm_fild64(x86emu_t* emu, int64_t* ed);
+void arm_fbstp(x86emu_t* emu, uint8_t* ed);
+void arm_fistp64(x86emu_t* emu, int64_t* ed);
+void arm_fld(x86emu_t* emu, uint8_t* ed);
+
+void arm_cpuid(x86emu_t* emu, uint32_t tmp32u);
+
+// Get an FPU single scratch reg
+int fpu_get_scratch_single(dynarec_arm_t* dyn);
+// Get an FPU double scratch reg
+int fpu_get_scratch_double(dynarec_arm_t* dyn);
+// Get an FPU quad scratch reg
+int fpu_get_scratch_quad(dynarec_arm_t* dyn);
+// Reset scratch regs counter
+int fpu_reset_scratch(dynarec_arm_t* dyn);
+// Get an FPU double reg
+int fpu_get_reg_double(dynarec_arm_t* dyn);
+// Free a FPU double reg
+void fpu_free_reg_double(dynarec_arm_t* dyn, int reg);
+// Get an FPU quad reg
+int fpu_get_reg_quad(dynarec_arm_t* dyn);
+// Free a FPU quad reg
+void fpu_free_reg_quad(dynarec_arm_t* dyn, int reg);
+// Reset fpu regs counter
+int fpu_reset_reg(dynarec_arm_t* dyn);
+
+#endif //__DYNAREC_ARM_FUNCTIONS_H__

src/dynarec/dynarec_arm_helper.h

@@ -0,0 +1,447 @@
+#ifndef __DYNAREC_ARM_HELPER_H__
+#define __DYNAREC_ARM_HELPER_H__
+
+#if STEP == 0
+#include "dynarec_arm_pass0.h"
+#elif STEP == 1
+#include "dynarec_arm_pass1.h"
+#elif STEP == 2
+#include "dynarec_arm_pass2.h"
+#elif STEP == 3
+#include "dynarec_arm_pass3.h"
+#endif
+
+#include "debug.h"
+#include "arm_emitter.h"
+#include "../emu/x86primop.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 PK(a)   *(uint8_t*)(addr+a)
+#define PK32(a)   *(uint32_t*)(addr+a)
+#define PK16(a)   *(uint16_t*)(addr+a)
+#define PKip(a)   *(uint8_t*)(ip+a)
+
+// GETGD    get x86 register in gd
+#define GETGD   gd = xEAX+((nextop&0x38)>>3)
+//GETED can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI
+#define GETED   if((nextop&0xC0)==0xC0) {   \
+                    ed = xEAX+(nextop&7);   \
+                    wback = 0;              \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 4095, 0); \
+                    LDR_IMM9(x1, wback, fixedaddress); \
+                    ed = x1;                \
+                }
+//GETEDH can use hint for ed, and r1 or r2 for wback (depending on hint). wback is 0 if ed is xEAX..xEDI
+#define GETEDH(hint)   if((nextop&0xC0)==0xC0) {   \
+                    ed = xEAX+(nextop&7);   \
+                    wback = 0;              \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, (hint==x2)?x1:x2, &fixedaddress, 4095, 0); \
+                    LDR_IMM9(hint, wback, fixedaddress); \
+                    ed = hint;              \
+                }
+//GETEDW can use hint for wback and ret for ed. wback is 0 if ed is xEAX..xEDI
+#define GETEDW(hint, ret)   if((nextop&0xC0)==0xC0) {   \
+                    ed = xEAX+(nextop&7);   \
+                    MOV_REG(ret, ed);       \
+                    wback = 0;              \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, hint, &fixedaddress, 4095, 0); \
+                    ed = ret;               \
+                    LDR_IMM9(ed, wback, fixedaddress); \
+                }
+// Write back ed in wback (if wback not 0)
+#define WBACK       if(wback) {STR_IMM9(ed, wback, fixedaddress);}
+// Send back wb to either ed or wback
+#define SBACK(wb)   if(wback) {STR_IMM9(wb, wback, fixedaddress);} else {MOV_REG(ed, wb);}
+//GETEDO can use r1 for ed, and r2 for wback. wback is 0 if ed is xEAX..xEDI
+#define GETEDO(O)   if((nextop&0xC0)==0xC0) {   \
+                    ed = xEAX+(nextop&7);   \
+                    wback = 0;              \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 0, 0); \
+                    LDR_REG_LSL_IMM5(x1, wback, O, 0);  \
+                    ed = x1;                 \
+                }
+#define WBACKO(O)   if(wback) {STR_REG_LSL_IMM5(ed, wback, O, 0);}
+//FAKEELike GETED, but doesn't get anything
+#define FAKEED  if((nextop&0xC0)!=0xC0) {   \
+                    addr = fakeed(dyn, addr, ninst, nextop); \
+                }
+// GETGW extract x86 register in gd, that is i
+#define GETGW(i) gd = xEAX+((nextop&0x38)>>3); UXTH(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);\
+                    UXTH(i, wback, 0);      \
+                    ed = i;                 \
+                    wb1 = 0;                \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, w, &fixedaddress, 255, 0); \
+                    LDRH_IMM8(i, wback, fixedaddress); \
+                    ed = i;                 \
+                    wb1 = 1;                \
+                }
+//GETEW will use i for ed, and can use r3 for wback.
+#define GETEW(i) if((nextop&0xC0)==0xC0) {  \
+                    wback = xEAX+(nextop&7);\
+                    UXTH(i, wback, 0);      \
+                    ed = i;                 \
+                    wb1 = 0;                \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0); \
+                    LDRH_IMM8(i, wback, fixedaddress); \
+                    ed = i;                 \
+                    wb1 = 1;                \
+                }
+//GETSEW will use i for ed, and can use r3 for wback. This is the Signed version
+#define GETSEW(i) if((nextop&0xC0)==0xC0) {  \
+                    wback = xEAX+(nextop&7);\
+                    SXTH(i, wback, 0);      \
+                    ed = i;                 \
+                    wb1 = 0;                \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0); \
+                    LDRSH_IMM8(i, wback, fixedaddress);\
+                    ed = i;                 \
+                    wb1 = 1;                \
+                }
+// Write ed back to original register / memory
+#define EWBACK   if(wb1) {STRH_IMM8(ed, wback, fixedaddress);} else {BFI(wback, ed, 0, 16);}
+// Write w back to original register / memory
+#define EWBACKW(w)   if(wb1) {STRH_IMM8(w, wback, fixedaddress);} else {BFI(wback, w, 0, 16);}
+// Write back gd in correct register
+#define GWBACK       BFI((xEAX+((nextop&0x38)>>3)), gd, 0, 16);
+//GETEB will use i for ed, and can use r3 for wback.
+#define GETEB(i) if((nextop&0xC0)==0xC0) {  \
+                    wback = (nextop&7);     \
+                    wb2 = (wback>>2);       \
+                    wback = xEAX+(wback&3); \
+                    UXTB(i, wback, wb2);    \
+                    wb1 = 0;                \
+                    ed = i;                 \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 4095, 0); \
+                    LDRB_IMM9(i, wback, fixedaddress); \
+                    wb1 = 1;                \
+                    ed = i;                 \
+                }
+//GETSEB sign extend EB, will use i for ed, and can use r3 for wback.
+#define GETSEB(i) if((nextop&0xC0)==0xC0) { \
+                    wback = (nextop&7);     \
+                    wb2 = (wback>>2);       \
+                    wback = xEAX+(wback&3); \
+                    SXTB(i, wback, wb2);    \
+                    wb1 = 0;                \
+                    ed = i;                 \
+                } else {                    \
+                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0); \
+                    LDRSB_IMM8(i, wback, fixedaddress);\
+                    wb1 = 1;                \
+                    ed = i;                 \
+                }
+// Write eb (ed) back to original register / memory
+#define EBBACK   if(wb1) {STRB_IMM9(ed, wback, fixedaddress);} else {BFI(wback, ed, wb2*8, 8);}
+//GETGB will use i for gd
+#define GETGB(i)    gd = (nextop&0x38)>>3;  \
+                    gb2 = ((gd&4)>>2);      \
+                    gb1 = xEAX+(gd&3);      \
+                    gd = i;                 \
+                    UXTB(gd, gb1, gb2);
+//GETSGB signe extend GB, will use i for gd
+#define GETSGB(i)    gd = (nextop&0x38)>>3; \
+                    gb2 = ((gd&4)>>2);      \
+                    gb1 = xEAX+(gd&3);      \
+                    gd = i;                 \
+                    SXTB(gd, gb1, gb2);
+// Write gb (gd) back to original register / memory
+#define GBBACK   BFI(gb1, gd, gb2*8, 8);
+
+// Get Direction with size Z and based of F_DF flag, on register r ready for LDR/STR fetching
+#define GETDIR(r, A)    \
+    LDR_IMM9(r, xEmu, offsetof(x86emu_t, flags[F_DF]));     \
+    CMPS_IMM8(r, 1);                                        \
+    MOVW(r, A);                                             \
+    RSB_COND_IMM8(cEQ, r, r, 0)
+
+// CALL will use x12 for the call address. Return value can be put in ret (unless ret is -1)
+#define CALL(F, ret, M) call_c(dyn, ninst, F, x12, ret, M)
+// CALL_ will use x3 for the call address. Return value can be put in ret (unless ret is -1)
+#define CALL_(F, ret, M) call_c(dyn, ninst, F, x3, ret, M)
+#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))
+
+// Branch to MARK if cond (use i32)
+#define B_MARK(cond)    \
+    i32 = GETMARK-(dyn->arm_size+8);    \
+    Bcond(cond, i32)
+// Branch to MARK2 if cond (use i32)
+#define B_MARK2(cond)    \
+    i32 = GETMARK2-(dyn->arm_size+8);   \
+    Bcond(cond, i32)
+// Branch to MARK3 if cond (use i32)
+#define B_MARK3(cond)    \
+    i32 = GETMARK3-(dyn->arm_size+8);   \
+    Bcond(cond, i32)
+// Branch to next instruction if cond (use i32)
+#define B_NEXT(cond)     \
+    i32 = (dyn->insts)?(dyn->insts[ninst].epilog-(dyn->arm_size+8)):0; \
+    Bcond(cond, i32)
+
+// Generate FCOM with s1 and s2 scratch regs (the VCMP is already done)
+#define FCOM(s1, s2)    \
+    VMRS_APSR();    /* 0b0100011100000000 */                                    \
+    LDRH_IMM8(s2, xEmu, offsetof(x86emu_t, sw));   /*offset is 8bits right?*/   \
+    MOVW(s1, 0b0100011100000000);                                               \
+    BIC_REG_LSL_IMM8(s2, s2, s1, 0);                                            \
+    MOVW_COND(cVS, s1, 0b0100010100000000); /* unordered */                     \
+    MOVW_COND(cEQ, s1, 0b0100000000000000); /* zero */                          \
+    MOVW_COND(cGT, s1, 0b0000000000000000); /* greater than */                  \
+    MOVW_COND(cLO, s1, 0b0000000100000000); /* less than */                     \
+    ORR_REG_LSL_IMM8(s2, s2, s1, 0);                                            \
+    STRH_IMM8(s2, xEmu, offsetof(x86emu_t, sw))
+
+// Generate FCOMI with s1 and s2 scratch regs (the VCMP is already done)
+#define FCOMI(s1, s2)    \
+    VMRS_APSR();    /* 0b111 */                             \
+    XOR_REG_LSL_IMM8(s2, s2, s2, 0);                        \
+    MOVW_COND(cVS, s1, 0b111); /* unordered */              \
+    MOVW_COND(cEQ, s1, 0b100); /* zero */                   \
+    MOVW_COND(cGT, s1, 0b000); /* greater than */           \
+    MOVW_COND(cLO, s1, 0b001); /* less than */              \
+    UBFX(s2, s1, 0, 1);                                     \
+    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_CF]));    \
+    UBFX(s2, s1, 1, 1);                                     \
+    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_PF]));    \
+    UBFX(s2, s1, 2, 1);                                     \
+    STR_IMM9(s2, xEmu, offsetof(x86emu_t, flags[F_ZF]));    \
+    MOVW(s2, d_none);                                       \
+    STR_IMM9(s2, xEmu, offsetof(x86emu_t, df));
+
+
+#ifndef UFLAGS
+#define UFLAGS(A)  dyn->cleanflags=A
+#endif
+#ifndef USEFLAG
+// USEFLAG will check status of defered flags the call update flags if needed. x3 will be used
+#define USEFLAG(A)   \
+    if(!dyn->cleanflags) {  \
+        if(A) {             \
+            LDR_IMM9(x3, xEmu, offsetof(x86emu_t, df)); \
+            TSTS_REG_LSL_IMM8(x3, x3, 0);               \
+            i32 = (GETMARKF)-(dyn->arm_size+8);         \
+            Bcond(cEQ, i32);                            \
+            CALL_(UpdateFlags, -1, 0);                  \
+            MARKF;                                      \
+            dyn->cleanflags=1;                          \
+        }                   \
+    }
+#endif
+#ifndef JUMP
+#define JUMP(A) 
+#endif
+#define BARRIER(A) if (dyn->insts && !dyn->insts[ninst].x86.barrier) dyn->insts[ninst].x86.barrier = A
+#define UFLAG_OP1(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, op1));}
+#define UFLAG_OP2(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, op2));}
+#define UFLAG_OP12(A1, A2) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A1, 0, offsetof(x86emu_t, op1));STR_IMM9(A2, 0, offsetof(x86emu_t, op2));}
+#define UFLAG_RES(A) if(dyn->insts && dyn->insts[ninst].x86.flags) {STR_IMM9(A, 0, offsetof(x86emu_t, res));}
+#define UFLAG_DF(r, A) if(dyn->insts && dyn->insts[ninst].x86.flags) {MOVW(r, A); STR_IMM9(r, 0, offsetof(x86emu_t, df));}
+#define UFLAG_IF if(dyn->insts && dyn->insts[ninst].x86.flags)
+#ifndef DEFAULT
+#define DEFAULT      BARRIER(2)
+#endif
+// Emit the LOCK mutex (x1, x2 and x3 are lost)
+#define LOCK        emit_lock(dyn, addr, ninst)
+// Emit the UNLOCK mutex (x1, x2 and x3 are lost)
+#define UNLOCK      emit_unlock(dyn, addr, ninst)
+
+
+void arm_epilog();
+void* arm_linker(x86emu_t* emu, void** table, uintptr_t addr);
+
+#ifndef STEPNAME
+#define STEPNAME3(N,M) N##M
+#define STEPNAME2(N,M) STEPNAME3(N,M)
+#define STEPNAME(N) STEPNAME2(N, STEP)
+#endif
+
+#define arm_pass        STEPNAME(arm_pass)
+
+#define dynarec00       STEPNAME(dynarec00)
+#define dynarec0F       STEPNAME(dynarec0F)
+#define dynarecGS       STEPNAME(dynarecGS)
+#define dynarec66       STEPNAME(dynarec66)
+#define dynarec67       STEPNAME(dynarec67)
+#define dynarecD8       STEPNAME(dynarecD8)
+#define dynarecD9       STEPNAME(dynarecD9)
+#define dynarecDA       STEPNAME(dynarecDA)
+#define dynarecDB       STEPNAME(dynarecDB)
+#define dynarecDC       STEPNAME(dynarecDC)
+#define dynarecDD       STEPNAME(dynarecDD)
+#define dynarecDE       STEPNAME(dynarecDE)
+#define dynarecDF       STEPNAME(dynarecDF)
+#define dynarecF0       STEPNAME(dynarecF0)
+#define dynarec660F     STEPNAME(dynarec660F)
+#define dynarecF20F     STEPNAME(dynarecF20F)
+#define dynarecF30F     STEPNAME(dynarecF30F)
+
+#define geted           STEPNAME(geted_)
+#define fakeed          STEPNAME(fakeed_)
+#define jump_to_epilog  STEPNAME(jump_to_epilog_)
+#define jump_to_linker  STEPNAME(jump_to_linker_)
+#define ret_to_epilog   STEPNAME(ret_to_epilog_)
+#define retn_to_epilog  STEPNAME(retn_to_epilog_)
+#define call_c          STEPNAME(call_c_)
+#define grab_tlsdata    STEPNAME(grab_tlsdata_)
+#define isNativeCall    STEPNAME(isNativeCall_)
+#define emit_lock       STEPNAME(emit_lock)
+#define emit_unlock     STEPNAME(emit_unlock)
+#define emit_cmp8       STEPNAME(emit_cmp8)
+#define emit_cmp16      STEPNAME(emit_cmp16)
+#define emit_cmp32      STEPNAME(emit_cmp32)
+#define emit_cmp8_0     STEPNAME(emit_cmp8_0)
+#define emit_cmp16_0    STEPNAME(emit_cmp16_0)
+#define emit_cmp32_0    STEPNAME(emit_cmp32_0)
+#define emit_test8     STEPNAME(emit_test8)
+#define emit_test16     STEPNAME(emit_test16)
+#define emit_test32     STEPNAME(emit_test32)
+
+#define x87_do_push     STEPNAME(x87_do_push)
+#define x87_do_push_empty STEPNAME(x87_do_push_empty)
+#define x87_do_pop      STEPNAME(x87_do_pop)
+#define x87_get_cache   STEPNAME(x87_get_cache)
+#define x87_get_st      STEPNAME(x87_get_st)
+#define x87_refresh     STEPNAME(x87_refresh)
+#define x87_forget      STEPNAME(x87_forget)
+#define x87_reget_st    STEPNAME(x87_reget_st)
+#define x87_stackcount  STEPNAME(x87_stackcount)
+#define x87_setround    STEPNAME(x87_setround)
+#define x87_restoreround STEPNAME(x87_restoreround)
+#define mmx_get_reg     STEPNAME(mmx_get_reg)
+#define mmx_get_reg_empty STEPNAME(mmx_get_reg_empty)
+#define sse_get_reg     STEPNAME(sse_get_reg)
+#define sse_get_reg_empty STEPNAME(sse_get_reg_empty)
+
+#define fpu_pushcache   STEPNAME(fpu_pushcache)
+#define fpu_popcache    STEPNAME(fpu_popcache)
+#define fpu_reset       STEPNAME(fpu_reset)
+#define fpu_purgecache  STEPNAME(fpu_purgecache)
+#ifdef HAVE_TRACE
+#define fpu_reflectcache STEPNAME(fpu_reflectcache)
+#endif
+
+// get the single reg that from the double "reg" (so Dx[idx])
+#define fpu_get_single_reg      STEPNAME(fpu_get_single_reg)
+// put back (if needed) the single reg in place
+#define fpu_putback_single_reg  STEPNAME(fpu_putback_single_reg)
+
+/* setup r2 to address pointed by */
+uintptr_t geted(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop, uint8_t* ed, uint8_t hint, int* fixedaddress, uint32_t absmax, uint32_t mask);
+
+// Do the GETED, but don't emit anything...
+uintptr_t fakeed(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop);
+
+// generic x86 helper
+void jump_to_epilog(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst);
+void jump_to_linker(dynarec_arm_t* dyn, uintptr_t ip, int reg, int ninst);
+void ret_to_epilog(dynarec_arm_t* dyn, int ninst);
+void retn_to_epilog(dynarec_arm_t* dyn, int ninst, int n);
+void call_c(dynarec_arm_t* dyn, int ninst, void* fnc, int reg, int ret, uint32_t mask);
+void grab_tlsdata(dynarec_arm_t* dyn, uintptr_t addr, int ninst, int reg);
+int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int* retn);
+void emit_lock(dynarec_arm_t* dyn, uintptr_t addr, int ninst);
+void emit_unlock(dynarec_arm_t* dyn, uintptr_t addr, int ninst);
+void emit_cmp8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_cmp16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_cmp32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_cmp8_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_cmp16_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_cmp32_0(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4);
+void emit_test8(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_test16(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+void emit_test32(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3, int s4);
+
+// x87 helper
+// cache of the local stack counter, to avoid upadte at every call
+void x87_stackcount(dynarec_arm_t* dyn, int ninst, int scratch);
+// fpu push. Return the Dd value to be used
+int x87_do_push(dynarec_arm_t* dyn, int ninst);
+// fpu push. Do not allocate a cache register. Needs a scratch register to do x87stack synch (or 0 to not do it)
+void x87_do_push_empty(dynarec_arm_t* dyn, int ninst, int s1);
+// fpu pop. All previous returned Dd should be considered invalid
+void x87_do_pop(dynarec_arm_t* dyn, int ninst);
+// get cache index for a x87 reg, create the entry if needed
+int x87_get_cache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
+// get vfpu register for a x87 reg, create the entry if needed
+int x87_get_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int a);
+// refresh a value from the cache ->emu (nothing done if value is not cached)
+void x87_refresh(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
+// refresh a value from the cache ->emu and then forget the cache (nothing done if value is not cached)
+void x87_forget(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
+// refresh the cache value from emu
+void x87_reget_st(dynarec_arm_t* dyn, int ninst, int s1, int s2, int st);
+// Set rounding according to cw flags, return reg to restore flags
+int x87_setround(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+// Restore round flag
+void x87_restoreround(dynarec_arm_t* dyn, int ninst, int s1);
+
+//MMX helpers
+// get neon register for a MMX reg, create the entry if needed
+int mmx_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a);
+// get neon register for a MMX reg, but don't try to synch it if it needed to be created
+int mmx_get_reg_empty(dynarec_arm_t* dyn, int ninst, int s1, int a);
+
+//SSE/SSE2 helpers
+// get neon register for a SSE reg, create the entry if needed
+int sse_get_reg(dynarec_arm_t* dyn, int ninst, int s1, int a);
+// 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);
+
+// common coproc helpers
+// reset the cache
+void fpu_reset(dynarec_arm_t* dyn, int ninst);
+// purge the FPU cache (needs 3 scratch registers)
+void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+#ifdef HAVE_TRACE
+void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
+#endif
+void fpu_pushcache(dynarec_arm_t* dyn, int ninst, int s1);
+void fpu_popcache(dynarec_arm_t* dyn, int ninst, int s1);
+
+// get the single reg that from the double "reg" (so Dx[idx])
+int fpu_get_single_reg(dynarec_arm_t* dyn, int ninst, int reg, int idx);
+// put back (if needed) the single reg in place
+void fpu_putback_single_reg(dynarec_arm_t* dyn, int ninst, int reg, int idx, int s);
+
+
+uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecGS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarec66(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarec67(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecD8(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDA(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDB(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDC(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDD(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDE(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecF0(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecF20F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst, int* ok, int* need_epilog);
+
+#endif //__DYNAREC_ARM_HELPER_H__

src/dynarec/dynarec_arm_pass.c

@@ -0,0 +1,88 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <stddef.h>
+#include <pthread.h>
+#include <errno.h>
+
+#include "debug.h"
+#include "box86context.h"
+#include "dynarec.h"
+#include "emu/x86emu_private.h"
+#include "emu/x86run_private.h"
+#include "x86run.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "emu/x86run_private.h"
+#include "x86trace.h"
+#include "dynarec_arm.h"
+#include "dynarec_arm_private.h"
+#include "arm_printer.h"
+#include "dynarec_arm_functions.h"
+#include "dynarec_arm_helper.h"
+
+#ifndef STEP
+#error No STEP defined
+#endif
+
+#include "dynarec_arm_helper.h"
+
+void arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
+{
+    uint8_t nextop, opcode;
+    int ok = 1;
+    int ninst = 0;
+    uintptr_t ip = addr;
+    uintptr_t natcall;
+    int retn;
+    uint8_t gd, ed;
+    int8_t i8;
+    int32_t i32, tmp;
+    uint8_t u8;
+    uint8_t gb1, gb2, eb1, eb2;
+    uint32_t u32;
+    int need_epilog = 1;
+    uint8_t wback, wb1, wb2;
+    int fixedaddress;
+    dyn->tablei = 0;
+    // Clean up (because there are multiple passes)
+    dyn->cleanflags = 0;
+    fpu_reset(dyn, ninst);
+    // ok, go now
+    INIT;
+    while(ok) {
+        ip = addr;
+        NEW_INST;
+        fpu_reset_scratch(dyn);
+#ifdef HAVE_TRACE
+        if(dyn->emu->dec && box86_dynarec_trace) {
+        if((dyn->emu->trace_end == 0) 
+            || ((ip >= dyn->emu->trace_start) && (ip <= dyn->emu->trace_end)))  {
+                MESSAGE(LOG_DUMP, "TRACE ----\n");
+                fpu_reflectcache(dyn, ninst, x1, x2, x3);
+                MOV32(1, ip);
+                STM(0, (1<<4)|(1<<5)|(1<<6)|(1<<7)|(1<<8)|(1<<9)|(1<<10)|(1<<11));
+                STR_IMM9(1, 0, offsetof(x86emu_t, ip));
+                MOV32(2, 1);
+                CALL(PrintTrace, -1, 0);
+                MESSAGE(LOG_DUMP, "----------\n");
+            }
+        }
+#endif
+
+        addr = dynarec00(dyn, addr, ip, ninst, &ok, &need_epilog);
+
+        INST_EPILOG;
+
+        if(dyn->insts && dyn->insts[ninst+1].x86.barrier) {
+            fpu_purgecache(dyn, ninst, x1, x2, x3);
+            if(dyn->insts[ninst+1].x86.barrier!=2)
+                dyn->cleanflags = 0;
+        }
+        ++ninst;
+    }
+    if(need_epilog) {
+        fpu_purgecache(dyn, ninst, x1, x2, x3);
+        jump_to_epilog(dyn, ip, 0, ninst);  // no linker here, it's an unknow instruction
+    }
+    FINI;
+}

src/dynarec/dynarec_arm_pass0.h

@@ -0,0 +1,16 @@
+
+#define INIT    uintptr_t sav_addr=addr
+#define FINI    dyn->isize = addr-sav_addr
+#define MESSAGE(A, ...)  
+#define EMIT(A)     
+#define UFLAGS(A)          {}
+#define USEFLAGS(A)        {}
+#define NEW_INST        ++dyn->size
+#define INST_EPILOG         
+#define INST_NAME(name) 
+#define DEFAULT         \
+        --dyn->size;    \
+        dynarec_log(LOG_INFO, "%p: Dynarec stopped because of Opcode %02X %02X %02X %02X %02X %02X %02X\n", \
+        ip, PKip(0),                   \
+        PKip(1), PKip(2), PKip(3),     \
+        PKip(4), PKip(5), PKip(6))

src/dynarec/dynarec_arm_pass1.h

@@ -0,0 +1,15 @@
+#define INIT    
+#define FINI     \
+    dyn->insts[ninst].x86.addr = addr; \
+    dyn->insts[ninst].x86.flags = X86_FLAGS_USE; \
+    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr;
+#define MESSAGE(A, ...)  
+#define EMIT(A)     
+#define UFLAGS(A)   dyn->insts[ninst+(A?1:0)].x86.flags = X86_FLAGS_CHANGE
+#define USEFLAG(A)  dyn->insts[ninst].x86.flags = X86_FLAGS_USE
+#define JUMP(A)     dyn->insts[ninst].x86.jmp = A
+#define NEW_INST \
+    dyn->insts[ninst].x86.addr = ip; \
+    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr;
+#define INST_EPILOG 
+#define INST_NAME(name) 

src/dynarec/dynarec_arm_pass2.h

@@ -0,0 +1,8 @@
+#define INIT        dyn->arm_size = 0
+#define FINI        if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
+
+#define MESSAGE(A, ...)  
+#define EMIT(A)     dyn->insts[ninst].size+=4; dyn->arm_size+=4
+#define NEW_INST    if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
+#define INST_EPILOG dyn->insts[ninst].epilog = dyn->arm_size; 
+#define INST_NAME(name) 

src/dynarec/dynarec_arm_pass3.h

@@ -0,0 +1,12 @@
+#define INIT    
+#define FINI
+#define EMIT(A)     \
+    if(box86_dynarec_log>=LOG_DUMP) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (A), arm_print(A));} \
+    *(uint32_t*)(dyn->block) = A;   \
+    dyn->block += 4; dyn->arm_size += 4
+
+#define MESSAGE(A, ...)  dynarec_log(A, __VA_ARGS__);
+#define NEW_INST        
+#define INST_EPILOG     
+#define INST_NAME(name) if(box86_dynarec_log>=LOG_DUMP) printf_x86_instruction(dyn->emu->dec, &dyn->insts[ninst].x86, name)
+

src/dynarec/dynarec_arm_private.h

@@ -0,0 +1,41 @@
+#ifndef __DYNAREC_ARM_PRIVATE_H_
+#define __DYNAREC_ARM_PRIVATE_H_
+
+#include "dynarec_private.h"
+
+typedef struct x86emu_s x86emu_t;
+
+typedef struct instruction_arm_s {
+    instruction_x86_t   x86;
+    uintptr_t           address;    // (start) address of the arm emited instruction
+    uintptr_t           epilog;     // epilog of current instruction (can be start of next, of barrier stuff)
+    int                 size;       // size of the arm emited instruction
+    uintptr_t           mark, mark2, mark3;
+    uintptr_t           markf;
+} instruction_arm_t;
+
+typedef struct dynarec_arm_s {
+    instruction_arm_t   *insts;
+    int32_t             size;
+    int32_t             cap;
+    uintptr_t           start;      // start of the block
+    uint32_t            isize;      // size in byte of x86 instructions included
+    void*               block;      // memory pointer where next instruction is emited
+    int                 arm_size;   // size of emitted arm code
+    x86emu_t*           emu;
+    uintptr_t*          table;      // jump table
+    int                 tablesz;    // size of the jump table
+    int                 tablei;     // index
+    int                 cleanflags; // flags are clean (CMP or TST have been just executed)
+    int                 x87cache[8];// cache status for the 8 x87 register behind the fpu stack
+    int                 x87reg[8];  // reg used for x87cache entry
+    int                 mmxcache[8];// cache status for the 8 MMX registers
+    int                 ssecache[8];// cache status for the 8 SSE(2) registers
+    int                 fpuused[24];// all 8..31 double reg from fpu, used by x87, sse and mmx
+    int                 x87stack;   // cache stack counter
+    int                 fpu_scratch;// scratch counter
+    int                 fpu_reg;    // x87/sse/mmx reg counter
+} dynarec_arm_t;
+
+
+#endif //__DYNAREC_ARM_PRIVATE_H_

src/dynarec/dynarec_private.h

@@ -0,0 +1,19 @@
+#ifndef __DYNAREC_PRIVATE_H_
+#define __DYNAREC_PRIVATE_H_
+
+#define X86_FLAGS_NONE      0
+#define X86_FLAGS_CHANGE    1
+#define X86_FLAGS_USE       2
+
+typedef struct instruction_x86_s {
+    uintptr_t   addr;   //address of the instruction
+    int32_t     size;   // size of the instruction
+    int         flags;  // flags for this instruction (see X86_FLAGS_XXXX)
+    int         barrier; // next instruction is a jump point, so no optim allowed
+    uintptr_t   jmp;    // offset to jump to, even if conditionnal (0 if not), no relative offset here
+    int         jmp_insts;  // instuction to jump to (-1 if out of the block)
+} instruction_x86_t;
+
+void printf_x86_instruction(zydis_dec_t* dec, instruction_x86_t* inst, const char* name);
+
+#endif //__DYNAREC_PRIVATE_H_

src/elfloader.c

@@ -1,494 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <elf.h>
-#include <sys/mman.h>
-
-#include "box86version.h"
-#include "elfloader.h"
-#include "debug.h"
-#include "elfload_dump.h"
-#include "elfloader_private.h"
-#include "librarian.h"
-#include "x86run.h"
-#include "bridge.h"
-#include "wrapper.h"
-#include "box86context.h"
-#include "library.h"
-#include "x86emu.h"
-
-void FreeElfHeader(elfheader_t** head)
-{
-    if(!head || !*head)
-        return;
-    elfheader_t *h = *head;
-    free(h->name);
-    free(h->PHEntries);
-    free(h->SHEntries);
-    free(h->SHStrTab);
-    free(h->StrTab);
-    free(h->Dynamic);
-    free(h->DynStr);
-    free(h->SymTab);
-    free(h->DynSym);
-    FreeElfMemory(h);
-    free(h);
-
-    *head = NULL;
-}
-
-int CalcLoadAddr(elfheader_t* head)
-{
-    head->memsz = 0;
-    head->paddr = head->vaddr = (uintptr_t)~0;
-    head->align = 1;
-    for (int i=0; i<head->numPHEntries; ++i)
-        if(head->PHEntries[i].p_type == PT_LOAD) {
-            if(head->paddr > head->PHEntries[i].p_paddr)
-                head->paddr = head->PHEntries[i].p_paddr;
-            if(head->vaddr > head->PHEntries[i].p_vaddr)
-                head->vaddr = head->PHEntries[i].p_vaddr;
-        }
-    
-    if(head->vaddr==~0 || head->paddr==~0) {
-        printf_log(LOG_NONE, "Error: v/p Addr for Elf Load not set\n");
-        return 1;
-    }
-
-    head->stacksz = 1024*1024;          //1M stack size default?
-    head->stackalign = 4;   // default align for stack
-    for (int i=0; i<head->numPHEntries; ++i) {
-        if(head->PHEntries[i].p_type == PT_LOAD) {
-            uintptr_t phend = head->PHEntries[i].p_vaddr - head->vaddr + head->PHEntries[i].p_memsz;
-            if(phend > head->memsz)
-                head->memsz = phend;
-            if(head->PHEntries[i].p_align > head->align)
-                head->align = head->PHEntries[i].p_align;
-        }
-        if(head->PHEntries[i].p_type == PT_GNU_STACK) {
-            if(head->stacksz < head->PHEntries[i].p_memsz)
-                head->stacksz = head->PHEntries[i].p_memsz;
-            if(head->stackalign < head->PHEntries[i].p_align)
-                head->stackalign = head->PHEntries[i].p_align;
-        }
-    }
-    printf_log(LOG_DEBUG, "Elf Addr(v/p)=%p/%p Memsize=0x%x (align=0x%x)\n", (void*)head->vaddr, (void*)head->paddr, head->memsz, head->align);
-    printf_log(LOG_DEBUG, "Elf Stack Memsize=%u (align=%u)\n", head->stacksz, head->stackalign);
-
-    return 0;
-}
-
-const char* ElfName(elfheader_t* head)
-{
-    return head->name;
-}
-
-int AllocElfMemory(elfheader_t* head)
-{
-    #if 0
-    printf_log(LOG_DEBUG, "Allocating memory for Elf \"%s\"\n", head->name);
-    if (posix_memalign((void**)&head->memory, head->align, head->memsz)) {
-        printf_log(LOG_NONE, "Cannot allocate aligned memory (0x%x/0x%x) for elf \"%s\"\n", head->memsz, head->align, head->name);
-        return 1;
-    }
-    printf_log(LOG_DEBUG, "Address is %p\n", head->memory);
-    printf_log(LOG_DEBUG, "And setting memory access to PROT_READ | PROT_WRITE | PROT_EXEC\n");
-    if (mprotect(head->memory, head->memsz, PROT_READ | PROT_WRITE | PROT_EXEC)) {
-        printf_log(LOG_NONE, "Cannot protect memory for elf \"%s\"\n", head->name);
-        // memory protect error not fatal for now....
-    }
-    #else
-    printf_log(LOG_DEBUG, "Allocating 0x%x memory @%p for Elf \"%s\"\n", head->memsz, (void*)head->vaddr, head->name);
-    void* p = mmap((void*)head->vaddr, head->memsz
-        , PROT_READ | PROT_WRITE | PROT_EXEC, MAP_SHARED | MAP_ANONYMOUS | ((head->vaddr)?MAP_FIXED:0)
-        , -1, 0);
-    if(p==MAP_FAILED) {
-        printf_log(LOG_NONE, "Cannot create memory map (@%p 0x%x/0x%x) for elf \"%s\"\n", (void*)head->vaddr, head->memsz, head->align, head->name);
-        return 1;
-    }
-    head->memory = p;
-    memset(p, 0, head->memsz);
-    head->delta = (intptr_t)p - (intptr_t)head->vaddr;
-    printf_log(LOG_DEBUG, "Got %p (delta=%p)\n", p, (void*)head->delta);
-    #endif
-
-    return 0;
-}
-
-void FreeElfMemory(elfheader_t* head)
-{
-    if(head->memory)
-        munmap((void*)head->memory, head->memsz);
-}
-
-int LoadElfMemory(FILE* f, elfheader_t* head)
-{
-    for (int i=0; i<head->numPHEntries; ++i) {
-        if(head->PHEntries[i].p_type == PT_LOAD) {
-            Elf32_Phdr * e = &head->PHEntries[i];
-            char* dest = (char*)e->p_paddr + head->delta;
-            printf_log(LOG_DEBUG, "Loading block #%i @%p (0x%x/0x%x)\n", i, dest, e->p_filesz, e->p_memsz);
-            fseek(f, e->p_offset, SEEK_SET);
-            if(fread(dest, e->p_filesz, 1, f)!=1) {
-                printf_log(LOG_NONE, "Fail to read PT_LOAD part #%d\n", i);
-                return 1;
-            }
-            // zero'd difference between filesz and memsz
-            if(e->p_filesz != e->p_memsz)
-                memset(dest+e->p_filesz, 0, e->p_memsz - e->p_filesz);
-        }
-    }
-    return 0;
-}
-
-int RelocateElfREL(lib_t *maplib, elfheader_t* head, int cnt, Elf32_Rel *rel)
-{
-    for (int i=0; i<cnt; ++i) {
-        Elf32_Sym *sym = &head->DynSym[ELF32_R_SYM(rel[i].r_info)];
-        const char* symname = SymName(head, sym);
-        uint32_t *p = (uint32_t*)(rel[i].r_offset + head->delta);
-        uintptr_t offs = 0;
-        uint32_t sz = 0;
-        uintptr_t end = 0;
-        GetGlobalSymbolStartEnd(maplib, symname, &offs, &end, 0);
-        uintptr_t globoffs, globend;
-        int t = ELF32_R_TYPE(rel[i].r_info);
-        switch(t) {
-            case R_386_NONE:
-                // can be ignored
-                printf_log(LOG_DEBUG, "Ignoring %s @%p (%p)\n", DumpRelType(t), p, (void*)(p?(*p):0));
-                break;
-            case R_386_PC32:
-                    offs = (*p) + (offs - (uintptr_t)p);
-                    printf_log(LOG_DEBUG, "Apply R_386_PC32 @%p with sym=%s (%p -> %p)\n", p, symname, *(void**)p, (void*)(*(uintptr_t*)p+offs));
-                    *p += offs;
-                break;
-            case R_386_GLOB_DAT:
-                // Look for same symbol already loaded but not in self
-                if (GetGlobalSymbolStartEnd(maplib, symname, &globoffs, &globend, 1)) {
-                    offs = globoffs;
-                    end = globend;
-                }
-                if (!offs) {
-                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_GLOB_DAT @%p (%p)\n", symname, p, *(void**)p);
-//                    return -1;
-                } else {
-                    printf_log(LOG_DEBUG, "Apply R_386_GLOB_DAT @%p (%p -> %p) on sym=%s\n", p, (void*)(p?(*p):0), (void*)offs, symname);
-                    *p = offs;
-                }
-                break;
-            case R_386_RELATIVE:
-                printf_log(LOG_DEBUG, "Apply R_386_RELATIVE @%p (%p -> %p)\n", p, *(void**)p, (void*)((*p)+head->delta));
-                *p += head->delta;
-                break;
-            case R_386_32:
-                if (!offs) {
-                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_32 @%p (%p)\n", symname, p, *(void**)p);
-//                    return -1;
-                } else {
-                    printf_log(LOG_DEBUG, "Apply R_386_32 @%p with sym=%s (%p -> %p)\n", p, symname, *(void**)p, (void*)(offs+*(uint32_t*)p));
-                    *p += offs;
-                }
-                break;
-            //case R_386_TLS_DTPMOD32:
-            // try to use _dl_next_tls_modid() ?
-            case R_386_TLS_DTPOFF32:
-            case R_386_JMP_SLOT:
-                if (!offs) {
-                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_JMP_SLOT @%p (%p)\n", symname, p, *(void**)p);
-//                    return -1;
-                } else {
-                    if(p) {
-                        printf_log(LOG_DEBUG, "Apply R_386_JMP_SLOT @%p with sym=%s (%p -> %p)\n", p, symname, *(void**)p, (void*)offs);
-                        *p = offs;
-                    } else {
-                        printf_log(LOG_NONE, "Warning, Symbol %s found, but Jump Slot Offset is NULL \n", symname);
-                    }
-                }
-                break;
-            case R_386_COPY:
-                if(offs) {
-                    GetGlobalSymbolStartEnd(maplib, symname, &offs, &end, 1);   // get original copy if any
-                    printf_log(LOG_DEBUG, "Apply R_386_COPY @%p with sym=%s, @%p size=%d (", p, symname, (void*)offs, sym->st_size);
-                    memmove(p, (void*)offs, sym->st_size);
-                    if(LOG_DEBUG<=box86_log) {
-                        uint32_t*k = (uint32_t*)p;
-                        for (int i=0; i<sym->st_size; i+=4, ++k)
-                            printf_log(LOG_DEBUG, "%s0x%08X", i?" ":"", *k);
-                        printf_log(LOG_DEBUG, ")\n");
-                    }
-                } else {
-                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_COPY @%p (%p)\n", symname, p, *(void**)p);
-                }
-                break;
-            default:
-                printf_log(LOG_INFO, "Warning, don't know of to handle rel #%d %s (%p)\n", i, DumpRelType(ELF32_R_TYPE(rel[i].r_info)), p);
-        }
-    }
-    return 0;
-}
-
-int RelocateElfRELA(lib_t *maplib, elfheader_t* head, int cnt, Elf32_Rela *rela)
-{
-    for (int i=0; i<cnt; ++i) {
-        Elf32_Sym *sym = &head->DynSym[ELF32_R_SYM(rela[i].r_info)];
-        const char* symname = SymName(head, sym);
-        uint32_t *p = (uint32_t*)(rela[i].r_offset + head->delta);
-        uintptr_t offs = 0;
-        uint32_t sz = 0;
-        uintptr_t end = 0;
-        switch(ELF32_R_TYPE(rela[i].r_info)) {
-            case R_386_NONE:
-            case R_386_PC32:
-                // can be ignored
-                break;
-            case R_386_COPY:
-                GetGlobalSymbolStartEnd(maplib, symname, &offs, &end,1);
-                if(offs) {
-                    // add r_addend to p?
-                    printf_log(LOG_DEBUG, "Apply R_386_COPY @%p with sym=%s, @%p size=%d\n", p, symname, (void*)offs, sym->st_size);
-                    memcpy(p, (void*)(offs+rela[i].r_addend), sym->st_size);
-                } else {
-                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_COPY @%p (%p)\n", symname, p, *(void**)p);
-                }
-                break;
-            default:
-                printf_log(LOG_INFO, "Warning, don't know of to handle rel #%d %s on %s\n", i, DumpRelType(ELF32_R_TYPE(rela[i].r_info)), symname);
-        }
-    }
-    return 0;
-}
-int RelocateElf(lib_t *maplib, elfheader_t* head)
-{
-    if(head->rel) {
-        int cnt = head->relsz / head->relent;
-        DumpRelTable(head, cnt, (Elf32_Rel *)(head->rel + head->delta), "Rel");
-        printf_log(LOG_DEBUG, "Applying %d Relocation(s) for %s\n", cnt, head->name);
-        if(RelocateElfREL(maplib, head, cnt, (Elf32_Rel *)(head->rel + head->delta)))
-            return -1;
-    }
-    if(head->rela) {
-        int cnt = head->relasz / head->relaent;
-        DumpRelATable(head, cnt, (Elf32_Rela *)(head->rela + head->delta), "RelA");
-        printf_log(LOG_DEBUG, "Applying %d Relocation(s) with Addend for %s\n", cnt, head->name);
-        if(RelocateElfRELA(maplib, head, cnt, (Elf32_Rela *)(head->rela + head->delta)))
-            return -1;
-    }
-   
-    return 0;
-}
-
-int RelocateElfPlt(box86context_t* context, lib_t *maplib, elfheader_t* head)
-{
-    if(head->pltrel) {
-        int cnt = head->pltsz / head->pltent;
-        if(head->pltrel==DT_REL) {
-            DumpRelTable(head, cnt, (Elf32_Rel *)(head->jmprel + head->delta), "PLT");
-            printf_log(LOG_DEBUG, "Applying %d PLT Relocation(s) for %s\n", cnt, head->name);
-            if(RelocateElfREL(maplib, head, cnt, (Elf32_Rel *)(head->jmprel + head->delta)))
-                return -1;
-        } else if(head->pltrel==DT_RELA) {
-            DumpRelATable(head, cnt, (Elf32_Rela *)(head->jmprel + head->delta), "PLT");
-            printf_log(LOG_DEBUG, "Applying %d PLT Relocation(s) with Addend for %s\n", cnt, head->name);
-            if(RelocateElfRELA(maplib, head, cnt, (Elf32_Rela *)(head->jmprel + head->delta)))
-                return -1;
-        }
-        if(head->gotplt) {
-            if(pltResolver==~0) {
-                pltResolver = AddBridge(context->system, vFEpp, PltResolver);   // should be vFEuu
-            }
-            *(uintptr_t*)(head->gotplt+head->delta+8) = pltResolver;
-            printf_log(LOG_DEBUG, "PLT Resolver injected at %p\n", (void*)(head->gotplt+head->delta+8));
-        }
-    }
-   
-    return 0;
-}
-
-void CalcStack(elfheader_t* elf, uint32_t* stacksz, int* stackalign)
-{
-    if(*stacksz < elf->stacksz)
-        *stacksz = elf->stacksz;
-    if(*stackalign < elf->stackalign)
-        *stackalign = elf->stackalign;
-}
-
-Elf32_Sym* GetFunction(elfheader_t* h, const char* name)
-{
-    // TODO: create a hash on named to avoid this loop
-    for (int i=0; i<h->numSymTab; ++i) {
-        if(h->SymTab[i].st_info == 18) {    // TODO: this "18" is probably defined somewhere
-            const char * symname = h->StrTab+h->SymTab[i].st_name;
-            if(strcmp(symname, name)==0) {
-                return h->SymTab+i;
-            }
-        }
-    }
-    return NULL;
-}
-
-uintptr_t GetFunctionAddress(elfheader_t* h, const char* name)
-{
-    Elf32_Sym* sym = GetFunction(h, name);
-    if(sym) return sym->st_value;
-    return 0;
-}
-
-uintptr_t GetEntryPoint(lib_t* maplib, elfheader_t* h)
-{
-    uintptr_t ep = h->entrypoint + h->delta;
-    printf_log(LOG_DEBUG, "Entry Point is %p\n", (void*)ep);
-    if(box86_log>=LOG_DUMP) {
-        printf_log(LOG_DUMP, "(short) Dump of Entry point\n");
-        int sz = 64;
-        uintptr_t lastbyte = GetLastByte(h);
-        if (ep + sz >  lastbyte)
-            sz = lastbyte - ep;
-        DumpBinary((char*)ep, sz);
-    }
-    /*
-    // but instead of regular entrypoint, lets grab "main", it will be easier to manage I guess
-    uintptr_t m = FindSymbol(maplib, "main");
-    if(m) {
-        ep = m;
-        printf_log(LOG_DEBUG, "Using \"main\" as Entry Point @%p\n", ep);
-        if(box86_log>=LOG_DUMP) {
-            printf_log(LOG_DUMP, "(short) Dump of Entry point\n");
-            int sz = 64;
-            uintptr_t lastbyte = GetLastByte(h);
-            if (ep + sz >  lastbyte)
-                sz = lastbyte - ep;
-            DumpBinary((char*)ep, sz);
-        }
-    }
-    */
-    return ep;
-}
-
-uintptr_t GetLastByte(elfheader_t* h)
-{
-    return (uintptr_t)h->memory + h->delta + h->memsz;
-}
-
-void AddGlobalsSymbols(lib_t *maplib, kh_mapsymbols_t* mapsymbols, elfheader_t* h)
-{
-    // info == 161 => OBJECT UNIQUE
-    // info == 33  => OBJECT WEAK
-    printf_log(LOG_DUMP, "Will look for Symbol to add in SymTable(%d)\n", h->numSymTab);
-    for (int i=0; i<h->numSymTab; ++i) {
-        const char * symname = h->StrTab+h->SymTab[i].st_name;
-        if(((h->SymTab[i].st_info & 2) || (h->SymTab[i].st_info == 17)
-             || (h->SymTab[i].st_info == 161 && !FindGlobalSymbol(maplib, symname)) 
-             || (h->SymTab[i].st_info == 33)
-            )
-            && (h->SymTab[i].st_other==0) && (h->SymTab[i].st_shndx!=0)) {
-            if(!h->SymTab[i].st_value && FindGlobalSymbol(maplib, symname))
-                break;
-            uintptr_t offs = h->SymTab[i].st_value + h->delta;
-            uint32_t sz = h->SymTab[i].st_size;
-            printf_log(LOG_DUMP, "Adding Symbol \"%s\" with offset=%p sz=%d\n", symname, (void*)offs, sz);
-            AddSymbol(mapsymbols, symname, offs, sz);
-        }
-    }
-    
-    printf_log(LOG_DUMP, "Will look for Symbol to add in DynSym (%d)\n", h->numDynSym);
-    for (int i=0; i<h->numDynSym; ++i) {
-        const char * symname = h->DynStr+h->DynSym[i].st_name;
-        if(((h->DynSym[i].st_info & 2) || (h->DynSym[i].st_info == 17)
-             || (h->DynSym[i].st_info == 161 && !FindGlobalSymbol(maplib, symname))
-             || (h->DynSym[i].st_info == 33) 
-            )
-            && (h->DynSym[i].st_other==0) && (h->DynSym[i].st_shndx!=0 && h->DynSym[i].st_shndx<62521)) {
-            if(!h->DynSym[i].st_value && FindGlobalSymbol(maplib, symname))
-                break;
-            uintptr_t offs = h->DynSym[i].st_value + h->delta;
-            uint32_t sz = h->DynSym[i].st_size;
-            printf_log(LOG_DUMP, "Adding Symbol \"%s\" with offset=%p sz=%d\n", symname, (void*)offs, sz);
-            AddSymbol(mapsymbols, symname, offs, sz);
-        }
-    }
-    
-}
-
-/*
-$ORIGIN – Provides the directory the object was loaded from. This token is typical
-used for locating dependencies in unbundled packages. For more details of this
-token expansion, see “Locating Associated Dependencies”
-$OSNAME – Expands to the name of the operating system (see the uname(1) man
-page description of the -s option). For more details of this token expansion, see
-“System Specific Shared Objects”
-$OSREL – Expands to the operating system release level (see the uname(1) man
-page description of the -r option). For more details of this token expansion, see
-“System Specific Shared Objects”
-$PLATFORM – Expands to the processor type of the current machine (see the
-uname(1) man page description of the -i option). For more details of this token
-expansion, see “System Specific Shared Objects”
-*/
-int LoadNeededLib(elfheader_t* h, lib_t *maplib, box86context_t *box86, x86emu_t* emu)
-{
-   DumpDynamicNeeded(h);
-   // reverse order on needed libs
-   for (int i=h->numDynamic-1; i>=0; --i)
-        if(h->Dynamic[i].d_tag==DT_NEEDED) {
-            char *needed = h->DynStrTab+h->delta+h->Dynamic[i].d_un.d_val;
-            // TODO: Add LD_LIBRARY_PATH and RPATH Handling
-            if(AddNeededLib(maplib, needed, box86, emu)) {
-                printf_log(LOG_INFO, "Error loading needed lib: \"%s\"\n", needed);
-                return 1;   //error...
-            }
-        }
-    return 0;
-}
-
-void RunElfInit(elfheader_t* h, x86emu_t *emu)
-{
-    if(h->init_done || !h->initentry)
-        return;
-    uintptr_t p = h->initentry + h->delta;
-    box86context_t* context = GetEmuContext(emu);
-    if(context->deferedInit) {
-        if(context->deferedInitSz==context->deferedInitCap) {
-            context->deferedInitCap += 4;
-            context->deferedInitList = (elfheader_t**)realloc(context->deferedInitList, context->deferedInitCap*sizeof(elfheader_t*));
-        }
-        context->deferedInitList[context->deferedInitSz++] = h;
-        return;
-    }
-    printf_log(LOG_DEBUG, "Calling Init for %s @%p\n", ElfName(h), (void*)p);
-    EmuCall(emu, p);
-    h->init_done = 1;
-    return;
-}
-void RunDeferedElfInit(x86emu_t *emu)
-{
-    box86context_t* context = GetEmuContext(emu);
-    if(!context->deferedInit)
-        return;
-    context->deferedInit = 0;
-    if(!context->deferedInitList)
-        return;
-    for (int i=0; i<context->deferedInitSz; ++i)
-        RunElfInit(context->deferedInitList[i], emu);
-    free(context->deferedInitList);
-    context->deferedInitList = NULL;
-    context->deferedInitCap = context->deferedInitSz = 0;
-}
-void RunElfFini(elfheader_t* h, x86emu_t *emu)
-{
-    if(h->fini_done || !h->finientry)
-        return;
-    uintptr_t p = h->finientry + h->delta;
-    printf_log(LOG_DEBUG, "Calling Fini for %s @%p\n", ElfName(h), (void*)p);
-    EmuCall(emu, p);
-    h->fini_done = 1;
-    return;
-}
-
-uintptr_t GetElfInit(elfheader_t* h)
-{
-    return h->initentry + h->delta;
-}
-uintptr_t GetElfFini(elfheader_t* h)
-{
-    return h->finientry + h->delta;
-}

src/elfloader.h

@@ -1,33 +0,0 @@
-#ifndef __ELF_LOADER_H_
-#define __ELF_LOADER_H_
-#include <stdio.h>
-
-typedef struct elfheader_s elfheader_t;
-typedef struct lib_s lib_t;
-typedef struct kh_mapsymbols_s kh_mapsymbols_t;
-typedef struct box86context_s box86context_t;
-typedef struct x86emu_s x86emu_t;
-
-void* LoadAndCheckElfHeader(FILE* f, const char* name, int exec); // exec : 0 = lib, 1 = exec
-void FreeElfHeader(elfheader_t** head);
-const char* ElfName(elfheader_t* head);
-
-// return 0 if OK
-int CalcLoadAddr(elfheader_t* head);
-int AllocElfMemory(elfheader_t* head);
-void FreeElfMemory(elfheader_t* head);
-int LoadElfMemory(FILE* f, elfheader_t* head);
-int RelocateElf(lib_t *maplib, elfheader_t* head);
-int RelocateElfPlt(box86context_t* context, lib_t *maplib, elfheader_t* head);
-void CalcStack(elfheader_t* h, uint32_t* stacksz, int* stackalign);
-uintptr_t GetEntryPoint(lib_t* maplib, elfheader_t* h);
-uintptr_t GetLastByte(elfheader_t* h);
-void AddGlobalsSymbols(lib_t *maplib, kh_mapsymbols_t* mapsymbols, elfheader_t* h);
-int LoadNeededLib(elfheader_t* h, lib_t *maplib, box86context_t* box86, x86emu_t* emu);
-uintptr_t GetElfInit(elfheader_t* h);
-uintptr_t GetElfFini(elfheader_t* h);
-void RunElfInit(elfheader_t* h, x86emu_t *emu);
-void RunElfFini(elfheader_t* h, x86emu_t *emu);
-void RunDeferedElfInit(x86emu_t *emu);
-
-#endif //__ELF_LOADER_H_

src/elfparser.c

@@ -1,271 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <elf.h>
-
-#include "box86version.h"
-#include "elfloader.h"
-#include "debug.h"
-#include "elfload_dump.h"
-#include "elfloader_private.h"
-
-#ifndef PN_XNUM 
-#define PN_XNUM (0xffff)
-#endif
-
-int LoadSH(FILE *f, Elf32_Shdr *s, void** SH, const char* name, uint32_t type)
-{
-    if(type && (s->sh_type != type)) {
-        printf_log(LOG_INFO, "Section Header \"%s\" (off=%d, size=%d) has incorect type (%d != %d)\n", name, s->sh_offset, s->sh_size, s->sh_type, type);
-        return -1;
-    }
-    if (type==SHT_SYMTAB && s->sh_size%sizeof(Elf32_Sym)) {
-        printf_log(LOG_INFO, "Section Header \"%s\" (off=%d, size=%d) has size (not multiple of %d)\n", name, s->sh_offset, s->sh_size, sizeof(Elf32_Sym));
-    }
-    *SH = calloc(1, s->sh_size);
-    fseek(f, s->sh_offset ,SEEK_SET);
-    if(fread(*SH, s->sh_size, 1, f)!=1) {
-            printf_log(LOG_INFO, "Cannot read Section Header \"%s\" (off=%d, size=%d)\n", name, s->sh_offset, s->sh_size);
-            return -1;
-    }
-
-    return 0;
-}
-
-int FindSection(Elf32_Shdr *s, int n, char* SHStrTab, const char* name)
-{
-    for (int i=0; i<n; ++i) {
-        if(s[i].sh_type!=SHT_NULL)
-            if(!strcmp(SHStrTab+s[i].sh_name, name))
-                return i;
-    }
-    return 0;
-}
-
-void LoadNamedSection(FILE *f, Elf32_Shdr *s, int size, char* SHStrTab, const char* name, const char* clearname, uint32_t type, void** what, int* num)
-{
-    int n = FindSection(s, size, SHStrTab, name);
-    printf_log(LOG_DEBUG, "Loading %s (idx = %d)\n", clearname, n);
-    if(n)
-        LoadSH(f, s+n, what, name, type);
-    if(type==SHT_SYMTAB || type==SHT_DYNSYM) {
-        if(*what && num)
-            *num = s[n].sh_size / sizeof(Elf32_Sym);
-    } else if(type==SHT_DYNAMIC) {
-        if(*what && num)
-            *num = s[n].sh_size / sizeof(Elf32_Dyn);
-    }
-}
-
-void* LoadAndCheckElfHeader(FILE* f, const char* name, int exec)
-{
-    Elf32_Ehdr header;
-    if(fread(&header, sizeof(Elf32_Ehdr), 1, f)!=1) {
-        printf_log(LOG_INFO, "Cannot read ELF Header\n");
-        return NULL;
-    }
-    if(memcmp(header.e_ident, ELFMAG, SELFMAG)!=0) {
-        printf_log(LOG_INFO, "Not an ELF file (sign=%c%c%c%c)\n", header.e_ident[0], header.e_ident[1], header.e_ident[2], header.e_ident[3]);
-        return NULL;
-    }
-    if(header.e_ident[EI_CLASS]!=ELFCLASS32) {
-        printf_log(LOG_INFO, "Not an 32bits ELF (%d)\n", header.e_ident[EI_CLASS]);
-        return NULL;
-    }
-    if(header.e_ident[EI_DATA]!=ELFDATA2LSB) {
-        printf_log(LOG_INFO, "Not an LittleEndian ELF (%d)\n", header.e_ident[EI_DATA]);
-        return NULL;
-    }
-    if(header.e_ident[EI_VERSION]!=EV_CURRENT) {
-        printf_log(LOG_INFO, "Incorrect ELF version (%d)\n", header.e_ident[EI_VERSION]);
-        return NULL;
-    }
-    if(header.e_ident[EI_OSABI]!=ELFOSABI_LINUX && header.e_ident[EI_OSABI]!=ELFOSABI_NONE && header.e_ident[EI_OSABI]!=ELFOSABI_SYSV) {
-        printf_log(LOG_INFO, "Not a Linux ELF (%d)\n",header.e_ident[EI_OSABI]);
-        return NULL;
-    }
-
-    if(header.e_type != ET_EXEC && header.e_type != ET_DYN) {
-        printf_log(LOG_INFO, "Not an Executable (%d)\n", header.e_type);
-        return NULL;
-    }
-
-    if(header.e_machine != EM_386) {
-        printf_log(LOG_INFO, "Not an i386 ELF (%d)\n", header.e_machine);
-        return NULL;
-    }
-
-    if(header.e_entry == 0) {
-        printf_log(LOG_INFO, "No entry point in ELF\n");
-        return NULL;
-    }
-    if(header.e_phentsize != sizeof(Elf32_Phdr)) {
-        printf_log(LOG_INFO, "Program Header Entry size incorrect (%d != %d)\n", header.e_phentsize, sizeof(Elf32_Phdr));
-        return NULL;
-    }
-    if(header.e_shentsize != sizeof(Elf32_Shdr)) {
-        printf_log(LOG_INFO, "Section Header Entry size incorrect (%d != %d)\n", header.e_shentsize, sizeof(Elf32_Shdr));
-        return NULL;
-    }
-
-    elfheader_t *h = calloc(1, sizeof(elfheader_t));
-    h->name = strdup(name);
-    h->entrypoint = header.e_entry;
-    h->numPHEntries = header.e_phnum;
-    h->numSHEntries = header.e_shnum;
-    h->SHIdx = header.e_shstrndx;
-    // special cases for nums
-    if(h->numSHEntries == 0) {
-        printf_log(LOG_DEBUG, "Read number of Sections in 1st Section\n");
-        // read 1st section header and grab actual number from here
-        fseek(f, header.e_shoff, SEEK_SET);
-        Elf32_Shdr section;
-        if(fread(&section, sizeof(Elf32_Shdr), 1, f)!=1) {
-            free(h);
-            printf_log(LOG_INFO, "Cannot read Initial Section Header\n");
-            return NULL;
-        }
-        h->numSHEntries = section.sh_size;
-    }
-    // now read all section headers
-    printf_log(LOG_DEBUG, "Read %d Section header\n", h->numSHEntries);
-    h->SHEntries = (Elf32_Shdr*)calloc(h->numSHEntries, sizeof(Elf32_Shdr));
-    fseek(f, header.e_shoff ,SEEK_SET);
-    if(fread(h->SHEntries, sizeof(Elf32_Shdr), h->numSHEntries, f)!=h->numSHEntries) {
-            FreeElfHeader(&h);
-            printf_log(LOG_INFO, "Cannot read all Section Header\n");
-            return NULL;
-    }
-
-    if(h->numPHEntries == PN_XNUM) {
-        printf_log(LOG_DEBUG, "Read number of Program Header in 1st Section\n");
-        // read 1st section header and grab actual number from here
-        h->numPHEntries = h->SHEntries[0].sh_info;
-    }
-
-    printf_log(LOG_DEBUG, "Read %d Program header\n", h->numPHEntries);
-    h->PHEntries = (Elf32_Phdr*)calloc(h->numPHEntries, sizeof(Elf32_Phdr));
-    fseek(f, header.e_phoff ,SEEK_SET);
-    if(fread(h->PHEntries, sizeof(Elf32_Phdr), h->numPHEntries, f)!=h->numPHEntries) {
-            FreeElfHeader(&h);
-            printf_log(LOG_INFO, "Cannot read all Program Header\n");
-            return NULL;
-    }
-
-    if(h->SHIdx == SHN_XINDEX) {
-        printf_log(LOG_DEBUG, "Read number of String Table in 1st Section\n");
-        h->SHIdx = h->SHEntries[0].sh_link;
-    }
-    if(h->SHIdx > h->numSHEntries) {
-        printf_log(LOG_INFO, "Incoherent Section String Table Index : %d / %d\n", h->SHIdx, h->numSHEntries);
-        FreeElfHeader(&h);
-        return NULL;
-    }
-    // load Section table
-    printf_log(LOG_DEBUG, "Loading Sections Table String (idx = %d)\n", h->SHIdx);
-    if(LoadSH(f, h->SHEntries+h->SHIdx, (void*)&h->SHStrTab, ".shstrtab", SHT_STRTAB)) {
-        FreeElfHeader(&h);
-        return NULL;
-    }
-    if(box86_log>=LOG_DUMP) DumpMainHeader(&header, h);
-
-    LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".strtab", "SymTab Strings", SHT_STRTAB, (void**)&h->StrTab, NULL);
-    LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".symtab", "SymTab", SHT_SYMTAB, (void**)&h->SymTab, &h->numSymTab);
-    if(box86_log>=LOG_DUMP && h->SymTab) DumpSymTab(h);
-
-    LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".dynamic", "Dynamic", SHT_DYNAMIC, (void**)&h->Dynamic, &h->numDynamic);
-    if(box86_log>=LOG_DUMP && h->Dynamic) DumpDynamicSections(h);
-    // grab DT_REL & DT_RELA stuffs
-    // also grab the DT_STRTAB string table
-    {
-        for (int i=0; i<h->numDynamic; ++i) {
-            if(h->Dynamic[i].d_tag == DT_REL)
-                h->rel = h->Dynamic[i].d_un.d_ptr;
-            else if(h->Dynamic[i].d_tag == DT_RELSZ)
-                h->relsz = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_RELENT)
-                h->relent = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_RELA)
-                h->rela = h->Dynamic[i].d_un.d_ptr;
-            else if(h->Dynamic[i].d_tag == DT_RELASZ)
-                h->relasz = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_RELAENT)
-                h->relaent = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_PLTREL)
-                h->pltrel = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_PLTRELSZ)
-                h->pltsz = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_JMPREL)
-                h->jmprel = h->Dynamic[i].d_un.d_val;
-            else if(h->Dynamic[i].d_tag == DT_STRTAB)
-                h->DynStrTab = (char*)(h->Dynamic[i].d_un.d_ptr);
-            else if(h->Dynamic[i].d_tag == DT_STRSZ)
-                h->szDynStrTab = h->Dynamic[i].d_un.d_val;
-        }
-        if(h->rel) {
-            if(h->relent != sizeof(Elf32_Rel)) {
-                printf_log(LOG_NONE, "Rel Table Entry size invalid (0x%x should be 0x%x)\n", h->relent, sizeof(Elf32_Rel));
-                FreeElfHeader(&h);
-                return NULL;
-            }
-            printf_log(LOG_DEBUG, "Rel Table @%p (0x%x/0x%x)\n", (void*)h->rel, h->relsz, h->relent);
-        }
-        if(h->rela) {
-            if(h->relaent != sizeof(Elf32_Rela)) {
-                printf_log(LOG_NONE, "RelA Table Entry size invalid (0x%x should be 0x%x)\n", h->relaent, sizeof(Elf32_Rela));
-                FreeElfHeader(&h);
-                return NULL;
-            }
-            printf_log(LOG_DEBUG, "RelA Table @%p (0x%x/0x%x)\n", (void*)h->rela, h->relasz, h->relaent);
-        }
-        if(h->jmprel) {
-            if(h->pltrel == DT_REL) {
-                h->pltent = sizeof(Elf32_Rel);
-            } else if(h->pltrel == DT_RELA) {
-                h->pltent = sizeof(Elf32_Rela);
-            } else {
-                printf_log(LOG_NONE, "PLT Table type is unknown (size = 0x%x, type=%d)\n", h->pltsz, h->pltrel);
-                FreeElfHeader(&h);
-                return NULL;
-            }
-            if((h->pltsz / h->pltent)*h->pltent != h->pltsz) {
-                printf_log(LOG_NONE, "PLT Table Entry size invalid (0x%x, ent=0x%x, type=%d)\n", h->pltsz, h->pltent, h->pltrel);
-                FreeElfHeader(&h);
-                return NULL;
-            }
-            printf_log(LOG_DEBUG, "PLT Table @%p (type=%d 0x%x/0x%0x)\n", (void*)h->jmprel, h->pltrel, h->pltsz, h->pltent);
-        }
-        if(h->DynStrTab && h->szDynStrTab) {
-            //DumpDynamicNeeded(h); cannot dump now, it's not loaded yet
-        }
-    }
-    // look for PLT Offset
-    int ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got.plt");
-    if(ii) {
-        h->gotplt = h->SHEntries[ii].sh_addr;
-        printf_log(LOG_DEBUG, "The GOT.PLT Table is at address %p\n", (void*)h->gotplt);
-    }
-    ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got");
-    if(ii) {
-        h->got = h->SHEntries[ii].sh_addr;
-        printf_log(LOG_DEBUG, "The GOT Table is at address %p\n", (void*)h->got);
-    }
-    // look for .init entry point
-    ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".init");
-    if(ii) {
-        h->initentry = h->SHEntries[ii].sh_addr;
-        printf_log(LOG_DEBUG, "The .init is at address %p\n", (void*)h->initentry);
-    }
-    // look for .fini entry point
-    ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".fini");
-    if(ii) {
-        h->finientry = h->SHEntries[ii].sh_addr;
-        printf_log(LOG_DEBUG, "The .fini is at address %p\n", (void*)h->finientry);
-    }
-
-    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);
-    if(box86_log>=LOG_DUMP && h->DynSym) DumpDynSym(h);
-
-    return h;
-}

src/elfs/elfload_dump.c

@@ -183,6 +183,7 @@ const char* DumpRelType(int t)
         GO(R_386_GOTOFF);
         GO(R_386_GOTPC);
         GO(R_386_PC8);
+        GO(R_386_TLS_TPOFF);
         GO(R_386_TLS_GD_32);
         GO(R_386_TLS_DTPMOD32);
         GO(R_386_TLS_DTPOFF32);

src/elfs/elfloader.c

@@ -0,0 +1,956 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <elf.h>
+#include <sys/mman.h>
+#include <link.h>
+
+#include "box86version.h"
+#include "elfloader.h"
+#include "debug.h"
+#include "elfload_dump.h"
+#include "elfloader_private.h"
+#include "librarian.h"
+#include "x86run.h"
+#include "bridge.h"
+#include "wrapper.h"
+#include "box86context.h"
+#include "library.h"
+#include "x86emu.h"
+#include "box86stack.h"
+#include "callback.h"
+#include "dynarec.h"
+#ifdef DYNAREC
+#include "dynablock.h"
+#endif
+#include "../emu/x86emu_private.h"
+
+void* my__IO_2_1_stderr_ = NULL;
+void* my__IO_2_1_stdin_  = NULL;
+void* my__IO_2_1_stdout_ = NULL;
+
+// return the index of header (-1 if it doesn't exist)
+int getElfIndex(box86context_t* ctx, elfheader_t* head) {
+    for (int i=0; i<ctx->elfsize; ++i)
+        if(ctx->elfs[i]==head)
+            return i;
+    return -1;
+}
+
+elfheader_t* LoadAndCheckElfHeader(FILE* f, const char* name, int exec)
+{
+    elfheader_t *h = ParseElfHeader(f, name, exec);
+    if(!h)
+        return NULL;
+
+    return h;
+}
+
+void FreeElfHeader(elfheader_t** head)
+{
+    if(!head || !*head)
+        return;
+    elfheader_t *h = *head;
+#ifdef DYNAREC
+    dynarec_log(LOG_INFO, "Free Dynarec block for %s\n", h->name);
+    FreeDynablockList(&h->blocks);
+#endif
+    free(h->name);
+    free(h->PHEntries);
+    free(h->SHEntries);
+    free(h->SHStrTab);
+    free(h->StrTab);
+    free(h->Dynamic);
+    free(h->DynStr);
+    free(h->SymTab);
+    free(h->DynSym);
+
+    FreeElfMemory(h);
+    free(h);
+
+    *head = NULL;
+}
+
+int CalcLoadAddr(elfheader_t* head)
+{
+    head->memsz = 0;
+    head->paddr = head->vaddr = ~(uintptr_t)0;
+    head->align = 1;
+    for (int i=0; i<head->numPHEntries; ++i)
+        if(head->PHEntries[i].p_type == PT_LOAD) {
+            if(head->paddr > (uintptr_t)head->PHEntries[i].p_paddr)
+                head->paddr = (uintptr_t)head->PHEntries[i].p_paddr;
+            if(head->vaddr > (uintptr_t)head->PHEntries[i].p_vaddr)
+                head->vaddr = (uintptr_t)head->PHEntries[i].p_vaddr;
+        }
+    
+    if(head->vaddr==~(uintptr_t)0 || head->paddr==~(uintptr_t)0) {
+        printf_log(LOG_NONE, "Error: v/p Addr for Elf Load not set\n");
+        return 1;
+    }
+
+    head->stacksz = 1024*1024;          //1M stack size default?
+    head->stackalign = 4;   // default align for stack
+    for (int i=0; i<head->numPHEntries; ++i) {
+        if(head->PHEntries[i].p_type == PT_LOAD) {
+            uintptr_t phend = head->PHEntries[i].p_vaddr - head->vaddr + head->PHEntries[i].p_memsz;
+            if(phend > head->memsz)
+                head->memsz = phend;
+            if(head->PHEntries[i].p_align > head->align)
+                head->align = head->PHEntries[i].p_align;
+        }
+        if(head->PHEntries[i].p_type == PT_GNU_STACK) {
+            if(head->stacksz < head->PHEntries[i].p_memsz)
+                head->stacksz = head->PHEntries[i].p_memsz;
+            if(head->stackalign < head->PHEntries[i].p_align)
+                head->stackalign = head->PHEntries[i].p_align;
+        }
+        if(head->PHEntries[i].p_type == PT_TLS) {
+            head->tlssize = head->PHEntries[i].p_memsz;
+            head->tlsalign = head->PHEntries[i].p_align;
+            // force alignement...
+            if(head->tlsalign>1)
+                while(head->tlssize&(head->tlsalign-1))
+                    head->tlssize++;
+        }
+    }
+    printf_log(LOG_DEBUG, "Elf Addr(v/p)=%p/%p Memsize=0x%x (align=0x%x)\n", (void*)head->vaddr, (void*)head->paddr, head->memsz, head->align);
+    printf_log(LOG_DEBUG, "Elf Stack Memsize=%u (align=%u)\n", head->stacksz, head->stackalign);
+    printf_log(LOG_DEBUG, "Elf TLS Memsize=%u (align=%u)\n", head->tlssize, head->tlsalign);
+
+    return 0;
+}
+
+const char* ElfName(elfheader_t* head)
+{
+    return head->name;
+}
+
+int AllocElfMemory(box86context_t* context, elfheader_t* head, int mainbin)
+{
+    uintptr_t offs = 0;
+    if(mainbin && head->vaddr==0) {
+        char* load_addr = getenv("BOX86_LOAD_ADDR");
+        if(load_addr)
+            if(sscanf(load_addr, "0x%x", &offs)!=1)
+                offs = 0;
+    }
+    if(!offs)
+        offs = head->vaddr;
+    printf_log(LOG_DEBUG, "Allocating 0x%x memory @%p for Elf \"%s\"\n", head->memsz, (void*)offs, head->name);
+    void* p = mmap((void*)offs, head->memsz
+        , PROT_READ | PROT_WRITE | PROT_EXEC, MAP_PRIVATE | MAP_ANONYMOUS | ((offs)?MAP_FIXED:0)
+        , -1, 0);
+    if(p==MAP_FAILED) {
+        printf_log(LOG_NONE, "Cannot create memory map (@%p 0x%x/0x%x) for elf \"%s\"\n", (void*)offs, head->memsz, head->align, head->name);
+        return 1;
+    }
+    head->memory = p;
+    memset(p, 0, head->memsz);
+    head->delta = (intptr_t)p - (intptr_t)head->vaddr;
+    printf_log(LOG_DEBUG, "Got %p (delta=%p)\n", p, (void*)head->delta);
+
+    head->tlsbase = AddTLSPartition(context, head->tlssize);
+
+#ifdef DYNAREC
+    head->blocks = NewDynablockList((uintptr_t)GetBaseAddress(head), head->text + head->delta, head->textsz);
+#endif
+
+    return 0;
+}
+
+void FreeElfMemory(elfheader_t* head)
+{
+    if(head->memory)
+        munmap((void*)head->memory, head->memsz);
+}
+
+int LoadElfMemory(FILE* f, box86context_t* context, elfheader_t* head)
+{
+    for (int i=0; i<head->numPHEntries; ++i) {
+        if(head->PHEntries[i].p_type == PT_LOAD) {
+            Elf32_Phdr * e = &head->PHEntries[i];
+            char* dest = (char*)e->p_paddr + head->delta;
+            printf_log(LOG_DEBUG, "Loading block #%i @%p (0x%x/0x%x)\n", i, dest, e->p_filesz, e->p_memsz);
+            fseek(f, e->p_offset, SEEK_SET);
+            if(e->p_filesz) {
+                if(fread(dest, e->p_filesz, 1, f)!=1) {
+                    printf_log(LOG_NONE, "Fail to read PT_LOAD part #%d\n", i);
+                    return 1;
+                }
+            }
+            // zero'd difference between filesz and memsz
+            /*if(e->p_filesz != e->p_memsz)
+                memset(dest+e->p_filesz, 0, e->p_memsz - e->p_filesz);*/    //block is already 0'd at creation
+        }
+        if(head->PHEntries[i].p_type == PT_TLS) {
+            Elf32_Phdr * e = &head->PHEntries[i];
+            char* dest = (char*)context->tlsdata;
+            printf_log(LOG_DEBUG, "Loading TLS block #%i @%p (0x%x/0x%x)\n", i, dest, e->p_filesz, e->p_memsz);
+            fseek(f, e->p_offset, SEEK_SET);
+            if(e->p_filesz) {
+                if(fread(dest, e->p_filesz, 1, f)!=1) {
+                    printf_log(LOG_NONE, "Fail to read PT_TLS part #%d\n", i);
+                    return 1;
+                }
+            }
+            // zero'd difference between filesz and memsz
+            if(e->p_filesz != e->p_memsz)
+                memset(dest+e->p_filesz, 0, e->p_memsz - e->p_filesz);
+        }
+    }
+    return 0;
+}
+
+int ReloadElfMemory(FILE* f, box86context_t* context, elfheader_t* head)
+{
+    for (int i=0; i<head->numPHEntries; ++i) {
+        if(head->PHEntries[i].p_type == PT_LOAD) {
+            Elf32_Phdr * e = &head->PHEntries[i];
+            char* dest = (char*)e->p_paddr + head->delta;
+            printf_log(LOG_DEBUG, "Re-loading block #%i @%p (0x%x/0x%x)\n", i, dest, e->p_filesz, e->p_memsz);
+            fseek(f, e->p_offset, SEEK_SET);
+            if(e->p_filesz) {
+                if(fread(dest, e->p_filesz, 1, f)!=1) {
+                    printf_log(LOG_NONE, "Fail to read PT_LOAD part #%d\n", i);
+                    return 1;
+                }
+            }
+            // zero'd difference between filesz and memsz
+            if(e->p_filesz != e->p_memsz)
+                memset(dest+e->p_filesz, 0, e->p_memsz - e->p_filesz);
+        }
+    }
+    // TLS data are just a copy, no need to re-load it
+    return 0;
+}
+
+int RelocateElfREL(lib_t *maplib, elfheader_t* head, int cnt, Elf32_Rel *rel)
+{
+    for (int i=0; i<cnt; ++i) {
+        Elf32_Sym *sym = &head->DynSym[ELF32_R_SYM(rel[i].r_info)];
+        int type = ELF32_ST_TYPE(sym->st_info);
+        int bind = ELF32_ST_BIND(sym->st_info);
+        const char* symname = SymName(head, sym);
+        uint32_t *p = (uint32_t*)(rel[i].r_offset + head->delta);
+        uintptr_t offs = 0;
+        uint32_t sz = 0;
+        uintptr_t end = 0;
+        if(bind==STB_LOCAL)
+            GetLocalSymbolStartEnd(maplib, symname, &offs, &end, head);
+        else
+            GetGlobalSymbolStartEnd(maplib, symname, &offs, &end);
+        uintptr_t globoffs, globend;
+        int delta;
+        int t = ELF32_R_TYPE(rel[i].r_info);
+        switch(t) {
+            case R_386_NONE:
+                // can be ignored
+                printf_log(LOG_DEBUG, "Ignoring %s @%p (%p)\n", DumpRelType(t), p, (void*)(p?(*p):0));
+                break;
+            case R_386_TLS_TPOFF:
+                // Negated offset in static TLS block
+                {
+                    elfheader_t *h = GetGlobalSymbolElf(maplib, symname);
+                    if(h) {
+                        delta = *(int*)p;
+                        printf_log(LOG_DEBUG, "Applying %s %s on %s @%p (%d -> %d)\n", (bind==STB_LOCAL)?"Local":"Global", DumpRelType(t), symname, p, delta, (int32_t)offs + h->tlsbase);
+                        *p = (uint32_t)((int32_t)offs + h->tlsbase);
+                    }
+                }
+                break;
+            case R_386_PC32:
+                    if (!offs) {
+                        printf_log(LOG_NONE, "Error: Global Symbol %s not found, cannot apply R_386_PC32 @%p (%p)\n", symname, p, *(void**)p);
+                    }
+                    offs = (offs - (uintptr_t)p);
+                    if(!offs)
+                    printf_log(LOG_DEBUG, "Apply %s R_386_PC32 @%p with sym=%s (%p -> %p)\n", (bind==STB_LOCAL)?"Local":"Global", p, symname, *(void**)p, (void*)(*(uintptr_t*)p+offs));
+                    *p += offs;
+                break;
+            case R_386_GLOB_DAT:
+                // Look for same symbol already loaded but not in self
+                if (GetGlobalNoWeakSymbolStartEnd(maplib, symname, &globoffs, &globend)) {
+                    offs = globoffs;
+                    end = globend;
+                }
+                if (!offs) {
+                    printf_log(LOG_NONE, "Error: Global Symbol %s not found, cannot apply R_386_GLOB_DAT @%p (%p)\n", symname, p, *(void**)p);
+//                    return -1;
+                } else {
+                    printf_log(LOG_DEBUG, "Apply %s R_386_GLOB_DAT @%p (%p -> %p) on sym=%s\n", (bind==STB_LOCAL)?"Local":"Global", p, (void*)(p?(*p):0), (void*)offs, symname);
+                    *p = offs;
+                }
+                break;
+            case R_386_RELATIVE:
+                printf_log(LOG_DEBUG, "Apply %s R_386_RELATIVE @%p (%p -> %p)\n", (bind==STB_LOCAL)?"Local":"Global", p, *(void**)p, (void*)((*p)+head->delta));
+                *p += head->delta;
+                break;
+            case R_386_32:
+                if (!offs) {
+                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_32 @%p (%p)\n", symname, p, *(void**)p);
+//                    return -1;
+                } else {
+                    printf_log(LOG_DEBUG, "Apply %s R_386_32 @%p with sym=%s (%p -> %p)\n", (bind==STB_LOCAL)?"Local":"Global", p, symname, *(void**)p, (void*)(offs+*(uint32_t*)p));
+                    *p += offs;
+                }
+                break;
+            case R_386_TLS_DTPMOD32:
+                // ID of module containing symbol
+                if(!symname || symname[0]=='\0' || bind==STB_LOCAL)
+                    offs = getElfIndex(GetLibrarianContext(maplib), head);
+                else {
+                    elfheader_t *h = GetGlobalSymbolElf(maplib, symname);
+                    offs = getElfIndex(GetLibrarianContext(maplib), h);
+                }
+                if(p) {
+                    printf_log(LOG_DEBUG, "Apply %s %s @%p with sym=%s (%p -> %p)\n", "R_386_TLS_DTPMOD32", (bind==STB_LOCAL)?"Local":"Global", p, symname, *(void**)p, (void*)offs);
+                    *p = offs;
+                } else {
+                    printf_log(LOG_NONE, "Warning, Symbol %s or Elf not found, but R_386_TLS_DTPMOD32 Slot Offset is NULL \n", symname);
+                }
+                break;
+            case R_386_TLS_DTPOFF32:
+                // Offset in TLS block
+                if (!offs) {
+                    if(bind==STB_WEAK) {
+                        printf_log(LOG_INFO, "Warning: Weak Symbol %s not found, cannot apply R_386_TLS_DTPOFF32 @%p (%p)\n", symname, p, *(void**)p);
+                    } else {
+                        printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_TLS_DTPOFF32 @%p (%p)\n", symname, p, *(void**)p);
+                    }
+//                    return -1;
+                } else {
+                    if(p) {
+                        int tlsoffset = offs;    // it's not an offset in elf memory
+                        printf_log(LOG_DEBUG, "Apply %s R_386_TLS_DTPOFF32 @%p with sym=%s (%p -> %p)\n", (bind==STB_LOCAL)?"Local":"Global", p, symname, (void*)tlsoffset, (void*)offs);
+                        *p = tlsoffset;
+                    } else {
+                        printf_log(LOG_NONE, "Warning, Symbol %s found, but R_386_TLS_DTPOFF32 Slot Offset is NULL \n", symname);
+                    }
+                }
+                break;
+            case R_386_JMP_SLOT:
+                if (!offs) {
+                    if(bind==STB_WEAK) {
+                        printf_log(LOG_INFO, "Warning: Weak Symbol %s not found, cannot apply R_386_JMP_SLOT @%p (%p)\n", symname, p, *(void**)p);
+                    } else {
+                        printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_JMP_SLOT @%p (%p)\n", symname, p, *(void**)p);
+                    }
+//                    return -1;
+                } else {
+                    if(p) {
+                        printf_log(LOG_DEBUG, "Apply %s R_386_JMP_SLOT @%p with sym=%s (%p -> %p)\n", (bind==STB_LOCAL)?"Local":"Global", p, symname, *(void**)p, (void*)offs);
+                        *p = offs;
+                    } else {
+                        printf_log(LOG_NONE, "Warning, Symbol %s found, but Jump Slot Offset is NULL \n", symname);
+                    }
+                }
+                break;
+            case R_386_COPY:
+                if(offs) {
+                    GetNoSelfSymbolStartEnd(maplib, symname, &offs, &end, head);   // get original copy if any
+                    printf_log(LOG_DEBUG, "Apply %s R_386_COPY @%p with sym=%s, @%p size=%d (", (bind==STB_LOCAL)?"Local":"Global", p, symname, (void*)offs, sym->st_size);
+                    memmove(p, (void*)offs, sym->st_size);
+                    if(LOG_DEBUG<=box86_log) {
+                        uint32_t*k = (uint32_t*)p;
+                        for (int i=0; i<((sym->st_size>128)?128:sym->st_size); i+=4, ++k)
+                            printf_log(LOG_DEBUG, "%s0x%08X", i?" ":"", *k);
+                        printf_log(LOG_DEBUG, "%s)\n", (sym->st_size>128)?" ...":"");
+                    }
+                } else {
+                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_COPY @%p (%p)\n", symname, p, *(void**)p);
+                }
+                break;
+            default:
+                printf_log(LOG_INFO, "Warning, don't know of to handle rel #%d %s (%p)\n", i, DumpRelType(ELF32_R_TYPE(rel[i].r_info)), p);
+        }
+    }
+    return 0;
+}
+
+int RelocateElfRELA(lib_t *maplib, elfheader_t* head, int cnt, Elf32_Rela *rela)
+{
+    for (int i=0; i<cnt; ++i) {
+        Elf32_Sym *sym = &head->DynSym[ELF32_R_SYM(rela[i].r_info)];
+        const char* symname = SymName(head, sym);
+        uint32_t *p = (uint32_t*)(rela[i].r_offset + head->delta);
+        uintptr_t offs = 0;
+        uint32_t sz = 0;
+        uintptr_t end = 0;
+        switch(ELF32_R_TYPE(rela[i].r_info)) {
+            case R_386_NONE:
+            case R_386_PC32:
+                // can be ignored
+                break;
+            case R_386_COPY:
+                GetNoSelfSymbolStartEnd(maplib, symname, &offs, &end, head);
+                if(offs) {
+                    // add r_addend to p?
+                    printf_log(LOG_DEBUG, "Apply R_386_COPY @%p with sym=%s, @%p size=%d\n", p, symname, (void*)offs, sym->st_size);
+                    memcpy(p, (void*)(offs+rela[i].r_addend), sym->st_size);
+                } else {
+                    printf_log(LOG_NONE, "Error: Symbol %s not found, cannot apply R_386_COPY @%p (%p)\n", symname, p, *(void**)p);
+                }
+                break;
+            default:
+                printf_log(LOG_INFO, "Warning, don't know of to handle rela #%d %s on %s\n", i, DumpRelType(ELF32_R_TYPE(rela[i].r_info)), symname);
+        }
+    }
+    return 0;
+}
+int RelocateElf(lib_t *maplib, elfheader_t* head)
+{
+    if(head->rel) {
+        int cnt = head->relsz / head->relent;
+        DumpRelTable(head, cnt, (Elf32_Rel *)(head->rel + head->delta), "Rel");
+        printf_log(LOG_DEBUG, "Applying %d Relocation(s) for %s\n", cnt, head->name);
+        if(RelocateElfREL(maplib, head, cnt, (Elf32_Rel *)(head->rel + head->delta)))
+            return -1;
+    }
+    if(head->rela) {
+        int cnt = head->relasz / head->relaent;
+        DumpRelATable(head, cnt, (Elf32_Rela *)(head->rela + head->delta), "RelA");
+        printf_log(LOG_DEBUG, "Applying %d Relocation(s) with Addend for %s\n", cnt, head->name);
+        if(RelocateElfRELA(maplib, head, cnt, (Elf32_Rela *)(head->rela + head->delta)))
+            return -1;
+    }
+   
+    return 0;
+}
+
+int RelocateElfPlt(box86context_t* context, lib_t *maplib, elfheader_t* head)
+{
+    if(head->pltrel) {
+        int cnt = head->pltsz / head->pltent;
+        if(head->pltrel==DT_REL) {
+            DumpRelTable(head, cnt, (Elf32_Rel *)(head->jmprel + head->delta), "PLT");
+            printf_log(LOG_DEBUG, "Applying %d PLT Relocation(s) for %s\n", cnt, head->name);
+            if(RelocateElfREL(maplib, head, cnt, (Elf32_Rel *)(head->jmprel + head->delta)))
+                return -1;
+        } else if(head->pltrel==DT_RELA) {
+            DumpRelATable(head, cnt, (Elf32_Rela *)(head->jmprel + head->delta), "PLT");
+            printf_log(LOG_DEBUG, "Applying %d PLT Relocation(s) with Addend for %s\n", cnt, head->name);
+            if(RelocateElfRELA(maplib, head, cnt, (Elf32_Rela *)(head->jmprel + head->delta)))
+                return -1;
+        }
+        if(head->gotplt) {
+            if(pltResolver==~0) {
+                pltResolver = AddBridge(context->system, vFEpp, PltResolver, 0);   // should be vFEuu
+            }
+            *(uintptr_t*)(head->gotplt+head->delta+8) = pltResolver;
+            printf_log(LOG_DEBUG, "PLT Resolver injected at %p\n", (void*)(head->gotplt+head->delta+8));
+        }
+    }
+   
+    return 0;
+}
+
+void CalcStack(elfheader_t* elf, uint32_t* stacksz, int* stackalign)
+{
+    if(*stacksz < elf->stacksz)
+        *stacksz = elf->stacksz;
+    if(*stackalign < elf->stackalign)
+        *stackalign = elf->stackalign;
+}
+
+Elf32_Sym* GetFunction(elfheader_t* h, const char* name)
+{
+    // TODO: create a hash on named to avoid this loop
+    for (int i=0; i<h->numSymTab; ++i) {
+        if(h->SymTab[i].st_info == 18) {    // TODO: this "18" is probably defined somewhere
+            const char * symname = h->StrTab+h->SymTab[i].st_name;
+            if(strcmp(symname, name)==0) {
+                return h->SymTab+i;
+            }
+        }
+    }
+    return NULL;
+}
+
+Elf32_Sym* GetElfObject(elfheader_t* h, const char* name)
+{
+    for (int i=0; i<h->numSymTab; ++i) {
+        if(h->SymTab[i].st_info == 16) {
+            const char * symname = h->StrTab+h->SymTab[i].st_name;
+            if(strcmp(symname, name)==0) {
+                return h->SymTab+i;
+            }
+        }
+    }
+    return NULL;
+}
+
+
+uintptr_t GetFunctionAddress(elfheader_t* h, const char* name)
+{
+    Elf32_Sym* sym = GetFunction(h, name);
+    if(sym) return sym->st_value;
+    return 0;
+}
+
+uintptr_t GetEntryPoint(lib_t* maplib, elfheader_t* h)
+{
+    uintptr_t ep = h->entrypoint + h->delta;
+    printf_log(LOG_DEBUG, "Entry Point is %p\n", (void*)ep);
+    if(box86_log>=LOG_DUMP) {
+        printf_log(LOG_DUMP, "(short) Dump of Entry point\n");
+        int sz = 64;
+        uintptr_t lastbyte = GetLastByte(h);
+        if (ep + sz >  lastbyte)
+            sz = lastbyte - ep;
+        DumpBinary((char*)ep, sz);
+    }
+    /*
+    // but instead of regular entrypoint, lets grab "main", it will be easier to manage I guess
+    uintptr_t m = FindSymbol(maplib, "main");
+    if(m) {
+        ep = m;
+        printf_log(LOG_DEBUG, "Using \"main\" as Entry Point @%p\n", ep);
+        if(box86_log>=LOG_DUMP) {
+            printf_log(LOG_DUMP, "(short) Dump of Entry point\n");
+            int sz = 64;
+            uintptr_t lastbyte = GetLastByte(h);
+            if (ep + sz >  lastbyte)
+                sz = lastbyte - ep;
+            DumpBinary((char*)ep, sz);
+        }
+    }
+    */
+    return ep;
+}
+
+uintptr_t GetLastByte(elfheader_t* h)
+{
+    return (uintptr_t)h->memory/* + h->delta*/ + h->memsz;
+}
+
+void AddSymbols(lib_t *maplib, kh_mapsymbols_t* mapsymbols, kh_mapsymbols_t* weaksymbols, kh_mapsymbols_t* localsymbols, elfheader_t* h)
+{
+    printf_log(LOG_DUMP, "Will look for Symbol to add in SymTable(%d)\n", h->numSymTab);
+    for (int i=0; i<h->numSymTab; ++i) {
+        const char * symname = h->StrTab+h->SymTab[i].st_name;
+        int bind = ELF32_ST_BIND(h->SymTab[i].st_info);
+        int type = ELF32_ST_TYPE(h->SymTab[i].st_info);
+        if((type==STT_OBJECT || type==STT_FUNC || type==STT_COMMON || type==STT_TLS  || type==STT_NOTYPE) 
+        && (h->SymTab[i].st_other==0) && (h->SymTab[i].st_shndx!=0)) {
+            if(bind==10/*STB_GNU_UNIQUE*/ && FindGlobalSymbol(maplib, symname))
+                continue;
+            uintptr_t offs = (type==STT_TLS)?h->SymTab[i].st_value:(h->SymTab[i].st_value + h->delta);
+            uint32_t sz = h->SymTab[i].st_size;
+            printf_log(LOG_DUMP, "Adding Symbol(bind=%s) \"%s\" with offset=%p sz=%d\n", (bind==STB_LOCAL)?"LOCAL":((bind==STB_WEAK)?"WEAK":"GLOBAL"), symname, (void*)offs, sz);
+            if(bind==STB_LOCAL)
+                AddSymbol(localsymbols, symname, offs, sz);
+            else    // add in local and global map 
+                if(bind==STB_WEAK) {
+                    AddSymbol(weaksymbols, symname, offs, sz);
+                } else {
+                    AddSymbol(mapsymbols, symname, offs, sz);
+                }
+        }
+    }
+    
+    printf_log(LOG_DUMP, "Will look for Symbol to add in DynSym (%d)\n", h->numDynSym);
+    for (int i=0; i<h->numDynSym; ++i) {
+        const char * symname = h->DynStr+h->DynSym[i].st_name;
+        int bind = ELF32_ST_BIND(h->DynSym[i].st_info);
+        int type = ELF32_ST_TYPE(h->DynSym[i].st_info);
+        if((type==STT_OBJECT || type==STT_FUNC || type==STT_COMMON || type==STT_TLS  || type==STT_NOTYPE) 
+        && (h->DynSym[i].st_other==0) && (h->DynSym[i].st_shndx!=0 && h->DynSym[i].st_shndx<62521)) {
+            if(bind==10/*STB_GNU_UNIQUE*/ && FindGlobalSymbol(maplib, symname))
+                continue;
+            uintptr_t offs = (type==STT_TLS)?h->DynSym[i].st_value:(h->DynSym[i].st_value + h->delta);
+            uint32_t sz = h->DynSym[i].st_size;
+            printf_log(LOG_DUMP, "Adding Symbol(bind=%s) \"%s\" with offset=%p sz=%d\n", (bind==STB_LOCAL)?"LOCAL":((bind==STB_WEAK)?"WEAK":"GLOBAL"), symname, (void*)offs, sz);
+            if(bind==STB_LOCAL)
+                AddSymbol(localsymbols, symname, offs, sz);
+            else // add in local and global map 
+                if(bind==STB_WEAK) {
+                    AddSymbol(weaksymbols, symname, offs, sz);
+                } else {
+                    AddSymbol(mapsymbols, symname, offs, sz);
+                }
+        }
+    }
+    
+}
+
+/*
+$ORIGIN – Provides the directory the object was loaded from. This token is typical
+used for locating dependencies in unbundled packages. For more details of this
+token expansion, see “Locating Associated Dependencies”
+$OSNAME – Expands to the name of the operating system (see the uname(1) man
+page description of the -s option). For more details of this token expansion, see
+“System Specific Shared Objects”
+$OSREL – Expands to the operating system release level (see the uname(1) man
+page description of the -r option). For more details of this token expansion, see
+“System Specific Shared Objects”
+$PLATFORM – Expands to the processor type of the current machine (see the
+uname(1) man page description of the -i option). For more details of this token
+expansion, see “System Specific Shared Objects”
+*/
+int LoadNeededLibs(elfheader_t* h, lib_t *maplib, library_t* parent, box86context_t *box86, x86emu_t* emu)
+{
+   DumpDynamicNeeded(h);
+   for (int i=0; i<h->numDynamic; ++i)
+        if(h->Dynamic[i].d_tag==DT_NEEDED) {
+            char *needed = h->DynStrTab+h->delta+h->Dynamic[i].d_un.d_val;
+            // TODO: Add LD_LIBRARY_PATH and RPATH Handling
+            if(AddNeededLib(maplib, parent, needed, box86, emu)) {
+                printf_log(LOG_INFO, "Error loading needed lib: \"%s\"\n", needed);
+                if(!allow_missing_libs)
+                    return 1;   //error...
+            }
+        }
+    return 0;
+}
+int FinalizeNeededLibs(elfheader_t* h, lib_t *maplib, box86context_t *box86, x86emu_t* emu)
+{
+   DumpDynamicNeeded(h);
+   for (int i=0; i<h->numDynamic; ++i)
+        if(h->Dynamic[i].d_tag==DT_NEEDED) {
+            char *needed = h->DynStrTab+h->delta+h->Dynamic[i].d_un.d_val;
+            // TODO: Add LD_LIBRARY_PATH and RPATH Handling
+            if(FinalizeLibrary(GetLib(maplib, needed), emu)) {
+                printf_log(LOG_INFO, "Error finalizing needed lib: \"%s\"\n", needed);
+                return 1;   //error...
+            }
+        }
+    return 0;
+}
+
+void RunElfInit(elfheader_t* h, x86emu_t *emu)
+{
+    if(h->init_done || !h->initentry)
+        return;
+    uintptr_t p = h->initentry + h->delta;
+    box86context_t* context = GetEmuContext(emu);
+    if(context->deferedInit) {
+        if(context->deferedInitSz==context->deferedInitCap) {
+            context->deferedInitCap += 4;
+            context->deferedInitList = (elfheader_t**)realloc(context->deferedInitList, context->deferedInitCap*sizeof(elfheader_t*));
+        }
+        context->deferedInitList[context->deferedInitSz++] = h;
+        return;
+    }
+    printf_log(LOG_DEBUG, "Calling Init for %s @%p\n", ElfName(h), (void*)p);
+    Push32(emu, (uintptr_t)context->envv);
+    Push32(emu, (uintptr_t)context->argv);
+    Push32(emu, context->argc);
+    DynaCall(emu, p);
+    printf_log(LOG_DEBUG, "Done Init for %s\n", ElfName(h));
+    // and check init array now
+    Elf32_Addr *addr = (Elf32_Addr*)(h->initarray + h->delta);
+    for (int i=0; i<h->initarray_sz; ++i) {
+        printf_log(LOG_DEBUG, "Calling Init[%d] for %s @%p\n", i, ElfName(h), (void*)addr[i]);
+        DynaCall(emu, (uintptr_t)addr[i]);
+    }
+    Pop32(emu);
+    Pop32(emu);
+    Pop32(emu);
+    h->init_done = 1;
+    h->fini_done = 0;   // can be fini'd now (in case it was re-inited)
+    printf_log(LOG_DEBUG, "All Init Done for %s\n", ElfName(h));
+    return;
+}
+void RunDeferedElfInit(x86emu_t *emu)
+{
+    box86context_t* context = GetEmuContext(emu);
+    if(!context->deferedInit)
+        return;
+    context->deferedInit = 0;
+    if(!context->deferedInitList)
+        return;
+    for (int i=0; i<context->deferedInitSz; ++i)
+        RunElfInit(context->deferedInitList[i], emu);
+    free(context->deferedInitList);
+    context->deferedInitList = NULL;
+    context->deferedInitCap = context->deferedInitSz = 0;
+}
+void RunElfFini(elfheader_t* h, x86emu_t *emu)
+{
+    if(h->fini_done || !h->finientry)
+        return;
+    uintptr_t p = h->finientry + h->delta;
+    printf_log(LOG_DEBUG, "Calling Fini for %s @%p\n", ElfName(h), (void*)p);
+    uint32_t sESP = GetESP(emu);
+    Push32(emu, (uintptr_t)GetEmuContext(emu)->envv);
+    Push32(emu, (uintptr_t)GetEmuContext(emu)->argv);
+    Push32(emu, GetEmuContext(emu)->argc);
+    DynaCall(emu, p);
+    // and check fini array now
+    Elf32_Addr *addr = (Elf32_Addr*)(h->finiarray + h->delta);
+    for (int i=0; i<h->finiarray_sz; ++i) {
+        printf_log(LOG_DEBUG, "Calling Fini[%d] for %s @%p\n", i, ElfName(h), (void*)addr[i]);
+        DynaCall(emu, (uintptr_t)addr[i]);
+    }
+    SetESP(emu, sESP);
+    h->fini_done = 1;
+    h->init_done = 0;   // can be re-inited again...
+    return;
+}
+
+uintptr_t GetElfInit(elfheader_t* h)
+{
+    return h->initentry + h->delta;
+}
+uintptr_t GetElfFini(elfheader_t* h)
+{
+    return h->finientry + h->delta;
+}
+
+void* GetBaseAddress(elfheader_t* h)
+{
+    return h->memory;
+}
+
+uint32_t GetBaseSize(elfheader_t* h)
+{
+    return h->memsz;
+}
+
+int IsAddressInElfSpace(elfheader_t* h, uintptr_t addr)
+{
+    if(!h)
+        return 0;
+    uintptr_t base = (uintptr_t)GetBaseAddress(h);
+    uintptr_t end = GetLastByte(h);
+    if(addr>=base && addr<=end)
+        return 1;
+    return 0;
+}
+elfheader_t* FindElfAddress(box86context_t *context, uintptr_t addr)
+{
+    for (int i=0; i<context->elfsize; ++i)
+        if(IsAddressInElfSpace(context->elfs[i], addr))
+            return context->elfs[i];
+    
+    return NULL;
+}
+
+const char* FindNearestSymbolName(elfheader_t* h, void* p, uintptr_t* start, uint32_t* sz)
+{
+    uintptr_t addr = (uintptr_t)p;
+
+    uint32_t distance = 0x7fffffff;
+    const char* ret = NULL;
+    uintptr_t s = 0;
+    uint32_t size = 0;
+    if(!h)
+        return ret;
+
+    for (int i=0; i<h->numSymTab && distance!=0; ++i) {   
+        const char * symname = h->StrTab+h->SymTab[i].st_name;
+        uintptr_t offs = h->SymTab[i].st_value + h->delta;
+
+        if(offs<addr) {
+            if(distance>addr-offs) {
+                distance = addr-offs;
+                ret = symname;
+                s = offs;
+                size = h->SymTab[i].st_size;
+            }
+        }
+    }
+    for (int i=0; i<h->numDynSym && distance!=0; ++i) {   
+        const char * symname = h->DynStr+h->DynSym[i].st_name;
+        uintptr_t offs = h->DynSym[i].st_value + h->delta;
+
+        if(offs<addr) {
+            if(distance>addr-offs) {
+                distance = addr-offs;
+                ret = symname;
+                s = offs;
+                size = h->DynSym[i].st_size;
+            }
+        }
+    }
+
+    if(start)
+        *start = s;
+    if(sz)
+        *sz = size;
+
+    return ret;
+}
+
+static void* fillTLSData(box86context_t *context)
+{
+        void* ptr = NULL;
+        pthread_mutex_lock(&context->mutex_lock);
+        // Setup the GS segment:
+        int dtsize = context->elfsize*8;
+        ptr = (char*)malloc(context->tlssize+0x50+dtsize+24); // plan a 8*4 dtp table
+        memcpy(ptr, context->tlsdata, context->tlssize);
+        tlsdatasize_t *data = (tlsdatasize_t*)calloc(1, sizeof(tlsdatasize_t));
+        data->tlsdata = ptr;
+        data->tlssize = context->tlssize;
+        pthread_setspecific(context->tlskey, data);
+        // copy canary...
+        memset((void*)((uintptr_t)ptr+context->tlssize), 0, 0x50+dtsize+24);        // set to 0 remining bytes
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x14), context->canary, 4);  // put canary in place
+        uintptr_t unknown = (uintptr_t)ptr+context->tlssize;
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x0), &unknown, 4);
+        uintptr_t dtp = (uintptr_t)ptr+context->tlssize+0x50;
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x4), &dtp, 4);
+        if(dtsize) {
+            for (int i=0; i<context->elfsize; ++i) {
+                // set pointer
+                dtp = (uintptr_t)ptr + (context->tlssize + GetTLSBase(context->elfs[i]));
+                memcpy((void*)((uintptr_t)ptr+context->tlssize+0x50+i*8), &dtp, 4);
+                memcpy((void*)((uintptr_t)ptr+context->tlssize+0x50+i*8+4), &i, 4); // index
+            }
+        }
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x10), &context->vsyscall, 4);  // address of vsyscall
+        pthread_mutex_unlock(&context->mutex_lock);
+        return data;
+}
+
+static void* resizeTLSData(box86context_t *context, void* oldptr)
+{
+        void* ptr = NULL;
+        pthread_mutex_lock(&context->mutex_lock);
+        tlsdatasize_t* oldata = (tlsdatasize_t*)oldptr;
+        tlsdatasize_t *data = (tlsdatasize_t*)calloc(1, sizeof(tlsdatasize_t));
+        // Setup the GS segment:
+        int dtsize = context->elfsize*8;
+        ptr = (char*)malloc(context->tlssize+0x50+dtsize+24); // plan a 8*4 dtp table
+        data->tlsdata = ptr;
+        data->tlssize = context->tlssize;
+
+        memcpy(ptr, context->tlsdata, context->tlssize);
+        pthread_setspecific(context->tlskey, data);
+        // copy canary...
+        memset((void*)((uintptr_t)ptr+context->tlssize), 0, 0x50+dtsize+24);        // set to 0 remining bytes
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x14), context->canary, 4);  // put canary in place
+        uintptr_t unknown = (uintptr_t)ptr+context->tlssize;
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x0), &unknown, 4);
+        uintptr_t dtp = (uintptr_t)ptr+context->tlssize+0x50;
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x4), &dtp, 4);
+        if(dtsize) {
+            for (int i=0; i<context->elfsize; ++i) {
+                // set pointer
+                dtp = (uintptr_t)ptr + (context->tlssize + GetTLSBase(context->elfs[i]));
+                memcpy((void*)((uintptr_t)ptr+context->tlssize+0x50+i*8), &dtp, 4);
+                memcpy((void*)((uintptr_t)ptr+context->tlssize+0x50+i*8+4), &i, 4); // index
+            }
+        }
+        memcpy((void*)((uintptr_t)ptr+context->tlssize+0x10), &context->vsyscall, 4);  // address of vsyscall
+        // copy the relevent old part, in case something changed
+        memcpy((void*)((uintptr_t)ptr+(context->tlssize-oldata->tlssize)), oldata->tlsdata, oldata->tlssize);
+        // all done, update new size, free old pointer and exit
+        pthread_mutex_unlock(&context->mutex_lock);
+        free_tlsdatasize(oldptr);
+        return data;
+}
+
+void* GetGSBase(box86context_t *context)
+{
+    tlsdatasize_t* ptr;
+    if ((ptr = (tlsdatasize_t*)pthread_getspecific(context->tlskey)) == NULL) {
+        ptr = (tlsdatasize_t*)fillTLSData(context);
+    }
+    if(ptr->tlssize != context->tlssize)
+        ptr = (tlsdatasize_t*)resizeTLSData(context, ptr);
+    return ptr->tlsdata+ptr->tlssize;
+}
+
+void* GetDTatOffset(box86context_t* context, int index, int offset)
+{
+    return (void*)((char*)GetTLSPointer(context, context->elfs[index])+offset);
+}
+
+int32_t GetTLSBase(elfheader_t* h)
+{
+    return h->tlsbase;
+}
+
+uint32_t GetTLSSize(elfheader_t* h)
+{
+    return h->tlssize;
+}
+
+void* GetTLSPointer(box86context_t* context, elfheader_t* h)
+{
+    if(!h->tlssize)
+        return NULL;
+    tlsdatasize_t* ptr;
+    if ((ptr = (tlsdatasize_t*)pthread_getspecific(context->tlskey)) == NULL) {
+        ptr = (tlsdatasize_t*)fillTLSData(context);
+    }
+    if(ptr->tlssize != context->tlssize)
+        ptr = (tlsdatasize_t*)resizeTLSData(context, ptr);
+    return ptr->tlsdata+(ptr->tlssize+h->tlsbase);
+}
+
+#ifdef DYNAREC
+dynablocklist_t* GetDynablocksFromAddress(box86context_t *context, uintptr_t addr)
+{
+    elfheader_t* elf = FindElfAddress(context, addr);
+    if(!elf) {
+        if((*(uint8_t*)addr)==0xCC)
+            return context->dynablocks;
+        if(box86_dynarec_forced)
+            return context->dynablocks;
+        dynarec_log(LOG_INFO, "Address %p not found in Elf memory and is not a native call wrapper\n", addr);
+        return NULL;
+    }
+    return elf->blocks;
+}
+#endif
+
+typedef struct my_dl_phdr_info_s {
+    void*           dlpi_addr;
+    const char*     dlpi_name;
+    Elf32_Phdr*     dlpi_phdr;
+    int             dlpi_phnum;
+} my_dl_phdr_info_t;
+
+static int dl_iterate_phdr_callback(x86emu_t *emu, my_dl_phdr_info_t *info, size_t size, void* data)
+{
+    SetCallbackArg(emu, 0, info);
+    SetCallbackArg(emu, 1, (void*)size);
+    int ret = RunCallback(emu);
+    return ret;
+}
+
+EXPORT int my_dl_iterate_phdr(x86emu_t *emu, void* F, void *data) {
+    printf_log(LOG_INFO, "Warning: call to partially implemented dl_iterate_phdr(%p, %p)\n", F, data);
+    x86emu_t* cbemu = AddSharedCallback(emu, (uintptr_t)F, 3, NULL, NULL, data, NULL);
+    box86context_t *context = GetEmuContext(emu);
+    const char* empty = "";
+    for (int idx=0; idx<context->elfsize; ++idx) {
+        my_dl_phdr_info_t info;
+        info.dlpi_addr = GetBaseAddress(context->elfs[idx]);
+        info.dlpi_name = idx?context->elfs[idx]->name:empty;    //1st elf is program, and this one doesn't get a name
+        info.dlpi_phdr = context->elfs[idx]->PHEntries;
+        info.dlpi_phnum = context->elfs[idx]->numPHEntries;
+        if(dl_iterate_phdr_callback(cbemu, &info, sizeof(info), data))
+            break;
+    }
+    FreeCallback(cbemu);
+    return 0;
+}
+
+void ResetSpecialCaseMainElf(elfheader_t* h)
+{
+    Elf32_Sym *sym = NULL;
+     for (int i=0; i<h->numDynSym; ++i) {
+        if(h->DynSym[i].st_info == 17) {
+            sym = h->DynSym+i;
+            const char * symname = h->DynStr+sym->st_name;
+            if(strcmp(symname, "_IO_2_1_stderr_")==0) {
+                memcpy((void*)sym->st_value+h->delta, stderr, sym->st_size);
+                my__IO_2_1_stderr_ = (void*)sym->st_value+h->delta;
+            } else
+            if(strcmp(symname, "_IO_2_1_stdin_")==0) {
+                memcpy((void*)sym->st_value+h->delta, stdin, sym->st_size);
+                my__IO_2_1_stdin_ = (void*)sym->st_value+h->delta;
+            } else
+            if(strcmp(symname, "_IO_2_1_stdout_")==0) {
+                memcpy((void*)sym->st_value+h->delta, stdout, sym->st_size);
+                my__IO_2_1_stdout_ = (void*)sym->st_value+h->delta;
+            }
+        }
+    }
+}

src/elfs/elfloader_private.h

@@ -1,6 +1,12 @@
 #ifndef __ELFLOADER_PRIVATE_H_
 #define __ELFLOADER_PRIVATE_H_
 
+#ifdef DYNAREC
+typedef struct dynablocklist_s dynablocklist_t;
+#endif
+
+#include <pthread.h>
+
 struct elfheader_s {
     char*       name;
     int         numPHEntries;
@@ -25,7 +31,11 @@ struct elfheader_s {
 
     uintptr_t   entrypoint;
     uintptr_t   initentry;
+    uintptr_t   initarray;
+    int         initarray_sz;
     uintptr_t   finientry;
+    uintptr_t   finiarray;
+    int         finiarray_sz;
 
     uintptr_t   rel;
     int         relsz;
@@ -39,6 +49,8 @@ struct elfheader_s {
     uint32_t    pltrel;
     uintptr_t   gotplt;
     uintptr_t   got;
+    uintptr_t   text;
+    int         textsz;
 
     uintptr_t   paddr;
     uintptr_t   vaddr;
@@ -46,11 +58,18 @@ struct elfheader_s {
     uint32_t    memsz;
     uint32_t    stacksz;
     int         stackalign;
+    uint32_t    tlssize;
+    int         tlsalign;
+
+    int32_t     tlsbase;    // the base of the tlsdata in the global tlsdata (always negative)
 
     int         init_done;
     int         fini_done;
 
     char*       memory; // char* and not void* to allow math on memory pointer
+#ifdef DYNAREC
+    dynablocklist_t *blocks;
+#endif
 };
 
 #define R_386_NONE	0
@@ -65,4 +84,6 @@ struct elfheader_s {
 #define R_386_GOTOFF	9
 #define R_386_GOTPC	10
 
+elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec);
+
 #endif //__ELFLOADER_PRIVATE_H_

src/elfs/elfparser.c

@@ -0,0 +1,296 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <elf.h>
+
+#include "box86version.h"
+#include "elfloader.h"
+#include "debug.h"
+#include "elfload_dump.h"
+#include "elfloader_private.h"
+
+#ifndef PN_XNUM 
+#define PN_XNUM (0xffff)
+#endif
+
+int LoadSH(FILE *f, Elf32_Shdr *s, void** SH, const char* name, uint32_t type)
+{
+    if(type && (s->sh_type != type)) {
+        printf_log(LOG_INFO, "Section Header \"%s\" (off=%d, size=%d) has incorect type (%d != %d)\n", name, s->sh_offset, s->sh_size, s->sh_type, type);
+        return -1;
+    }
+    if (type==SHT_SYMTAB && s->sh_size%sizeof(Elf32_Sym)) {
+        printf_log(LOG_INFO, "Section Header \"%s\" (off=%d, size=%d) has size (not multiple of %d)\n", name, s->sh_offset, s->sh_size, sizeof(Elf32_Sym));
+    }
+    *SH = calloc(1, s->sh_size);
+    fseek(f, s->sh_offset ,SEEK_SET);
+    if(fread(*SH, s->sh_size, 1, f)!=1) {
+            printf_log(LOG_INFO, "Cannot read Section Header \"%s\" (off=%d, size=%d)\n", name, s->sh_offset, s->sh_size);
+            return -1;
+    }
+
+    return 0;
+}
+
+int FindSection(Elf32_Shdr *s, int n, char* SHStrTab, const char* name)
+{
+    for (int i=0; i<n; ++i) {
+        if(s[i].sh_type!=SHT_NULL)
+            if(!strcmp(SHStrTab+s[i].sh_name, name))
+                return i;
+    }
+    return 0;
+}
+
+void LoadNamedSection(FILE *f, Elf32_Shdr *s, int size, char* SHStrTab, const char* name, const char* clearname, uint32_t type, void** what, int* num)
+{
+    int n = FindSection(s, size, SHStrTab, name);
+    printf_log(LOG_DEBUG, "Loading %s (idx = %d)\n", clearname, n);
+    if(n)
+        LoadSH(f, s+n, what, name, type);
+    if(type==SHT_SYMTAB || type==SHT_DYNSYM) {
+        if(*what && num)
+            *num = s[n].sh_size / sizeof(Elf32_Sym);
+    } else if(type==SHT_DYNAMIC) {
+        if(*what && num)
+            *num = s[n].sh_size / sizeof(Elf32_Dyn);
+    }
+}
+
+elfheader_t* ParseElfHeader(FILE* f, const char* name, int exec)
+{
+    Elf32_Ehdr header;
+    if(fread(&header, sizeof(Elf32_Ehdr), 1, f)!=1) {
+        printf_log(LOG_INFO, "Cannot read ELF Header\n");
+        return NULL;
+    }
+    if(memcmp(header.e_ident, ELFMAG, SELFMAG)!=0) {
+        printf_log(LOG_INFO, "Not an ELF file (sign=%c%c%c%c)\n", header.e_ident[0], header.e_ident[1], header.e_ident[2], header.e_ident[3]);
+        return NULL;
+    }
+    if(header.e_ident[EI_CLASS]!=ELFCLASS32) {
+        printf_log(LOG_INFO, "Not an 32bits ELF (%d)\n", header.e_ident[EI_CLASS]);
+        return NULL;
+    }
+    if(header.e_ident[EI_DATA]!=ELFDATA2LSB) {
+        printf_log(LOG_INFO, "Not an LittleEndian ELF (%d)\n", header.e_ident[EI_DATA]);
+        return NULL;
+    }
+    if(header.e_ident[EI_VERSION]!=EV_CURRENT) {
+        printf_log(LOG_INFO, "Incorrect ELF version (%d)\n", header.e_ident[EI_VERSION]);
+        return NULL;
+    }
+    if(header.e_ident[EI_OSABI]!=ELFOSABI_LINUX && header.e_ident[EI_OSABI]!=ELFOSABI_NONE && header.e_ident[EI_OSABI]!=ELFOSABI_SYSV) {
+        printf_log(LOG_INFO, "Not a Linux ELF (%d)\n",header.e_ident[EI_OSABI]);
+        return NULL;
+    }
+
+    if(header.e_type != ET_EXEC && header.e_type != ET_DYN) {
+        printf_log(LOG_INFO, "Not an Executable (%d)\n", header.e_type);
+        return NULL;
+    }
+
+    if(header.e_machine != EM_386) {
+        printf_log(LOG_INFO, "Not an i386 ELF (%d)\n", header.e_machine);
+        return NULL;
+    }
+
+    if(header.e_entry == 0 && exec) {
+        printf_log(LOG_INFO, "No entry point in ELF\n");
+        return NULL;
+    }
+    if(header.e_phentsize != sizeof(Elf32_Phdr)) {
+        printf_log(LOG_INFO, "Program Header Entry size incorrect (%d != %d)\n", header.e_phentsize, sizeof(Elf32_Phdr));
+        return NULL;
+    }
+    if(header.e_shentsize != sizeof(Elf32_Shdr) && header.e_shentsize != 0) {
+        printf_log(LOG_INFO, "Section Header Entry size incorrect (%d != %d)\n", header.e_shentsize, sizeof(Elf32_Shdr));
+        return NULL;
+    }
+
+    elfheader_t *h = calloc(1, sizeof(elfheader_t));
+    h->name = strdup(name);
+    h->entrypoint = header.e_entry;
+    h->numPHEntries = header.e_phnum;
+    h->numSHEntries = header.e_shnum;
+    h->SHIdx = header.e_shstrndx;
+    if(header.e_shentsize!=0) {
+        // special cases for nums
+        if(h->numSHEntries == 0) {
+            printf_log(LOG_DEBUG, "Read number of Sections in 1st Section\n");
+            // read 1st section header and grab actual number from here
+            fseek(f, header.e_shoff, SEEK_SET);
+            Elf32_Shdr section;
+            if(fread(&section, sizeof(Elf32_Shdr), 1, f)!=1) {
+                free(h);
+                printf_log(LOG_INFO, "Cannot read Initial Section Header\n");
+                return NULL;
+            }
+            h->numSHEntries = section.sh_size;
+        }
+        // now read all section headers
+        printf_log(LOG_DEBUG, "Read %d Section header\n", h->numSHEntries);
+        h->SHEntries = (Elf32_Shdr*)calloc(h->numSHEntries, sizeof(Elf32_Shdr));
+        fseek(f, header.e_shoff ,SEEK_SET);
+        if(fread(h->SHEntries, sizeof(Elf32_Shdr), h->numSHEntries, f)!=h->numSHEntries) {
+                FreeElfHeader(&h);
+                printf_log(LOG_INFO, "Cannot read all Section Header\n");
+                return NULL;
+        }
+
+        if(h->numPHEntries == PN_XNUM) {
+            printf_log(LOG_DEBUG, "Read number of Program Header in 1st Section\n");
+            // read 1st section header and grab actual number from here
+            h->numPHEntries = h->SHEntries[0].sh_info;
+        }
+    }
+
+    printf_log(LOG_DEBUG, "Read %d Program header\n", h->numPHEntries);
+    h->PHEntries = (Elf32_Phdr*)calloc(h->numPHEntries, sizeof(Elf32_Phdr));
+    fseek(f, header.e_phoff ,SEEK_SET);
+    if(fread(h->PHEntries, sizeof(Elf32_Phdr), h->numPHEntries, f)!=h->numPHEntries) {
+            FreeElfHeader(&h);
+            printf_log(LOG_INFO, "Cannot read all Program Header\n");
+            return NULL;
+    }
+
+    if(header.e_shentsize!=0) {
+        if(h->SHIdx == SHN_XINDEX) {
+            printf_log(LOG_DEBUG, "Read number of String Table in 1st Section\n");
+            h->SHIdx = h->SHEntries[0].sh_link;
+        }
+        if(h->SHIdx > h->numSHEntries) {
+            printf_log(LOG_INFO, "Incoherent Section String Table Index : %d / %d\n", h->SHIdx, h->numSHEntries);
+            FreeElfHeader(&h);
+            return NULL;
+        }
+        // load Section table
+        printf_log(LOG_DEBUG, "Loading Sections Table String (idx = %d)\n", h->SHIdx);
+        if(LoadSH(f, h->SHEntries+h->SHIdx, (void*)&h->SHStrTab, ".shstrtab", SHT_STRTAB)) {
+            FreeElfHeader(&h);
+            return NULL;
+        }
+        if(box86_log>=LOG_DUMP) DumpMainHeader(&header, h);
+
+        LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".strtab", "SymTab Strings", SHT_STRTAB, (void**)&h->StrTab, NULL);
+        LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".symtab", "SymTab", SHT_SYMTAB, (void**)&h->SymTab, &h->numSymTab);
+        if(box86_log>=LOG_DUMP && h->SymTab) DumpSymTab(h);
+
+        LoadNamedSection(f, h->SHEntries, h->numSHEntries, h->SHStrTab, ".dynamic", "Dynamic", SHT_DYNAMIC, (void**)&h->Dynamic, &h->numDynamic);
+        if(box86_log>=LOG_DUMP && h->Dynamic) DumpDynamicSections(h);
+        // grab DT_REL & DT_RELA stuffs
+        // also grab the DT_STRTAB string table
+        {
+            for (int i=0; i<h->numDynamic; ++i) {
+                if(h->Dynamic[i].d_tag == DT_REL)
+                    h->rel = h->Dynamic[i].d_un.d_ptr;
+                else if(h->Dynamic[i].d_tag == DT_RELSZ)
+                    h->relsz = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_RELENT)
+                    h->relent = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_RELA)
+                    h->rela = h->Dynamic[i].d_un.d_ptr;
+                else if(h->Dynamic[i].d_tag == DT_RELASZ)
+                    h->relasz = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_RELAENT)
+                    h->relaent = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_PLTREL)
+                    h->pltrel = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_PLTRELSZ)
+                    h->pltsz = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_JMPREL)
+                    h->jmprel = h->Dynamic[i].d_un.d_val;
+                else if(h->Dynamic[i].d_tag == DT_STRTAB)
+                    h->DynStrTab = (char*)(h->Dynamic[i].d_un.d_ptr);
+                else if(h->Dynamic[i].d_tag == DT_STRSZ)
+                    h->szDynStrTab = h->Dynamic[i].d_un.d_val;
+            }
+            if(h->rel) {
+                if(h->relent != sizeof(Elf32_Rel)) {
+                    printf_log(LOG_NONE, "Rel Table Entry size invalid (0x%x should be 0x%x)\n", h->relent, sizeof(Elf32_Rel));
+                    FreeElfHeader(&h);
+                    return NULL;
+                }
+                printf_log(LOG_DEBUG, "Rel Table @%p (0x%x/0x%x)\n", (void*)h->rel, h->relsz, h->relent);
+            }
+            if(h->rela) {
+                if(h->relaent != sizeof(Elf32_Rela)) {
+                    printf_log(LOG_NONE, "RelA Table Entry size invalid (0x%x should be 0x%x)\n", h->relaent, sizeof(Elf32_Rela));
+                    FreeElfHeader(&h);
+                    return NULL;
+                }
+                printf_log(LOG_DEBUG, "RelA Table @%p (0x%x/0x%x)\n", (void*)h->rela, h->relasz, h->relaent);
+            }
+            if(h->jmprel) {
+                if(h->pltrel == DT_REL) {
+                    h->pltent = sizeof(Elf32_Rel);
+                } else if(h->pltrel == DT_RELA) {
+                    h->pltent = sizeof(Elf32_Rela);
+                } else {
+                    printf_log(LOG_NONE, "PLT Table type is unknown (size = 0x%x, type=%d)\n", h->pltsz, h->pltrel);
+                    FreeElfHeader(&h);
+                    return NULL;
+                }
+                if((h->pltsz / h->pltent)*h->pltent != h->pltsz) {
+                    printf_log(LOG_NONE, "PLT Table Entry size invalid (0x%x, ent=0x%x, type=%d)\n", h->pltsz, h->pltent, h->pltrel);
+                    FreeElfHeader(&h);
+                    return NULL;
+                }
+                printf_log(LOG_DEBUG, "PLT Table @%p (type=%d 0x%x/0x%0x)\n", (void*)h->jmprel, h->pltrel, h->pltsz, h->pltent);
+            }
+            if(h->DynStrTab && h->szDynStrTab) {
+                //DumpDynamicNeeded(h); cannot dump now, it's not loaded yet
+            }
+        }
+        // look for PLT Offset
+        int ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got.plt");
+        if(ii) {
+            h->gotplt = h->SHEntries[ii].sh_addr;
+            printf_log(LOG_DEBUG, "The GOT.PLT Table is at address %p\n", (void*)h->gotplt);
+        }
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".got");
+        if(ii) {
+            h->got = h->SHEntries[ii].sh_addr;
+            printf_log(LOG_DEBUG, "The GOT Table is at address %p\n", (void*)h->got);
+        }
+        // look for .init entry point
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".init");
+        if(ii) {
+            h->initentry = h->SHEntries[ii].sh_addr;
+            printf_log(LOG_DEBUG, "The .init is at address %p\n", (void*)h->initentry);
+        }
+        // and .init_array
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".init_array");
+        if(ii) {
+            h->initarray_sz = h->SHEntries[ii].sh_size / sizeof(Elf32_Addr);
+            h->initarray = (uintptr_t)(h->SHEntries[ii].sh_addr);
+            printf_log(LOG_DEBUG, "The .init_array is at address %p, and have %d elements\n", (void*)h->initarray, h->initarray_sz);
+        }
+        // look for .fini entry point
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".fini");
+        if(ii) {
+            h->finientry = h->SHEntries[ii].sh_addr;
+            printf_log(LOG_DEBUG, "The .fini is at address %p\n", (void*)h->finientry);
+        }
+        // and .fini_array
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".fini_array");
+        if(ii) {
+            h->finiarray_sz = h->SHEntries[ii].sh_size / sizeof(Elf32_Addr);
+            h->finiarray = (uintptr_t)(h->SHEntries[ii].sh_addr);
+            printf_log(LOG_DEBUG, "The .fini_array is at address %p, and have %d elements\n", (void*)h->finiarray, h->finiarray_sz);
+        }
+        // grab .text for main code
+        ii = FindSection(h->SHEntries, h->numSHEntries, h->SHStrTab, ".text");
+        if(ii) {
+            h->text = (uintptr_t)(h->SHEntries[ii].sh_addr);
+            h->textsz = h->SHEntries[ii].sh_size;
+            printf_log(LOG_DEBUG, "The .text is at address %p, and is %d big\n", (void*)h->text, h->textsz);
+        }
+
+        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);
+        if(box86_log>=LOG_DUMP && h->DynSym) DumpDynSym(h);
+    }
+    
+    return h;
+}

src/emu/modrm.h

@@ -0,0 +1,88 @@
+// ModRM utilities macros
+#define getecommon(A, T) \
+    if(!(nextop&0xC0)) { \
+        if((nextop&7)==4) { \
+            uint8_t sib = F8; \
+            uintptr_t base = ((sib&0x7)==5)?(F32):(emu->regs[(sib&0x7)].dword[0]); \
+            base += (emu->sbiidx[(sib>>3)&7]->sdword[0] << (sib>>6)); \
+            A = (T*)base; \
+        } else if((nextop&7)==5) { \
+            A = (T*)(F32); \
+        } else { \
+            A = (T*)(emu->regs[nextop&7].dword[0]); \
+        } \
+    } else { \
+        uintptr_t base; \
+        if((nextop&7)==4) { \
+            uint8_t sib = F8;   \
+            base = emu->regs[(sib&0x7)].dword[0]; \
+            base += (emu->sbiidx[(sib>>3)&7]->sdword[0] << (sib>>6));   \
+        } else { \
+            base = emu->regs[(nextop&0x7)].dword[0];    \
+        } \
+        base+=(nextop&0x80)?(F32S):(F8S); \
+        A = (T*)base; \
+    }
+#define getecommono(A, T, O) \
+    if(!(nextop&0xC0)) { \
+        if((nextop&7)==4) { \
+            uint8_t sib = F8; \
+            uintptr_t base = ((sib&0x7)==5)?(F32):(emu->regs[(sib&0x7)].dword[0]); \
+            base += (emu->sbiidx[(sib>>3)&7]->sdword[0] << (sib>>6)); \
+            A = (T*)(base+O); \
+        } else if((nextop&7)==5) { \
+            A = (T*)(F32+O); \
+        } else { \
+            A = (T*)(emu->regs[nextop&7].dword[0]+O); \
+        } \
+    } else { \
+        uintptr_t base; \
+        if((nextop&7)==4) { \
+            uint8_t sib = F8;   \
+            base = emu->regs[(sib&0x7)].dword[0]; \
+            base += (emu->sbiidx[(sib>>3)&7]->sdword[0] << (sib>>6));   \
+        } else { \
+            base = emu->regs[(nextop&0x7)].dword[0];    \
+        } \
+        base+=(nextop&0x80)?(F32S):(F8S); \
+        A = (T*)(base+O); \
+    }
+#define geteb(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
+    } else getecommon(A, reg32_t)
+#define geted(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->regs[(nextop&7)]; \
+    } else getecommon(A, reg32_t)
+#define getedo(A, O) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->regs[(nextop&7)]; \
+    } else getecommono(A, reg32_t, O)
+#define getem(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->mmx[(nextop&7)]; \
+    } else getecommon(A, mmx_regs_t)
+#define getex(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->xmm[(nextop&7)]; \
+    } else getecommon(A, sse_regs_t)
+// Macros for ModR/M gets
+#define GET_EB      geteb(oped)
+#define GET_ED      geted(oped)
+#define GET_ED_OFFS(o) getedo(oped, o)
+#define GET_EM      getem(opem)
+#define GET_EX      getex(opex)
+#define EB          oped
+#define ED          oped
+#define EM          opem
+#define EX          opex
+#define GB          emu->regs[(nextop>>3)&3].byte[(nextop>>5)&0x1]
+#define GD          emu->regs[((nextop&0x38)>>3)]
+#define GM          emu->mmx[((nextop&0x38)>>3)]
+#define GX          emu->xmm[((nextop&0x38)>>3)]
+
+// Alias
+#define GET_EW      GET_ED
+#define EW          ED
+#define GW          GD

src/emu/run0f.h

@@ -1,6 +1,8 @@
     opcode = F8;
     goto *opcodes0f[opcode];
-
+        _0f_0x0B:                      /* UD2 */
+            kill(getpid(), SIGILL);              // this is undefined instruction
+            NEXT;
         _0f_0x10:                      /* MOVUPS Gx,Ex */
             nextop = F8;
             GET_EX;
@@ -49,6 +51,11 @@
             GET_EX;
             EX->q[0] = GX.q[1];
             NEXT;
+        _0f_0x18:                       /* PREFETCHh Ed */
+            nextop = F8;
+            GET_ED;
+            __builtin_prefetch((void*)ED->dword[0], 0, 0); // inoring wich level of cache (roughly (nextop>>3)&7)
+            NEXT;
 
         _0f_0x1F:                      /* NOP (multi-byte) */
             nextop = F8;
@@ -73,15 +80,41 @@
             GX.f[0] = EM->sd[0];
             GX.f[1] = EM->sd[1];
             NEXT;
-
+        _0f_0x2B:                      /* MOVNTPS Ex,Gx */
+            nextop = F8;
+            GET_EX;
+            EX->q[0] = GX.q[0];
+            EX->q[1] = GX.q[1];
+            NEXT;
         _0f_0x2C:                      /* CVTTPS2PI Gm, Ex */
-        _0f_0x2D:                      /* CVTPS2PI Gm, Ex */
-            // rounding should be done; and indefinite integer should also be assigned if overflow or NaN/Inf
             nextop = F8;
             GET_EX;
             GM.sd[1] = EX->f[1];
             GM.sd[0] = EX->f[0];
             NEXT;
+        _0f_0x2D:                      /* CVTPS2PI Gm, Ex */
+            // rounding should be done; and indefinite integer should also be assigned if overflow or NaN/Inf
+            nextop = F8;
+            GET_EX;
+            switch(emu->round) {
+                case ROUND_Nearest:
+                    GM.sd[1] = floorf(EX->f[1]+0.5f);
+                    GM.sd[0] = floorf(EX->f[0]+0.5f);
+                    break;
+                case ROUND_Down:
+                    GM.sd[1] = floorf(EX->f[1]);
+                    GM.sd[0] = floorf(EX->f[0]);
+                    break;
+                case ROUND_Up:
+                    GM.sd[1] = ceilf(EX->f[1]);
+                    GM.sd[0] = ceilf(EX->f[0]);
+                    break;
+                case ROUND_Chop:
+                    GM.sd[1] = EX->f[1];
+                    GM.sd[0] = EX->f[0];
+                    break;
+            }
+            NEXT;
         _0f_0x2E:                      /* UCOMISS Gx, Ex */
             // same for now
         _0f_0x2F:                      /* COMISS Gx, Ex */
@@ -97,6 +130,13 @@
             } else {
                 SET_FLAG(F_ZF); CLEAR_FLAG(F_PF); CLEAR_FLAG(F_CF);
             }
+            CLEAR_FLAG(F_OF); CLEAR_FLAG(F_AF); CLEAR_FLAG(F_SF);
+            NEXT;
+
+        _0f_0x31:                   /* RDTSC */
+            tmp64u = ReadTSC(emu);
+            R_EDX = tmp64u>>32;
+            R_EAX = tmp64u&0xFFFFFFFF;
             NEXT;
         
         #define GOCOND(BASE, PREFIX, CONDITIONAL) \
@@ -212,7 +252,18 @@
             for(int i=0; i<4; ++i)
                 GX.f[i] = sqrtf(EX->f[i]);
             NEXT;
-
+        _0f_0x52:                      /* RSQRTPS Gx, Ex */
+            nextop = F8;
+            GET_EX;
+            for(int i=0; i<4; ++i)
+                GX.f[i] = 1.0f/sqrtf(EX->f[i]);
+            NEXT;
+        _0f_0x53:                      /* RCPPS Gx, Ex */
+            nextop = F8;
+            GET_EX;
+            for(int i=0; i<4; ++i)
+                GX.f[i] = 1.0f/EX->f[i];
+            NEXT;
         _0f_0x54:                      /* ANDPS Gx, Ex */
             nextop = F8;
             GET_EX;
@@ -314,17 +365,206 @@
             GET_EM;
             GM.ud[1] = EM->ud[0];
             NEXT;
+        _0f_0x63:                      /* PACKSSWB Gm, Em */
+            nextop = F8;
+            GET_EM;
+            GM.sb[0] = (GM.sw[0] > 127) ? 127 : ((GM.sw[0] < -128) ? -128 : GM.sw[0]);
+            GM.sb[1] = (GM.sw[1] > 127) ? 127 : ((GM.sw[1] < -128) ? -128 : GM.sw[1]);
+            GM.sb[2] = (GM.sw[2] > 127) ? 127 : ((GM.sw[2] < -128) ? -128 : GM.sw[2]);
+            GM.sb[3] = (GM.sw[3] > 127) ? 127 : ((GM.sw[3] < -128) ? -128 : GM.sw[3]);
+            GM.sb[4] = (EM->sw[0] > 127) ? 127 : ((EM->sw[0] < -128) ? -128 : EM->sw[0]);
+            GM.sb[5] = (EM->sw[1] > 127) ? 127 : ((EM->sw[1] < -128) ? -128 : EM->sw[1]);
+            GM.sb[6] = (EM->sw[2] > 127) ? 127 : ((EM->sw[2] < -128) ? -128 : EM->sw[2]);
+            GM.sb[7] = (EM->sw[3] > 127) ? 127 : ((EM->sw[3] < -128) ? -128 : EM->sw[3]);
+            NEXT;
+        _0f_0x64:                       /* PCMPGTB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 8; i++) {
+                GM.ub[i] = (GM.sb[i] > EM->sb[i]) ? 0xFF : 0;
+            }
+            NEXT;
+        _0f_0x65:                       /* PCMPGTW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 4; i++) {
+                GM.uw[i] = (GM.sw[i] > EM->sw[i]) ? 0xFFFF : 0;
+            }
+            NEXT;
+        _0f_0x66:                       /* PCMPGTD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 2; i++) {
+                GM.ud[i] = (GM.sd[i] > EM->sd[i]) ? 0xFFFFFFFF : 0;
+            }
+            NEXT;
+        _0f_0x67:                       /* PACKUSWB Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i)
+                GM.ub[i] = (GM.sw[i]<0)?0:((GM.sw[i]>0xff)?0xff:GM.sw[i]);
+            for(int i=0; i<4; ++i)
+                GM.ub[4+i] = (EM->sw[i]<0)?0:((EM->sw[i]>0xff)?0xff:EM->sw[i]);
+            NEXT;
+        _0f_0x68:                       /* PUNPCKHBW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
+            for(int i=0; i<4; ++i)
+                GM.ub[2 * i] = GM.ub[i + 4];
+            for(int i=0; i<4; ++i)
+                GM.ub[2 * i + 1] = EM->ub[i + 4];
+            NEXT;
+        _0f_0x69:                       /* PUNPCKHWD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
+            for(int i=0; i<2; ++i)
+                GM.uw[2 * i] = GM.uw[i + 2];
+            for(int i=0; i<2; ++i)
+                GM.uw[2 * i + 1] = EM->uw[i + 2];
+            NEXT;
+        _0f_0x6A:                       /* PUNPCKHDQ Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
+            GM.ud[0] = GM.ud[1];
+            GM.ud[1] = EM->ud[1];
+            NEXT;
+        _0f_0x6B:                       /* PACKSSDW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
+            for(int i=0; i<2; ++i)
+                GM.sw[i] = (GM.sd[i]<-32768)?-32768:((GM.sd[i]>32767)?32767:GM.sd[i]);
+            for(int i=0; i<2; ++i)
+                GM.sw[2+i] = (EM->sd[i]<-32768)?-32768:((EM->sd[i]>32767)?32767:EM->sd[i]);
+            NEXT;
 
+        _0f_0x6E:                      /* MOVD Gm, Ed */
+            nextop = F8;
+            GET_ED;
+            GM.q = ED->dword[0];    // zero extended
+            NEXT;
         _0f_0x6F:                      /* MOVQ Gm, Em */
             nextop = F8;
             GET_EM;
             GM.q = EM->q;
             NEXT;
-
+        _0f_0x70:                       /* PSHUFW Gm, Em, Ib */
+            nextop = F8;
+            GET_EM;
+            tmp8u = F8;
+            if(EM==&GM) {eam1 = GM; EM = &eam1;}   // copy is needed
+            for(int i=0; i<4; ++i)
+                GM.uw[i] = EM->uw[(tmp8u>>(i*2))&3];
+            NEXT;
+        _0f_0x71:  /* GRP */
+            nextop = F8;
+            GET_EM;
+            switch((nextop>>3)&7) {
+                case 2:                 /* PSRLW Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>15)
+                        {EM->q = 0;}
+                    else
+                        for (int i=0; i<4; ++i) EM->uw[i] >>= tmp8u;
+                    break;
+                case 4:                 /* PSRAW Em, Ib */
+                    tmp8u = F8;
+                    for (int i=0; i<4; ++i) EM->sw[i] >>= tmp8u;
+                    break;
+                case 6:                 /* PSLLW Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>15)
+                        {EM->q = 0;}
+                    else
+                        for (int i=0; i<4; ++i) EM->uw[i] <<= tmp8u;
+                    break;
+                default:
+                    goto _default;
+            }
+            NEXT;
+        _0f_0x72:  /* GRP */
+            nextop = F8;
+            GET_EM;
+            switch((nextop>>3)&7) {
+                case 2:                 /* PSRLD Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>31)
+                        {EM->q = 0;}
+                    else
+                        for (int i=0; i<2; ++i) EM->ud[i] >>= tmp8u;
+                    break;
+                case 4:                 /* PSRAD Em, Ib */
+                    tmp8u = F8;
+                    for (int i=0; i<2; ++i) EM->sd[i] >>= tmp8u;
+                    break;
+                case 6:                 /* PSLLD Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>31)
+                        {EM->q = 0;}
+                    else
+                        for (int i=0; i<2; ++i) EM->ud[i] <<= tmp8u;
+                    break;
+                default:
+                    goto _default;
+            }
+            NEXT;
+        _0f_0x73:  /* GRP */
+            nextop = F8;
+            GET_EM;
+            switch((nextop>>3)&7) {
+                case 2:                 /* PSRLQ Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>63)
+                        {EM->q = 0;}
+                    else
+                        {EM->q >>= tmp8u;}
+                    break;
+                case 6:                 /* PSLLQ Em, Ib */
+                    tmp8u = F8;
+                    if(tmp8u>63)
+                        {EM->q = 0;}
+                    else
+                        {EM->q <<= tmp8u;}
+                    break;
+                default:
+                    goto _default;
+            }
+            NEXT;
+        _0f_0x74:                       /* PCMPEQB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 8; i++) {
+                GM.ub[i] = (GM.sb[i] == EM->sb[i]) ? 0xFF : 0;
+            }
+            NEXT;
+        _0f_0x75:                       /* PCMPEQW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 4; i++) {
+                GM.uw[i] = (GM.sw[i] == EM->sw[i]) ? 0xFFFF : 0;
+            }
+            NEXT;
+        _0f_0x76:                       /* PCMPEQD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for (int i = 0; i < 2; i++) {
+                GM.ud[i] = (GM.sd[i] == EM->sd[i]) ? 0xFFFFFFFF : 0;
+            }
+            NEXT;
         _0f_0x77:                      /* EMMS */
             // empty MMX, FPU now usable
+            emu->top = 0;
+            emu->fpu_stack = 0;
             NEXT;
 
+        _0f_0x7E:                       /* MOVD Ed, Gm */
+            nextop = F8;
+            GET_ED;
+            ED->dword[0] = GM.ud[0];
+            NEXT;
         _0f_0x7F:                      /* MOVQ Em, Gm */
             nextop = F8;
             GET_EM;
@@ -337,7 +577,7 @@
                 case 0x0:
                     // emulate a P4
                     R_EAX = 0x80000004;
-                    // return GuenuineIntel
+                    // return GenuineIntel
                     R_EBX = 0x756E6547;
                     R_EDX = 0x49656E69;
                     R_ECX = 0x6C65746E;
@@ -351,7 +591,7 @@
                             | 1<<15     // cmov
                             | 1<<19     // clflush (seems to be with SSE2)
                             | 1<<23     // mmx
-                            | 1<<24     // fxsr (fxsave, fxrestore)
+                            //| 1<<24     // fxsr (fxsave, fxrestore)
                             | 1<<25     // SSE
                             | 1<<26     // SSE2
                             ;
@@ -359,8 +599,65 @@
                             | 1<<13     // cx16 (cmpxchg16)
                             ;           // also 1<<0 is SSE3 and 1<<9 is SSSE3
                     break;
+                case 0x2:   // TLB and Cache info. Sending 1st gen P4 info...
+                    R_EAX = 0x665B5001;
+                    R_EBX = 0x00000000;
+                    R_ECX = 0x00000000;
+                    R_EDX = 0x007A7000;
+                    break;
+                
+                case 0x4:   // Cache info
+                    switch (R_ECX) {
+                        case 0: // L1 data cache
+                            R_EAX = (1 | (1<<5) | (1<<8));   //type
+                            R_EBX = (63 | (7<<22)); // size
+                            R_ECX = 63;
+                            R_EDX = 1;
+                            break;
+                        case 1: // L1 inst cache
+                            R_EAX = (2 | (1<<5) | (1<<8)); //type
+                            R_EBX = (63 | (7<<22)); // size
+                            R_ECX = 63;
+                            R_EDX = 1;
+                            break;
+                        case 2: // L2 cache
+                            R_EAX = (3 | (2<<5) | (1<<8)); //type
+                            R_EBX = (63 | (15<<22));    // size
+                            R_ECX = 4095;
+                            R_EDX = 1;
+                            break;
+
+                        default:
+                            R_EAX = 0x00000000;
+                            R_EBX = 0x00000000;
+                            R_ECX = 0x00000000;
+                            R_EDX = 0x00000000;
+                            break;
+                    }
+                    break;
+                case 0x5:   //mwait info
+                    R_EAX = 0;
+                    R_EBX = 0;
+                    R_ECX = 1 | 2;
+                    R_EDX = 0;
+                    break;
+                case 0x6:   // thermal
+                case 0x9:   // direct cache access
+                case 0xA:   // Architecture performance monitor
+                    R_EAX = 0;
+                    R_EBX = 0;
+                    R_ECX = 0;
+                    R_EDX = 0;
+                    break;
+                case 0x7:   // extended bits...
+                    /*if(R_ECX==0)    R_EAX = 0;
+                    else*/ R_EAX = R_ECX = R_EBX = R_EDX = 0;
+                    break;
+
+                case 0x80000000:        // max extended
+                    break;              // no extended, so return same value as beeing the max value!
                 default:
-                    printf_log(LOG_INFO, "Warning, CPUID command %X unsupported\n", tmp32u);
+                    printf_log(LOG_INFO, "Warning, CPUID command %X unsupported (ECX=%08x)\n", tmp32u, R_ECX);
                     R_EAX = 0;
             }
             NEXT;
@@ -425,6 +722,20 @@
             }
             GET_ED;
             switch((nextop>>3)&7) {
+                case 0:                 /* FXSAVE m512byte */
+                    // should save flags & all
+                    // copy MMX regs...
+                    memcpy(ED->dword+32/4, &emu->mmx[0], sizeof(emu->mmx));
+                    // copy SSE regs
+                    memcpy(ED->dword+160/4, &emu->xmm[0], sizeof(emu->xmm));
+                    break;
+                case 1:                 /* FXRSTOR m512byte */
+                    // should restore flags & all
+                    // copy MMX regs...
+                    memcpy(&emu->mmx[0], ED->dword+32/4, sizeof(emu->mmx));
+                    // copy SSE regs
+                    memcpy(&emu->xmm[0], ED->dword+160/4, sizeof(emu->xmm));
+                    break;
                 case 2:                 /* LDMXCSR Md */
                     emu->mxcsr = ED->dword[0];
                     break;
@@ -440,7 +751,6 @@
             GET_ED;
             GD.dword[0] = imul32(emu, GD.dword[0], ED->dword[0]);
             NEXT;
-
         _0f_0xB0:                      /* CMPXCHG Eb,Gb */
             CHECK_FLAGS(emu);
             nextop = F8;
@@ -491,23 +801,6 @@
             GD.dword[0] = EW->word[0];
             NEXT;
 
-        _0f_0xBB:                      /* BTC Ed,Gd */
-            CHECK_FLAGS(emu);
-            nextop = F8;
-            GET_ED;
-            tmp8u = GD.byte[0];
-            if((nextop&0xC0)!=0xC0)
-            {
-                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
-            }
-            tmp8u&=31;
-            if(ED->dword[0] & (1<<tmp8u))
-                SET_FLAG(F_CF);
-            else
-                CLEAR_FLAG(F_CF);
-            ED->dword[0] ^= (1<<tmp8u);
-            NEXT;
-
         _0f_0xBA:                      
             nextop = F8;
             switch((nextop>>3)&7) {
@@ -524,11 +817,43 @@
                         SET_FLAG(F_CF);
                     else
                         CLEAR_FLAG(F_CF);
-                    break;;
+                    break;
+                case 6:             /* BTR Ed, Ib */
+                    CHECK_FLAGS(emu);
+                    GET_ED;
+                    tmp8u = F8;
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                    }
+                    tmp8u&=31;
+                    if(ED->dword[0] & (1<<tmp8u)) {
+                        SET_FLAG(F_CF);
+                        ED->dword[0] ^= (1<<tmp8u);
+                    } else
+                        CLEAR_FLAG(F_CF);
+                    break;
+
                 default:
                     goto _default;
             }
             NEXT;
+        _0f_0xBB:                      /* BTC Ed,Gd */
+            CHECK_FLAGS(emu);
+            nextop = F8;
+            GET_ED;
+            tmp8u = GD.byte[0];
+            if((nextop&0xC0)!=0xC0)
+            {
+                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+            }
+            tmp8u&=31;
+            if(ED->dword[0] & (1<<tmp8u))
+                SET_FLAG(F_CF);
+            else
+                CLEAR_FLAG(F_CF);
+            ED->dword[0] ^= (1<<tmp8u);
+            NEXT;
         _0f_0xBC:                      /* BSF Ed,Gd */
             CHECK_FLAGS(emu);
             nextop = F8;
@@ -602,6 +927,18 @@
             }
             NEXT;
 
+        _0f_0xC4:                       /* PINSRW Gm,Ew,Ib */
+            nextop = F8;
+            GET_ED;
+            tmp8u = F8;
+            GM.uw[tmp8u&3] = ED->word[0];   // only low 16bits
+            NEXT;
+        _0f_0xC5:                       /* PEXTRW Gw,Em,Ib */
+            nextop = F8;
+            GET_EM;
+            tmp8u = F8;
+            GD.dword[0] = EM->uw[tmp8u&3];  // 16bits extract, 0 extended
+            NEXT;
         _0f_0xC6:                      /* SHUFPS Gx, Ex, Ib */
             nextop = F8;
             GET_EX;
@@ -643,3 +980,276 @@
             emu->regs[tmp8s].dword[0] = __builtin_bswap32(emu->regs[tmp8s].dword[0]);
             NEXT;
 
+        _0f_0xD1:                   /* PSRLW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>15)
+                GM.q=0; 
+            else {
+                tmp8u = EM->ub[0];
+                for(int i=0; i<4; ++i)
+                    GM.uw[i] >>= tmp8u;
+            }
+            NEXT;
+        _0f_0xD2:                   /* PSRLD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>31)
+                GM.q=0;
+            else {
+                tmp8u = EM->ub[0];
+                for(int i=0; i<2; ++i)
+                    GM.ud[i] >>= tmp8u;
+            }
+            NEXT;
+        _0f_0xD3:                   /* PSRLQ Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.q = (EM->q > 63) ? 0 : (GM.q >> EM->q);
+            NEXT;
+        _0f_0xD4:                   /* PADDQ Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.sq += EM->sq;
+            NEXT;
+        _0f_0xD5:                   /* PMULLW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32s = (int32_t)GM.sw[i] * EM->sw[i];
+                GM.sw[i] = tmp32s;
+            }
+            NEXT;
+
+        _0f_0xD8:                   /* PSUBUSB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i) {
+                tmp32s = (int32_t)GM.ub[i] - EM->ub[i];
+                GM.ub[i] = (tmp32s < 0) ? 0 : tmp32s;
+            }
+            NEXT;
+        _0f_0xD9:                   /* PSUBUSW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32s = (int32_t)GM.uw[i] - EM->uw[i];
+                GM.uw[i] = (tmp32s < 0) ? 0 : tmp32s;
+            }
+            NEXT;
+
+        _0f_0xDB:                   /* PAND Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.q &= EM->q;
+            NEXT;
+        _0f_0xDC:                   /* PADDUSB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i) {
+                tmp32u = (uint32_t)GM.ub[i] + EM->ub[i];
+                GM.ub[i] = (tmp32u>255) ? 255 : tmp32u;
+            }
+            NEXT;
+        _0f_0xDD:                   /* PADDUSW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32u = (uint32_t)GM.uw[i] + EM->uw[i];
+                GM.uw[i] = (tmp32u>65535) ? 65535 : tmp32u;
+            }
+            NEXT;
+
+        _0f_0xDF:                   /* PANDN Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.q = (~GM.q) & EM->q;
+            NEXT;
+        _0f_0xE0:                   /* PAVGB Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i)
+                    GM.ub[i] = ((uint32_t)GM.ub[i]+EM->ub[i]+1)>>1;
+            NEXT;
+        _0f_0xE1:                   /* PSRAW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>15)
+                tmp8u = 16;
+            else
+                tmp8u = EM->ub[0];
+            for(int i=0; i<4; ++i)
+                GM.sw[i] >>= tmp8u;
+            NEXT;
+        _0f_0xE2:                   /* PSRAD Gm, Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>31) {
+                for(int i=0; i<2; ++i)
+                    GM.sd[i] = (GM.sd[i]<0)?-1:0;
+            } else {
+                tmp8u = EM->ub[0];
+                for(int i=0; i<2; ++i)
+                    GM.sd[i] >>= tmp8u;
+            }
+            NEXT;
+        _0f_0xE3:                   /* PSRAQ Gm, Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>63)
+                tmp8u = 64;
+            else
+                tmp8u = EM->ub[0];
+            GM.sq >>= tmp8u;
+            NEXT;
+        _0f_0xE4:                   /* PMULHUW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32u = (int32_t)GM.uw[i] * EM->uw[i];
+                GM.uw[i] = (tmp32u>>16)&0xffff;
+            }
+            NEXT;
+        _0f_0xE5:                   /* PMULHW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32s = (int32_t)GM.sw[i] * EM->sw[i];
+                GM.uw[i] = (tmp32s>>16)&0xffff;
+            }
+            NEXT;
+
+        _0f_0xE8:                   /* PSUBSB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i) {
+                tmp32s = (int32_t)GM.sb[i] - EM->sb[i];
+                GM.sb[i] = (tmp32s>127)?127:((tmp32s<-128)?-128:tmp32s);
+            }
+            NEXT;
+        _0f_0xE9:                   /* PSUBSW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32s = (int32_t)GM.sw[i] - EM->sw[i];
+                GM.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
+            }
+            NEXT;
+
+        _0f_0xEB:                   /* POR Gm, Em */
+            nextop = F8;
+            GET_EM;
+            GM.q |= EM->q;
+            NEXT;
+        _0f_0xEC:                   /* PADDSB Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i) {
+                tmp32s = (int32_t)GM.sb[i] + EM->sb[i];
+                GM.sb[i] = (tmp32s>127)?127:((tmp32s<-128)?-128:tmp32s);
+            }
+            NEXT;
+        _0f_0xED:                   /* PADDSW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i) {
+                tmp32s = (int32_t)GM.sw[i] + EM->sw[i];
+                GM.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
+            }
+            NEXT;
+
+        _0f_0xEF:                   /* PXOR Gm, Em */
+            nextop = F8;
+            GET_EM;
+            GM.q ^= EM->q;
+            NEXT;
+
+        _0f_0xF1:                   /* PSLLW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>15)
+                GM.q = 0;
+            else {
+                tmp8u = EM->ub[0];
+                for(int i=0; i<4; ++i)
+                    GM.sw[i] <<= tmp8u;
+            }
+            NEXT;
+        _0f_0xF2:                   /* PSLLD Gm, Em */
+            nextop = F8;
+            GET_EM;
+            if(EM->q>31)
+                GM.q = 0;
+            else {
+                tmp8u = EM->ub[0];
+                for(int i=0; i<2; ++i)
+                    GM.sd[i] <<= tmp8u;
+            }
+            NEXT;
+        _0f_0xF3:                   /* PSLLQ Gm, Em */
+            nextop = F8;
+            GET_EM;
+            GM.q = (EM->q > 63) ? 0 : (GM.q << EM->ub[0]);
+            NEXT;
+        _0f_0xF4:                   /* PMULUDQ Gm,Em */
+            nextop = F8;
+            GET_EM;
+            GM.q = (uint64_t)GM.ud[0] * (uint64_t)EM->ud[0];
+            NEXT;
+        _0f_0xF5:                   /* PMADDWD Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for (int i=0; i<2; ++i) {
+                int offset = i * 2;
+
+                tmp32s = (int32_t)GM.sw[offset + 0] * EM->sw[offset + 0];
+                tmp32s2 = (int32_t)GM.sw[offset + 1] * EM->sw[offset + 1];
+                GM.sd[i] = tmp32s + tmp32s2;
+            }
+            NEXT;
+        _0f_0xF6:                   /* PSADBW Gm, Em */
+            nextop = F8;
+            GET_EM;
+            tmp32u = 0;
+            for (int i=0; i<8; ++i)
+                tmp32u += (GM.ub[i]>EM->ub[i])?(GM.ub[i] - EM->ub[i]):(EM->ub[i] - GM.ub[i]);
+            GM.q = tmp32u;
+            NEXT;
+
+        _0f_0xF8:                   /* PSUBB Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i)
+                GM.sb[i] -= EM->sb[i];
+            NEXT;
+        _0f_0xF9:                   /* PSUBW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i)
+                GM.sw[i] -= EM->sw[i];
+            NEXT;
+        _0f_0xFA:                   /* PSUBD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<2; ++i)
+                GM.sd[i] -= EM->sd[i];
+            NEXT;
+
+        _0f_0xFC:                   /* PADDB Gm, Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<8; ++i)
+                GM.sb[i] += EM->sb[i];
+            NEXT;
+        _0f_0xFD:                   /* PADDW Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<4; ++i)
+                GM.sw[i] += EM->sw[i];
+            NEXT;
+        _0f_0xFE:                   /* PADDD Gm,Em */
+            nextop = F8;
+            GET_EM;
+            for(int i=0; i<2; ++i)
+                GM.sd[i] += EM->sd[i];
+            NEXT;

src/emu/run66.h

@@ -25,8 +25,19 @@
     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....
+        NEXT;
+
     _66_0x0F:                      /* 66 0f prefix */
         #include "run660f.h"
+        
+    _66_0x26:                      /* ES: */
+        // ignored
+        NEXT;
     _66_0x2E:                      /* CS: */
         // ignored
         NEXT;
@@ -66,17 +77,46 @@
         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_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;
@@ -131,11 +171,21 @@
         GET_EW;
         GW.word[0] = EW->word[0];
         NEXT;
+    _66_0x8C:                              /* MOV Ew,Seg */
+        nextop = F8;
+        GET_EW;
+        EW->word[0] = emu->segs[(nextop&38)>>3];
+        NEXT;
     
+    _66_0x8E:                               /* MOV Seg,Ew */
+        nextop = F8;
+        GET_EW;
+        emu->segs[((nextop&0x38)>>3)] = EW->word[0];
+        NEXT;
     _66_0x8F:                              /* POP Ew */
         nextop = F8;
         GET_EW;
-        EW->dword[0] = Pop16(emu);
+        EW->word[0] = Pop16(emu);
         NEXT;
     _66_0x90:                              /* NOP */
         NEXT;
@@ -150,7 +200,13 @@
         emu->regs[_AX].sword[0] = emu->regs[_AX].sbyte[0];
         NEXT;
     _66_0x99:                              /* CWD */
-        R_DX=(R_AX & 0x8000)?0xFFFF:0x0000;
+        R_DX=((R_AX & 0x8000)?0xFFFF:0x0000);
+        NEXT;
+
+    _66_0x9C:                              /* PUSHFW */
+        CHECK_FLAGS(emu);
+        PackFlags(emu);
+        Push16(emu, (uint16_t)emu->packed_eflags.x32);
         NEXT;
 
     _66_0xA1:                              /* MOV AX,Ow */
@@ -174,6 +230,11 @@
         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;
@@ -272,6 +333,8 @@
                 }
                 break;
             case 0xA7:              /* REP(N)Z CMPSW */
+                tmp16u = 0;
+                tmp16u2 = 0;
                 while(tmp32u) {
                     --tmp32u;
                     tmp16u  = *(uint16_t*)R_EDI;
@@ -298,6 +361,7 @@
                 }
                 break;
             case 0xAF:              /* REP(N)Z SCASW */
+                tmp16u = 0;
                 while(tmp32u) {
                     --tmp32u;
                     tmp16u = *(uint16_t*)R_EDI;
@@ -361,6 +425,3 @@
                 goto fini;
         }
         NEXT;
-                
-
-

src/emu/run660f.h

@@ -101,7 +101,16 @@
         GET_EX;
         memcpy(EX, &GX, 16); // unaligned...
         NEXT;
-
+    _6f_0x12:                      /* MOVLPD Gx, Eq */
+        nextop = F8;
+        GET_ED;
+        GX.q[0] = *(uint64_t*)ED;
+        NEXT;
+    _6f_0x13:                      /* MOVLPD Eq, Gx */
+        nextop = F8;
+        GET_ED;
+        *(uint64_t*)ED = GX.q[0];
+        NEXT;
     _6f_0x14:                      /* UNPCKLPD Gx, Ex */
         nextop = F8;
         GET_EX;
@@ -150,12 +159,33 @@
 
 
     _6f_0x2C:                      /* CVTTPD2PI Gm, Ex */
-    _6f_0x2D:                      /* CVTPD2PI Gm, Ex */
         nextop = F8;
         GET_EX;
         GM.sd[0] = EX->d[0];
         GM.sd[1] = EX->d[1];
         NEXT;
+    _6f_0x2D:                      /* CVTPD2PI Gm, Ex */
+        nextop = F8;
+        GET_EX;
+        switch(emu->round) {
+            case ROUND_Nearest:
+                GM.sd[0] = floor(EX->d[0]+0.5);
+                GM.sd[1] = floor(EX->d[1]+0.5);
+                break;
+            case ROUND_Down:
+                GM.sd[0] = floor(EX->d[0]);
+                GM.sd[1] = floor(EX->d[1]);
+                break;
+            case ROUND_Up:
+                GM.sd[0] = ceil(EX->d[0]);
+                GM.sd[1] = ceil(EX->d[1]);
+                break;
+            case ROUND_Chop:
+                GM.sd[0] = EX->d[0];
+                GM.sd[1] = EX->d[1];
+                break;
+        }
+        NEXT;
     _6f_0x2E:                      /* UCOMISD Gx, Ex */
         // no special check...
     _6f_0x2F:                      /* COMISD Gx, Ex */
@@ -171,6 +201,60 @@
         } else {
             SET_FLAG(F_ZF); CLEAR_FLAG(F_PF); CLEAR_FLAG(F_CF);
         }
+        CLEAR_FLAG(F_OF); CLEAR_FLAG(F_AF); CLEAR_FLAG(F_SF);
+        NEXT;
+
+    _6f_0x38:  // these are some SSE3 opcodes
+        opcode = F8;
+        switch(opcode) {
+            case 0x00:  // PSHUFB
+                nextop = F8;
+                GET_EX;
+                eax1 = GX;
+                for (int i=0; i<15; ++i) {
+                    if(EX->ub[i]&128)
+                        GX.ub[i] = 0;
+                    else
+                        GX.ub[i] = eax1.ub[EX->ub[i]&15];
+                }
+                break;
+            default:
+                goto _default;
+        }
+        NEXT;
+
+    _6f_0x3A:  // these are some SSE3 opcodes
+        opcode = F8;
+        switch(opcode) {
+            case 0x0F:          // PALIGNR
+                nextop = F8;
+                GET_EX;
+                tmp8u = F8;
+                if(tmp8u>31)
+                    {EX->q[0] = EX->q[1] = 0;}
+                else if(tmp8u>15) {
+                    tmp8u=(tmp8u-16)*8;
+                    if (tmp8u < 64) {
+                        GX.q[0] = (EX->q[0] >> tmp8u) | (EX->q[1] << (64 - tmp8u));
+                        GX.q[1] = (EX->q[1] >> tmp8u);
+                    } else {
+                        GX.q[0] = EX->q[1] >> (tmp8u - 64);
+                        GX.q[1] = 0;
+                    }                    
+                } else {
+                    tmp8u*=8;
+                    if (tmp8u < 64) {
+                        GX.q[0] = (GX.q[0] >> tmp8u) | (GX.q[1] << (64 - tmp8u));
+                        GX.q[1] = (GX.q[1] >> tmp8u) | (EX->q[0] << (64-tmp8u));
+                    } else {
+                        GX.q[0] = GX.q[1] >> (tmp8u - 64) | (EX->q[0] << (tmp8u));
+                        GX.q[1] = (EX->q[0] << (64-tmp8u)) | (EX->q[1] >> (tmp8u - 64));
+                    }                    
+                }
+                break;
+            default:
+                goto _default;
+        }
         NEXT;
 
     _6f_0x50:                      /* MOVMSKPD Gd, Ex */
@@ -290,36 +374,44 @@
         GX.ud[2] = GX.ud[1];
         GX.ud[1] = EX->ud[0];
         NEXT;
-
+    _6f_0x63:  /* PACKSSWB Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i)
+            GX.sb[i] = (GX.sw[i]<-128)?-128:((GX.sw[i]>127)?127:GX.sw[i]);
+        for(int i=0; i<8; ++i)
+            GX.sb[8+i] = (EX->sw[i]<-128)?-128:((EX->sw[i]>127)?127:EX->sw[i]);
+        NEXT;
     _6f_0x64:  /* PCMPGTB Gx,Ex */
         nextop = F8;
         GET_EX;
         for(int i=0; i<16; ++i)
-            GX.sb[i] = (GX.sb[i]>EX->sb[i])?0xFF:0x00;
+            GX.ub[i] = (GX.sb[i]>EX->sb[i])?0xFF:0x00;
         NEXT;
     _6f_0x65:  /* PCMPGTW Gx,Ex */
         nextop = F8;
         GET_EX;
         for(int i=0; i<8; ++i)
-            GX.sw[i] = (GX.sw[i]>EX->sw[i])?0xFFFF:0x0000;
+            GX.uw[i] = (GX.sw[i]>EX->sw[i])?0xFFFF:0x0000;
         NEXT;
     _6f_0x66:  /* PCMPGTD Gx,Ex */
         nextop = F8;
         GET_EX;
         for(int i=0; i<4; ++i)
-            GX.sd[i] = (GX.sd[i]>EX->sd[i])?0xFFFFFFFF:0x00000000;
+            GX.ud[i] = (GX.sd[i]>EX->sd[i])?0xFFFFFFFF:0x00000000;
         NEXT;
     _6f_0x67:  /* PACKUSWB */
         nextop = F8;
         GET_EX;
-        for(int i=0; i<4; ++i)
+        for(int i=0; i<8; ++i)
             GX.ub[i] = (GX.sw[i]<0)?0:((GX.sw[i]>0xff)?0xff:GX.sw[i]);
-        for(int i=0; i<4; ++i)
-            GX.ub[4+i] = (EX->sw[i]<0)?0:((EX->sw[i]>0xff)?0xff:EX->sw[i]);
+        for(int i=0; i<8; ++i)
+            GX.ub[8+i] = (EX->sw[i]<0)?0:((EX->sw[i]>0xff)?0xff:EX->sw[i]);
         NEXT;
     _6f_0x68:  /* PUNPCKHBW Gx,Ex */
         nextop = F8;
         GET_EX;
+        if(EX==&GX) {eax1 = GX; EX = &eax1;}   // copy is needed
         for(int i=0; i<8; ++i)
             GX.ub[2 * i] = GX.ub[i + 8];
         for(int i=0; i<8; ++i)
@@ -328,6 +420,7 @@
     _6f_0x69:  /* PUNPCKHWD Gx,Ex */
         nextop = F8;
         GET_EX;
+        if(EX==&GX) {eax1 = GX; EX = &eax1;}   // copy is needed
         for(int i=0; i<4; ++i)
             GX.uw[2 * i] = GX.uw[i + 4];
         for(int i=0; i<4; ++i)
@@ -336,6 +429,7 @@
     _6f_0x6A:  /* PUNPCKHDQ Gx,Ex */
         nextop = F8;
         GET_EX;
+        if(EX==&GX) {eax1 = GX; EX = &eax1;}   // copy is needed
         GX.ud[0] = GX.ud[2];
         GX.ud[2] = GX.ud[3];
         GX.ud[1] = EX->ud[2];
@@ -344,6 +438,7 @@
     _6f_0x6B:  /* PACKSSDW Gx,Ex */
         nextop = F8;
         GET_EX;
+        if(EX==&GX) {eax1 = GX; EX = &eax1;}   // copy is needed
         for(int i=0; i<4; ++i)
             GX.sw[i] = (GX.sd[i]<-32768)?-32768:((GX.sd[i]>32767)?32767:GX.sd[i]);
         for(int i=0; i<4; ++i)
@@ -376,31 +471,26 @@
         nextop = F8;
         GET_EX;
         tmp8u = F8;
-        if(&GX!=EX)
-            for (int i=0; i<4; ++i)
-                GX.ud[i] = EX->ud[(tmp8u>>(i*2))&3];
-        else {
-            for (int i=0; i<4; ++i)
-                eax1.ud[i] = EX->ud[(tmp8u>>(i*2))&3];
-            memcpy(&GX, &eax1, sizeof(eax1));
-        }
+        if(EX==&GX) {eax1 = GX; EX = &eax1;}   // copy is needed
+        for (int i=0; i<4; ++i)
+            GX.ud[i] = EX->ud[(tmp8u>>(i*2))&3];
         NEXT;
     _6f_0x71:  /* GRP */
         nextop = F8;
         GET_EX;
         switch((nextop>>3)&7) {
-            case 2:                 /* PSRLW Gx, Ib */
+            case 2:                 /* PSRLW Ex, Ib */
                 tmp8u = F8;
                 if(tmp8u>15)
                     {EX->q[0] = EX->q[1] = 0;}
                 else
                     for (int i=0; i<8; ++i) EX->uw[i] >>= tmp8u;
                 break;
-            case 4:                 /* PSRAW Gx, Ib */
+            case 4:                 /* PSRAW Ex, Ib */
                 tmp8u = F8;
                 for (int i=0; i<8; ++i) EX->sw[i] >>= tmp8u;
                 break;
-            case 6:                 /* PSLLW Gx, Ib */
+            case 6:                 /* PSLLW Ex, Ib */
                 tmp8u = F8;
                 if(tmp8u>15)
                     {EX->q[0] = EX->q[1] = 0;}
@@ -453,10 +543,14 @@
                 if(tmp8u>15)
                     {EX->q[0] = EX->q[1] = 0;}
                 else {
-                    for (int i=tmp8u; i<16; ++i)
-                        EX->ub[i-tmp8u] = EX->ub[i];
-                    for (int i=16-tmp8u; i<16; ++i)
-                        EX->ub[i] = 0;
+                    tmp8u*=8;
+                    if (tmp8u < 64) {
+                        EX->q[0] = (EX->q[0] >> tmp8u) | (EX->q[1] << (64 - tmp8u));
+                        EX->q[1] = (EX->q[1] >> tmp8u);
+                    } else {
+                        EX->q[0] = EX->q[1] >> (tmp8u - 64);
+                        EX->q[1] = 0;
+                    }                    
                 }
                 break;
             case 6:                 /* PSLLQ Ex, Ib */
@@ -471,10 +565,14 @@
                 if(tmp8u>15)
                     {EX->q[0] = EX->q[1] = 0;}
                 else {
-                    for (int i=16-tmp8u; i>=0; --i)
-                        EX->ub[i+tmp8u] = EX->ub[i];
-                    for (int i=0; i<tmp8u; ++i)
-                        EX->ub[i] = 0;
+                    tmp8u*=8;
+                    if (tmp8u < 64) {
+                        EX->q[1] = (EX->q[1] << tmp8u) | (EX->q[0] >> (64 - tmp8u));
+                        EX->q[0] = (EX->q[0] << tmp8u);
+                    } else {
+                        EX->q[1] = EX->q[0] << (tmp8u - 64);
+                        EX->q[0] = 0;
+                    }
                 }
                 break;
             default:
@@ -655,15 +753,15 @@
 
     _6f_0xC4:  /* PINSRW Gx,Ew,Ib */
         nextop = F8;
-        GET_EW;
+        GET_ED;
         tmp8u = F8;
-        GX.uw[tmp8u&7] = EW->word[0];
+        GX.uw[tmp8u&7] = ED->word[0];   // only low 16bits
         NEXT;
     _6f_0xC5:  /* PEXTRW Gw,Ex,Ib */
         nextop = F8;
         GET_EX;
         tmp8u = F8;
-        GW.word[0] = EX->uw[tmp8u&7];
+        GD.dword[0] = EX->uw[tmp8u&7];  // 16bits extract, 0 extended
         NEXT;
     _6f_0xC6:  /* SHUFPD Gx, Ex, Ib */
         nextop = F8;
@@ -681,7 +779,7 @@
         if(EX->q[0]>15)
             {GX.q[0] = GX.q[1] = 0;}
         else 
-            {tmp8u=EX->q[0]; for (int i=0; i<8; ++i) GX.uw[i] >>= tmp8u;}
+            {tmp8u=EX->ub[0]; for (int i=0; i<8; ++i) GX.uw[i] >>= tmp8u;}
         NEXT;
     _6f_0xD2:  /* PSRLD Gx, Ex */
         nextop = F8;
@@ -689,7 +787,7 @@
         if(EX->q[0]>31)
             {GX.q[0] = GX.q[1] = 0;}
         else 
-            {tmp8u=EX->q[0]; for (int i=0; i<4; ++i) GX.ud[i] >>= tmp8u;}
+            {tmp8u=EX->ub[0]; for (int i=0; i<4; ++i) GX.ud[i] >>= tmp8u;}
         NEXT;
     _6f_0xD3:  /* PSRLQ Gx, Ex */
         nextop = F8;
@@ -697,15 +795,22 @@
         if(EX->q[0]>63)
             {GX.q[0] = GX.q[1] = 0;}
         else 
-            {tmp8u=EX->q[0]; for (int i=0; i<2; ++i) GX.q[i] >>= tmp8u;}
+            {tmp8u=EX->ub[0]; for (int i=0; i<2; ++i) GX.q[i] >>= tmp8u;}
         NEXT;
     _6f_0xD4:  /* PADDQ Gx,Ex */
         nextop = F8;
         GET_EX;
-        GX.q[0] += EX->q[0];
-        GX.q[1] += EX->q[1];
+        GX.sq[0] += EX->sq[0];
+        GX.sq[1] += EX->sq[1];
+        NEXT;
+    _6f_0xD5:  /* PMULLW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i) {
+            tmp32s = (int32_t)GX.sw[i] * EX->sw[i];
+            GX.sw[i] = tmp32s;
+        }
         NEXT;
-
     _6f_0xD6:  /* MOVQ Ex,Gx */
         nextop = F8;
         GET_EX;
@@ -713,6 +818,30 @@
         if((nextop&0xC0)==0xC0)
             EX->q[1] = 0;
         NEXT;
+    _6f_0xD7:  /* PMOVMSKB Gd,Ex */
+        nextop = F8;
+        GET_EX;
+        GD.dword[0] = 0;
+        for (int i=0; i<16; ++i)
+            if(EX->ub[i]&0x80)
+                GD.dword[0] |= (1<<i);
+        NEXT;
+    _6f_0xD8:  /* PSUBUSB Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<16; ++i) {
+            tmp16s = (int16_t)GX.ub[i] - EX->ub[i];
+            GX.ub[i] = (tmp16s<0)?0:((tmp16s>255)?255:tmp16s);
+        }
+        NEXT;
+    _6f_0xD9:  /* PSUBUSW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i) {
+            tmp32s = (int32_t)GX.uw[i] - EX->uw[i];
+            GX.uw[i] = (tmp32s<0)?0:((tmp32s>0xffff)?0xffff:tmp32s);
+        }
+        NEXT;
 
     _6f_0xDB:  /* PAND Gx,Ex */
         nextop = F8;
@@ -720,7 +849,21 @@
         GX.q[0] &= EX->q[0];
         GX.q[1] &= EX->q[1];
         NEXT;
+    _6f_0xDC:  /* PADDUSB Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<16; ++i) {
+            tmp16s = (int16_t)GX.ub[i] + EX->ub[i];
+            GX.ub[i] = (tmp16s>255)?255:tmp16s;
+        }
+        NEXT;
 
+    _6f_0xDE:  /* PMAXUB Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        for (int i=0; i<16; ++i)
+            if(EX->ub[i]>GX.ub[i]) GX.ub[i] = EX->ub[i];
+        NEXT;
     _6f_0xDF:  /* PANDN Gx,Ex */
         nextop = F8;
         GET_EX;
@@ -731,15 +874,24 @@
     _6f_0xE1:  /* PSRAW Gx, Ex */
         nextop = F8;
         GET_EX;
-        tmp8u=EX->q[0];
+        tmp8u=(EX->q[0]>15)?16:EX->ub[0];
         for (int i=0; i<8; ++i) GX.sw[i] >>= tmp8u;
         NEXT;
     _6f_0xE2:  /* PSRAD Gx, Ex */
         nextop = F8;
         GET_EX;
-        tmp8u=EX->q[0]; for (int i=0; i<4; ++i) GX.sd[i] >>= tmp8u;
+        tmp8u=(EX->q[0]>31)?32:EX->ub[0];
+        for (int i=0; i<4; ++i) GX.sd[i] >>= tmp8u;
         NEXT;
 
+    _6f_0xE5:  /* PMULHW Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i) {
+            tmp32s = (int32_t)GX.sw[i] * EX->sw[i];
+            GX.uw[i] = (tmp32s>>16)&0xffff;
+        }
+        NEXT;
     _6f_0xE6:  /* CVTTPD2DQ Gx, Ex */
         nextop = F8;
         GET_EX;
@@ -747,7 +899,34 @@
         GX.sd[1] = EX->d[1];
         GX.q[1] = 0;
         NEXT;
-
+    _6f_0xE7:                      /* MOVNTDQ Ex, Gx */
+        nextop = F8;
+        GET_EX;
+        EX->q[0] = GX.q[0];
+        EX->q[1] = GX.q[1];
+        NEXT;
+    _6f_0xE8:  /* PSUBSB Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<16; ++i) {
+            tmp16s = (int16_t)GX.sb[i] - EX->sb[i];
+            GX.ub[i] = (tmp16s<-128)?-128:((tmp16s>127)?127:tmp16s);
+        }
+        NEXT;
+    _6f_0xE9:  /* PSUBSW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i) {
+            tmp32s = (int32_t)GX.sw[i] - EX->sw[i];
+            GX.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
+        }
+        NEXT;
+    _6f_0xEA:  /* PMINSW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for (int i=0; i<8; ++i)
+            GX.sw[i] = (GX.sw[i]<EX->sw[i])?GX.sw[i]:EX->sw[i];
+        NEXT;
     _6f_0xEB:  /* POR Gx,Ex */
         nextop = F8;
         GET_EX;
@@ -760,7 +939,7 @@
         nextop = F8;
         GET_EX;
         for(int i=0; i<16; ++i) {
-            tmp16s = GX.sb[i] + EX->sb[i];
+            tmp16s = (int16_t)GX.sb[i] + EX->sb[i];
             GX.sb[i] = (tmp16s>127)?127:((tmp16s<-128)?-128:tmp16s);
         }
         NEXT;
@@ -768,7 +947,7 @@
         nextop = F8;
         GET_EX;
         for(int i=0; i<8; ++i) {
-            tmp32s = GX.sw[i] + EX->sw[i];
+            tmp32s = (int32_t)GX.sw[i] + EX->sw[i];
             GX.sw[i] = (tmp32s>32767)?32767:((tmp32s<-32768)?-32768:tmp32s);
         }
         NEXT;
@@ -793,7 +972,7 @@
         if(EX->q[0]>15)
             {GX.q[0] = GX.q[1] = 0;}
         else 
-            {tmp8u=EX->q[0]; for (int i=0; i<8; ++i) GX.uw[i] <<= tmp8u;}
+            {tmp8u=EX->ub[0]; for (int i=0; i<8; ++i) GX.uw[i] <<= tmp8u;}
         NEXT;
     _6f_0xF2:  /* PSLLD Gx, Ex */
         nextop = F8;
@@ -801,7 +980,7 @@
         if(EX->q[0]>31)
             {GX.q[0] = GX.q[1] = 0;}
         else 
-            {tmp8u=EX->q[0]; for (int i=0; i<4; ++i) GX.ud[i] <<= tmp8u;}
+            {tmp8u=EX->ub[0]; for (int i=0; i<4; ++i) GX.ud[i] <<= tmp8u;}
         NEXT;
     _6f_0xF3:  /* PSLLQ Gx, Ex */
         nextop = F8;
@@ -817,35 +996,73 @@
         GX.q[1] = (uint64_t)EX->ud[2]*GX.ud[2];
         GX.q[0] = (uint64_t)EX->ud[0]*GX.ud[0];
         NEXT;
+    _6f_0xF5:  /* PMADDWD Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for (int i=0; i<4; ++i)
+            GX.sd[i] = (int32_t)(GX.sw[i*2+0])*EX->sw[i*2+0] + (int32_t)(GX.sw[i*2+1])*EX->sw[i*2+1];
+        NEXT;
+    _6f_0xF6:  /* PSADBW Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        tmp32u = 0;
+        for (int i=0; i<8; ++i)
+            tmp32u += (GX.ub[i]>EX->ub[i])?(GX.ub[i] - EX->ub[i]):(EX->ub[i] - GX.ub[i]);
+        GX.q[0] = tmp32u;
+        tmp32u = 0;
+        for (int i=8; i<16; ++i)
+            tmp32u += (GX.ub[i]>EX->ub[i])?(GX.ub[i] - EX->ub[i]):(EX->ub[i] - GX.ub[i]);
+        GX.q[1] = tmp32u;
+        NEXT;
+    _6f_0xF7:  /* MASKMOVDQU Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        opx2 = (sse_regs_t *)(R_EDI);
+        for (int i=0; i<16; ++i) {
+            if(EX->ub[i]&0x80)
+                opx2->ub[i] = GX.ub[i];
+        }
+        NEXT;
 
+    _6f_0xF9:  /* PSUBW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i)
+            GX.uw[i] -= EX->uw[i];
+        NEXT;
     _6f_0xFA:  /* PSUBD Gx,Ex */
         nextop = F8;
         GET_EX;
-        GX.ud[0] -= EX->ud[0];
-        GX.ud[1] -= EX->ud[1];
-        GX.ud[2] -= EX->ud[2];
-        GX.ud[3] -= EX->ud[3];
+        GX.sd[0] -= EX->sd[0];
+        GX.sd[1] -= EX->sd[1];
+        GX.sd[2] -= EX->sd[2];
+        GX.sd[3] -= EX->sd[3];
         NEXT;
     _6f_0xFB:  /* PSUBQ Gx,Ex */
         nextop = F8;
         GET_EX;
-        GX.q[0] -= EX->q[0];
-        GX.q[1] -= EX->q[1];
+        GX.sq[0] -= EX->sq[0];
+        GX.sq[1] -= EX->sq[1];
         NEXT;
     _6f_0xFC:  /* PADDB Gx,Ex */
         nextop = F8;
         GET_EX;
         for(int i=0; i<16; ++i)
-            GX.ub[i] += EX->ub[i];
+            GX.sb[i] += EX->sb[i];
+        NEXT;
+    _6f_0xFD:  /* PADDW Gx,Ex */
+        nextop = F8;
+        GET_EX;
+        for(int i=0; i<8; ++i)
+            GX.sw[i] += EX->sw[i];
         NEXT;
-
     _6f_0xFE:  /* PADDD Gx,Ex */
         nextop = F8;
         GET_EX;
-        GX.ud[0] += EX->ud[0];
-        GX.ud[1] += EX->ud[1];
-        GX.ud[2] += EX->ud[2];
-        GX.ud[3] += EX->ud[3];
+        GX.sd[0] += EX->sd[0];
+        GX.sd[1] += EX->sd[1];
+        GX.sd[2] += EX->sd[2];
+        GX.sd[3] += EX->sd[3];
         NEXT;
 
 

src/emu/rund8.h

@@ -1,3 +1,5 @@
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wstrict-aliasing"
     nextop = F8;
     switch(nextop) {
         case 0xC0:
@@ -8,7 +10,11 @@
         case 0xC5:
         case 0xC6:
         case 0xC7:  /* FADD */
+            #ifdef USE_FLOAT
+            ST0.f += ST(nextop&7).f;
+            #else
             ST0.d += ST(nextop&7).d;
+            #endif
             break;
         case 0xC8:
         case 0xC9:
@@ -18,7 +24,11 @@
         case 0xCD:
         case 0xCE:
         case 0xCF:  /* FMUL */
+            #ifdef USE_FLOAT
+            ST0.f *= ST(nextop&7).f;
+            #else
             ST0.d *= ST(nextop&7).d;
+            #endif
             break;
         case 0xD0:
         case 0xD1:
@@ -28,7 +38,11 @@
         case 0xD5:
         case 0xD6:
         case 0xD7:  /* FCOM */
+            #ifdef USE_FLOAT
+            fpu_fcom(emu, ST(nextop&7).f);
+            #else
             fpu_fcom(emu, ST(nextop&7).d);
+            #endif
             break;
         case 0xD8:
         case 0xD9:
@@ -38,7 +52,11 @@
         case 0xDD:
         case 0xDE:
         case 0xDF:  /* FCOMP */
+            #ifdef USE_FLOAT
+            fpu_fcom(emu, ST(nextop&7).f);
+            #else
             fpu_fcom(emu, ST(nextop&7).d);
+            #endif
             fpu_do_pop(emu);
             break;
         case 0xE0:
@@ -49,7 +67,11 @@
         case 0xE5:
         case 0xE6:
         case 0xE7:  /* FSUB */
+            #ifdef USE_FLOAT
+            ST0.f -= ST(nextop&7).f;
+            #else
             ST0.d -= ST(nextop&7).d;
+            #endif
             break;
         case 0xE8:
         case 0xE9:
@@ -59,7 +81,11 @@
         case 0xED:
         case 0xEE:
         case 0xEF:  /* FSUBR */
+            #ifdef USE_FLOAT
+            ST0.f = ST(nextop&7).f - ST0.f;
+            #else
             ST0.d = ST(nextop&7).d - ST0.d;
+            #endif
             break;
         case 0xF0:
         case 0xF1:
@@ -69,7 +95,11 @@
         case 0xF5:
         case 0xF6:
         case 0xF7:  /* FDIV */
+            #ifdef USE_FLOAT
+            ST0.f /= ST(nextop&7).f;
+            #else
             ST0.d /= ST(nextop&7).d;
+            #endif
             break;
         case 0xF8:
         case 0xF9:
@@ -79,26 +109,46 @@
         case 0xFD:
         case 0xFE:
         case 0xFF:  /* FDIVR */
+            #ifdef USE_FLOAT
+            ST0.f = ST(nextop&7).f / ST0.f;
+            #else
             ST0.d = ST(nextop&7).d / ST0.d;
+            #endif
             break;
         default:
         switch((nextop>>3)&7) {
             case 0:         /* FADD ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f += *(float*)ED;
+                    #else
                     ST0.d += *(float*)ED;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f += f;
+                    #else
                     ST0.d += f;
+                    #endif
                 }
                 break;
             case 1:         /* FMUL ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f *= *(float*)ED;
+                    #else
                     ST0.d *= *(float*)ED;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f *= f;
+                    #else
                     ST0.d *= f;
+                    #endif
                 }
                 break;
             case 2:      /* FCOM ST0, float */
@@ -123,40 +173,73 @@
             case 4:         /* FSUB ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f -= *(float*)ED;
+                    #else
                     ST0.d -= *(float*)ED;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f -= f;
+                    #else
                     ST0.d -= f;
+                    #endif
                 }
                 break;
             case 5:         /* FSUBR ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f = *(float*)ED - ST0.f;
+                    #else
                     ST0.d = *(float*)ED - ST0.d;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f = f - ST0.f;
+                    #else
                     ST0.d = f - ST0.d;
+                    #endif
                 }
                 break;
             case 6:         /* FDIV ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f /= *(float*)ED;
+                    #else
                     ST0.d /= *(float*)ED;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f /= f;
+                    #else
                     ST0.d /= f;
+                    #endif
                 }
                 break;
             case 7:         /* FDIVR ST0, float */
                 GET_ED;
                 if(!(((uintptr_t)ED)&3))
+                    #ifdef USE_FLOAT
+                    ST0.f = *(float*)ED / ST0.f;
+                    #else
                     ST0.d = *(float*)ED / ST0.d;
+                    #endif
                 else {
                     *(uint32_t*)&f = ED->dword[0];
+                    #ifdef USE_FLOAT
+                    ST0.f = f / ST0.f;
+                    #else
                     ST0.d = f / ST0.d;
+                    #endif
                 }
                 break;
             default:
                 goto _default;
         }
     }
+#pragma GCC diagnostic pop

src/emu/rund9.h

@@ -1,3 +1,5 @@
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wstrict-aliasing"
     nextop = F8;
     switch (nextop) {
         case 0xC0:
@@ -29,80 +31,198 @@
             break;
 
         case 0xE0:  /* FCHS */
+            #ifdef USE_FLOAT
+            ST0.f = -ST0.f;
+            #else
             ST0.d = -ST0.d;
+            #endif
             break;
         case 0xE1:  /* FABS */
+            #ifdef USE_FLOAT
+            ST0.f = fabsf(ST0.f);
+            #else
             ST0.d = fabs(ST0.d);
+            #endif
             break;
         
+        case 0xE4:  /* FTST */
+            fpu_ftst(emu);
+            break;
         case 0xE5:  /* FXAM */
             fpu_fxam(emu);
             break;
 
         case 0xE8:  /* FLD1 */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = 1.0f;
+            #else
             ST0.d = 1.0;
+            #endif
             break;
         case 0xE9:  /* FLDL2T */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = L2T;
+            #else
             ST0.d = L2T;
+            #endif
             break;
         case 0xEA:  /* FLDL2E */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = L2E;
+            #else
             ST0.d = L2E;
+            #endif
             break;
         case 0xEB:  /* FLDPI */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = PI;
+            #else
             ST0.d = PI;
+            #endif
             break;
         case 0xEC:  /* FLDLG2 */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = LG2;
+            #else
             ST0.d = LG2;
+            #endif
             break;
         case 0xED:  /* FLDLN2 */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = LN2;
+            #else
             ST0.d = LN2;
+            #endif
             break;
         case 0xEE:  /* FLDZ */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = 0.0f;
+            #else
             ST0.d = 0.0;
+            #endif
+            break;
+
+        case 0xF0:  /* F2XM1 */
+            #ifdef USE_FLOAT
+            ST0.f = exp2f(ST0.f) - 1.0f;
+            #else
+            ST0.d = exp2(ST0.d) - 1.0;
+            #endif
+            break;
+        case 0xF1:  /* FYL2X */
+            #ifdef USE_FLOAT
+            ST(1).f = log2f(ST0.f)*ST(1).f;
+            #else
+            ST(1).d = log2(ST0.d)*ST(1).d;
+            #endif
+            fpu_do_pop(emu);
             break;
-        
         case 0xF2:  /* FTAN */
+            #ifdef USE_FLOAT
+            ST0.f = tanf(ST0.f);
+            fpu_do_push(emu);
+            ST0.f = 1.0f;
+            #else
             ST0.d = tan(ST0.d);
             fpu_do_push(emu);
             ST0.d = 1.0;
+            #endif
             break;
         case 0xF3:  /* FPATAN */
+            #ifdef USE_FLOAT
+            ST1.f = atan2f(ST1.f, ST0.f);
+            #else
             ST1.d = atan2(ST1.d, ST0.d);
+            #endif
             fpu_do_pop(emu);
             break;
+        case 0xF4:  /* FXTRACT */
+            #ifdef USE_FLOAT
+            tmp32s = (ST0.ll&0x7f700000)>>23;
+            tmp32s -= 127;
+            ST0.f /= exp2f(tmp32s);
+            fpu_do_push(emu);
+            ST0.f = tmp32s;
+            #else
+            tmp32s = (ST0.ll&0x7ff0000000000000LL)>>52;
+            tmp32s -= 1023;
+            ST0.d /= exp2(tmp32s);
+            fpu_do_push(emu);
+            ST0.d = tmp32s;
+            #endif
+            break;
 
         case 0xF8:  /* FPREM */
+            #ifdef USE_FLOAT
+            tmp32s = ST0.f / ST1.f;
+            ST0.f -= ST1.f * tmp32s;
+            #else
             tmp32s = ST0.d / ST1.d;
             ST0.d -= ST1.d * tmp32s;
+            #endif
             emu->sw.f.F87_C2 = 0;
             emu->sw.f.F87_C0 = (tmp32s&1);
             emu->sw.f.F87_C3 = ((tmp32s>>1)&1);
             emu->sw.f.F87_C1 = ((tmp32s>>2)&1);
             break;
-
+        case 0xF9:  /* FYL2XP1 */
+            #ifdef USE_FLOAT
+            ST(1).f = log2f(ST0.f + 1.0f)*ST(1).f;
+            #else
+            ST(1).d = log2(ST0.d + 1.0)*ST(1).d;
+            #endif
+            fpu_do_pop(emu);
+            break;
         case 0xFA:  /* FSQRT */
+            #ifdef USE_FLOAT
+            ST0.f = sqrtf(ST0.f);
+            #else
             ST0.d = sqrt(ST0.d);
+            #endif
             break;
         case 0xFB:  /* FSINCOS */
             fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            sincosf(ST1.f, &ST1.f, &ST0.f);
+            #else
             sincos(ST1.d, &ST1.d, &ST0.d);
+            #endif
             break;
         case 0xFC:  /* FRNDINT */
+            #ifdef USE_FLOAT
+            ST0.f = fpu_round(emu, ST0.f);
+            #else
             ST0.d = fpu_round(emu, ST0.d);
+            #endif
+            break;
+        case 0xFD:  /* FSCALE */
+            // this could probably be done by just altering the exponant part of the float...
+            #ifdef USE_FLOAT
+            ST0.f *= exp2f(truncf(ST1.f));
+            #else
+            ST0.d *= exp2(trunc(ST1.d));
+            #endif
             break;
-
         case 0xFE:  /* FSIN */
+            #ifdef USE_FLOAT
+            ST0.f = sinf(ST0.f);
+            #else
             ST0.d = sin(ST0.d);
+            #endif
             break;
         case 0xFF:  /* FCOS */
+            #ifdef USE_FLOAT
+            ST0.f = cosf(ST0.f);
+            #else
             ST0.d = cos(ST0.d);
+            #endif
             break;
 
 
@@ -121,33 +241,54 @@
         case 0xDF:
         case 0xE2:
         case 0xE3:
-        case 0xE4:
         case 0xE6:
         case 0xE7:
         case 0xEF:
-        case 0xF0:
-        case 0xF1:
-        case 0xF4:
         case 0xF5:
         case 0xF6:
         case 0xF7:
-        case 0xF9:
-        case 0xFD:
             goto _default;
         default:
         switch((nextop>>3)&7) {
             case 0:     /* FLD ST0, Ed float */
                 GET_ED;
                 fpu_do_push(emu);
-                ST0.d = *(float*)ED;
+                #ifdef USE_FLOAT
+                ST0.ll = ED->dword[0];
+                #else
+                if(!(((uintptr_t)ED)&3))
+                    ST0.d = *(float*)ED;
+                else {
+                    memcpy(&f, ED->dword, sizeof(float));
+                    ST0.d = f;
+                }
+                #endif
                 break;
             case 2:     /* FST Ed, ST0 */
                 GET_ED;
-                *(float*)ED = ST0.d;
+                #ifdef USE_FLOAT
+                *(float*)ED = ST0.f;
+                #else
+                if(!(((uintptr_t)ED)&3))
+                    *(float*)ED = ST0.d;
+                else {
+                    f = ST0.d;
+                    memcpy(ED->dword, &f, sizeof(float));
+                }
+                #endif
                 break;
             case 3:     /* FSTP Ed, ST0 */
                 GET_ED;
-                *(float*)ED = ST0.d;
+                #ifdef USE_FLOAT
+                ED->dword[0] = ST0.ll;
+                #else
+                if(!(((uintptr_t)ED)&3))
+                    *(float*)ED = ST0.d;
+                else {
+                    f = ST0.d;
+                    memcpy(ED->dword, &f, sizeof(float));
+                }
+                #endif
                 fpu_do_pop(emu);
                 break;
             case 4:     /* FLDENV m */
@@ -180,3 +321,4 @@
                 goto _default;
         }
     }
+#pragma GCC diagnostic pop

src/emu/runda.h

@@ -51,7 +51,11 @@
         break;
     
     case 0xE9:      /* FUCOMPP */
+        #ifdef USE_FLOAT
+        fpu_fcom(emu, ST1.f);
+        #else
         fpu_fcom(emu, ST1.d);   // bad, should handle QNaN and IA interrupt
+        #endif
         fpu_do_pop(emu);
         fpu_do_pop(emu);
         break;
@@ -72,25 +76,60 @@
         switch((nextop>>3)&7) {
             case 0:     /* FIADD ST0, Ed int */
                 GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f += ED->sdword[0];
+                #else
                 ST0.d += ED->sdword[0];
+                #endif
                 break;
             case 1:     /* FIMUL ST0, Ed int */
                 GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f *= ED->sdword[0];
+                #else
                 ST0.d *= ED->sdword[0];
+                #endif
+                break;
+            case 2:     /* FICOM ST0, Ed int */
+                GET_ED;
+                fpu_fcom(emu, ED->sdword[0]);
+                break;
+            case 3:     /* FICOMP ST0, Ed int */
+                GET_ED;
+                fpu_fcom(emu, ED->sdword[0]);
+                fpu_do_pop(emu);
                 break;
             case 4:     /* FISUB ST0, Ed int */
                 GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f -= ED->sdword[0];
+                #else
                 ST0.d -= ED->sdword[0];
+                #endif
                 break;
             case 5:     /* FISUBR ST0, Ed int */
                 GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f = ED->sdword[0] - ST0.f;
+                #else
                 ST0.d = ED->sdword[0] - ST0.d;
+                #endif
+                break;
+            case 6:     /* FIDIV ST0, Ed int */
+                GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f /= ED->sdword[0];
+                #else
+                ST0.d /= ED->sdword[0];
+                #endif
                 break;
             case 7:     /* FIDIVR ST0, Ed int */
                 GET_ED;
+                #ifdef USE_FLOAT
+                ST0.f = ED->sdword[0] / ST0.f;
+                #else
                 ST0.d = ED->sdword[0] / ST0.d;
+                #endif
                 break;
-            default:
-                goto _default;
         }
     }

src/emu/rundb.h

@@ -0,0 +1,241 @@
+    nextop = F8;
+    switch(nextop) {
+    case 0xC0:      /* FCMOVNB ST(0), ST(i) */
+    case 0xC1:
+    case 0xC2:
+    case 0xC3:
+    case 0xC4:
+    case 0xC5:
+    case 0xC6:
+    case 0xC7:
+        CHECK_FLAGS(emu);
+        if(!ACCESS_FLAG(F_CF))
+            ST0.ll = ST(nextop&7).ll;
+        break;
+    case 0xC8:      /* FCMOVNE ST(0), ST(i) */
+    case 0xC9:
+    case 0xCA:
+    case 0xCB:
+    case 0xCC:
+    case 0xCD:
+    case 0xCE:
+    case 0xCF:
+        CHECK_FLAGS(emu);
+        if(!ACCESS_FLAG(F_ZF))
+            ST0.ll = ST(nextop&7).ll;
+        break;
+    case 0xD0:      /* FCMOVNBE ST(0), ST(i) */
+    case 0xD1:
+    case 0xD2:
+    case 0xD3:
+    case 0xD4:
+    case 0xD5:
+    case 0xD6:
+    case 0xD7:
+        CHECK_FLAGS(emu);
+        if(!(ACCESS_FLAG(F_CF) || ACCESS_FLAG(F_ZF)))
+            ST0.ll = ST(nextop&7).ll;
+        break;
+    case 0xD8:      /* FCMOVNU ST(0), ST(i) */
+    case 0xD9:
+    case 0xDA:
+    case 0xDB:
+    case 0xDC:
+    case 0xDD:
+    case 0xDE:
+    case 0xDF:
+        CHECK_FLAGS(emu);
+        if(!ACCESS_FLAG(F_PF))
+            ST0.ll = ST(nextop&7).ll;
+        break;
+
+    case 0xE2:      /* FNCLEX */
+        //Clears the floating-point exception flags (PE, UE, OE, ZE, DE, and IE), 
+        // the exception summary status flag (ES), the stack fault flag (SF), and the busy flag (B) in the FPU status word
+        emu->sw.f.F87_PE = 0;
+        emu->sw.f.F87_UE = 0;
+        emu->sw.f.F87_OE = 0;
+        emu->sw.f.F87_ZE = 0;
+        emu->sw.f.F87_DE = 0;
+        emu->sw.f.F87_IE = 0;
+        emu->sw.f.F87_ES = 0;
+        emu->sw.f.F87_SF = 0;
+        emu->sw.f.F87_B = 0;
+        break;
+    case 0xE3:      /* FNINIT */
+        reset_fpu(emu);
+        break;
+    case 0xE8:  /* FUCOMI ST0, STx */
+    case 0xE9:
+    case 0xEA:
+    case 0xEB:
+    case 0xEC:
+    case 0xED:
+    case 0xEE:
+    case 0xEF:
+        #ifdef USE_FLOAT
+        fpu_fcomi(emu, ST(nextop&7).f);
+        #else
+        fpu_fcomi(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
+        #endif
+        break;
+
+    case 0xF0:  /* FCOMI ST0, STx */
+    case 0xF1:
+    case 0xF2:
+    case 0xF3:
+    case 0xF4:
+    case 0xF5:
+    case 0xF6:
+    case 0xF7:
+        #ifdef USE_FLOAT
+        fpu_fcomi(emu, ST(nextop&7).f);
+        #else
+        fpu_fcomi(emu, ST(nextop&7).d);
+        #endif
+        break;
+    case 0xE0:
+    case 0xE1:
+    case 0xE4:
+    case 0xE5:
+    case 0xE6:
+    case 0xE7:
+        goto _default;
+    default:
+        switch((nextop>>3)&7) {
+            case 0: /* FILD ST0, Gd */
+                GET_ED;
+                fpu_do_push(emu);
+                #ifdef USE_FLOAT
+                ST0.f = ED->sdword[0];
+                #else
+                ST0.d = ED->sdword[0];
+                #endif
+                break;
+            case 1: /* FISTTP Ed, ST0 */
+                GET_ED;
+                #ifdef USE_FLOAT
+                tmp32s = ST0.f; // TODO: Handling of FPU Exception
+                if(tmp32s==0x7fffffff && isgreater(ST0.f, (float)(int32_t)0x7fffffff))
+                    tmp32s = 0x80000000;
+                #else
+                tmp32s = ST0.d; // TODO: Handling of FPU Exception
+                if(tmp32s==0x7fffffff && isgreater(ST0.d, (double)(int32_t)0x7fffffff))
+                    tmp32s = 0x80000000;
+                #endif
+                fpu_do_pop(emu);
+                ED->sdword[0] = tmp32s;
+                break;
+            case 2: /* FIST Ed, ST0 */
+                GET_ED;
+                #ifdef USE_FLOAT
+                if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
+                    ED->sdword[0] = 0x80000000;
+                else {
+                    switch(emu->round) {
+                        case ROUND_Nearest:
+                            ED->sdword[0] = floorf(ST0.f+0.5);
+                            break;
+                        case ROUND_Down:
+                            ED->sdword[0] = floorf(ST0.f);
+                            break;
+                        case ROUND_Up:
+                            ED->sdword[0] = ceilf(ST0.f);
+                            break;
+                        case ROUND_Chop:
+                            ED->sdword[0] = ST0.f;
+                            break;
+                    }
+                }
+                #else
+                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
+                    ED->sdword[0] = 0x80000000;
+                else {
+                    switch(emu->round) {
+                        case ROUND_Nearest:
+                            ED->sdword[0] = floor(ST0.d+0.5);
+                            break;
+                        case ROUND_Down:
+                            ED->sdword[0] = floor(ST0.d);
+                            break;
+                        case ROUND_Up:
+                            ED->sdword[0] = ceil(ST0.d);
+                            break;
+                        case ROUND_Chop:
+                            ED->sdword[0] = ST0.d;
+                            break;
+                    }
+                }
+                #endif
+                break;
+            case 3: /* FISTP Ed, ST0 */
+                GET_ED;
+                #ifdef USE_FLOAT
+                if(isgreater(ST0.f, (float)(int32_t)0x7fffffff) || isless(ST0.f, -(float)(int32_t)0x7fffffff))
+                    ED->sdword[0] = 0x80000000;
+                else {
+                    switch(emu->round) {
+                        case ROUND_Nearest:
+                            ED->sdword[0] = floorf(ST0.f+0.5);
+                            break;
+                        case ROUND_Down:
+                            ED->sdword[0] = floorf(ST0.f);
+                            break;
+                        case ROUND_Up:
+                            ED->sdword[0] = ceilf(ST0.f);
+                            break;
+                        case ROUND_Chop:
+                            ED->sdword[0] = ST0.f;
+                            break;
+                    }
+                }
+                #else
+                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff))
+                    ED->sdword[0] = 0x80000000;
+                else {
+                    switch(emu->round) {
+                        case ROUND_Nearest:
+                            ED->sdword[0] = floor(ST0.d+0.5);
+                            break;
+                        case ROUND_Down:
+                            ED->sdword[0] = floor(ST0.d);
+                            break;
+                        case ROUND_Up:
+                            ED->sdword[0] = ceil(ST0.d);
+                            break;
+                        case ROUND_Chop:
+                            ED->sdword[0] = ST0.d;
+                            break;
+                    }
+                }
+                #endif
+                fpu_do_pop(emu);
+                break;
+            case 5: /* FLD ST0, Gt */
+                GET_ED;
+                fpu_do_push(emu);
+                memcpy(&STld(0).ld, ED, 10);
+                #ifdef USE_FLOAT
+                LD2D(&STld(0), &d);
+                ST0.f = d;
+                #else
+                LD2D(&STld(0), &ST(0).d);
+                #endif
+                STld(0).ref = ST0.ll;
+                break;
+            case 7: /* FSTP tbyte */
+                GET_ED;
+                if(ST0.ll!=STld(0).ref)
+                    #ifdef USE_FLOAT
+                    {d = ST0.f; D2LD(&d, ED);}
+                    #else
+                    D2LD(&ST0.d, ED);
+                    #endif
+                else
+                    memcpy(ED, &STld(0).ld, 10);
+                fpu_do_pop(emu);
+                break;
+            default:
+                goto _default;
+        }
+    }

src/emu/rundc.h

@@ -1,3 +1,5 @@
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wstrict-aliasing"
     nextop = F8;
     switch(nextop) {
         case 0xC0:
@@ -8,7 +10,11 @@
         case 0xC5:
         case 0xC6:
         case 0xC7:  /* FADD */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f += ST0.f;
+            #else
             ST(nextop&7).d += ST0.d;
+            #endif
             break;
         case 0xC8:
         case 0xC9:
@@ -18,7 +24,11 @@
         case 0xCD:
         case 0xCE:
         case 0xCF:  /* FMUL */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f *= ST0.f;
+            #else
             ST(nextop&7).d *= ST0.d;
+            #endif
             break;
         case 0xD0:
         case 0xD1:
@@ -28,7 +38,11 @@
         case 0xD5:
         case 0xD6:
         case 0xD7:  /* FCOM */
-            fpu_fcom(emu, ST(nextop&7).d);  // TODO: is this ok?
+            #ifdef USE_FLOAT
+            fpu_fcom(emu, ST(nextop&7).f);
+            #else
+            fpu_fcom(emu, ST(nextop&7).d);
+            #endif
             break;
         case 0xD8:
         case 0xD9:
@@ -38,7 +52,11 @@
         case 0xDD:
         case 0xDE:
         case 0xDF:  /* FCOMP */
+            #ifdef USE_FLOAT
+            fpu_fcom(emu, ST(nextop&7).f);
+            #else
             fpu_fcom(emu, ST(nextop&7).d);  // TODO: is this ok?
+            #endif
             fpu_do_pop(emu);
             break;
         case 0xE0:
@@ -49,7 +67,11 @@
         case 0xE5:
         case 0xE6:
         case 0xE7:  /* FSUBR */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f = ST0.f -ST(nextop&7).f;
+            #else
             ST(nextop&7).d = ST0.d -ST(nextop&7).d;
+            #endif
             break;
         case 0xE8:
         case 0xE9:
@@ -59,7 +81,11 @@
         case 0xED:
         case 0xEE:
         case 0xEF:  /* FSUB */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f -= ST0.f;
+            #else
             ST(nextop&7).d -= ST0.d;
+            #endif
             break;
         case 0xF0:
         case 0xF1:
@@ -69,7 +95,11 @@
         case 0xF5:
         case 0xF6:
         case 0xF7:  /* FDIVR */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f = ST0.f / ST(nextop&7).f;
+            #else
             ST(nextop&7).d = ST0.d / ST(nextop&7).d;
+            #endif
             break;
         case 0xF8:
         case 0xF9:
@@ -79,26 +109,48 @@
         case 0xFD:
         case 0xFE:
         case 0xFF:  /* FDIV */
+            #ifdef USE_FLOAT
+            ST(nextop&7).f /=  ST0.f;
+            #else
             ST(nextop&7).d /=  ST0.d;
+            #endif
             break;
         default:
             GET_ED;
             switch((nextop>>3)&7) {
             case 0:         /* FADD ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f += *(double*)ED;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f += d;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d += *(double*)ED;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d += d;
                 }
+                #endif
                 break;
             case 1:         /* FMUL ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f *= *(double*)ED;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f *= d;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d *= *(double*)ED;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d *= d;
                 }
+                #endif
                 break;
             case 2:      /* FCOM ST0, double */
                 if(!(((uintptr_t)ED)&7))
@@ -118,38 +170,75 @@
                 fpu_do_pop(emu);
                 break;
             case 4:         /* FSUB ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f -= *(double*)ED;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f -= d;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d -= *(double*)ED;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d -= d;
                 }
+                #endif
                 break;
             case 5:         /* FSUBR ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f = *(double*)ED - ST0.f;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f = d - ST0.f;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d = *(double*)ED - ST0.d;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d = d - ST0.d;
                 }
+                #endif
                 break;
             case 6:         /* FDIV ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f /= *(double*)ED;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f /= d;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d /= *(double*)ED;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d /= d;
                 }
+                #endif
                 break;
             case 7:         /* FDIVR ST0, double */
+                #ifdef USE_FLOAT
+                if(!(((uintptr_t)ED)&7))
+                    ST0.f = *(double*)ED / ST0.f;
+                else {
+                    *(uint64_t*)&d = *(uint64_t*)ED;
+                    ST0.f = d / ST0.f;
+                }
+                #else
                 if(!(((uintptr_t)ED)&7))
                     ST0.d = *(double*)ED / ST0.d;
                 else {
                     *(uint64_t*)&d = *(uint64_t*)ED;
                     ST0.d = d / ST0.d;
                 }
+                #endif
                 break;
             default:
                 goto _default;
         }
     }
+#pragma GCC diagnostic pop

src/emu/rundd.h

@@ -40,7 +40,11 @@
     case 0xE5:
     case 0xE6:
     case 0xE7:
+        #ifdef USE_FLOAT
+        fpu_fcom(emu, ST(nextop&7).f);
+        #else
         fpu_fcom(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
+        #endif
         break;
     case 0xE8:  /* FUCOMP ST0, STx */
     case 0xE9:
@@ -50,7 +54,11 @@
     case 0xED:
     case 0xEE:
     case 0xEF:
+        #ifdef USE_FLOAT
+        fpu_fcom(emu, ST(nextop&7).f);
+        #else
         fpu_fcom(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
+        #endif
         fpu_do_pop(emu);
         break;
 
@@ -85,15 +93,39 @@
             case 0: /* FLD double */
                 GET_ED;
                 fpu_do_push(emu);
+                #ifdef USE_FLOAT
+                *(uint64_t*)&d = *(int64_t*)ED;
+                ST0.f = d;
+                #else
                 ST0.ll = *(int64_t*)ED;
+                #endif
+                break;
+            case 1: /* FISTTP ED qword */
+                GET_ED;
+                #ifdef USE_FLOAT
+                *(int64_t*)ED = ST0.f;
+                #else
+                *(int64_t*)ED = ST0.d;
+                #endif
+                fpu_do_pop(emu);
                 break;
             case 2: /* FST double */
                 GET_ED;
+                #ifdef USE_FLOAT
+                d = ST0.f;
+                *(int64_t*)ED = *(uint64_t*)&d;
+                #else
                 *(int64_t*)ED = ST0.ll;
+                #endif
                 break;
             case 3: /* FSTP double */
                 GET_ED;
+                #ifdef USE_FLOAT
+                d = ST0.f;
+                *(int64_t*)ED = *(uint64_t*)&d;
+                #else
                 *(int64_t*)ED = ST0.ll;
+                #endif
                 fpu_do_pop(emu);
                 break;
             case 4: /* FRSTOR m108byte */
@@ -104,7 +136,12 @@
                     char* p =(char*)ED;
                     p += 28;
                     for (int i=0; i<8; ++i) {
+                        #ifdef USE_FLOAT
+                        d = ST(i).f;
+                        LD2D(p, &d);
+                        #else
                         LD2D(p, &ST(i).d);
+                        #endif
                         p+=10;
                     }
                 }
@@ -119,12 +156,22 @@
                     char* p =(char*)ED;
                     p += 28;
                     for (int i=0; i<8; ++i) {
+                        #ifdef USE_FLOAT
+                        D2LD(&d, p);
+                        ST(i).f = d;
+                        #else
                         D2LD(&ST(i).d, p);
+                        #endif
                         p+=10;
                     }
                 }
                 reset_fpu(emu);
                 break;
+            case 7: /* FNSTSW m2byte */
+                GET_ED;
+                emu->sw.f.F87_TOP = emu->top&7;
+                *(uint16_t*)ED = emu->sw.x16;
+                break;
             default:
                 goto _default;
         }

src/emu/runde.h

@@ -1,7 +1,11 @@
     nextop = F8;
     switch (nextop) {
     case 0xC1:  /* FADDP ST1, ST0 */
+        #ifdef USE_FLOAT
+        ST(1).f += ST0.f;
+        #else
         ST(1).d += ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
     case 0xC0:  /* FADDP STx, ST0 */
@@ -11,11 +15,19 @@
     case 0xC5:
     case 0xC6:
     case 0xC7:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f += ST0.f;
+        #else
         ST(nextop&7).d += ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
     case 0xC9:  /* FMULP ST1, ST0 */
+        #ifdef USE_FLOAT
+        ST1.f *= ST0.f;
+        #else
         ST1.d *= ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
     case 0xC8:  /* FMULP STx, ST0 */
@@ -25,11 +37,19 @@
     case 0xCD:
     case 0xCE:
     case 0xCF:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f *= ST0.f;
+        #else
         ST(nextop&7).d *= ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
     case 0xD9:  /* FCOMPP */
+        #ifdef USE_FLOAT
+        fpu_fcom(emu, ST1.f);
+        #else
         fpu_fcom(emu, ST1.d);
+        #endif
         fpu_do_pop(emu);
         fpu_do_pop(emu);
         break;
@@ -45,11 +65,19 @@
     case 0xE5:
     case 0xE6:
     case 0xE7:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f = ST0.f - ST(nextop&7).f;
+        #else
         ST(nextop&7).d = ST0.d - ST(nextop&7).d;
+        #endif
         fpu_do_pop(emu);
         break;
    case 0xE9:  /* FSUBP ST1, ST0 */
+        #ifdef USE_FLOAT
+        ST1.f -= ST0.f;
+        #else
         ST1.d -= ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
     case 0xE8:  /* FSUBP STx, ST0 */
@@ -59,12 +87,13 @@
     case 0xED:
     case 0xEE:
     case 0xEF:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f -= ST0.f;
+        #else
         ST(nextop&7).d -= ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
-        /*ST(1).d = ST0.d / ST(1).d;
-        fpu_do_pop(emu);
-        break;*/
     case 0xF0:  /* FDIVRP STx, ST0 */
     case 0xF1:  /* FDIVRP ST1, ST0 */
     case 0xF2:
@@ -73,13 +102,13 @@
     case 0xF5:
     case 0xF6:
     case 0xF7:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f = ST0.f / ST(nextop&7).f;
+        #else
         ST(nextop&7).d = ST0.d / ST(nextop&7).d;
+        #endif
         fpu_do_pop(emu);
         break;
-        /*d = ST0.d;
-        fpu_do_pop(emu);
-        ST0.d /= d;
-        break;*/
     case 0xF8:  /* FDIVP STx, ST0 */
     case 0xF9:  /* FDIVP ST1, ST0 */
     case 0xFA:
@@ -88,7 +117,11 @@
     case 0xFD:
     case 0xFE:
     case 0xFF:
+        #ifdef USE_FLOAT
+        ST(nextop&7).f /= ST0.f;
+        #else
         ST(nextop&7).d /= ST0.d;
+        #endif
         fpu_do_pop(emu);
         break;
 
@@ -113,23 +146,51 @@
         switch((nextop>>3)&7) {
             case 0:     /* FIADD ST0, Ew int */
                 GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f += EW->sword[0];
+                #else
                 ST0.d += EW->sword[0];
+                #endif
                 break;
             case 1:     /* FIMUL ST0, Ew int */
                 GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f *= EW->sword[0];
+                #else
                 ST0.d *= EW->sword[0];
+                #endif
                 break;
             case 4:     /* FISUB ST0, Ew int */
                 GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f -= EW->sword[0];
+                #else
                 ST0.d -= EW->sword[0];
+                #endif
                 break;
             case 5:     /* FISUBR ST0, Ew int */
                 GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f = EW->sword[0] - ST0.f;
+                #else
                 ST0.d = EW->sword[0] - ST0.d;
+                #endif
+                break;
+            case 6:     /* FIDIV ST0, Ew int */
+                GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f /= EW->sword[0];
+                #else
+                ST0.d /= EW->sword[0];
+                #endif
                 break;
             case 7:     /* FIDIVR ST0, Ew int */
                 GET_EW;
+                #ifdef USE_FLOAT
+                ST0.f = EW->sword[0] / ST0.f;
+                #else
                 ST0.d = EW->sword[0] / ST0.d;
+                #endif
                 break;
         default:
             goto _default;

src/emu/rundf.h

@@ -0,0 +1,240 @@
+    nextop = F8;
+    switch (nextop) {
+    case 0xE0:  /* FNSTSW AX */
+        emu->sw.f.F87_TOP = emu->top&7;
+        R_AX = emu->sw.x16;
+        break;
+
+    case 0xE8:  /* FUCOMIP ST0, STx */
+    case 0xE9:
+    case 0xEA:
+    case 0xEB:
+    case 0xEC:
+    case 0xED:
+    case 0xEE:
+    case 0xEF:
+        #ifdef USE_FLOAT
+        fpu_fcomi(emu, ST(nextop&7).f);
+        #else
+        fpu_fcomi(emu, ST(nextop&7).d);   // bad, should handle QNaN and IA interrupt
+        #endif
+        fpu_do_pop(emu);
+        break;
+
+    case 0xF0:  /* FCOMIP ST0, STx */
+    case 0xF1:
+    case 0xF2:
+    case 0xF3:
+    case 0xF4:
+    case 0xF5:
+    case 0xF6:
+    case 0xF7:
+        #ifdef USE_FLOAT
+        fpu_fcomi(emu, ST(nextop&7).f);
+        #else
+        fpu_fcomi(emu, ST(nextop&7).d);
+        #endif
+        fpu_do_pop(emu);
+        break;
+
+    case 0xC0:
+    case 0xC1:
+    case 0xC2:
+    case 0xC3:
+    case 0xC4:
+    case 0xC5:
+    case 0xC6:
+    case 0xC7:
+    case 0xC8:
+    case 0xC9:
+    case 0xCA:
+    case 0xCB:
+    case 0xCC:
+    case 0xCD:
+    case 0xCE:
+    case 0xCF:
+    case 0xD0:
+    case 0xD1:
+    case 0xD2:
+    case 0xD3:
+    case 0xD4:
+    case 0xD5:
+    case 0xD6:
+    case 0xD7:
+    case 0xD8:
+    case 0xD9:
+    case 0xDA:
+    case 0xDB:
+    case 0xDC:
+    case 0xDD:
+    case 0xDE:
+    case 0xDF:
+    case 0xE1:
+    case 0xE2:
+    case 0xE3:
+    case 0xE4:
+    case 0xE5:
+    case 0xE6:
+    case 0xE7:
+    case 0xF8:
+    case 0xF9:
+    case 0xFA:
+    case 0xFB:
+    case 0xFC:
+    case 0xFD:
+    case 0xFE:
+    case 0xFF:
+        goto _default;
+
+    default:
+        switch((nextop>>3)&7) {
+        case 0: /* FILD ST0, Gw */
+            GET_EW;
+            tmp16s = EW->sword[0];
+            fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = tmp16s;
+            #else
+            ST0.d = tmp16s;
+            #endif
+            break;
+        case 1: /* FISTTP Ew, ST0 */
+            GET_EW;
+            #ifdef USE_FLOAT
+            tmp16s = ST0.f;
+            #else
+            tmp16s = ST0.d;
+            #endif
+            EW->sword[0] = tmp16s;
+            fpu_do_pop(emu);
+            break;
+        case 2: /* FIST Ew, ST0 */
+            GET_EW;
+            #ifdef USE_FLOAT
+            if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
+                EW->sword[0] = 0x8000;
+            else {
+                switch(emu->round) {
+                    case ROUND_Nearest:
+                        EW->sword[0] = floorf(ST0.f+0.5);
+                        break;
+                    case ROUND_Down:
+                        EW->sword[0] = floorf(ST0.f);
+                        break;
+                    case ROUND_Up:
+                        EW->sword[0] = ceilf(ST0.f);
+                        break;
+                    case ROUND_Chop:
+                        EW->sword[0] = ST0.f;
+                        break;
+                }
+            }
+            #else
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
+                EW->sword[0] = 0x8000;
+            else {
+                switch(emu->round) {
+                    case ROUND_Nearest:
+                        EW->sword[0] = floor(ST0.d+0.5);
+                        break;
+                    case ROUND_Down:
+                        EW->sword[0] = floor(ST0.d);
+                        break;
+                    case ROUND_Up:
+                        EW->sword[0] = ceil(ST0.d);
+                        break;
+                    case ROUND_Chop:
+                        EW->sword[0] = ST0.d;
+                        break;
+                }
+            }
+            #endif
+            break;
+        case 3: /* FISTP Ew, ST0 */
+            GET_EW;
+            #ifdef USE_FLOAT
+            if(isgreater(ST0.f, (float)(int32_t)0x7fff) || isless(ST0.f, -(float)(int32_t)0x7fff))
+                EW->sword[0] = 0x8000;
+            else {
+                switch(emu->round) {
+                    case ROUND_Nearest:
+                        EW->sword[0] = floorf(ST0.f+0.5);
+                        break;
+                    case ROUND_Down:
+                        EW->sword[0] = floorf(ST0.f);
+                        break;
+                    case ROUND_Up:
+                        EW->sword[0] = ceilf(ST0.f);
+                        break;
+                    case ROUND_Chop:
+                        EW->sword[0] = ST0.f;
+                        break;
+                }
+            }
+            #else
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff))
+                EW->sword[0] = 0x8000;
+            else {
+                switch(emu->round) {
+                    case ROUND_Nearest:
+                        EW->sword[0] = floor(ST0.d+0.5);
+                        break;
+                    case ROUND_Down:
+                        EW->sword[0] = floor(ST0.d);
+                        break;
+                    case ROUND_Up:
+                        EW->sword[0] = ceil(ST0.d);
+                        break;
+                    case ROUND_Chop:
+                        EW->sword[0] = ST0.d;
+                        break;
+                }
+            }
+            #endif
+            fpu_do_pop(emu);
+            break;
+        case 4: /* FBLD ST0, tbytes */
+            GET_ED;
+            fpu_do_push(emu);
+            fpu_fbld(emu, (uint8_t*)ED);
+            break;
+        case 5: /* FILD ST0, Gq */
+            GET_ED;
+            tmp64s = *(int64_t*)ED;
+            fpu_do_push(emu);
+            #ifdef USE_FLOAT
+            ST0.f = tmp64s;
+            #else
+            ST0.d = tmp64s;
+            #endif
+            STll(0).ll = tmp64s;
+            STll(0).ref = ST0.ll;
+            break;
+        case 6: /* FBSTP tbytes, ST0 */
+            GET_ED;
+            fpu_fbst(emu, (uint8_t*)ED);
+            fpu_do_pop(emu);
+            break;
+        case 7: /* FISTP i64 */
+            GET_ED;
+            if(STll(0).ref==ST(0).ll) {
+                *(int64_t*)ED = STll(0).ll;
+            } else {
+                #ifdef USE_FLOAT
+                if(isgreater(ST0.f, (float)(int64_t)0x7fffffffffffffffLL) || isless(ST0.f, (float)(int64_t)0x8000000000000000LL))
+                    *(int64_t*)ED = 0x8000000000000000LL;
+                else
+                    *(int64_t*)ED = (int64_t)ST0.f;
+                #else
+                if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL))
+                    *(int64_t*)ED = 0x8000000000000000LL;
+                else
+                    *(int64_t*)ED = (int64_t)ST0.d;
+                #endif
+            }
+            fpu_do_pop(emu);
+            break;
+        default:
+            goto _default;
+        }
+    }

src/emu/runf20f.h

@@ -15,6 +15,11 @@
         GET_EX;
         EX->q[0] = GX.q[0];
         break;
+    case 0x12:  /* MOVDDUP Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.q[1] = GX.q[0] = EX->q[0];
+        break;
 
     case 0x2A:  /* CVTSI2SD Gx, Ed */
         nextop = F8;
@@ -23,12 +28,28 @@
         break;
 
     case 0x2C:  /* CVTTSD2SI Gd, Ex */
-    case 0x2D:  /* CVTSD2SI Gd, Ex */
         nextop = F8;
         GET_EX;
         GD.sdword[0] = EX->d[0];
         break;
-
+    case 0x2D:  /* CVTSD2SI Gd, Ex */
+        nextop = F8;
+        GET_EX;
+        switch(emu->round) {
+            case ROUND_Nearest:
+                GD.sdword[0] = floor(EX->d[0]+0.5);
+                break;
+            case ROUND_Down:
+                GD.sdword[0] = floor(EX->d[0]);
+                break;
+            case ROUND_Up:
+                GD.sdword[0] = ceil(EX->d[0]);
+                break;
+            case ROUND_Chop:
+                GD.sdword[0] = EX->d[0];
+                break;
+        }
+        break;
     case 0x51:  /* SQRTSD Gx, Ex */
         nextop = F8;
         GET_EX;

src/emu/runf30f.h

@@ -23,17 +23,44 @@
         break;
 
     case 0x2C:  /* CVTTSS2SI Gd, Ex */
-    case 0x2D:  /* CVTSS2SI Gd, Ex */
         nextop = F8;
         GET_EX;
         GD.sdword[0] = EX->f[0];
         break;
+    case 0x2D:  /* CVTSS2SI Gd, Ex */
+        nextop = F8;
+        GET_EX;
+        switch(emu->round) {
+            case ROUND_Nearest:
+                GD.sdword[0] = floor(EX->f[0]+0.5);
+                break;
+            case ROUND_Down:
+                GD.sdword[0] = floor(EX->f[0]);
+                break;
+            case ROUND_Up:
+                GD.sdword[0] = ceil(EX->f[0]);
+                break;
+            case ROUND_Chop:
+                GD.sdword[0] = EX->f[0];
+                break;
+        }
+        break;
 
     case 0x51:  /* SQRTSS Gx, Ex */
         nextop = F8;
         GET_EX;
         GX.f[0] = sqrtf(EX->f[0]);
         break;
+    case 0x52:  /* RSQRTSS Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.f[0] = 1.0f/sqrtf(EX->f[0]);
+        break;
+    case 0x53:  /* RCPSS Gx, Ex */
+        nextop = F8;
+        GET_EX;
+        GX.f[0] = 1.0f/EX->f[0];
+        break;
 
     case 0x58:  /* ADDSS Gx, Ex */
         nextop = F8;
@@ -114,6 +141,25 @@
         memcpy(EX, &GX, 16);    // unaligned...
         break;
 
+    case 0xBC:  /* TZCNT Ew,Gw */
+        CHECK_FLAGS(emu);
+        nextop = F8;
+        GET_ED;
+        tmp32u = ED->dword[0];
+        if(tmp32u) {
+            CLEAR_FLAG(F_ZF);
+            tmp8u = 0;
+            while(!(tmp32u&(1<<tmp8u))) ++tmp8u;
+            GD.dword[0] = tmp8u;
+            CONDITIONAL_SET_FLAG(tmp8u, F_ZF);
+            CLEAR_FLAG(F_CF);
+        } else {
+            CLEAR_FLAG(F_ZF);
+            SET_FLAG(F_CF);
+            GD.dword[0] = 32;
+        }
+        break;
+
     case 0xC2:  /* CMPSS Gx, Ex, Ib */
         nextop = F8;
         GET_EX;

src/emu/x86emu.c

@@ -2,9 +2,12 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <sys/time.h>
 
 #include "debug.h"
-#include "stack.h"
+#include "box86stack.h"
 #include "x86emu.h"
 #include "x86emu_private.h"
 #include "x87emu_private.h"
@@ -19,81 +22,97 @@ typedef struct cleanup_s {
     void*       a;
 } cleanup_t;
 
+static uint32_t x86emu_parity_tab[8] =
+{
+	0x96696996,
+	0x69969669,
+	0x69969669,
+	0x96696996,
+	0x69969669,
+	0x96696996,
+	0x96696996,
+	0x69969669,
+};
+
 static uint8_t EndEmuMarker[] = {0xcc, 'S', 'C', 0, 0, 0, 0};
 void PushExit(x86emu_t* emu)
 {
     Push(emu, (uint32_t)&EndEmuMarker);
 }
-/// maxval not inclusive
-int getrand(int maxval)
+
+void* GetExit()
 {
-    if(maxval<1024) {
-        return ((random()&0x7fff)*maxval)/0x7fff;
-    } else {
-        uint64_t r = random();
-        r = (r*maxval) / RAND_MAX;
-        return r;
-    }
+    return &EndEmuMarker;
 }
 
-x86emu_t *NewX86Emu(box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize)
+static void internalX86Setup(x86emu_t* emu, box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
 {
-    printf_log(LOG_DEBUG, "Allocate a new X86 Emu, with EIP=%p and Stack=%p/0x%X\n", (void*)start, (void*)stack, stacksize);
-
-    x86emu_t *emu = (x86emu_t*)calloc(1, sizeof(x86emu_t));
     emu->context = context;
     // setup cpu helpers
     for (int i=0; i<8; ++i)
         emu->sbiidx[i] = &emu->regs[i];
     emu->sbiidx[4] = &emu->zero;
-    emu->packed_eflags.x32 = 0x02; // default flags?
+    emu->x86emu_parity_tab = x86emu_parity_tab;
+    emu->packed_eflags.x32 = 0x202; // default flags?
     UnpackFlags(emu);
+    // own stack?
+    if(ownstack)
+        emu->stack = (void*)stack;
+    emu->init_stack = (void*)stack;
+    emu->size_stack = stacksize;
     // set default value
     R_EIP = start;
-    R_ESP = stack + stacksize;
+    R_ESP = (stack + stacksize) & ~7;   // align stack start, always
+    // fake init of segments...
+    emu->segs[_CS] = 0x73;
+    emu->segs[_DS] = emu->segs[_ES] = emu->segs[_SS] = 0x7b;
+    emu->segs[_FS] = 0;
+    emu->segs[_GS] = 0x33;
     // setup fpu regs
     reset_fpu(emu);
-    // that should be enough
-    emu->scratch = (uint32_t*)calloc(200, sizeof(uint32_t));
-
-
     // if trace is activated
     if(context->x86trace) {
         emu->dec = InitX86TraceDecoder(context);
         if(!emu->dec)
             printf_log(LOG_INFO, "Failed to initialize Zydis decoder and formater, no trace activated\n");
     }
+}
+
+x86emu_t *NewX86Emu(box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
+{
+    printf_log(LOG_DEBUG, "Allocate a new X86 Emu, with EIP=%p and Stack=%p/0x%X\n", (void*)start, (void*)stack, stacksize);
+
+    x86emu_t *emu = (x86emu_t*)calloc(1, sizeof(x86emu_t));
+
+    internalX86Setup(emu, context, start, stack, stacksize, ownstack);
+
+    return emu;
+}
+
+x86emu_t *NewX86EmuFromStack(x86emu_t* emu, box86context_t *context, uintptr_t start, uintptr_t stack, int stacksize, int ownstack)
+{
+    printf_log(LOG_DEBUG, "New X86 Emu from stack, with EIP=%p and Stack=%p/0x%X\n", (void*)start, (void*)stack, stacksize);
+
+    internalX86Setup(emu, context, start, stack, stacksize, ownstack);
     
     return emu;
 }
 
-void SetupX86Emu(x86emu_t *emu, int* shared_global, void* globals)
+void SetupX86Emu(x86emu_t *emu)
 {
     printf_log(LOG_DEBUG, "Setup X86 Emu\n");
-
-    // Setup the GS segment:
-    if(shared_global) {
-        emu->globals = globals;
-        emu->shared_global = shared_global;
-    } else {
-        emu->globals = calloc(1, 256);  // arbitrary 256 byte size?
-        // calc canary...
-        uint8_t canary[4];
-        for (int i=0; i<4; ++i) canary[i] = 1 +  getrand(255);
-        canary[getrand(4)] = 0;
-        memcpy(emu->globals+0x14, canary, sizeof(canary));  // put canary in place
-        printf_log(LOG_DEBUG, "Setting up canary (for Stack protector) at GS:0x14, value:%08X\n", *(uint32_t*)canary);
-        emu->shared_global = (int*)calloc(1, sizeof(int));
-    }
-    (*emu->shared_global)++;
 }
 
 void SetTraceEmu(x86emu_t *emu, uintptr_t trace_start, uintptr_t trace_end)
 {
-    if (trace_end == 0) {
-        printf_log(LOG_INFO, "Setting trace\n");
-    } else {
-        printf_log(LOG_INFO, "Setting trace only between %p and %p\n", (void*)trace_start, (void*)trace_end);
+    if(emu->dec) {
+        if (trace_end == 0) {
+            printf_log(LOG_INFO, "Setting trace\n");
+        } else {
+            if(trace_end!=1) {  // 0-1 is basically no trace, so don't printf it...
+                printf_log(LOG_INFO, "Setting trace only between %p and %p\n", (void*)trace_start, (void*)trace_end);
+            }
+        }
     }
     emu->trace_start = trace_start;
     emu->trace_end = trace_end;
@@ -129,7 +148,7 @@ void CallCleanup(x86emu_t *emu, void* p)
                 Push(emu, (uintptr_t)emu->cleanups[i].a);
             PushExit(emu);
             emu->ip.dword[0] = (uintptr_t)(emu->cleanups[i].f);
-            Run(emu);
+            Run(emu, 0);
             emu->quit = 0;
             // now remove the cleanup
             if(i!=emu->clean_sz-1)
@@ -142,6 +161,7 @@ void CallCleanup(x86emu_t *emu, void* p)
 
 void CallAllCleanup(x86emu_t *emu)
 {
+    /* disabling for now...
     for(int i=emu->clean_sz-1; i>=0; --i) {
         printf_log(LOG_DEBUG, "Call cleanup #%d\n", i);
         if(emu->cleanups[i].arg)
@@ -151,35 +171,46 @@ void CallAllCleanup(x86emu_t *emu)
         Run(emu);
         emu->quit = 0;
     }
+    */
     emu->clean_sz = 0;
 }
 
+static void internalFreeX86(x86emu_t* emu)
+{
+    // stop trace now
+    if(emu->dec)
+        DeleteX86TraceDecoder(&emu->dec);
+    emu->dec = NULL;
+    // call atexit and fini first!
+    CallAllCleanup(emu);
+    free(emu->cleanups);
+
+    free(emu->stack);
+}
+
 void FreeX86Emu(x86emu_t **emu)
 {
     if(!emu)
         return;
     printf_log(LOG_DEBUG, "Free a X86 Emu (%p)\n", *emu);
-    // stop trace now
-    if((*emu)->dec)
-        DeleteX86TraceDecoder(&(*emu)->dec);
-    (*emu)->dec = NULL;
-    // call atexit and fini first!
-    CallAllCleanup(*emu);
-    free((*emu)->cleanups);
-    if((*emu)->shared_global && !(*(*emu)->shared_global)--) {
-        if((*emu)->globals)
-            free((*emu)->globals);
-        free((*emu)->shared_global);
-    }
 
-    free((*emu)->scratch);
-
-    free((*emu)->stack);
+    internalFreeX86(*emu);
 
     free(*emu);
     *emu = NULL;
 }
 
+void FreeX86EmuFromStack(x86emu_t **emu)
+{
+    if(!emu)
+        return;
+    printf_log(LOG_DEBUG, "Free a X86 Emu from stack (%p)\n", *emu);
+
+    internalFreeX86(*emu);
+
+    *emu = NULL;
+}
+
 void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
 {
 	memcpy(newemu->regs, emu->regs, sizeof(emu->regs));
@@ -201,10 +232,9 @@ void CloneEmu(x86emu_t *newemu, const x86emu_t* emu)
     memcpy(newemu->mmx, emu->mmx, sizeof(emu->mmx));
     memcpy(newemu->xmm, emu->xmm, sizeof(emu->xmm));
     newemu->mxcsr = emu->mxcsr;
-    memcpy(&newemu->zero, &emu->zero, sizeof(emu->zero));
-    memcpy(newemu->sbiidx, emu->sbiidx, sizeof(emu->sbiidx));
     newemu->quit = emu->quit;
     newemu->error = emu->error;
+    SetTraceEmu(newemu, emu->trace_start, emu->trace_end);
     // addapt R_ESP to new stack frame
     uintptr_t oldst = (uintptr_t)((emu->stack)?emu->stack:emu->context->stack);
     uintptr_t newst = (uintptr_t)((newemu->stack)?newemu->stack:newemu->context->stack);
@@ -220,6 +250,10 @@ uint32_t GetEAX(x86emu_t *emu)
 {
     return R_EAX;
 }
+uint64_t GetEDXEAX(x86emu_t *emu)
+{
+    return ((uint64_t)R_EAX)|(((uint64_t)R_EDX)<<32);
+}
 void SetEAX(x86emu_t *emu, uint32_t v)
 {
     R_EAX = v;
@@ -240,6 +274,27 @@ void SetEIP(x86emu_t *emu, uint32_t v)
 {
     R_EIP = v;
 }
+void SetESI(x86emu_t *emu, uint32_t v)
+{
+    R_ESI = v;
+}
+void SetEDI(x86emu_t *emu, uint32_t v)
+{
+    R_EDI = v;
+}
+void SetEBP(x86emu_t *emu, uint32_t v)
+{
+    R_EBP = v;
+}
+void SetESP(x86emu_t *emu, uint32_t v)
+{
+    R_ESP = v;
+}
+uint32_t GetESP(x86emu_t *emu)
+{
+    return R_ESP;
+}
+
 
 void ResetFlags(x86emu_t *emu)
 {
@@ -252,10 +307,18 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
     char* regname[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
     char tmp[80];
     buff[0] = '\0';
+    if(trace_emm) {
+        // do emm reg is needed
+        for(int i=0; i<8; ++i) {
+            sprintf(tmp, "mm%d:%016llx", i, emu->mmx[i].q);
+            strcat(buff, tmp);
+            if ((i&3)==3) strcat(buff, "\n"); else strcat(buff, " ");
+        }
+    }
     if(trace_xmm) {
         // do xmm reg is needed
         for(int i=0; i<8; ++i) {
-            sprintf(tmp, "%d:%016llX%016llX", i, emu->xmm[i].q[1], emu->xmm[i].q[0]);
+            sprintf(tmp, "%d:%016llx%016llx", i, emu->xmm[i].q[1], emu->xmm[i].q[0]);
             strcat(buff, tmp);
             if ((i&3)==3) strcat(buff, "\n"); else strcat(buff, " ");
         }
@@ -263,7 +326,11 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
     // start with FPU regs...
     if(emu->fpu_stack) {
         for (int i=0; i<emu->fpu_stack; i++) {
+            #ifdef USE_FLOAT
+            sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].f);
+            #else
             sprintf(tmp, "ST%d=%f", i, emu->fpu[(emu->top+i)&7].d);
+            #endif
             strcat(buff, tmp);
             int c = 10-strlen(tmp);
             if(c<1) c=1;
@@ -273,7 +340,7 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
         strcat(buff, "\n");
     }
     for (int i=_AX; i<=_DI; ++i) {
-        sprintf(tmp, "%s=%08X ", regname[i], emu->regs[i].dword[0]);
+        sprintf(tmp, "%s=%08x ", regname[i], emu->regs[i].dword[0]);
         strcat(buff, tmp);
 
         if (i==3) {
@@ -287,7 +354,7 @@ const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
             }
         }
     }
-    sprintf(tmp, "EIP=%08X ", ip);
+    sprintf(tmp, "EIP=%08x ", ip);
     strcat(buff, tmp);
     return buff;
 }
@@ -305,8 +372,9 @@ void UnimpOpcode(x86emu_t* emu)
 {
     R_EIP = emu->old_ip;
 
-    printf_log(LOG_NONE, "%p: Unimplemented Opcode %02X %02X %02X %02X %02X %02X %02X %02X\n", 
-        (void*)emu->old_ip,
+    int tid = syscall(SYS_gettid);
+    printf_log(LOG_NONE, "%04d|%p: Unimplemented Opcode %02X %02X %02X %02X %02X %02X %02X %02X\n", 
+        tid, (void*)emu->old_ip,
         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;
@@ -315,11 +383,100 @@ void UnimpOpcode(x86emu_t* emu)
 
 void EmuCall(x86emu_t* emu, uintptr_t addr)
 {
+    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);
-    uintptr_t oldip = R_EIP;
     R_EIP = addr;
     emu->df = d_none;
-    Run(emu);
+    Run(emu, 0);
     emu->quit = 0;  // reset Quit flags...
-    R_EIP = oldip;  // and set back instruction pointer
-}
+    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
+    }
+}
+
+#if defined(__ARM_ARCH)
+#if (__ARM_ARCH >= 6)
+static inline unsigned int arm_perf (void)
+{
+  unsigned int value;
+  // Read CCNT Register
+  asm volatile ("MRC p15, 0, %0, c9, c13, 0\t\n": "=r"(value));
+  return value;
+}
+static inline void init_perfcounters (int32_t do_reset, int32_t enable_divider)
+{
+  // in general enable all counters (including cycle counter)
+  int32_t value = 1;
+
+  // peform reset:
+  if (do_reset)
+  {
+    value |= 2;     // reset all counters to zero.
+    value |= 4;     // reset cycle counter to zero.
+  }
+
+  if (enable_divider)
+    value |= 8;     // enable "by 64" divider for CCNT.
+
+  value |= 16;
+
+  // program the performance-counter control-register:
+  asm volatile ("MCR p15, 0, %0, c9, c12, 0\t\n" :: "r"(value));
+
+  // enable all counters:
+  asm volatile ("MCR p15, 0, %0, c9, c12, 1\t\n" :: "r"(0x8000000f));
+
+  // clear overflows:
+  asm volatile ("MCR p15, 0, %0, c9, c12, 3\t\n" :: "r"(0x8000000f));
+}
+
+#endif
+#endif
+
+uint64_t ReadTSC(x86emu_t* emu)
+{
+    // Read the TimeStamp Counter as 64bits.
+    // this is supposed to be the number of instrunctions executed since last reset
+#if defined(__i386__)
+  uint64_t ret;
+  __asm__ volatile("rdtsc" : "=A"(ret));
+  return ret;
+#if 0
+#elif defined(__ARM_ARCH)
+#if (__ARM_ARCH >= 6)
+    static int init = 1;
+    if(init) {
+        init_perfcounters(1, 1);
+        init = 0;
+    }
+    uint32_t perf = arm_perf();
+    if(perf) {
+        return ((uint64_t)perf)*64;
+    }
+#endif
+#endif
+#endif
+// fall back to gettime...
+#ifndef NOGETCLOCK
+  struct timespec ts;
+  clock_gettime(CLOCK_MONOTONIC_COARSE, &ts);
+  return (uint64_t)(ts.tv_sec) * 1000000000LL + ts.tv_nsec;
+#else
+  struct timeval tv;
+  gettimeofday(&tv, NULL);
+  return (uint64_t)(tv.tv_sec) * 1000000 + tv.tv_usec;
+#endif
+}

src/emu/x86emu_private.h

@@ -5,6 +5,7 @@
 
 typedef struct zydis_dec_s zydis_dec_t;
 typedef struct box86context_s box86context_t;
+typedef struct i386_ucontext_s i386_ucontext_t;
 
 #define ERR_UNIMPL  1
 #define ERR_DIVBY0  2
@@ -12,57 +13,74 @@ typedef struct box86context_s box86context_t;
 
 typedef struct cleanup_s cleanup_t;
 
+typedef struct forkpty_s {
+    void*    amaster;
+    void*   name;
+    void*   termp;
+    void*   winp;
+    void*   f;  // forkpty function
+} forkpty_t;
+
 typedef struct x86emu_s {
     // cpu
 	reg32_t     regs[8],ip;
     int         flags[F_LAST];
 	x86flags_t  packed_eflags;
     uintptr_t   old_ip;
-    // segments
-    uint32_t    segs[6];    // only 32bits value?
     // fpu
 	fpu_reg_t   fpu[9];
-    fpu_ld_t    fpu_ld[9]; // for long double emulation / 80bits fld fst
-    fpu_ll_t    fpu_ll[9]; // for 64bits fild / fist sequence
-	fpu_p_reg_t p_regs[9];
 	uint16_t    cw,cw_mask_all;
 	x87flags_t  sw;
 	uint32_t    top;        // top is part of sw, but it's faster to have it separatly
     int         fpu_stack;
 	fpu_round_t round;
+    fpu_ld_t    fpu_ld[9]; // for long double emulation / 80bits fld fst
+    fpu_ll_t    fpu_ll[9]; // for 64bits fild / fist sequence
+	fpu_p_reg_t p_regs[9];
     // mmx
     mmx_regs_t  mmx[8];
     // sse
     sse_regs_t  xmm[8];
     uint32_t    mxcsr;
-    // cpu helpers
-    reg32_t     zero;
-    reg32_t     *sbiidx[8];
     // defered flags
     defered_flags_t df;
     uint32_t    op1;
     uint32_t    op2;
     uint32_t    res;
+    uint32_t    *x86emu_parity_tab; // helper
+    #ifdef HAVE_TRACE
+    uintptr_t   prev2_ip, prev_ip;
+    #endif
+    // segments
+    uint32_t    segs[6];    // only 32bits value?
     // 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
     // trace
     zydis_dec_t *dec;
     uintptr_t   trace_start, trace_end;
-    // global stuffs, pointed with GS: segment
-    void        *globals;
-    int         *shared_global;
     // parent context
     box86context_t *context;
+    // cpu helpers
+    reg32_t     zero;
+    reg32_t     *sbiidx[8];
     // atexit and fini functions
     cleanup_t   *cleanups;
     int         clean_sz;
     int         clean_cap;
-    // scratch stack, used for alignement of double and 64bits ints on arm
-    uint32_t    *scratch;
+    // scratch stack, used for alignement of double and 64bits ints on arm. 200 elements should be enough
+    uint32_t    scratch[200];
     // local stack, do be deleted when emu is freed
     void*       stack;
+    void*       init_stack; // initial stack (owned or not)
+    uint32_t    size_stack; // stack size (owned or not)
+
+    i386_ucontext_t *uc_link; // to handle setcontext
 
 } x86emu_t;
 

src/emu/x86int3.c

@@ -0,0 +1,216 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <sys/syscall.h>
+#include <sys/types.h>
+#include <pthread.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 "wrapper.h"
+#include "box86context.h"
+#include "librarian.h"
+
+typedef int32_t (*iFpppp_t)(void*, void*, void*, void*);
+
+x86emu_t* x86emu_fork(x86emu_t* emu, int forktype)
+{
+    // execute atforks prepare functions, in reverse order
+    for (int i=emu->context->atfork_sz-1; i>=0; --i)
+        EmuCall(emu, emu->context->atforks[i].prepare);
+    int v;
+    if(forktype==2) {
+        iFpppp_t forkpty = (iFpppp_t)emu->forkpty_info->f;
+        v = forkpty(emu->forkpty_info->amaster, emu->forkpty_info->name, emu->forkpty_info->termp, emu->forkpty_info->winp);
+        emu->forkpty_info = NULL;
+    } else
+        v = fork();
+    if(v==EAGAIN || v==ENOMEM) {
+        // error...
+    } else if(v!=0) {  
+        // execute atforks parent functions
+        for (int i=0; i<emu->context->atfork_sz; --i)
+            EmuCall(emu, emu->context->atforks[i].parent);
+
+    } else if(v==0) {
+        // execute atforks child functions
+        for (int i=0; i<emu->context->atfork_sz; --i)
+            EmuCall(emu, emu->context->atforks[i].child);
+    }
+    R_EAX = v;
+    return emu;
+}
+
+extern int errno;
+void x86Int3(x86emu_t* emu)
+{
+    if(Peek(emu, 0)=='S' && Peek(emu, 1)=='C') // Signature for "Out of x86 door"
+    {
+        R_EIP += 2;
+        uint32_t addr = Fetch32(emu);
+        if(addr==0) {
+            //printf_log(LOG_INFO, "%p:Exit x86 emu (emu=%p)\n", *(void**)(R_ESP), emu);
+            emu->quit=1; // normal quit
+        } else {
+            RESET_FLAGS(emu);
+            wrapper_t w = (wrapper_t)addr;
+            addr = Fetch32(emu);
+            if(box86_log>=LOG_DEBUG /*&& emu->trace_end==0 && !emu->context->x86trace*/) {
+                pthread_mutex_lock(&emu->context->mutex_trace);
+                int tid = syscall(SYS_gettid);
+                char buff[256] = "\0";
+                char buff2[64] = "\0";
+                char buff3[64] = "\0";
+                char *tmp;
+                int post = 0;
+                int perr = 0;
+                uint32_t *pu32 = NULL;
+                const char *s = GetNativeName(emu, (void*)addr);
+                if(addr==(uintptr_t)PltResolver) {
+                    snprintf(buff, 256, "%s", " ... ");
+                } else
+                if(strstr(s, "SDL_RWFromFile")==s || strstr(s, "SDL_RWFromFile")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%s, %s)", tid, *(void**)(R_ESP), s, *(char**)(R_ESP+4), *(char**)(R_ESP+8));
+                } else  if(strstr(s, "glColor4f")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f, %f, %f)", tid, *(void**)(R_ESP), "glColor4f", *(float*)(R_ESP+4), *(float*)(R_ESP+8), *(float*)(R_ESP+12), *(float*)(R_ESP+16));
+                } else  if(strstr(s, "glTexCoord2f")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f)", tid, *(void**)(R_ESP), "glTexCoord2f", *(float*)(R_ESP+4), *(float*)(R_ESP+8));
+                } else  if(strstr(s, "glVertex2f")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f)", tid, *(void**)(R_ESP), "glVertex2f", *(float*)(R_ESP+4), *(float*)(R_ESP+8));
+                } else  if(strstr(s, "glVertex3f")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f, %f)", tid, *(void**)(R_ESP), "glVertex3f", *(float*)(R_ESP+4), *(float*)(R_ESP+8), *(float*)(R_ESP+12));
+                } else  if(strstr(s, "__open64")==s || strstr(s, "open64")==s || strstr(s, "open64")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", %d, %d)", tid, *(void**)(R_ESP), "open64", *(char**)(R_ESP+4), *(int*)(R_ESP+8), *(int*)(R_ESP+12));
+                    perr = 1;
+                } else  if(strstr(s, "__open")==s || strstr(s, "open")==s || strstr(s, "open")==s) {
+                    tmp = *(char**)(R_ESP+4);
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", %d)", tid, *(void**)(R_ESP), s, (tmp)?tmp:"(nil)", *(int*)(R_ESP+8));
+                    perr = 1;
+                } else  if(strstr(s, "fopen")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", \"%s\")", tid, *(void**)(R_ESP), "fopen", *(char**)(R_ESP+4), *(char**)(R_ESP+8));
+                    perr = 1;
+                } else  if(strstr(s, "chdir")==s) {
+                    pu32=*(uint32_t**)(R_ESP+4);
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\")", tid, *(void**)(R_ESP), "chdir", pu32?((pu32==(uint32_t*)1)?"/1/":(char*)pu32):"/0/");
+                } else  if(strstr(s, "getenv")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\")", tid, *(void**)(R_ESP), "getenv", *(char**)(R_ESP+4));
+                    post = 2;
+                } else  if(strstr(s, "pread")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%d, %p, %u, %i)", tid, *(void**)(R_ESP), "pread", *(int32_t*)(R_ESP+4), *(void**)(R_ESP+8), *(uint32_t*)(R_ESP+12), *(int32_t*)(R_ESP+16));
+                } else  if(strstr(s, "statvfs64")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p(\"%s\"), %p)", tid, *(void**)(R_ESP), "statvfs64", *(void**)(R_ESP+4), *(char**)(R_ESP+4), *(void**)(R_ESP+8));
+                } else  if(strstr(s, "index")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p(\"%s\"), %i(%c))", tid, *(void**)(R_ESP), "index", *(char**)(R_ESP+4), *(char**)(R_ESP+4), *(int32_t*)(R_ESP+8), *(int32_t*)(R_ESP+8));
+                } else  if(strstr(s, "rindex")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p(\"%s\"), %i(%c))", tid, *(void**)(R_ESP), "rindex", *(char**)(R_ESP+4), *(char**)(R_ESP+4), *(int32_t*)(R_ESP+8), *(int32_t*)(R_ESP+8));
+                } else  if(strstr(s, "__xstat64")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%d, %p(\"%s\"), %p)", tid, *(void**)(R_ESP), "__xstat64", *(int32_t*)(R_ESP+4), *(char**)(R_ESP+8), *(char**)(R_ESP+8), *(void**)(R_ESP+12));
+                    perr = 1;
+                } else  if(strstr(s, "__lxstat64")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%d, %p(\"%s\"), %p)", tid, *(void**)(R_ESP), "__lxstat64", *(int32_t*)(R_ESP+4), *(char**)(R_ESP+8), *(char**)(R_ESP+8), *(void**)(R_ESP+12));
+                    perr = 1;
+                } else  if(strstr(s, "sem_timedwait")==s) {
+                    pu32 = *(uint32_t**)(R_ESP+8);
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p, %p[%d sec %d ns])", tid, *(void**)(R_ESP), "sem_timedwait", *(void**)(R_ESP+4), *(void**)(R_ESP+8), pu32?pu32[0]:-1, pu32?pu32[1]:-1);
+                    perr = 1;
+                } else  if(strstr(s, "clock_gettime")==s || strstr(s, "__clock_gettime")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%d, %p)", tid, *(void**)(R_ESP), "clock_gettime", *(uint32_t*)(R_ESP+4), *(void**)(R_ESP+8));
+                    post = 1;
+                    pu32 = *(uint32_t**)(R_ESP+8);
+                } else  if(strstr(s, "strcasecmp")==s || strstr(s, "__strcasecmp")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", \"%s\")", tid, *(void**)(R_ESP), "strcasecmp", *(char**)(R_ESP+4), *(char**)(R_ESP+8));
+                } else  if(strstr(s, "gtk_signal_connect_full")) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p, \"%s\", %p, %p, %p, %p, %d, %d)", tid, *(void**)(R_ESP), "gtk_signal_connect_full", *(void**)(R_ESP+4), *(char**)(R_ESP+8), *(void**)(R_ESP+12), *(void**)(R_ESP+16), *(void**)(R_ESP+20), *(void**)(R_ESP+24), *(int32_t*)(R_ESP+28), *(int32_t*)(R_ESP+32));
+                } else  if(strstr(s, "strncasecmp")==s || strstr(s, "__strncasecmp")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", \"%s\", %u)", tid, *(void**)(R_ESP), "strcasecmp", *(char**)(R_ESP+4), *(char**)(R_ESP+8), *(uint32_t*)(R_ESP+12));
+                } else  if(strstr(s, "strcmp")==s || strstr(s, "__strcmp")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", \"%s\")", tid, *(void**)(R_ESP), "strcmp", *(char**)(R_ESP+4), *(char**)(R_ESP+8));
+                } else  if(strstr(s, "strstr")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(\"%s\", \"%s\")", tid, *(void**)(R_ESP), "strstr", *(char**)(R_ESP+4), *(char**)(R_ESP+8));
+                } else  if(strstr(s, "strlen")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p(\"%s\"))", tid, *(void**)(R_ESP), "strlen", *(char**)(R_ESP+4), ((R_ESP+4))?(*(char**)(R_ESP+4)):"nil");
+                } else  if(strstr(s, "vsnprintf")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%08X, %u, %08X...)", tid, *(void**)(R_ESP), "vsnprintf", *(uint32_t*)(R_ESP+4), *(uint32_t*)(R_ESP+8), *(uint32_t*)(R_ESP+12));
+                    pu32 = *(uint32_t**)(R_ESP+4);
+                    post = 3;
+                } else  if(strstr(s, "vsprintf")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%08X, \"%s\"...)", tid, *(void**)(R_ESP), "vsprintf", *(uint32_t*)(R_ESP+4), *(char**)(R_ESP+8));
+                    pu32 = *(uint32_t**)(R_ESP+4);
+                    post = 3;
+                } else  if(strstr(s, "snprintf")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%08X, %u, %08X...)", tid, *(void**)(R_ESP), "snprintf", *(uint32_t*)(R_ESP+4), *(uint32_t*)(R_ESP+8), *(uint32_t*)(R_ESP+12));
+                    pu32 = *(uint32_t**)(R_ESP+4);
+                    post = 3;
+                } else  if(strstr(s, "sprintf")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%08X, %08X...)", tid, *(void**)(R_ESP), "sprintf", *(uint32_t*)(R_ESP+4), *(uint32_t*)(R_ESP+8));
+                    pu32 = *(uint32_t**)(R_ESP+4);
+                    post = 3;
+                } else  if(strstr(s, "fprintf")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%08X, \"%s\", ...)", tid, *(void**)(R_ESP), "fprintf", *(uint32_t*)(R_ESP+4), *(char**)(R_ESP+8));
+                } else if(strstr(s, "XCreateWindow")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p, %p, %d, %d, %u, %u, %u, %d, %u, %p, %u, %p)", tid, *(void**)(R_ESP), "XCreateWindow", *(void**)(R_ESP+4), *(void**)(R_ESP+8), *(int*)(R_ESP+12), *(int*)(R_ESP+16), *(uint32_t*)(R_ESP+20), *(uint32_t*)(R_ESP+24), *(uint32_t*)(R_ESP+28), *(int32_t*)(R_ESP+32), *(uint32_t*)(R_ESP+36), *(void**)(R_ESP+40), *(uint32_t*)(R_ESP+44), *(void**)(R_ESP+48));
+                } else if(strstr(s, "XLoadQueryFont")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p, \"%s\")", tid, *(void**)(R_ESP), "XLoadQueryFont", *(void**)(R_ESP+4), *(char**)(R_ESP+8));
+                } else if(strstr(s, "pthread_mutex_lock")==s) {
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p)", tid, *(void**)(R_ESP), "pthread_mutex_lock", *(void**)(R_ESP+4));
+                } else if(strstr(s, "fmodf")==s) {
+                    post = 4;
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f)", tid, *(void**)(R_ESP), "fmodf", *(float*)(R_ESP+4), *(float*)(R_ESP+8));
+                } else if(strstr(s, "fmod")==s) {
+                    post = 4;
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%f, %f)", tid, *(void**)(R_ESP), "fmod", *(double*)(R_ESP+4), *(double*)(R_ESP+12));
+                } else if(strstr(s, "SDL_GetWindowSurface")==s) {
+                    post = 5;
+                    snprintf(buff, 255, "%04d|%p: Calling %s(%p)", tid, *(void**)(R_ESP), "SDL_GetWindowSurface", *(void**)(R_ESP+4));
+                } else {
+                    snprintf(buff, 255, "%04d|%p: Calling %s (%08X, %08X, %08X...)", tid, *(void**)(R_ESP), s, *(uint32_t*)(R_ESP+4), *(uint32_t*)(R_ESP+8), *(uint32_t*)(R_ESP+12));
+                }
+                printf_log(LOG_DEBUG, "%s =>", buff);
+                pthread_mutex_unlock(&emu->context->mutex_trace);
+                w(emu, addr);   // some function never come back, so unlock the mutex first!
+                pthread_mutex_lock(&emu->context->mutex_trace);
+                if(post)
+                    switch(post) {
+                    case 1: snprintf(buff2, 63, " [%d sec %d nsec]", pu32?pu32[0]:-1, pu32?pu32[1]:-1);
+                            break;
+                    case 2: snprintf(buff2, 63, "(%s)", R_EAX?((char*)R_EAX):"nil");
+                            break;
+                    case 3: snprintf(buff2, 63, "(%s)", pu32?((char*)pu32):"nil");
+                            break;
+                    #ifdef USE_FLOAT
+                    case 4: snprintf(buff2, 63, " (%f)", ST0.f);
+                    #else
+                    case 4: snprintf(buff2, 63, " (%f)", ST0.d);
+                    #endif
+                            break;
+                    case 5: {
+                            uint32_t* p = (uint32_t*)R_EAX;
+                            if(p)
+                                snprintf(buff2, 63, " size=%dx%d, pitch=%d, pixels=%p", p[2], p[3], p[4], p+5);
+                            else
+                                snprintf(buff2, 63, "NULL Surface");
+                            }
+                            break;
+                }
+                if(perr && ((int)R_EAX)<0)
+                    snprintf(buff3, 63, " (errno=%d)", errno);
+                printf_log(LOG_DEBUG, " return 0x%08X%s%s\n", R_EAX, buff2, buff3);
+                pthread_mutex_unlock(&emu->context->mutex_trace);
+            } else
+                w(emu, addr);
+        }
+        return;
+    }
+    printf_log(LOG_INFO, "Warning, ignoring unsupported Int 3 call\n");
+    //emu->quit = 1;
+}

src/emu/x86primop.c

@@ -105,19 +105,7 @@
 
 /*------------------------- Global Variables ------------------------------*/
 
-static uint32_t x86emu_parity_tab[8] =
-{
-	0x96696996,
-	0x69969669,
-	0x69969669,
-	0x96696996,
-	0x69969669,
-	0x96696996,
-	0x96696996,
-	0x69969669,
-};
-
-#define PARITY(x)   (((x86emu_parity_tab[(x) / 32] >> ((x) % 32)) & 1) == 0)
+#define PARITY(x)   (((emu->x86emu_parity_tab[(x) / 32] >> ((x) % 32)) & 1) == 0)
 #define XOR2(x) 	(((x) ^ ((x)>>1)) & 0x1)
 
 /*----------------------------- Implementation ----------------------------*/
@@ -223,8 +211,8 @@ Implements the ADC instruction and side effects.
 ****************************************************************************/
 uint8_t adc8(x86emu_t *emu, uint8_t d, uint8_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t cc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t cc;
 
 	CHECK_FLAGS(emu);
 
@@ -251,8 +239,8 @@ Implements the ADC instruction and side effects.
 ****************************************************************************/
 uint16_t adc16(x86emu_t *emu, uint16_t d, uint16_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t cc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t cc;
 
 	CHECK_FLAGS(emu);
 
@@ -279,10 +267,10 @@ Implements the ADC instruction and side effects.
 ****************************************************************************/
 uint32_t adc32(x86emu_t *emu, uint32_t d, uint32_t s)
 {
-	register uint32_t lo;	/* all operands in native machine order */
-	register uint32_t hi;
-	register uint32_t res;
-	register uint32_t cc;
+	uint32_t lo;	/* all operands in native machine order */
+	uint32_t hi;
+	uint32_t res;
+	uint32_t cc;
 
 	CHECK_FLAGS(emu);
 
@@ -297,7 +285,7 @@ uint32_t adc32(x86emu_t *emu, uint32_t d, uint32_t s)
 	hi = (lo >> 16) + (d >> 16) + (s >> 16);
 
 	CONDITIONAL_SET_FLAG(hi & 0x10000, F_CF);
-	CONDITIONAL_SET_FLAG((res & 0xffffffff) == 0, F_ZF);
+	CONDITIONAL_SET_FLAG(!res, F_ZF);
 	CONDITIONAL_SET_FLAG(res & 0x80000000, F_SF);
 	CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
 
@@ -314,8 +302,8 @@ Implements the CMP instruction and side effects.
 ****************************************************************************/
 uint8_t cmp8(x86emu_t *emu, uint8_t d, uint8_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t bc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t bc;
 
 	RESET_FLAGS(emu);
 
@@ -339,8 +327,8 @@ Implements the CMP instruction and side effects.
 ****************************************************************************/
 uint16_t cmp16(x86emu_t *emu, uint16_t d, uint16_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t bc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t bc;
 
 	RESET_FLAGS(emu);
 
@@ -363,14 +351,14 @@ Implements the CMP instruction and side effects.
 ****************************************************************************/
 uint32_t cmp32(x86emu_t *emu, uint32_t d, uint32_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t bc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t bc;
 
 	RESET_FLAGS(emu);
 
 	res = d - s;
 	CONDITIONAL_SET_FLAG(res & 0x80000000, F_SF);
-	CONDITIONAL_SET_FLAG((res & 0xffffffff) == 0, F_ZF);
+	CONDITIONAL_SET_FLAG(!res, F_ZF);
 	CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
 
 	/* calculate the borrow chain.  See note at top */
@@ -393,10 +381,11 @@ uint8_t daa8(x86emu_t *emu, uint8_t d)
 		res += 6;
 		SET_FLAG(F_AF);
 	}
-	if (res > 0x9F || ACCESS_FLAG(F_CF)) {
+	if (d > 0x99 || ACCESS_FLAG(F_CF)) {
 		res += 0x60;
 		SET_FLAG(F_CF);
-	}
+	} else
+		CLEAR_FLAG(F_CF);
 	CONDITIONAL_SET_FLAG(res & 0x80, F_SF);
 	CONDITIONAL_SET_FLAG((res & 0xFF) == 0, F_ZF);
 	CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
@@ -409,21 +398,24 @@ Implements the DAS instruction and side effects.
 ****************************************************************************/
 uint8_t das8(x86emu_t *emu, uint8_t d)
 {
+	uint32_t res = d;
 	CHECK_FLAGS(emu);
+	uint32_t newcf = 0;
 	if ((d & 0xf) > 9 || ACCESS_FLAG(F_AF)) {
-		d -= 6;
+		res -= 6;
+		newcf = (d < 6);
 		SET_FLAG(F_AF);
 	} else
 		CLEAR_FLAG(F_AF);
-	if (d > 0x9F || ACCESS_FLAG(F_CF)) {
-		d -= 0x60;
-		SET_FLAG(F_CF);
-	} else
-		CLEAR_FLAG(F_CF);
-	CONDITIONAL_SET_FLAG(d & 0x80, F_SF);
-	CONDITIONAL_SET_FLAG(d == 0, F_ZF);
-	CONDITIONAL_SET_FLAG(PARITY(d & 0xff), F_PF);
-	return d;
+	if (d > 0x99 || ACCESS_FLAG(F_CF)) {
+		res -= 0x60;
+		newcf = 1;
+	}
+	CONDITIONAL_SET_FLAG(newcf, F_CF);
+	CONDITIONAL_SET_FLAG(res & 0x80, F_SF);
+	CONDITIONAL_SET_FLAG((res & 0xFF) == 0, F_ZF);
+	CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
+	return res;
 }
 
 /****************************************************************************
@@ -432,7 +424,7 @@ Implements the RCL instruction and side effects.
 ****************************************************************************/
 uint8_t rcl8(x86emu_t *emu, uint8_t d, uint8_t s)
 {
-    register unsigned int res, cnt, mask, cf;
+    unsigned int res, cnt, mask, cf;
 	CHECK_FLAGS(emu);
 	s = s&0x1f;
 
@@ -507,7 +499,7 @@ Implements the RCL instruction and side effects.
 ****************************************************************************/
 uint16_t rcl16(x86emu_t *emu, uint16_t d, uint8_t s)
 {
-	register unsigned int res, cnt, mask, cf;
+	unsigned int res, cnt, mask, cf;
 	CHECK_FLAGS(emu);
 	s = s&0x1f;
 
@@ -533,14 +525,14 @@ Implements the RCL instruction and side effects.
 ****************************************************************************/
 uint32_t rcl32(x86emu_t *emu, uint32_t d, uint8_t s)
 {
-	register uint32_t res, cnt, mask, cf;
+	uint32_t res, cnt, mask, cf;
 	CHECK_FLAGS(emu);
 	s = s&0x1f;
 
 	res = d;
 	if ((cnt = s % 33) != 0) {
 		cf = (d >> (32 - cnt)) & 0x1;
-		res = (d << cnt) & 0xffffffff;
+		res = (d << cnt);
 		mask = (1 << (cnt - 1)) - 1;
 		res |= (d >> (33 - cnt)) & mask;
 		if (ACCESS_FLAG(F_CF)) {     /* carry flag is set */
@@ -899,7 +891,7 @@ uint16_t shld16 (x86emu_t *emu, uint16_t d, uint16_t fill, uint8_t s)
 			CLEAR_FLAG(F_OF);
 		}
 	} else {
-		res = (fill << (cnt)) | (d >> (16 - cnt));
+		res = (fill << (cnt)) | (fill >> (16 - cnt));
 		if(s==16)
 			cf = d & 1;
 		else
@@ -928,7 +920,7 @@ uint32_t shld32 (x86emu_t *emu, uint32_t d, uint32_t fill, uint8_t s)
 		res = (d << cnt) | (fill >> (32-cnt));
 		cf = d & (1 << (32 - cnt));
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		CONDITIONAL_SET_FLAG((res & 0xffffffff) == 0, F_ZF);
+		CONDITIONAL_SET_FLAG(!res, F_ZF);
 		CONDITIONAL_SET_FLAG(res & 0x80000000, F_SF);
 		CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
 	} else {
@@ -973,7 +965,7 @@ uint16_t shrd16 (x86emu_t *emu, uint16_t d, uint16_t fill, uint8_t s)
         }
 	} else {
 		cf = fill & (1 << (cnt - 1));
-		res = (fill >> cnt) | (d << (16 - cnt));
+		res = (fill >> cnt) | (fill << (16 - cnt));
 		CONDITIONAL_SET_FLAG(cf, F_CF);
 		CONDITIONAL_SET_FLAG((res & 0xffff) == 0, F_ZF);
 		CONDITIONAL_SET_FLAG(res & 0x8000, F_SF);
@@ -1006,7 +998,7 @@ uint32_t shrd32 (x86emu_t *emu, uint32_t d, uint32_t fill, uint8_t s)
 		cf = d & (1 << (cnt - 1));
 		res = (d >> cnt) | (fill << (32 - cnt));
 		CONDITIONAL_SET_FLAG(cf, F_CF);
-		CONDITIONAL_SET_FLAG((res & 0xffffffff) == 0, F_ZF);
+		CONDITIONAL_SET_FLAG(!res, F_ZF);
 		CONDITIONAL_SET_FLAG(res & 0x80000000, F_SF);
 		CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
 	} else {
@@ -1026,8 +1018,8 @@ Implements the SBB instruction and side effects.
 ****************************************************************************/
 uint8_t sbb8(x86emu_t *emu, uint8_t d, uint8_t s)
 {
-    register uint32_t res;   /* all operands in native machine order */
-    register uint32_t bc;
+    uint32_t res;   /* all operands in native machine order */
+    uint32_t bc;
 	CHECK_FLAGS(emu);
 
 	if (ACCESS_FLAG(F_CF))
@@ -1052,8 +1044,8 @@ Implements the SBB instruction and side effects.
 ****************************************************************************/
 uint16_t sbb16(x86emu_t *emu, uint16_t d, uint16_t s)
 {
-    register uint32_t res;   /* all operands in native machine order */
-    register uint32_t bc;
+    uint32_t res;   /* all operands in native machine order */
+    uint32_t bc;
 	CHECK_FLAGS(emu);
 
 	if (ACCESS_FLAG(F_CF))
@@ -1078,8 +1070,8 @@ Implements the SBB instruction and side effects.
 ****************************************************************************/
 uint32_t sbb32(x86emu_t *emu, uint32_t d, uint32_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
-	register uint32_t bc;
+	uint32_t res;   /* all operands in native machine order */
+	uint32_t bc;
 	CHECK_FLAGS(emu);
 
 	if (ACCESS_FLAG(F_CF))
@@ -1087,7 +1079,7 @@ uint32_t sbb32(x86emu_t *emu, uint32_t d, uint32_t s)
     else
         res = d - s;
 	CONDITIONAL_SET_FLAG(res & 0x80000000, F_SF);
-	CONDITIONAL_SET_FLAG((res & 0xffffffff) == 0, F_ZF);
+	CONDITIONAL_SET_FLAG(!res, F_ZF);
 	CONDITIONAL_SET_FLAG(PARITY(res & 0xff), F_PF);
 
 	/* calculate the borrow chain.  See note at top */
@@ -1104,7 +1096,7 @@ Implements the TEST instruction and side effects.
 ****************************************************************************/
 void test8(x86emu_t *emu, uint8_t d, uint8_t s)
 {
-    register uint32_t res;   /* all operands in native machine order */
+    uint32_t res;   /* all operands in native machine order */
 	RESET_FLAGS(emu);
 
     res = d & s;
@@ -1123,7 +1115,7 @@ Implements the TEST instruction and side effects.
 ****************************************************************************/
 void test16(x86emu_t *emu, uint16_t d, uint16_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
+	uint32_t res;   /* all operands in native machine order */
 	RESET_FLAGS(emu);
 
 	res = d & s;
@@ -1142,7 +1134,7 @@ Implements the TEST instruction and side effects.
 ****************************************************************************/
 void test32(x86emu_t *emu, uint32_t d, uint32_t s)
 {
-	register uint32_t res;   /* all operands in native machine order */
+	uint32_t res;   /* all operands in native machine order */
 	RESET_FLAGS(emu);
 
 	res = d & s;
@@ -1155,189 +1147,6 @@ void test32(x86emu_t *emu, uint32_t d, uint32_t s)
 	CLEAR_FLAG(F_CF);
 }
 
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-void imul8(x86emu_t *emu, uint8_t s)
-{
-	int16_t res = (int16_t)(int8_t)R_AL * (int8_t)s;
-	RESET_FLAGS(emu);
-
-	R_AX = res;
-	if (((R_AL & 0x80) == 0 && R_AH == 0x00) ||
-		((R_AL & 0x80) != 0 && R_AH == 0xFF)) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-void imul16_eax(x86emu_t *emu, uint16_t s)
-{
-	int32_t res = (int32_t)(int16_t)R_AX * (int16_t)s;
-	RESET_FLAGS(emu);
-
-	R_AX = (uint16_t)res;
-	R_DX = (uint16_t)(res >> 16);
-	if (((R_AX & 0x8000) == 0 && R_DX == 0x00) ||
-		((R_AX & 0x8000) != 0 && R_DX == 0xFFFF)) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-uint16_t imul16(x86emu_t *emu, uint16_t op1, uint16_t op2)
-{
-	int32_t res = (int32_t)(int16_t)op1 * (int16_t)op2;
-	RESET_FLAGS(emu);
-
-	op1 = (uint16_t)res;
-	uint16_t tmp = (uint16_t)(res >> 16);
-	if (((op1 & 0x8000) == 0 && tmp == 0x00) ||
-		((op1 & 0x8000) != 0 && tmp == 0xFFFF)) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-
-	return op1;
-}
-
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-void imul32_direct(uint32_t *res_lo, uint32_t* res_hi,uint32_t d, uint32_t s)
-{
-	int64_t res = (int64_t)(int32_t)d * (int32_t)s;
-
-	*res_lo = (uint32_t)res;
-	*res_hi = (uint32_t)(res >> 32);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-uint32_t imul32(x86emu_t *emu, uint32_t op1, uint32_t op2)
-{
-	uint32_t s;
-	uint32_t r;
-	RESET_FLAGS(emu);
-	imul32_direct(&r,&s,op1,op2);
-	if (((r & 0x80000000) == 0 && s == 0x00) ||
-		((r & 0x80000000) != 0 && s == 0xFFFFFFFF)) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(r & 0xff), F_PF);
-	return r;
-}
-
-/****************************************************************************
-REMARKS:
-Implements the IMUL instruction and side effects.
-****************************************************************************/
-void imul32_eax(x86emu_t *emu, uint32_t s)
-{
-	RESET_FLAGS(emu);
-	imul32_direct(&R_EAX,&R_EDX,R_EAX,s);
-	if (((R_EAX & 0x80000000) == 0 && R_EDX == 0x00) ||
-		((R_EAX & 0x80000000) != 0 && R_EDX == 0xFFFFFFFF)) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the MUL instruction and side effects.
-****************************************************************************/
-void mul8(x86emu_t *emu, uint8_t s)
-{
-	uint16_t res = (uint16_t)(R_AL) * s;
-	RESET_FLAGS(emu);
-
-	R_AX = res;
-	if (R_AH == 0) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-	}
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the MUL instruction and side effects.
-****************************************************************************/
-void mul16(x86emu_t *emu, uint16_t s)
-{
-	uint32_t res = (uint32_t)R_AX * s;
-	RESET_FLAGS(emu);
-
-	R_AX = (uint16_t)res;
-	R_DX = (uint16_t)(res >> 16);
-	if (R_DX == 0) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-    } else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-    }
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
-/****************************************************************************
-REMARKS:
-Implements the MUL instruction and side effects.
-****************************************************************************/
-void mul32_eax(x86emu_t *emu, uint32_t s)
-{
-	uint64_t res = (uint64_t)R_EAX * s;
-	RESET_FLAGS(emu);
-
-	R_EAX = (uint32_t)res;
-	R_EDX = (uint32_t)(res >> 32);
-
-	if (R_EDX == 0) {
-		CLEAR_FLAG(F_CF);
-		CLEAR_FLAG(F_OF);
-	} else {
-		SET_FLAG(F_CF);
-		SET_FLAG(F_OF);
-    }
-	CONDITIONAL_SET_FLAG(PARITY(R_AL & 0xff), F_PF);
-}
-
 /****************************************************************************
 REMARKS:
 Implements the IDIV instruction and side effects.

src/emu/x86primop.h

@@ -239,7 +239,7 @@ static inline uint32_t neg32(x86emu_t *emu, uint32_t s)
 
 /****************************************************************************
 REMARKS:
-Implements the NOT instruction and side effects.
+Implements the NOT instruction
 ****************************************************************************/
 static inline uint8_t not8(x86emu_t *emu, uint8_t s)
 {
@@ -248,7 +248,7 @@ static inline uint8_t not8(x86emu_t *emu, uint8_t s)
 
 /****************************************************************************
 REMARKS:
-Implements the NOT instruction and side effects.
+Implements the NOT instruction
 ****************************************************************************/
 static inline uint16_t not16(x86emu_t *emu, uint16_t s)
 {
@@ -257,7 +257,7 @@ static inline uint16_t not16(x86emu_t *emu, uint16_t s)
 
 /****************************************************************************
 REMARKS:
-Implements the NOT instruction and side effects.
+Implements the NOT instruction
 ****************************************************************************/
 static inline uint32_t not32(x86emu_t *emu, uint32_t s)
 {
@@ -503,15 +503,100 @@ static inline uint32_t xor32(x86emu_t *emu, uint32_t d, uint32_t s)
 	return emu->res;
 }
 
-void         imul8  (x86emu_t *emu, uint8_t s);
-void         imul16_eax (x86emu_t *emu, uint16_t s);
-uint16_t     imul16 (x86emu_t *emu, uint16_t op1, uint16_t op2);
-void         imul32_eax (x86emu_t *emu, uint32_t s);
-uint32_t     imul32(x86emu_t *emu, uint32_t op1, uint32_t op2);
-void 	     imul32_direct(uint32_t *res_lo, uint32_t* res_hi, uint32_t d, uint32_t s);
-void         mul8  (x86emu_t *emu, uint8_t s);
-void         mul16 (x86emu_t *emu, uint16_t s);
-void         mul32_eax (x86emu_t *emu, uint32_t s);
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static void imul8(x86emu_t *emu, uint8_t s)
+{
+	emu->df = d_imul8;
+	R_AX = emu->res = (int16_t)(int8_t)R_AL * (int8_t)s;
+}
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static void imul16_eax(x86emu_t *emu, uint16_t s)
+{
+	emu->df = d_imul16;
+	emu->res = (int32_t)(int16_t)R_AX * (int16_t)s;
+	R_AX = (uint16_t)emu->res;
+	R_DX = (uint16_t)(emu->res >> 16);
+}
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static uint16_t imul16(x86emu_t *emu, uint16_t op1, uint16_t op2)
+{
+	emu->df = d_imul16;
+	emu->res = (int32_t)(int16_t)op1 * (int16_t)op2;
+	return (uint16_t)emu->res;
+}
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static void imul32_direct(uint32_t *res_lo, uint32_t* res_hi,uint32_t d, uint32_t s)
+{
+	int64_t res = (int64_t)(int32_t)d * (int32_t)s;
+
+	*res_lo = (uint32_t)res;
+	*res_hi = (uint32_t)(res >> 32);
+}
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static uint32_t imul32(x86emu_t *emu, uint32_t op1, uint32_t op2)
+{
+	emu->df = d_imul32;
+	imul32_direct(&emu->res,&emu->op1,op1,op2);
+	return emu->res;
+}
+/****************************************************************************
+REMARKS:
+Implements the IMUL instruction and side effects.
+****************************************************************************/
+static void imul32_eax(x86emu_t *emu, uint32_t s)
+{
+	emu->df = d_imul32;
+	imul32_direct(&R_EAX,&R_EDX,R_EAX,s);
+	emu->res = R_EAX;
+	emu->op1 = R_EDX;
+}
+/****************************************************************************
+REMARKS:
+Implements the MUL instruction and side effects.
+****************************************************************************/
+static void mul8(x86emu_t *emu, uint8_t s)
+{
+	emu->df = d_mul8;
+	R_AX = emu->res = (uint16_t)(R_AL) * s;
+}
+/****************************************************************************
+REMARKS:
+Implements the MUL instruction and side effects.
+****************************************************************************/
+static void mul16(x86emu_t *emu, uint16_t s)
+{
+	emu->df = d_mul16;
+	emu->res = (uint32_t)R_AX * s;
+	R_AX = (uint16_t)emu->res;
+	R_DX = (uint16_t)(emu->res >> 16);
+}
+/****************************************************************************
+REMARKS:
+Implements the MUL instruction and side effects.
+****************************************************************************/
+static void mul32_eax(x86emu_t *emu, uint32_t s)
+{
+	emu->df = d_mul32;
+	uint64_t res = (uint64_t)R_EAX * s;
+	emu->res = R_EAX = (uint32_t)res;
+	emu->op1 = R_EDX = (uint32_t)(res >> 32);
+}
+
 void         idiv8  (x86emu_t *emu, uint8_t s);
 void         idiv16 (x86emu_t *emu, uint16_t s);
 void         idiv32 (x86emu_t *emu, uint32_t s);
@@ -519,4 +604,4 @@ void         div8  (x86emu_t *emu, uint8_t s);
 void         div16 (x86emu_t *emu, uint16_t s);
 void         div32 (x86emu_t *emu, uint32_t s);
 
-#endif //__X86PRIMOP_H_
+#endif //__X86PRIMOP_H_

src/emu/x86run66.c

@@ -0,0 +1,479 @@
+#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"
+
+
+#define F8      *(uint8_t*)(R_EIP++)
+#define F8S     *(int8_t*)(R_EIP++)
+#define F16     *(uint16_t*)(R_EIP+=2, R_EIP-2)
+#define F32     *(uint32_t*)(R_EIP+=4, R_EIP-4)
+#define F32S    *(int32_t*)(R_EIP+=4, R_EIP-4)
+#define PK(a)   *(uint8_t*)(R_EIP+a)
+
+#include "modrm.h"
+
+void Run6766(x86emu_t *emu)
+{
+    uint8_t opcode = Fetch8(emu);
+    uint8_t nextop;
+    reg32_t *op1, *op2, *op3, *op4;
+    reg32_t ea1, ea2, ea3, ea4;
+    uint8_t tmp8u;
+    int8_t tmp8s;
+    uint16_t tmp16u, tmp16u2;
+    int16_t tmp16s;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    switch(opcode) {
+
+    case 0x8D:                              /* LEA Gw,Ew */
+        nextop = Fetch8(emu);
+        op1=GetEw16(emu, nextop);
+        op2=GetG(emu, nextop);
+        op2->word[0] = (uint16_t)(uintptr_t)op1;
+        break;
+    
+                
+    default:
+        UnimpOpcode(emu);
+    }
+}
+
+void Run67(x86emu_t *emu)
+{
+    uint8_t opcode = Fetch8(emu);
+    uint8_t nextop;
+    reg32_t *op1, *op2, *op3, *op4;
+    reg32_t ea1, ea2, ea3, ea4;
+    uint8_t tmp8u;
+    int8_t tmp8s;
+    uint16_t tmp16u;
+    int16_t tmp16s;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    switch(opcode) {
+
+    case 0x66:                      /* MoooRE opcodes */
+        Run6766(emu);
+        break;
+
+    case 0xE0:                      /* LOOPNZ */
+        CHECK_FLAGS(emu);
+        tmp8s = Fetch8s(emu);
+        --R_CX; // don't update flags
+        if(R_CX && !ACCESS_FLAG(F_ZF))
+            R_EIP += tmp8s;
+        break;
+    case 0xE1:                      /* LOOPZ */
+        CHECK_FLAGS(emu);
+        tmp8s = Fetch8s(emu);
+        --R_CX; // don't update flags
+        if(R_CX && ACCESS_FLAG(F_ZF))
+            R_EIP += tmp8s;
+        break;
+    case 0xE2:                      /* LOOP */
+        tmp8s = Fetch8s(emu);
+        --R_CX; // don't update flags
+        if(R_CX)
+            R_EIP += tmp8s;
+        break;
+    case 0xE3:                      /* JCXZ */
+        tmp8s = Fetch8s(emu);
+        if(!R_CX)
+            R_EIP += tmp8s;
+        break;
+
+    default:
+        UnimpOpcode(emu);
+    }
+}
+
+void RunLock(x86emu_t *emu)
+{
+    uint8_t opcode = Fetch8(emu);
+    uint8_t nextop;
+    reg32_t *oped;
+    reg32_t ea1;
+    sse_regs_t *opex, eax1;
+    mmx_regs_t *opem;
+    uint8_t tmp8u;
+    int8_t tmp8s;
+    uint16_t tmp16u;
+    int16_t tmp16s;
+    uint32_t tmp32u, tmp32u2;
+    int32_t tmp32s;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    pthread_mutex_lock(&emu->context->mutex_lock);
+    switch(opcode) {
+        #define GO(B, OP)                      \
+        case B+0: \
+            nextop = F8;               \
+            GET_EB;             \
+            EB->byte[0] = OP##8(emu, EB->byte[0], GB);  \
+            break;                              \
+        case B+1: \
+            nextop = F8;               \
+            GET_ED;             \
+            ED->dword[0] = OP##32(emu, ED->dword[0], GD.dword[0]); \
+            break;                              \
+        case B+2: \
+            nextop = F8;               \
+            GET_EB;                   \
+            GB = OP##8(emu, GB, EB->byte[0]); \
+            break;                              \
+        case B+3: \
+            nextop = F8;               \
+            GET_ED;         \
+            GD.dword[0] = OP##32(emu, GD.dword[0], ED->dword[0]); \
+            break;                              \
+        case B+4: \
+            R_AL = OP##8(emu, R_AL, F8); \
+            break;                              \
+        case B+5: \
+            R_EAX = OP##32(emu, R_EAX, F32); \
+            break;
+
+        GO(0x00, add)                   /* ADD 0x00 -> 0x05 */
+        GO(0x08, or)                    /*  OR 0x08 -> 0x0D */
+        GO(0x10, adc)                   /* ADC 0x10 -> 0x15 */
+        GO(0x18, sbb)                   /* SBB 0x18 -> 0x1D */
+        GO(0x20, and)                   /* AND 0x20 -> 0x25 */
+        GO(0x28, sub)                   /* SUB 0x28 -> 0x2D */
+        GO(0x30, xor)                   /* XOR 0x30 -> 0x35 */
+        #undef GO
+
+        case 0x0f:
+            opcode = F8;
+            switch (opcode) { 
+                case 0xB0:                      /* CMPXCHG Eb,Gb */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_EB;
+                    cmp8(emu, R_AL, EB->byte[0]);
+                    if(ACCESS_FLAG(F_ZF)) {
+                        EB->byte[0] = GB;
+                    } else {
+                        R_AL = EB->byte[0];
+                    }
+                    break;
+                case 0xB1:                      /* CMPXCHG Ed,Gd */
+                    nextop = F8;
+                    GET_ED;
+                    cmp32(emu, R_EAX, ED->dword[0]);
+                    if(ACCESS_FLAG(F_ZF)) {
+                        ED->dword[0] = GD.dword[0];
+                    } else {
+                        R_EAX = ED->dword[0];
+                    }
+                    break;
+                case 0xB3:                      /* BTR Ed,Gd */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_ED;
+                    tmp8u = GD.byte[0];
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                    }
+                    tmp8u&=31;
+                    if(ED->dword[0] & (1<<tmp8u)) {
+                        SET_FLAG(F_CF);
+                        ED->dword[0] ^= (1<<tmp8u);
+                    } else
+                        CLEAR_FLAG(F_CF);
+                    break;
+                case 0xBA:                      
+                    nextop = F8;
+                    switch((nextop>>3)&7) {
+                        case 4:                 /* BT Ed,Ib */
+                            CHECK_FLAGS(emu);
+                            GET_ED;
+                            tmp8u = F8;
+                            if((nextop&0xC0)!=0xC0)
+                            {
+                                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                            }
+                            tmp8u&=31;
+                            if(ED->dword[0] & (1<<tmp8u))
+                                SET_FLAG(F_CF);
+                            else
+                                CLEAR_FLAG(F_CF);
+                            break;
+                        case 6:             /* BTR Ed, Ib */
+                            CHECK_FLAGS(emu);
+                            GET_ED;
+                            tmp8u = F8;
+                            if((nextop&0xC0)!=0xC0)
+                            {
+                                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                            }
+                            tmp8u&=31;
+                            if(ED->dword[0] & (1<<tmp8u)) {
+                                SET_FLAG(F_CF);
+                                ED->dword[0] ^= (1<<tmp8u);
+                            } else
+                                CLEAR_FLAG(F_CF);
+                            break;
+
+                        default:
+                            R_EIP -= 3; //unfetch
+                            break;
+                    }
+                    break;
+                case 0xBB:                      /* BTC Ed,Gd */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_ED;
+                    tmp8u = GD.byte[0];
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                    }
+                    tmp8u&=31;
+                    if(ED->dword[0] & (1<<tmp8u))
+                        SET_FLAG(F_CF);
+                    else
+                        CLEAR_FLAG(F_CF);
+                    ED->dword[0] ^= (1<<tmp8u);
+                    break;
+                case 0xC0:                      /* XADD Gb,Eb */
+                    nextop = F8;
+                    GET_EB;
+                    tmp8u = add8(emu, EB->byte[0], GB);
+                    GB = EB->byte[0];
+                    EB->byte[0] = tmp8u;
+                    break;
+                case 0xC1:                      /* XADD Gd,Ed */
+                    nextop = F8;
+                    GET_ED;
+                    tmp32u = add32(emu, ED->dword[0], GD.dword[0]);
+                    GD.dword[0] = ED->dword[0];
+                    ED->dword[0] = tmp32u;
+                    break;
+                case 0xC7:                      /* CMPXCHG8B Gq */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_ED;
+                    tmp32u = ED->dword[0];
+                    tmp32u2= ED->dword[1];
+                    if(R_EAX == tmp32u && R_EDX == tmp32u2) {
+                        SET_FLAG(F_ZF);
+                        ED->dword[0] = R_EBX;
+                        ED->dword[1] = R_ECX;
+                    } else {
+                        CLEAR_FLAG(F_ZF);
+                        R_EAX = tmp32u;
+                        R_EDX = tmp32u2;
+                    }
+                    break;
+                default:
+                    // trigger invalid lock?
+                    R_EIP -= 2; // unfetch
+                    break;
+            }
+            break;
+
+        case 0x81:              /* GRP Ed,Id */
+        case 0x83:              /* GRP Ed,Ib */
+            nextop = F8;
+            GET_ED;
+            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 0x86:                      /* XCHG Eb,Gb */
+            nextop = F8;
+            GET_EB;
+            tmp8u = GB;
+            GB = EB->byte[0];
+            EB->byte[0] = tmp8u;
+            break;
+        case 0x87:                      /* XCHG Ed,Gd */
+            nextop = F8;
+            GET_ED;
+            tmp32u = GD.dword[0];
+            GD.dword[0] = ED->dword[0];
+            ED->dword[0] = tmp32u;
+            break;
+        case 0xFF:              /* GRP 5 Ed */
+            nextop = F8;
+            GET_ED;
+            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;
+                default:
+                    printf_log(LOG_NONE, "Illegal Opcode 0xF0 0x%02X 0x%02X\n", opcode, PK(0));
+                    emu->quit=1;
+                    emu->error |= ERR_ILLEGAL;
+                    break;
+            }
+            break;
+        default:
+            //UnimpOpcode(emu);
+            // should trigger invalid unlock ?
+            R_EIP--;    // "unfetch" to use normal instruction
+    }
+    pthread_mutex_unlock(&emu->context->mutex_lock);
+}
+
+void RunGS(x86emu_t *emu)
+{
+    uint8_t opcode = Fetch8(emu);
+    uint8_t nextop;
+    reg32_t *oped;
+    reg32_t ea1;
+    sse_regs_t *opex, eax1;
+    mmx_regs_t *opem;
+    uint8_t tmp8u;
+    int8_t tmp8s;
+    uint16_t tmp16u;
+    int16_t tmp16s;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    uintptr_t tlsdata = GetGSBaseEmu(emu);
+    switch(opcode) {
+        case 0x33:              /* XOR Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = xor32(emu, GD.dword[0], ED->dword[0]);
+            break;
+        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 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 0x89:              /* MOV Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = GD.dword[0];
+            break;
+        case 0x8B:              /* MOV Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = ED->dword[0];
+            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 0xC7:              /* MOV Ed,Id */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = F32;
+            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 */
+                    Push(emu, R_EIP);
+                    R_EIP = ED->dword[0];  // should get value in temp var. in case ED use ESP?
+                    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;
+                    } else {
+                        Push16(emu, R_CS);
+                        Push(emu, R_EIP);
+                        R_EIP = ED->dword[0];
+                        R_CS = (ED+1)->word[0];
+                    }
+                    break;
+                case 4:                 /* JMP NEAR Ed */
+                    R_EIP = 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;
+                    } else {
+                        R_EIP = 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;
+            }
+            break;
+        default:
+            UnimpOpcode(emu);
+    }
+}

src/emu/x86run_private.c

@@ -4,61 +4,66 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#ifdef HAVE_TRACE
+#include <unistd.h>
+#include <sys/syscall.h>
+#endif
 
 #include "debug.h"
-#include "stack.h"
+#include "box86stack.h"
 #include "x86emu.h"
 #include "x86run.h"
 #include "x86run_private.h"
 #include "x86emu_private.h"
 #include "box86context.h"
 #include "x86run.h"
+#include "librarian.h"
+#include "elfloader.h"
+#ifdef HAVE_TRACE
+#include "x86trace.h"
+#endif
 
-static uint32_t x86emu_parity_tab[8] =
-{
-	0x96696996,
-	0x69969669,
-	0x69969669,
-	0x96696996,
-	0x69969669,
-	0x96696996,
-	0x96696996,
-	0x69969669,
-};
-
-#define PARITY(x)   (((x86emu_parity_tab[(x) / 32] >> ((x) % 32)) & 1) == 0)
+#define PARITY(x)   (((emu->x86emu_parity_tab[(x) / 32] >> ((x) % 32)) & 1) == 0)
 #define XOR2(x) 	(((x) ^ ((x)>>1)) & 0x1)
 
 int32_t EXPORT my___libc_start_main(x86emu_t* emu, int *(main) (int, char * *, char * *), int argc, char * * ubp_av, void (*init) (void), void (*fini) (void), void (*rtld_fini) (void), void (* stack_end))
 {
     //TODO: register rtld_fini
     //TODO: register fini
+    // let's cheat and set all args...
+    Push(emu, (uint32_t)emu->context->envv);
+    Push(emu, (uint32_t)emu->context->argv);
+    Push(emu, (uint32_t)emu->context->argc);
     if(init) {
         PushExit(emu);
         R_EIP=(uint32_t)*init;
         printf_log(LOG_DEBUG, "Calling init(%p) from __libc_start_main\n", *init);
-        Run(emu);
+        DynaRun(emu);
         if(emu->error)  // any error, don't bother with more
             return 0;
         emu->quit = 0;
     }
-    // let's cheat and set all args...
+    printf_log(LOG_DEBUG, "Transfert to main(%d, %p, %p)=>%p from __libc_start_main\n", emu->context->argc, emu->context->argv, emu->context->envv, main);
     // call main and finish
-    Push(emu, (uint32_t)emu->context->envv);
-    Push(emu, (uint32_t)emu->context->argv);
-    Push(emu, (uint32_t)emu->context->argc);
     PushExit(emu);
     R_EIP=(uint32_t)main;
-    printf_log(LOG_DEBUG, "Transfert to main(%d, %p, %p)=>%p from __libc_start_main\n", emu->context->argc, emu->context->argv, emu->context->envv, main);
+#ifdef DYNAREC
+    DynaRun(emu);
+#endif
     return 0;
 }
 
-const char* GetNativeName(void* p)
+const char* GetNativeName(x86emu_t* emu, void* p)
 {
+    static char unknown[10] = "???";
+    const char *ret = GetNameOffset(emu->context->maplib, p);
+    if(ret)
+        return ret;
+
     static char buff[500] = {0};
     Dl_info info;
     if(dladdr(p, &info)==0)
-        strcpy(buff, "???");
+        return unknown;
     else {
         if(info.dli_sname) {
             strcpy(buff, info.dli_sname);
@@ -66,7 +71,7 @@ const char* GetNativeName(void* p)
                 strcat(buff, " ("); strcat(buff, info.dli_fname); strcat(buff, ")");
             }
         } else
-            strcpy(buff, "unknown");
+            return unknown;
     }
     return buff;
 }
@@ -74,7 +79,7 @@ const char* GetNativeName(void* p)
 uintptr_t pltResolver = ~0;
 void PltResolver(x86emu_t* emu, uint32_t id, uintptr_t ofs)
 {
-    printf_log(LOG_INFO, "PltResolver: Ofs=%p, Id=%d (IP=%p)", (void*)ofs, id, *(void**)(R_ESP));
+    printf_log(LOG_INFO, "PltResolver: Ofs=%p, Id=%d (IP=%p, ESP+4=%p)", (void*)ofs, id, *(void**)(R_ESP), *(void**)(R_ESP+4));
     emu->quit=1;
 }
 
@@ -111,7 +116,7 @@ void UpdateFlags(x86emu_t *emu)
             lo = (emu->op2 & 0xFFFF) + (emu->op1 & 0xFFFF);
             hi = (lo >> 16) + (emu->op2 >> 16) + (emu->op1 >> 16);
             CONDITIONAL_SET_FLAG(hi & 0x10000, F_CF);
-            CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+            CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
             CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             cc = (emu->op1 & emu->op2) | ((~emu->res) & (emu->op1 | emu->op2));
@@ -124,7 +129,7 @@ void UpdateFlags(x86emu_t *emu)
             CLEAR_FLAG(F_AF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80, F_SF);
             CONDITIONAL_SET_FLAG(emu->res == 0, F_ZF);
-            CONDITIONAL_SET_FLAG(PARITY(emu->res), F_PF);
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             break;
         case d_and16:
             CLEAR_FLAG(F_OF);
@@ -160,7 +165,7 @@ void UpdateFlags(x86emu_t *emu)
             break;
         case d_dec32:
             CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
-            CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+            CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
             CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             bc = (emu->res & (~emu->op1 | 1)) | (~emu->op1 & 1);
             CONDITIONAL_SET_FLAG(XOR2(bc >> 30), F_OF);
@@ -183,20 +188,91 @@ void UpdateFlags(x86emu_t *emu)
             CONDITIONAL_SET_FLAG(cc & 0x8, F_AF);
             break;
         case d_inc32:
-            CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+            CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
             CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             cc = (1 & emu->op1) | ((~emu->res) & (1 | emu->op1));
             CONDITIONAL_SET_FLAG(XOR2(cc >> 30), F_OF);
             CONDITIONAL_SET_FLAG(cc & 0x8, F_AF);
             break;
+        case d_imul8:
+            lo = emu->res & 0xff;
+            hi = (emu->res>>8)&0xff;
+            if (((lo & 0x80) == 0 && hi == 0x00) ||
+                ((lo & 0x80) != 0 && hi == 0xFF)) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(lo & 0xff), F_PF);
+            break;
+        case d_imul16:
+            lo = (uint16_t)emu->res;
+            hi = (uint16_t)(emu->res >> 16);
+            if (((lo & 0x8000) == 0 && hi == 0x00) ||
+                ((lo & 0x8000) != 0 && hi == 0xFFFF)) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(lo & 0xff), F_PF);
+            break;
+        case d_imul32:
+            if (((emu->res & 0x80000000) == 0 && emu->op1 == 0x00) ||
+                ((emu->res & 0x80000000) != 0 && emu->op1 == 0xFFFFFFFF)) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+            break;
+        case d_mul8:
+            lo = emu->res & 0xff;
+            hi = (emu->res>>8)&0xff;
+            if (hi == 0) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(lo & 0xff), F_PF);
+            break;
+        case d_mul16:
+            lo = (uint16_t)emu->res;
+            hi = (uint16_t)(emu->res >> 16);
+            if (hi == 0) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(lo & 0xff), F_PF);
+            break;
+        case d_mul32:
+            if (emu->op1 == 0) {
+                CLEAR_FLAG(F_CF);
+                CLEAR_FLAG(F_OF);
+            } else {
+                SET_FLAG(F_CF);
+                SET_FLAG(F_OF);
+            }
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
+            break;
         case d_or8:
             CLEAR_FLAG(F_OF);
             CLEAR_FLAG(F_CF);
             CLEAR_FLAG(F_AF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80, F_SF);
             CONDITIONAL_SET_FLAG(emu->res == 0, F_ZF);
-            CONDITIONAL_SET_FLAG(PARITY(emu->res), F_PF);
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             break;
         case d_or16:
             CLEAR_FLAG(F_OF);
@@ -218,7 +294,7 @@ void UpdateFlags(x86emu_t *emu)
             CONDITIONAL_SET_FLAG(emu->op1 != 0, F_CF);
             CONDITIONAL_SET_FLAG((emu->res & 0xff) == 0, F_ZF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80, F_SF);
-            CONDITIONAL_SET_FLAG(PARITY(emu->res), F_PF);
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             bc = emu->res | emu->op1;
             CONDITIONAL_SET_FLAG(XOR2(bc >> 6), F_OF);
             CONDITIONAL_SET_FLAG(bc & 0x8, F_AF);
@@ -234,7 +310,7 @@ void UpdateFlags(x86emu_t *emu)
             break;
         case d_neg32:
             CONDITIONAL_SET_FLAG(emu->op1 != 0, F_CF);
-            CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+            CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
             CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             bc = emu->res | emu->op1;
@@ -291,7 +367,7 @@ void UpdateFlags(x86emu_t *emu)
             if (emu->op2 > 0) {
                 cc = emu->op1 & (1 << (32 - emu->op2));
                 CONDITIONAL_SET_FLAG(cc, F_CF);
-                CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+                CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
                 CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
                 CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             }
@@ -402,7 +478,7 @@ void UpdateFlags(x86emu_t *emu)
             if (cnt > 0) {
                 cc = emu->op1 & (1 << (cnt - 1));
                 CONDITIONAL_SET_FLAG(cc, F_CF);
-                CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+                CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
                 CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
                 CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             }
@@ -430,7 +506,7 @@ void UpdateFlags(x86emu_t *emu)
             break;
         case d_sub32:
             CONDITIONAL_SET_FLAG(emu->res & 0x80000000, F_SF);
-            CONDITIONAL_SET_FLAG((emu->res & 0xffffffff) == 0, F_ZF);
+            CONDITIONAL_SET_FLAG(!emu->res, F_ZF);
             CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             bc = (emu->res & (~emu->op1 | emu->op2)) | (~emu->op1 & emu->op2);
             CONDITIONAL_SET_FLAG(bc & 0x80000000, F_CF);
@@ -441,7 +517,7 @@ void UpdateFlags(x86emu_t *emu)
             CLEAR_FLAG(F_OF);
             CONDITIONAL_SET_FLAG(emu->res & 0x80, F_SF);
             CONDITIONAL_SET_FLAG(emu->res == 0, F_ZF);
-            CONDITIONAL_SET_FLAG(PARITY(emu->res), F_PF);
+            CONDITIONAL_SET_FLAG(PARITY(emu->res & 0xff), F_PF);
             CLEAR_FLAG(F_CF);
             CLEAR_FLAG(F_AF);
             break;
@@ -461,6 +537,9 @@ void UpdateFlags(x86emu_t *emu)
             CLEAR_FLAG(F_CF);
             CLEAR_FLAG(F_AF);
             break;
+        case d_unknown:
+            printf_log(LOG_NONE, "Box86: %p trying to evaluate Unknown defered Flags\n", (void*)R_EIP);
+            break;
     }
     RESET_FLAGS(emu);
 }
@@ -518,3 +597,80 @@ void UnpackFlags(x86emu_t* emu)
     GO(ID);
     #undef GO
 }
+
+uintptr_t GetGSBaseEmu(x86emu_t* emu)
+{
+    return (uintptr_t)GetGSBase(emu->context);
+}
+
+#ifdef HAVE_TRACE
+extern uint64_t start_cnt;
+#define PK(a)   *(uint8_t*)(ip+a)
+#define PK32(a)   *(uint32_t*)(uint8_t*)(ip+a)
+
+static void printFunctionAddr(x86emu_t* emu, uintptr_t nextaddr, const char* text)
+{
+    uint32_t sz = 0;
+    uintptr_t start = 0;
+    const char* symbname = FindNearestSymbolName(FindElfAddress(emu->context, nextaddr), (void*)nextaddr, &start, &sz);
+    if(symbname && nextaddr>=start && nextaddr<(start+sz)) {
+        if(nextaddr==start)
+            printf_log(LOG_NONE, " (%s%s)", text, symbname);
+        else
+            printf_log(LOG_NONE, " (%s%s + 0x%x)", text, symbname, nextaddr - start);
+    }
+}
+
+void PrintTrace(x86emu_t* emu, uintptr_t ip, int dynarec)
+{
+    if(start_cnt) --start_cnt;
+    if(!start_cnt && emu->dec && (
+            (emu->trace_end == 0) 
+            || ((ip >= emu->trace_start) && (ip < emu->trace_end))) ) {
+        pthread_mutex_lock(&emu->context->mutex_trace);
+        int tid = syscall(SYS_gettid);
+#ifdef DYNAREC
+        if((emu->context->trace_tid != tid) || (emu->context->trace_dynarec!=dynarec)) {
+            printf_log(LOG_NONE, "Thread %04d| (%s) |\n", tid, dynarec?"dyn":"int");
+            emu->context->trace_tid = tid;
+            emu->context->trace_dynarec = dynarec;
+        }
+#else
+        if(emu->context->trace_tid != tid) {
+            printf_log(LOG_NONE, "Thread %04d|\n", tid);
+            emu->context->trace_tid = tid;
+        }
+#endif
+        printf_log(LOG_NONE, "%s", DumpCPURegs(emu, ip));
+        if(PK(0)==0xcc && PK(1)=='S' && PK(2)=='C') {
+            uint32_t a = *(uint32_t*)(ip+3);
+            if(a==0) {
+                printf_log(LOG_NONE, "0x%p: Exit x86emu\n", (void*)ip);
+            } else {
+                printf_log(LOG_NONE, "0x%p: Native call to %p => %s\n", (void*)ip, (void*)a, GetNativeName(emu, *(void**)(ip+7)));
+            }
+        } else {
+            printf_log(LOG_NONE, "%s", DecodeX86Trace(emu->dec, ip));
+            uint8_t peek = PK(0);
+            if(peek==0xC3 || peek==0xC2) {
+                printf_log(LOG_NONE, " => %p", *(void**)(R_ESP));
+                printFunctionAddr(emu, *(uintptr_t*)(R_ESP), "=> ");
+            } else if(peek==0x55) {
+                printf_log(LOG_NONE, " => STACK_TOP: %p", *(void**)(R_ESP));
+                printFunctionAddr(emu, ip, "here: ");
+            } else if(peek==0xE8) { // Call
+                uintptr_t nextaddr = ip + PK32(1);
+                printFunctionAddr(emu, nextaddr, "=> ");
+            } else if(peek==0xFF) {
+                if(PK(1)==0x25) {
+                    uintptr_t nextaddr = ip + PK32(2);
+                    printFunctionAddr(emu, nextaddr, "=> ");
+                }
+            }
+            printf_log(LOG_NONE, "\n");
+        }
+        pthread_mutex_unlock(&emu->context->mutex_trace);
+    }
+}
+
+#endif

src/emu/x86run_private.h

@@ -6,42 +6,42 @@
 #include "x86emu_private.h"
 typedef struct x86emu_s x86emu_t;
 
-inline uint8_t Fetch8(x86emu_t *emu) {return *(uint8_t*)(R_EIP++);}
-inline int8_t Fetch8s(x86emu_t *emu) {return *(int8_t*)(R_EIP++);}
-inline uint16_t Fetch16(x86emu_t *emu)
+static inline uint8_t Fetch8(x86emu_t *emu) {return *(uint8_t*)(R_EIP++);}
+static inline int8_t Fetch8s(x86emu_t *emu) {return *(int8_t*)(R_EIP++);}
+static inline uint16_t Fetch16(x86emu_t *emu)
 {
     uint16_t val = *(uint16_t*)R_EIP;
     R_EIP+=2;
     return val;
 }
-inline int16_t Fetch16s(x86emu_t *emu)
+static inline int16_t Fetch16s(x86emu_t *emu)
 {
     int16_t val = *(int16_t*)R_EIP;
     R_EIP+=2;
     return val;
 }
-inline uint32_t Fetch32(x86emu_t *emu)
+static inline uint32_t Fetch32(x86emu_t *emu)
 {
     uint32_t val = *(uint32_t*)R_EIP;
     R_EIP+=4;
     return val;
 }
-inline int32_t Fetch32s(x86emu_t *emu)
+static inline int32_t Fetch32s(x86emu_t *emu)
 {
     int32_t val = *(int32_t*)R_EIP;
     R_EIP+=4;
     return val;
 }
-inline uint8_t Peek(x86emu_t *emu, int offset){return *(uint8_t*)(R_EIP + offset);}
+static inline uint8_t Peek(x86emu_t *emu, int offset){return *(uint8_t*)(R_EIP + offset);}
 
-inline uint32_t Pop(x86emu_t *emu)
+static inline uint32_t Pop(x86emu_t *emu)
 {
     uint32_t* st = ((uint32_t*)(R_ESP));
     R_ESP += 4;
     return *st;
 }
 
-inline void Push(x86emu_t *emu, uint32_t v)
+static inline void Push(x86emu_t *emu, uint32_t v)
 {
     R_ESP -= 4;
     *((uint32_t*)R_ESP) = v;
@@ -172,15 +172,22 @@ 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 RunLock(x86emu_t *emu);
 
 void x86Syscall(x86emu_t *emu);
 void x86Int3(x86emu_t* emu);
-x86emu_t* x86emu_fork(x86emu_t* e);
+x86emu_t* x86emu_fork(x86emu_t* e, int forktype);
 
 void PackFlags(x86emu_t* emu);
 void UnpackFlags(x86emu_t* emu);
 
-const char* GetNativeName(void* p);
+uintptr_t GetGSBaseEmu(x86emu_t* emu);
 
+const char* GetNativeName(x86emu_t* emu, void* p);
+
+#ifdef HAVE_TRACE
+void PrintTrace(x86emu_t* emu, uintptr_t ip, int dynarec);
+#endif
 
 #endif //__X86RUN_PRIVATE_H_

src/emu/x86syscall.c

@@ -0,0 +1,581 @@
+#define _GNU_SOURCE         /* See feature_test_macros(7) */
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <sys/syscall.h>   /* For SYS_xxx definitions */
+#include <unistd.h>
+#include <time.h>
+#include <sys/mman.h>
+#include <sys/select.h>
+#include <sys/types.h>
+#include <sys/ioctl.h>
+#include <asm/stat.h>
+#include <errno.h>
+#include <sched.h>
+#include <sys/wait.h>
+#include <sys/utsname.h>
+#ifndef __NR_socketcall
+#include <linux/net.h>
+#include <sys/socket.h>
+#endif
+#include <sys/resource.h>
+#include <poll.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 "myalign.h"
+#include "box86context.h"
+#include "callback.h"
+#include "signals.h"
+
+#ifndef __NR_socketcall
+//#ifndef SYS_RECVMMSG
+//#define SYS_RECVMMSG    19
+//#endif
+//#ifndef SYS_SENDMMSG
+//#define SYS_SENDMMSG    20
+//#endif
+int32_t my_accept4(x86emu_t* emu, int32_t fd, void* a, void* l, int32_t flags); // not always present, so used wrapped version
+#endif
+
+
+int32_t my_getrandom(x86emu_t* emu, void* buf, uint32_t buflen, uint32_t flags);
+int of_convert(int flag);
+int32_t my_open(x86emu_t* emu, void* pathname, int32_t flags, uint32_t mode);
+
+// cannot include <fcntl.h>, it conflict with some asm includes...
+#ifndef O_NONBLOCK
+#define O_NONBLOCK 04000
+#endif
+#undef fcntl
+int fcntl(int fd, int cmd, ... /* arg */ );
+
+// Syscall table for x86 can be found here: http://shell-storm.org/shellcode/files/syscalls.html
+typedef struct scwrap_s {
+    int x86s;
+    int nats;
+    int nbpars;
+} scwrap_t;
+
+scwrap_t syscallwrap[] = {
+    { 2, __NR_fork, 1 },    // should wrap this one, because of the struct pt_regs (the only arg)?
+    //{ 3, __NR_read, 3 },  // wrapped so SA_RESTART can be handled by libc
+    //{ 4, __NR_write, 3 }, // same
+    //{ 5, __NR_open, 3 },  // flags need transformation
+    //{ 6, __NR_close, 1 },   // wrapped so SA_RESTART can be handled by libc
+#ifdef __NR_waitpid
+    { 7, __NR_waitpid, 3 },
+#endif
+    { 10, __NR_unlink, 1 },
+#ifdef __NR_time
+    { 13, __NR_time, 1 },
+#endif
+    { 15, __NR_chmod, 2 },
+    { 19, __NR_lseek, 3 },
+    { 20, __NR_getpid, 0 },
+    { 24, __NR_getuid, 0 },
+    { 33, __NR_access, 2 },
+    { 37, __NR_kill, 2 },
+    { 38, __NR_rename, 2 },
+    { 39, __NR_mkdir, 2 },
+    { 40, __NR_rmdir, 1 },
+    { 41, __NR_dup, 1 },
+    { 42, __NR_pipe, 1 },
+    { 45, __NR_brk, 1 },
+    { 47, __NR_getgid, 0 },
+    { 49, __NR_geteuid, 0 },
+    { 50, __NR_getegid, 0 },
+    //{ 54, __NR_ioctl, 5 },    // wrapped to allow SA_RESTART handling by libc
+    //{ 55, __NR_fcntl, 3 },    // wrapped to allow filter of F_SETFD
+    { 60, __NR_umask, 1 },
+    { 63, __NR_dup2, 2 },
+    { 64, __NR_getppid, 0 },
+    { 75, __NR_setrlimit, 2 },
+#ifdef __NR_getrlimit
+    { 76, __NR_getrlimit, 2 },
+#endif
+    { 77, __NR_getrusage, 2 },
+    { 78, __NR_gettimeofday, 2 },
+    { 83, __NR_symlink, 2 },
+#ifdef __NR_select
+    { 82, __NR_select, 5 },
+#endif
+    { 85, __NR_readlink, 3 },
+    { 91, __NR_munmap, 2 },
+    { 94, __NR_fchmod, 2 },
+    { 99, __NR_statfs, 2 },
+#ifdef __NR_socketcall
+    { 102, __NR_socketcall, 2 },
+#endif
+    { 104, __NR_setitimer, 3 },
+    { 105, __NR_getitimer, 2 },
+#ifdef __NR_newstat
+    { 106, __NR_newstat, 2 },
+#else
+    { 106, __NR_stat, 2 },
+#endif
+#ifdef __NR_newlstat
+    { 107, __NR_newlstat, 2 },
+#else
+    { 107, __NR_lstat, 2 },
+#endif
+#ifdef __NR_newfstat
+    { 108, __NR_newfstat, 2 },
+#else
+    { 108, __NR_fstat, 2 },
+#endif
+#ifdef __NR_olduname
+    { 109, __NR_olduname, 1 },
+#endif
+#ifdef __NR_iopl
+    { 110, __NR_iopl, 1 },
+#endif
+    { 114, __NR_wait4, 4 }, //TODO: check struct rusage alignment
+#ifdef __NR_ipc
+    { 117, __NR_ipc, 6 },
+#endif
+    //{ 119, __NR_sigreturn, 0},
+    //{ 120, __NR_clone, 5 },    // need works
+    //{ 122, __NR_uname, 1 },
+    { 125, __NR_mprotect, 3 },
+    { 136, __NR_personality, 1 },
+    { 140, __NR__llseek, 5 },
+    { 141, __NR_getdents, 3 },
+    { 142, __NR__newselect, 5 },
+    { 143, __NR_flock,  2 },
+    { 144, __NR_msync, 3 },
+    { 145, __NR_readv, 3 },
+    { 146, __NR_writev, 3 },
+    { 148, __NR_fdatasync, 1 },
+    { 149, __NR__sysctl, 1 },    // need wrapping?
+    { 156, __NR_sched_setscheduler, 3 },
+    { 157, __NR_sched_getscheduler, 1 },
+    { 158, __NR_sched_yield, 0 },
+    { 162, __NR_nanosleep, 2 },
+    { 164, __NR_setresuid, 3 },
+    //{ 168, __NR_poll, 3 },    // wrapped to allow SA_RESTART wrapping by libc
+    { 172, __NR_prctl, 5 },
+    //{ 173, __NR_rt_sigreturn, 0 },
+    { 175, __NR_rt_sigprocmask, 4 },
+    { 179, __NR_rt_sigsuspend, 2 },
+    { 183, __NR_getcwd, 2 },
+    { 186, __NR_sigaltstack, 2 },    // neeed wrap or something?
+    { 191, __NR_ugetrlimit, 2 },
+    { 192, __NR_mmap2, 6},
+    //{ 195, __NR_stat64, 2 },  // need proprer wrap because of structure size change
+    //{ 197, __NR_fstat64, 2 },  // need proprer wrap because of structure size change
+    { 199, __NR_getuid32, 0 },
+    { 200, __NR_getgid32, 0 },
+    { 201, __NR_geteuid32, 0 },
+    { 202, __NR_getegid32, 0 },
+    { 208, __NR_setresuid32, 3 },
+    { 220, __NR_getdents64, 3 },
+    //{ 221, __NR_fcntl64, 3 },
+    { 224, __NR_gettid, 0 },
+    { 240, __NR_futex, 6 },
+    { 252, __NR_exit_group, 1 },
+    { 265, __NR_clock_gettime, 2 },
+    //{ 270, __NR_tgkill, 3 },
+    { 311, __NR_set_robust_list, 2 },
+    { 312, __NR_get_robust_list, 4 },
+};
+
+struct mmap_arg_struct {
+    unsigned long addr;
+    unsigned long len;
+    unsigned long prot;
+    unsigned long flags;
+    unsigned long fd;
+    unsigned long offset;
+};
+
+#undef st_atime
+#undef st_ctime
+#undef st_mtime
+
+struct x86_pt_regs {
+	long ebx;
+	long ecx;
+	long edx;
+	long esi;
+	long edi;
+	long ebp;
+	long eax;
+	int  xds;
+	int  xes;
+	int  xfs;
+	int  xgs;
+	long orig_eax;
+	long eip;
+	int  xcs;
+	long eflags;
+	long esp;
+	int  xss;
+};
+
+#ifndef __NR_olduname
+struct oldold_utsname {
+        char sysname[9];
+        char nodename[9];
+        char release[9];
+        char version[9];
+        char machine[9];
+};
+#endif
+struct old_utsname {
+        char sysname[65];
+        char nodename[65];
+        char release[65];
+        char version[65];
+        char machine[65];
+};
+
+struct i386_user_desc {
+    unsigned int  entry_number;
+    unsigned long base_addr;
+    unsigned int  limit;
+    unsigned int  seg_32bit:1;
+    unsigned int  contents:2;
+    unsigned int  read_exec_only:1;
+    unsigned int  limit_in_pages:1;
+    unsigned int  seg_not_present:1;
+    unsigned int  useable:1;
+};
+
+int clone_fn(void* arg)
+{
+    x86emu_t *emu = (x86emu_t*)arg;
+    R_EAX = 0;
+    DynaRun(emu);
+    int ret = R_EAX;
+    FreeX86Emu(&emu);
+    return ret;
+}
+
+void EXPORT x86Syscall(x86emu_t *emu)
+{
+    RESET_FLAGS(emu);
+    uint32_t s = R_EAX;
+    printf_log(LOG_DEBUG, "%p: Calling syscall 0x%02X (%d) %p %p %p %p %p", (void*)R_EIP, s, s, (void*)R_EBX, (void*)R_ECX, (void*)R_EDX, (void*)R_ESI, (void*)R_EDI); 
+    // check wrapper first
+    int cnt = sizeof(syscallwrap) / sizeof(scwrap_t);
+    for (int i=0; i<cnt; i++) {
+        if(syscallwrap[i].x86s == s) {
+            int sc = syscallwrap[i].nats;
+            switch(syscallwrap[i].nbpars) {
+                case 0: *(int32_t*)&R_EAX = syscall(sc); break;
+                case 1: *(int32_t*)&R_EAX = syscall(sc, R_EBX); break;
+                case 2: if(s==33) {printf_log(LOG_DUMP, " => sys_access(\"%s\", %d)\n", (char*)R_EBX, R_ECX);}; *(int32_t*)&R_EAX = syscall(sc, R_EBX, R_ECX); break;
+                case 3: *(int32_t*)&R_EAX = syscall(sc, R_EBX, R_ECX, R_EDX); break;
+                case 4: *(int32_t*)&R_EAX = syscall(sc, R_EBX, R_ECX, R_EDX, R_ESI); break;
+                case 5: *(int32_t*)&R_EAX = syscall(sc, R_EBX, R_ECX, R_EDX, R_ESI, R_EDI); break;
+                case 6: *(int32_t*)&R_EAX = syscall(sc, R_EBX, R_ECX, R_EDX, R_ESI, R_EDI, R_EBP); break;
+                default:
+                   printf_log(LOG_NONE, "ERROR, Unimplemented syscall wrapper (%d, %d)\n", s, syscallwrap[i].nbpars); 
+                   emu->quit = 1;
+                   return;
+            }
+            printf_log(LOG_DEBUG, " => 0x%x\n", R_EAX);
+            return;
+        }
+    }
+    switch (s) {
+        case 1: // sys_exit
+            emu->quit = 1;
+            emu->exit = 1;
+            R_EAX = R_EBX; // faking the syscall here, we don't want to really terminate the program now
+            break;
+        case 3:  // sys_read
+            R_EAX = (uint32_t)read((int)R_EBX, (void*)R_ECX, (size_t)R_EDX);
+            break;
+        case 4:  // sys_write
+            R_EAX = (uint32_t)write((int)R_EBX, (void*)R_ECX, (size_t)R_EDX);
+            break;
+        case 5: // sys_open
+            if(s==5) {printf_log(LOG_DUMP, " => sys_open(\"%s\", %d, %d)", (char*)R_EBX, of_convert(R_ECX), R_EDX);}; 
+            //R_EAX = (uint32_t)open((void*)R_EBX, of_convert(R_ECX), R_EDX);
+            R_EAX = (uint32_t)my_open(emu, (void*)R_EBX, of_convert(R_ECX), R_EDX);
+            break;
+        case 6:  // sys_close
+            R_EAX = (uint32_t)close((int)R_EBX);
+            break;
+#ifndef __NR_waitpid
+        case 7: //sys_waitpid
+            R_EAX = waitpid((pid_t)R_EBX, (int*)R_ECX, (int)R_EDX);
+            break;
+#endif
+        case 11: // sys_execve
+            {
+                char* prog = (char*)R_EBX;
+                char** argv = (char**)R_ECX;
+                char** envv = (char**)R_EDX;
+                printf_log(LOG_DUMP, " => sys_execve(\"%s\", %p(\"%s\", \"%s\", \"%s\"...), %p)\n", prog, argv, (argv && argv[0])?argv[0]:"nil", (argv && argv[0] && argv[1])?argv[1]:"nil", (argv && argv[0] && argv[1] && argv[2])?argv[2]:"nil", envv);
+                R_EAX = execve((const char*)R_EBX, (void*)R_ECX, (void*)R_EDX);
+            }
+            break;
+#ifndef __NR_time
+        case 13:    // sys_time (it's deprecated and remove on ARM EABI it seems)
+            R_EAX = time(NULL);
+            break;
+#endif
+        case 54: // sys_ioctl
+            R_EAX = (uint32_t)ioctl((int)R_EBX, R_ECX, R_EDX, R_ESI, R_EDI);
+            break;
+        case 55: // sys_fcntl
+            if(R_ECX==4) {
+                // filter out O_NONBLOCK so old stacally linked games that access X11 don't get EAGAIN error sometimes
+                int tmp = of_convert((int)R_EDX)&(~O_NONBLOCK);
+                R_EAX = (uint32_t)fcntl((int)R_EBX, (int)R_ECX, tmp);
+            } else
+                R_EAX = (uint32_t)fcntl((int)R_EBX, (int)R_ECX, R_EDX);
+            break;
+#ifndef __NR_getrlimit
+        case 76:    // sys_getrlimit... this is the old version, using the new one. Maybe some tranform is needed?
+            R_EAX = getrlimit(R_EBX, (void*)R_ECX);
+            break;
+#endif
+#ifndef __NR_select
+        case 82:   // select
+            R_EAX = select(R_EBX, (fd_set*)R_ECX, (fd_set*)R_EDX, (fd_set*)R_ESI, (struct timeval*)R_EDI);
+            break;
+#endif
+        case 90:    // old_mmap
+            {
+                struct mmap_arg_struct *st = (struct mmap_arg_struct*)R_EBX;
+                R_EAX = (uintptr_t)mmap((void*)st->addr, st->len, st->prot, st->flags, st->fd, st->offset);
+            }
+            break;
+#ifndef __NR_socketcall
+        case 102: {
+                unsigned long *args = (unsigned long *)R_ECX;
+                // need to do all call "by hand"
+                switch(R_EBX) {
+                    case SYS_SOCKET: R_EAX = socket(args[0], args[1], args[2]); break;
+                    case SYS_BIND: R_EAX = bind(args[0], (void*)args[1], args[2]); break;
+                    case SYS_CONNECT: R_EAX = connect(args[0], (void*)args[1], args[2]); break;
+                    case SYS_LISTEN: R_EAX = listen(args[0], args[1]); break;
+                    case SYS_ACCEPT: R_EAX = accept(args[0], (void*)args[1], (void*)args[2]); break;
+                    case SYS_GETSOCKNAME: R_EAX = getsockname(args[0], (void*)args[1], (void*)args[2]); break;
+                    case SYS_GETPEERNAME: R_EAX = getpeername(args[0], (void*)args[1], (void*)args[2]); break;
+                    case SYS_SOCKETPAIR: R_EAX = socketpair(args[0], args[1], args[2], (int*)args[3]); break;
+                    case SYS_SEND: R_EAX = send(args[0], (void*)args[1], args[2], args[3]); break;
+                    case SYS_RECV: R_EAX = recv(args[0], (void*)args[1], args[2], args[3]); break;
+                    case SYS_SENDTO: R_EAX = sendto(args[0], (void*)args[1], args[2], args[3], (void*)args[4], args[5]); break;
+                    case SYS_RECVFROM: R_EAX = recvfrom(args[0], (void*)args[1], args[2], args[3], (void*)args[4], (void*)args[5]); break;
+                    case SYS_SHUTDOWN: R_EAX = shutdown(args[0], args[1]); break;
+                    case SYS_SETSOCKOPT: R_EAX = setsockopt(args[0], args[1], args[2], (void*)args[3], args[4]); break;
+                    case SYS_GETSOCKOPT: R_EAX = getsockopt(args[0], args[1], args[2], (void*)args[3], (void*)args[4]); break;
+                    case SYS_SENDMSG: R_EAX = sendmsg(args[0], (void*)args[1], args[2]); break;
+                    case SYS_RECVMSG: R_EAX = recvmsg(args[0], (void*)args[1], args[2]); break;
+                    case SYS_ACCEPT4: R_EAX = my_accept4(emu, args[0], (void*)args[1], (void*)args[2], args[3]); break;
+                    #ifdef SYS_RECVMMSG
+                    // TODO: Create my_ version of recvmmsg and sendmmsg
+                    case SYS_RECVMMSG: R_EAX = recvmmsg(args[0], (void*)args[1], args[2], args[3], (void*)args[4]); break;
+                    case SYS_SENDMMSG: R_EAX = sendmmsg(args[0], (void*)args[1], args[2], args[3]); break;
+                    #endif
+                    default:
+                        printf_log(LOG_DEBUG, "BOX86 Error on Syscall 102: Unknown Soket command %d\n", R_EBX);
+                        R_EAX = -1;
+                }
+            }
+            break;
+#endif
+#ifndef __NR_olduname
+        case 109:   // olduname
+            {
+                struct utsname un;
+                R_EAX = uname(&un);
+                if(!R_EAX) {
+                    struct oldold_utsname *old = (struct oldold_utsname*)R_EBX;
+                    memcpy(old->sysname, un.sysname, 9);
+                    memcpy(old->nodename, un.nodename, 9);
+                    memcpy(old->release, un.release, 9);
+                    memcpy(old->version, un.version, 9);
+                    strcpy(old->machine, "i686");
+                }
+            }
+            break;
+#endif
+#ifndef __NR_iopl
+        case 110:   // iopl
+            R_EAX = 0;  // only on x86, so return 0...
+            break;
+#endif
+        case 119: // sys_sigreturn
+            emu->quit = 1;  // we should be inside a DynaCall in a sigaction callback....
+            break;
+        case 120:   // clone
+            {
+                //struct x86_pt_regs *regs = (struct x86_pt_regs *)R_EDI;
+                // lets duplicate the emu
+                void* stack_base = (void*)R_ECX;
+                int stack_size = 0;
+                if(!R_ECX) {
+                    // allocate a new stack...
+                    stack_size = 1024*1024;
+                    stack_base = malloc(stack_size); // why not 1M... (normal operation do copy on write, simpler to just copy)
+                    // copy value from old stack to new stack
+                    int size_to_copy = (uintptr_t)emu->init_stack + emu->size_stack - (R_ESP);
+                    memcpy(stack_base-size_to_copy, (void*)R_ESP, size_to_copy);
+                }
+                x86emu_t * newemu = NewX86Emu(emu->context, R_EIP, (uintptr_t)stack_base, stack_size, (R_ECX)?0:1);
+                SetupX86Emu(newemu);
+                CloneEmu(newemu, emu);
+                SetESP(newemu, (uintptr_t)stack_base);
+                // setup registers
+                /*if(regs) {
+                    SetEAX(newemu, regs->eax);
+                    SetEBX(newemu, regs->ebx);
+                    SetECX(newemu, regs->ecx);
+                    SetEDX(newemu, regs->edx);
+                    SetEDI(newemu, regs->edi);
+                    SetESI(newemu, regs->esi);
+                    SetEBP(newemu, regs->ebp);
+                    SetESP(newemu, (uintptr_t)stack_base - size_to_copy);
+                }
+                SetESP(newemu, (uintptr_t)stack_base - size_to_copy);*/
+                void* mystack = (R_EBX&CLONE_VM)?malloc(1024*1024):NULL;  // stack for own process... memory leak, but no practical way to remove it
+                int ret = clone(clone_fn, (void*)((R_EBX&CLONE_VM)?((uintptr_t)mystack+1024*1024):0), R_EBX, newemu, R_ESI, R_EDI, R_EBP);
+                //int r = syscall(__NR_clone, R_EBX, (R_EBX&CLONE_VM)?((uintptr_t)mystack):0, R_EDX, R_ESI, NULL);  // cannot use that syscall in C: how to setup the stack?!
+                R_EAX = ret;
+            }
+            break;
+        case 122:   // uname
+            {
+                struct utsname un;
+                R_EAX = uname(&un);
+                if(!R_EAX) {
+                    struct old_utsname *old = (struct old_utsname*)R_EBX;
+                    memcpy(old->sysname, un.sysname, 65);
+                    memcpy(old->nodename, un.nodename, 65);
+                    memcpy(old->release, un.release, 65);
+                    memcpy(old->version, un.version, 65);
+                    strcpy(old->machine, "i686");
+                }
+            }
+            break;
+        case 168: // sys_poll
+            R_EAX = (uint32_t)poll((void*)R_EBX, R_ECX, (int)R_EDX);
+            break;
+        case 173: // sys_rt_sigreturn
+            emu->quit = 1;  // we should be inside a DynaCall in a sigaction callback....
+            break;
+        case 174: // sys_rt_sigaction
+            //printf_log(LOG_NONE, "Warning, Ignoring sys_rt_sigaction(0x%02X, %p, %p)\n", R_EBX, (void*)R_ECX, (void*)R_EDX);
+            R_EAX = my_syscall_sigaction(emu, R_EBX, (void*)R_ECX, (void*)R_EDX, R_ESI);
+            break;
+        case 190:   // vfork
+            {
+                int r = vfork();
+                R_EAX = r;
+            }
+            break;
+        case 195:   // stat64
+            {   
+                struct stat64 st;
+                unsigned int r = syscall(__NR_stat64, R_EBX, &st);
+                UnalignStat64(&st, (void*)R_ECX);
+                
+                R_EAX = r;
+            }
+            break;
+        case 197:   // fstat64
+            {   
+                struct stat64 st;
+                unsigned int r = syscall(__NR_fstat64, R_EBX, &st);
+                UnalignStat64(&st, (void*)R_ECX);
+                
+                R_EAX = r;
+            }
+            break;
+        case 221: // sys_fcntl64
+            if(R_ECX==4) {
+                int tmp = of_convert((int)R_EDX);
+                R_EAX = (uint32_t)fcntl((int)R_EBX, R_ECX, tmp);
+            } else
+                R_EAX = (uint32_t)fcntl((int)R_EBX, (int)R_ECX, R_EDX);
+            break;
+        case 270:   // tgkill
+            R_EAX = syscall(__NR_tgkill, R_EBX, R_ECX, R_EDX);
+            break;
+        case 355:  // get_random
+            R_EAX = my_getrandom(emu, (void*)R_EBX, R_ECX, R_EDX);
+            break;
+        default:
+            printf_log(LOG_INFO, "Error: Unsupported Syscall 0x%02Xh (%d)\n", s, s);
+            emu->quit = 1;
+            emu->error |= ERR_UNIMPL;
+            return;
+    }
+    printf_log(LOG_DEBUG, " => 0x%x\n", R_EAX);
+}
+
+#define stack(n) (R_ESP+4+n)
+#define i32(n)  *(int32_t*)stack(n)
+#define u32(n)  *(uint32_t*)stack(n)
+#define p(n)    *(void**)stack(n)
+
+uint32_t EXPORT my_syscall(x86emu_t *emu)
+{
+    uint32_t s = u32(0);
+    printf_log(LOG_DUMP, "%p: Calling libc syscall 0x%02X (%d) %p %p %p %p %p\n", (void*)R_EIP, s, s, (void*)u32(4), (void*)u32(8), (void*)u32(12), (void*)u32(16), (void*)u32(20)); 
+    // check wrapper first
+    int cnt = sizeof(syscallwrap) / sizeof(scwrap_t);
+    for (int i=0; i<cnt; i++) {
+        if(syscallwrap[i].x86s == s) {
+            int sc = syscallwrap[i].nats;
+            switch(syscallwrap[i].nbpars) {
+                case 0: return syscall(sc);
+                case 1: return syscall(sc, u32(4));
+                case 2: return syscall(sc, u32(4), u32(8));
+                case 3: return syscall(sc, u32(4), u32(8), u32(12));
+                case 4: return syscall(sc, u32(4), u32(8), u32(12), u32(16));
+                case 5: return syscall(sc, u32(4), u32(8), u32(12), u32(16), u32(20));
+                case 6: return syscall(sc, u32(4), u32(8), u32(12), u32(16), u32(20), u32(24));
+                default:
+                   printf_log(LOG_NONE, "ERROR, Unimplemented syscall wrapper (%d, %d)\n", s, syscallwrap[i].nbpars); 
+                   emu->quit = 1;
+                   return 0;
+            }
+        }
+    }
+    switch (s) {
+        case 1: // __NR_exit
+            emu->quit = 1;
+            return u32(4); // faking the syscall here, we don't want to really terminate the program now
+        case 3:  // sys_read
+            return (uint32_t)read(i32(4), p(8), u32(12));
+            break;
+        case 4:  // sys_write
+            return (uint32_t)write(i32(4), p(8), u32(12));
+            break;
+        case 5: // sys_open
+            return my_open(emu, p(4), of_convert(u32(8)), u32(12));
+        case 6:  // sys_close
+            return (uint32_t)close(i32(4));
+            break;
+        case 270: //_NR_tgkill
+            if(!u32(12)) {
+                //printf("tgkill(%u, %u, %u) => ", u32(4), u32(8), u32(12));
+                uint32_t ret = (uint32_t)syscall(__NR_tgkill, u32(4), u32(8), u32(12));
+                //printf("%u (errno=%d)\n", ret, (ret==(uint32_t)-1)?errno:0);
+                return ret;
+            } else {
+                printf_log(LOG_INFO, "Warning: ignoring libc Syscall tgkill (%u, %u, %u)\n", u32(4), u32(8), u32(12));
+            }
+            return 0;
+        default:
+            printf_log(LOG_INFO, "Error: Unsupported libc Syscall 0x%02X (%d)\n", s, s);
+            emu->quit = 1;
+            emu->error |= ERR_UNIMPL;
+    }
+    return 0;
+}

src/emu/x86trace.c

@@ -39,6 +39,8 @@ typedef struct zydis_dec_s {
 
 int InitX86Trace(box86context_t *context)
 {
+    if(context->zydis)
+        return 0;
     context->zydis = (zydis_t*)calloc(1, sizeof(zydis_t));
     if(!context->zydis)
         return 1;

src/emu/x86zydis.h

@@ -929,6 +929,9 @@ typedef struct ZydisFormatter_
      * @brief   The `ZYDIS_FORMATTER_FUNC_PRINT_DECORATOR` function.
      */
     ZydisFormatterDecoratorFunc func_print_decorator;
+
+    // ummy padding, in case the structure growth...
+    uint32_t    padding[20];
 } ZydisFormatter;
 
 /**

src/emu/x87emu_private.c

@@ -23,8 +23,14 @@ void fpu_fbst(x86emu_t* emu, uint8_t* d) {
     // very aproximative... but should not be much used...
     uint8_t p;
     uint8_t sign = 0x00;
+    #ifdef USE_FLOAT
+    double tmp, v = ST0.f;
+    if(ST0.f<0.0f) 
+    #else
     double tmp, v = ST0.d;
-    if(ST0.d<0.0) {
+    if(ST0.d<0.0) 
+    #endif
+    {
         sign = 0x80;
         v = -v;
     }
@@ -56,23 +62,47 @@ void fpu_fbld(x86emu_t* emu, uint8_t* s) {
         tmp += m * ((p>>4)&0x0f);
         m *= 10;
     }
+    #ifdef USE_FLOAT
+    ST0.f = tmp;
+    p =*(s++);
+    ST0.f += m * (p&0x0f);
+    if(p&0x80)
+        ST0.f = -ST0.f;
+    #else
     ST0.d = tmp;
     p =*(s++);
     ST0.d += m * (p&0x0f);
     if(p&0x80)
         ST0.d = -ST0.d;
+    #endif
 }
 
 
+#ifdef USE_FLOAT
+typedef union {
+    double d;
+    struct {
+        uint32_t lower;
+        uint32_t upper;
+    } l;
+    struct {
+        float lower;
+        float upper;
+    } f;
+    int64_t ll;
+} FPU_t;
+#else
+#define FPU_t fpu_reg_t
+#endif
 #define BIAS80 16383
 #define BIAS64 1023
 // long double (80bits) -> double (64bits)
 void LD2D(void* ld, void* d)
 {
-	fpu_reg_t result;
+	FPU_t result;
     #pragma pack(push, 1)
 	struct {
-		fpu_reg_t f;
+		FPU_t f;
 		int16_t b;
 	} val;
     #pragma pack(pop)
@@ -82,7 +112,7 @@ void LD2D(void* ld, void* d)
 	int32_t exp64 = (((uint32_t)(val.b&0x7fff) - BIAS80) + BIAS64);
 	int32_t exp64final = exp64&0x7ff;
     // do specific value first (0, infinite...)
-    // bit 63 is "intetger part"
+    // bit 63 is "integer part"
     // bit 62 is sign
     if((uint32_t)(val.b&0x7fff)==0x7fff) {
         // infinity and nans
@@ -96,7 +126,7 @@ void LD2D(void* ld, void* d)
         if(t) {    // infinite
             result.d = HUGE_VAL;
         } else {      // NaN
-            result.l.upper |= 0x7ff << 20;
+            result.l.upper = 0x7ff << 20;
             result.l.lower = 0;
         }
         if(val.b&0x8000)
@@ -138,11 +168,11 @@ void D2LD(void* d, void* ld)
 {
     #pragma pack(push, 1)
 	struct {
-		fpu_reg_t f;
+		FPU_t f;
 		int16_t b;
 	} val;
     #pragma pack(pop)
-    fpu_reg_t s;
+    FPU_t s;
     s.ll = *(uint64_t*)d;   // use memcpy to avoid risk of Bus Error?
     // do special value first
     if((s.ll&0x7fffffffffffffffLL)==0) {
@@ -188,7 +218,6 @@ double FromLD(void* ld)
     return ret;
 }
 
-
 void fpu_loadenv(x86emu_t* emu, char* p, int b16)
 {
     emu->cw = *(uint16_t*)p;

src/emu/x87emu_private.h

@@ -50,8 +50,28 @@ static inline void fpu_do_pop(x86emu_t* emu)
 
 void reset_fpu(x86emu_t* emu);
 
+#ifdef USE_FLOAT
+static inline void fpu_fcom(x86emu_t* emu, float b)
+#else
 static inline void fpu_fcom(x86emu_t* emu, double b)
+#endif
 {
+    emu->sw.f.F87_C1 = 0;
+    #ifdef USE_FLOAT
+    if(isnan(ST0.f) || isnan(b)) {
+        emu->sw.f.F87_C0 = 1;
+        emu->sw.f.F87_C2 = 1;
+        emu->sw.f.F87_C3 = 1;
+    } else if (isgreater(ST0.f, b)) {
+        emu->sw.f.F87_C0 = 0;
+        emu->sw.f.F87_C2 = 0;
+        emu->sw.f.F87_C3 = 0;
+    } else if (isless(ST0.f, b)) {
+        emu->sw.f.F87_C0 = 1;
+        emu->sw.f.F87_C2 = 0;
+        emu->sw.f.F87_C3 = 0;
+    } 
+    #else
     if(isnan(ST0.d) || isnan(b)) {
         emu->sw.f.F87_C0 = 1;
         emu->sw.f.F87_C2 = 1;
@@ -64,16 +84,38 @@ static inline void fpu_fcom(x86emu_t* emu, double b)
         emu->sw.f.F87_C0 = 1;
         emu->sw.f.F87_C2 = 0;
         emu->sw.f.F87_C3 = 0;
-    } else {
+    } 
+    #endif
+    else {
         emu->sw.f.F87_C0 = 0;
         emu->sw.f.F87_C2 = 0;
         emu->sw.f.F87_C3 = 1;
     }
 }
 
+#ifdef USE_FLOAT
+static inline void fpu_fcomi(x86emu_t* emu, float b)
+#else
 static inline void fpu_fcomi(x86emu_t* emu, double b)
+#endif
 {
     RESET_FLAGS(emu);
+    emu->sw.f.F87_C1 = 0;
+    #ifdef USE_FLOAT
+    if(isnan(ST0.f) || isnan(b)) {
+        emu->flags[F_CF] = 1;
+        emu->flags[F_PF] = 1;
+        emu->flags[F_ZF] = 1;
+    } else if (isgreater(ST0.f, b)) {
+        emu->flags[F_CF] = 0;
+        emu->flags[F_PF] = 0;
+        emu->flags[F_ZF] = 0;
+    } else if (isless(ST0.f, b)) {
+        emu->flags[F_CF] = 1;
+        emu->flags[F_PF] = 0;
+        emu->flags[F_ZF] = 0;
+    } 
+    #else
     if(isnan(ST0.d) || isnan(b)) {
         emu->flags[F_CF] = 1;
         emu->flags[F_PF] = 1;
@@ -86,41 +128,89 @@ static inline void fpu_fcomi(x86emu_t* emu, double b)
         emu->flags[F_CF] = 1;
         emu->flags[F_PF] = 0;
         emu->flags[F_ZF] = 0;
-    } else {
+    } 
+    #endif
+    else {
         emu->flags[F_CF] = 0;
         emu->flags[F_PF] = 0;
         emu->flags[F_ZF] = 1;
     }
 }
 
+#ifdef USE_FLOAT
+static inline float fpu_round(x86emu_t* emu, float d) {
+    if (!isfinite(d))
+        return d;
+    switch(emu->round) {
+        case ROUND_Nearest:
+            return floorf(d+0.5);
+        case ROUND_Down:
+            return floorf(d);
+        case ROUND_Up:
+            return ceilf(d);
+        case ROUND_Chop:
+        default:
+            return roundf(d);
+    }
+}
+#else
 static inline double fpu_round(x86emu_t* emu, double d) {
     if (!isfinite(d))
         return d;
-    //switch(emu->cw)   // TODO: implement Rounding...
-    return round(d);
+    switch(emu->round) {
+        case ROUND_Nearest:
+            return floor(d+0.5);
+        case ROUND_Down:
+            return floor(d);
+        case ROUND_Up:
+            return ceil(d);
+        case ROUND_Chop:
+        default:
+            return round(d);
+    }
 }
+#endif
 
 static inline void fpu_fxam(x86emu_t* emu) {
+    #ifdef USE_FLOAT
+    emu->sw.f.F87_C1 = (ST0.ll&0x8000)?1:0;
+    #else
     emu->sw.f.F87_C1 = (ST0.l.upper&0x80000000)?1:0;
+    #endif
     if(!emu->fpu_stack) {
         emu->sw.f.F87_C3 = 1;
         emu->sw.f.F87_C2 = 0;
         emu->sw.f.F87_C0 = 1;
         return;
     }
-    if(isinf(ST0.d)) {  // TODO: Unsuported and denormal not analysed...
+    #ifdef USE_FLOAT
+    if(isinf(ST0.f))
+    #else
+    if(isinf(ST0.d)) 
+    #endif
+    {  // TODO: Unsuported and denormal not analysed...
         emu->sw.f.F87_C3 = 0;
         emu->sw.f.F87_C2 = 1;
         emu->sw.f.F87_C0 = 1;
         return;
     }
-    if(isnan(ST0.d)) {  // TODO: Unsuported and denormal not analysed...
+    #ifdef USE_FLOAT
+    if(isnan(ST0.f))
+    #else
+    if(isnan(ST0.d))
+    #endif
+    {  // TODO: Unsuported and denormal not analysed...
         emu->sw.f.F87_C3 = 0;
         emu->sw.f.F87_C2 = 0;
         emu->sw.f.F87_C0 = 1;
         return;
     }
-    if(ST0.d==0.0) {
+    #ifdef USE_FLOAT
+    if(ST0.f==0.0f)
+    #else
+    if(ST0.d==0.0)
+    #endif
+    {
         emu->sw.f.F87_C3 = 1;
         emu->sw.f.F87_C2 = 0;
         emu->sw.f.F87_C0 = 0;
@@ -133,10 +223,44 @@ static inline void fpu_fxam(x86emu_t* emu) {
 
 }
 
+static inline void fpu_ftst(x86emu_t* emu) {
+    emu->sw.f.F87_C1 = 0;
+    #ifdef USE_FLOAT
+    if(isinf(ST0.f) || isnan(ST0.f))
+    #else
+    if(isinf(ST0.d) || isnan(ST0.d)) 
+    #endif
+    {  // TODO: Unsuported and denormal not analysed...
+        emu->sw.f.F87_C3 = 1;
+        emu->sw.f.F87_C2 = 1;
+        emu->sw.f.F87_C0 = 1;
+        return;
+    }
+    #ifdef USE_FLOAT
+    if(ST0.f==0.0f)
+    #else
+    if(ST0.d==0.0)
+    #endif
+    {
+        emu->sw.f.F87_C3 = 1;
+        emu->sw.f.F87_C2 = 0;
+        emu->sw.f.F87_C0 = 0;
+        return;
+    }
+    // normal...
+    emu->sw.f.F87_C3 = 0;
+    emu->sw.f.F87_C2 = 0;
+    #ifdef USE_FLOAT
+    emu->sw.f.F87_C0 = (ST0.ll&0x8000)?1:0;
+    #else
+    emu->sw.f.F87_C0 = (ST0.l.upper&0x80000000)?1:0;
+    #endif
+}
+
 void fpu_fbst(x86emu_t* emu, uint8_t* d);
 void fpu_fbld(x86emu_t* emu, uint8_t* s);
 
 void fpu_loadenv(x86emu_t* emu, char* p, int b16);
 void fpu_savenv(x86emu_t* emu, char* p, int b16);
 
-#endif //__X87RUN_PRIVATE_H_
+#endif //__X87RUN_PRIVATE_H_

src/emu/x87run.c

@@ -6,7 +6,7 @@
 #include <string.h>
 
 #include "debug.h"
-#include "stack.h"
+#include "box86stack.h"
 #include "x86emu.h"
 #include "x86run.h"
 #include "x86emu_private.h"
@@ -159,7 +159,12 @@ void Run66DD(x86emu_t *emu)
                     char* p =(char*)op1;
                     p += 14;
                     for (int i=0; i<8; ++i) {
+                        #ifdef USE_FLOAT
+                        LD2D(p, &d);
+                        ST(i).f = d;
+                        #else
                         LD2D(p, &ST(i).d);
+                        #endif
                         p+=10;
                     }
                 }
@@ -173,7 +178,12 @@ void Run66DD(x86emu_t *emu)
                     char* p =(char*)op1;
                     p += 14;
                     for (int i=0; i<8; ++i) {
+                        #ifdef USE_FLOAT
+                        d = ST(i).f;
+                        D2LD(&d, p);
+                        #else
                         D2LD(&ST(i).d, p);
+                        #endif
                         p+=10;
                     }
                 }

src/fileutils.c

@@ -1,48 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <sys/types.h> 
-#include <sys/stat.h> 
-#include <unistd.h>
-#include <limits.h>
-#include <stdint.h>
-
-#ifndef MAX_PATH
-#define MAX_PATH 4096
-#endif
-
-#include "debug.h"
-#include "fileutils.h"
-
-int FileExist(const char* filename, int flags)
-{
-    struct stat sb;
-    if (stat(filename, &sb) == -1)
-        return 0;
-    // check type of file? should be executable, or folder
-    if(flags&IS_FILE) {
-        if(!S_ISREG(sb.st_mode))
-            return 0;
-    } else {
-        if(!S_ISDIR(sb.st_mode))
-            return 0;
-    }
-    if(flags&IS_EXECUTABLE) {
-        if((sb.st_mode&S_IXUSR)!=S_IXUSR)
-            return 0;   // nope
-    }
-    return 1;
-}
-
-char* ResolveFile(const char* filename, path_collection_t* paths)
-{
-    char p[MAX_PATH];
-    for (int i=0; i<paths->size; ++i) {
-        strcpy(p, paths->paths[i]);
-        strcat(p, filename);
-        if(FileExist(p, IS_FILE))
-            return strdup(p);
-    }
-
-    return NULL;
-}

src/fileutils.h

@@ -1,13 +0,0 @@
-#ifndef __FILEUTILS_H_
-#define __FILEUTILS_H_
-
-#include "pathcoll.h"
-
-#define IS_EXECUTABLE   (1<<0)
-#define IS_FILE         (1<<1)
-       
-
-int FileExist(const char* filename, int flags);    // 0 : doesn't exist, 1: Does exist
-char* ResolveFile(const char* filename, path_collection_t* paths); // find a file, using Path if needed
-
-#endif //__FILEUTILS_H_

src/include/box86context.h

@@ -14,14 +14,39 @@ typedef struct dlprivate_s dlprivate_t;
 typedef struct kh_symbolmap_s kh_symbolmap_t;
 typedef struct callbacklist_s callbacklist_t;
 typedef struct library_s library_t;
+typedef struct kh_fts_s kh_fts_t;
+typedef struct atfork_fnc_s {
+    uintptr_t prepare;
+    uintptr_t parent;
+    uintptr_t child;
+} atfork_fnc_t;
+#ifdef DYNAREC
+typedef struct dynablocklist_s dynablocklist_t;
+typedef struct mmaplist_s mmaplist_t;
+#endif
 
 typedef void* (*procaddess_t)(const char* name);
 
+#define MAX_SIGNAL 64
+
+typedef struct tlsdatasize_s {
+    int32_t     tlssize;
+    void*       tlsdata;
+} tlsdatasize_t;
+
+void free_tlsdatasize(void* p);
+
 typedef struct box86context_s {
     path_collection_t   box86_path;     // PATH env. variable
     path_collection_t   box86_ld_lib;   // LD_LIBRARY_PATH env. variable
 
+    path_collection_t   box86_emulated_libs;    // Collection of libs that should not be wrapped
+
     int                 x86trace;
+    int                 trace_tid;
+#ifdef DYNAREC
+    int                 trace_dynarec;
+#endif
     zydis_t             *zydis;         // dlopen the zydis dissasembler
     void*               box86lib;       // dlopen on box86 itself
 
@@ -32,6 +57,7 @@ typedef struct box86context_s {
     char**              envv;
 
     char*               fullpath;
+    char*               box86path;      // path of current box86 executable
 
     uint32_t            stacksz;
     int                 stackalign;
@@ -58,26 +84,75 @@ typedef struct box86context_s {
     callbacklist_t      *callbacks;     // all callbacks
 
     pthread_mutex_t     mutex_once;
+    pthread_mutex_t     mutex_once2;
+    pthread_mutex_t     mutex_trace;
+    pthread_mutex_t     mutex_lock;
 
     library_t           *libclib;       // shortcut to libc library (if loaded, so probably yes)
     library_t           *sdl1lib;       // shortcut to SDL1 library (if loaded)
+    void*               sdl1allocrw;
+    void*               sdl1freerw;
     library_t           *sdl1mixerlib;
     library_t           *sdl1imagelib;
+    library_t           *sdl1ttflib;
     library_t           *sdl2lib;       // shortcut to SDL2 library (if loaded)
+    void*               sdl2allocrw;
+    void*               sdl2freerw;
     library_t           *sdl2mixerlib;
     library_t           *sdl2imagelib;
+    library_t           *sdl2ttflib;
+    library_t           *x11lib;
+    library_t           *zlib;
+    library_t           *vorbisfile;
+    library_t           *vorbis;
+    library_t           *asound;
+    library_t           *monounity;
 
     int                 deferedInit;
     elfheader_t         **deferedInitList;
     int                 deferedInitSz;
     int                 deferedInitCap;
 
+    pthread_key_t       tlskey;     // then tls key to have actual tlsdata
+    void*               tlsdata;    // the initial global tlsdata
+    int32_t             tlssize;    // wanted size of tlsdata
+
+#ifdef DYNAREC
+    pthread_mutex_t     mutex_blocks;
+    pthread_mutex_t     mutex_mmap;
+    dynablocklist_t     *dynablocks;
+    mmaplist_t          *mmaplist;
+    int                 mmapsize;
+#endif
+#ifndef NOALIGN
+    kh_fts_t            *ftsmap;
+#endif
+
     int                 forked;         //  how many forks... cleanup only when < 0
+
+    atfork_fnc_t        *atforks;       // fnc for atfork...
+    int                 atfork_sz;
+    int                 atfork_cap;
+
+    uint8_t             canary[4];
+
+    uintptr_t           signals[MAX_SIGNAL];
+    uintptr_t           restorer[MAX_SIGNAL];
+    int                 no_sigsegv;
+
 } box86context_t;
 
 box86context_t *NewBox86Context(int argc);
 void FreeBox86Context(box86context_t** context);
 
-int AddElfHeader(box86context_t* ctx, elfheader_t* head);    // return the index of header
+// return the index of the added header
+int AddElfHeader(box86context_t* ctx, elfheader_t* head);
+
+// return the tlsbase (negative) for the new TLS partition created (no partition index is stored in the context)
+int AddTLSPartition(box86context_t* context, int tlssize);
+
+#ifdef DYNAREC
+uintptr_t AllocDynarecMap(box86context_t *context, int size);
+#endif
 
 #endif //__BOX86CONTEXT_H_

src/include/box86lib.h

@@ -0,0 +1,13 @@
+#ifndef _BOX86LIB_H_
+#define _BOX86LIB_H_
+
+typedef struct path_collection_s path_collection_t;
+typedef struct box86context_s box86context_t;
+
+void LoadLogEnv();
+void LoadEnvPath(path_collection_t *col, const char* defpath, const char* env);
+int CountEnv(const char** env);
+int GatherEnv(char*** dest, const char** env, const char* prog);
+void LoadEnvVars(box86context_t *context);
+
+#endif //_BOX86LIB_H_

src/include/box86stack.h

@@ -1,5 +1,5 @@
-#ifndef __STACK_H_
-#define __STACK_H_
+#ifndef __BOX86_STACK_H_
+#define __BOX86_STACK_H_
 
 #include <stdint.h>
 
@@ -11,5 +11,7 @@ void SetupInitialStack(box86context_t *context);
 
 uint16_t Pop16(x86emu_t *emu);
 void Push16(x86emu_t *emu, uint16_t v);
+uint32_t Pop32(x86emu_t *emu);
+void Push32(x86emu_t *emu, uint32_t v);
 
-#endif //__STACK_H_
+#endif //__BOX86_STACK_H_

src/include/bridge.h

@@ -2,16 +2,17 @@
 #define __BRIDGE_H_
 #include <stdint.h>
 
-#include "wrapper.h"
-
 typedef struct x86emu_s x86emu_t;
 typedef struct bridge_s bridge_t;
+typedef void (*wrapper_t)(x86emu_t* emu, uintptr_t fnc);
 
 bridge_t *NewBridge();
 void FreeBridge(bridge_t** bridge);
 
-uintptr_t AddBridge(bridge_t* bridge, wrapper_t w, void* fnc);
+uintptr_t AddBridge(bridge_t* bridge, wrapper_t w, void* fnc, int N);
 uintptr_t CheckBridged(bridge_t* bridge, void* fnc);
+uintptr_t AddCheckBridge(bridge_t* bridge, wrapper_t w, void* fnc, int N);
 void* GetNativeFnc(uintptr_t fnc);
+void* GetNativeFncOrFnc(uintptr_t fnc);
 
 #endif //__BRIDGE_H_

src/include/callback.h

@@ -0,0 +1,30 @@
+#ifndef __CALLBACK_H__
+#define __CALLBACK_H__
+
+#include <stdint.h>
+
+typedef struct callbacklist_s callbacklist_t;
+typedef struct x86emu_s x86emu_t;
+
+callbacklist_t* NewCallbackList();
+void FreeCallbackList(callbacklist_t** callbacks);
+
+x86emu_t* AddCallback(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4);
+x86emu_t* AddVariableCallback(x86emu_t* emu, int stsize, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4);
+x86emu_t* AddSmallCallback(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4);
+x86emu_t* AddSharedCallback(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1, void* arg2, void* arg3, void* arg4);
+x86emu_t* GetCallback1Arg(x86emu_t* emu, uintptr_t fnc, int nb_args, void* arg1);
+x86emu_t* FreeCallback(x86emu_t* emu);
+uint32_t RunCallback(x86emu_t* emu);
+void SetCallbackArg(x86emu_t* emu, int arg, void* val);
+void SetCallbackArgs(x86emu_t* emu, int nargs, ...);            // set args starting from 0
+void SetCallbackNArgs(x86emu_t* emu, int N, int nargs, ...);    // set args starting from N
+void* GetCallbackArg(x86emu_t* emu, int arg);
+void SetCallbackNArg(x86emu_t* emu, int narg);
+void SetCallbackAddress(x86emu_t* emu, uintptr_t address);
+uintptr_t GetCallbackAddress(x86emu_t* emu);
+int IsCallback(box86context_t* context, x86emu_t* cb);
+uint32_t RunFunction(box86context_t *context, uintptr_t fnc, int nargs, ...);   // use a local stack from the stack of 128K
+uint32_t RunFunctionWithEmu(x86emu_t *emu, uintptr_t fnc, int nargs, ...);
+
+#endif //__CALLBACK_H__

src/include/debug.h

@@ -0,0 +1,31 @@
+#ifndef __DEBUG_H_
+#define __DEBUG_H_
+
+extern int box86_log;    // log level
+#ifdef DYNAREC
+extern int box86_dynarec_log;
+extern int box86_dynarec;
+extern int box86_dynarec_linker;
+extern int box86_dynarec_trace;
+extern int box86_dynarec_forced;
+#endif
+extern int dlsym_error;  // log dlsym error
+extern int trace_xmm;    // include XMM reg in trace?
+extern int trace_emm;    // include EMM reg in trace?
+extern int allow_missing_libs;
+#define LOG_NONE 0
+#define LOG_INFO 1
+#define LOG_DEBUG 2
+#define LOG_DUMP 3
+
+extern FILE* ftrace;
+
+#define printf_log(L, ...) do {if(L<=box86_log) {fprintf(ftrace, __VA_ARGS__); fflush(ftrace);}} while(0)
+
+#ifdef DYNAREC
+#define dynarec_log(L, ...) do {if(L<=box86_dynarec_log) {fprintf(ftrace, __VA_ARGS__); fflush(ftrace);}} while(0)
+#endif
+
+#define EXPORT __attribute__((visibility("default")))
+
+#endif //__DEBUG_H_

src/include/dynablock.h

@@ -0,0 +1,17 @@
+#ifndef __DYNABLOCK_H_
+#define __DYNABLOCK_H_
+
+typedef struct x86emu_s x86emu_t;
+
+typedef struct dynablock_s dynablock_t;
+
+typedef struct dynablocklist_s dynablocklist_t;
+
+dynablocklist_t* NewDynablockList(uintptr_t base, uintptr_t text, int textsz);
+void FreeDynablockList(dynablocklist_t** dynablocks);
+
+
+// Handling of Dynarec block (i.e. an exectable chunk of x86 translated code)
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t* current);   // return NULL if block is not found / cannot be created. Don't create if create==0
+
+#endif //__DYNABLOCK_H_