Optimized Exit bridge managment

Improved custom memory allocator
Fixed a nasty small memory leak
2026-11-07 17:43:14 -10:00 · 2023-05-10 17:27:31 +02:00 · 2023-05-10 16:23:28 +02:00 · 2023-05-10 16:22:56 +02:00 · 2023-05-09 14:53:13 +02:00 · 2023-05-09 14:45:08 +02:00
141 changed files with 15844 additions and 9051 deletions
@@ -29,6 +29,9 @@ option(HAVE_TRACE "Set to ON to have Trace ability (needs ZydisInfo library)" ${
 option(NOLOADADDR "Set to ON to avoid fixing the load address of Box86" ${NO_LOADAADR})
 option(NOGETCLOCK "Set to ON to avoid using clock_gettime with CLOCK_MONOTONIC_COARSE for RDTSC opcode (use gltimeofday instead)" ${NOGETCLOCK})
 option(NOGIT "Set to ON if not building from a git clone repo (like when building from a zip download from github)" ${NOGIT})
+option(NO_LIB_INSTALL "Set ON to not install a few x86 libs that are used by many program" ${NO_LIB_INSTALL})
+option(NO_CONF_INSTALL "Set ON to not install config files" ${NO_CONF_INSTALL})
+
 if(PANDORA OR PYRA OR RPI2 OR RPI3 OR RPI4 OR GAMESHELL OR ODROID OR GOA_CLONE OR RK3288 OR RK3399 OR ODROIDN2 OR TEGRAX1 OR TEGRAX1ARM32 OR RPI3ARM64 OR RPI4ARM64 OR PHYTIUM OR SD845 OR A64 OR LX2160A OR ARM64 OR RK3588)
    set(LD80BITS OFF CACHE BOOL "")
    set(NOALIGN OFF CACHE BOOL "")
@@ -297,9 +300,6 @@ set(ELFLOADER_SRC
    "${BOX86_ROOT}/src/librarian/symbols.c"
    "${BOX86_ROOT}/src/librarian/globalsymbols.c"
    "${BOX86_ROOT}/src/emu/x86emu.c"
-    "${BOX86_ROOT}/src/emu/x86run.c"
-    "${BOX86_ROOT}/src/emu/x86run66.c"
-    "${BOX86_ROOT}/src/emu/x86run660f.c"
    "${BOX86_ROOT}/src/emu/x86run_private.c"
    "${BOX86_ROOT}/src/emu/x86syscall.c"
    "${BOX86_ROOT}/src/emu/x86primop.c"
@@ -307,7 +307,6 @@ set(ELFLOADER_SRC
    "${BOX86_ROOT}/src/emu/x86int3.c"
    "${BOX86_ROOT}/src/emu/x86tls.c"
    "${BOX86_ROOT}/src/emu/x87emu_private.c"
-    "${BOX86_ROOT}/src/emu/x87run.c"
    "${BOX86_ROOT}/src/libtools/auxval.c"
    "${BOX86_ROOT}/src/libtools/myalign.c"
    "${BOX86_ROOT}/src/libtools/myalign64.c"
@@ -324,6 +323,33 @@ if(NOT ANDROID)
    list(APPEND ELFLOADER_SRC "${BOX86_ROOT}/src/libtools/obstack.c")
 endif()

+set(INTERPRETER
+    "${BOX86_ROOT}/src/emu/x86run.c"
+    "${BOX86_ROOT}/src/emu/x86run0f.c"
+    "${BOX86_ROOT}/src/emu/x86run64.c"
+    "${BOX86_ROOT}/src/emu/x86run640f.c"
+    "${BOX86_ROOT}/src/emu/x86run6466.c"
+    "${BOX86_ROOT}/src/emu/x86run6467.c"
+    "${BOX86_ROOT}/src/emu/x86run66.c"
+    "${BOX86_ROOT}/src/emu/x86run660f.c"
+    "${BOX86_ROOT}/src/emu/x86run66d9.c"
+    "${BOX86_ROOT}/src/emu/x86run66dd.c"
+    "${BOX86_ROOT}/src/emu/x86run67.c"
+    "${BOX86_ROOT}/src/emu/x86run6766.c"
+    "${BOX86_ROOT}/src/emu/x86rund8.c"
+    "${BOX86_ROOT}/src/emu/x86rund9.c"
+    "${BOX86_ROOT}/src/emu/x86runda.c"
+    "${BOX86_ROOT}/src/emu/x86rundb.c"
+    "${BOX86_ROOT}/src/emu/x86rundc.c"
+    "${BOX86_ROOT}/src/emu/x86rundd.c"
+    "${BOX86_ROOT}/src/emu/x86runde.c"
+    "${BOX86_ROOT}/src/emu/x86rundf.c"
+    "${BOX86_ROOT}/src/emu/x86runf0.c"
+    "${BOX86_ROOT}/src/emu/x86runf066.c"
+    "${BOX86_ROOT}/src/emu/x86runf20f.c"
+    "${BOX86_ROOT}/src/emu/x86runf30f.c"
+)
+
 set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedlibc.c"
    "${BOX86_ROOT}/src/wrapped/wrappedlibpthread.c"
@@ -492,6 +518,7 @@ set(WRAPPEDS
    "${BOX86_ROOT}/src/wrapped/wrappedpcap.c"
    "${BOX86_ROOT}/src/wrapped/wrappedfaudio.c"
    "${BOX86_ROOT}/src/wrapped/wrappedtbbmallocproxy.c"
+    "${BOX86_ROOT}/src/wrapped/wrappedlibfuse.c"
 )
 if(ANDROID)
    list(APPEND WRAPPEDS
@@ -545,7 +572,9 @@ if(ARM_DYNAREC)

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

        "${BOX86_ROOT}/src/dynarec/arm_prolog.S"
        "${BOX86_ROOT}/src/dynarec/arm_epilog.S"
@@ -593,6 +622,8 @@ if(ARM_DYNAREC)
    set_target_properties(arm_pass2 PROPERTIES COMPILE_FLAGS "-DSTEP=2")
    add_library(arm_pass3 OBJECT ${DYNAREC_PASS})
    set_target_properties(arm_pass3 PROPERTIES COMPILE_FLAGS "-DSTEP=3")
+    add_library(test_interpreter OBJECT ${INTERPRETER})
+    set_target_properties(test_interpreter PROPERTIES COMPILE_FLAGS "-DTEST_INTERPRETER")
    add_dependencies(arm_pass0 WRAPPERS)
    add_dependencies(arm_pass1 WRAPPERS)
    add_dependencies(arm_pass2 WRAPPERS)
@@ -604,6 +635,7 @@ if(ARM_DYNAREC)
        $<TARGET_OBJECTS:arm_pass1>
        $<TARGET_OBJECTS:arm_pass2>
        $<TARGET_OBJECTS:arm_pass3>
+        $<TARGET_OBJECTS:test_interpreter>
    )

 endif()
@@ -613,16 +645,18 @@ if(ARM_DYNAREC)
    add_custom_command(
        OUTPUT "${BOX86_ROOT}/src/git_head.h"
        COMMAND bash -c "echo \\\#define GITREV \\\"$(git rev-parse --short HEAD)\\\">\"${BOX86_ROOT}/src/git_head.h\""
-        DEPENDS dynarec ${ELFLOADER_SRC} ${WRAPPEDS}
+        DEPENDS dynarec ${ELFLOADER_SRC} ${INTERPRETER} ${WRAPPEDS}
        VERBATIM)
 else()
    add_custom_command(
        OUTPUT "${BOX86_ROOT}/src/git_head.h"
        COMMAND bash -c "echo \\\#define GITREV \\\"$(git rev-parse --short HEAD)\\\">\"${BOX86_ROOT}/src/git_head.h\""
-        DEPENDS ${ELFLOADER_SRC} ${WRAPPEDS}
+        DEPENDS ${ELFLOADER_SRC} ${INTERPRETER} ${WRAPPEDS}
        VERBATIM)
 endif()

+add_library(interpreter OBJECT ${INTERPRETER})
+
 if(BOX86LIB)
 #build a static or dynamic lib...
 if(BOX86DYNAMIC)
@@ -632,9 +666,9 @@ else()
 endif()
 if(ARM_DYNAREC)
    if(ANDROID)
-        target_link_libraries(${BOX86} dynarec m dl)
+        target_link_libraries(${BOX86} dynarec interpreter m dl)
    else()
-        target_link_libraries(${BOX86} dynarec m dl rt pthread resolv)
+        target_link_libraries(${BOX86} dynarec interpreter m dl rt pthread resolv)
    endif()
 endif()
 if(${CMAKE_VERSION} VERSION_LESS "3.13")
@@ -660,7 +694,9 @@ else()
    set_target_properties(${BOX86} PROPERTIES LINK_FLAGS "-Wl,--no-as-needed -lc -lm -ldl -lrt -lpthread -lresolv -Wl,--as-needed -pthread")
 endif()
 if(ARM_DYNAREC)
-    target_link_libraries(${BOX86} dynarec)
+    target_link_libraries(${BOX86} dynarec interpreter)
+else()
+    target_link_libraries(${BOX86} interpreter)
 endif()

 if(${CMAKE_VERSION} VERSION_LESS "3.13")
@@ -679,13 +715,17 @@ endif()
 if(NOT _x86 AND NOT _x86_64)
  install(TARGETS ${BOX86}
    RUNTIME DESTINATION bin)
-  configure_file(system/box86.conf.cmake system/box86.conf)
-  install(FILES ${CMAKE_BINARY_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
-  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.6 DESTINATION /usr/lib/i386-linux-gnu/)
-  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.5 DESTINATION /usr/lib/i386-linux-gnu/)
-  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libgcc_s.so.1 DESTINATION /usr/lib/i386-linux-gnu/)
-  install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libpng12.so.0 DESTINATION /usr/lib/i386-linux-gnu/)
-  install(FILES ${CMAKE_SOURCE_DIR}/system/box86.box86rc DESTINATION /etc/)
+  if(NOT NO_CONF_INSTALL)
+    configure_file(system/box86.conf.cmake system/box86.conf)
+    install(FILES ${CMAKE_BINARY_DIR}/system/box86.conf DESTINATION /etc/binfmt.d/)
+    install(FILES ${CMAKE_SOURCE_DIR}/system/box86.box86rc DESTINATION /etc/)
+  endif()
+  if(NOT NO_LIB_INSTALL)
+    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.6 DESTINATION /usr/lib/i386-linux-gnu/)
+    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libstdc++.so.5 DESTINATION /usr/lib/i386-linux-gnu/)
+    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libgcc_s.so.1 DESTINATION /usr/lib/i386-linux-gnu/)
+    install(FILES ${CMAKE_SOURCE_DIR}/x86lib/libpng12.so.0 DESTINATION /usr/lib/i386-linux-gnu/)
+  endif()
 endif()

 if(NOT TARGET uninstall)
@@ -6,7 +6,8 @@
 override_dh_auto_configure:
 	dh_auto_configure -- \
 	      -DCMAKE_BUILD_TYPE=RelWithDebInfo \
-	      -DARM_DYNAREC=1
+	      -DARM_DYNAREC=1 \
+		  -DNOGIT=1

 override_dh_shlibdeps:
 	dh_shlibdeps --exclude=libgcc_s.so.1 --exclude=libpng12.so.0 --exclude=libstdc++.so.5 --exclude=libstdc++.so.6 
@@ -6,7 +6,7 @@ You can use [@Itai-Nelken](https://github.com/Itai-Nelken)'s apt repository to i

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

@@ -117,7 +117,7 @@ Also, on armbian, you may need to install `libc6-dev-armhf-cross` or you may hav

 #### for Other ARM64 64bits Linux platform

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

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

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

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

@@ -127,6 +127,11 @@ Enables/Disables trace for generated code.
 Forbid dynablock creation in the interval specified (helpfull for debugging behaviour difference between Dynarec and Interpretor)
 * 0xXXXXXXXX-0xYYYYYYYY : define the interval where dynablock cannot start (inclusive-exclusive)

+#### BOX86_DYNAREC_TEST *
+Dynarec will compare it's execution with the interpreter (super slow, only for testing)
+ * 0 : No comparison
+ * 1 : Each opcode runs on interepter and on Dynarec, and regs and memory are compared and print if different
+
 #### BOX86_DYNAREC_BIGBLOCK *
 Enables/Disables Box86's Dynarec building BigBlock.
 * 0 : Don't try to build block as big as possible (can help program using lots of thread and a JIT, like C#/Unity) (Default when libmonobdwgc-2.0.so is loaded)
@@ -170,11 +175,21 @@ Handling of HotPage (Page beeing both executed and writen)
 * 0 : Don't track hotpage
 * 1-255 : Trak HotPage, and disable execution of a page beeing writen for N attempts (default is 16)

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

+#### BOX86_DYNAREC_WAIT *
+Behavior with FillBlock is not availble (FillBlock build Dynarec blocks and is not multithreaded)
+* 0 : Dynarec will not wait for FillBlock to ready and use Interpreter instead (might speedup a bit massive multithread or JIT programs)
+* 1 : Dynarec will wait for FillBlock to be ready (Default)
+
 #### BOX86_SSE_FLUSHTO0 *
 Handling of SSE Flush to 0 flags
 * 0 : Just track the flag (Default)
@@ -28,14 +28,21 @@ export DBUS_FATAL_WARNINGS=0
 chmod +x steam
 sudo mv steam /usr/local/bin/

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

@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
-pkgname=box86-rk3399-git
-pkgver=r3858.5368285
+pkgname=box86-a64-git
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -9,21 +9,28 @@ url="https://github.com/ptitSeb/box86"
 license=('MIT')
 optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
 makedepends=('git' 'cmake' 'make')
-provides=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
-conflicts=("${pkgname%-git}" "${pkgname%-rk3399-git}-git" "${pkgname%-rk3399-git}")
+provides=("${pkgname%-git}" "${pkgname%-a64-git}-git" "${pkgname%-a64-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-a64-git}-git" "${pkgname%-a64-git}")
 source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

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

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

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

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

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

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

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

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

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-gameshell-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-goaclone-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-goaclone-git}-git" "${pkgname%-goaclone-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-goaclone-git}-git" "${pkgname%-goaclone-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-goaclone-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-goaclone-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DGOA_CLONE=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-goaclone-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-goaclone-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-lx2610a-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-lx2610a-git}-git" "${pkgname%-lx2610a-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-lx2610a-git}-git" "${pkgname%-lx2610a-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DLX2610A=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-lx2610a-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-odroid-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

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

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

 package() {
-	make -C build DESTDIR="${pkgdir}/" install
-}
+	cd "${srcdir}/${pkgname%-odroid-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -0,0 +1,36 @@
+# Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
+# Author: Sebastien Chevalier <[email protected]>
+pkgname=box86-odroidn2-git
+pkgver=r4581.d698861
+pkgrel=1
+pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
+arch=('armv7h')
+url="https://github.com/ptitSeb/box86"
+license=('MIT')
+optdepends=('gl4es: OpenGL 2 for GLES 2 devices')
+makedepends=('git' 'cmake' 'make')
+provides=("${pkgname%-git}" "${pkgname%-odroidn2-git}-git" "${pkgname%-odroidn2-git}")
+conflicts=("${pkgname%-git}" "${pkgname%-odroidn2-git}-git" "${pkgname%-odroidn2-git}")
+source=('git+https://github.com/ptitSeb/box86')
+md5sums=('SKIP')
+
+pkgver() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}"
+	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+prepare() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}"
+	[[ ! -d ./build ]] && mkdir build
+	cmake -B build -DODROID=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_INSTALL_PREFIX=/usr
+}
+
+build() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}/build"
+	make -j$(nproc)
+}
+
+package() {
+	cd "${srcdir}/${pkgname%-odroidn2-git}/build"
+	make DESTDIR="${pkgdir}/" install
+}
@@ -1,7 +1,7 @@
 # Maintainer: Jai-JAP <[email protected]>, SpacingBat3 <[email protected]>
 # Author: Sebastien Chevalier <[email protected]>
 pkgname=box86-pandora-git
-pkgver=r3858.5368285
+pkgver=r4581.d698861
 pkgrel=1
 pkgdesc="Linux Userspace x86 Emulator with a twist, targeted at ARM Linux devices."
 arch=('armv7h')
@@ -15,15 +15,22 @@ source=('git+https://github.com/ptitSeb/box86')
 md5sums=('SKIP')

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

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

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

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

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

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

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

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

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

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

 pkgver() {
-	cd "$srcdir/${pkgname%-x86-git}"
+	cd "${srcdir}/${pkgname%-x86-git}"
 	printf "r%s.%s" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
 }

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

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

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

 ### 3. Any advantages with that sollution?
 This might give you some advantages like those:
@@ -23,6 +23,7 @@
 #include "custommem.h"
 #include "dictionnary.h"
 #include "rcfile.h"
+#include "gltools.h"
 #ifdef DYNAREC
 #include "dynablock.h"
 #include "dynarec/arm_lock_helper.h"
@@ -146,6 +147,7 @@ static void init_mutexes(box86context_t* context)
    arm_lock_stored(&context->mutex_tls, 0);
    arm_lock_stored(&context->mutex_thread, 0);
    arm_lock_stored(&context->mutex_bridge, 0);
+    pthread_mutex_init(&context->mutex_lock, NULL);
 #else
    pthread_mutexattr_t attr;
    pthread_mutexattr_init(&attr);
@@ -222,9 +224,12 @@ box86context_t *NewBox86Context(int argc)
    context->system = NewBridge();
    context->globaldefver = NewDefaultVersion();
    context->weakdefver = NewDefaultVersion();
+    // Cannot use Bridge name as the map is not initialized yet
    // create vsyscall
    context->vsyscall = AddBridge(context->system, iFEv, x86Syscall, 0, NULL);
    addAlternate((void*)0xffffe400, (void*)context->vsyscall);
+    // create exit bridge
+    context->exit_bridge = AddBridge(context->system, NULL, NULL, 0, NULL);
 #ifdef BUILD_LIB
    context->box86lib = RTLD_DEFAULT;   // not ideal
 #else
@@ -250,7 +255,7 @@ box86context_t *NewBox86Context(int argc)

    return context;
 }
-
+void freeALProcWrapper(box86context_t* context);
 EXPORTDYN
 void FreeBox86Context(box86context_t** context)
 {
@@ -359,7 +364,9 @@ void FreeBox86Context(box86context_t** context)

    finiAllHelpers(ctx);

-#ifndef DYNAREC
+#ifdef DYNAREC
+    pthread_mutex_destroy(&ctx->mutex_lock);
+#else
    pthread_mutex_destroy(&ctx->mutex_once);
    pthread_mutex_destroy(&ctx->mutex_once2);
    pthread_mutex_destroy(&ctx->mutex_trace);
@@ -3,6 +3,6 @@

 #define BOX86_MAJOR 0
 #define BOX86_MINOR 3
-#define BOX86_REVISION 0
+#define BOX86_REVISION 1

 #endif //__BOX86_VERSION_H_
@@ -390,6 +390,8 @@ Op is 20-27
 #define LDRD_IMM8(reg, addr, imm8) EMIT(c__ | 0b000<<25 | 1<<24  | (((imm8)<0)?0:1)<<23 | 1<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | ((abs(imm8))&0xf0)<<(8-4) | (0b1101<<4) | ((abs(imm8))&0x0f) )
 // ldrd reg, reg+1, [addr, +rm], reg must be even, reg+1 is implicit
 #define LDRD_REG(reg, addr, rm) EMIT(c__ | 0b000<<25 | 1<<24  | 1<<23 | 0<<22 | 0<<21 | 0<<20 | ((reg) << 12) | ((addr) << 16) | 0b1101<<4 | (rm) )
+// ldr reg, [pc+imm]
+#define LDR_literal(reg, imm12) EMIT(c__ | 0b010<<25 | 1<<24 | (((imm12)>=0)?1:0)<<23 | 0<<21 | 1<<20 | 0b1111<<16 | (reg)<<12 | (abs(imm12)))

 // str reg, [addr, #+/-imm9]
 #define STR_IMM9(reg, addr, imm9) EMIT(0xe5000000 | (((imm9)<0)?0:1)<<23 | ((reg) << 12) | ((addr) << 16) | brIMM(imm9) )
@@ -9,6 +9,7 @@
 .extern LinkNext

 .global arm_next
+    .4byte  0   // NULL pointer before arm64_next, for getDB
 arm_next:
    // emu is r0
    //stm     r0, {r4-r12,r14}   // don't put put back reg value in emu, faster but more tricky to debug
@@ -41,21 +41,37 @@ uint32_t X31_hash_code(void* addr, int len)
 	return (uint32_t)h;
 }

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

 void FreeDynablock(dynablock_t* db, int need_lock)
@@ -63,237 +79,100 @@ void FreeDynablock(dynablock_t* db, int need_lock)
    if(db) {
        if(db->gone)
            return; // already in the process of deletion!
-        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x86=%p:%p father=%p, with %d son(s) already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size, db->father, db->sons_size, db->gone);
+        dynarec_log(LOG_DEBUG, "FreeDynablock(%p), db->block=%p x86=%p:%p already gone=%d\n", db, db->block, db->x86_addr, db->x86_addr+db->x86_size-1, db->gone);
        if(need_lock)
            mutex_lock(&my_context->mutex_dyndump);
-        db->done = 0;
-        db->gone = 1;
-        // remove from direct if there
-        uintptr_t startdb = db->parent->text;
-        uintptr_t enddb = db->parent->text + db->parent->textsz;
-        if(db->parent->direct) {
-            uintptr_t addr = (uintptr_t)db->x86_addr;
-            if(addr>=startdb && addr<enddb)
-                db->parent->direct[addr-startdb] = NULL;
-        }
        // remove jumptable
        setJumpTableDefault(db->x86_addr);
-        // remove and free the sons
-        for (int i=0; i<db->sons_size; ++i) {
-            dynablock_t *son = (dynablock_t*)arm_lock_xchg(&db->sons[i], 0);
-            FreeDynablock(son, 0);
-        }
-        // only the father free the DynarecMap
-        if(!db->father) {
-            dynarec_log(LOG_DEBUG, " -- FreeDyrecMap(%p, %d)\n", db->block, db->size);
-            FreeDynarecMap(db, (uintptr_t)db->block, db->size);
-            box_free(db->sons);
-            box_free(db->instsize);
-        }
-        box_free(db);
+        dynarec_log(LOG_DEBUG, " -- FreeDyrecMap(%p, %d)\n", db->actual_block, db->size);
+        db->done = 0;
+        db->gone = 1;
+        if(db->previous)
+            FreeInvalidDynablock(db->previous, 0);
+        FreeDynarecMap((uintptr_t)db->actual_block);
+        customFree(db);
        if(need_lock)
            mutex_unlock(&my_context->mutex_dyndump);
    }
 }

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

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

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

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

-void MarkDirectDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+static int MarkedDynablock(dynablock_t* db)
 {
-    if(!dynablocks)
-        return;
-    if(!dynablocks->direct)
-        return;
-    uintptr_t startdb = dynablocks->text;
-    uintptr_t enddb = startdb + dynablocks->textsz -1;
-    uintptr_t start = addr;
-    uintptr_t end = addr+size-1;
-    if(start<startdb)
-        start = startdb;
-    if(end>enddb)
-        end = enddb;
-    dynablock_t *db;
-    if(end>startdb && start<enddb)
-        for(uintptr_t i = start; i<end; ++i)
-            if((db=dynablocks->direct[i-startdb]))
-                if(IntervalIntersects((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1))
-                    MarkDynablock(db);
-}
-
-void FreeRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
-{
-    if(!dynablocks)
-        return;
-
-    if(dynablocks->direct) {
-        int need_lock = my_context?1:0;
-        dynablock_t* db;
-        int ret;
-        khint_t k;
-        kh_dynablocks_t *blocks = kh_init(dynablocks);
-        // copy in a temporary list
-        if(dynablocks->direct) {
-            uintptr_t startdb = dynablocks->text;
-            uintptr_t enddb = startdb + dynablocks->textsz;
-            uintptr_t start = addr;
-            uintptr_t end = addr+size;
-            if(start<startdb)
-                start = startdb;
-            if(end>enddb)
-                end = enddb;
-            if(end>startdb && start<enddb)
-                for(uintptr_t i = start; i<end; ++i) {
-                    db = (dynablock_t*)arm_lock_xchg(&dynablocks->direct[i-startdb], 0);
-                    if(db) {
-                        if(db->father)
-                            db = db->father;
-                        if(db->parent==dynablocks) {
-                            k = kh_put(dynablocks, blocks, (uintptr_t)db, &ret);
-                            kh_value(blocks, k) = db;
-                        }
-                    }
-                }
-        }
-        // purge the list
-        kh_foreach_value(blocks, db,
-            FreeDynablock(db, need_lock);
-        );
-        kh_destroy(dynablocks, blocks);
+    if(db) {
+        if(getNeedTest((uintptr_t)db->x86_addr))
+            return 1; // already done
    }
+    return 0;
 }
-void MarkRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size)
+
+void MarkRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
 {
-    if(!dynablocks)
+    // Mark will try to find *any* blocks that intersect the range to mark
+    if(!db)
        return;
-    if(dynablocks->direct) {
-        uintptr_t new_addr = addr - dynablocks->maxsz;
-        uintptr_t new_size = size + dynablocks->maxsz;
-        MarkDirectDynablock(dynablocks, new_addr, new_size);
-        // the blocks check before
-        for(int idx=(new_addr)>>DYNAMAP_SHIFT; idx<(addr>>DYNAMAP_SHIFT); ++idx)
-            MarkDirectDynablock(getDB(idx), new_addr, new_size);
+    dynarec_log(LOG_DEBUG, "MarkRangeDynablock %p-%p .. startdb=%p, sizedb=%p\n", (void*)addr, (void*)addr+size-1, (void*)db->x86_addr, (void*)db->x86_size);
+    if(IntervalIntersects((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1))
+        MarkDynablock(db);
+}
+
+int FreeRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size)
+{
+    if(!db)
+        return 1;
+
+    int need_lock = my_context?1:0;
+    if(IntervalIntersects((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr+db->x86_size-1, addr, addr+size+1)) {
+        FreeDynablock(db, need_lock);
+        return 0;
    }
+    return 1;
 }

-dynablock_t* FindDynablockDynablocklist(void* addr, kh_dynablocks_t* dynablocks)
+dynablock_t *AddNewDynablock(uintptr_t addr)
 {
-    if(!dynablocks)
-        return NULL;
-    dynablock_t* db;
-    kh_foreach_value(dynablocks, db, 
-        const uintptr_t s = (uintptr_t)db->block;
-        const uintptr_t e = (uintptr_t)db->block+db->size;
-        if((uintptr_t)addr>=s && (uintptr_t)addr<e)
-            return db->father?db->father:db;
-    )
-    return NULL;
-}
-
-static dynablocklist_t* getDBFromAddress(uintptr_t addr)
-{
-    const int idx = (addr>>DYNAMAP_SHIFT);
-    return getDB(idx);
-}
-
-dynablock_t *AddNewDynablock(dynablocklist_t* dynablocks, uintptr_t addr, int* created)
-{
-    if(!dynablocks) {
-        dynarec_log(LOG_INFO, "Warning: Ask to create a dynablock with a NULL dynablocklist (addr=%p)\n", (void*)addr);
-        *created = 0;
+    dynablock_t* block;
+    #if 0
+    // check if memory as the correct flags
+    int prot = getProtection(addr);
+    if(!(prot&(PROT_EXEC|PROT_DYNAREC|PROT_DYNAREC_R))) {
+        dynarec_log(LOG_VERBOSE, "Block asked on a memory with no execution flags 0x%02X\n", prot);
        return NULL;
    }
-    if((addr<dynablocks->text) || (addr>=(dynablocks->text+dynablocks->textsz))) {
-        // this should be useless
-        //dynarec_log(LOG_INFO, "Warning: Refused to create a Direct Block that is out-of-bound: dynablocks=%p (%p:%p), addr=%p\n", dynablocks, (void*)(dynablocks->text), (void*)(dynablocks->text+dynablocks->textsz), (void*)addr);
-        //*created = 0;
-        //return NULL;
-        return AddNewDynablock(getDBFromAddress(addr), addr, created);
-    }
-    dynablock_t* block = NULL;
-    // first, check if it exist in direct access mode
-    if(dynablocks->direct) {
-        block = dynablocks->direct[addr-dynablocks->text];
-        if(block) {
-            dynarec_log(LOG_VERBOSE, "Block already exist in Direct Map\n");
-            *created = 0;
-            return block;
-        }
-    }
    
-    if (!*created)
-        return block;
-    
-    if(!dynablocks->direct) {
-        dynablock_t** p = (dynablock_t**)box_calloc(dynablocks->textsz, sizeof(dynablock_t*));
-        if(arm_lock_storeifnull(&dynablocks->direct, p)!=NULL)
-            box_free(p);    // someone already create the direct array, too late...
-    }
-
+    #endif
    // create and add new block
    dynarec_log(LOG_VERBOSE, "Ask for DynaRec Block creation @%p\n", (void*)addr);
-
-    block = (dynablock_t*)box_calloc(1, sizeof(dynablock_t));
-    block->parent = dynablocks; 
-    dynablock_t* tmp = (dynablock_t*)arm_lock_storeifref(&dynablocks->direct[addr-dynablocks->text], block, NULL);
-    if(tmp !=  block) {
-        // a block appeard!
-        box_free(block);
-        *created = 0;
-        return tmp;
-    }
-    *created = 1;
+    block = (dynablock_t*)customCalloc(1, sizeof(dynablock_t));
    return block;
 }

@@ -309,121 +188,120 @@ void cancelFillBlock()
    return NULL if block is not found / cannot be created. 
    Don't create if create==0
 */
-static dynablock_t* internalDBGetBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, dynablock_t* current, int need_lock)
+static dynablock_t* internalDBGetBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr, int create, int need_lock)
 {
    if(hasAlternate((void*)addr))
        return NULL;
-    // try the quickest way first: get parent of current and check if ok!
-    dynablocklist_t *dynablocks = NULL;
-    dynablock_t* block = NULL;
-    if(current && current->done) {
-        dynablocks = current->parent;
-        if(dynablocks && !(addr>=dynablocks->text && addr<(dynablocks->text+dynablocks->textsz)))
-            dynablocks = NULL;
-    }
-    // nope, lets do the long way
-    if(!dynablocks) {
-        dynablocks = getDBFromAddress(addr);
-        if(!dynablocks) {
-            dynablocks = GetDynablocksFromAddress(emu->context, addr);
-            if(!dynablocks)
+    dynablock_t* block = getDB(addr);
+    if(block || !create)
+        return block;
+
+    if(need_lock) {
+        if(box86_dynarec_wait) {
+            mutex_lock(&my_context->mutex_dyndump);
+        } else {
+            if(mutex_trylock(&my_context->mutex_dyndump))   // FillBlock not available for now
                return NULL;
        }
-    }
-    // check direct first, without lock
-    if(dynablocks->direct/* && (addr>=dynablocks->text) && (addr<(dynablocks->text+dynablocks->textsz))*/)
-        if((block = dynablocks->direct[addr-dynablocks->text]))
-            return block;
-
-    int created = create;
-    if(need_lock)
-        mutex_lock(&my_context->mutex_dyndump);
-    block = AddNewDynablock(dynablocks, addr, &created);
-    if(!created) {
-        if(need_lock)
+        block = getDB(addr);    // just in case
+        if(block) {
            mutex_unlock(&my_context->mutex_dyndump);
-        return block;   // existing block...
+            return block;
+        }
    }
+    
+    block = AddNewDynablock(addr);

    // fill the block
    block->x86_addr = (void*)addr;
    if(sigsetjmp(&dynarec_jmpbuf, 1)) {
        printf_log(LOG_INFO, "FillBlock at %p triggered a segfault, cancelling\n", (void*)addr);
+        FreeDynablock(block, 0);
        if(need_lock)
            mutex_unlock(&my_context->mutex_dyndump);
        return NULL;
    }
    void* ret = FillBlock(block, filladdr);
-    if(need_lock)
-        mutex_unlock(&my_context->mutex_dyndump);
    if(!ret) {
        dynarec_log(LOG_DEBUG, "Fillblock of block %p for %p returned an error\n", block, (void*)addr);
-        void* old = (void*)arm_lock_storeifref(&dynablocks->direct[addr-dynablocks->text], 0, block);
-        if(old!=block && old) { // put it back in place, strange things are happening here!
-            dynarec_log(LOG_INFO, "Warning, a wild block appeared at %p: %p\n", (void*)addr, old);
-            // Doing nothing else, the block has not been written
-        }
-        box_free(block);
+        customFree(block);
        block = NULL;
    }
    // check size
-    if(block && block->x86_size) {
+    if(block) {
        int blocksz = block->x86_size;
-        if(dynablocks->maxsz<blocksz) {
-            dynablocks->maxsz = blocksz;
-            for(int idx=(addr>>DYNAMAP_SHIFT)+1; idx<=((addr+blocksz)>>DYNAMAP_SHIFT); ++idx) {
-                dynablocklist_t* dblist;
-                if((dblist = getDB(idx)))
-                    if(dblist->maxsz<blocksz)
-                        dblist->maxsz = blocksz;
-            }
-        }
-        //protectDB((uintptr_t)block->x86_addr, block->x86_size);   // already protected in FillBlock
-        // but check if it already need test
-        if(block->need_test || (block->father && block->father->need_test))
-            return NULL;
+        if(blocksz>my_context->max_db_size)
+            my_context->max_db_size = blocksz;
        // fill-in jumptable
-        addJumpTableIfDefault(block->x86_addr, block->block);
-        for(int i=0; i<block->sons_size; ++i)
-            addJumpTableIfDefault(block->sons[i]->x86_addr, block->sons[i]->block);
+        if(!addJumpTableIfDefault(block->x86_addr, block->dirty?block->jmpnext:block->block)) {
+            FreeDynablock(block, 0);
+            block = getDB(addr);
+            MarkDynablock(block);   // just in case...
+        } else {
+            if(block->x86_size)
+                block->done = 1;    // don't validate the block if the size is null, but keep the block
+        }
    }
+    if(need_lock)
+        mutex_unlock(&my_context->mutex_dyndump);

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

    return block;
 }

-dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create, dynablock_t** current)
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create)
 {
-    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, *current, 1);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
+    dynablock_t *db = internalDBGetBlock(emu, addr, addr, create, 1);
+    if(db && db->done && db->block && getNeedTest(addr)) {
+        if(db->always_test)
+            sched_yield();  // just calm down...
        if(AreaInHotPage((uintptr_t)db->x86_addr, (uintptr_t)db->x86_addr + db->x86_size - 1)) {
-            dynarec_log(LOG_DEBUG, "Not running block %p from %p:%p with for %p because it's in a hotpage\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, (void*)addr);
-            return NULL;
+            emu->test.test = 0;
+            if(box86_dynarec_fastpage) {
+                uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+                if(hash==db->hash) { // seems ok, run it without reprotecting it
+                    setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+                    return db;
+                }
+                db->done = 0;   // invalidating the block, it's already not good
+                dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, hash, db->hash, (void*)addr);
+                // Free db, it's now invalid!
+                FreeDynablock(db, 1);
+                return NULL;    // not building a new one, it's still a hotpage
+            } else {
+                dynarec_log(LOG_INFO, "Not running block %p from %p:%p with for %p because it's in a hotpage\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, (void*)addr);
+                return NULL;
+            }
        }
-        if(mutex_trylock(&my_context->mutex_dyndump))
-            return NULL;
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = (getProtection((uintptr_t)father->x86_addr)&PROT_READ)?X31_hash_code(father->x86_addr, father->x86_size):0;
-        if(hash!=father->hash) {
-            dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) with %d son(s) for %p\n", father, father->x86_addr, father->x86_addr+father->x86_size, hash, father->hash, father->sons_size, (void*)addr);
-            // no more current if it gets invalidated too
-            if(*current && father->x86_addr>=(*current)->x86_addr && (father->x86_addr+father->x86_size)<(*current)->x86_addr)
-                *current = NULL;
-            // Free father, it's now invalid!
-            FreeDynablock(father, 0);
+        uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+        int need_lock = mutex_trylock(&my_context->mutex_dyndump);
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            dynablock_t* old = InvalidDynablock(db, need_lock);
            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, addr, create, *current, 0);
+            db = internalDBGetBlock(emu, addr, addr, create, need_lock);
+            if(db) {
+                if(db->previous)
+                    FreeInvalidDynablock(db->previous, need_lock);
+                db->previous = old;
+            } else
+                FreeInvalidDynablock(old, need_lock);
        } else {
-            father->need_test = 0;
-            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            dynarec_log(LOG_DEBUG, "Validating block %p from %p:%p (hash:%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size-1, db->hash, (void*)addr);
+            protectDB((uintptr_t)db->x86_addr, db->x86_size);
            // fill back jumptable
-            addJumpTableIfDefault(father->x86_addr, father->block);
-            for(int i=0; i<father->sons_size; ++i)
-                addJumpTableIfDefault(father->sons[i]->x86_addr, father->sons[i]->block);
+            if(isprotectedDB((uintptr_t)db->x86_addr, db->x86_size) && !db->always_test) {
+                setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+            }
        }
-        mutex_unlock(&my_context->mutex_dyndump);
+        if(!need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    } 
+    if(!db || !db->block || !db->done)
+        emu->test.test = 0;
    return db;
 }

@@ -431,27 +309,36 @@ dynablock_t* DBAlternateBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr)
 {
    dynarec_log(LOG_DEBUG, "Creating AlternateBlock at %p for %p\n", (void*)addr, (void*)filladdr);
    int create = 1;
-    dynablock_t *db = internalDBGetBlock(emu, addr, filladdr, create, NULL, 1);
-    if(db && db->done && db->block && (db->need_test || (db->father && db->father->need_test))) {
-        if(mutex_trylock(&my_context->mutex_dyndump))
-            return NULL;
-        dynablock_t *father = db->father?db->father:db;
-        uint32_t hash = (getProtection((uintptr_t)father->x86_addr)&PROT_READ)?X31_hash_code(father->x86_addr, father->x86_size):0;
-        if(hash!=father->hash) {
-            dynarec_log(LOG_DEBUG, "Invalidating alt block %p from %p:%p (hash:%X/%X) with %d son(s) for %p\n", father, father->x86_addr, father->x86_addr+father->x86_size, hash, father->hash, father->sons_size, (void*)addr);
-            // Free father, it's now invalid!
-            FreeDynablock(father, 0);
+    dynablock_t *db = internalDBGetBlock(emu, addr, filladdr, create, 1);
+    if(db && db->done && db->block && getNeedTest(filladdr)) {
+        if(db->always_test)
+            sched_yield();  // just calm down...
+        int need_lock = mutex_trylock(&my_context->mutex_dyndump);
+        uint32_t hash = X31_hash_code(db->x86_addr, db->x86_size);
+        if(hash!=db->hash) {
+            db->done = 0;   // invalidating the block
+            dynarec_log(LOG_DEBUG, "Invalidating alt block %p from %p:%p (hash:%X/%X) for %p\n", db, db->x86_addr, db->x86_addr+db->x86_size, hash, db->hash, (void*)addr);
+            // Free db, it's now invalid!
+            dynablock_t* old = InvalidDynablock(db, need_lock);
            // start again... (will create a new block)
-            db = internalDBGetBlock(emu, addr, filladdr, create, NULL, 0);
+            db = internalDBGetBlock(emu, addr, filladdr, create, need_lock);
+            if(db) {
+                if(db->previous)
+                    FreeInvalidDynablock(db->previous, need_lock);
+                db->previous = old;
+            } else
+                FreeInvalidDynablock(old, need_lock);
        } else {
-            father->need_test = 0;
-            protectDB((uintptr_t)father->x86_addr, father->x86_size);
+            protectDB((uintptr_t)db->x86_addr, db->x86_size);
            // fill back jumptable
-            addJumpTableIfDefault(father->x86_addr, father->block);
-            for(int i=0; i<father->sons_size; ++i)
-                addJumpTableIfDefault(father->sons[i]->x86_addr, father->sons[i]->block);
+            if(isprotectedDB((uintptr_t)db->x86_addr, db->x86_size) && !db->always_test) {
+                setJumpTableIfRef(db->x86_addr, db->block, db->jmpnext);
+            }
        }
-        mutex_unlock(&my_context->mutex_dyndump);
+        if(!need_lock)
+            mutex_unlock(&my_context->mutex_dyndump);
    } 
+    if(!db || !db->block || !db->done)
+        emu->test.test = 0;
    return db;
 }
@@ -1,36 +1,26 @@
 #ifndef __DYNABLOCK_PRIVATE_H_
 #define __DYNABLOCK_PRIVATE_H_

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

 typedef struct dynablock_s {
-    dynablocklist_t* parent;
-    void*           block;
+    void*           block;  // block-sizeof(void*) == self
+    void*           actual_block;   // the actual start of the block (so block-sizeof(void*))
+    struct dynablock_s*    previous;   // a previous block that might need to be freed
    int             size;
    void*           x86_addr;
    uintptr_t       x86_size;
    uint32_t        hash;
-    uint8_t         need_test;
    uint8_t         done;
    uint8_t         gone;
-    uint8_t         dummy;
+    uint8_t         always_test;
+    uint8_t         dirty;      // if need to be tested as soon as it's created
    int             isize;
-    dynablock_t**   sons;   // sons (kind-of dummy dynablock...)
-    int             sons_size;
-    dynablock_t*    father; // set only in the case of a son
    instsize_t*     instsize;
+    void*           jmpnext;    // a branch jmpnext code when block is marked
 } dynablock_t;

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

 #ifdef ARM
@@ -44,9 +45,8 @@ void* LinkNext(x86emu_t* emu, uintptr_t addr, void* x2)
        printf_log(LOG_NONE, "Warning, jumping to NULL address from %p (db=%p, x86addr=%p/%s)\n", x2, db, x86addr, pcname);
    }
    #endif
-    dynablock_t* current = NULL;
    void * jblock;
-    dynablock_t* block = DBGetBlock(emu, addr, 1, &current);
+    dynablock_t* block = DBGetBlock(emu, addr, 1);
    if(!block) {
        // no block, let link table as is...
        if(hasAlternate((void*)addr)) {
@@ -54,12 +54,15 @@ void* LinkNext(x86emu_t* emu, uintptr_t addr, void* x2)
            addr = (uintptr_t)getAlternate((void*)addr);
            R_EIP = addr;
            printf_log(LOG_INFO, " -> %p\n", (void*)addr);
-            block = DBGetBlock(emu, addr, 1, &current);
+            block = DBGetBlock(emu, addr, 1);
        }
        if(!block) {
            #ifdef HAVE_TRACE
-            dynablock_t* db = FindDynablockFromNativeAddress(x2);
-            dynarec_log(LOG_INFO, "Warning, jumping to a no-block address %p from %p (db=%p, x86addr=%p)\n", (void*)addr, x2, db, db?(void*)getX86Address(db, (uintptr_t)x2):NULL);
+            if(LOG_INFO<=box86_dynarec_log) {
+                dynablock_t* db = FindDynablockFromNativeAddress(x2);
+                elfheader_t* h = FindElfAddress(my_context, (uintptr_t)x2);
+                dynarec_log(LOG_INFO, "Warning, jumping to a no-block address %p from %p (db=%p, x64addr=%p(elf=%s))\n", (void*)addr, x2, db, db?(void*)getX86Address(db, (uintptr_t)x2-4):NULL, h?ElfName(h):"(none)");
+            }
            #endif
            //tableupdate(arm_epilog, addr, table);
            return arm_epilog;
@@ -84,18 +87,25 @@ void DynaCall(x86emu_t* emu, uintptr_t addr)
    // prepare setjump for signal handling
    emu_jmpbuf_t *ejb = NULL;
    int jmpbuf_reset = 0;
+    int skip = 0;
    if(emu->type == EMUTYPE_MAIN) {
        ejb = GetJmpBuf();
        if(!ejb->jmpbuf_ok) {
            ejb->emu = emu;
            ejb->jmpbuf_ok = 1;
            jmpbuf_reset = 1;
-            int a;
-            if((a=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
-                printf_log(LOG_DEBUG, "Setjmp DynaCall %d, fs=0x%x\n", a, ejb->emu->segs[_FS]);
+            if((skip=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
+                printf_log(LOG_DEBUG, "Setjmp DynaCall %d, fs=0x%x\n", skip, ejb->emu->segs[_FS]);
                addr = R_EIP;   // not sure if it should still be inside DynaCall!
-                if(a==2)
-                    Run(emu, 1);    // "single step" next instruction that is doing auto-smc
+                #ifdef DYNAREC
+                if(box86_dynarec_test) {
+                    if(emu->test.clean)
+                        x86test_check(emu, R_EIP);
+                    emu->test.clean = 0;
+                }
+                #endif
+                if(skip!=2)
+                    skip = 0;
            }
        }
    }
@@ -117,15 +127,18 @@ void DynaCall(x86emu_t* emu, uintptr_t addr)
        dynablock_t* block = NULL;
        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_EIP, 1, &current);
+            block = (skip==2)?NULL:DBGetBlock(emu, R_EIP, 1);
            current = block;
            if(!block || !block->block || !block->done) {
+                skip = 0;
                // no block, of block doesn't have DynaRec content (yet, temp is not null)
                // Use interpreter (should use single instruction step...)
                dynarec_log(LOG_DEBUG, "%04d|Calling Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
+                if(box86_dynarec_test)
+                    emu->test.clean = 0;
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "%04d|Calling DynaRec Block @%p (%p) of %d x86 instructions (father=%p) emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize ,block->father, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Calling DynaRec Block @%p (%p) of %d x86 instructions emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize ,emu);
                CHECK_FLAGS(emu);
                // block is here, let's run it!
                #ifdef ARM
@@ -173,6 +186,7 @@ int DynaRun(x86emu_t* emu)
 {
    // prepare setjump for signal handling
    emu_jmpbuf_t *ejb = NULL;
+    int skip = 0;
 #ifdef DYNAREC
    int jmpbuf_reset = 1;
 #endif
@@ -185,11 +199,17 @@ int DynaRun(x86emu_t* emu)
 #ifdef DYNAREC
            jmpbuf_reset = 1;
 #endif
-            if((a=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
-                printf_log(LOG_DEBUG, "Setjmp DynaRun %d, fs=0x%x\n", a, ejb->emu->segs[_FS]);
-                if(a==2)
-                    Run(emu, 1);    // "single step" next instruction that is doing auto-smc
-
+            if((skip=sigsetjmp((struct __jmp_buf_tag*)ejb->jmpbuf, 1))) {
+                printf_log(LOG_DEBUG, "Setjmp DynaRun %d, fs=0x%x\n", skip, ejb->emu->segs[_FS]);
+                #ifdef DYNAREC
+                if(box86_dynarec_test) {
+                    if(emu->test.clean)
+                        x86test_check(emu, R_EIP);
+                    emu->test.clean = 0;
+                }
+                #endif
+                if(skip!=2)
+                    skip = 0;
            }
        }
    }
@@ -202,15 +222,18 @@ int DynaRun(x86emu_t* emu)
        dynablock_t* block = NULL;
        dynablock_t* current = NULL;
        while(!emu->quit) {
-            block = DBGetBlock(emu, R_EIP, 1, &current);
+            block = (skip==2)?NULL:DBGetBlock(emu, R_EIP, 1);
            current = block;
            if(!block || !block->block || !block->done) {
+                skip = 0;
                // no block, of block doesn't have DynaRec content (yet, temp is not null)
                // Use interpreter (should use single instruction step...)
                dynarec_log(LOG_DEBUG, "%04d|Running Interpretor @%p, emu=%p\n", GetTID(), (void*)R_EIP, emu);
+                if(box86_dynarec_test)
+                    emu->test.clean = 0;
                Run(emu, 1);
            } else {
-                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x86 insts (father=%p) emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize, block->father, emu);
+                dynarec_log(LOG_DEBUG, "%04d|Running DynaRec Block @%p (%p) of %d x86 insts emu=%p\n", GetTID(), (void*)R_EIP, block->block, block->isize, emu);
                // block is here, let's run it!
                #ifdef ARM
                arm_prolog(emu, block->block);
@@ -73,8 +73,8 @@ void add_next(dynarec_arm_t *dyn, uintptr_t addr) {
        }
    // add slots
    if(dyn->next_sz == dyn->next_cap) {
-        dyn->next_cap += 16;
-        dyn->next = (uintptr_t*)box_realloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
+        dyn->next_cap += 64;
+        dyn->next = (uintptr_t*)customRealloc(dyn->next, dyn->next_cap*sizeof(uintptr_t));
    }
    dyn->next[dyn->next_sz++] = addr;
 }
@@ -83,11 +83,13 @@ uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr) {
    uintptr_t best = 0;
    int i = 0;
    while((i<dyn->next_sz) && (best!=addr)) {
-        if(dyn->next[i]<addr) { // remove the address, it's before current address
-            dyn->next[i] = 0;
-        } else {
-            if((dyn->next[i]<best) || !best)
-                best = dyn->next[i];
+        if(dyn->next[i]) {
+            if(dyn->next[i]<addr) { // remove the address, it's before current address
+                dyn->next[i] = 0;
+            } else {
+                if((dyn->next[i]<best) || !best)
+                    best = dyn->next[i];
+            }
        }
        ++i;
    }
@@ -241,35 +243,31 @@ int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n)
    return (i==n)?1:0;
 }

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

+
 static void fillPredecessors(dynarec_arm_t* dyn)
 {
    int pred_sz = 1;    // to be safe
@@ -288,7 +286,7 @@ static void fillPredecessors(dynarec_arm_t* dyn)
            ++dyn->insts[i+1].pred_sz;
        }
    }
-    dyn->predecessor = (int*)box_malloc(pred_sz*sizeof(int));
+    dyn->predecessor = (int*)customMalloc(pred_sz*sizeof(int));
    // fill pred pointer
    int* p = dyn->predecessor;
    for(int i=0; i<dyn->size; ++i) {
@@ -356,36 +354,70 @@ static int updateNeed(dynarec_arm_t* dyn, int ninst, uint8_t need) {
    return ninst;
 }

+void* current_helper = NULL;
+
+void CancelBlock(int need_lock)
+{
+    if(need_lock)
+        mutex_lock(&my_context->mutex_dyndump);
+    dynarec_arm_t* helper = (dynarec_arm_t*)current_helper;
+    current_helper = NULL;
+    if(helper) {
+        customFree(helper->next);
+        customFree(helper->insts);
+        customFree(helper->predecessor);
+        if(helper->dynablock && helper->dynablock->actual_block) {
+            FreeDynarecMap((uintptr_t)helper->dynablock->actual_block);
+            helper->dynablock->actual_block = NULL;
+        }
+    }
+    if(need_lock)
+        mutex_unlock(&my_context->mutex_dyndump);
+}
+
 uintptr_t arm_pass0(dynarec_arm_t* dyn, uintptr_t addr);
 uintptr_t arm_pass1(dynarec_arm_t* dyn, uintptr_t addr);
 uintptr_t arm_pass2(dynarec_arm_t* dyn, uintptr_t addr);
 uintptr_t arm_pass3(dynarec_arm_t* dyn, uintptr_t addr);
-
-void* current_helper = NULL;
-
-void CancelBlock()
-{
-    dynarec_arm_t* helper = (dynarec_arm_t*)current_helper;
-    current_helper = NULL;
-    if(!helper)
-        return;
-    box_free(helper->next);
-    box_free(helper->insts);
-    box_free(helper->predecessor);
-    if(helper->dynablock && helper->dynablock->block)
-        FreeDynarecMap(helper->dynablock, (uintptr_t)helper->dynablock->block, helper->dynablock->size);
-}
+void arm_epilog();
+void arm_next();

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

 void* FillBlock(dynablock_t* block, uintptr_t addr) {
+    /*
+        A Block must have this layout:
+
+        0x0000..0x0003  : dynablock_t* : self
+        0x0004..4+4*n   : actual Native instructions, (n is the total number)
+        B ..    B+3     : dynablock_t* : self (as part of JmpNext, that simulate another block)
+        B+4 ..  B+15    : 3 Native code for jmpnext (or jmp epilog in case of empty block)
+        B+16 .. B+19    : jmpnext (or jmp_epilog) address
+        B+20 .. B+31    : empty (in case an architecture needs more than 3 opcodes)
+        B+32 .. B+32+sz : instsize (compressed array with each instruction lenght on x86 and native side)
+
+    */
 dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    if(IsInHotPage(addr)) {
        dynarec_log(LOG_DEBUG, "Cancelling dynarec FillBlock on hotpage for %p\n", (void*)addr);
@@ -395,10 +427,6 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
        dynarec_log(LOG_DEBUG, "Asked to fill a block in fobidden zone\n");
        return CreateEmptyBlock(block, addr);
    }
-    if(!isJumpTableDefault((void*)addr)) {
-        dynarec_log(LOG_DEBUG, "Asked to fill a block at %p, but JumpTable is not default\n", (void*)addr);
-        return NULL;
-    }
    if(current_helper) {
        dynarec_log(LOG_DEBUG, "Cancelling dynarec FillBlock at %p as anothor one is going on\n", (void*)addr);
        return NULL;
@@ -412,23 +440,23 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    helper.start = addr;
    uintptr_t start = addr;
    helper.cap = 64; // needs epilog handling
-    helper.insts = (instruction_arm_t*)box_calloc(helper.cap, sizeof(instruction_arm_t));
+    helper.insts = (instruction_arm_t*)customCalloc(helper.cap, sizeof(instruction_arm_t));
    // pass 0, addresses, x86 jump addresses, overall size of the block
    uintptr_t end = arm_pass0(&helper, addr);
    // no need for next anymore
-    box_free(helper.next);
+    customFree(helper.next);
    helper.next_sz = helper.next_cap = 0;
    helper.next = NULL;
    // basic checks
    if(!helper.size) {
        dynarec_log(LOG_DEBUG, "Warning, null-sized dynarec block (%p)\n", (void*)addr);
-        CancelBlock();
-        return (void*)block;
+        CancelBlock(0);
+        return CreateEmptyBlock(block, addr);
    }
    if(!isprotectedDB(addr, 1)) {
        dynarec_log(LOG_INFO, "Warning, write on current page on pass0, aborting dynablock creation (%p)\n", (void*)addr);
        AddHotPage(addr);
-        CancelBlock();
+        CancelBlock(0);
        return NULL;
    }
    // already protect the block and compute hash signature
@@ -441,6 +469,8 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
        if(helper.insts[i].x86.jmp) {
            uintptr_t j = helper.insts[i].x86.jmp;
            if(j<start || j>=end || j==helper.insts[i].x86.addr) {
+                if(j==helper.insts[i].x86.addr) // if there is a loop on some opcode, make the block "always to tested"
+                    helper.always_test = 1;
                helper.insts[i].x86.jmp_insts = -1;
                helper.insts[i].x86.need_after |= X_PEND;
            } else {
@@ -468,29 +498,68 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
    // pass 2, instruction size
    arm_pass2(&helper, addr);
    // ok, now allocate mapped memory, with executable flag on
-    int sz = helper.arm_size;
-    void* p = (void*)AllocDynarecMap(block, sz);
-    if(p==NULL) {
-        dynarec_log(LOG_DEBUG, "AllocDynarecMap(%p, %d) failed, cancelling block\n", block, sz);
-        CancelBlock();
+    size_t insts_rsize = (helper.insts_size+2)*sizeof(instsize_t);
+    insts_rsize = (insts_rsize+7)&~7;   // round the size...
+    size_t arm_size = (helper.arm_size+7)&~7;   // round the size...
+    // ok, now allocate mapped memory, with executable flag on
+    size_t sz = sizeof(void*) + arm_size + 8*sizeof(void*) + insts_rsize;
+    //           dynablock_t* block (arm insts) jmpnext code    instsize
+    void* actual_p = (void*)AllocDynarecMap(sz);
+    void* p = (void*)(((uintptr_t)actual_p) + sizeof(void*));
+    void* next = p + arm_size;
+    void* instsize = next + 8*sizeof(void*);
+    if(actual_p==NULL) {
+        dynarec_log(LOG_INFO, "AllocDynarecMap(%p, %zu) failed, cancelling block\n", block, sz);
+        CancelBlock(0);
        return NULL;
    }
    helper.block = p;
+    block->actual_block = actual_p;
    helper.arm_start = (uintptr_t)p;
-    if(helper.sons_size) {
-        helper.sons_x86 = (uintptr_t*)alloca(helper.sons_size*sizeof(uintptr_t));
-        helper.sons_arm = (void**)alloca(helper.sons_size*sizeof(void*));
-    }
-    // pass 3, emit (log emit arm opcode)
+    helper.jmp_next = (uintptr_t)next+sizeof(void*);
+    helper.insts_size = 0;  // reset
+    helper.instsize = (instsize_t*)instsize;
+    *(dynablock_t**)actual_p = block;
+    // pass 3, emit (log emit native opcode)
    if(box86_dynarec_dump) {
        dynarec_log(LOG_NONE, "%s%04d|Emitting %d bytes for %d x86 bytes", (box86_dynarec_dump>1)?"\e[01;36m":"", GetTID(), helper.arm_size, helper.isize); 
        printFunctionAddr(helper.start, " => ");
        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");
    }
+    size_t oldarmsize = helper.arm_size;
    helper.arm_size = 0;
    arm_pass3(&helper, addr);
-    if(sz!=helper.arm_size) {
-        printf_log(LOG_NONE, "BOX86: Warning, size difference in block between pass2 (%d) & pass3 (%d)!\n", sz, helper.arm_size);
+    // keep size of instructions for signal handling
+    block->instsize = instsize;
+    // ok, free the helper now
+    customFree(helper.insts);
+    helper.insts = NULL;
+    helper.instsize = NULL;
+    customFree(helper.predecessor);
+    helper.predecessor = NULL;
+    block->size = sz;
+    block->isize = helper.size;
+    block->block = p;
+    block->jmpnext = next+sizeof(void*);
+    block->always_test = helper.always_test;
+    block->dirty = block->always_test;
+    *(dynablock_t**)next = block;
+    *(void**)(next+5*sizeof(void*)) = arm_next;
+    CreateJmpNext(block->jmpnext, next+5*sizeof(void*));
+    //block->x86_addr = (void*)start;
+    block->x86_size = end-start;
+    // all done...
+    __clear_cache(actual_p, actual_p+sz);   // need to clear the cache before execution...
+    block->hash = X31_hash_code(block->x86_addr, block->x86_size);
+    // Check if something changed, to abbort if it as
+    if((block->hash != hash)) {
+        dynarec_log(LOG_DEBUG, "Warning, a block changed while being processed hash(%p:%ld)=%x/%x\n", block->x86_addr, block->x86_size, block->hash, hash);
+        AddHotPage(addr);
+        CancelBlock(0);
+        return NULL;
+    }
+    if((oldarmsize!=helper.arm_size)) {
+        printf_log(LOG_NONE, "BOx86: Warning, size difference in block between pass2 (%zu) & pass3 (%zu)!\n", sz, helper.arm_size);
        uint8_t *dump = (uint8_t*)helper.start;
        printf_log(LOG_NONE, "Dump of %d x86 opcodes:\n", helper.size);
        for(int i=0; i<helper.size; ++i) {
@@ -500,76 +569,16 @@ dynarec_log(LOG_DEBUG, "Asked to Fill block %p with %p\n", block, (void*)addr);
            printf_log(LOG_NONE, "\t%d -> %d\n", helper.insts[i].size2, helper.insts[i].size);
        }
        printf_log(LOG_NONE, " ------------\n");
-    }
-    // all done...
-    __clear_cache(p, p+sz);   // need to clear the cache before execution...
-    // keep size of instructions for signal handling
-    {
-        size_t cap = 1;
-        for(int i=0; i<helper.size; ++i)
-            cap += 1 + ((helper.insts[i].x86.size>helper.insts[i].size)?helper.insts[i].x86.size:helper.insts[i].size)/15;
-        size_t size = 0;
-        block->instsize = (instsize_t*)box_calloc(cap, sizeof(instsize_t));
-        for(int i=0; i<helper.size; ++i)
-            block->instsize = addInst(block->instsize, &size, &cap, helper.insts[i].x86.size, helper.insts[i].size/4);
-        block->instsize = addInst(block->instsize, &size, &cap, 0, 0);    // add a "end of block" mark, just in case
-    }
-    // ok, free the helper now
-    box_free(helper.insts);
-    helper.insts = NULL;
-    box_free(helper.next);
-    helper.next = NULL;
-    block->size = sz;
-    block->isize = helper.size;
-    block->block = p;
-    block->need_test = 0;
-    //block->x86_addr = (void*)start;
-    block->x86_size = end-start;
-    if(box86_dynarec_largest<block->x86_size)
-        box86_dynarec_largest = block->x86_size;
-    block->hash = X31_hash_code(block->x86_addr, block->x86_size);
-    // Check if something changed, to abbort if it as
-    if(block->hash != hash) {
-        dynarec_log(LOG_INFO, "Warning, a block changed while beeing processed hash(%p:%d)=%x/%x\n", block->x86_addr, block->x86_size, block->hash, hash);
-        CancelBlock();
-        AddHotPage(addr);
+        CancelBlock(0);
        return NULL;
    }
    if(!isprotectedDB(addr, end-addr)) {
-        dynarec_log(LOG_DEBUG, "Warning, block unprotected while beeing processed %p:%zd, cancelling\n", block->x86_addr, block->x86_size);
+        dynarec_log(LOG_DEBUG, "Warning, block unprotected while being processed %p:%ld, marking as need_test\n", block->x86_addr, block->x86_size);
        AddHotPage(addr);
-        block->need_test = 1;
+        block->dirty = 1;
        //protectDB(addr, end-addr);
    }
-    // fill sons if any
-    dynablock_t** sons = NULL;
-    int sons_size = 0;
-    if(helper.sons_size) {
-        sons = (dynablock_t**)box_calloc(helper.sons_size, sizeof(dynablock_t*));
-        for (int i=0; i<helper.sons_size; ++i) {
-            int created = 1;
-            dynablock_t *son = AddNewDynablock(block->parent, helper.sons_x86[i], &created);
-            if(created) {    // avoid breaking a working block!
-                son->block = helper.sons_arm[i];
-                son->x86_addr = (void*)helper.sons_x86[i];
-                son->x86_size = end-helper.sons_x86[i];
-                if(!son->x86_size) {printf_log(LOG_NONE, "Warning, son with null x86 size! (@%p / ARM=%p)", son->x86_addr, son->block);}
-                son->father = block;
-                son->done = 1;
-                sons[sons_size++] = son;
-                if(!son->parent)
-                    son->parent = block->parent;
-            }
-        }
-        if(sons_size) {
-            block->sons = sons;
-            block->sons_size = sons_size;
-        } else
-            box_free(sons);
-    }
-    box_free(helper.predecessor);
-    helper.predecessor = NULL;
    current_helper = NULL;
-    block->done = 1;
+    //block->done = 1;
    return (void*)block;
 }
@@ -1663,6 +1663,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                if((PK(0)==0) && (PK(1)==0) && (PK(2)==0) && (PK(3)==0))
                {
                    addr+=4;
+                    SKIPTEST(x14);
                    MESSAGE(LOG_DEBUG, "Exit x86 Emu\n");
                    MOV32(x14, ip+1+2);
                    STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
@@ -1672,6 +1673,7 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    *need_epilog = 1;
                } else {
                    MESSAGE(LOG_DUMP, "Native Call to %s\n", GetNativeName(GetNativeFnc(ip)));
+                    SKIPTEST(x14);
                    x87_forget(dyn, ninst, x3, x14, 0);
                    if((box86_log<2) && !cycle_log) {   // call the wrapper directly
                        uintptr_t ncall[2]; // to avoid BUSERROR!!!
@@ -1715,10 +1717,11 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xCD:
            SETFLAGS(X_ALL, SF_SET);    // Hack, set all flags (to an unknown state...)
+            SKIPTEST(x14);
            SMEND();
-            if(PK(0)==0x80) {
+            u8 = F8;
+            if(u8==0x80) {
                INST_NAME("Syscall");
-                u8 = F8;
                BARRIER(BARRIER_FLOAT);
                MOV32(xEIP, ip+2);
                STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP)|(1<<xEIP)|(1<<xFlags));
@@ -1734,7 +1737,11 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                jump_to_epilog(dyn, 0, xEIP, ninst);
            } else {
                INST_NAME("INT Ib");
-                DEFAULT;
+                if(box86_wine && u8==0x2D) {
+                    MESSAGE(LOG_DEBUG, "Hack for wine/int 2d\n");
+                } else {
+                    DEFAULT;
+                }
            }
            break;

@@ -2158,8 +2165,8 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                tmp = dyn->insts[ninst].pass2choice = 1;
            else if ((getProtection(u32)&PROT_READ) && (PKa(u32+0)==0x8B) && (((PKa(u32+1))&0xC7)==0x04) && (PKa(u32+2)==0x24) && (PKa(u32+3)==0xC3))
                tmp = dyn->insts[ninst].pass2choice = 2;
-            /*else if(isNativeCall(dyn, u32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
-                tmp = dyn->insts[ninst].pass2choice = 3;*/
+            else if(isNativeCall(dyn, u32, &dyn->insts[ninst].natcall, &dyn->insts[ninst].retn))
+                tmp = dyn->insts[ninst].pass2choice = 3;
            else 
                tmp = dyn->insts[ninst].pass2choice = 0;
            #else
@@ -2169,18 +2176,19 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                case 1:
                    //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    MESSAGE(LOG_DUMP, "Hack for Call 0\n");
+                    SKIPTEST(x14);
                    MOV32(xEIP, addr);
                    PUSH1(xEIP);
                    break;
                case 2:
                    //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    MESSAGE(LOG_DUMP, "Hack for Call x86.get_pc_thunk.reg\n");
+                    SKIPTEST(x14);
                    u8 = PK(i32+1);
                    gd = xEAX+((u8&0x38)>>3);
                    MOV32(gd, addr);
                    break;
-                // disabling this to avoid fetching data outside current block (in case this part changed, this block will not been marck as dirty)
-                /*case 3:
+                case 3:
                    SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    BARRIER(BARRIER_FULL);
                    BARRIER_NEXT(BARRIER_FULL);
@@ -2219,16 +2227,22 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    B_NEXT(cNE);    // not quitting, so lets continue
                    MARK;
                    jump_to_epilog(dyn, 0, xEIP, ninst);
-                    break;*/
+                    break;
                default:
                    if((box86_dynarec_safeflags>1) || (ninst && dyn->insts[ninst-1].x86.set_flags)) {
                        READFLAGS(X_PEND);  // that's suspicious
                    } else {
-                        //SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
+                        SETFLAGS(X_ALL, SF_SET);    // Hack to set flags to "dont'care" state
                    }
                    // regular call
-                    BARRIER(BARRIER_FLOAT);
                    //BARRIER_NEXT(BARRIER_FULL);
+                    if(/*box86_dynarec_callret &&*/ box86_dynarec_bigblock>1) {
+                        BARRIER(BARRIER_FULL);
+                    } else {
+                        BARRIER(BARRIER_FLOAT);
+                        *need_epilog = 0;
+                        *ok = 0;
+                    }
                    *need_epilog = 0;
                    *ok = 0;
                    MOV32(x2, addr);
@@ -2823,10 +2837,13 @@ uintptr_t dynarec00(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2: // CALL Ed
                    INST_NAME("CALL Ed");
-                    PASS2IF((box86_dynarec_safeflags>1) || (ninst && dyn->insts[ninst-1].x86.set_flags), 1) {
+                    PASS2IF((box86_dynarec_safeflags>1) || 
+                        ((ninst && dyn->insts[ninst-1].x86.set_flags)
+                        || ((ninst>1) && dyn->insts[ninst-2].x86.set_flags)), 1)
+                    {
                        READFLAGS(X_PEND);          // that's suspicious
                    } else {
-                        //SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
+                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
                    }
                    GETEDH(xEIP);
                    BARRIER(BARRIER_FLOAT);
@@ -90,6 +90,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            INST_NAME("UD2");
            SETFLAGS(X_ALL, SF_SET);    // Hack to set flags in "don't care" state
            //CALL(arm_ud, -1, 0);
+            SKIPTEST(x14);
            UDF(0);
            break;

@@ -421,6 +422,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
        case 0x31:
            INST_NAME("RDTSC");
            MESSAGE(LOG_DUMP, "Need Optimization\n");
+            SKIPTEST(x14);
            CALL(ReadTSC, xEAX, 0);   // will return the u64 in x1:xEAX
            MOV_REG(xEDX, x1);
            break;
@@ -656,6 +658,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x51:
            INST_NAME("SQRTPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
@@ -695,6 +698,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x52:
            INST_NAME("RSQRTPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
@@ -703,6 +707,7 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0x53:
            INST_NAME("RCPPS Gx, Ex");
+            SKIPTEST(x1);
            nextop = F8;
            GETEX(q0, 0);
            gd = (nextop&0x38)>>3;
@@ -791,7 +796,23 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGX(v0, 1);
            GETEX(v1, 0);
+            // VMIN/VMAX wll put default NaN if any input is NaN
+            // but x86 will copy if either v0[x] or v1[x] is NaN, so lets force a copy if source is NaN
+            if(!box86_dynarec_fastnan) {
+                if(v0==v1) {
+                    q1 = fpu_get_scratch_quad(dyn);
+                    VMOVQ(q1, v1);
+                } else
+                    q1 = v1;
+            }
            VMINQ_F32(v0, v0, v1);
+            if(!box86_dynarec_fastnan) {
+                q0 = fpu_get_scratch_quad(dyn);
+                VCEQQ_F32(q0, v0, v0);   // 0 is NaN, -1 is not NaN
+                VANDQ(v0, v0, q0);
+                VBICQ(q0, q1, q0);
+                VORRQ(v0, v0, q0);
+            }
            break;
        case 0x5E:
            INST_NAME("DIVPS Gx, Ex");
@@ -815,7 +836,23 @@ uintptr_t dynarec0F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            nextop = F8;
            GETGX(v0, 1);
            GETEX(v1, 0);
+            // VMIN/VMAX wll put default NaN if any input is NaN
+            // but x86 will copy if either v0[x] or v1[x] is NaN, so lets force a copy if source is NaN
+            if(!box86_dynarec_fastnan) {
+                if(v0==v1) {
+                    q1 = fpu_get_scratch_quad(dyn);
+                    VMOVQ(q1, v1);
+                } else
+                    q1 = v1;
+            }
            VMAXQ_F32(v0, v0, v1);
+            if(!box86_dynarec_fastnan) {
+                q0 = fpu_get_scratch_quad(dyn);
+                VCEQQ_F32(q0, v0, v0);   // 0 is NaN, -1 is not NaN
+                VANDQ(v0, v0, q0);
+                VBICQ(q0, q1, q0);
+                VORRQ(v0, v0, q0);
+            }
            break;
        case 0x60:
            INST_NAME("PUNPCKLBW Gm,Em");
@@ -313,7 +313,10 @@ uintptr_t dynarecFS(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 2: // CALL Ed
                    INST_NAME("CALL FS:Ed");
-                    PASS2IF(ninst && dyn->insts[ninst-1].x86.set_flags, 1) {
+                    PASS2IF((box86_dynarec_safeflags>1) ||
+                        ((ninst && dyn->insts[ninst-1].x86.set_flags)
+                        || ((ninst>1) && dyn->insts[ninst-2].x86.set_flags)), 1)
+                    {
                        READFLAGS(X_PEND);          // that's suspicious
                    } else {
                        SETFLAGS(X_ALL, SF_SET);    //Hack to put flag in "don't care" state
@@ -230,7 +230,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            } else {
                v0 = sse_get_reg_empty(dyn, ninst, x1, gd);
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
-                VLD1Q_32(v0, ed);
+                VLD1Q_64(v0, ed);
                SMWRITE2();
            }
            break;
@@ -244,7 +244,23 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
                VMOVQ(v1, v0);
            } else {
                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
-                VST1Q_32(v0, ed);
+                VST1Q_64(v0, ed);
+                SMWRITE2();
+            }
+            break;
+
+
+        case 0x2B:
+            INST_NAME("MOVNTPD Ex, Gx");
+            nextop = F8;
+            gd = (nextop&0x38)>>3;
+            v0 = sse_get_reg(dyn, ninst, x1, gd, 0);
+            if((nextop&0xC0)==0xC0) {
+                v1 = sse_get_reg_empty(dyn, ninst, x1, nextop&7);
+                VMOVQ(v1, v0);
+            } else {
+                addr = geted(dyn, addr, ninst, nextop, &ed, x1, &fixedaddress, 0, 0, 0, NULL);
+                VST1Q_64(v0, ed);
                SMWRITE2();
            }
            break;
@@ -2295,7 +2311,7 @@ uintptr_t dynarec660F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
            VPADDLQ_U32(q0, q0);
            break;
        case 0xF7:
-            INST_NAME("MASKMOVDQU Gx, Ex")
+            INST_NAME("MASKMOVDQU Gx, Ex");
            nextop = F8;
            GETGX(q0, 1);
            GETEX(q1, 0);
@@ -194,7 +194,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,

        case 0xE8:
            INST_NAME("FLD1");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                VMOV_i_32(v1, 0b01110000);
            } else {
@@ -204,7 +204,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xE9:
            INST_NAME("FLDL2T");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_l2t));
                VLDR_32(v1, x2, 0);
@@ -216,7 +216,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEA:     
            INST_NAME("FLDL2E");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_l2e));
                VLDR_32(v1, x2, 0);
@@ -228,7 +228,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEB:
            INST_NAME("FLDPI");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_pi));
                VLDR_32(v1, x2, 0);
@@ -240,7 +240,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEC:
            INST_NAME("FLDLG2");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_lg2));
                VLDR_32(v1, x2, 0);
@@ -252,7 +252,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xED:
            INST_NAME("FLDLN2");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                MOV32(x2, (&f_ln2));
                VLDR_32(v1, x2, 0);
@@ -264,7 +264,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xEE:
            INST_NAME("FLDZ");
-            v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            if(ST_IS_F(0)) {
                VMOV_8(v1/2, 0);  // float is *2...
            } else {
@@ -309,7 +309,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            break;
        case 0xF2:
            INST_NAME("FPTAN");
-            v2 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+            v2 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
            v1 = x87_get_st(dyn, ninst, x1, x2, 1, NEON_CACHE_ST_D);
            VMOV_64(0, v1);    // prepare call to tan
            CALL_1D(tan, 0);
@@ -589,7 +589,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2:
                    INST_NAME("FST float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    if(ST_IS_F(0))
                        s0 = v1;
                    else {
@@ -608,7 +608,7 @@ uintptr_t dynarecD9(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 3:
                    INST_NAME("FSTP float[ED], ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    if(ST_IS_F(0))
                        s0 = v1;
                    else {
@@ -161,7 +161,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
            switch((nextop>>3)&7) {
                case 0:
                    INST_NAME("FILD ST0, Ew");
-                    v1 = x87_do_push(dyn, ninst, x1, NEON_CACHE_ST_F);
+                    v1 = x87_do_push(dyn, ninst, x1, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x3, &fixedaddress, 255, 0, 0, NULL);
                    LDRSH_IMM8(x1, wback, fixedaddress);
                    if(ST_IS_F(0)) {
@@ -175,7 +175,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 1:
                    INST_NAME("FISTTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
                    s0 = fpu_get_scratch_single(dyn);
@@ -202,7 +202,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 2:
                    INST_NAME("FIST Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    u8 = x87_setround_reset(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
@@ -224,7 +224,7 @@ uintptr_t dynarecDF(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int ninst,
                    break;
                case 3:
                    INST_NAME("FISTP Ew, ST0");
-                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, NEON_CACHE_ST_F);
+                    v1 = x87_get_st(dyn, ninst, x1, x2, 0, box86_dynarec_x87double?NEON_CACHE_ST_D:NEON_CACHE_ST_F);
                    u8 = x87_setround_reset(dyn, ninst, x1, x2, x14);
                    addr = geted(dyn, addr, ninst, nextop, &wback, x2, &fixedaddress, 255, 0, 0, NULL);
                    ed = x1;
@@ -27,6 +27,8 @@
 #include "dynarec_arm_private.h"
 #include "dynarec_arm_functions.h"
 #include "bridge.h"
+#include "arm_printer.h"
+#include "custommem.h"

 void arm_fstp(x86emu_t* emu, void* p)
 {
@@ -109,7 +111,7 @@ void arm_fistp64(x86emu_t* emu, int64_t* ed)
        memcpy(ed, &STll(0).sq, sizeof(int64_t));
    } else {
        int64_t tmp;
-        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, (double)(int64_t)0x8000000000000000LL) || !isfinite(ST0.d))
+        if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(ST0.d))
            tmp = 0x8000000000000000LL;
        else
            tmp = fpu_round(emu, ST0.d);
@@ -120,7 +122,7 @@ void arm_fistp64(x86emu_t* emu, int64_t* ed)
 int64_t arm_fist64_0(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = nearbyint(d);
@@ -129,7 +131,7 @@ int64_t arm_fist64_0(double d)
 int64_t arm_fist64_1(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = floor(d);
@@ -138,7 +140,7 @@ int64_t arm_fist64_1(double d)
 int64_t arm_fist64_2(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = ceil(d);
@@ -147,7 +149,7 @@ int64_t arm_fist64_2(double d)
 int64_t arm_fist64_3(double d)
 {
    int64_t tmp;
-    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, (double)(int64_t)0x8000000000000000LL) || !isfinite(d))
+    if(isgreater(d, (double)(int64_t)0x7fffffffffffffffLL) || isless(d, -(double)(uint64_t)0x8000000000000000LL) || !isfinite(d))
        tmp = 0x8000000000000000LL;
    else
        tmp = trunc(d);
@@ -170,6 +172,7 @@ void arm_fld(x86emu_t* emu, uint8_t* ed)

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

@@ -608,6 +611,7 @@ void neoncacheUnwind(neoncache_t* cache)
                case NEON_CACHE_XMMW:
                    cache->ssecache[cache->neoncache[i].n].reg = i+FPUFIRST;
                    cache->ssecache[cache->neoncache[i].n].write = (cache->neoncache[i].t==NEON_CACHE_XMMW)?1:0;
+                    cache->neoncache[i] = cache->neoncache[i+1];
                    ++cache->fpu_reg;
                    ++i;    // next next
                    break;
@@ -702,13 +706,13 @@ int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int
 #define PK(a)       *(uint8_t*)(addr+a)
 #define PK32(a)     *(uint32_t*)(addr+a)

-    if(!addr)
+    if(!addr || !getProtection(addr))
        return 0;
    if(PK(0)==0xff && PK(1)==0x25) {  // absolute jump, maybe the GOT
        uintptr_t a1 = (PK32(2));   // need to add a check to see if the address is from the GOT !
        addr = (uintptr_t)getAlternate(*(void**)a1); 
    }
-    if(addr<0x10000)    // too low, that is suspicious
+    if(addr<0x10000 || !getProtection(addr))    // too low, that is suspicious
        return 0;
    onebridge_t *b = (onebridge_t*)(addr);
    if(b->CC==0xCC && b->S=='S' && b->C=='C' && b->w!=(wrapper_t)0 && b->f!=(uintptr_t)PltResolver) {
@@ -737,24 +741,6 @@ const char* getCacheName(int t, int n)
    return buff;
 }

-// is inst clean for a son branch?
-int isInstClean(dynarec_arm_t* dyn, int ninst)
-{
-    // check flags cache
-    if(dyn->insts[ninst].f_entry.dfnone || dyn->insts[ninst].f_entry.pending)
-        return 0;
-    if(dyn->insts[ninst].x86.state_flags)
-        return 0;
-    // check neoncache
-    neoncache_t* n = &dyn->insts[ninst].n;
-    if(n->news || n->stack || n->stack_next)
-        return 0;
-    for(int i=0; i<24; ++i)
-        if(n->neoncache[i].v)
-            return 0;
-    return 1;
-}
-
 static int flagsCacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
    int jmp = dyn->insts[ninst].x86.jmp_insts;
    if(jmp<0)
@@ -795,4 +781,58 @@ int CacheNeedsTransform(dynarec_arm_t* dyn, int ninst) {
    if (fpuCacheNeedsTransform(dyn, ninst)) ret|=1;
    if (flagsCacheNeedsTransform(dyn, ninst)) ret|=2;
    return ret;
-}
+}
+
+void inst_name_pass3(dynarec_arm_t* dyn, int ninst, const char* name)
+{
+    if(box86_dynarec_dump) {
+        printf_x86_instruction(my_context->dec, &dyn->insts[ninst].x86, name);
+        dynarec_log(LOG_NONE, "%s%p: %d emited opcodes, inst=%d, barrier=%d state=%d/%d(%d), %s=%X/%X, use=%X, need=%X/%X sm=%d/%d",
+            (box86_dynarec_dump>1)?"\e[32m":"",
+            (void*)(dyn->arm_start+dyn->insts[ninst].address),
+            dyn->insts[ninst].size/4,
+            ninst,
+            dyn->insts[ninst].x86.barrier,
+            dyn->insts[ninst].x86.state_flags,
+            dyn->f.pending,
+            dyn->f.dfnone,
+            dyn->insts[ninst].x86.may_set?"may":"set",
+            dyn->insts[ninst].x86.set_flags,
+            dyn->insts[ninst].x86.gen_flags,
+            dyn->insts[ninst].x86.use_flags,
+            dyn->insts[ninst].x86.need_before,
+            dyn->insts[ninst].x86.need_after,
+            dyn->smread, dyn->smwrite);
+        if(dyn->insts[ninst].pred_sz) {
+            dynarec_log(LOG_NONE, ", pred=");
+            for(int ii=0; ii<dyn->insts[ninst].pred_sz; ++ii)
+                dynarec_log(LOG_NONE, "%s%d", ii?"/":"", dyn->insts[ninst].pred[ii]);
+        }
+        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts>=0)
+            dynarec_log(LOG_NONE, ", jmp=%d", dyn->insts[ninst].x86.jmp_insts);
+        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts==-1)
+            dynarec_log(LOG_NONE, ", jmp=out");
+        for(int ii=0; ii<24; ++ii) {
+            switch(dyn->insts[ninst].n.neoncache[ii].t) {
+                case NEON_CACHE_ST_D: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_ST_F: dynarec_log(LOG_NONE, " S%d:%s", (ii+8)*2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_MM: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_XMMW: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;
+                case NEON_CACHE_XMMR: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;
+                case NEON_CACHE_SCR: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;
+                case NEON_CACHE_NONE:
+                default:    break;
+            }
+        }
+        if(dyn->n.stack || dyn->insts[ninst].n.stack_next || dyn->insts[ninst].n.x87stack)
+            dynarec_log(LOG_NONE, " X87:%d/%d(+%d/-%d)%d", dyn->n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, dyn->insts[ninst].n.stack_pop, dyn->insts[ninst].n.x87stack);
+        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
+            dynarec_log(LOG_NONE, " %s:%d/%d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
+        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");
+    }
+}
+
+void print_opcode(dynarec_arm_t* dyn, int ninst, uint32_t opcode)
+{
+    dynarec_log(LOG_NONE, "\t%08x\t%s\n", opcode, arm_print(opcode));
+}
@@ -71,9 +71,6 @@ int CacheNeedsTransform(dynarec_arm_t* dyn, int i1);
 // Undo the changes of a neoncache to get the status before the instruction
 void neoncacheUnwind(neoncache_t* cache);

-// is inst clean for a son branch?
-int isInstClean(dynarec_arm_t* dyn, int ninst);
-
 // predecessor access
 int isPred(dynarec_arm_t* dyn, int ninst, int pred);
 int getNominalPred(dynarec_arm_t* dyn, int ninst);
@@ -88,4 +85,7 @@ int isNativeCall(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t* calladdress, int

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

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

 #include "debug.h"
 #include "box86context.h"
@@ -185,7 +186,7 @@ uintptr_t geted16(dynarec_arm_t* dyn, uintptr_t addr, int ninst, uint8_t nextop,
            case 1: offset = F8S; break;
            case 2: offset = F16S; break;
        }
-        if(offset && (abs(offset)>absmax || (offset&mask))) {
+        if(offset && (abs(offset)<=absmax && !(offset&mask))) {
            *fixaddress = offset;
            offset = 0;
        }
@@ -861,10 +862,29 @@ void x87_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int
    MESSAGE(LOG_DUMP, "\t---Purge x87 Cache and Synch Stackcount\n");
 }

-#ifdef HAVE_TRACE
 static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
 {
-    x87_stackcount(dyn, ninst, s1);
+    // Synch top & stack counter
+    int a = dyn->n.x87stack;
+    if(a) {
+    // Add x87stack to emu fpu_stack
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, fpu_stack));
+        if(a>0) {
+            ADD_IMM8(s2, s2, a);
+        } else {
+            SUB_IMM8(s2, s2, -a);
+        }
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, fpu_stack));
+        // Sub x87stack to top, with and 7
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+        if(a>0) {
+            SUB_IMM8(s2, s2, a);
+        } else {
+            ADD_IMM8(s2, s2, -a);
+        }
+        AND_IMM8(s2, s2, 7);
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    }
    int ret = 0;
    for (int i=0; (i<8) && (!ret); ++i)
        if(dyn->n.x87cache[i] != -1)
@@ -872,13 +892,20 @@ static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
    if(!ret)    // nothing to do
        return;
    // Get top
-    LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    if(!a) {
+        // already there
+        LDR_IMM9(s2, xEmu, offsetof(x86emu_t, top));
+    }
    // loop all cache entries
    for (int i=0; i<8; ++i)
        if(dyn->n.x87cache[i]!=-1) {
-            ADD_IMM8(s3, s2, dyn->n.x87cache[i]);
-            AND_IMM8(s3, s3, 7);    // (emu->top + i)&7
-            ADD_REG_LSL_IMM5(s3, xEmu, s3, 3);    // fpu[(emu->top+i)&7] lsl 3 because fpu are double, so 8 bytes
+            if(dyn->n.x87cache[i]) {
+                ADD_IMM8(s3, s2, dyn->n.x87cache[i]);
+                AND_IMM8(s3, s3, 7);    // (emu->top + i)&7
+                ADD_REG_LSL_IMM5(s3, xEmu, s3, 3);    // fpu[(emu->top+i)&7] lsl 3 because fpu are double, so 8 bytes
+            } else {
+                ADD_REG_LSL_IMM5(s3, xEmu, s2, 3);
+            }
            if(ST_IS_F(dyn->n.x87cache[i])) {
                VCVT_F64_F32(0, dyn->n.x87reg[i]*2);    // use D0 as scratch...
                VSTR_64(0, s3, offsetof(x86emu_t, x87));
@@ -887,7 +914,31 @@ static void x87_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
            }
        }
 }
-#endif
+
+static void x87_unreflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3)
+{
+    // go back with the top & stack counter
+    int a = dyn->n.x87stack;
+    if(a) {
+    // Add x87stack to emu fpu_stack
+        LDR_IMM9(s1, xEmu, offsetof(x86emu_t, fpu_stack));
+        if(a>0) {
+            SUB_IMM8(s1, s1, a);
+        } else {
+            ADD_IMM8(s1, s1, -a);
+        }
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, fpu_stack));
+        // Sub x87stack to top, with and 7
+        LDR_IMM9(s1, xEmu, offsetof(x86emu_t, top));
+        if(a>0) {
+            ADD_IMM8(s1, s1, a);
+        } else {
+            SUB_IMM8(s1, s1, -a);
+        }
+        AND_IMM8(s1, s1, 7);
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, top));
+    }
+}

 int x87_get_current_cache(dynarec_arm_t* dyn, int ninst, int st, int t)
 {
@@ -1223,7 +1274,6 @@ void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1)
        MESSAGE(LOG_DUMP, "\t------ Purge MMX Cache\n");
    }
 }
-#ifdef HAVE_TRACE
 static void mmx_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
    int old = -1;
@@ -1241,7 +1291,6 @@ static void mmx_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
            VST1_32_W(dyn->n.mmxcache[i], s1);
        }
 }
-#endif


 // SSE / SSE2 helpers
@@ -1321,7 +1370,6 @@ static void sse_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1)
        MESSAGE(LOG_DUMP, "\t------ Purge SSE Cache\n");
    }
 }
-#ifdef HAVE_TRACE
 static void sse_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
    int old = -1;
@@ -1345,7 +1393,6 @@ static void sse_reflectcache(dynarec_arm_t* dyn, int ninst, int s1)
            VST1Q_32_W(dyn->n.ssecache[i].reg, s1);
        }
 }
-#endif

 void fpu_pushcache(dynarec_arm_t* dyn, int ninst, int s1)
 {
@@ -1446,8 +1493,10 @@ static void swapCache(dynarec_arm_t* dyn, int ninst, int i, int j, neoncache_t *
        quad =1;
    if(cache->neoncache[j].t==NEON_CACHE_XMMR || cache->neoncache[j].t==NEON_CACHE_XMMW)
        quad =1;
-    if(quad)    // if quad, we need to swap the whole quad!
+    if(quad) {    // if quad, we need to swap the whole quad!
        i&=~1;
+        j&=~1;
+    }
    if(!cache->neoncache[i].v && ((quad && !cache->neoncache[i+1].v) || !quad)) {
        // a mov is enough, no need to swap
        MESSAGE(LOG_DUMP, "\t  - Moving %d <- %d\n", i, j);
@@ -1717,7 +1766,7 @@ static void fpuCacheTransform(dynarec_arm_t* dyn, int ninst, int s1, int s2, int
        stack_cnt = cache_i2.stack;
    }
    neoncache_t cache = dyn->n;
-    // sanitize XMM regs alocation
+    // sanitize XMM regs alocation, cache_i2 has been sanitized in neoncacheUnwind already
    for(int i=0; i<24; i+=2) {
        if(cache.neoncache[i].t == NEON_CACHE_XMMR || cache.neoncache[i].t == NEON_CACHE_XMMW)
            cache.neoncache[i+1] = cache.neoncache[i];
@@ -1825,11 +1874,11 @@ static void flagsCacheTransform(dynarec_arm_t* dyn, int ninst, int s1)
        if(dyn->f.pending!=SF_PENDING) {
            LDR_IMM9(s1, xEmu, offsetof(x86emu_t, df));
            TSTS_REG_LSL_IMM5(s1, s1, 0);
-            j32 = (GETMARK3)-(dyn->arm_size+8);
+            j32 = (GETMARKF2)-(dyn->arm_size+8);
            Bcond(cEQ, j32);
        }
        CALL_(UpdateFlags, -1, 0);
-        MARK3;
+        MARKF2;
    }
 #endif
 }
@@ -1841,16 +1890,17 @@ void CacheTransform(dynarec_arm_t* dyn, int ninst, int cacheupd, int s1, int s2,
        flagsCacheTransform(dyn, ninst, s1);
 }

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

 void fpu_reset(dynarec_arm_t* dyn)
 {
@@ -1887,3 +1937,64 @@ void emit_pf(dynarec_arm_t* dyn, int ninst, int s1, int s3, int s4)
    MVN_REG_LSR_REG(s4, s4, s3);
    BFI(xFlags, s4, F_PF, 1);
 }
+
+void fpu_reset_cache(dynarec_arm_t* dyn, int ninst, int reset_n)
+{
+    MESSAGE(LOG_DEBUG, "Reset Caches with %d\n",reset_n);
+    #if STEP > 1
+    // for STEP 2 & 3, just need to refrest with current, and undo the changes (push & swap)
+    dyn->n = dyn->insts[ninst].n;
+    neoncacheUnwind(&dyn->n);
+    #ifdef HAVE_TRACE
+    if(box86_dynarec_dump)
+        if(memcmp(&dyn->n, &dyn->insts[reset_n].n, sizeof(neon_cache_t))) {
+            MESSAGE(LOG_DEBUG, "Warning, difference in neoncache: reset=");
+            for(int i=0; i<24; ++i)
+                if(dyn->insts[reset_n].n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[reset_n].n.neoncache[i].t, dyn->insts[reset_n].n.neoncache[i].n));
+            if(dyn->insts[reset_n].n.combined1 || dyn->insts[reset_n].n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[reset_n].n.swapped?"SWP":"CMB", dyn->insts[reset_n].n.combined1, dyn->insts[reset_n].n.combined2);
+            if(dyn->insts[reset_n].n.stack_push || dyn->insts[reset_n].n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[reset_n].n.stack_push, -dyn->insts[reset_n].n.stack_pop);
+            MESSAGE(LOG_DEBUG, " ==> ");
+            for(int i=0; i<24; ++i)
+                if(dyn->insts[ninst].n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[ninst].n.neoncache[i].t, dyn->insts[ninst].n.neoncache[i].n));
+            if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
+            if(dyn->insts[ninst].n.stack_push || dyn->insts[ninst].n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
+            MESSAGE(LOG_DEBUG, " -> ");
+            for(int i=0; i<24; ++i)
+                if(dyn->n.neoncache[i].v)
+                    MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->n.neoncache[i].t, dyn->n.neoncache[i].n));
+            if(dyn->n.combined1 || dyn->n.combined2)
+                MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->n.swapped?"SWP":"CMB", dyn->n.combined1, dyn->n.combined2);
+            if(dyn->n.stack_push || dyn->n.stack_pop)
+                MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->n.stack_push, -dyn->n.stack_pop);
+            MESSAGE(LOG_DEBUG, "\n");
+        }
+    #endif //HAVE_TRACE
+    #else
+    dyn->n = dyn->insts[reset_n].n;
+    #endif
+}
+
+// propagate ST stack state, especial stack pop that are deferred
+void fpu_propagate_stack(dynarec_arm_t* dyn, int ninst)
+{
+    if(dyn->n.stack_pop) {
+        for(int j=0; j<24; ++j)
+            if((dyn->n.neoncache[j].t == NEON_CACHE_ST_D || dyn->n.neoncache[j].t == NEON_CACHE_ST_F)) {
+                if(dyn->n.neoncache[j].n<dyn->n.stack_pop)
+                    dyn->n.neoncache[j].v = 0;
+                else
+                    dyn->n.neoncache[j].n-=dyn->n.stack_pop;
+            }
+        dyn->n.stack_pop = 0;
+    }
+    dyn->n.stack = dyn->n.stack_next;
+    dyn->n.news = 0;
+    dyn->n.stack_push = 0;
+    dyn->n.swapped = 0;
+}
@@ -260,17 +260,19 @@
 // CALL_1RD will use S as scratch. Return value in D0, 1 ARG in R
 #define CALL_1RD(F, R, S, M) call_rd(dyn, ninst, F, R, S, M, 0);

-#define MARK    if(dyn->insts) {dyn->insts[ninst].mark = (uintptr_t)dyn->arm_size;}
-#define GETMARK ((dyn->insts)?dyn->insts[ninst].mark:(dyn->arm_size+4))
-#define MARK2   if(dyn->insts) {dyn->insts[ninst].mark2 = (uintptr_t)dyn->arm_size;}
-#define GETMARK2 ((dyn->insts)?dyn->insts[ninst].mark2:(dyn->arm_size+4))
-#define MARK3   if(dyn->insts) {dyn->insts[ninst].mark3 = (uintptr_t)dyn->arm_size;}
-#define GETMARK3 ((dyn->insts)?dyn->insts[ninst].mark3:(dyn->arm_size+4))
-#define MARKF   if(dyn->insts) {dyn->insts[ninst].markf = (uintptr_t)dyn->arm_size;}
-#define GETMARKF ((dyn->insts)?dyn->insts[ninst].markf:(dyn->arm_size+4))
-#define MARKSEG if(dyn->insts) {dyn->insts[ninst].markseg = (uintptr_t)dyn->arm_size;}
-#define GETMARKSEG ((dyn->insts)?dyn->insts[ninst].markseg:(dyn->arm_size+4))
-#define MARKLOCK if(dyn->insts) {dyn->insts[ninst].marklock = (uintptr_t)dyn->arm_size;}
+#define MARK        if(dyn->insts) {dyn->insts[ninst].mark = (uintptr_t)dyn->arm_size;}
+#define GETMARK     ((dyn->insts)?dyn->insts[ninst].mark:(dyn->arm_size+4))
+#define MARK2       if(dyn->insts) {dyn->insts[ninst].mark2 = (uintptr_t)dyn->arm_size;}
+#define GETMARK2    ((dyn->insts)?dyn->insts[ninst].mark2:(dyn->arm_size+4))
+#define MARK3       if(dyn->insts) {dyn->insts[ninst].mark3 = (uintptr_t)dyn->arm_size;}
+#define GETMARK3    ((dyn->insts)?dyn->insts[ninst].mark3:(dyn->arm_size+4))
+#define MARKF       if(dyn->insts) {dyn->insts[ninst].markf = (uintptr_t)dyn->arm_size;}
+#define GETMARKF    ((dyn->insts)?dyn->insts[ninst].markf:(dyn->arm_size+4))
+#define MARKF2      if(dyn->insts) {dyn->insts[ninst].markf2 = (uintptr_t)dyn->arm_size;}
+#define GETMARKF2   ((dyn->insts)?dyn->insts[ninst].markf2:(dyn->arm_size+4))
+#define MARKSEG     if(dyn->insts) {dyn->insts[ninst].markseg = (uintptr_t)dyn->arm_size;}
+#define GETMARKSEG  ((dyn->insts)?dyn->insts[ninst].markseg:(dyn->arm_size+4))
+#define MARKLOCK    if(dyn->insts) {dyn->insts[ninst].marklock = (uintptr_t)dyn->arm_size;}
 #define GETMARKLOCK ((dyn->insts)?dyn->insts[ninst].marklock:(dyn->arm_size+4))

 // Branch to MARK if cond (use j32)
@@ -547,12 +549,13 @@ void* arm_next(x86emu_t* emu, uintptr_t addr);
 #define fpu_pushcache   STEPNAME(fpu_pushcache)
 #define fpu_popcache    STEPNAME(fpu_popcache)
 #define fpu_reset       STEPNAME(fpu_reset)
+#define fpu_reset_cache STEPNAME(fpu_reset_cache)
+#define fpu_propagate_stack STEPNAME(fpu_propagate_stack)
 #define fpu_purgecache  STEPNAME(fpu_purgecache)
 #define x87_purgecache  STEPNAME(x87_purgecache)
 #define mmx_purgecache  STEPNAME(mmx_purgecache)
-#ifdef HAVE_TRACE
 #define fpu_reflectcache STEPNAME(fpu_reflectcache)
-#endif
+#define fpu_unreflectcache STEPNAME(fpu_unreflectcache)

 // get the single reg that from the double "reg" (so Dx[idx])
 #define fpu_get_single_reg      STEPNAME(fpu_get_single_reg)
@@ -741,13 +744,16 @@ 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);
+// reset the cache with n
+void fpu_reset_cache(dynarec_arm_t* dyn, int ninst, int reset_n);
+// propagate stack state
+void fpu_propagate_stack(dynarec_arm_t* dyn, int ninst);
 // purge the FPU cache (needs 3 scratch registers) next=1 if for a conditionnal branch jumping out of block (no tracking updated)
 void fpu_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 void x87_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1, int s2, int s3);
 void mmx_purgecache(dynarec_arm_t* dyn, int ninst, int next, int s1);
-#ifdef HAVE_TRACE
 void fpu_reflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
-#endif
+void fpu_unreflectcache(dynarec_arm_t* dyn, int ninst, int s1, int s2, int s3);
 void fpu_pushcache(dynarec_arm_t* dyn, int ninst, int s1);
 void fpu_popcache(dynarec_arm_t* dyn, int ninst, int s1);

@@ -794,4 +800,21 @@ uintptr_t dynarecF30F(dynarec_arm_t* dyn, uintptr_t addr, uintptr_t ip, int nins
 #define MAYUSE(A)   
 #endif

+#define NOTEST(s1)                                          \
+    if(box86_dynarec_test) {                                \
+        MOVW(s1, 0);                                        \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.test));  \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.clean)); \
+    }
+#define SKIPTEST(s1)                                        \
+    if(box86_dynarec_test) {                                \
+        MOVW(s1, 0);                                        \
+        STR_IMM9(s1, xEmu, offsetof(x86emu_t, test.clean)); \
+    }
+#define GOTEST(s1, s2)                                      \
+    if(box86_dynarec_test) {                                \
+        MOVW(s2, 1);                                        \
+        STR_IMM9(s2, xEmu, offsetof(x86emu_t, test.test));  \
+    }
+
 #endif //__DYNAREC_ARM_HELPER_H__
@@ -0,0 +1,12 @@
+#include <stdint.h>
+#include <stdlib.h>
+
+#include "arm_emitter.h"
+
+#define EMIT(A) *block = (A); ++block
+void CreateJmpNext(void* addr, void* next)
+{
+    uint32_t* block = (uint32_t*)addr;
+    LDR_literal(x2, ((intptr_t)next - (intptr_t)addr)-8);
+    BX(x2);
+}
@@ -36,7 +36,6 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
    uintptr_t init_addr = addr;
    MAYUSE(init_addr);
    int need_epilog = 1;
-    dyn->sons_size = 0;
    // Clean up (because there are multiple passes)
    dyn->f.pending = 0;
    dyn->f.dfnone = 0;
@@ -57,44 +56,7 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->f.pending = 0;
                fpu_reset(dyn);
            } else {
-                MESSAGE(LOG_DEBUG, "Reset Caches with %d\n",reset_n);
-                #if STEP > 1
-                // for STEP 2 & 3, just need to refrest with current, and undo the changes (push & swap)
-                dyn->n = dyn->insts[ninst].n;
-                neoncacheUnwind(&dyn->n);
-                #ifdef HAVE_TRACE
-                if(box86_dynarec_dump)
-                    if(memcmp(&dyn->n, &dyn->insts[reset_n].n, sizeof(neon_cache_t))) {
-                        MESSAGE(LOG_DEBUG, "Warning, difference in neoncache: reset=");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->insts[reset_n].n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[reset_n].n.neoncache[i].t, dyn->insts[reset_n].n.neoncache[i].n));
-                        if(dyn->insts[reset_n].n.combined1 || dyn->insts[reset_n].n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[reset_n].n.swapped?"SWP":"CMB", dyn->insts[reset_n].n.combined1, dyn->insts[reset_n].n.combined2);
-                        if(dyn->insts[reset_n].n.stack_push || dyn->insts[reset_n].n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[reset_n].n.stack_push, -dyn->insts[reset_n].n.stack_pop);
-                        MESSAGE(LOG_DEBUG, " ==> ");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->insts[ninst].n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->insts[ninst].n.neoncache[i].t, dyn->insts[ninst].n.neoncache[i].n));
-                        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);
-                        if(dyn->insts[ninst].n.stack_push || dyn->insts[ninst].n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->insts[ninst].n.stack_push, -dyn->insts[ninst].n.stack_pop);
-                        MESSAGE(LOG_DEBUG, " -> ");
-                        for(int i=0; i<24; ++i)
-                            if(dyn->n.neoncache[i].v)
-                                MESSAGE(LOG_DEBUG, " %02d:%s", i, getCacheName(dyn->n.neoncache[i].t, dyn->n.neoncache[i].n));
-                        if(dyn->n.combined1 || dyn->n.combined2)
-                            MESSAGE(LOG_DEBUG, " %s:%02d/%02d", dyn->n.swapped?"SWP":"CMB", dyn->n.combined1, dyn->n.combined2);
-                        if(dyn->n.stack_push || dyn->n.stack_pop)
-                            MESSAGE(LOG_DEBUG, " (%d:%d)", dyn->n.stack_push, -dyn->n.stack_pop);
-                        MESSAGE(LOG_DEBUG, "\n");
-                    }
-                #endif //HAVE_TRACE
-                #else
-                dyn->n = dyn->insts[reset_n].n;
-                #endif
+                fpu_reset_cache(dyn, ninst, reset_n);
                dyn->f = dyn->insts[reset_n].f_exit;
                if(dyn->insts[ninst].x86.barrier&BARRIER_FLOAT) {
                    MESSAGE(LOG_DEBUG, "Apply Barrier Float\n");
@@ -108,29 +70,29 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            }
            reset_n = -1;
        }
-        // propagate ST stack state, especial stack pop that are defered
-        if(dyn->n.stack_pop) {
-            for(int j=0; j<24; ++j)
-                if((dyn->n.neoncache[j].t == NEON_CACHE_ST_D || dyn->n.neoncache[j].t == NEON_CACHE_ST_F)) {
-                    if(dyn->n.neoncache[j].n<dyn->n.stack_pop)
-                        dyn->n.neoncache[j].v = 0;
-                    else
-                        dyn->n.neoncache[j].n-=dyn->n.stack_pop;
-                }
-            dyn->n.stack_pop = 0;
-        }
-        dyn->n.stack = dyn->n.stack_next;
-        dyn->n.news = 0;
-        dyn->n.stack_push = 0;
-        dyn->n.swapped = 0;
+        fpu_propagate_stack(dyn, ninst);
        NEW_INST;
+        if(!ninst) {
+            GOTEST(x1, x2);
+        }
        if(dyn->insts[ninst].pred_sz>1) {SMSTART();}
        fpu_reset_scratch(dyn);
        if((dyn->insts[ninst].x86.need_before&~X_PEND) && !dyn->insts[ninst].pred_sz) {
            READFLAGS(dyn->insts[ninst].x86.need_before&~X_PEND);
        }
+        if(box86_dynarec_test)  {
+            MESSAGE(LOG_DUMP, "TEST INIT  ----\n");
+            fpu_reflectcache(dyn, ninst, x1, x2, x3);
+            MOV32(x1, ip);
+            STM(xEmu, (1<<xEAX)|(1<<xEBX)|(1<<xECX)|(1<<xEDX)|(1<<xESI)|(1<<xEDI)|(1<<xESP)|(1<<xEBP));
+            STR_IMM9(x1, xEmu, offsetof(x86emu_t, ip));
+            MOVW(x2, 1);
+            CALL(x86test_init, -1, 0);
+            fpu_unreflectcache(dyn, ninst, x1, x2, x3);
+            MESSAGE(LOG_DUMP, "----------\n");
+        }
 #ifdef HAVE_TRACE
-        if(my_context->dec && box86_dynarec_trace) {
+        else if(my_context->dec && box86_dynarec_trace) {
        if((trace_end == 0) 
            || ((ip >= trace_start) && (ip < trace_end)))  {
                MESSAGE(LOG_DUMP, "TRACE ----\n");
@@ -140,6 +102,7 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                STR_IMM9(x1, xEmu, offsetof(x86emu_t, ip));
                MOVW(x2, 1);
                CALL(PrintTrace, -1, 0);
+                fpu_unreflectcache(dyn, ninst, x1, x2, x3);
                MESSAGE(LOG_DUMP, "----------\n");
            }
        }
@@ -162,13 +125,26 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->f.dfnone = 0;
            }
        }
+        #ifndef PROT_READ
+        #define PROT_READ 1
+        #endif
        #if STEP != 0
        if(!ok && !need_epilog && (addr < (dyn->start+dyn->isize))) {
            ok = 1;
            // we use the 1st predecessor here
            int ii = ninst+1;
-            while(ii<dyn->size && !dyn->insts[ii].pred_sz) 
-                ++ii;
+            if(ii<dyn->size && !dyn->insts[ii].pred_sz) {
+                while(ii<dyn->size && (!dyn->insts[ii].pred_sz || (dyn->insts[ii].pred_sz==1 && dyn->insts[ii].pred[0]==ii-1))) {
+                    // may need to skip opcodes to advance
+                    ++ninst;
+                    NEW_INST;
+                    MESSAGE(LOG_DEBUG, "Skipping unused opcode\n");
+                    INST_NAME("Skipped opcode");
+                    INST_EPILOG;
+                    addr += dyn->insts[ii].x86.size;
+                    ++ii;
+                }
+            }
            if((dyn->insts[ii].x86.barrier&BARRIER_FULL)==BARRIER_FULL)
                reset_n = -2;    // hack to say Barrier!
            else {
@@ -180,20 +156,18 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            }
        }
        #else
-        #ifndef PROT_READ
-        #define PROT_READ 1
-        #endif
        if(dyn->forward) {
-            if(dyn->forward_to == addr) {
+            if(dyn->forward_to == addr && !need_epilog) {
                // we made it!
-                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Forward extend block %p -> %p\n", (void*)dyn->forward, (void*)dyn->forward_to);
+                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Forward extend block for %d bytes %p -> %p\n", dyn->forward_to-dyn->forward, (void*)dyn->forward, (void*)dyn->forward_to);
                dyn->forward = 0;
                dyn->forward_to = 0;
                dyn->forward_size = 0;
                dyn->forward_ninst = 0;
+                ok = 1; // in case it was 0
            } else if ((dyn->forward_to < addr) || !ok) {
                // something when wrong! rollback
-                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Could not forward extend block %p -> %p\n", (void*)dyn->forward, (void*)dyn->forward_to);
+                if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Could not forward extend block for %d bytes %p -> %p\n", dyn->forward_to-dyn->forward, (void*)dyn->forward, (void*)dyn->forward_to);
                ok = 0;
                dyn->size = dyn->forward_size;
                ninst = dyn->forward_ninst;
@@ -204,12 +178,12 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                dyn->forward_ninst = 0;
            }
            // else just continue
-        } else  if(!ok && !need_epilog && box86_dynarec_bigblock && (getProtection(addr+3)&PROT_READ))
+        } else if(!ok && !need_epilog && box86_dynarec_bigblock && (getProtection(addr+3)&~PROT_READ))
            if(*(uint32_t*)addr!=0) {   // check if need to continue (but is next 4 bytes are 0, stop)
                uintptr_t next = get_closest_next(dyn, addr);
                if(next && (
                    (((next-addr)<15) && is_nops(dyn, addr, next-addr)) 
-                    /*&& box86_dynarec_bigblock*/))
+                    /*||(((next-addr)<30) && is_instructions(dyn, addr, next-addr))*/ ))
                {
                    ok = 1;
                    // need to find back that instruction to copy the caches, as previous version cannot be used anymore
@@ -220,30 +194,56 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
                            ii=ninst;
                        }
                    if(box86_dynarec_dump) dynarec_log(LOG_NONE, "Extend block %p, %p -> %p (ninst=%d, jump from %d)\n", dyn, (void*)addr, (void*)next, ninst, reset_n);
-                } else if(next && (next-addr<box86_dynarec_forward) 
-                && (getProtection(next)&PROT_READ)/*box86_dynarec_bigblock>=forwardblock*/) {
-                    dyn->forward = addr;
-                    dyn->forward_to = next;
-                    dyn->forward_size = dyn->size;
-                    dyn->forward_ninst = ninst;
-                    reset_n = -2;
-                    ok = 1;
+                } else if(next && (next-addr)<box86_dynarec_forward && (getProtection(next)&PROT_READ)/*box86_dynarec_bigblock>=stopblock*/) {
+                    if(!((box86_dynarec_bigblock<stopblock) && !isJumpTableDefault((void*)next))) {
+                        dyn->forward = addr;
+                        dyn->forward_to = next;
+                        dyn->forward_size = dyn->size;
+                        dyn->forward_ninst = ninst;
+                        reset_n = -2;
+                        ok = 1;
+                    }
                }
            }
        #endif
-        if(ok<0)  {ok = 0; need_epilog=1;}
+        if(ok<0)  {
+            ok = 0; need_epilog=1; 
+            #if STEP == 0
+            if(ninst) {
+                --ninst;
+                if(!dyn->insts[ninst].x86.barrier) {
+                    BARRIER(BARRIER_FLOAT);
+                }
+                dyn->insts[ninst].x86.need_after |= X_PEND;
+                ++ninst;
+            }
+            #endif
+        }
        ++ninst;
        #if STEP == 0
        if(ok && (((box86_dynarec_bigblock<stopblock) && !isJumpTableDefault((void*)addr)) 
-            || (addr>=box86_nodynarec_start && addr<box86_nodynarec_end)
-            /*|| (dyn->size >= 2000)*/))    // is this still needed? maybe it should be a parameter?
+            || (addr>=box86_nodynarec_start && addr<box86_nodynarec_end)))
        #else
        if(ok && (ninst==dyn->size))
        #endif
        {
+            #if STEP == 0
+            if(dyn->forward) {
+                // stopping too soon
+                dyn->size = dyn->forward_size;
+                ninst = dyn->forward_ninst+1;
+                addr = dyn->forward;
+                dyn->forward = 0;
+                dyn->forward_to = 0;
+                dyn->forward_size = 0;
+                dyn->forward_ninst = 0;
+            }
+            #endif
            int j32;
            MAYUSE(j32);
-            MESSAGE(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size); 
+            MESSAGE(LOG_DEBUG, "Stopping block %p (%d / %d)\n",(void*)init_addr, ninst, dyn->size);
+            if(!box86_dynarec_dump && addr>=box86_nodynarec_start && addr<box86_nodynarec_end)
+                dynarec_log(LOG_INFO, "Stopping block in no-dynarec zone\n");
            --ninst;
            if(!dyn->insts[ninst].x86.barrier) {
                BARRIER(BARRIER_FLOAT);
@@ -252,16 +252,18 @@ uintptr_t arm_pass(dynarec_arm_t* dyn, uintptr_t addr)
            dyn->insts[ninst].x86.need_after |= X_PEND;
            #endif
            ++ninst;
+            NOTEST(x3);
            fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
            jump_to_next(dyn, addr, 0, ninst);
            ok=0; need_epilog=0;
        }
    }
    if(need_epilog) {
+        NOTEST(x3);
        fpu_purgecache(dyn, ninst, 0, x1, x2, x3);
        jump_to_epilog(dyn, ip, 0, ninst);  // no linker here, it's an unknow instruction
    }
    FINI;
-    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d, %d sons)\n", dyn->size, dyn->sons_size);
+    MESSAGE(LOG_DUMP, "---- END OF BLOCK ---- (%d)\n", dyn->size);
    return addr;
 }
@@ -4,6 +4,7 @@
    dyn->isize = addr-sav_addr;\
    dyn->insts[ninst].x86.addr = addr;\
    if(ninst) dyn->insts[ninst-1].x86.size = dyn->insts[ninst].x86.addr - dyn->insts[ninst-1].x86.addr
+
 #define MESSAGE(A, ...)  
 #define MAYSETFLAGS()   dyn->insts[ninst].x86.may_set = 1
 #define READFLAGS(A)    \
@@ -15,13 +16,13 @@
        dyn->f.pending=(B)&SF_SET_PENDING;      \
        dyn->f.dfnone=((B)&SF_SET)?1:0;
 #define EMIT(A)     
-#define JUMP(A, C)      if((A)>addr) add_next(dyn, (uintptr_t)(A)); dyn->insts[ninst].x86.jmp = A; dyn->insts[ninst].x86.jmp_cond = C
+#define JUMP(A, C)      add_next(dyn, (uintptr_t)A); dyn->insts[ninst].x86.jmp = A; dyn->insts[ninst].x86.jmp_cond = C
 #define BARRIER(A)      if(A!=BARRIER_MAYBE) {fpu_purgecache(dyn, ninst, 0, x1, x2, x3); dyn->insts[ninst].x86.barrier = A;} else dyn->insts[ninst].barrier_maybe = 1
 #define BARRIER_NEXT(A) dyn->insts[ninst+1].x86.barrier = A
 #define NEW_INST \
        ++dyn->size;                            \
        if(dyn->size+3>=dyn->cap) {             \
-                dyn->insts = (instruction_arm_t*)box_realloc(dyn->insts, sizeof(instruction_arm_t)*dyn->cap*2); \
+                dyn->insts = (instruction_arm_t*)customRealloc(dyn->insts, sizeof(instruction_arm_t)*dyn->cap*2);\
                memset(&dyn->insts[dyn->cap], 0, sizeof(instruction_arm_t)*dyn->cap);   \
                dyn->cap *= 2;                  \
        }                                       \
@@ -2,7 +2,6 @@
 #define FINI
 #define MESSAGE(A, ...)  
 #define EMIT(A)     
-
 #define NEW_INST                                \
        dyn->insts[ninst].f_entry = dyn->f;     \
        dyn->n.combined1 = dyn->n.combined2 = 0;\
@@ -1,13 +1,16 @@
 #define INIT        dyn->arm_size = 0
-#define FINI        if(ninst) {dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);}
+#define FINI                                                                                            \
+        if(ninst) {                                                                                     \
+                dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);     \
+                dyn->insts_size += 1+((dyn->insts[ninst].x86.size>dyn->insts[ninst].size)?dyn->insts[ninst].x86.size:dyn->insts[ninst].size)/15; \
+        }

 #define MESSAGE(A, ...)  
-#define EMIT(A)     dyn->insts[ninst].size+=4; dyn->arm_size+=4
+#define EMIT(A)     do{dyn->insts[ninst].size+=4; dyn->arm_size+=4;}while(0)
 #define NEW_INST                                                                                        \
        if(ninst) {                                                                                     \
                dyn->insts[ninst].address = (dyn->insts[ninst-1].address+dyn->insts[ninst-1].size);     \
-                if(ninst && isInstClean(dyn, ninst))                                                    \
-                        ++dyn->sons_size;                                                               \
+                dyn->insts_size += 1+((dyn->insts[ninst-1].x86.size>dyn->insts[ninst-1].size)?dyn->insts[ninst-1].x86.size:dyn->insts[ninst-1].size)/15; \
        }
 #define INST_EPILOG dyn->insts[ninst].epilog = dyn->arm_size; 
 #define INST_NAME(name) 
@@ -1,65 +1,20 @@
 #define INIT    
-#define FINI
-#define EMIT(A)     \
-    if(box86_dynarec_dump) {dynarec_log(LOG_NONE, "\t%08x\t%s\n", (A), arm_print(A));} \
-    *(uint32_t*)(dyn->block) = A;       \
-    dyn->block += 4; dyn->arm_size += 4;\
-    dyn->insts[ninst].size2 += 4
+#define FINI        \
+    if(ninst)       \
+        addInst(dyn->instsize, &dyn->insts_size, dyn->insts[ninst].x86.size, dyn->insts[ninst].size/4); \
+    addInst(dyn->instsize, &dyn->insts_size, 0, 0);
+#define EMIT(A)                                         \
+    do{                                                 \
+        if(box86_dynarec_dump) print_opcode(dyn, ninst, (uint32_t)(A)); \
+        if((uintptr_t)dyn->block<(uintptr_t)dyn->next-sizeof(void*))\
+            *(uint32_t*)(dyn->block) = (uint32_t)(A);   \
+        dyn->block += 4; dyn->arm_size += 4;            \
+        dyn->insts[ninst].size2 += 4;                   \
+    }while(0)

 #define MESSAGE(A, ...)  if(box86_dynarec_dump) dynarec_log(LOG_NONE, __VA_ARGS__)
 #define NEW_INST        \
-    if(ninst && isInstClean(dyn, ninst)) {                      \
-        dyn->sons_x86[dyn->sons_size] = (uintptr_t)ip;          \
-        dyn->sons_arm[dyn->sons_size] = dyn->block;             \
-        MESSAGE(LOG_DUMP, "----> potential Son here %p/%p\n", (void*)ip, dyn->block);  \
-        ++dyn->sons_size;                                       \
-    }
-
+    if(ninst)                                                   \
+        addInst(dyn->instsize, &dyn->insts_size, dyn->insts[ninst-1].x86.size, dyn->insts[ninst-1].size/4);
 #define INST_EPILOG     
-#define INST_NAME(name) \
-    if(box86_dynarec_dump) {\
-        printf_x86_instruction(my_context->dec, &dyn->insts[ninst].x86, name); \
-        dynarec_log(LOG_NONE, "%s%p: %d emited opcodes, inst=%d, barrier=%d state=%d/%d(%d), %s=%X/%X, use=%X, need=%X/%X sm=%d/%d", \
-            (box86_dynarec_dump>1)?"\e[32m":"", \
-            (void*)(dyn->arm_start+dyn->insts[ninst].address),  \
-            dyn->insts[ninst].size/4,           \
-            ninst,                              \
-            dyn->insts[ninst].x86.barrier,      \
-            dyn->insts[ninst].x86.state_flags,  \
-            dyn->f.pending,                     \
-            dyn->f.dfnone,                      \
-            dyn->insts[ninst].x86.may_set?"may":"set",          \
-            dyn->insts[ninst].x86.set_flags,    \
-            dyn->insts[ninst].x86.gen_flags,    \
-            dyn->insts[ninst].x86.use_flags,    \
-            dyn->insts[ninst].x86.need_before,  \
-            dyn->insts[ninst].x86.need_after,   \
-            dyn->smread, dyn->smwrite);         \
-        if(dyn->insts[ninst].pred_sz) {         \
-            dynarec_log(LOG_NONE, ", pred=");     \
-            for(int ii=0; ii<dyn->insts[ninst].pred_sz; ++ii)\
-                dynarec_log(LOG_NONE, "%s%d", ii?"/":"", dyn->insts[ninst].pred[ii]);\
-        }                                       \
-        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts>=0)\
-            dynarec_log(LOG_NONE, ", jmp=%d", dyn->insts[ninst].x86.jmp_insts);\
-        if(dyn->insts[ninst].x86.jmp && dyn->insts[ninst].x86.jmp_insts==-1)\
-            dynarec_log(LOG_NONE, ", jmp=out");\
-        for(int ii=0; ii<24; ++ii) {            \
-            switch(dyn->insts[ninst].n.neoncache[ii].t) {    \
-                case NEON_CACHE_ST_D: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;            \
-                case NEON_CACHE_ST_F: dynarec_log(LOG_NONE, " S%d:%s", (ii+8)*2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;        \
-                case NEON_CACHE_MM: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;              \
-                case NEON_CACHE_XMMW: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;  \
-                case NEON_CACHE_XMMR: dynarec_log(LOG_NONE, " Q%d:%s", (ii+8)/2, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); ++ii; break;  \
-                case NEON_CACHE_SCR: dynarec_log(LOG_NONE, " D%d:%s", ii+8, getCacheName(dyn->insts[ninst].n.neoncache[ii].t, dyn->insts[ninst].n.neoncache[ii].n)); break;             \
-                case NEON_CACHE_NONE:           \
-                default:    break;              \
-            }                                   \
-        }                                       \
-        if(dyn->n.stack || dyn->insts[ninst].n.stack_next || dyn->insts[ninst].n.x87stack)     \
-            dynarec_log(LOG_NONE, " X87:%d/%d(+%d/-%d)%d", dyn->n.stack, dyn->insts[ninst].n.stack_next, dyn->insts[ninst].n.stack_push, dyn->insts[ninst].n.stack_pop, dyn->insts[ninst].n.x87stack); \
-        if(dyn->insts[ninst].n.combined1 || dyn->insts[ninst].n.combined2)                     \
-            dynarec_log(LOG_NONE, " %s:%d/%d", dyn->insts[ninst].n.swapped?"SWP":"CMB", dyn->insts[ninst].n.combined1, dyn->insts[ninst].n.combined2);   \
-        dynarec_log(LOG_NONE, "%s\n", (box86_dynarec_dump>1)?"\e[m":"");                       \
-    }
-
+#define INST_NAME(name) inst_name_pass3(dyn, ninst, name)
@@ -5,6 +5,7 @@

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

 #define BARRIER_MAYBE   8

@@ -68,7 +69,7 @@ typedef struct instruction_arm_s {
    int                 pred_sz;    // size of predecessor list
    int                 *pred;      // predecessor array
    uintptr_t           mark, mark2, mark3;
-    uintptr_t           markf;
+    uintptr_t           markf, markf2;
    uintptr_t           markseg;
    uintptr_t           marklock;
    int                 pass2choice;// value for choices that are fixed on pass2 for pass3
@@ -90,22 +91,23 @@ typedef struct dynarec_arm_s {
    void*               block;      // memory pointer where next instruction is emited
    uintptr_t           arm_start;  // start of the arm code
    int                 arm_size;   // size of emitted arm code
+    uintptr_t           jmp_next;   // address of the jump_next address
    flagcache_t         f;
    neoncache_t         n;          // cache for the 8..31 double reg from fpu, plus x87 stack delta
    uintptr_t*          next;       // variable array of "next" jump address
    int                 next_sz;
    int                 next_cap;
-    uintptr_t*          sons_x86;   // the x86 address of potential dynablock sons
-    void**              sons_arm;   // the arm address of potential dynablock sons
-    int                 sons_size;  // number of potential dynablock sons
    int*                predecessor;// single array of all predecessor
    dynablock_t*        dynablock;
+    instsize_t*         instsize;
+    size_t              insts_size; // size of the instruction size array (calculated)
    uint8_t             smread;    // for strongmem model emulation
    uint8_t             smwrite;    // for strongmem model emulation
    uintptr_t           forward;    // address of the last end of code while testing forward
    uintptr_t           forward_to; // address of the next jump to (to check if everything is ok)
    int32_t             forward_size;   // size at the forward point
    int                 forward_ninst;  // ninst at the forward point
+    uint8_t             always_test;
 } dynarec_arm_t;

 void add_next(dynarec_arm_t *dyn, uintptr_t addr);
@@ -113,4 +115,6 @@ uintptr_t get_closest_next(dynarec_arm_t *dyn, uintptr_t addr);
 int is_nops(dynarec_arm_t *dyn, uintptr_t addr, int n);
 int is_instructions(dynarec_arm_t *dyn, uintptr_t addr, int n);

+void CreateJmpNext(void* addr, void* next);
+
 #endif //__DYNAREC_ARM_PRIVATE_H_
@@ -1607,16 +1607,12 @@ void* ElfSetBrk(void* newbrk)
 }

 #ifdef DYNAREC
-dynablocklist_t* GetDynablocksFromAddress(box86context_t *context, uintptr_t addr)
+dynablock_t* GetDynablocksFromAddress(box86context_t *context, uintptr_t addr)
 {
-    // if we are here, the there is not block in standard "space"
-    /*dynablocklist_t* ret = getDBFromAddress(addr);
-    if(ret) {
-        return ret;
-    }*/
+    (void)context;
    if(box86_dynarec_forced) {
        addDBFromAddressRange(addr, 1);
-        return getDB(addr>>DYNAMAP_SHIFT);
+        return getDB(addr);
    }
    //check if address is in an elf... if yes, grant a block (should I warn)
    Dl_info info;
@@ -1,10 +1,6 @@
 #ifndef __ELFLOADER_PRIVATE_H_
 #define __ELFLOADER_PRIVATE_H_

-#ifdef DYNAREC
-typedef struct dynablocklist_s dynablocklist_t;
-#endif
-
 typedef struct library_s library_t;
 typedef struct needed_libs_s needed_libs_t;
 typedef struct kh_mapsymbols_s kh_mapsymbols_t;
@@ -1,3 +1,25 @@
+#include <string.h>
+
+#define F8      *(uint8_t*)(addr++)
+#define F8S     *(int8_t*)(addr++)
+#define F16     *(uint16_t*)(addr+=2, addr-2)
+#define F16S    *(int16_t*)(addr+=2, addr-2)
+#define F32     *(uint32_t*)(addr+=4, addr-4)
+#define F32S    *(int32_t*)(addr+=4, addr-4)
+#define F32S64  (uint64_t)(int64_t)F32S
+#define F64     *(uint64_t*)(addr+=8, addr-8)
+#define F64S    *(int64_t*)(addr+=8, addr-8)
+#define PK(a)   *(uint8_t*)(addr+a)
+#ifdef DYNAREC
+#define STEP if(step) return 0;
+#define STEP2 if(step) {R_EIP = addr; return 0;}
+#define STEP3 if(*step) (*step)++;
+#else
+#define STEP
+#define STEP2
+#define STEP3
+#endif
+
 // ModRM utilities macros
 #define getecommon(A, T) \
    if(!(nextop&0xC0)) { \
@@ -97,6 +119,17 @@
    if((nextop&0xC0)==0xC0) { \
        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
    } else getecommon(A, reg32_t)
+#define testeb(A, O) \
+    if((nextop&0xC0)==0xC0) { \
+        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
+    } else { \
+        reg32_t* ret; \
+        getecommono(ret, reg32_t, O);\
+        test->memsize = 1; \
+        test->memaddr = (uintptr_t)ret;\
+        test->mem[0] = ret->byte[0];\
+        A = (reg32_t*)test->mem;\
+    }
 #define getebo(A, O)          \
    if((nextop&0xC0)==0xC0) { \
        A = (reg32_t*)&emu->regs[(nextop&3)].byte[((nextop&0x4)>>2)]; \
@@ -109,14 +142,47 @@
    if((nextop&0xC0)==0xC0) { \
        A = &emu->regs[(nextop&7)]; \
    } else getecommono(A, reg32_t, O)
+#define tested(A, SZ, O) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->regs[(nextop&7)]; \
+    } else {\
+        reg32_t* ret; \
+        getecommono(ret, reg32_t, O);\
+        test->memsize = SZ; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->dword, SZ);\
+        A = (reg32_t*)test->mem;\
+    }
 #define getem(A) \
    if((nextop&0xC0)==0xC0) { \
        A = &emu->mmx[(nextop&7)]; \
    } else getecommon(A, mmx87_regs_t)
+#define testem(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->mmx[(nextop&7)]; \
+    } else {\
+        mmx87_regs_t* ret; \
+        getecommon(ret, mmx87_regs_t) \
+        test->memsize = 8; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->ub, 8);\
+        A = (mmx87_regs_t*)test->mem;\
+    }
 #define getex(A) \
    if((nextop&0xC0)==0xC0) { \
        A = &emu->xmm[(nextop&7)]; \
    } else getecommon(A, sse_regs_t)
+#define testex(A) \
+    if((nextop&0xC0)==0xC0) { \
+        A = &emu->xmm[(nextop&7)]; \
+    } else {\
+        sse_regs_t* ret; \
+        getecommon(ret, sse_regs_t);\
+        test->memsize = 16; \
+        test->memaddr = (uintptr_t)ret;\
+        memcpy(test->mem, ret->ub, 16);\
+        A = (sse_regs_t*)test->mem;\
+    }
 #define getew16(A)  \
    if((nextop&0xC0)==0xC0) { \
        A = &emu->regs[(nextop&7)]; \
@@ -125,16 +191,44 @@
    if((nextop&0xC0)==0xC0) {       \
        A = &emu->regs[(nextop&7)]; \
    } else getecommon16o(A, reg32_t, O)
+#define testtew16(A, O)                 \
+    if((nextop&0xC0)==0xC0) {           \
+        A = &emu->regs[(nextop&7)];     \
+    } else {                            \
+        reg32_t* ret;                   \
+        getecommon16o(A, reg32_t, O);   \
+        test->memsize = 2;              \
+        test->memaddr = (uintptr_t)ret; \
+        *(uint16_t*)test->mem = ret->word[0];\
+        A = (reg32_t*)test->mem;        \
+    }

 // Macros for ModR/M gets
+#ifdef TEST_INTERPRETER
+#define GET_EB      testeb(oped, 0)
+#define GET_ED      tested(oped, 4, 0)
+#define GET_ED8     tested(oped, 8, 0)
+#define GET_EDT     tested(oped, 10, 0)
+#define GET_ED_OFFS(o) tested(oped, 4, o)
+#define GET_EB_OFFS(o) testeb(oped, o)
+#define GET_EM      testem(opem)
+#define GET_EX      testex(opex)
+#define GET_EW16    testew16(oped, 0)
+#define GET_EW16_OFFS(o)    testtew16(oped, o)
+#else
 #define GET_EB      geteb(oped)
 #define GET_ED      geted(oped)
+#define GET_ED8     geted(oped)
+#define GET_EDT     geted(oped)
 #define GET_ED_OFFS(o) getedo(oped, o)
 #define GET_EB_OFFS(o) getebo(oped, o)
 #define GET_EM      getem(opem)
 #define GET_EX      getex(opex)
 #define GET_EW16    getew16(oped)
 #define GET_EW16_OFFS(o)    getew16o(oped, o)
+#endif
+#define GET_ED_     geted(oped)
+#define GET_EW16_   getew16(oped)
 #define EB          oped
 #define ED          oped
 #define EM          opem
@@ -1,555 +0,0 @@
-    opcode = F8;
-    goto *opcodes66[opcode];
-
-    #define GO(B, OP)                       \
-    _66_##B##_0: \
-        nextop = F8;               \
-        GET_EB;             \
-        EB->byte[0] = OP##8(emu, EB->byte[0], GB);  \
-        NEXT;                              \
-    _66_##B##_1:                               \
-        nextop = F8;                \
-        GET_EW;                   \
-        EW->word[0] = OP##16(emu, EW->word[0], GW.word[0]); \
-        NEXT;                              \
-    _66_##B##_2: \
-        nextop = F8;               \
-        GET_EB;                   \
-        GB = OP##8(emu, GB, EB->byte[0]); \
-        NEXT;                              \
-    _66_##B##_3:                               \
-        nextop = F8;                \
-        GET_EW;                   \
-        GW.word[0] = OP##16(emu, GW.word[0], EW->word[0]); \
-        NEXT;                              \
-    _66_##B##_4: \
-        R_AL = OP##8(emu, R_AL, F8); \
-        NEXT;                              \
-    _66_##B##_5:                               \
-        R_AX = OP##16(emu, R_AX, F16); \
-        NEXT;
-
-    GO(0x00, add)                   /* ADD 0x01 ~> 0x05 */
-    GO(0x08, or)                    /*  OR 0x09 ~> 0x0D */
-    GO(0x10, adc)                   /* ADC 0x11 ~> 0x15 */
-    GO(0x18, sbb)                   /* SBB 0x19 ~> 0x1D */
-    GO(0x20, and)                   /* AND 0x21 ~> 0x25 */
-    GO(0x28, sub)                   /* SUB 0x29 ~> 0x2D */
-    GO(0x30, xor)                   /* XOR 0x31 ~> 0x35 */
-    //GO(0x38, cmp)                   /* CMP 0x39 ~> 0x3D */
-    #undef GO
-    _66_0x06:                      /* PUSH ES */
-        Push16(emu, emu->segs[_ES]);
-        NEXT;
-    _66_0x07:                      /* POP ES */
-        emu->segs[_ES] = Pop16(emu);    // no check, no use....
-        emu->segs_serial[_ES] = 0;
-        NEXT;
-
-    _66_0x0F:                      /* 66 0f prefix */
-        emu->old_ip = R_EIP;
-        R_EIP = ip-2;   // don't count 66 0F yet
-        Run660F(emu); // implemented in Run660f.c
-        ip = R_EIP;
-        if(emu->quit) goto fini;
-        STEP
-        NEXT;
-
-    _66_0x1E:                      /* PUSH DS */
-        Push16(emu, emu->segs[_DS]);
-        NEXT;
-    _66_0x1F:                      /* POP DS */
-        emu->segs[_DS] = Pop16(emu);    // no check, no use....
-        emu->segs_serial[_DS] = 0;
-        NEXT;
-        
-    _66_0x26:                      /* ES: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    _66_0x2E:                      /* CS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    _66_0x36:                      /* SS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-
-    _66_0x39:
-        nextop = F8;
-        GET_EW;
-        cmp16(emu, EW->word[0], GW.word[0]);
-        NEXT;
-    _66_0x3B:
-        nextop = F8;
-        GET_EW;
-        cmp16(emu, GW.word[0], EW->word[0]);
-        NEXT;
-    _66_0x3D:
-        cmp16(emu, R_AX, F16);
-        NEXT;
-    _66_0x3E:                      /* DS: */
-        // ignored
-        opcode = F8;
-        goto *opcodes66[opcode];
-    
-    _66_0x40:
-    _66_0x41:
-    _66_0x42:
-    _66_0x43:
-    _66_0x44:
-    _66_0x45:
-    _66_0x46:
-    _66_0x47:                              /* INC Reg */
-        tmp8u = opcode&7;
-        emu->regs[tmp8u].word[0] = inc16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x48:
-    _66_0x49:
-    _66_0x4A:
-    _66_0x4B:
-    _66_0x4C:
-    _66_0x4D:
-    _66_0x4E:
-    _66_0x4F:                              /* DEC Reg */
-        tmp8u = opcode&7;
-        emu->regs[tmp8u].word[0] = dec16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x50:
-    _66_0x51:
-    _66_0x52:
-    _66_0x53:
-    _66_0x54:
-    _66_0x55:
-    _66_0x56:
-    _66_0x57:                              /* PUSH Reg */
-        tmp8u = opcode&7;
-        Push16(emu, emu->regs[tmp8u].word[0]);
-        NEXT;
-    _66_0x58:
-    _66_0x59:
-    _66_0x5A:
-    _66_0x5B:
-    _66_0x5C:                      
-    _66_0x5D:
-    _66_0x5E:
-    _66_0x5F:                              /* POP Reg */
-        tmp8u = opcode&7;
-        emu->regs[tmp8u].word[0] = Pop16(emu);
-        NEXT;
-    _66_0x60:                              /* PUSHA */
-        tmp16u = R_SP;
-        Push16(emu, R_AX);
-        Push16(emu, R_CX);
-        Push16(emu, R_DX);
-        Push16(emu, R_BX);
-        Push16(emu, tmp16u);
-        Push16(emu, R_BP);
-        Push16(emu, R_SI);
-        Push16(emu, R_DI);
-        NEXT;
-    _66_0x61:                              /* POPA */
-        R_DI = Pop16(emu);
-        R_SI = Pop16(emu);
-        R_BP = Pop16(emu);
-        R_ESP+=2;   // POP ESP
-        R_BX = Pop16(emu);
-        R_DX = Pop16(emu);
-        R_CX = Pop16(emu);
-        R_AX = Pop16(emu);
-        NEXT;
-
-    _66_0x64:
-        R_EIP = ip-2;
-        RunFS66(emu, GetFSBaseEmu(emu));
-        ip = R_EIP;
-        STEP;
-        NEXT;
-
-    _66_0x66:
-        goto _0x66; // 0x66 0x66 => can remove one 0x66
-
-    _66_0x68:                       /* PUSH u16 */
-        tmp16u = F16;
-        Push16(emu, tmp16u);
-        NEXT;
-    _66_0x69:                      /* IMUL Gw,Ew,Iw */
-        nextop = F8;
-        GET_EW;
-        tmp16u = F16;
-        GW.word[0] = imul16(emu, EW->word[0], tmp16u);
-        NEXT;
-    _66_0x6A:                      /* PUSH Ib (as signed word) */
-        tmp16s = F8S;
-        Push16(emu, (uint16_t)tmp16s);
-        NEXT;
-    _66_0x6B:                      /* IMUL Gw,Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        tmp16s = F8S;
-        GW.word[0] = imul16(emu, EW->word[0], (uint16_t)tmp16s);
-        NEXT;
-
-    _66_0x81:                              /* GRP3 Ew,Iw */
-    _66_0x83:                              /* GRP3 Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        if(opcode==0x81) 
-            tmp16u = F16;
-        else {
-            tmp16s = F8S;
-            tmp16u = (uint16_t)tmp16s;
-        }
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = add16(emu, EW->word[0], tmp16u); break;
-            case 1: EW->word[0] =  or16(emu, EW->word[0], tmp16u); break;
-            case 2: EW->word[0] = adc16(emu, EW->word[0], tmp16u); break;
-            case 3: EW->word[0] = sbb16(emu, EW->word[0], tmp16u); break;
-            case 4: EW->word[0] = and16(emu, EW->word[0], tmp16u); break;
-            case 5: EW->word[0] = sub16(emu, EW->word[0], tmp16u); break;
-            case 6: EW->word[0] = xor16(emu, EW->word[0], tmp16u); break;
-            case 7:               cmp16(emu, EW->word[0], tmp16u); break;
-        }
-        NEXT;
-
-    _66_0x85:                              /* TEST Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        test16(emu, EW->word[0], GW.word[0]);
-        NEXT;
-
-    _66_0x87:                              /* XCHG Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        tmp16u = GW.word[0];
-        GW.word[0] = EW->word[0];
-        EW->word[0] = tmp16u;
-        NEXT;
-
-    _66_0x89:                              /* MOV Ew,Gw */
-        nextop = F8;
-        GET_EW;
-        EW->word[0] = GW.word[0];
-        NEXT;
-
-    _66_0x8B:                              /* MOV Gw,Ew */
-        nextop = F8;
-        GET_EW;
-        GW.word[0] = EW->word[0];
-        NEXT;
-    _66_0x8C:                              /* MOV Ew,Seg */
-        nextop = F8;
-        GET_EW;
-        EW->word[0] = emu->segs[(nextop&0x38)>>3];
-        NEXT;
-    
-    _66_0x8E:                               /* MOV Seg,Ew */
-        nextop = F8;
-        GET_EW;
-        emu->segs[((nextop&0x38)>>3)] = EW->word[0];
-        emu->segs_serial[((nextop&0x38)>>3)] = 0;
-        NEXT;
-    _66_0x8F:                              /* POP Ew */
-        nextop = F8;
-        GET_EW;
-        EW->word[0] = Pop16(emu);
-        NEXT;
-    _66_0x90:                              /* NOP */
-        NEXT;
-
-    _66_0x91:
-    _66_0x92:
-    _66_0x93:
-    _66_0x94:
-    _66_0x95:
-    _66_0x96:
-    _66_0x97:                      /* XCHG reg,EAX */
-        tmp16u = R_AX;
-        R_AX = emu->regs[opcode&7].word[0];
-        emu->regs[opcode&7].word[0] = tmp16u;
-        NEXT;
-
-    _66_0x98:                               /* CBW */
-        emu->regs[_AX].sword[0] = emu->regs[_AX].sbyte[0];
-        NEXT;
-    _66_0x99:                              /* CWD */
-        R_DX=((R_AX & 0x8000)?0xFFFF:0x0000);
-        NEXT;
-
-    _66_0x9C:                              /* PUSHFW */
-        CHECK_FLAGS(emu);
-        Push16(emu, (uint16_t)emu->eflags.x32);
-        NEXT;
-    _66_0x9D:                              /* POPFW */
-        CHECK_FLAGS(emu);
-        emu->eflags.x32 &=0xffff0000;
-        emu->eflags.x32 |= (Pop16(emu) & 0x3F7FD7) | 0x2;
-        NEXT;
-
-    _66_0xA1:                              /* MOV AX,Ow */
-        R_AX = *(uint16_t*)F32;
-        NEXT;
-    _66_0xA3:                              /* MOV Ow,AX */
-        *(uint16_t*)F32 = R_AX;
-        NEXT;
-    _66_0xA5:                              /* MOVSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        *(uint16_t*)R_EDI = *(uint16_t*)R_ESI;
-        R_EDI += tmp8s;
-        R_ESI += tmp8s;
-        NEXT;
-
-    _66_0xA7:                              /* CMPSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        tmp16u  = *(uint16_t*)R_EDI;
-        tmp16u2 = *(uint16_t*)R_ESI;
-        R_EDI += tmp8s;
-        R_ESI += tmp8s;
-        cmp16(emu, tmp16u2, tmp16u);
-        NEXT;
-
-    _66_0xA9:                             /* TEST AX,Iw */
-        test16(emu, R_AX, F16);
-        NEXT;
-
-    _66_0xAB:                              /* STOSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        *(uint16_t*)R_EDI = R_AX;
-        R_EDI += tmp8s;
-        NEXT;
-    _66_0xAD:                              /* LODSW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        R_AX = *(uint16_t*)R_ESI;
-        R_ESI += tmp8s;
-        NEXT;
-    _66_0xAF:                              /* SCASW */
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        cmp16(emu, R_AX, *(uint16_t*)R_EDI);
-        R_EDI += tmp8s;
-        NEXT;
-
-    _66_0xB8:                              /* MOV AX,Iw */
-    _66_0xB9:                              /* MOV CX,Iw */
-    _66_0xBA:                              /* MOV DX,Iw */
-    _66_0xBB:                              /* MOV BX,Iw */
-    _66_0xBC:                              /*    ...     */
-    _66_0xBD:
-    _66_0xBE:
-    _66_0xBF:
-        emu->regs[opcode&7].word[0] = F16;
-        NEXT;
-
-    _66_0xC1:                              /* GRP2 Ew,Ib */
-        nextop = F8;
-        GET_EW;
-        tmp8u = F8 /*& 0x1f*/;
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = rol16(emu, EW->word[0], tmp8u); break;
-            case 1: EW->word[0] = ror16(emu, EW->word[0], tmp8u); break;
-            case 2: EW->word[0] = rcl16(emu, EW->word[0], tmp8u); break;
-            case 3: EW->word[0] = rcr16(emu, EW->word[0], tmp8u); break;
-            case 4:
-            case 6: EW->word[0] = shl16(emu, EW->word[0], tmp8u); break;
-            case 5: EW->word[0] = shr16(emu, EW->word[0], tmp8u); break;
-            case 7: EW->word[0] = sar16(emu, EW->word[0], tmp8u); break;
-        }
-        NEXT;
-
-    _66_0xC7:                              /* MOV Ew,Iw */
-        nextop = F8;
-        GET_EW;
-        EW->word[0] = F16;
-        NEXT;
-
-    _66_0xCB:                               /* FAR RET */
-        ip = Pop(emu);
-        emu->segs[_CS] = Pop(emu);    // no check, no use....
-        emu->segs_serial[_CS] = 0;
-        // need to check status of CS register!
-        STEP
-        NEXT;
-    _66_0xCC:                              /* INT3 */
-        emu->old_ip = R_EIP;
-        R_EIP = ip-1;
-        if(my_context->signals[SIGTRAP])
-            raise(SIGTRAP);
-        ip = R_EIP;
-        if(emu->quit) goto fini;
-        STEP
-        NEXT;
-    _66_0xD1:                              /* GRP2 Ew,1  */
-    _66_0xD3:                              /* GRP2 Ew,CL */
-        nextop = F8;
-        GET_EW;
-        tmp8u=(opcode==0xD3)?R_CL:1;
-        switch((nextop>>3)&7) {
-            case 0: EW->word[0] = rol16(emu, EW->word[0], tmp8u); break;
-            case 1: EW->word[0] = ror16(emu, EW->word[0], tmp8u); break;
-            case 2: EW->word[0] = rcl16(emu, EW->word[0], tmp8u); break;
-            case 3: EW->word[0] = rcr16(emu, EW->word[0], tmp8u); break;
-            case 4: 
-            case 6: EW->word[0] = shl16(emu, EW->word[0], tmp8u); break;
-            case 5: EW->word[0] = shr16(emu, EW->word[0], tmp8u); break;
-            case 7: EW->word[0] = sar16(emu, EW->word[0], tmp8u); break;
-        }
-        NEXT;
-    
-    _66_0xD9:
-        emu->old_ip = R_EIP;
-        R_EIP = ip;
-        Run66D9(emu);
-        ip = R_EIP;
-        if(emu->quit) goto fini;
-        STEP
-        NEXT;
-
-    _66_0xDD:
-        emu->old_ip = R_EIP;
-        R_EIP = ip;
-        Run66DD(emu);
-        ip = R_EIP;
-        if(emu->quit) goto fini;
-        STEP
-        NEXT;
-
-    _66_0xF0:                      /* LOCK prefix */
-        emu->old_ip = R_EIP;
-        R_EIP = ip - 2;
-        RunLock66(emu);
-        ip = R_EIP;
-        if(emu->quit) goto fini;
-        STEP
-        NEXT;
-
-    _66_0xF2:                      /* REPNZ prefix */
-    _66_0xF3:                      /* REPZ prefix */
-        nextop = F8;
-        tmp8s = ACCESS_FLAG(F_DF)?-2:+2;
-        tmp32u = R_ECX;
-        switch(nextop) {
-            case 0xA4:              /* REP MOVSB */
-                while(tmp32u) {
-                    --tmp32u;
-                    *(uint8_t*)R_EDI = *(uint8_t*)R_ESI;
-                    R_EDI += tmp8s;
-                    R_ESI += tmp8s;
-                }
-                break;
-            case 0xA5:              /* REP MOVSW */
-                while(tmp32u) {
-                    --tmp32u;
-                    *(uint16_t*)R_EDI = *(uint16_t*)R_ESI;
-                    R_EDI += tmp8s;
-                    R_ESI += tmp8s;
-                }
-                break;
-            case 0xA7:              /* REP(N)Z CMPSW */
-                tmp16u = 0;
-                tmp16u2 = 0;
-                while(tmp32u) {
-                    --tmp32u;
-                    tmp16u  = *(uint16_t*)R_EDI;
-                    tmp16u2 = *(uint16_t*)R_ESI;
-                    R_EDI += tmp8s;
-                    R_ESI += tmp8s;
-                    if((tmp16u==tmp16u2)==(opcode==0xF2))
-                        break;
-                }
-                if(R_ECX) cmp16(emu, tmp16u2, tmp16u);
-                break;
-            case 0xAB:              /* REP STOSW */
-                while(tmp32u) {
-                    --tmp32u;
-                    *(uint16_t*)R_EDI = R_AX;
-                    R_EDI += tmp8s;
-                }
-                break;
-            case 0xAD:              /* REP LODSW */
-                while(tmp32u) {
-                    --tmp32u;
-                    R_AX = *(uint16_t*)R_ESI;
-                    R_ESI += tmp8s;
-                }
-                break;
-            case 0xAF:              /* REP(N)Z SCASW */
-                tmp16u = 0;
-                while(tmp32u) {
-                    --tmp32u;
-                    tmp16u = *(uint16_t*)R_EDI;
-                    R_EDI += tmp8s;
-                    if((R_AX==tmp16u)==(opcode==0xF2))
-                        break;
-                }
-                if(R_ECX) cmp16(emu, R_AX, tmp16u);
-                break;
-            default:
-                goto _default;
-        }
-        R_ECX = tmp32u;
-        NEXT;
-
-    _66_0xF5:                      /* CMC */
-        CHECK_FLAGS(emu);
-        CONDITIONAL_SET_FLAG(!ACCESS_FLAG(F_CF), F_CF);
-        NEXT;
-
-    _66_0xF7:                      /* GRP3 Ew(,Iw) */
-        nextop = F8;
-        GET_EW;
-        switch((nextop>>3)&7) {
-            case 0: 
-            case 1:                 /* TEST Ew,Iw */
-                test16(emu, EW->word[0], F16);
-                break;
-            case 2:                 /* NOT Ew */
-                EW->word[0] = not16(emu, EW->word[0]);
-                break;
-            case 3:                 /* NEG Ew */
-                EW->word[0] = neg16(emu, EW->word[0]);
-                break;
-            case 4:                 /* MUL AX,Ew */
-                mul16(emu, EW->word[0]);
-                break;
-            case 5:                 /* IMUL AX,Ew */
-                imul16_eax(emu, EW->word[0]);
-                break;
-            case 6:                 /* DIV Ew */
-                div16(emu, EW->word[0]);
-                break;
-            case 7:                 /* IDIV Ew */
-                idiv16(emu, EW->word[0]);
-                break;
-        }
-        NEXT;
-    _66_0xF8:                       /* CLC */
-        CHECK_FLAGS(emu);
-        CLEAR_FLAG(F_CF);
-        NEXT;
-    _66_0xF9:                       /* STC */
-        CHECK_FLAGS(emu);
-        SET_FLAG(F_CF);
-        NEXT;
-
-    _66_0xFF:                      /* GRP 5 Ew */
-        nextop = F8;
-        GET_EW;
-        switch((nextop>>3)&7) {
-            case 0:                 /* INC Ed */
-                EW->word[0] = inc16(emu, EW->word[0]);
-                break;
-            case 1:                 /* DEC Ed */
-                EW->word[0] = dec16(emu, EW->word[0]);
-                break;
-            case 6:
-                Push16(emu, EW->word[0]);
-                break;
-            default:
-                emu->old_ip = R_EIP;
-                R_EIP = ip;
-                printf_log(LOG_NONE, "Illegal Opcode 66 %02X %02X\n", opcode, nextop);
-                emu->quit=1;
-                emu->error |= ERR_ILLEGAL;
-                goto fini;
-        }
-        NEXT;
@@ -54,17 +54,6 @@ uint32_t* GetParityTab()
    return x86emu_parity_tab;
 }

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

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

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

+void CopyEmu(x86emu_t *newemu, const x86emu_t* emu)
+{
+	memcpy(newemu->regs, emu->regs, sizeof(emu->regs));
+    memcpy(&newemu->ip, &emu->ip, sizeof(emu->ip));
+	memcpy(&newemu->eflags, &emu->eflags, sizeof(emu->eflags));
+    newemu->old_ip = emu->old_ip;
+    memcpy(newemu->segs, emu->segs, sizeof(emu->segs));
+    memcpy(newemu->segs_serial, emu->segs_serial, sizeof(emu->segs_serial));
+    memcpy(newemu->segs_offs, emu->segs_offs, sizeof(emu->segs_offs));
+	memcpy(newemu->x87, emu->x87, sizeof(emu->x87));
+	memcpy(newemu->mmx, emu->mmx, sizeof(emu->mmx));
+    memcpy(newemu->xmm, emu->xmm, sizeof(emu->xmm));
+    memcpy(newemu->fpu_ld, emu->fpu_ld, sizeof(emu->fpu_ld));
+    memcpy(newemu->fpu_ll, emu->fpu_ll, sizeof(emu->fpu_ll));
+	memcpy(newemu->p_regs, emu->p_regs, sizeof(emu->p_regs));
+	newemu->cw = emu->cw;
+	newemu->sw = emu->sw;
+	newemu->top = emu->top;
+    newemu->fpu_stack = emu->fpu_stack;
+    newemu->df = emu->df;
+    newemu->op1 = emu->op1;
+    newemu->op2 = emu->op2;
+    newemu->res = emu->res;
+    newemu->mxcsr = emu->mxcsr;
+    newemu->quit = emu->quit;
+    newemu->error = emu->error;
+}
+
 box86context_t* GetEmuContext(x86emu_t* emu)
 {
    return emu->context;
@@ -311,7 +337,7 @@ void ResetFlags(x86emu_t *emu)
 const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip)
 {
    static char buff[800];
-    char* regname[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
+    static const char* regname[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
    char tmp[80];
    buff[0] = '\0';
    if(trace_emm) {
@@ -23,6 +23,17 @@ typedef struct forkpty_s {
    void*   f;  // forkpty function
 } forkpty_t;

+typedef struct x86emu_s x86emu_t;
+
+typedef struct x86test_s {
+    x86emu_t*   emu;
+    uintptr_t   memaddr;
+    int         memsize;
+    int         test;
+    int         clean;
+    uint8_t     mem[16];
+} x86test_t;
+
 typedef struct x86emu_s {
    // cpu
 	reg32_t     regs[8];
@@ -63,6 +74,7 @@ typedef struct x86emu_s {
    int         exit;
    int         quitonlongjmp;  // quit if longjmp is called
    int         longjmp;        // if quit because of longjmp
+    x86test_t   test;       // used for dynarec testing
    // parent context
    box86context_t *context;
    // cpu helpers
@@ -0,0 +1,459 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test64(x86test_t *test, int seg, uintptr_t addr)
+#else
+uintptr_t Run64(x86emu_t *emu, int seg, uintptr_t addr)
+#endif
+{
+    uint8_t opcode = F8;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uintptr_t tlsdata = (seg==_FS)?GetFSBaseEmu(emu):GetGSBaseEmu(emu);
+    switch(opcode) {
+        case 0x01:              /* ADD GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = add32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x03:              /* ADD Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = add32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x06:              /* PUSH ES */
+            Push(emu, emu->segs[_ES]);
+            break;
+        case 0x07:             /* POP ES */
+            emu->segs[_ES] = Pop(emu);
+            emu->segs_serial[_ES] = 0;
+            break;
+
+        case 0x0f:
+            #ifdef TEST_INTERPRETER
+            return Test640F(test, tlsdata, addr);
+            #else
+            return Run640F(emu, tlsdata, addr);
+            #endif
+
+        case 0x11:              /* ADC GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = adc32(emu, ED->dword[0], GD.dword[0]);
+            break;
+        
+        case 0x13:              /* ADC Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = adc32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x21:              /* AND Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = and32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x23:              /* AND Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = and32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x29:              /* SUB GS:Ed, Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = sub32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x2B:              /* SUB Gd, GS:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = sub32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x31:              /* XOR Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = xor32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x33:              /* XOR Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = xor32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x39:              /* CMP Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            cmp32(emu, ED->dword[0], GD.dword[0]);
+            break;
+
+        case 0x3B:              /* CMP Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            cmp32(emu, GD.dword[0], ED->dword[0]);
+            break;
+        
+        case 0x40:
+        case 0x41:
+        case 0x42:
+        case 0x43:
+        case 0x44:
+        case 0x45:
+        case 0x46:
+        case 0x47:             /* INC Reg */
+            tmp8u = opcode&7;
+            emu->regs[tmp8u].dword[0] = inc32(emu, emu->regs[tmp8u].dword[0]);
+            break;
+        case 0x48:
+        case 0x49:
+        case 0x4A:
+        case 0x4B:
+        case 0x4C:
+        case 0x4D:
+        case 0x4E:
+        case 0x4F:            /* DEC Reg */
+            tmp8u = opcode&7;
+            emu->regs[tmp8u].dword[0] = dec32(emu, emu->regs[tmp8u].dword[0]);
+            break;
+        case 0x50:              /* PUSH Reg */
+        case 0x51:
+        case 0x52:
+        case 0x53:
+        case 0x54:               /* PUSH ESP */
+        case 0x55:
+        case 0x56:
+        case 0x57:              
+        case 0x58:              /* POP Reg */
+        case 0x59:
+        case 0x5A:
+        case 0x5B:
+        case 0x5C:              /* POP ESP */
+        case 0x5D:
+        case 0x5E:
+        case 0x5F:
+            // Segment override if for memory loc, no stack segment 
+            return addr-1;
+
+        case 0x64:              /* FS: */
+            // so just ignore that GS: prefix then
+            #ifdef TEST_INTERPRETER
+            return Test64(test, _FS, addr-1); // put FS back
+            #else
+            return Run64(emu, _FS, addr-1); // put FS back
+            #endif
+
+        case 0x66:
+            #ifdef TEST_INTERPRETER
+            return Test6466(test, tlsdata, addr);
+            #else
+            return Run6466(emu, tlsdata, addr);
+            #endif
+        case 0x67:
+            #ifdef TEST_INTERPRETER
+            return Test6467(test, tlsdata, addr);
+            #else
+            return Run6467(emu, tlsdata, addr);
+            #endif
+
+        case 0x69:              /* IMUL Gd,Ed,Id */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            tmp32u = F32;
+            GD.dword[0] = imul32(emu, ED->dword[0], tmp32u);
+            break;
+        
+        case 0x70:
+        case 0x71:
+        case 0x72:
+        case 0x73:
+        case 0x74:
+        case 0x75:
+        case 0x76:
+        case 0x77:
+        case 0x78:
+        case 0x79:
+        case 0x7A:
+        case 0x7B:
+        case 0x7C:
+        case 0x7D:
+        case 0x7E:
+        case 0x7F:
+            // just ignore the prefix
+            return addr-1;
+        case 0x80:             /* GRP Eb,Ib */
+        case 0x82:             // 0x82 and 0x80 are the same opcodes it seems?
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = F8;
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = add8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] =  or8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = adc8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = sbb8(emu, EB->byte[0], tmp8u); break;
+                case 4: EB->byte[0] = and8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = sub8(emu, EB->byte[0], tmp8u); break;
+                case 6: EB->byte[0] = xor8(emu, EB->byte[0], tmp8u); break;
+                case 7:               cmp8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+        case 0x81:              /* GRP Ed,Id */
+        case 0x83:              /* GRP Ed,Ib */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            if(opcode==0x83) {
+                tmp32s = F8S;
+                tmp32u = (uint32_t)tmp32s;
+            } else
+                tmp32u = F32;
+            switch((nextop>>3)&7) {
+                case 0: ED->dword[0] = add32(emu, ED->dword[0], tmp32u); break;
+                case 1: ED->dword[0] =  or32(emu, ED->dword[0], tmp32u); break;
+                case 2: ED->dword[0] = adc32(emu, ED->dword[0], tmp32u); break;
+                case 3: ED->dword[0] = sbb32(emu, ED->dword[0], tmp32u); break;
+                case 4: ED->dword[0] = and32(emu, ED->dword[0], tmp32u); break;
+                case 5: ED->dword[0] = sub32(emu, ED->dword[0], tmp32u); break;
+                case 6: ED->dword[0] = xor32(emu, ED->dword[0], tmp32u); break;
+                case 7:                cmp32(emu, ED->dword[0], tmp32u); break;
+            }
+            break;
+        case 0x87:
+            nextop = F8;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+            GET_ED_OFFS(tlsdata);
+            if((nextop&0xC0)==0xC0) {
+                tmp32u = GD.dword[0];
+                GD.dword[0] = ED->dword[0];
+                ED->dword[0] = tmp32u;
+            } else {
+                if(((uintptr_t)ED)&3)
+                {
+                    // not aligned, dont't try to "LOCK"
+                    tmp32u = ED->dword[0];
+                    ED->dword[0] = GD.dword[0];
+                    GD.dword[0] = tmp32u;
+                } else {
+                    // XCHG is supposed to automaticaly LOCK memory bus
+                    GD.dword[0] = arm_lock_xchg(ED, GD.dword[0]);
+                }
+            }
+#else
+            GET_ED_OFFS(tlsdata);
+            if((nextop&0xC0)!=0xC0)
+                pthread_mutex_lock(&emu->context->mutex_lock); // XCHG always LOCK (but when accessing memory only)
+            tmp32u = GD.dword[0];
+            GD.dword[0] = ED->dword[0];
+            ED->dword[0] = tmp32u;
+            if((nextop&0xC0)!=0xC0)
+                pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+            break;
+            
+        case 0x89:              /* MOV Ed,Gd */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = GD.dword[0];
+            break;
+        case 0x8A:              /* MOV Gb,GS:Eb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            GB = EB->byte[0];
+            break;
+        case 0x8B:              /* MOV Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = ED->dword[0];
+            break;
+        case 0x8C:              /* MOV Seg:Ed, Seg */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = emu->segs[((nextop&0x38)>>3)];
+            break;
+        case 0x8E:              /* MOV Seg,Seg:Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            emu->segs[((nextop&0x38)>>3)] = ED->word[0];
+            emu->segs_serial[((nextop&0x38)>>3)] = 0;
+            break;
+        case 0x8F:              /* POP Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = Pop(emu);
+            break;
+
+        case 0x9C:              /* PUSHFD */
+            // Segment override if for memory loc, no stack segment 
+            return addr-1;   // so ignore prefix and continue
+
+        case 0xA0:              /* MOV AL,Ob */
+            tmp32s = F32S;
+            R_AL = *(uint8_t*)((tlsdata) + tmp32s);
+            break;
+        case 0xA1:              /* MOV EAX,Ov */
+            tmp32s = F32S;
+            R_EAX = *(uint32_t*)((tlsdata) + tmp32s);
+            break;
+        case 0xA2:              /* MOV Ob,AL */
+            tmp32s = F32S;
+            *(uint8_t*)((tlsdata) + tmp32s) = R_AL;
+            break;
+        case 0xA3:             /* MOV Od,EAX */
+            tmp32s = F32S;
+            *(uint32_t*)((tlsdata) + tmp32s) = R_EAX;
+            break;
+
+        case 0xA8:            /* TEST AL, Ib */
+            test8(emu, R_AL, F8);
+            break;
+
+        case 0xC0:             /* GRP2 Eb,Ib */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = F8/* & 0x1f*/; // masking done in each functions
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = rol8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] = ror8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = rcl8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = rcr8(emu, EB->byte[0], tmp8u); break;
+                case 4:
+                case 6: EB->byte[0] = shl8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = shr8(emu, EB->byte[0], tmp8u); break;
+                case 7: EB->byte[0] = sar8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+
+        case 0xC6:              /* MOV Eb,Ib */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            EB->byte[0] = F8;
+            break;
+        case 0xC7:              /* MOV Ed,Id */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            ED->dword[0] = F32;
+            break;
+
+        case 0xD0:                      /* GRP2 Eb,1 */
+        case 0xD2:                      /* GRP2 Eb,CL */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            tmp8u = (opcode==0xD0)?1:R_CL;
+            switch((nextop>>3)&7) {
+                case 0: EB->byte[0] = rol8(emu, EB->byte[0], tmp8u); break;
+                case 1: EB->byte[0] = ror8(emu, EB->byte[0], tmp8u); break;
+                case 2: EB->byte[0] = rcl8(emu, EB->byte[0], tmp8u); break;
+                case 3: EB->byte[0] = rcr8(emu, EB->byte[0], tmp8u); break;
+                case 4: 
+                case 6: EB->byte[0] = shl8(emu, EB->byte[0], tmp8u); break;
+                case 5: EB->byte[0] = shr8(emu, EB->byte[0], tmp8u); break;
+                case 7: EB->byte[0] = sar8(emu, EB->byte[0], tmp8u); break;
+            }
+            break;
+        case 0xD1:                      /* GRP2 Ed,1 */
+        case 0xD3:                      /* GRP2 Ed,CL */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            tmp8u = (opcode==0xD1)?1:R_CL;
+            switch((nextop>>3)&7) {
+                case 0: ED->dword[0] = rol32(emu, ED->dword[0], tmp8u); break;
+                case 1: ED->dword[0] = ror32(emu, ED->dword[0], tmp8u); break;
+                case 2: ED->dword[0] = rcl32(emu, ED->dword[0], tmp8u); break;
+                case 3: ED->dword[0] = rcr32(emu, ED->dword[0], tmp8u); break;
+                case 4: 
+                case 6: ED->dword[0] = shl32(emu, ED->dword[0], tmp8u); break;
+                case 5: ED->dword[0] = shr32(emu, ED->dword[0], tmp8u); break;
+                case 7: ED->dword[0] = sar32(emu, ED->dword[0], tmp8u); break;
+            }
+            break;
+
+        case 0xE9:
+        case 0xEB:
+            return addr-1;       // ignore FS: to execute regular opcode
+            break;
+
+        case 0xFF:              /* GRP 5 Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            switch((nextop>>3)&7) {
+                case 0:                 /* INC Ed */
+                    ED->dword[0] = inc32(emu, ED->dword[0]);
+                    break;
+                case 1:                 /* DEC Ed */
+                    ED->dword[0] = dec32(emu, ED->dword[0]);
+                    break;
+                case 2:                 /* CALL NEAR Ed */
+                    R_EIP = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                    Push(emu, addr);
+                    addr = R_EIP;
+                    break;
+                case 3:                 /* CALL FAR Ed */
+                    if(nextop>0xc0) {
+                        printf_log(LOG_NONE, "Illegal Opcode %02X %02X\n", opcode, nextop);
+                        emu->quit=1;
+                        emu->error |= ERR_ILLEGAL;
+                        return 0;
+                    } else {
+                        Push16(emu, R_CS);
+                        Push(emu, addr);
+                        addr = ED->dword[0];
+                        R_CS = (ED+1)->word[0];
+                    }
+                    break;
+                case 4:                 /* JMP NEAR Ed */
+                    addr = (uintptr_t)getAlternate((void*)ED->dword[0]);
+                    break;
+                case 5:                 /* JMP FAR Ed */
+                    if(nextop>0xc0) {
+                        printf_log(LOG_NONE, "Illegal Opcode 0x%02X 0x%02X\n", opcode, nextop);
+                        emu->quit=1;
+                        emu->error |= ERR_ILLEGAL;
+                        return 0;
+                    } else {
+                        addr = ED->dword[0];
+                        R_CS = (ED+1)->word[0];
+                    }
+                    break;
+                case 6:                 /* Push Ed */
+                    Push(emu, ED->dword[0]);
+                    break;
+                default:
+                    printf_log(LOG_NONE, "Illegal Opcode 0x%02X 0x%02X\n", opcode, nextop);
+                    emu->quit=1;
+                    emu->error |= ERR_ILLEGAL;
+                    return 0;
+            }
+            break;
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,55 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test640F(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run640F(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0xAF:                      /* IMUL Gd,Ed */
+            nextop = F8;
+            GET_ED_OFFS(tlsdata);
+            GD.dword[0] = imul32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0xB6:                              /* MOVZX Gd,GS:Eb */
+            nextop = F8;
+            GET_EB_OFFS(tlsdata);
+            GD.dword[0] = EB->byte[0];
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,55 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6466(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run6466(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0x03:                              /* ADD Gw, FS:Ew */
+            nextop = F8;
+            GET_EW_OFFS(tlsdata);
+            GW.word[0] = add16(emu, GW.word[0], EW->word[0]);
+            break;
+
+        case 0x8B:                              /* MOV Gw,FS:Ew */
+            nextop = F8;
+            GET_EW_OFFS(tlsdata);
+            GW.word[0] = EW->word[0];
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,92 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6467(x86test_t *test, uintptr_t tlsdata, uintptr_t addr)
+#else
+uintptr_t Run6467(x86emu_t *emu, uintptr_t tlsdata, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0x3B:                              /* CMP GD, FS:Ed16 */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            cmp32(emu, GD.dword[0], ED->dword[0]);
+            break;
+
+        case 0x89:                              /* MOV ED16,Gd */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            ED->dword[0] = GD.dword[0];
+            break;
+
+        case 0x8B:                              /* MOV Gd,Ed16 */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            GD.dword[0] = ED->dword[0];
+            break;
+
+        case 0x8F:                              /* POP FS:Ed */
+            nextop = F8;
+            GET_EW16_OFFS(tlsdata);
+            ED->dword[0] = Pop(emu);
+            break;
+
+        case 0xA1:                              /* MOV EAX,Ov16 */
+            tmp32u = F16;
+            R_EAX = *(uint32_t*)(tlsdata + tmp32u);
+            break;
+        case 0xA3:                              /* MOV Ov16,EAX */
+            tmp32u = F16;
+            #ifdef TEST_INTERPRETER
+            test->memaddr = tlsdata + tmp32u;
+            test->memsize = 4;
+            *(uint32_t*)(test->mem) = R_EAX;
+            #else
+            *(uint32_t*)(tlsdata + tmp32u) = R_EAX;
+            #endif
+            break;
+        case 0xFF:                              /* GRP 5 Ed */
+            nextop = F8;
+            switch((nextop>>3)&7) {
+                case 6:                         /* Push Ed */
+                    GET_EW16_OFFS(tlsdata);
+                    Push(emu, ED->dword[0]);
+                    break;
+                default:
+                    return 0;
+            }
+            break;
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -14,19 +14,14 @@
 #include "x86trace.h"
 #include "box86context.h"

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

-void Run660F(x86emu_t *emu)
+#ifdef TEST_INTERPRETER
+uintptr_t Test660F(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run660F(x86emu_t *emu, uintptr_t addr)
+#endif
 {
-    uintptr_t ip = R_EIP+2; //skip 66 0F
    uint8_t opcode;
    uint8_t nextop;
    reg32_t *oped;
@@ -43,6 +38,9 @@ void Run660F(x86emu_t *emu)
    #endif


+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
    opcode = F8;
    switch(opcode) {

@@ -144,8 +142,7 @@ void Run660F(x86emu_t *emu)
                EW->word[0] = 0;    // dummy operation. Should be CR0 -> Ew
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x10:                      /* MOVUPD Gx, Ex */
@@ -160,7 +157,7 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x12:                      /* MOVLPD Gx, Eq */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        if((uintptr_t)ED & 7)
            memcpy(&GX.q[0], ED, 8);
        else
@@ -168,7 +165,7 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x13:                      /* MOVLPD Eq, Gx */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        if((uintptr_t)ED & 7)
            memcpy(ED, &GX.q[0], 8);
        else
@@ -187,18 +184,18 @@ void Run660F(x86emu_t *emu)
        break;
    case 0x16:                      /* MOVHPD Gx, Ed */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        GX.q[1] = *(uint64_t*)ED;
        break;
    case 0x17:                      /* MOVHPD Ed, Gx */
        nextop = F8;
-        GET_ED;
+        GET_ED8;
        *(uint64_t*)ED = GX.q[1];
        break;

    case 0x1F:                      /* NOP (multi-byte) */
        nextop = F8;
-        GET_ED;
+        GET_ED_;
        break;
    
    case 0x28:                      /* MOVAPD Gx, Ex */
@@ -219,16 +216,20 @@ void Run660F(x86emu_t *emu)
        GX.d[0] = EM->sd[0];
        GX.d[1] = EM->sd[1];
        break;
-
-
+    case 0x2B:                      /* MOVNTPD Ex, Gx */
+        nextop = F8;
+        GET_EX;
+        EX->q[0] = GX.q[0]; // Address needs to ba aligned for MOVNTPD
+        EX->q[1] = GX.q[1];
+        break;
    case 0x2C:                      /* CVTTPD2PI Gm, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(!isfinite(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GM.ud[0] = 0x80000000;
        else
            GM.sd[0] = EX->d[0];
-        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>0x7fffffff)
+        if(!isfinite(EX->d[1]) || EX->d[1]>(double)0x7fffffff || EX->d[1]<-(double)0x80000000U)
            GM.ud[1] = 0x80000000;
        else
            GM.sd[1] = EX->d[1];
@@ -237,7 +238,7 @@ void Run660F(x86emu_t *emu)
        nextop = F8;
        GET_EX;
        for(int i=0; i<2; ++i)
-            if(isnan(EX->d[i]) || isinf(EX->d[i]) || EX->d[i]>0x7fffffff)
+            if(!isfinite(EX->d[i]) || EX->d[i]>(double)0x7fffffff || EX->d[i]<-(double)0x80000000U)
                GM.ud[i] = 0x80000000;
            else
                switch(emu->mxcsr.f.MXCSR_RC) {
@@ -347,7 +348,34 @@ void Run660F(x86emu_t *emu)
                        GX.sw[4+i] = EX->sw[i*2+0] - EX->sw[i*2+1];
                }
                break;
-                
+            case 0x06:  /* PHSUBD Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<2; ++i)
+                    GX.sd[i] = GX.sd[i*2+0] - GX.sd[i*2+1];
+                if(&GX == EX) {
+                    GX.q[1] = GX.q[0];
+                } else {
+                    for (int i=0; i<2; ++i)
+                        GX.sd[2+i] = EX->sd[i*2+0] - EX->sd[i*2+1];
+                }
+                break;
+            case 0x07:  /* PHSUBSW Gx, Ex */
+                nextop = F8;
+                GET_EX;
+                for (int i=0; i<4; ++i) {
+                    tmp32s = GX.sw[i*2+0]-GX.sw[i*2+1];
+                    GX.sw[i] = (tmp32s<-32768)?-32768:((tmp32s>32767)?32767:tmp32s);
+                }
+                if(&GX == EX) {
+                    GX.q[1] = GX.q[0];
+                } else {
+                    for (int i=0; i<4; ++i) {
+                        tmp32s = EX->sw[i*2+0] - EX->sw[i*2+1];
+                        GX.sw[4+i] = (tmp32s<-32768)?-32768:((tmp32s>32767)?32767:tmp32s);
+                    }
+                }
+                break;
            case 0x08:  /* PSIGNB Gx, Ex */
                nextop = F8;
                GET_EX;
@@ -513,8 +541,7 @@ void Run660F(x86emu_t *emu)
                break;

            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;

@@ -528,28 +555,28 @@ void Run660F(x86emu_t *emu)
                switch((tmp8u & 4) ? (emu->mxcsr.f.MXCSR_RC) : (tmp8u & 3))
                {
                    case ROUND_Nearest:
-                        GX.f[0] = nearbyint(EX->f[0]);
-                        GX.f[1] = nearbyint(EX->f[1]);
-                        GX.f[2] = nearbyint(EX->f[2]);
-                        GX.f[3] = nearbyint(EX->f[3]);
+                        GX.f[0] = nearbyintf(EX->f[0]);
+                        GX.f[1] = nearbyintf(EX->f[1]);
+                        GX.f[2] = nearbyintf(EX->f[2]);
+                        GX.f[3] = nearbyintf(EX->f[3]);
                        break;
                    case ROUND_Down:
-                        GX.f[0] = floor(EX->f[0]);
-                        GX.f[1] = floor(EX->f[1]);
-                        GX.f[2] = floor(EX->f[2]);
-                        GX.f[3] = floor(EX->f[3]);
+                        GX.f[0] = floorf(EX->f[0]);
+                        GX.f[1] = floorf(EX->f[1]);
+                        GX.f[2] = floorf(EX->f[2]);
+                        GX.f[3] = floorf(EX->f[3]);
                        break;
                    case ROUND_Up:
-                        GX.f[0] = ceil(EX->f[0]);
-                        GX.f[1] = ceil(EX->f[1]);
-                        GX.f[2] = ceil(EX->f[2]);
-                        GX.f[3] = ceil(EX->f[3]);
+                        GX.f[0] = ceilf(EX->f[0]);
+                        GX.f[1] = ceilf(EX->f[1]);
+                        GX.f[2] = ceilf(EX->f[2]);
+                        GX.f[3] = ceilf(EX->f[3]);
                        break;
                    case ROUND_Chop:
-                        GX.f[0] = trunc(EX->f[0]);
-                        GX.f[1] = trunc(EX->f[1]);
-                        GX.f[2] = trunc(EX->f[2]);
-                        GX.f[3] = trunc(EX->f[3]);
+                        GX.f[0] = truncf(EX->f[0]);
+                        GX.f[1] = truncf(EX->f[1]);
+                        GX.f[2] = truncf(EX->f[2]);
+                        GX.f[3] = truncf(EX->f[3]);
                        break;
                }
                break;
@@ -584,16 +611,16 @@ void Run660F(x86emu_t *emu)
                switch((tmp8u & 4) ? (emu->mxcsr.f.MXCSR_RC) : (tmp8u & 3))
                {
                    case ROUND_Nearest:
-                        GX.f[0] = nearbyint(EX->f[0]);
+                        GX.f[0] = nearbyintf(EX->f[0]);
                        break;
                    case ROUND_Down:
-                        GX.f[0] = floor(EX->f[0]);
+                        GX.f[0] = floorf(EX->f[0]);
                        break;
                    case ROUND_Up:
-                        GX.f[0] = ceil(EX->f[0]);
+                        GX.f[0] = ceilf(EX->f[0]);
                        break;
                    case ROUND_Chop:
-                        GX.f[0] = trunc(EX->f[0]);
+                        GX.f[0] = truncf(EX->f[0]);
                        break;
                }
                break;
@@ -634,35 +661,12 @@ void Run660F(x86emu_t *emu)
                if(tmp8u>31)
                    {GX.q[0] = GX.q[1] = 0;}
                else
-                #if 0
-                    if(tmp8u>15) {
-                    tmp8u=(tmp8u-16)*8;
-                    if (tmp8u < 64) {
-                        GX.q[0] = (GX.q[0] >> tmp8u) | (GX.q[1] << (64 - tmp8u));
-                        GX.q[1] = (GX.q[1] >> tmp8u);
-                    } else {
-                        GX.q[0] = GX.q[1] >> (tmp8u - 64);
-                        GX.q[1] = 0;
-                    }                    
-                } else {
-                    tmp8u*=8;
-                    if (tmp8u < 64) {
-                        GX.q[0] = (EX->q[0] >> tmp8u) | (EX->q[1] << (64 - tmp8u));
-                        GX.q[1] = (EX->q[1] >> tmp8u) | (GX.q[0] << (64-tmp8u));
-                    } else {
-                        tmp8u -= 64;
-                        GX.q[0] = (EX->q[1] >> tmp8u) | (GX.q[0] << (64 - tmp8u));
-                        GX.q[1] = (GX.q[0] >> tmp8u) | (GX.q[1] >> (64 - tmp8u));
-                    }                    
-                }
-                #else
                {
                    for (int i=0; i<16; ++i, ++tmp8u)
                        eax1.ub[i] = (tmp8u>15)?((tmp8u>31)?0:GX.ub[tmp8u-16]):EX->ub[tmp8u];
                    GX.q[0] = eax1.q[0];
                    GX.q[1] = eax1.q[1];
                }
-                #endif
                break;

            case 0x16:      // PEXTRD ED, GX, u8
@@ -696,8 +700,7 @@ void Run660F(x86emu_t *emu)
                break;

            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;

@@ -782,23 +785,24 @@ void Run660F(x86emu_t *emu)
        nextop = F8;
        GET_EX;
        for(int i=0; i<4; ++i)
-            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>0x7fffffff)
+            // note, converting 0x7fffffff to float gives 0x80000000 basicaly, because of the loss or precision in the mantice. So uing >= here...
+            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>=(float)0x80000000U || EX->f[i]<-(float)0x80000000U)
                GX.sd[i] = 0x80000000;
-        else
-            switch(emu->mxcsr.f.MXCSR_RC) {
-                case ROUND_Nearest:
-                    GX.sd[i] = nearbyintf(EX->f[i]);
-                    break;
-                case ROUND_Down:
-                    GX.sd[i] = floorf(EX->f[i]);
-                    break;
-                case ROUND_Up:
-                    GX.sd[i] = ceilf(EX->f[i]);
-                    break;
-                case ROUND_Chop:
-                    GX.sd[i] = EX->f[i];
-                    break;
-            }
+            else
+                switch(emu->mxcsr.f.MXCSR_RC) {
+                    case ROUND_Nearest:
+                        GX.sd[i] = nearbyintf(EX->f[i]);
+                        break;
+                    case ROUND_Down:
+                        GX.sd[i] = floorf(EX->f[i]);
+                        break;
+                    case ROUND_Up:
+                        GX.sd[i] = ceilf(EX->f[i]);
+                        break;
+                    case ROUND_Chop:
+                        GX.sd[i] = EX->f[i];
+                        break;
+                }
        break;
    case 0x5C:                      /* SUBPD Gx, Ex */
        nextop = F8;
@@ -1016,8 +1020,7 @@ void Run660F(x86emu_t *emu)
                    for (int i=0; i<8; ++i) EX->uw[i] <<= tmp8u;
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x72:  /* GRP */
@@ -1043,8 +1046,7 @@ void Run660F(x86emu_t *emu)
                    for (int i=0; i<4; ++i) EX->ud[i] <<= tmp8u;
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x73:  /* GRP */
@@ -1096,8 +1098,7 @@ void Run660F(x86emu_t *emu)
                }
                break;
            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0x74:  /* PCMPEQB Gx,Ex */
@@ -1190,7 +1191,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1217,7 +1223,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1263,7 +1274,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u)) {
            SET_FLAG(F_CF);
@@ -1332,8 +1348,7 @@ void Run660F(x86emu_t *emu)
                break;

            default:
-                ip = R_EIP;
-                UnimpOpcode(emu);
+                return 0;
        }
        break;
    case 0xBB:                      /* BTC Ew,Gw */
@@ -1345,7 +1360,12 @@ void Run660F(x86emu_t *emu)
        GET_EW;
        if((nextop&0xC0)!=0xC0)
        {
+            #ifdef TEST_INTERPRETER
+            test->memaddr=((test->memaddr)+tmp32s*2);
+            *(uint16_t*)test->mem = *(uint16_t*)test->memaddr;
+            #else
            EW=(reg32_t*)(((uintptr_t)(EW))+(tmp32s*2));
+            #endif
        }
        if(EW->word[0] & (1<<tmp8u))
            SET_FLAG(F_CF);
@@ -1512,8 +1532,7 @@ void Run660F(x86emu_t *emu)
                if(EX->ub[i]&0x80)
                    GD.dword[0] |= (1<<i);
        } else {
-            ip = R_EIP;
-            UnimpOpcode(emu);
+            return 0;
        }
        break;
    case 0xD8:  /* PSUBUSB Gx,Ex */
@@ -1617,11 +1636,11 @@ void Run660F(x86emu_t *emu)
    case 0xE6:  /* CVTTPD2DQ Gx, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GX.sd[0] = 0x80000000;
        else
            GX.sd[0] = EX->d[0];
-        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>0x7fffffff)
+        if(isnan(EX->d[1]) || isinf(EX->d[1]) || EX->d[1]>(double)0x7fffffff || EX->d[1]<-(double)0x80000000U)
            GX.sd[1] = 0x80000000;
        else
            GX.sd[1] = EX->d[1];
@@ -1799,10 +1818,8 @@ void Run660F(x86emu_t *emu)
        break;
 
    default:
-        ip = R_EIP;
-        UnimpOpcode(emu);
+        return 0;
    }

-    R_EIP = ip;
-    return;
+    return addr;
 }
@@ -0,0 +1,103 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test66D9(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run66D9(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+
+    uint8_t opcode = F8;
+    switch(opcode) {
+        
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+        case 0xD0:
+        case 0xE0:
+        case 0xE5:
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xFC:
+        case 0xE1:
+        case 0xE4:
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            return 0;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 4:     /* FLDENV m */
+                    // warning, incomplete
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_loadenv(emu, (char*)ED, 1);
+                    #endif
+                    break;
+                case 6:     /* FNSTENV m */
+                    // warning, incomplete
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_savenv(emu, (char*)ED, 1);
+                    #endif
+                    break;
+                default:
+                    return 0;
+            }
+    }
+
+    return addr;
+}
@@ -0,0 +1,122 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test66DD(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t Run66DD(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+
+    uint8_t opcode = F8;
+    switch(opcode) {
+
+        case 0xC0:
+        case 0xC1:
+        case 0xC2:
+        case 0xC3:
+        case 0xC4:
+        case 0xC5:
+        case 0xC6:
+        case 0xC7:
+        case 0xC8:
+        case 0xC9:
+        case 0xCA:
+        case 0xCB:
+        case 0xCC:
+        case 0xCD:
+        case 0xCE:
+        case 0xCF:
+        case 0xD0:
+        case 0xE0:
+        case 0xE5:
+        case 0xE8:
+        case 0xE9:
+        case 0xEA:
+        case 0xEB:
+        case 0xEC:
+        case 0xED:
+        case 0xEE:
+        case 0xFC:
+        case 0xE1:
+        case 0xE4:
+        case 0xF0:
+        case 0xF1:
+        case 0xF2:
+        case 0xF3:
+        case 0xF4:
+        case 0xF5:
+        case 0xF6:
+        case 0xF7:
+        case 0xF8:
+        case 0xF9:
+        case 0xFA:
+        case 0xFB:
+        case 0xFD:
+        case 0xFE:
+        case 0xFF:
+            return 0;
+        break;
+
+        default:
+            switch((nextop>>3)&7) {
+                case 4: /* FRSTOR m94byte */
+                    GET_EW;
+                    #ifndef TEST_INTERPRETER
+                    fpu_loadenv(emu, (char*)ED, 1);
+                    // get the STx
+                    {
+                        char* p =(char*)ED;
+                        p += 14;
+                        for (int i=0; i<8; ++i) {
+                            LD2D(p, &ST(i).d);
+                            p+=10;
+                        }
+                    }
+                    #endif
+                    break;
+                case 6: /* FNSAVE m94byte */
+                    GET_EW;
+                    // ENV first...
+                    #ifndef TEST_INTERPRETER
+                    fpu_savenv(emu, (char*)ED, 1);
+                    // save the STx
+                    {
+                        char* p =(char*)ED;
+                        p += 14;
+                        for (int i=0; i<8; ++i) {
+                            D2LD(&ST(i).d, p);
+                            p+=10;
+                        }
+                    }
+                    #endif
+                    reset_fpu(emu);
+                    break;
+                default:
+                    return 0;
+            }
+    }
+
+    return addr;
+}
@@ -0,0 +1,147 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "my_cpuid.h"
+#include "bridge.h"
+#include "signals.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test67(x86test_t *test, int rep, uintptr_t addr)
+#else
+uintptr_t Run67(x86emu_t *emu, int rep, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u, tmp8u2;
+    int8_t tmp8s;
+    uint16_t tmp16u, tmp16u2;
+    int16_t tmp16s;
+    uint32_t tmp32u, tmp32u2, tmp32u3;
+    int32_t tmp32s, tmp32s2;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    double d;
+    float f;
+    int64_t ll;
+    sse_regs_t *opex, eax1;
+    mmx87_regs_t *opem, eam1;
+    uintptr_t tlsdata;
+
+#ifdef TEST_INTERPRETER
+    x86emu_t *emu = test->emu;
+#endif
+    opcode = F8;
+
+    while(opcode==0x67)
+        opcode = F8;
+
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+
+        case 0x64:
+            tlsdata = GetFSBaseEmu(emu);
+            #ifdef TEST_INTERPRETER
+            return Test6467(test, tlsdata, addr);
+            #else
+            return Run6467(emu, tlsdata, addr);
+            #endif
+
+        case 0x66:                      /* MoooRE opcodes */
+            #ifdef TEST_INTERPRETER
+            return Test6766(test, rep, addr);
+            #else
+            return Run6766(emu, rep, addr);
+            #endif
+
+        case 0x6C:                      /* INSB */
+            tmp32u = rep?R_CX:1;
+            while(tmp32u--) {
+                #ifdef TEST_INTERPRETER
+                *(int8_t*)(R_DI+GetESBaseEmu(emu)) = 0;         // faking port read, using actual segment ES, just in case
+                #endif
+                if(ACCESS_FLAG(F_DF))
+                    R_DI-=1;
+                else
+                    R_DI+=1;
+            }
+            if(rep) R_CX = 0;
+            break;
+
+        case 0xAC:                      /* LODSB */
+            tmp32u = rep?R_CX:1;
+            while(tmp32u--) {
+                R_AL = *(int8_t*)(R_SI+GetDSBaseEmu(emu));
+                if(ACCESS_FLAG(F_DF))
+                    R_SI-=1;
+                else
+                    R_SI+=1;
+            }
+            if(rep) R_CX = 0;
+            break;
+
+        case 0xE0:                      /* LOOPNZ */
+            CHECK_FLAGS(emu);
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX && !ACCESS_FLAG(F_ZF))
+                addr += tmp8s;
+            break;
+        case 0xE1:                      /* LOOPZ */
+            CHECK_FLAGS(emu);
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX && ACCESS_FLAG(F_ZF))
+                addr += tmp8s;
+            break;
+        case 0xE2:                      /* LOOP */
+            tmp8s = F8S;
+            --R_CX; // don't update flags
+            if(R_CX)
+                addr += tmp8s;
+            break;
+        case 0xE3:                      /* JCXZ */
+            tmp8s = F8S;
+            if(!R_CX)
+                addr += tmp8s;
+            break;
+
+        case 0xE8:                      /* CALL Id */
+            tmp32s = F32S; // call is relative
+            Push(emu, addr);
+            addr += tmp32s;
+            break;
+
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,58 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+#include "bridge.h"
+#ifdef DYNAREC
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+
+#ifdef TEST_INTERPRETER
+uintptr_t Test6766(x86test_t *test, int rep, uintptr_t addr)
+#else
+uintptr_t Run6766(x86emu_t *emu, int rep, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int32_t tmp32s;
+    #ifdef TEST_INTERPRETER
+    x86emu_t* emu = test->emu;
+    #endif
+    uint8_t opcode = F8;
+
+    while((opcode==0x2E) || (opcode==0x66))   // ignoring CS: or multiple 0x66
+        opcode = F8;
+
+    while((opcode==0xF2) || (opcode==0xF3)) {
+        rep = opcode-0xF1;
+        opcode = F8;
+    }
+
+    switch(opcode) {
+
+        case 0x8D:                              /* LEA Gw,Ew */
+            nextop = F8;
+            GET_EW16_;
+            GW.word[0] = (uint16_t)(uintptr_t)ED;
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -4,6 +4,7 @@
 #include <stdint.h>
 #include "regs.h"
 #include "x86emu_private.h"
+#include "box86context.h"
 typedef struct x86emu_s x86emu_t;

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

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

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

+#if 0
 // the op code definition can be found here: http://ref.x86asm.net/geek32.html
-
 static inline reg32_t* GetECommon(x86emu_t* emu, uint32_t m)
 {
    if (m<=7) {
@@ -188,23 +199,69 @@ static inline sse_regs_t* GetGx(x86emu_t *emu, uint32_t v)
    uint8_t m = (v&0x38)>>3;
    return &emu->xmm[m&7];
 }
-
+#endif
 void UpdateFlags(x86emu_t *emu);

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

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

 void x86Syscall(x86emu_t *emu);
 void x86Int3(x86emu_t* emu);
@@ -1,7 +1,40 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wstrict-aliasing"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestD8(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunD8(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    float f;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
-    switch(nextop) {
+    switch (nextop) {
        case 0xC0:
        case 0xC1:
        case 0xC2:
@@ -159,7 +192,9 @@
                }
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
+   return addr;
+}
 #pragma GCC diagnostic pop
@@ -1,3 +1,38 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestD9(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunD9(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    float f;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch (nextop) {
        case 0xC0:
@@ -206,7 +241,7 @@
        case 0xE6:
        case 0xE7:
        case 0xEF:
-            goto _default;
+            return 0;
        default:
        switch((nextop>>3)&7) {
            case 0:     /* FLD ST0, Ed float */
@@ -241,7 +276,9 @@
            case 4:     /* FLDENV m */
                // warning, incomplete
                GET_ED;
+                #ifndef TEST_INTERPRETER
                fpu_loadenv(emu, (char*)ED, 0);
+                #endif
                break;
            case 5:     /* FLDCW Ew */
                GET_EW;
@@ -251,7 +288,9 @@
            case 6:     /* FNSTENV m */
                // warning, incomplete
                GET_ED;
+                #ifndef TEST_INTERPRETER
                fpu_savenv(emu, (char*)ED, 0);
+                #endif
                // intruction pointer: 48bits
                // data (operand) pointer: 48bits
                // last opcode: 11bits save: 16bits restaured (1st and 2nd opcode only)
@@ -261,6 +300,8 @@
                EW->word[0] = emu->cw.x16;
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
+   return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDA(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDA(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    float f;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch(nextop) {

@@ -66,8 +100,7 @@
    case 0xF8:
    case 0xF9:
    case 0xFD:
-        UnimpOpcode(emu);
-        break;
+        return 0;
    default:
        switch((nextop>>3)&7) {
            case 0:     /* FIADD ST0, Ed int */
@@ -105,3 +138,5 @@
                break;
        }
    }
+    return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDB(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDB(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    float f;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch(nextop) {
    case 0xC0:      /* FCMOVNB ST(0), ST(i) */
@@ -93,7 +127,7 @@
    case 0xE5:
    case 0xE6:
    case 0xE7:
-        goto _default;
+        return 0;
    default:
        switch((nextop>>3)&7) {
            case 0: /* FILD ST0, Ed */
@@ -103,15 +137,16 @@
                break;
            case 1: /* FISTTP Ed, ST0 */
                GET_ED;
-                tmp32s = ST0.d; // TODO: Handling of FPU Exception
-                if(tmp32s==0x7fffffff && isgreater(ST0.d, (double)(int32_t)0x7fffffff))
+                if(isgreater(ST0.d, (double)0x7fffffff) || isless(ST0.d, -(double)0x80000000U) || !isfinite(ST0.d))
                    tmp32s = 0x80000000;
+                else
+                    tmp32s = ST0.d; // TODO: Handling of FPU Exception
                fpu_do_pop(emu);
                ED->sdword[0] = tmp32s;
                break;
            case 2: /* FIST Ed, ST0 */
                GET_ED;
-                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff) || !isfinite(ST0.d))
+                if(isgreater(ST0.d, (double)0x7fffffff) || isless(ST0.d, -(double)0x80000000U) || !isfinite(ST0.d))
                    ED->sdword[0] = 0x80000000;
                else {
                    volatile int32_t tmp = fpu_round(emu, ST0.d);    // tmp to avoid BUS ERROR
@@ -120,7 +155,7 @@
                break;
            case 3: /* FISTP Ed, ST0 */
                GET_ED;
-                if(isgreater(ST0.d, (double)(int32_t)0x7fffffff) || isless(ST0.d, -(double)(int32_t)0x7fffffff) || !isfinite(ST0.d))
+                if(isgreater(ST0.d, (double)0x7fffffff) || isless(ST0.d, -(double)0x80000000U) || !isfinite(ST0.d))
                    ED->sdword[0] = 0x80000000;
                else {
                    volatile int32_t tmp = fpu_round(emu, ST0.d);    // tmp to avoid BUS ERROR
@@ -129,14 +164,14 @@
                fpu_do_pop(emu);
                break;
            case 5: /* FLD ST0, Et */
-                GET_ED;
+                GET_EDT;
                fpu_do_push(emu);
                memcpy(&STld(0).ld, ED, 10);
                LD2D(&STld(0).ld, &ST(0).d);
                STld(0).uref = ST0.q;
                break;
            case 7: /* FSTP tbyte */
-                GET_ED;
+                GET_EDT;
                if(ST0.q!=STld(0).uref)
                    D2LD(&ST0.d, ED);
                else
@@ -144,6 +179,8 @@
                fpu_do_pop(emu);
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
+    return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDC(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDC(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    double d;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch(nextop) {
        case 0xC0:
@@ -82,7 +116,7 @@
            ST(nextop&7).d /=  ST0.d;
            break;
        default:
-            GET_ED;
+            GET_ED8;
            switch((nextop>>3)&7) {
            case 0:         /* FADD ST0, double */
                if(!(((uintptr_t)ED)&7))
@@ -150,7 +184,8 @@
                }
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
-#pragma GCC diagnostic pop
+    return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDD(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDD(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    double d;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch(nextop) {
    
@@ -79,12 +113,12 @@
    case 0xFD:
    case 0xFE:
    case 0xFF:
-        goto _default;
+        return 0;

    default:
        switch((nextop>>3)&7) {
            case 0: /* FLD double */
-                GET_ED;
+                GET_ED8;
                fpu_do_push(emu);
                if(!(((uintptr_t)ED)&7))
                    ST0.d = *(double*)ED;
@@ -93,17 +127,32 @@
                }
                break;
            case 1: /* FISTTP ED qword */
-                GET_ED;
-                if(!(((uintptr_t)ED)&7))
-                    *(int64_t*)ED = ST0.d;
-                else {
-                    int64_t i64 = ST0.d;
+                GET_ED8;
+                if(!(((uintptr_t)ED)&7)) {
+                    if(STll(0).sref==ST(0).sq)
+                        *(int64_t*)ED = STll(0).sq;
+                    else {
+                        if(isgreater(ST0.d, (double)0x7fffffffffffffffLL) || isless(ST0.d, -(double)0x8000000000000000LL) || !isfinite(ST0.d))
+                            *(uint64_t*)ED = 0x8000000000000000LL;
+                        else
+                            *(int64_t*)ED = ST0.d;
+                    }
+                } else {
+                    int64_t i64;
+                    if(STll(0).sref==ST(0).sq)
+                        i64 = STll(0).sq;
+                    else {
+                        if(isgreater(ST0.d, (double)0x7fffffffffffffffLL) || isless(ST0.d, -(double)0x8000000000000000LL) || !isfinite(ST0.d))
+                            i64 = 0x8000000000000000LL;
+                        else
+                            i64 = ST0.d;
+                    }
                    memcpy(ED, &i64, sizeof(int64_t));
                }
                fpu_do_pop(emu);
                break;
            case 2: /* FST double */
-                GET_ED;
+                GET_ED8;
                if(!(((uintptr_t)ED)&7))
                    *(double*)ED = ST0.d;
                else {
@@ -111,7 +160,7 @@
                }
                break;
            case 3: /* FSTP double */
-                GET_ED;
+                GET_ED8;
                if(!(((uintptr_t)ED)&7))
                    *(double*)ED = ST0.d;
                else {
@@ -120,8 +169,10 @@
                fpu_do_pop(emu);
                break;
            case 4: /* FRSTOR m108byte */
-                GET_ED;
+                GET_ED_;
+                #ifndef TEST_INTERPRETER
                fpu_loadenv(emu, (char*)ED, 0);
+                #endif
                // get the STx
                {
                    char* p =(char*)ED;
@@ -133,9 +184,10 @@
                }
                break;
            case 6: /* FNSAVE m108byte */
-                GET_ED;
+                GET_ED_;
                // ENV first...
                // warning, incomplete
+                #ifndef TEST_INTERPRETER
                fpu_savenv(emu, (char*)ED, 0);
                // save the STx
                {
@@ -146,6 +198,7 @@
                        p+=10;
                    }
                }
+                #endif
                reset_fpu(emu);
                break;
            case 7: /* FNSTSW m2byte */
@@ -154,6 +207,8 @@
                *(uint16_t*)ED = emu->sw.x16;
                break;
            default:
-                goto _default;
+                return 0;
        }
    }
+    return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDE(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDE(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int32_t tmp32s;
+    int64_t ll;
+    double d;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch (nextop) {
    case 0xC0:  /* FADDP STx, ST0 */
@@ -95,7 +129,7 @@
    case 0xDD:
    case 0xDE:
    case 0xDF:
-        goto _default;
+        return 0;
    
    default:
        switch((nextop>>3)&7) {
@@ -124,6 +158,8 @@
                ST0.d = (double)EW->sword[0] / ST0.d;
                break;
        default:
-            goto _default;
+            return 0;
        }
    }
+    return addr;
+}
@@ -1,3 +1,37 @@
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x87emu_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "box86context.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestDF(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunDF(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t nextop;
+    int16_t tmp16s;
+    int64_t tmp64s;
+    double d;
+    reg32_t *oped;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    nextop = F8;
    switch (nextop) {
    case 0xC0:  /* FFREEP STx */
@@ -25,8 +59,11 @@
        break;

    case 0xE0:  /* FNSTSW AX */
-        emu->sw.f.F87_TOP = emu->top&7;
-        R_AX = emu->sw.x16;
+        {
+            x87flags_t sw = emu->sw;
+            sw.f.F87_TOP = emu->top&7;
+            R_AX = sw.x16;
+        }
        break;

    case 0xE8:  /* FUCOMIP ST0, STx */
@@ -84,7 +121,7 @@
    case 0xFD:
    case 0xFE:
    case 0xFF:
-        goto _default;
+        return 0;

    default:
        switch((nextop>>3)&7) {
@@ -96,32 +133,35 @@
            break;
        case 1: /* FISTTP Ew, ST0 */
            GET_EW;
-            tmp16s = ST0.d;
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x8000) || !isfinite(ST0.d))
+                tmp16s = 0x8000;
+            else
+                tmp16s = ST0.d;
            EW->sword[0] = tmp16s;
            fpu_do_pop(emu);
            break;
        case 2: /* FIST Ew, ST0 */
            GET_EW;
-            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff) || !isfinite(ST0.d))
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x8000) || !isfinite(ST0.d))
                EW->sword[0] = 0x8000;
            else
                EW->sword[0] = fpu_round(emu, ST0.d);
            break;
        case 3: /* FISTP Ew, ST0 */
            GET_EW;
-            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x7fff) || !isfinite(ST0.d))
+            if(isgreater(ST0.d, (double)(int32_t)0x7fff) || isless(ST0.d, -(double)(int32_t)0x8000) || !isfinite(ST0.d))
                EW->sword[0] = 0x8000;
            else
                EW->sword[0] = fpu_round(emu, ST0.d);
            fpu_do_pop(emu);
            break;
        case 4: /* FBLD ST0, tbytes */
-            GET_ED;
+            GET_EDT;
            fpu_do_push(emu);
            fpu_fbld(emu, (uint8_t*)ED);
            break;
        case 5: /* FILD ST0, Gq */
-            GET_ED;
+            GET_ED8;
            tmp64s = *(int64_t*)ED;
            fpu_do_push(emu);
            ST0.d = tmp64s;
@@ -129,19 +169,19 @@
            STll(0).sref = ST0.sq;
            break;
        case 6: /* FBSTP tbytes, ST0 */
-            GET_ED;
+            GET_EDT;
            fpu_fbst(emu, (uint8_t*)ED);
            fpu_do_pop(emu);
            break;
        case 7: /* FISTP i64 */
-            GET_ED;
+            GET_ED8;
            if((uintptr_t)ED & 0x7) {
                // un-aligned!
                if(STll(0).sref==ST(0).sq)
                    memcpy(ED, &STll(0).sq, sizeof(int64_t));
                else {
                    int64_t i64;
-                    if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(int64_t)0x7fffffffffffffffLL) || !isfinite(ST0.d))
+                    if(isgreater(ST0.d, (double)0x7fffffffffffffffLL) || isless(ST0.d, -(double)0x8000000000000000LL) || !isfinite(ST0.d))
                        i64 = 0x8000000000000000LL;
                    else
                        i64 = fpu_round(emu, ST0.d);
@@ -151,7 +191,7 @@
                if(STll(0).sref==ST(0).sq)
                    *(int64_t*)ED = STll(0).sq;
                else {
-                    if(isgreater(ST0.d, (double)(int64_t)0x7fffffffffffffffLL) || isless(ST0.d, -(double)(int64_t)0x7fffffffffffffffLL) || !isfinite(ST0.d))
+                    if(isgreater(ST0.d, (double)0x7fffffffffffffffLL) || isless(ST0.d, -(double)0x8000000000000000LL) || !isfinite(ST0.d))
                        *(int64_t*)ED = 0x8000000000000000LL;
                    else
                        *(int64_t*)ED = fpu_round(emu, ST0.d);
@@ -160,6 +200,8 @@
            fpu_do_pop(emu);
            break;
        default:
-            goto _default;
+            return 0;
        }
    }
+    return addr;
+}
@@ -0,0 +1,625 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "my_cpuid.h"
+#include "bridge.h"
+#include "signals.h"
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestF0(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunF0(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u, tmp8u2;
+    int8_t tmp8s;
+    uint16_t tmp16u, tmp16u2;
+    int16_t tmp16s;
+    uint32_t tmp32u, tmp32u2, tmp32u3;
+    int32_t tmp32s, tmp32s2;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    double d;
+    float f;
+    int64_t ll;
+    sse_regs_t *opex, eax1;
+    mmx87_regs_t *opem, eam1;
+
+#ifdef TEST_INTERPRETER
+    x86emu_t *emu = test->emu;
+#endif
+    opcode = F8;
+
+    switch(opcode) {
+
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+        #define GO(B, OP)                      \
+        case B+0: \
+            nextop = F8;               \
+            GET_EB;             \
+            do {                \
+            tmp8u = arm_lock_read_b(EB);     \
+            tmp8u = OP##8(emu, tmp8u, GB);  \
+            } while (arm_lock_write_b(EB, tmp8u));   \
+            break;                              \
+        case B+1: \
+            nextop = F8;               \
+            GET_ED;             \
+            do {                \
+            tmp32u = arm_lock_read_d(ED);     \
+            tmp32u = OP##32(emu, tmp32u, GD.dword[0]);  \
+            } while (arm_lock_write_d(ED, tmp32u));   \
+            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;
+#else
+        #define GO(B, OP)                      \
+        case B+0: \
+            nextop = F8;               \
+            GET_EB;             \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            EB->byte[0] = OP##8(emu, EB->byte[0], GB);  \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;                              \
+        case B+1: \
+            nextop = F8;               \
+            GET_ED;             \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            ED->dword[0] = OP##32(emu, ED->dword[0], GD.dword[0]); \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;                              \
+        case B+2: \
+            nextop = F8;               \
+            GET_EB;                   \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            GB = OP##8(emu, GB, EB->byte[0]); \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;                              \
+        case B+3: \
+            nextop = F8;               \
+            GET_ED;         \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            GD.dword[0] = OP##32(emu, GD.dword[0], ED->dword[0]); \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;                              \
+        case B+4: \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            R_AL = OP##8(emu, R_AL, F8); \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;                              \
+        case B+5: \
+            pthread_mutex_lock(&emu->context->mutex_lock);\
+            R_EAX = OP##32(emu, R_EAX, F32); \
+            pthread_mutex_unlock(&emu->context->mutex_lock);\
+            break;
+#endif
+        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;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    do {
+                        tmp8u = arm_lock_read_b(EB);
+                        cmp8(emu, R_AL, tmp8u);
+                        if(ACCESS_FLAG(F_ZF)) {
+                            tmp32s = arm_lock_write_b(EB, GB);
+                        } else {
+                            R_AL = tmp8u;
+                            tmp32s = 0;
+                        }
+                    } while(tmp32s);
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    cmp8(emu, R_AL, EB->byte[0]);
+                    if(ACCESS_FLAG(F_ZF)) {
+                        EB->byte[0] = GB;
+                    } else {
+                        R_AL = EB->byte[0];
+                    }
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 0xB1:                      /* CMPXCHG Ed,Gd */
+                    nextop = F8;
+                    GET_ED;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    if(((uintptr_t)ED)&3) {
+                        do {
+                            tmp32u = ED->dword[0] & ~0xff;
+                            tmp32u |= arm_lock_read_b(ED);
+                            cmp32(emu, R_EAX, tmp32u);
+                            if(ACCESS_FLAG(F_ZF)) {
+                                tmp32s = arm_lock_write_b(ED, GD.dword[0] & 0xff);
+                                if(!tmp32s)
+                                    ED->dword[0] = GD.dword[0];
+                            } else {
+                                R_EAX = tmp32u;
+                                tmp32s = 0;
+                            }
+                        } while(tmp32s);
+                    } else {
+                        do {
+                            tmp32u = arm_lock_read_d(ED);
+                            cmp32(emu, R_EAX, tmp32u);
+                            if(ACCESS_FLAG(F_ZF)) {
+                                tmp32s = arm_lock_write_d(ED, GD.dword[0]);
+                            } else {
+                                R_EAX = tmp32u;
+                                tmp32s = 0;
+                            }
+                        } while(tmp32s);
+                    }
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    cmp32(emu, R_EAX, ED->dword[0]);
+                    if(ACCESS_FLAG(F_ZF)) {
+                        ED->dword[0] = GD.dword[0];
+                    } else {
+                        R_EAX = ED->dword[0];
+                    }
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 0xB3:                      /* BTR Ed,Gd */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_ED;
+                    tmp8u = GD.byte[0];
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        #ifdef TEST_INTERPRETER
+                        test->memaddr=((test->memaddr)+(tmp8u>>5));
+                        *(uint32_t*)test->mem = *(uint32_t*)test->memaddr;
+                        #else
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                        #endif
+                    }
+                    tmp8u&=31;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    do {
+                        tmp32u = arm_lock_read_d(ED);
+                        if(tmp32u & (1<<tmp8u)) {
+                            SET_FLAG(F_CF);
+                            tmp32u ^= (1<<tmp8u);
+                            tmp32s = arm_lock_write_d(ED, tmp32u);
+                        } else {
+                            CLEAR_FLAG(F_CF);
+                            tmp32s = 0;
+                        }
+                    } while(tmp32s);
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    if(ED->dword[0] & (1<<tmp8u)) {
+                        SET_FLAG(F_CF);
+                        ED->dword[0] ^= (1<<tmp8u);
+                    } else
+                        CLEAR_FLAG(F_CF);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    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)
+                            {
+                                #ifdef TEST_INTERPRETER
+                                test->memaddr=((test->memaddr)+(tmp8u>>5));
+                                *(uint32_t*)test->mem = *(uint32_t*)test->memaddr;
+                                #else
+                                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                                #endif
+                            }
+                            tmp8u&=31;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                            if(arm_lock_read_d(ED) & (1<<tmp8u))
+                                SET_FLAG(F_CF);
+                            else
+                                CLEAR_FLAG(F_CF);
+#else
+                            pthread_mutex_lock(&emu->context->mutex_lock);
+                            if(ED->dword[0] & (1<<tmp8u))
+                                SET_FLAG(F_CF);
+                            else
+                                CLEAR_FLAG(F_CF);
+                            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                            break;
+                        case 5:             /* BTS Ed, Ib */
+                            CHECK_FLAGS(emu);
+                            GET_ED;
+                            tmp8u = F8;
+                            if((nextop&0xC0)!=0xC0)
+                            {
+                                #ifdef TEST_INTERPRETER
+                                test->memaddr=((test->memaddr)+(tmp8u>>5));
+                                *(uint32_t*)test->mem = *(uint32_t*)test->memaddr;
+                                #else
+                                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                                #endif
+                            }
+                            tmp8u&=31;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                            do {
+                                tmp32u = arm_lock_read_d(ED);
+                                if(tmp32u & (1<<tmp8u)) {
+                                    SET_FLAG(F_CF);
+                                    tmp32s = 0;
+                                } else {
+                                    CLEAR_FLAG(F_CF);
+                                    tmp32u ^= (1<<tmp8u);
+                                    tmp32s = arm_lock_write_d(ED, tmp32u);
+                                }
+                            } while(tmp32s);
+#else
+                            pthread_mutex_lock(&emu->context->mutex_lock);
+                            if(ED->dword[0] & (1<<tmp8u)) {
+                                SET_FLAG(F_CF);
+                            } else {
+                                CLEAR_FLAG(F_CF);
+                                ED->dword[0] ^= (1<<tmp8u);
+                            }
+                            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                            break;
+                        case 6:             /* BTR Ed, Ib */
+                            CHECK_FLAGS(emu);
+                            GET_ED;
+                            tmp8u = F8;
+                            if((nextop&0xC0)!=0xC0)
+                            {
+                                #ifdef TEST_INTERPRETER
+                                test->memaddr=((test->memaddr)+(tmp8u>>5));
+                                *(uint32_t*)test->mem = *(uint32_t*)test->memaddr;
+                                #else
+                                ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                                #endif
+                            }
+                            tmp8u&=31;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                            do {
+                                tmp32u = arm_lock_read_d(ED);
+                                if(tmp32u & (1<<tmp8u)) {
+                                    SET_FLAG(F_CF);
+                                    tmp32u ^= (1<<tmp8u);
+                                    tmp32s = arm_lock_write_d(ED, tmp32u);
+                                } else {
+                                    CLEAR_FLAG(F_CF);
+                                    tmp32s = 0;
+                                }
+                            } while(tmp32s);
+#else
+                            pthread_mutex_lock(&emu->context->mutex_lock);
+                            if(ED->dword[0] & (1<<tmp8u)) {
+                                SET_FLAG(F_CF);
+                                ED->dword[0] ^= (1<<tmp8u);
+                            } else
+                                CLEAR_FLAG(F_CF);
+                            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                            break;
+
+                        default:
+                            return addr-3; //unfetchall 0F BA nextop but not F0, continue normal without LOCK
+                    }
+                    break;
+                case 0xBB:                      /* BTC Ed,Gd */
+                    CHECK_FLAGS(emu);
+                    nextop = F8;
+                    GET_ED;
+                    tmp8u = GD.byte[0];
+                    if((nextop&0xC0)!=0xC0)
+                    {
+                        #ifdef TEST_INTERPRETER
+                        test->memaddr=((test->memaddr)+(tmp8u>>5));
+                        *(uint32_t*)test->mem = *(uint32_t*)test->memaddr;
+                        #else
+                        ED=(reg32_t*)(((uint32_t*)(ED))+(tmp8u>>5));
+                        #endif
+                    }
+                    tmp8u&=31;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    do {
+                        tmp32u = arm_lock_read_d(ED);
+                        if(tmp32u & (1<<tmp8u)) {
+                            SET_FLAG(F_CF);
+                        } else {
+                            CLEAR_FLAG(F_CF);
+                        }
+                        tmp32u ^= (1<<tmp8u);
+                    } while(arm_lock_write_d(ED, tmp32u));
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    if(ED->dword[0] & (1<<tmp8u))
+                        SET_FLAG(F_CF);
+                    else
+                        CLEAR_FLAG(F_CF);
+                    ED->dword[0] ^= (1<<tmp8u);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 0xC0:                      /* XADD Gb,Eb */
+                    nextop = F8;
+                    GET_EB;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    do {
+                        tmp8u = arm_lock_read_b(EB);
+                        tmp8u2 = add8(emu, tmp8u, GB);
+                    } while (arm_lock_write_b(EB, tmp8u2));
+                    GB = tmp8u;
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    tmp8u = add8(emu, EB->byte[0], GB);
+                    GB = EB->byte[0];
+                    EB->byte[0] = tmp8u;
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 0xC1:                      /* XADD Gd,Ed */
+                    nextop = F8;
+                    GET_ED;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    if(((uintptr_t)ED)&3) {
+                        do {
+                            tmp32u = ED->dword[0] & ~0xff;
+                            tmp32u |= arm_lock_read_b(ED);
+                            tmp32u2 = add32(emu, tmp32u, GD.dword[0]);
+                        } while(arm_lock_write_b(ED, tmp32u2&0xff));
+                        ED->dword[0] = tmp32u2;
+                    } else {
+                        do {
+                            tmp32u = arm_lock_read_d(ED);
+                            tmp32u2 = add32(emu, tmp32u, GD.dword[0]);
+                        } while(arm_lock_write_d(ED, tmp32u2));
+                    }
+                    GD.dword[0] = tmp32u;
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    tmp32u = add32(emu, ED->dword[0], GD.dword[0]);
+                    GD.dword[0] = ED->dword[0];
+                    ED->dword[0] = tmp32u;
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 0xC7:                      /* CMPXCHG8B Gq */
+                    nextop = F8;
+                    GET_ED8;
+                    switch((nextop>>3)&7) {
+                        case 1:
+                            CHECK_FLAGS(emu);
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                            if(((uintptr_t)ED)&7) {
+                                // unaligned!!!
+                                uint64_t ref = R_EDX;
+                                ref = ref<<32 | R_EAX;
+                                void* p = (void*)(((uintptr_t)ED + 3)&~3);
+                                int d = (((uintptr_t)p - (uintptr_t)ED))*8;
+                                do {
+                                    uint64_t m = *(uint64_t*)ED;
+                                    m&=~((0xffffffffLL)<<d);
+                                    tmp32u = arm_lock_read_d(p);
+                                    m|=((uint64_t)tmp32u)<<d;
+                                    if(ref == m) {
+                                        SET_FLAG(F_ZF);
+                                        m = ((uint64_t)R_ECX)<<32 | R_EBX;
+                                        tmp32u2 = m>>d;
+                                        tmp32s = arm_lock_write_d(p, tmp32u2);
+                                        if(!tmp32s)
+                                            *(uint64_t*)ED = m;
+                                    } else {
+                                        CLEAR_FLAG(F_ZF);
+                                        R_EAX = m&0xffffffff;
+                                        R_EDX = m>>32;
+                                        tmp32s = 0;
+                                    }
+                                } while(tmp32s);
+                            } else
+                            do {
+                                arm_lock_read_dd(&tmp32u, &tmp32u2, ED);
+                                if(R_EAX == tmp32u && R_EDX == tmp32u2) {
+                                    SET_FLAG(F_ZF);
+                                    tmp32s = arm_lock_write_dd(R_EBX, R_ECX, ED);
+                                } else {
+                                    CLEAR_FLAG(F_ZF);
+                                    R_EAX = tmp32u;
+                                    R_EDX = tmp32u2;
+                                    tmp32s = 0;
+                                }
+                            } while(tmp32s);
+#else
+                            pthread_mutex_lock(&emu->context->mutex_lock);
+                            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;
+                            }
+                            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                            break;
+                        default:
+                            printf_log(LOG_NONE, "Illegal Opcode 0xF0 0xC7 0x%02X 0x%02X\n", nextop, PK(0));
+                                emu->quit=1;
+                                emu->error |= ERR_ILLEGAL;
+                            return 0;
+                    }
+                    break;
+                default:
+                    // trigger invalid lock?
+                    return addr-2; // unfetch 0F and nextop but not F0
+            }
+            break;
+
+        case 0x66:
+            #ifdef TEST_INTERPRETER
+            return TestF066(test, addr);
+            #else
+            return RunF066(emu, addr);
+            #endif
+
+        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;
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+            switch((nextop>>3)&7) {
+                case 0: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, add32(emu, tmp32u2, tmp32u))); break;
+                case 1: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED,  or32(emu, tmp32u2, tmp32u))); break;
+                case 2: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, adc32(emu, tmp32u2, tmp32u))); break;
+                case 3: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, sbb32(emu, tmp32u2, tmp32u))); break;
+                case 4: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, and32(emu, tmp32u2, tmp32u))); break;
+                case 5: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, sub32(emu, tmp32u2, tmp32u))); break;
+                case 6: do { tmp32u2 = arm_lock_read_d(ED);} while(arm_lock_write_d(ED, xor32(emu, tmp32u2, tmp32u))); break;
+                case 7:                cmp32(emu, ED->dword[0], tmp32u); break;
+            }
+#else
+            pthread_mutex_lock(&emu->context->mutex_lock);
+            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;
+            }
+            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+            break;
+        case 0x86:                      /* XCHG Eb,Gb */
+        case 0x87:                      /* XCHG Ed,Gd */
+            return addr-1;   // let the normal XCHG execute, it have integrated LOCK
+        case 0x90:  // This is xchg reg, reg, so let's ignore the LOCK prefix...
+        case 0x91:
+        case 0x92:
+        case 0x93:
+        case 0x94:
+        case 0x95:
+        case 0x96:
+        case 0x97:
+            return addr-1;
+
+        case 0xFF:              /* GRP 5 Ed */
+            nextop = F8;
+            GET_ED;
+            switch((nextop>>3)&7) {
+                case 0:                 /* INC Ed */
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    if((uintptr_t)ED&3) { 
+                        //meh.
+                        do {
+                            tmp32u = ED->dword[0];
+                            tmp32u &=~0xff;
+                            tmp32u |= arm_lock_read_b(ED);
+                            tmp32u = inc32(emu, tmp32u);
+                        } while(arm_lock_write_b(ED, tmp32u&0xff));
+                        ED->dword[0] = tmp32u;
+                    } else
+                        do {
+                            tmp32u = arm_lock_read_d(ED);
+                        } while(arm_lock_write_d(ED, inc32(emu, tmp32u)));
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    ED->dword[0] = inc32(emu, ED->dword[0]);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 1:                 /* DEC Ed */
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+                    if((uintptr_t)ED&3) { 
+                        //meh.
+                        do {
+                            tmp32u = ED->dword[0];
+                            tmp32u &=~0xff;
+                            tmp32u |= arm_lock_read_b(ED);
+                            tmp32u = dec32(emu, tmp32u);
+                        } while(arm_lock_write_b(ED, tmp32u&0xff));
+                        ED->dword[0] = tmp32u;
+                    } else
+                        do {
+                            tmp32u = arm_lock_read_d(ED);
+                        } while(arm_lock_write_d(ED, dec32(emu, tmp32u)));
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    ED->dword[0] = dec32(emu, ED->dword[0]);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                default:
+                    printf_log(LOG_NONE, "Illegal Opcode 0xF0 0xFF 0x%02X 0x%02X\n", nextop, PK(0));
+                    emu->quit=1;
+                    emu->error |= ERR_ILLEGAL;
+                    return 0;
+            }
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -0,0 +1,184 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "my_cpuid.h"
+#include "bridge.h"
+#include "signals.h"
+#if defined(DYNAREC) && !defined(TEST_INTERPRETER)
+#include "../dynarec/arm_lock_helper.h"
+#endif
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestF066(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunF066(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    reg32_t *oped;
+    uint8_t tmp8u, tmp8u2;
+    int8_t tmp8s;
+    uint16_t tmp16u, tmp16u2;
+    int16_t tmp16s;
+    uint32_t tmp32u, tmp32u2, tmp32u3;
+    int32_t tmp32s, tmp32s2;
+    uint64_t tmp64u;
+    int64_t tmp64s;
+    double d;
+    float f;
+    int64_t ll;
+    sse_regs_t *opex, eax1;
+    mmx87_regs_t *opem, eam1;
+
+#ifdef TEST_INTERPRETER
+    x86emu_t *emu = test->emu;
+#endif
+    opcode = F8;
+
+    switch(opcode) {
+
+        case 0x0f:
+            opcode = F8;
+            switch (opcode) { 
+                case 0xB1:                      /* CMPXCHG Ew,Gw */
+                    nextop = F8;
+                    GET_EW;
+#ifdef DYNAREC
+                    do {
+                        tmp16u = arm_lock_read_h(ED);
+                        cmp16(emu, R_AX, tmp16u);
+                        if(ACCESS_FLAG(F_ZF)) {
+                            tmp32s = arm_lock_write_h(ED, GW.word[0]);
+                        } else {
+                            R_AX = tmp16u;
+                            tmp32s = 0;
+                        }
+                    } while(tmp32s);
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    GET_EW;
+                    cmp16(emu, R_AX, EW->word[0]);
+                    if(ACCESS_FLAG(F_ZF)) {
+                        EW->word[0] = GW.word[0];
+                    } else {
+                        R_AX = EW->word[0];
+                    }
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+            default:
+                return 0;
+            }
+            break;
+        case 0x81:              /* GRP Ed,Iw */
+        case 0x83:              /* GRP Ed,Ib */
+            nextop = F8;
+            GET_EW;
+            if(opcode==0x83) {
+                tmp16s = F8S;
+                tmp16u = (uint32_t)tmp16s;
+            } else
+                tmp16u = F16;
+#ifdef DYNAREC
+            switch((nextop>>3)&7) {
+                case 0: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, add16(emu, tmp16u2, tmp16u))); break;
+                case 1: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW,  or16(emu, tmp16u2, tmp16u))); break;
+                case 2: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, adc16(emu, tmp16u2, tmp16u))); break;
+                case 3: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, sbb16(emu, tmp16u2, tmp16u))); break;
+                case 4: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, and16(emu, tmp16u2, tmp16u))); break;
+                case 5: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, sub16(emu, tmp16u2, tmp16u))); break;
+                case 6: do { tmp16u2 = arm_lock_read_h(EW);} while(arm_lock_write_h(EW, xor16(emu, tmp16u2, tmp16u))); break;
+                case 7:                cmp16(emu, ED->dword[0], tmp16u); break;
+            }
+#else
+            pthread_mutex_lock(&emu->context->mutex_lock);
+            switch((nextop>>3)&7) {
+                case 0: EW->word[0] = add16(emu, EW->word[0], tmp16u); break;
+                case 1: EW->word[0] =  or16(emu, EW->word[0], tmp16u); break;
+                case 2: EW->word[0] = adc16(emu, EW->word[0], tmp16u); break;
+                case 3: EW->word[0] = sbb16(emu, EW->word[0], tmp16u); break;
+                case 4: EW->word[0] = and16(emu, EW->word[0], tmp16u); break;
+                case 5: EW->word[0] = sub16(emu, EW->word[0], tmp16u); break;
+                case 6: EW->word[0] = xor16(emu, EW->word[0], tmp16u); break;
+                case 7:               cmp16(emu, EW->word[0], tmp16u); break;
+            }
+            pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+            break;
+        
+        case 0xFF:
+            nextop = F8;
+            GET_EW;
+            switch((nextop>>3)&7) {
+                case 0:                 /* INC Ed */
+#ifdef DYNAREC
+                    if((uintptr_t)ED&1) { 
+                        //meh.
+                        do {
+                            tmp16u = ED->dword[0];
+                            tmp16u &=~0xff;
+                            tmp16u |= arm_lock_read_b(ED);
+                            tmp16u = inc16(emu, tmp16u);
+                        } while(arm_lock_write_b(ED, tmp16u&0xff));
+                        ED->dword[0] = tmp16u;
+                    } else
+                        do {
+                            tmp16u = arm_lock_read_h(ED);
+                        } while(arm_lock_write_h(ED, inc16(emu, tmp16u)));
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    ED->dword[0] = inc32(emu, ED->dword[0]);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                case 1:                 /* DEC Ed */
+#ifdef DYNAREC
+                    if((uintptr_t)ED&1) { 
+                        //meh.
+                        do {
+                            tmp16u = ED->dword[0];
+                            tmp16u &=~0xff;
+                            tmp16u |= arm_lock_read_b(ED);
+                            tmp16u = dec16(emu, tmp16u);
+                        } while(arm_lock_write_b(ED, tmp16u&0xff));
+                        ED->dword[0] = tmp16u;
+                    } else
+                        do {
+                            tmp16u = arm_lock_read_h(ED);
+                        } while(arm_lock_write_h(ED, dec16(emu, tmp16u)));
+#else
+                    pthread_mutex_lock(&emu->context->mutex_lock);
+                    ED->dword[0] = dec32(emu, ED->dword[0]);
+                    pthread_mutex_unlock(&emu->context->mutex_lock);
+#endif
+                    break;
+                default:
+                    return 0;
+            }
+            break;
+
+        default:
+            return 0;
+    }
+    return addr;
+}
@@ -1,4 +1,51 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "bridge.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestF20F(x86test_t *test, uintptr_t addr, int *step)
+#else
+uintptr_t RunF20F(x86emu_t *emu, uintptr_t addr, int *step)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    int8_t tmp8s;
+    uint8_t tmp8u;
+    int32_t tmp32s;
+    int64_t tmp64s0, tmp64s1;
+    reg32_t *oped, *opgd;
+    sse_regs_t *opex, *opgx, eax1;
+    mmx87_regs_t *opgm;
+    #ifndef NOALIGN
+    int is_nan;
+    #endif
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    opcode = F8;
+
    switch(opcode) {

    case 0x10:  /* MOVSD Gx, Ex */
@@ -30,7 +77,7 @@
    case 0x2C:  /* CVTTSD2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(!isfinite(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GD.dword[0] = 0x80000000;
        else
            GD.sdword[0] = EX->d[0];
@@ -38,7 +85,7 @@
    case 0x2D:  /* CVTSD2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(EX->d[0]) || isinf(EX->d[0]) || EX->d[0]>0x7fffffff)
+        if(!isfinite(EX->d[0]) || EX->d[0]>(double)0x7fffffff || EX->d[0]<-(double)0x80000000U)
            GD.dword[0] = 0x80000000;
        else
            switch(emu->mxcsr.f.MXCSR_RC) {
@@ -255,7 +302,7 @@
        
    GOCOND(0x80
        , tmp32s = F32S; CHECK_FLAGS(emu);
-        , ip += tmp32s;
+        , addr += tmp32s; STEP3;
    )                               /* 0x80 -> 0x8F Jxx */
        
    case 0xC2:  /* CMPSD Gx, Ex, Ib */
@@ -295,7 +342,7 @@
        nextop = F8;
        GET_EX;
        for(int i=0; i<2; ++i)
-            if(isnan(EX->d[i]) || isinf(EX->d[i]) || EX->d[i]>0x7fffffff)
+            if(!isfinite(EX->d[i]) || EX->d[i]>(double)0x7fffffff || EX->d[i]<-(double)0x80000000U)
                GX.ud[i] = 0x80000000;
            else
                switch(emu->mxcsr.f.MXCSR_RC) {
@@ -322,6 +369,7 @@
        break;

    default:
-        goto _default;
+        return 0;
    }
-
+    return addr;
+}
@@ -1,4 +1,49 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "bridge.h"
+
+#include "modrm.h"
+
+#ifdef TEST_INTERPRETER
+uintptr_t TestF30F(x86test_t *test, uintptr_t addr)
+#else
+uintptr_t RunF30F(x86emu_t *emu, uintptr_t addr)
+#endif
+{
+    uint8_t opcode;
+    uint8_t nextop;
+    int8_t tmp8s;
+    uint8_t tmp8u;
+    uint32_t tmp32u;
+    int64_t tmp64s;
+    uint64_t tmp64u;
+    reg32_t *oped, *opgd;
+    sse_regs_t *opex, *opgx, eax1;
+    mmx87_regs_t *opem;
+    #ifdef TEST_INTERPRETER
+    x86emu_t*emu = test->emu;
+    #endif
+
    opcode = F8;
+
    switch(opcode) {

    case 0x10:  /* MOVSS Gx Ex */
@@ -47,7 +92,7 @@
    case 0x2C:  /* CVTTSS2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>0x7fffffff)
+        if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>=(float)0x80000000U || EX->f[0]<-(float)0x80000000U)
            GD.dword[0] = 0x80000000;
        else
            GD.sdword[0] = EX->f[0];
@@ -55,7 +100,7 @@
    case 0x2D:  /* CVTSS2SI Gd, Ex */
        nextop = F8;
        GET_EX;
-        if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>0x7fffffff)
+        if(isnanf(EX->f[0]) || isinff(EX->f[0]) || EX->f[0]>=(float)0x80000000U || EX->f[0]<-(float)0x80000000U)
            GD.dword[0] = 0x80000000;
        else
            switch(emu->mxcsr.f.MXCSR_RC) {
@@ -109,7 +154,7 @@
        nextop = F8;
        GET_EX;
        for(int i=0; i<4; ++i)
-            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>0x7fffffff)
+            if(isnanf(EX->f[i]) || isinff(EX->f[i]) || EX->f[i]>=(float)0x80000000U || EX->f[i]<-(float)0x80000000U)
                GX.ud[i] = 0x80000000;
            else
                GX.sd[i] = EX->f[i];
@@ -123,7 +168,7 @@
    case 0x5D:  /* MINSS Gx, Ex */
        nextop = F8;
        GET_EX;
-        if(isnan(GX.f[0]) || isnan(EX->f[0]) || islessequal(EX->f[0], GX.f[0]))
+        if(isnanf(GX.f[0]) || isnanf(EX->f[0]) || islessequal(EX->f[0], GX.f[0]))
            GX.f[0] = EX->f[0];
        break;
    case 0x5E:  /* DIVSS Gx, Ex */
@@ -134,7 +179,7 @@
    case 0x5F:  /* MAXSS Gx, Ex */
        nextop = F8;
        GET_EX;
-        if (isnan(GX.f[0]) || isnan(EX->f[0]) || isgreaterequal(EX->f[0], GX.f[0]))
+        if (isnanf(GX.f[0]) || isnanf(EX->f[0]) || isgreaterequal(EX->f[0], GX.f[0]))
            GX.f[0] = EX->f[0];
        break;

@@ -240,5 +285,7 @@
        break;

    default:
-        goto _default;
+        return 0;
    }
+    return addr;
+}
@@ -75,6 +75,39 @@ int my_mprotect(x86emu_t* emu, void *addr, unsigned long len, int prot);
 void* ElfSetBrk(void* newbrk);
 int my_vfork(x86emu_t* emu);

+#ifndef __NR_socketcall
+int my_socketcall(x86emu_t* emu, int n, unsigned long *args) {
+    int ret = 0;
+    switch(n) {
+        case SYS_SOCKET: ret = socket(args[0], args[1], args[2]); break;
+        case SYS_BIND: ret = bind(args[0], (void*)args[1], args[2]); break;
+        case SYS_CONNECT: ret = connect(args[0], (void*)args[1], args[2]); break;
+        case SYS_LISTEN: ret = listen(args[0], args[1]); break;
+        case SYS_ACCEPT: ret = accept(args[0], (void*)args[1], (void*)args[2]); break;
+        case SYS_GETSOCKNAME: ret = getsockname(args[0], (void*)args[1], (void*)args[2]); break;
+        case SYS_GETPEERNAME: ret = getpeername(args[0], (void*)args[1], (void*)args[2]); break;
+        case SYS_SOCKETPAIR: ret = socketpair(args[0], args[1], args[2], (int*)args[3]); break;
+        case SYS_SEND: ret = send(args[0], (void*)args[1], args[2], args[3]); break;
+        case SYS_RECV: ret = recv(args[0], (void*)args[1], args[2], args[3]); break;
+        case SYS_SENDTO: ret = sendto(args[0], (void*)args[1], args[2], args[3], (void*)args[4], args[5]); break;
+        case SYS_RECVFROM: ret = recvfrom(args[0], (void*)args[1], args[2], args[3], (void*)args[4], (void*)args[5]); break;
+        case SYS_SHUTDOWN: ret = shutdown(args[0], args[1]); break;
+        case SYS_SETSOCKOPT: ret = setsockopt(args[0], args[1], args[2], (void*)args[3], args[4]); break;
+        case SYS_GETSOCKOPT: ret = getsockopt(args[0], args[1], args[2], (void*)args[3], (void*)args[4]); break;
+        case SYS_SENDMSG: ret = sendmsg(args[0], (void*)args[1], args[2]); break;
+        case SYS_RECVMSG: ret = recvmsg(args[0], (void*)args[1], args[2]); break;
+        case SYS_ACCEPT4: ret = my_accept4(emu, args[0], (void*)args[1], (void*)args[2], args[3]); break;
+        #ifdef SYS_RECVMMSG
+        case SYS_RECVMMSG: ret = my_recvmmsg(emu, args[0], (void*)args[1], args[2], args[3], (void*)args[4]); break;
+        case SYS_SENDMMSG: ret = my___sendmmsg(emu, 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", n);
+            ret = -EINVAL;
+    }
+};
+#endif
+
 // cannot include <fcntl.h>, it conflict with some asm includes...
 #ifndef O_NONBLOCK
 #define O_NONBLOCK 04000
@@ -491,39 +524,10 @@ void EXPORT x86Syscall(x86emu_t *emu)
                R_EAX = (uint32_t)-errno;
            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
-                    case SYS_RECVMMSG: R_EAX = my_recvmmsg(emu, args[0], (void*)args[1], args[2], args[3], (void*)args[4]); break;
-                    case SYS_SENDMMSG: R_EAX = my___sendmmsg(emu, 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 = -EINVAL;
-                }
+        case 102:
+            R_EAX = my_socketcall(emu, R_EBX, (unsigned long *)R_ECX);
            if(R_EAX==0xffffffff && errno>0)
                R_EAX = (uint32_t)-errno;
-            }
            break;
 #endif
 #ifndef __NR_olduname
@@ -748,9 +752,8 @@ void EXPORT x86Syscall(x86emu_t *emu)
            break;
 #endif
        default:
-            printf_log(LOG_INFO, "Error: Unsupported Syscall 0x%02Xh (%d)\n", s, s);
-            emu->quit = 1;
-            emu->error |= ERR_UNIMPL;
+            printf_log(LOG_INFO, "Warning: Unsupported Syscall 0x%02Xh (%d)\n", s, s);
+            R_EAX = (uint32_t)-ENOSYS;
            return;
    }
    printf_log(LOG_DEBUG, " => 0x%x\n", R_EAX);
@@ -804,6 +807,10 @@ uint32_t EXPORT my_syscall(x86emu_t *emu)
            return (uintptr_t)ElfSetBrk(p(4));
        case 91:   // munmap
            return (uint32_t)my_munmap(emu, p(4), u32(8));
+#ifndef __NR_socketcall
+        case 102:
+            return my_socketcall(emu, i32(4), (unsigned long *)p(8));
+#endif
 #ifndef __NR_ipc
        case 117:   // ipc
            return (uint32_t)my_ipc(u32(4), i32(8), i32(12), i32(16), p(20), i32(24));
@@ -0,0 +1,159 @@
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include <string.h>
+#include <signal.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <stddef.h>
+
+#include "debug.h"
+#include "box86stack.h"
+#include "x86emu.h"
+#include "x86run.h"
+#include "x86emu_private.h"
+#include "x86run_private.h"
+#include "x86primop.h"
+#include "x86trace.h"
+#include "x87emu_private.h"
+#include "box86context.h"
+#include "bridge.h"
+#include "signals.h"
+
+void print_banner(x86emu_t* ref)
+{
+    printf_log(LOG_NONE, "Warning, difference between Interpreter and Dynarec in %p (%02x %02x %02x %02x %02x %02x %02x %02x)\n"\
+        "=======================================\n", 
+        (void*)ref->old_ip, 
+        ((uint8_t*)ref->old_ip)[0], ((uint8_t*)ref->old_ip)[1], ((uint8_t*)ref->old_ip)[2], ((uint8_t*)ref->old_ip)[3],
+        ((uint8_t*)ref->old_ip)[4], ((uint8_t*)ref->old_ip)[5], ((uint8_t*)ref->old_ip)[6], ((uint8_t*)ref->old_ip)[7]
+    );
+    printf_log(LOG_NONE, "DIFF: Dynarec |  Interpreter\n----------------------\n");
+}
+#define BANNER if(!banner) {banner=1; print_banner(ref);}
+void x86test_check(x86emu_t* ref, uintptr_t ip)
+{
+    int banner = 0;
+    x86test_t* test = &ref->test;
+    x86emu_t* emu = test->emu;
+    if(memcmp(ref->regs, emu->regs, sizeof(emu->regs))) {
+        static const char* regname[] = {"RAX", "RCX", "RDX", "RBX", "RSP", "RBP", "RSI", "RDI"};
+        BANNER;
+        for(int i=0; i<8; ++i) {
+            if(ref->regs[i].dword[0]!=emu->regs[i].dword[0]) {
+                printf_log(LOG_NONE, "%s: %08x | %08x\n", regname[i], ref->regs[i].dword[0], emu->regs[i].dword[0]);
+            }
+        }
+    }
+    if(ip!=emu->ip.dword[0]) {
+        BANNER;
+        printf_log(LOG_NONE, "RIP: %08x | %08x\n", ip, emu->ip.dword[0]);
+    }
+    // flags are volatile, so don't test them
+	//memcpy(&ref->eflags, &emu->eflags, sizeof(emu->eflags));
+    if(memcmp(ref->segs, emu->segs, sizeof(emu->segs))) {
+        static const char* segname[] = {"CS", "DS", "ES", "SS", "FS", "GS"};
+        BANNER;
+        for(int i=0; i<6; ++i) {
+            if(ref->segs[i]!=emu->segs[i]) {
+                printf_log(LOG_NONE, "%s: %04x | %04x\n", segname[i], ref->segs[i], emu->segs[i]);
+            }
+        }
+    }
+    if(ref->top != emu->top) {
+        BANNER;
+        printf_log(LOG_NONE, "X87 TOP: %d | %d\n", ref->top, emu->top);
+    }
+    if(ref->fpu_stack != emu->fpu_stack) {
+        BANNER;
+        printf_log(LOG_NONE, "X87 STACK: %d | %d\n", ref->fpu_stack, emu->fpu_stack);
+    }
+	if(ref->fpu_stack && memcmp(ref->x87, emu->x87, sizeof(emu->x87))) {
+        // need to check each regs, unused one might have different left over value
+        for(int i=0; i<ref->fpu_stack; ++i) {
+            if(ref->x87[(ref->top+i)&7].d != emu->x87[(emu->top+i)&7].d) {
+                BANNER;
+                printf_log(LOG_NONE, "ST%d: %g | %g\n", i, ref->x87[(ref->top+i)&7].d, emu->x87[(emu->top+i)&7].d);
+            }
+        }
+    }
+    //memcpy(ref->fpu_ld, emu->fpu_ld, sizeof(emu->fpu_ld));
+    //memcpy(ref->fpu_ll, emu->fpu_ll, sizeof(emu->fpu_ll));
+	/*if(ref->p_regs != emu->p_regs) {
+        BANNER;
+        printf_log(LOG_NONE, "X87 PREG: %x | %x\n", ref->p_regs, emu->p_regs);
+    }*/
+	if(ref->cw.x16 != emu->cw.x16) {
+        BANNER;
+        printf_log(LOG_NONE, "X87 CW: %x | %x\n", ref->cw.x16, emu->cw.x16);
+    }
+	if(ref->sw.x16 != emu->sw.x16) {
+        BANNER;
+        printf_log(LOG_NONE, "X87 SW: %x | %x\n", ref->sw.x16, emu->sw.x16);
+    }
+	if(memcmp(ref->mmx, emu->mmx, sizeof(emu->mmx))) {
+        BANNER;
+        for(int i=0; i<8; ++i) {
+            if(ref->mmx[i].q!=emu->mmx[i].q) {
+                printf_log(LOG_NONE, "EMM[%d]: %016llx | %016llx\n", i, ref->mmx[i].q, emu->mmx[i].q);
+            }
+        }
+    }
+    if(ref->mxcsr.x32 != emu->mxcsr.x32) {
+        BANNER;
+        printf_log(LOG_NONE, "MXCSR: %x | %x\n", ref->mxcsr.x32, emu->mxcsr.x32);
+    }
+    if(memcmp(ref->xmm, emu->xmm, sizeof(emu->xmm))) {
+        BANNER;
+        for(int i=0; i<8; ++i) {
+            if(ref->xmm[i].q[0]!=emu->xmm[i].q[0] || ref->xmm[i].q[1]!=emu->xmm[i].q[1] ) {
+                printf_log(LOG_NONE, "XMM[%02d]: %016llx-%016llx | %016llx-%016llx\n", i, ref->xmm[i].q[1], ref->xmm[i].q[0], emu->xmm[i].q[1], emu->xmm[i].q[0]);
+            }
+        }
+    }
+    if(test->memsize) {
+        if(memcmp(test->mem, (void*)test->memaddr, test->memsize)) {
+            BANNER;
+            printf_log(LOG_NONE, "MEM: @%p :", (void*)test->memaddr);
+            for(int i=0; i<test->memsize; ++i)
+                printf_log(LOG_NONE, " %02x", test->mem[i]);
+            printf_log(LOG_NONE, " |");
+            for(int i=0; i<test->memsize; ++i)
+                printf_log(LOG_NONE, " %02x", ((uint8_t*)test->memaddr)[i]);
+            printf_log(LOG_NONE, "\n");
+        }
+    }
+    if(banner)  // there was an error, re-sync!
+        CopyEmu(emu, ref);
+}
+#undef BANNER
+
+void x86test_init(x86emu_t* ref, uintptr_t ip)
+{
+    x86test_t* test = &ref->test;
+    if(!test->test) {
+        test->clean = 0;
+        return;
+    }
+    // check if test as a valid emu struct
+    if(!test->emu) {
+        test->emu = NewX86Emu(my_context, ip, (uintptr_t)ref->init_stack, ref->size_stack, 0);
+        CopyEmu(test->emu, ref);
+    } else {
+        // check if IP is same, else, sync
+        uintptr_t prev_ip = test->emu->ip.dword[0];
+        if(test->clean)
+            x86test_check(ref, ip);
+        if(ip != prev_ip || !test->test || !test->clean) {
+            CopyEmu(test->emu, ref);
+        }
+    }
+    // Do a Dry single Step
+    test->memsize = 0;
+    test->clean = 1;
+    ref->old_ip = ip;
+    RunTest(test);
+    // this will be anakyzed next step
+}
@@ -304,6 +304,7 @@ void fpu_fxsave(x86emu_t* emu, void* ed)
    emu->sw.f.F87_TOP = top;
    p->ControlWord = emu->cw.x16;
    p->StatusWord = emu->sw.x16;
+    p->MxCsr = emu->mxcsr.x32;
    uint8_t tags = 0;
    for (int i=0; i<8; ++i)
        tags |= (emu->p_regs[i].tag==0b11)?0:(1<<(i*2));
@@ -313,8 +314,6 @@ void fpu_fxsave(x86emu_t* emu, void* ed)
    p->ErrorSelector = 0;
    p->DataOffset = 0;
    p->DataSelector = 0;
-    p->MxCsr = 0;
-    p->MxCsr_Mask = 0;
    // copy MMX regs...
    for(int i=0; i<8; ++i)
        memcpy(&p->FloatRegisters[i].q[0], (i<stack)?&ST(i):&emu->mmx[i], sizeof(mmx87_regs_t));
@@ -327,6 +326,7 @@ void fpu_fxrstor(x86emu_t* emu, void* ed)
    xsave_t *p = (xsave_t*)ed;
    emu->cw.x16 = p->ControlWord;
    emu->sw.x16 = p->StatusWord;
+    emu->mxcsr.x32 = p->MxCsr;
    emu->top = emu->sw.f.F87_TOP;
    uint8_t tags = p->TagWord;
    for(int i=0; i<8; ++i)
@@ -14,10 +14,6 @@ typedef struct x86emu_s x86emu_t;
 #define LN2		0.69314718055994531
 #define LG2		0.3010299956639812

-void Run66D9(x86emu_t *emu);
-void Run66DD(x86emu_t *emu);
-void RunDF(x86emu_t *emu);
-
 #define ST0 emu->x87[emu->top]
 #define ST1 emu->x87[(emu->top+1)&7]
 #define ST(a) emu->x87[(emu->top+(a))&7]
@@ -1,193 +0,0 @@
-#define _GNU_SOURCE
-#include <stdint.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <math.h>
-#include <string.h>
-
-#include "debug.h"
-#include "box86stack.h"
-#include "x86emu.h"
-#include "x86run.h"
-#include "x86emu_private.h"
-#include "x86run_private.h"
-#include "x87emu_private.h"
-#include "x86primop.h"
-#include "x86trace.h"
-#include "x86run.h"
-
-#define F8      *(uint8_t*)(ip++)
-#define F8S     *(int8_t*)(ip++)
-#define F16     *(uint16_t*)(ip+=2, ip-2)
-#define F16S    *(int16_t*)(ip+=2, ip-2)
-#define F32     *(uint32_t*)(ip+=4, ip-4)
-#define F32S    *(int32_t*)(ip+=4, ip-4)
-#define PK(a)   *(uint8_t*)(ip+a)
-
-#include "modrm.h"
-
-void Run66D9(x86emu_t *emu)
-{
-    uintptr_t ip = R_EIP;
-    uint8_t nextop = F8;
-    reg32_t *oped;
-    switch (nextop) {
-        case 0xC0:
-        case 0xC1:
-        case 0xC2:
-        case 0xC3:
-        case 0xC4:
-        case 0xC5:
-        case 0xC6:
-        case 0xC7:
-        case 0xC8:
-        case 0xC9:
-        case 0xCA:
-        case 0xCB:
-        case 0xCC:
-        case 0xCD:
-        case 0xCE:
-        case 0xCF:
-        case 0xD0:
-        case 0xE0:
-        case 0xE5:
-        case 0xE8:
-        case 0xE9:
-        case 0xEA:
-        case 0xEB:
-        case 0xEC:
-        case 0xED:
-        case 0xEE:
-        case 0xFC:
-        case 0xE1:
-        case 0xE4:
-        case 0xF0:
-        case 0xF1:
-        case 0xF2:
-        case 0xF3:
-        case 0xF4:
-        case 0xF5:
-        case 0xF6:
-        case 0xF7:
-        case 0xF8:
-        case 0xF9:
-        case 0xFA:
-        case 0xFB:
-        case 0xFD:
-        case 0xFE:
-        case 0xFF:
-            ip = R_EIP-2;
-            UnimpOpcode(emu);
-            break;
-        default:
-        switch((nextop>>3)&7) {
-            case 4:     /* FLDENV m */
-                // warning, incomplete
-                GET_EW;
-                fpu_loadenv(emu, (char*)ED, 1);
-                break;
-            case 6:     /* FNSTENV m */
-                // warning, incomplete
-                GET_EW;
-                fpu_savenv(emu, (char*)ED, 1);
-                break;
-            default:
-                ip = R_EIP-2;
-                UnimpOpcode(emu);
-        }
-    }
-    R_EIP = ip;
-}
-
-void Run66DD(x86emu_t *emu)
-{
-    uintptr_t ip = R_EIP;
-    uint8_t nextop = F8;
-    reg32_t *oped;
-    switch(nextop) {
-    
-        case 0xC0:
-        case 0xC1:
-        case 0xC2:
-        case 0xC3:
-        case 0xC4:
-        case 0xC5:
-        case 0xC6:
-        case 0xC7:
-        case 0xC8:
-        case 0xC9:
-        case 0xCA:
-        case 0xCB:
-        case 0xCC:
-        case 0xCD:
-        case 0xCE:
-        case 0xCF:
-        case 0xD0:
-        case 0xE0:
-        case 0xE5:
-        case 0xE8:
-        case 0xE9:
-        case 0xEA:
-        case 0xEB:
-        case 0xEC:
-        case 0xED:
-        case 0xEE:
-        case 0xFC:
-        case 0xE1:
-        case 0xE4:
-        case 0xF0:
-        case 0xF1:
-        case 0xF2:
-        case 0xF3:
-        case 0xF4:
-        case 0xF5:
-        case 0xF6:
-        case 0xF7:
-        case 0xF8:
-        case 0xF9:
-        case 0xFA:
-        case 0xFB:
-        case 0xFD:
-        case 0xFE:
-        case 0xFF:
-            ip = R_EIP-2;
-            UnimpOpcode(emu);
-        break;
-
-    default:
-        switch((nextop>>3)&7) {
-            case 4: /* FRSTOR m94byte */
-                GET_EW;
-                fpu_loadenv(emu, (char*)ED, 1);
-                // get the STx
-                {
-                    char* p =(char*)ED;
-                    p += 14;
-                    for (int i=0; i<8; ++i) {
-                        LD2D(p, &ST(i).d);
-                        p+=10;
-                    }
-                }
-                break;
-            case 6: /* FNSAVE m94byte */
-                GET_EW;
-                // ENV first...
-                fpu_savenv(emu, (char*)ED, 1);
-                // save the STx
-                {
-                    char* p =(char*)ED;
-                    p += 14;
-                    for (int i=0; i<8; ++i) {
-                        D2LD(&ST(i).d, p);
-                        p+=10;
-                    }
-                }
-                reset_fpu(emu);
-                break;
-            default:
-                ip = R_EIP-2; // unfetch
-                UnimpOpcode(emu);
-        }
-    }
-    R_EIP = ip;
-}
@@ -30,7 +30,6 @@ typedef struct atfork_fnc_s {
 } atfork_fnc_t;
 #ifdef DYNAREC
 typedef struct dynablock_s      dynablock_t;
-typedef struct dynablocklist_s  dynablocklist_t;
 typedef struct mmaplist_s       mmaplist_t;
 typedef struct kh_dynablocks_s  kh_dynablocks_t;
 #endif
@@ -38,9 +37,10 @@ typedef struct kh_dynablocks_s  kh_dynablocks_t;
 #define DYNAMAP_SIZE (1<<(32-DYNAMAP_SHIFT))
 #define JMPTABL_SHIFT 16
 #define JMPTABL_SIZE (1<<(32-JMPTABL_SHIFT))
+#define JMPTABLE_MASK ((1<<JMPTABL_SHIFT)-1)

-typedef void* (*procaddess_t)(const char* name);
-typedef void* (*vkprocaddess_t)(void* instance, const char* name);
+typedef void* (*procaddress_t)(const char* name);
+typedef void* (*vkprocaddress_t)(void* instance, const char* name);

 #define MAX_SIGNAL 64

@@ -113,18 +113,16 @@ typedef struct box86context_s {

    kh_threadstack_t    *stacksizes;    // stack sizes attributes for thread (temporary)
    bridge_t            *system;        // other bridges
+    uintptr_t           exit_bridge;    // exit bridge value
    uintptr_t           vsyscall;       // vsyscall bridge value
    dlprivate_t         *dlprivate;     // dlopen library map
-    kh_symbolmap_t      *glwrappers;    // the map of wrapper for glProcs (for GLX or SDL1/2)
-    kh_symbolmap_t      *glmymap;       // link to the mysymbolmap of libGL
-    procaddess_t        glxprocaddress;
    kh_symbolmap_t      *alwrappers;    // the map of wrapper for alGetProcAddress
    kh_symbolmap_t      *almymap;       // link to the mysymbolmap if libOpenAL
    kh_symbolmap_t      *vkwrappers;    // the map of wrapper for VulkanProcs (TODO: check SDL2)
    kh_symbolmap_t      *vkmymap;       // link to the mysymbolmap of libGL
    kh_defaultversion_t *globaldefver;  // the global default version for symbols (the XXX@@vvvv of symbols)
    kh_defaultversion_t *weakdefver;    // the weak default version for symbols (the XXX@@vvvv of symbols)
-    vkprocaddess_t      vkprocaddress;
+    vkprocaddress_t     vkprocaddress;

    #ifndef DYNAREC
    pthread_mutex_t     mutex_once;
@@ -142,7 +140,9 @@ typedef struct box86context_s {
    uint32_t            mutex_thread;
    uint32_t            mutex_bridge;
    uint32_t            mutex_dyndump;
+    uintptr_t           max_db_size;    // the biggest (in x86_64 instructions bytes) built dynablock
    int                 trace_dynarec;
+    pthread_mutex_t     mutex_lock;     // this is for the Test interpreter
    #endif

    library_t           *libclib;       // shortcut to libc library (if loaded, so probably yes)
@@ -208,7 +208,7 @@ typedef struct box86context_s {
 #else
 int GetTID();
 #define mutex_lock(A)       {int tid = GetTID(); while(arm_lock_storeifnull(A, (void*)tid)) sched_yield();}
-#define mutex_trylock(A)    arm_lock_storeifnull(A, (void*)GetTID())
+#define mutex_trylock(A)    (uintptr_t)arm_lock_storeifnull(A, (void*)GetTID())
 #define mutex_unlock(A)     arm_lock_storeifref(A, NULL, (void*)GetTID())
 #endif

@@ -18,35 +18,43 @@ void customFree(void* p);

 #ifdef DYNAREC
 typedef struct dynablock_s dynablock_t;
-typedef struct dynablocklist_s dynablocklist_t;
 // custom protection flag to mark Page that are Write protected for Dynarec purpose
-uintptr_t AllocDynarecMap(dynablock_t* db, int size);
-void FreeDynarecMap(dynablock_t* db, uintptr_t addr, uint32_t size);
+uintptr_t AllocDynarecMap(size_t size);
+void FreeDynarecMap(uintptr_t addr);

-void addDBFromAddressRange(uintptr_t addr, uintptr_t size);
-void cleanDBFromAddressRange(uintptr_t addr, uintptr_t size, int destroy);
+void addDBFromAddressRange(uintptr_t addr, size_t size);
+void cleanDBFromAddressRange(uintptr_t addr, size_t size, int destroy);

-dynablocklist_t* getDB(uintptr_t idx);
-void addJumpTableIfDefault(void* addr, void* jmp);
+dynablock_t* getDB(uintptr_t idx);
+int getNeedTest(uintptr_t idx);
+int addJumpTableIfDefault(void* addr, void* jmp); // return 1 if write was succesfull
+int setJumpTableIfRef(void* addr, void* jmp, void* ref); // return 1 if write was succesfull
 void setJumpTableDefault(void* addr);
+void setJumpTableDefaultRef(void* addr, void* jmp);
 int isJumpTableDefault(void* addr);
 uintptr_t getJumpTable();
 uintptr_t getJumpTableAddress(uintptr_t addr);
+uintptr_t getJumpAddress(uintptr_t addr);
 #endif

 #define PROT_DYNAREC    0x80
 #define PROT_DYNAREC_R  0x40
-#define PROT_CUSTOM     (PROT_DYNAREC | PROT_DYNAREC_R)
+#define PROT_NOPROT     0x20
+#define PROT_MMAP       0x10
+#define PROT_DYN        (PROT_DYNAREC | PROT_DYNAREC_R | PROT_NOPROT)
+#define PROT_CUSTOM     (PROT_DYNAREC | PROT_DYNAREC_R | PROT_MMAP | PROT_NOPROT)

-void updateProtection(uintptr_t addr, uintptr_t size, uint32_t prot);
-void setProtection(uintptr_t addr, uintptr_t size, uint32_t prot);
+void updateProtection(uintptr_t addr, size_t size, uint32_t prot);
+void setProtection(uintptr_t addr, size_t size, uint32_t prot);
+void setProtection_mmap(uintptr_t addr, size_t size, uint32_t prot);
+void freeProtection(uintptr_t addr, size_t size);
+void refreshProtection(uintptr_t addr);
 uint32_t getProtection(uintptr_t addr);
-void forceProtection(uintptr_t addr, uintptr_t size, uint32_t prot);
-void freeProtection(uintptr_t addr, uintptr_t size);
+int getMmapped(uintptr_t addr);
 void loadProtectionFromMap();
 #ifdef DYNAREC
-void protectDB(uintptr_t addr, uintptr_t size);
-void unprotectDB(uintptr_t addr, uintptr_t size, int mark); // if mark==0, the blocks are not marked as potentially dirty
+void protectDB(uintptr_t addr, size_t size);
+void unprotectDB(uintptr_t addr, size_t size, int mark);    // if mark==0, the blocks are not marked as potentially dirty
 int isprotectedDB(uintptr_t addr, size_t size);
 int IsInHotPage(uintptr_t addr);
 int AreaInHotPage(uintptr_t start, uintptr_t end);
@@ -24,7 +24,10 @@ extern uintptr_t box86_nodynarec_start, box86_nodynarec_end;
 extern int box86_dynarec_fastnan;
 extern int box86_dynarec_fastround;
 extern int box86_dynarec_hotpage;
+extern int box86_dynarec_wait;
+extern int box86_dynarec_fastpage;
 extern int box86_dynarec_bleeding_edge;
+extern int box86_dynarec_test;
 #ifdef ARM
 extern int arm_vfp;     // vfp version (3 or 4), with 32 registers is mendatory
 extern int arm_swap;
@@ -3,31 +3,22 @@

 typedef struct x86emu_s x86emu_t;
 typedef struct dynablock_s dynablock_t;
-typedef struct dynablocklist_s dynablocklist_t;
 typedef struct kh_dynablocks_s  kh_dynablocks_t;

 uint32_t X31_hash_code(void* addr, int len);
-dynablocklist_t* NewDynablockList(uintptr_t text, int textsz, int direct);
-void FreeDynablockList(dynablocklist_t** dynablocks);
 void FreeDynablock(dynablock_t* db, int need_lock);
 void MarkDynablock(dynablock_t* db);
-void FreeRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size);
-void MarkRangeDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size);
+void MarkRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size);
+int FreeRangeDynablock(dynablock_t* db, uintptr_t addr, uintptr_t size);
+void FreeInvalidDynablock(dynablock_t* db, int need_lock);
+dynablock_t* InvalidDynablock(dynablock_t* db, int need_lock);

-dynablock_t* FindDynablockFromNativeAddress(void* addr);    // defined in box86context.h
-dynablock_t* FindDynablockDynablocklist(void* addr, kh_dynablocks_t* dynablocks);
+dynablock_t* FindDynablockFromNativeAddress(void* addr);    // defined in box64context.h

-uintptr_t StartDynablockList(dynablocklist_t* db);
-uintptr_t EndDynablockList(dynablocklist_t* db);
-void MarkDirectDynablock(dynablocklist_t* dynablocks, uintptr_t addr, uintptr_t size);
-
-// 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
+// Handling of Dynarec block (i.e. an exectable chunk of x64 translated code)
+dynablock_t* DBGetBlock(x86emu_t* emu, uintptr_t addr, int create);   // return NULL if block is not found / cannot be created. Don't create if create==0
 dynablock_t* DBAlternateBlock(x86emu_t* emu, uintptr_t addr, uintptr_t filladdr);

-// Create and Add an new dynablock in the list, handling direct/map
-dynablock_t *AddNewDynablock(dynablocklist_t* dynablocks, uintptr_t addr, int* created);
-
 // for use in signal handler
 void cancelFillBlock();

@@ -4,5 +4,6 @@
 typedef struct x86emu_s x86emu_t;

 void DynaCall(x86emu_t* emu, uintptr_t addr); // try to use DynaRec... Fallback to EmuCall if no dynarec available
+void x86test_init(x86emu_t* ref, uintptr_t ip);

 #endif // __DYNAREC_H_
@@ -3,6 +3,9 @@

 typedef struct dynablock_s dynablock_t;
 typedef struct x86emu_s x86emu_t;
+typedef struct instsize_s instsize_t;
+
+void addInst(instsize_t* insts, size_t* size, int x86_size, int arm_size);

 void CancelBlock();
 void* FillBlock(dynablock_t* block, uintptr_t addr);
@@ -10,7 +10,7 @@ typedef struct box86context_s box86context_t;
 typedef struct x86emu_s x86emu_t;
 typedef struct needed_libs_s needed_libs_t;
 #ifdef DYNAREC
-typedef struct dynablocklist_s dynablocklist_t;
+typedef struct dynablock_s dynablock_t;
 #endif

 elfheader_t* LoadAndCheckElfHeader(FILE* f, const char* name, int exec); // exec : 0 = lib, 1 = exec
@@ -49,8 +49,7 @@ uint32_t GetTLSSize(elfheader_t* h);
 void* GetTLSPointer(box86context_t* context, elfheader_t* h);
 void* GetDTatOffset(box86context_t* context, int index, int offset);
 #ifdef DYNAREC
-dynablocklist_t* GetDynablocksFromAddress(box86context_t* context, uintptr_t addr);
-dynablocklist_t* GetDynablocksFromElf(elfheader_t* h);
+dynablock_t* GetDynablocksFromAddress(box86context_t* context, uintptr_t addr);
 #endif
 void ResetSpecialCaseMainElf(elfheader_t* h);
 void CreateMemorymapFile(box86context_t* context, int fd);
@@ -0,0 +1,12 @@
+#ifndef __GL_TOOLS_H__
+#define __GL_TOOLS_H__
+
+typedef struct box86context_s box86context_t;
+
+typedef void* (*glprocaddress_t)(const char* name);
+
+void freeGLProcWrapper(box86context_t* context);
+
+void* getGLProcAddress(x86emu_t* emu, glprocaddress_t procaddr, const char* rname);
+
+#endif //__GL_TOOLS_H__
@@ -457,6 +457,19 @@ typedef struct my_GDBusObjectManagerClientClass_s
  void* padding[8];
 } my_GDBusObjectManagerClientClass_t;

+typedef struct my_GstObjectClass_s {
+  my_GInitiallyUnownedClass_t parent;
+  const char*   path_string_separator;
+  void          (*deep_notify)      (void* object, void* orig, void* pspec);
+  void*        _gst_reserved[4];
+} my_GstObjectClass_t;
+
+typedef struct my_GstAllocatorClass_s {
+  my_GstObjectClass_t parent;
+  void*     (*alloc)      (void *allocator, size_t size, void *params);
+  void      (*free)       (void *allocator, void *memory);
+  void*    _gst_reserved[4];
+} my_GstAllocatorClass_t;

 // GTypeValueTable
 typedef struct my_GTypeValueTable_s {
@@ -540,6 +553,8 @@ GTKCLASS(GtkMenuShell2)             \
 GTKCLASS(GtkMenuBar2)               \
 GTKCLASS(MetaFrames2)               \
 GTKCLASS(GDBusObjectManagerClient)  \
+GTKCLASS(GstObject)                 \
+GTKCLASS(GstAllocator)              \

 #define GTKCLASS(A) void Set##A##ID(int id);
 GTKCLASSES()
@@ -27,10 +27,6 @@ int IsSameLib(library_t* lib, const char* path);    // check if lib is same (pat
 int GetLibGlobalSymbolStartEnd(library_t* lib, const char* name, uintptr_t* start, uintptr_t* end, size_t size, int *weak, int version, const char* vername, int local);
 int GetLibWeakSymbolStartEnd(library_t* lib, const char* name, uintptr_t* start, uintptr_t* end, size_t size, int* weak, int version, const char* vername, int local);
 int GetLibLocalSymbolStartEnd(library_t* lib, const char* name, uintptr_t* start, uintptr_t* end, size_t size, int* weak, int version, const char* vername, int local);
-void fillGLProcWrapper(box86context_t* context);
-void freeGLProcWrapper(box86context_t* context);
-void fillALProcWrapper(box86context_t* context);
-void freeALProcWrapper(box86context_t* context);
 needed_libs_t* GetNeededLibs(library_t* lib);
 int GetNeededLibN(library_t* lib);
 library_t* GetNeededLib(library_t* lib, int idx);
@@ -23,6 +23,7 @@ void myStackAlign(const char* fmt, uint32_t* st, uint32_t* mystack);
 void myStackAlignGVariantNew(const char* fmt, uint32_t* st, uint32_t* mystack);
 void myStackAlignW(const char* fmt, uint32_t* st, uint32_t* mystack);

+void AlignStat64(const void* source, void* dest);
 void UnalignStat64(const void* source, void* dest);

 void UnalignStatFS64(const void* source, void* dest);
@@ -10,6 +10,7 @@ void SetupX86Emu(x86emu_t *emu);
 void FreeX86Emu(x86emu_t **x86emu);
 void FreeX86EmuFromStack(x86emu_t **emu);
 void CloneEmu(x86emu_t *newemu, const x86emu_t* emu);
+void CopyEmu(x86emu_t *newemu, const x86emu_t* emu);
 void SetTraceEmu(uintptr_t trace_start, uintptr_t trace_end);

 box86context_t* GetEmuContext(x86emu_t* emu);
@@ -34,8 +35,6 @@ void ResetSegmentsCache(x86emu_t *emu);
 const char* DumpCPURegs(x86emu_t* emu, uintptr_t ip);

 void StopEmu(x86emu_t* emu, const char* reason);
-void PushExit(x86emu_t* emu);
-void* GetExit();
 void EmuCall(x86emu_t* emu, uintptr_t addr);
 void AddCleanup(x86emu_t *emu, void *p);
 void AddCleanup1Arg(x86emu_t *emu, void *p, void* a);
@@ -3,7 +3,9 @@
 #include <stdint.h>

 typedef struct x86emu_s x86emu_t;
+typedef struct x86test_s x86test_t;
 int Run(x86emu_t *emu, int step); // 0 if run was successfull, 1 if error in x86 world
+int RunTest(x86test_t *test);
 int DynaRun(x86emu_t *emu);

 uint32_t LibSyscall(x86emu_t *emu);
@@ -173,6 +173,7 @@ GO("libcap.so.2", cap)
 GO("libpcap.so.0", pcap)
 GO("libFAudio.so.0", faudio)
 GO("libtbbmalloc_proxy.so.2", tbbmallocproxy)
+GO("libfuse.so.2", libfuse)

 #ifdef ANDROID
 GO("libc.so", libc)
@@ -43,6 +43,35 @@ void UnalignStat64(const void* source, void* dest)
    i386st->st_ino      = st->st_ino;
 }

+void AlignStat64(const void* source, void* dest)
+{
+    struct stat64 *st = (struct stat64*) dest;
+    struct i386_stat64 *i386st = (struct i386_stat64*)source;
+    
+    st->st_dev          = i386st->st_dev;
+#ifndef POWERPCLE
+    st->__st_ino        = i386st->__st_ino;
+#else
+    st->st_ino          = i386st->__st_ino; // Separate __st_ino doesn't 
+                                            // exist on powerpc
+#endif
+    st->st_mode         = i386st->st_mode;
+    st->st_nlink        = i386st->st_nlink;
+    st->st_uid          = i386st->st_uid;
+    st->st_gid          = i386st->st_gid;
+    st->st_rdev         = i386st->st_rdev;
+    st->st_size         = i386st->st_size;
+    st->st_blksize      = i386st->st_blksize;
+    st->st_blocks       = i386st->st_blocks;
+    st->st_atime        = i386st->st_atime;
+    st->st_atime_nsec   = i386st->st_atime_nsec;
+    st->st_mtime        = i386st->st_mtime;
+    st->st_mtime_nsec   = i386st->st_mtime_nsec;
+    st->st_ctime        = i386st->st_ctime;
+    st->st_ctime_nsec   = i386st->st_ctime_nsec;
+    st->st_ino          = i386st->st_ino;
+}
+
 struct native_fsid {
  int     val[2];
 };
@@ -277,6 +277,10 @@ uint32_t RunFunctionHandler(int* exit, int dynarec, i386_ucontext_t* sigcontext,
    //trace_start = 0; trace_end = 1; // disabling trace, globably for now...

    x86emu_t *emu = thread_get_emu();
+    #ifdef DYNAREC
+    if(box86_dynarec_test)
+        emu->test.test = 0;
+    #endif

    printf_log(LOG_DEBUG, "%04d|signal function handler %p called, ESP=%p\n", GetTID(), (void*)fnc, (void*)R_ESP);
    
@@ -308,6 +312,12 @@ uint32_t RunFunctionHandler(int* exit, int dynarec, i386_ucontext_t* sigcontext,
    uint32_t ret = R_EAX;
    emu->quitonlongjmp = oldquitonlongjmp;

+    #ifdef DYNAREC
+    if(box86_dynarec_test)
+        emu->test.test = 0;
+        emu->test.clean = 0;
+    #endif
+
    if(emu->longjmp) {
        // longjmp inside signal handler, lets grab all relevent value and do the actual longjmp in the signal handler
        emu->longjmp = 0;
@@ -676,7 +686,7 @@ void my_sigactionhandler_oldcode(int32_t sig, int simple, int Locks, siginfo_t*
            //relockMutex(Locks);   // do not relock mutex, because of the siglongjmp, whatever was running is canceled
            #ifdef DYNAREC
            if(Locks & is_dyndump_locked)
-                CancelBlock();
+                CancelBlock(1);
            #endif
            siglongjmp(ejb->jmpbuf, 1);
        }
@@ -718,6 +728,9 @@ void my_sigactionhandler_oldcode(int32_t sig, int simple, int Locks, siginfo_t*
 }

 extern void* current_helper;
+#ifdef DYNAREC
+static uint32_t mutex_dynarec_prot = 0;
+#endif

 void my_box86signalhandler(int32_t sig, siginfo_t* info, void * ucntx)
 {
@@ -749,18 +762,20 @@ void my_box86signalhandler(int32_t sig, siginfo_t* info, void * ucntx)
    #endif
 #ifdef DYNAREC
    if((Locks & is_dyndump_locked) && (sig==SIGSEGV) && current_helper) {
-        relockMutex(Locks);
-        CancelBlock();
+        CancelBlock(0);
        cancelFillBlock();  // Segfault inside a Fillblock
+        relockMutex(Locks);
    }
    dynablock_t* db = NULL;
    int db_searched = 0;
    if ((sig==SIGSEGV) && (addr) && (info->si_code == SEGV_ACCERR) && (prot&PROT_CUSTOM)) {
+        mutex_lock(&mutex_dynarec_prot);
        // access error, unprotect the block (and mark them dirty)
        unprotectDB((uintptr_t)addr, 1, 1);    // unprotect 1 byte... But then, the whole page will be unprotected
        // check if SMC inside block
        db = FindDynablockFromNativeAddress(pc);
        db_searched = 1;
+        int db_need_test = (db && !box86_dynarec_fastpage)?getNeedTest((uintptr_t)db->x86_addr):0;
        dynarec_log(LOG_INFO/*LOG_DEBUG*/, "SIGSEGV with Access error on %p for %p , db=%p(%p)\n", pc, addr, db, db?((void*)db->x86_addr):NULL);
        static uintptr_t repeated_page = 0;
        static int repeated_count = 0;
@@ -772,7 +787,7 @@ void my_box86signalhandler(int32_t sig, siginfo_t* info, void * ucntx)
            repeated_page = (uintptr_t)addr&~0xfff;
            repeated_count = 0;
        }
-        if(db && ((addr>=db->x86_addr && addr<(db->x86_addr+db->x86_size)) || db->need_test)) {
+        if(db && ((addr>=db->x86_addr && addr<(db->x86_addr+db->x86_size)) || db_need_test)) {
            // dynablock got auto-dirty! need to get out of it!!!
            emu_jmpbuf_t* ejb = GetJmpBuf();
            if(ejb->jmpbuf_ok) {
@@ -799,40 +814,46 @@ void my_box86signalhandler(int32_t sig, siginfo_t* info, void * ucntx)
                    dynarec_log(LOG_INFO, "Dynablock %p(%p) unprotected, getting out (arm pc=%p, x86_pc=%p, special=%d)!\n", db, db->x86_addr, pc, (void*)ejb->emu->ip.dword[0], special);
                }
                //relockMutex(Locks);   // do not relock because of he siglongjmp
-                #ifdef DYNAREC
+                mutex_unlock(&mutex_dynarec_prot);
                if(Locks & is_dyndump_locked)
-                    CancelBlock();
-                #endif
+                    CancelBlock(1);
+                ejb->emu->test.clean = 0;
                siglongjmp(ejb->jmpbuf, 2);
            }
            dynarec_log(LOG_INFO, "Warning, Auto-SMC (%p for db %p/%p) detected, but jmpbuffer not ready!\n", (void*)addr, db, (void*)db->x86_addr);
        }
        // done
        if(prot&PROT_WRITE) {
+            mutex_unlock(&mutex_dynarec_prot);
            // if there is no write permission, don't return and continue to program signal handling
            relockMutex(Locks);
            return;
        }
+        mutex_unlock(&mutex_dynarec_prot);
    } else if ((sig==SIGSEGV) && (addr) && (info->si_code == SEGV_ACCERR) && (prot&(PROT_READ|PROT_WRITE))) {
+        mutex_lock(&mutex_dynarec_prot);
        db = FindDynablockFromNativeAddress(pc);
        db_searched = 1;
        if(db && db->x86_addr>= addr && (db->x86_addr+db->x86_size)<addr) {
            dynarec_log(LOG_INFO, "Warning, addr inside current dynablock!\n");
        }
-        if(addr && pc && db) {
-            static void* glitch_pc = NULL;
-            static void* glitch_addr = NULL;
-            static int glitch_prot = 0;
+        // mark stuff as unclean
+        cleanDBFromAddressRange(((uintptr_t)addr)&~(box86_pagesize-1), box86_pagesize, 0);
+        static void* glitch_pc = NULL;
+        static void* glitch_addr = NULL;
+        static int glitch_prot = 0;
+        if(addr && pc /*&& db*/) {
            if((glitch_pc!=pc || glitch_addr!=addr || glitch_prot!=prot)) {
                // probably a glitch due to intensive multitask...
-                dynarec_log(/*LOG_DEBUG*/LOG_INFO, "SIGSEGV with Access error on %p for %p , db=%p, retrying\n", pc, addr, db);
+                dynarec_log(/*LOG_DEBUG*/LOG_INFO, "SIGSEGV with Access error on %p for %p, db=%p, prot=0x%x, retrying\n", pc, addr, db, prot);
                glitch_pc = pc;
                glitch_addr = addr;
                glitch_prot = prot;
                relockMutex(Locks);
-                sched_yield();  // give time to the other process
+                mutex_unlock(&mutex_dynarec_prot);
                return; // try again
            }
+dynarec_log(/*LOG_DEBUG*/LOG_INFO, "Repeated SIGSEGV with Access error on %p for %p, db=%p, prot=0x%x\n", pc, addr, db, prot);
            glitch_pc = NULL;
            glitch_addr = NULL;
            glitch_prot = 0;
@@ -848,16 +869,23 @@ void my_box86signalhandler(int32_t sig, siginfo_t* info, void * ucntx)
                glitch2_addr = addr;
                glitch2_prot = prot;
                sched_yield();  // give time to the other process
-                forceProtection((uintptr_t)addr, 1, prot); // force the protection
+                refreshProtection((uintptr_t)addr);
                relockMutex(Locks);
                sched_yield();  // give time to the other process
+                mutex_unlock(&mutex_dynarec_prot);
                return; // try again
            }
            glitch2_pc = NULL;
            glitch2_addr = NULL;
            glitch2_prot = 0;
        }
+        mutex_unlock(&mutex_dynarec_prot);
+    } else if ((sig==SIGSEGV) && (addr) && (info->si_code == SEGV_ACCERR) && (prot&PROT_DYNAREC_R)) {
+        // unprotect and continue to signal handler, because Write is not there on purpose
+        unprotectDB((uintptr_t)addr, 1, 1);    // unprotect 1 byte... But then, the whole page will be unprotected
    }
+    if(!db_searched)
+        db = FindDynablockFromNativeAddress(pc);
 #else
    void* db = NULL;
 #endif
@@ -928,7 +956,7 @@ exit(-1);
            emu->init_stack, emu->init_stack+emu->size_stack, emu->stack2free, (void*)R_EBP, 
            addr, info->si_code, prot, db, db?db->block:0, db?(db->block+db->size):0, 
            db?db->x86_addr:0, db?(db->x86_addr+db->x86_size):0, 
-            getAddrFunctionName((uintptr_t)(db?db->x86_addr:0)), (db?db->need_test:0)?"need_stest":"clean", db?db->hash:0, hash);
+            getAddrFunctionName((uintptr_t)(db?db->x86_addr:0)), (db?getNeedTest((uintptr_t)db->x86_addr):0)?"need_stest":"clean", db?db->hash:0, hash);
 #if defined(ARM)
        static const char* reg_name[] = {"EAX", "ECX", "EDX", "EBX", "ESP", "EBP", "ESI", "EDI"};
        if(db)
@@ -1309,7 +1337,7 @@ EXPORT int my_makecontext(x86emu_t* emu, void* ucp, void* fnc, int32_t argc, int
    }
    // push the return value
    --esp;
-    *esp = (uintptr_t)GetExit();
+    *esp = my_context->exit_bridge;
    u->uc_mcontext.gregs[REG_ESP] = (uintptr_t)esp;
    
    return 0;
@@ -1324,7 +1352,12 @@ EXPORT int my_swapcontext(x86emu_t* emu, void* ucp1, void* ucp2)
    my_setcontext(emu, ucp2);
    return 0;
 }
-
+#ifdef DYNAREC
+static void atfork_child_dynarec_prot(void)
+{
+    arm_lock_stored(&mutex_dynarec_prot, 0);
+}
+#endif
 void init_signal_helper(box86context_t* context)
 {
    // setup signal handling
@@ -1343,6 +1376,10 @@ void init_signal_helper(box86context_t* context)
    sigaction(SIGILL, &action, NULL);

 	pthread_once(&sigstack_key_once, sigstack_key_alloc);
+#ifdef DYNAREC
+    atfork_child_dynarec_prot();
+    pthread_atfork(NULL, NULL, atfork_child_dynarec_prot);
+#endif
 }

 void fini_signal_helper()
Author	SHA1	Message	Date
ptitSeb	157d368939	Optimized Exit bridge managment	2023-05-10 17:27:31 +02:00
ptitSeb	ea7eb6f866	Improved custom memory allocator	2023-05-10 16:23:28 +02:00
ptitSeb	c6cdde4b53	Fixed a nasty small memory leak	2023-05-10 16:22:56 +02:00
ptitSeb	df71c7da43	[DYNAREC] Removed all remaining trace of dynablocklist_t	2023-05-09 14:53:13 +02:00
ptitSeb	ec8f7a6179	Added some more libgstreamer wrapped functions	2023-05-09 14:45:08 +02:00
ptitSeb	cbbe23606d	[DYNAREC] Fixed some remaining issues (PvZ works now)	2023-05-08 13:30:43 +02:00
ptitSeb	1c4898754f	[DYNAREC] Removed DynablockList, backported from Box64 (PvZ broken for now)	2023-05-08 13:30:43 +02:00
ptitSeb	bcb19dbe5d	Improved fpu_fxsave/fpu_fxrstor	2023-05-08 13:30:10 +02:00
ptitSeb	039c62f740	[DYNAREC] Removed dynablocks sons handling, it makes things compicated for not enough benefits (fixed GoG Windows installers)	2023-05-04 21:30:07 +02:00
ptitSeb	7663aacd1b	Added some Hack for wine int 2d calls	2023-05-04 21:28:41 +02:00
ptitSeb	fd5bf804ec	[DYNAREC] Slightly better handling of dynablock getting dirty during signal	2023-05-04 17:26:28 +02:00
ptitSeb	70d18cfb7c	[DYNAREC] Slightly better signal handling for running dynablock getting dirty durring signal	2023-05-04 17:25:40 +02:00
ptitSeb	1e32ccc9c7	Try to fix a wine error message	2023-05-04 17:24:56 +02:00
ptitSeb	23fa99d148	[DYNAREC_TEST] Fixed 0F AE opcodes handling	2023-05-04 17:23:58 +02:00
josch	0e3ce6d8b1	add NO_LIB_INSTALL and NO_CONF_INSTALL (#822 ) This adds commits 6ff68849f88cbec622fe7a485e301111319ab8d6 and 53687387ebca091043cecbfc2d99494865336b94 from box64. This allows installation of box86 without any configuration files or libraries involved. Intended usecases are * packaging box86 with multiple variants in one package, where multiple variants would step on each other during installation. * packaging box86 for distributions shipping their own armhf shared library versions of libc and c++ (for example via multiarch)	2023-05-03 21:43:16 +02:00
ptitSeb	970384559f	[DYNAREC] Another fix for DYNAEC_TEST	2023-05-01 09:42:59 +02:00
ptitSeb	2f8a939485	[DYNAREC] Some more improvement to DYNAREC_TEST to avoid some false-positive	2023-04-30 19:05:45 +02:00
ptitSeb	4f13c33628	[DYNAREC] Various small improvement for DYNAREC_TEST and DYNARCE_X87DOUBLE	2023-04-30 12:21:44 +02:00
ptitSeb	4ef5da404c	Small optimto some float 66 0F opcodes	2023-04-30 11:02:08 +02:00
ptitSeb	15fa9f687c	[DYNAREC] Added fastnan=0 handling for MINPS/MAXPS opcodes	2023-04-30 11:00:42 +02:00
ptitSeb	331e5700a2	[DYNAREC] Fixed DYAREC_TEST false-positive when opcode use NOTEST and next opcode found a difference	2023-04-30 11:00:14 +02:00
ptitSeb	1411d44b51	[DYNAREC] Backported DYNAREC_TEST to compare dynarec vs interpreter state	2023-04-29 19:30:35 +02:00
ptitSeb	bf53a76223	[INTERPRETER] Reverted the label jumtables and go back to classic switch/case, and prepare for TEST_DYNAREC too	2023-04-29 14:14:37 +02:00
ptitSeb	ecbc1e6642	Small improment on x87 interpreter	2023-04-29 10:07:29 +02:00
ptitSeb	7dc5b501f3	Fixed regression on wrapped libGL	2023-04-29 09:13:37 +02:00
ptitSeb	af67f69ee2	Refactored libgL symbol fetching (backported from box64)	2023-04-23 17:39:18 +02:00
ptitSeb	8f2a4ea880	Added 1 libatk wrapped function	2023-04-23 17:02:07 +02:00
ptitSeb	eb44fd62fb	Added more gstvideo wrapped function and fixed wrapped gst_pad_set_getrange_function_full callback (for #815 and #811 )	2023-04-19 11:43:13 +02:00
ptitSeb	46f015da9d	Added 1 gtk2 function	2023-04-19 10:33:22 +02:00
ptitSeb	705414a3d0	[DYNAREC] Reorganized pass3 to reduce size of box binary and lower compilation time	2023-04-15 09:20:04 +02:00
Ryan Fortner	4591d337cb	add missing \ to debian rules (#812 )	2023-04-09 22:22:24 +02:00
ptitSeb	54e13be02a	Added some more gstreamer wrapped function (for #811 )	2023-04-09 09:53:15 +02:00
ptitSeb	1524e883cd	Added 0F 08 opcode	2023-04-08 17:02:33 +02:00
ptitSeb	7a65bd60e9	Added 66 0F 07 opcode	2023-04-08 16:58:32 +02:00
ptitSeb	8c1694e530	Added 0F 01/02/03/05/06/07 opcodes	2023-04-08 16:56:51 +02:00
ptitSeb	292d1455b6	Added 66 0F 2B MOVNTPD opcode ([DYNAREC] too)	2023-04-08 16:34:57 +02:00
ptitSeb	9c69702011	Fixed various rounding error in interpreter	2023-04-08 16:26:10 +02:00
ptitSeb	3554d6decb	Added 66 0F 38 06 opcode	2023-04-08 15:28:48 +02:00
ptitSeb	7fb36acf85	Added libfuse support, and so appimage support (back ported from box64)	2023-04-06 21:09:33 +02:00
ptitSeb	ed3296978e	[DYNAREC] Never protect bridge memory (improve Steam stability a lot)	2023-04-02 18:44:44 +02:00
Jai A P	4b212f91bf	Latest Batch of PKGBUILDS (#804 ) * Latest Batch of PKGBUILDS * Fix * Fix	2023-03-31 14:33:50 +02:00
wannacu	94f451c4b7	Add a tool to generate wrapping library (#803 ) This tool will automatically generate part of the wrapping library, certainly to avoid a lot of manual writing hoping to bring some help.	2023-03-30 09:57:06 +02:00
ptitSeb	13bbf105ad	Better wrapping for syscall 102	2023-03-23 10:36:54 +01:00
ptitSeb	3eb16fb7bd	Wrapped __isoc99_swscanf libc function	2023-03-23 10:36:39 +01:00
wannacu	e85bcb671a	Fix libjpeg62 wrapping error (#793 )	2023-03-23 08:06:28 +01:00
Sevan Janiyan	1a0888ca51	docs/COMPILE.md: Use trusted.gpg.d (#792 ) add-key(8) is deprecated and causes warnings to be reported when running apt-get update after adding the repo.	2023-03-22 18:27:50 +01:00
Sevan Janiyan	3043f5066f	install_steam.sh: More deps (#790 ) Needed to get steam started.	2023-03-21 08:26:15 +01:00
ptitSeb	b3b55d2d74	[DYNAREC] Fixed a potential issue when Neon Cache is converted and contains both single and double (quad) elements (for #789 )	2023-03-19 16:31:29 +01:00
ptitSeb	58426cbfb6	[DYNAREC] Roolback previous commit and do a better one (still for #789 , probably others too)	2023-03-18 18:57:27 +01:00
ptitSeb	4a88ff38ea	[DYNAREC] Fixed a type in a SSE2 helper function (should help #789 , probably others too)	2023-03-18 18:44:38 +01:00
ptitSeb	8d96ff02e3	Added a few wrapped function to libgio2	2023-03-18 18:43:50 +01:00
Alex	eaa6ecb9cc	steam: install mesa for armhf if present for arm64 (#786 ) * steam: install mesa for armhf if present for arm64 * fix a copy-pasting issue	2023-03-14 16:06:54 +01:00
Alex	027909bddc	steam: install additional libraries to suppress wrong ELF class errors (#784 )	2023-03-14 15:29:51 +01:00
ptitSeb	0eca6a8b94	[DYNAREC] Fix some potential issue in the next jump handling	2023-03-12 18:12:30 +01:00
ptitSeb	24494ba143	Don't quit on unsupported (non-libc) syscalls (for #782 )	2023-03-12 12:26:52 +01:00
ptitSeb	a813321ad4	Start of a new dev. cycle	2023-03-12 12:00:22 +01:00
Fancy2209	a43884ff66	Fix ARM64 Compile Instructions (#779 ) CMake needs it to be -DARM64=1 from my experience at least	2023-03-10 21:30:29 +01:00
Ryan Fortner	27b269f53a	add -dnogit=1 flag to debian rules (#778 )	2023-03-10 20:16:18 +01:00