Squashed 'externals/xbyak/' changes from 590c10e37..9357732aa

9357732aa v6.06 b161a3eb7 update changelog e5453b064 Merge branch 'dev' 2a265d9d9 memfd: keep file descriptor open during allocation lifetime 31ff018ed replace unsigned int with uint32_t in xbyak_util.h e427b2231 move Type into Cpu 8cf41063b refactoring Cpu 66d62968d add -cpuid to test_util cf7cb744c add comments 028112949 add detection of clzero c88007b03 update doc 4cb2e77a9 v6.052 015c27cf6 Merge branch 'dev' d808f9ecb add test of Cpu::has b48a7bb1b add operator== to Type 1a90f456c v6.051 96cb1d660 update changelog 1f5a77f56 Merge branch 'dev' 87e14f02a add test with noexception 30144f809 fix error when XBYAK_NO_EXCEPTION is defined ec15751df Merge branch 'dev' 4831b3fb3 v6.05 9ddf251f3 remove debug code 905b31bab fix typo 72d1ac118 add movdir64b a6665996b add detection of movdir64b 92ddc6dfd remove Type::operator<<() because it's ambiguous 9cdd40f5e add detection of movdiri 379f8bf37 add movdiri 0ad6db138 fix cldemote test for 32-bit 84ab46bb3 add cldemote to Cpu a84ddc12d support cldemote 3a6cc626e add clwb 38c40c02e detection of clflushopt c061ac839 avoid unnecessary replacement in readme.md 095ebbff1 extend Cpu::Type to 128 bit 3ea8e45d3 Merge branch 'dev' f7bfc2634 v6.041 ab4e52d67 update doc 2bef54399 include intrin.h on mingw 621ed0775 Merge branch 'dev' 4f5893e1e memfd_create: disable for ANDROID_API < 30 684d69db0 Merge branch 'dev' ea88b6d85 disable warning on mingw 7e8923fa5 add colon 9914216bd Merge branch 'dev' b335602ff [skip ci] tweak doc 25fbb3996 tweak doc 122054dc4 Formatted Supported OS Section d4fa7e46b Reworked Authors Section 9e4b64a74 Adjusted Spacing ff9fdb45f Use Github Sponsor Widget 893b31d46 Added Quicklinks 2d2adf78b Added License Badge 335b5941b Formatted Header 70603addf Moved Install / Usage Into Dedicated Files 1659d28fe Moved History Into Dedicated File fb953284b Merge branch 'dev' dbd96b277 add news 898c354e6 v6.04 f8e2ad1e9 add waitpkg detection to Cpu a220fd69a add umwait 64ec053e6 add umonitor 764d54f6f add tpause c68646e8d remove warning of vc 4e8214ca2 Merge branch 'dev' 360f4b673 test badSSE only for 32-bit mode a9fddc454 fix typo of readme 250b5dc39 fix typo of readme c46e92756 Merge branch 'dev' 31b7cd350 v6.03 a2f9ed085 rename isValidXMm to isValidSSE 2bc8fcbd3 add tests to badSSE 6de93fb88 add baseSSE test e2eda384c update doc 71a7b1773 fix condition to throw error for SSE instructions when using XMM16-XMM31 615b665cc sample/memfd shows /proc/self/maps 2861517f2 add memfd sample 507b0285e apt update at first 452c07f77 typedef for no-MmapAllocator 8af6e2026 minimize diff 0af3b5d07 Allocator: take optional name parameter and use it with memfd a67e24505 Merge branch 'dev' b3892c15e Merge pull request #138 from Tachi107/make-cxx-cpp-ld-flags b4eddaced build(make): honour CXXFLAGS, CPPFLAGS and LDFLAGS fbe60e590 test generates only a.asm 0a4a7571d Merge branch 'patch-1' of https://github.com/Tachi107/xbyak into Tachi107-patch-1 55b9f131b build(make): fix clean target in test/Makefile 7aef3ff54 Merge branch 'dev' 82e0deb8a v6.02 4d9906a94 fix condition to throw error for invalid displacements c79311a51 fix test_util.cpp 9b2c175b3 Merge branch 'dev' 06d797e33 Allow parallel feature checks 2a85bba3f Merge branch 'dev' 8d5af80a6 v6.01 df39606fe update doc 08f11817c supprt retf 1abfc3465 support call(mem, T_FAR) fb158f901 support jmp(mem, T_FAR) 9be47ceb2 Merge branch 'lioncash-hlt' into dev 3162eb16f add test of hlt bb55725a9 xbyak: Add hlt opcode function 47cf85fdb fix conflict b29e471ea build(meson): fix CMake Config file include dir 2cc21925f delete the sentences translated into Japanese from COPYRIGHT dca3930de remove tabs cecd204a5 Merge pull request #129 from Tachi107/meson-cmake-config 345de8a54 build(meson): generate CMake package config files e831805cc revert change of the type of Pack::util::operator[] fc9d953e5 util::Pack has not pointer but instance of Reg64 0868c54a9 use _WIN32 instead of _MSC_VER for mingw64 894a1b14a use original uint8_t dfc079ca1 add cstrs of Xbyak::util::Pack for 11 or 12 args 06e8f531f Merge branch 'Tachi107-meson' into dev ef90b6bd2 Makefile updates the version of meson.build 0000938f5 build: add Meson support This allows Meson users to use xbyak more easly, while it also provides a pkg-config file generator, useful to Linux distributions 757e4063f v6.00 b3489f548 add vcvtusi2sh 16d9898ab add vcvtsi2sh cfc03cb8f unify T_66, T_F3, T_F2 flags 02fa7057d add vcvttph2qq 4e72a9dc4 add vcvttsh2usi 05d08e05b add vcvttsh2si 621e6548c add vcvtsh2usi 34abda5c5 extend vcvtps2ph facd622b2 add vcvtw2ph 270af1cb2 add vcvtuw2ph 17dc697cd add vcvttph2w 62f022aea vcvttph2uw 836346bfd add vcvtph2w 300edb37b add vcvtph2uw 737904b5a add vcvtuqq2ph 52b2ebf18 add vcvtqq2ph 6761f1e05 add vcvtpd2ph 03f95b7e3 add vcvtudq2ph 64430b2bb add vcvtps2phx 88e426aa4 add vcvtdq2ph bf28a94a5 add vcvttph2uqq 4c4e665d3 add vcvtph2uqq 2b0099b63 add vcvtph2qq 3ff69a474 add vcvtph2pd 678b52956 add vcvttph2udq 72a5717e4 add vcvttph2dq 524d52bf6 add vcvtph2udq d1cf4db97 add vcvtph2psx fccd2c49b add vcvtph2dq 6530f4099 refactor gen-cvt c51e16156 tweak ed9f6c72a add vcvtsh2si ac1407bd1 add vcvtss2sh 2958a19bb add vcvtsh2ss 10d683303 add vcvtsh2sd 3920c950c add vcvtsd2sh 9a1b73932 add vmovw d86e4882f vmovsh xmm, addr 61f85a204 add vmovsh 847166cef add tests of vgetmant{ph,sh} c11a21cf4 add vgetmantph 55ab361f0 add tests of vgetexp{ph,sh} 8653f4152 add vgetexpsh 697eeb627 add vgetexpph ee920a991 add vfpclasssh 7aed436e6 add vfpclassph 56dca14e4 v5.997 b3b1e4e3c fix vrndscale* to support {sae} 681077eb1 add vrndscalesh 2f14eae85 add vrndscaleph f96870a44 add vreducesh 043c94ae9 add vreduceph 37bf3bb49 add vscalefsh c16f91c59 add vscalefph 61ad45935 add vsqrtsh e00b508d5 add vsqrtph 52765d54f add vrsqrtsh 2ec6a7ab4 add vrsqrtph b1ff7891d add vrcpph, vrcpsh 4f543ca0e add vucomish 412b95f02 add vf{,c}mulcph 9fc53baed add vf{,c}maddcph 0098ce98d add vf{,n}m{sub,add}sh 849dffb10 add vfmadd, vfnmadd, vfnmsub for avx512-fp16 eadb93d0b add vfmsubaddcph 9b8802cba add vfmaddsub{132,213,231}ph ab9481b4c add vcomish cd036ea17 fix vcmpsh b494b321a vcmpsh supports ptr_b eb2d63c83 vcmpph uses T_B16 a480b3dd9 v5.996 16d18b1d4 fix v{add,sub,mul,...}{sd,ss} to support T_rd_sae etc. 66c6ca1ab support vaddsh with T_rd_sae 6333ec099 add m16bcst 84053c8e7 add vcmpph 5df23d263 move FP16 to AVX-512 d5c7336f8 fix disp scaling of v{add,sub,mul,div,max,min}sh ebf29542d add vaddsh test f29689d02 add test of vaddph 2c4b6ac16 add v{add,sub,mul,div,min,max}{ph,sh} edf3c2f6f remove mask for mmm 2c561aad6 add T_MAP5 and T_MAP6 acd360c38 add Cpu::tAVX512_FP16 1554f479c remove unused flags(tSSE4a and tSSE5) of Cpu c313a8758 update CMakeLists.txt version 740b39e24 Merge pull request #122 from abouvier/cmake 10a1e5759 v5.995 0a557a099 rename XBYAK_MEMFD_CREATE to XBYAK_USE_MEMFD 0ad9b9bbe Merge branch 'captain5050-master' into dev c7bb66383 fix for mac and enable it if XBYAK_MEMFD_CREATE is defined c1e6569f1 fix cmake config files ccd4130ec Name mmap pages on Linux with memfd_create 2fb843c32 v5.994 413a66b44 add alias of vcmpXX{ps,pd,ss,sd} for mask register 6f4f76890 v5.993 19043cb9a add test for gather/scatter b5acb1d1c gather test does not generate bad combination of regs 67ec1674a check restriction of gather/scatter regs ea9814f4f check bad reg combination of gather a34850b2d add endbr32 and endbr64 git-subtree-dir: externals/xbyak git-subtree-split: 9357732aa2aa3cf97809027596dfa5c61d1515b2
2022-06-05 13:31:49 +01:00 · 2022-06-05 13:31:49 +01:00 · c926d9f409
commit c926d9f409
parent 2f9dea5cc3
29 changed files with 2936 additions and 861 deletions
--- a/.github/FUNDING.yml
+++ b/.github/FUNDING.yml
@ -0,0 +1 @@
+github: herumi
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@ -7,5 +7,7 @@ jobs:
    runs-on: ubuntu-latest
    steps:
    - uses: actions/checkout@v2
+    - run: sudo apt update
    - run: sudo apt install nasm yasm g++-multilib tcsh
    - run: make test
+    - run: make -C sample CXXFLAGS="-DXBYAK_NO_EXCEPTION"
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 2.6...3.0.2)

-project(xbyak CXX)
+project(xbyak LANGUAGES CXX VERSION 6.06)

 file(GLOB headers xbyak/*.h)

@ -18,17 +18,26 @@ if (DEFINED CMAKE_VERSION AND CMAKE_VERSION VERSION_GREATER_EQUAL 3.0.2)
 	install(
 		TARGETS ${PROJECT_NAME}
 		EXPORT ${PROJECT_NAME}-targets
+		INCLUDES DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/${PROJECT_NAME}
 	)

-	configure_file(
+	include(CMakePackageConfigHelpers)
+	configure_package_config_file(
 		cmake/config.cmake.in
-		${PROJECT_NAME}Config.cmake
-		@ONLY
+		"${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}-config.cmake"
+		INSTALL_DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}
+	)
+	write_basic_package_version_file(
+		"${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}-config-version.cmake"
+		COMPATIBILITY SameMajorVersion
 	)

 	install(
-		FILES ${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}Config.cmake
-		DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}
+		FILES
+			"${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}-config.cmake"
+			"${CMAKE_CURRENT_BINARY_DIR}/${PROJECT_NAME}-config-version.cmake"
+		DESTINATION
+			${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}
 	)

 	install(
--- a/20
+++ b/20
@ -25,23 +25,3 @@ INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
 THE POSSIBILITY OF SUCH DAMAGE.
-----------------------------------------------------------------------------
-ソースコード形式かバイナリ形式か、変更するかしないかを問わず、以下の条件を満た
-す場合に限り、再頒布および使用が許可されます。
-
-ソースコードを再頒布する場合、上記の著作権表示、本条件一覧、および下記免責条項
-を含めること。
-バイナリ形式で再頒布する場合、頒布物に付属のドキュメント等の資料に、上記の著作
-権表示、本条件一覧、および下記免責条項を含めること。
-書面による特別の許可なしに、本ソフトウェアから派生した製品の宣伝または販売促進
-に、著作権者の名前またはコントリビューターの名前を使用してはならない。
-本ソフトウェアは、著作権者およびコントリビューターによって「現状のまま」提供さ
-れており、明示黙示を問わず、商業的な使用可能性、および特定の目的に対する適合性
-に関する暗黙の保証も含め、またそれに限定されない、いかなる保証もありません。
-著作権者もコントリビューターも、事由のいかんを問わず、 損害発生の原因いかんを
-問わず、かつ責任の根拠が契約であるか厳格責任であるか（過失その他の）不法行為で
-あるかを問わず、仮にそのような損害が発生する可能性を知らされていたとしても、
-本ソフトウェアの使用によって発生した（代替品または代用サービスの調達、使用の
-喪失、データの喪失、利益の喪失、業務の中断も含め、またそれに限定されない）直接
-損害、間接損害、偶発的な損害、特別損害、懲罰的損害、または結果損害について、
-一切責任を負わないものとします。
--- a/cmake/config.cmake.in
+++ b/cmake/config.cmake.in
@ -1 +1,3 @@
+@PACKAGE_INIT@
+
 include("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@-targets.cmake")
--- a/cmake/meson-config.cmake.in
+++ b/cmake/meson-config.cmake.in
@ -0,0 +1,8 @@
+@PACKAGE_INIT@
+
+if(NOT TARGET @TARGET_NAME@)
+	add_library(@TARGET_NAME@ INTERFACE IMPORTED)
+	set_target_properties(@TARGET_NAME@ PROPERTIES
+		INTERFACE_INCLUDE_DIRECTORIES "@ABSOLUTE_INCLUDE_DIR@"
+	)
+endif()
--- a/doc/changelog.md
+++ b/doc/changelog.md
@ -0,0 +1,180 @@
+# History
+
+* 2022/Jun/01 ver 6.06 refactor Cpu::Type class and improve MmapAllocator when XBYAK_USE_MEMFD is defined.
+* 2022/Mar/20 ver 6.052 add Cpu::operator==()
+* 2022/Mar/13 ver 6.051 fix compile error when XBYAK_NO_EXCEPTION is defined
+* 2022/Mar/12 ver 6.05 add movdiri, movdir64b, clwb, cldemote
+* 2022/Apr/22 ver 6.041 consider Android and mingw
+* 2022/Apr/05 ver 6.04 add tpause, umonitor, umwait
+* 2022/Mar/08 ver 6.03 MmapAllocator supports memfd with user-defined strings.
+* 2022/Jan/28 ver 6.02 strict check the range of 32-bit dispacement
+* 2021/Dec/14 ver 6.01 support T_FAR jump/call and retf
+* 2021/Sep/14 ver 6.00 fully support AVX512-FP16
+* 2021/Sep/09 ver 5.997 fix vrndscale* to support {sae}
+* 2021/Sep/03 ver 5.996 fix v{add,sub,mul,div,max,min}{sd,ss} to support T_rd_sae.
+* 2021/Aug/15 ver 5.995 add a label to /proc/self/maps if XBYAK_USE_MEMFD is defined on Linux
+* 2021/Jun/17 ver 5.994 add alias of vcmpXX{ps,pd,ss,sd} with mask register
+* 2021/Jun/06 ver 5.993 strict check of gather/scatter register combination
+* 2021/May/09 ver 5.992 support endbr32 and endbr64
+* 2020/Nov/16 ver 5.991 disable constexpr for gcc-5 with -std=c++-14
+* 2020/Oct/19 ver 5.99 support VNNI instructions(Thanks to akharito)
+* 2020/Oct/17 ver 5.98 support the form of [scale * reg]
+* 2020/Sep/08 ver 5.97 replace uint32 with uint32_t etc.
+* 2020/Aug/28 ver 5.95 some constructors of register classes support constexpr if C++14 or later
+* 2020/Aug/04 ver 5.941 `CodeGenerator::reset()` calls `ClearError()`.
+* 2020/Jul/28 ver 5.94 remove #include <winsock2.h> (only windows)
+* 2020/Jul/21 ver 5.93 support exception-less mode
+* 2020/Jun/30 ver 5.92 support Intel AMX instruction set (Thanks to nshustrov)
+* 2020/Jun/22 ver 5.913 fix mov(r64, imm64) on 32-bit env with XBYAK64
+* 2020/Jun/19 ver 5.912 define MAP_JIT on macOS regardless of Xcode version (Thanks to rsdubtso)
+* 2020/May/10 ver 5.911 XBYAK_USE_MMAP_ALLOCATOR is defined unless XBYAK_DONT_USE_MMAP_ALLOCATOR is defined.
+* 2020/Apr/20 ver 5.91 accept mask register k0 (it means no mask)
+* 2020/Apr/09 ver 5.90 kmov{b,d,w,q} throws exception for an unsupported register
+* 2020/Feb/26 ver 5.891 fix typo of type
+* 2020/Jan/03 ver 5.89 fix error of vfpclasspd
+* 2019/Dec/20 ver 5.88 fix compile error on Windows
+* 2019/Dec/19 ver 5.87 add setDefaultJmpNEAR(), which deals with `jmp` of an undefined label as T_NEAR if no type is specified.
+* 2019/Dec/13 ver 5.86 [changed] revert to the behavior before v5.84 if -fno-operator-names is defined (and() is available)
+* 2019/Dec/07 ver 5.85 append MAP_JIT flag to mmap for macOS mojave or later
+* 2019/Nov/29 ver 5.84 [changed] XBYAK_NO_OP_NAMES is defined unless XBYAK_USE_OP_NAMES is defined
+* 2019/Oct/12 ver 5.83 exit(1) was removed
+* 2019/Sep/23 ver 5.82 support monitorx, mwaitx, clzero (thanks to @MagurosanTeam)
+* 2019/Sep/14 ver 5.81 support some generic mnemonics.
+* 2019/Aug/01 ver 5.802 fix detection of AVX512_BF16 (thanks to vpirogov)
+* 2019/May/27 support vp2intersectd, vp2intersectq (not tested)
+* 2019/May/26 ver 5.80 support vcvtne2ps2bf16, vcvtneps2bf16, vdpbf16ps
+* 2019/Apr/27 ver 5.79 vcmppd/vcmpps supports ptr_b(thanks to jkopinsky)
+* 2019/Apr/15 ver 5.78 rewrite Reg::changeBit() (thanks to MerryMage)
+* 2019/Mar/06 ver 5.77 fix number of cores that share LLC cache by densamoilov
+* 2019/Jan/17 ver 5.76 add Cpu::getNumCores() by shelleygoel
+* 2018/Oct/31 ver 5.751 recover Xbyak::CastTo for compatibility
+* 2018/Oct/29 ver 5.75 unlink LabelManager from Label when msg is destroyed
+* 2018/Oct/21 ver 5.74 support RegRip +/- int. Xbyak::CastTo is removed
+* 2018/Oct/15 util::AddressFrame uses push/pop instead of mov
+* 2018/Sep/19 ver 5.73 fix evex encoding of vpslld, vpslldq, vpsllw, etc for (reg, mem, imm8)
+* 2018/Sep/19 ver 5.72 fix the encoding of vinsertps for disp8N(Thanks to petercaday)
+* 2018/Sep/04 ver 5.71 L() returns a new label instance
+* 2018/Aug/27 ver 5.70 support setProtectMode() and DontUseProtect for read/exec setting
+* 2018/Aug/24 ver 5.68 fix wrong VSIB encoding with vector index >= 16(thanks to petercaday)
+* 2018/Aug/14 ver 5.67 remove mutable in Address ; fix setCacheHierarchy for cloud vm
+* 2018/Jul/26 ver 5.661 support mingw64
+* 2018/Jul/24 ver 5.66 add CodeArray::PROTECT_RE to mode of protect()
+* 2018/Jun/26 ver 5.65 fix push(qword [mem])
+* 2018/Mar/07 ver 5.64 fix zero division in Cpu() on some cpu
+* 2018/Feb/14 ver 5.63 fix Cpu::setCacheHierarchy() and fix EvexModifierZero for clang<3.9(thanks to mgouicem)
+* 2018/Feb/13 ver 5.62 Cpu::setCacheHierarchy() by mgouicem and rsdubtso
+* 2018/Feb/07 ver 5.61 vmov* supports mem{k}{z}(I forgot it)
+* 2018/Jan/24 ver 5.601 add xword, yword, etc. into Xbyak::util namespace
+* 2018/Jan/05 ver 5.60 support AVX-512 for Ice lake(319433-030.pdf)
+* 2017/Aug/22 ver 5.53 fix mpx encoding, add bnd() prefix
+* 2017/Aug/18 ver 5.52 fix align (thanks to MerryMage)
+* 2017/Aug/17 ver 5.51 add multi-byte nop and align() uses it(thanks to inolen)
+* 2017/Aug/08 ver 5.50 add mpx(thanks to magurosan)
+* 2017/Aug/08 ver 5.45 add sha(thanks to magurosan)
+* 2017/Aug/08 ver 5.44 add prefetchw(thanks to rsdubtso)
+* 2017/Jul/12 ver 5.432 reduce warnings of PVS studio
+* 2017/Jul/09 ver 5.431 fix hasRex() (no affect) (thanks to drillsar)
+* 2017/May/14 ver 5.43 fix CodeGenerator::resetSize() (thanks to gibbed)
+* 2017/May/13 ver 5.42 add movs{b,w,d,q}
+* 2017/Jan/26 ver 5.41 add prefetchwt1 and support for scale == 0(thanks to rsdubtso)
+* 2016/Dec/14 ver 5.40 add Label::getAddress() method to get the pointer specified by the label
+* 2016/Dec/09 ver 5.34 fix handling of negative offsets when encoding disp8N(thanks to rsdubtso)
+* 2016/Dec/08 ver 5.33 fix encoding of vpbroadcast{b,w,d,q}, vpinsr{b,w}, vpextr{b,w} for disp8N
+* 2016/Dec/01 ver 5.32 rename __xgetbv() to _xgetbv() to support clang for Visual Studio(thanks to freiro)
+* 2016/Nov/27 ver 5.31 rename AVX512_4VNNI to AVX512_4VNNIW
+* 2016/Nov/27 ver 5.30 add AVX512_4VNNI, AVX512_4FMAPS instructions(thanks to rsdubtso)
+* 2016/Nov/26 ver 5.20 add detection of AVX512_4VNNI and AVX512_4FMAPS(thanks to rsdubtso)
+* 2016/Nov/20 ver 5.11 lost vptest for ymm(thanks to gregory38)
+* 2016/Nov/20 ver 5.10 add addressing [rip+&var]
+* 2016/Sep/29 ver 5.03 fix detection ERR_INVALID_OPMASK_WITH_MEMORY(thanks to PVS-Studio)
+* 2016/Aug/15 ver 5.02 xbyak does not include xbyak_bin2hex.h
+* 2016/Aug/15 ver 5.011 fix detection of version of gcc 5.4
+* 2016/Aug/03 ver 5.01 disable omitted operand
+* 2016/Jun/24 ver 5.00 support avx-512 instruction set
+* 2016/Jun/13 avx-512 add mask instructions
+* 2016/May/05 ver 4.91 add detection of AVX-512 to Xbyak::util::Cpu
+* 2016/Mar/14 ver 4.901 comment to ready() function(thanks to skmp)
+* 2016/Feb/04 ver 4.90 add jcc(const void *addr);
+* 2016/Jan/30 ver 4.89 vpblendvb supports ymm reg(thanks to John Funnell)
+* 2016/Jan/24 ver 4.88 lea, cmov supports 16-bit register(thanks to whyisthisfieldhere)
+* 2015/Oct/05 ver 4.87 support segment selectors
+* 2015/Aug/18 ver 4.86 fix [rip + label] addressing with immediate value(thanks to whyisthisfieldhere)
+* 2015/Aug/10 ver 4.85 Address::operator==() is not correct(thanks to inolen)
+* 2015/Jun/22 ver 4.84 call() support variadic template if available(thanks to randomstuff)
+* 2015/Jun/16 ver 4.83 support movbe(thanks to benvanik)
+* 2015/May/24 ver 4.82 support detection of F16C
+* 2015/Apr/25 ver 4.81 fix the condition to throw exception for setSize(thanks to whyisthisfieldhere)
+* 2015/Apr/22 ver 4.80 rip supports label(thanks to whyisthisfieldhere)
+* 2015/Jar/28 ver 4.71 support adcx, adox, cmpxchg, rdseed, stac
+* 2014/Oct/14 ver 4.70 support MmapAllocator
+* 2014/Jun/13 ver 4.62 disable warning of VC2014
+* 2014/May/30 ver 4.61 support bt, bts, btr, btc
+* 2014/May/28 ver 4.60 support vcvtph2ps, vcvtps2ph
+* 2014/Apr/11 ver 4.52 add detection of rdrand
+* 2014/Mar/25 ver 4.51 remove state information of unreferenced labels
+* 2014/Mar/16 ver 4.50 support new Label
+* 2014/Mar/05 ver 4.40 fix wrong detection of BMI/enhanced rep on VirtualBox
+* 2013/Dec/03 ver 4.30 support Reg::cvt8(), cvt16(), cvt32(), cvt64()
+* 2013/Oct/16 ver 4.21 label support std::string
+* 2013/Jul/30 ver 4.20 [break backward compatibility] split Reg32e class into RegExp(base+index*scale+disp) and Reg32e(means Reg32 or Reg64)
+* 2013/Jul/04 ver 4.10 [break backward compatibility] change the type of Xbyak::Error from enum to a class
+* 2013/Jun/21 ver 4.02 add putL(LABEL) function to put the address of the label
+* 2013/Jun/21 ver 4.01 vpsllw, vpslld, vpsllq, vpsraw, vpsrad, vpsrlw, vpsrld, vpsrlq support (ymm, ymm, xmm). support vpbroadcastb, vpbroadcastw, vpbroadcastd, vpbroadcastq(thanks to Gabest).
+* 2013/May/30 ver 4.00 support AVX2, VEX-encoded GPR-instructions
+* 2013/Mar/27 ver 3.80 support mov(reg, "label");
+* 2013/Mar/13 ver 3.76 add cqo(), jcxz(), jecxz(), jrcxz()
+* 2013/Jan/15 ver 3.75 add setSize() to modify generated code
+* 2013/Jan/12 ver 3.74 add CodeGenerator::reset() ; add Allocator::useProtect()
+* 2013/Jan/06 ver 3.73 use unordered_map if possible
+* 2012/Dec/04 ver 3.72 eax, ebx, ... are member variables of CodeGenerator(revert), Xbyak::util::eax, ... are static const.
+* 2012/Nov/17 ver 3.71 and_(), or_(), xor_(), not_() are available if XBYAK_NO_OP_NAMES is not defined.
+* 2012/Nov/17 change eax, ebx, ptr and so on in CodeGenerator as static member and alias of them are defined in Xbyak::util.
+* 2012/Nov/09 ver 3.70 XBYAK_NO_OP_NAMES macro is added to use and_() instead of and() (thanks to Mattias)
+* 2012/Nov/01 ver 3.62 add fwait/fnwait/finit/fninit
+* 2012/Nov/01 ver 3.61 add fldcw/fstcw
+* 2012/May/03 ver 3.60 change interface of Allocator
+* 2012/Mar/23 ver 3.51 fix userPtr mode
+* 2012/Mar/19 ver 3.50 support AutoGrow mode
+* 2011/Nov/09 ver 3.05 fix bit property of rip addresing / support movsxd
+* 2011/Aug/15 ver 3.04 fix dealing with imm8 such as add(dword [ebp-8], 0xda); (thanks to lolcat)
+* 2011/Jun/16 ver 3.03 fix __GNUC_PREREQ macro for Mac gcc(thanks to t_teruya)
+* 2011/Apr/28 ver 3.02 do not use xgetbv on Mac gcc
+* 2011/May/24 ver 3.01 fix typo of OSXSAVE
+* 2011/May/23 ver 3.00 add vcmpeqps and so on
+* 2011/Feb/16 ver 2.994 beta add vmovq for 32-bit mode(I forgot it)
+* 2011/Feb/16 ver 2.993 beta remove cvtReg to avoid thread unsafe
+* 2011/Feb/10 ver 2.992 beta support one argument syntax for fadd like nasm
+* 2011/Feb/07 ver 2.991 beta fix pextrw reg, xmm, imm(Thanks to Gabest)
+* 2011/Feb/04 ver 2.99 beta support AVX
+* 2010/Dec/08 ver 2.31 fix ptr [rip + 32bit offset], support rdtscp
+* 2010/Oct/19 ver 2.30 support pclmulqdq, aesdec, aesdeclast, aesenc, aesenclast, aesimc, aeskeygenassist
+* 2010/Jun/07 ver 2.29 fix call(<label>)
+* 2010/Jun/17 ver 2.28 move some member functions to public
+* 2010/Jun/01 ver 2.27 support encoding of mov(reg64, imm) like yasm(not nasm)
+* 2010/May/24 ver 2.26 fix sub(rsp, 1000)
+* 2010/Apr/26 ver 2.25 add jc/jnc(I forgot to implement them...)
+* 2010/Apr/16 ver 2.24 change the prototype of rewrite() method
+* 2010/Apr/15 ver 2.23 fix align() and xbyak_util.h for Mac
+* 2010/Feb/16 ver 2.22 fix inLocalLabel()/outLocalLabel()
+* 2009/Dec/09 ver 2.21 support cygwin(gcc 4.3.2)
+* 2009/Nov/28 support a part of FPU
+* 2009/Jun/25 fix mov(qword[rax], imm); (thanks to Martin)
+* 2009/Mar/10 fix redundant REX.W prefix on jmp/call reg64
+* 2009/Feb/24 add movq reg64, mmx/xmm; movq mmx/xmm, reg64
+* 2009/Feb/13 movd(xmm7, dword[eax]) drops 0x66 prefix (thanks to Gabest)
+* 2008/Dec/30 fix call in short relative address(thanks to kato san)
+* 2008/Sep/18 support @@, @f, @b and localization of label(thanks to nobu-q san)
+* 2008/Sep/18 support (ptr[rip + 32bit offset]) (thanks to Dango-Chu san)
+* 2008/Jun/03 fix align(). mov(ptr[eax],1) throws ERR_MEM_SIZE_IS_NOT_SPECIFIED.
+* 2008/Jun/02 support memory interface allocated by user
+* 2008/May/26 fix protect() to avoid invalid setting(thanks to shinichiro_h san)
+* 2008/Apr/30 add cmpxchg16b, cdqe
+* 2008/Apr/29 support x64
+* 2008/Apr/14 code refactoring
+* 2008/Mar/12 add bsr/bsf
+* 2008/Feb/14 fix output of sub eax, 1234 (thanks to Robert)
+* 2007/Nov/5  support lock, xadd, xchg
+* 2007/Nov/2  support SSSE3/SSE4 (thanks to Dango-Chu san)
+* 2007/Feb/4  fix the bug that exception doesn't occur under the condition which the offset of jmp mnemonic without T_NEAR is over 127.
+* 2007/Jan/21 fix the bug to create address like [disp] select smaller representation for mov (eax|ax|al, [disp])
+* 2007/Jan/4  first version
--- a/doc/install.md
+++ b/doc/install.md
@ -0,0 +1,14 @@
+# Install
+
+The following files are necessary. Please add the path to your compile directory.
+
+* xbyak.h
+* xbyak_mnemonic.h
+* xbyak_util.h
+
+Linux:
+```
+make install
+```
+
+These files are copied into `/usr/local/include/xbyak`.
--- a/doc/usage.md
+++ b/doc/usage.md
@ -0,0 +1,409 @@
+# Usage
+
+Inherit `Xbyak::CodeGenerator` class and make the class method.
+```
+#include <xbyak/xbyak.h>
+
+struct Code : Xbyak::CodeGenerator {
+    Code(int x)
+    {
+        mov(eax, x);
+        ret();
+    }
+};
+```
+Or you can pass the instance of CodeGenerator without inheriting.
+```
+void genCode(Xbyak::CodeGenerator& code, int x) {
+    using namespace Xbyak::util;
+    code.mov(eax, x);
+    code.ret();
+}
+```
+
+Make an instance of the class and get the function
+pointer by calling `getCode()` and call it.
+```
+Code c(5);
+int (*f)() = c.getCode<int (*)()>();
+printf("ret=%d\n", f()); // ret = 5
+```
+
+## Syntax
+Similar to MASM/NASM syntax with parentheses.
+
+```
+NASM              Xbyak
+mov eax, ebx  --> mov(eax, ebx);
+inc ecx           inc(ecx);
+ret           --> ret();
+```
+
+## Addressing
+Use `qword`, `dword`, `word` and `byte` if it is necessary to specify the size of memory,
+otherwise use `ptr`.
+
+```
+(ptr|qword|dword|word|byte) [base + index * (1|2|4|8) + displacement]
+                            [rip + 32bit disp] ; x64 only
+
+NASM                   Xbyak
+mov eax, [ebx+ecx] --> mov(eax, ptr [ebx+ecx]);
+mov al, [ebx+ecx]  --> mov(al, ptr [ebx + ecx]);
+test byte [esp], 4 --> test(byte [esp], 4);
+inc qword [rax]    --> inc(qword [rax]);
+```
+**Note**: `qword`, ... are member variables, then don't use `dword` as unsigned int type.
+
+### How to use Selector (Segment Register)
+```
+mov eax, [fs:eax] --> putSeg(fs);
+                      mov(eax, ptr [eax]);
+mov ax, cs        --> mov(ax, cs);
+```
+**Note**: Segment class is not derived from `Operand`.
+
+## AVX
+
+```
+vaddps(xmm1, xmm2, xmm3); // xmm1 <- xmm2 + xmm3
+vaddps(xmm2, xmm3, ptr [rax]); // use ptr to access memory
+vgatherdpd(xmm1, ptr [ebp + 256 + xmm2*4], xmm3);
+```
+
+**Note**:
+If `XBYAK_ENABLE_OMITTED_OPERAND` is defined, then you can use two operand version for backward compatibility.
+But the newer version will not support it.
+```
+vaddps(xmm2, xmm3); // xmm2 <- xmm2 + xmm3
+```
+
+## AVX-512
+
+```
+vaddpd zmm2, zmm5, zmm30                --> vaddpd(zmm2, zmm5, zmm30);
+vaddpd xmm30, xmm20, [rax]              --> vaddpd(xmm30, xmm20, ptr [rax]);
+vaddps xmm30, xmm20, [rax]              --> vaddps(xmm30, xmm20, ptr [rax]);
+vaddpd zmm2{k5}, zmm4, zmm2             --> vaddpd(zmm2 | k5, zmm4, zmm2);
+vaddpd zmm2{k5}{z}, zmm4, zmm2          --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2);
+vaddpd zmm2{k5}{z}, zmm4, zmm2,{rd-sae} --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2 | T_rd_sae);
+                                            vaddpd(zmm2 | k5 | T_z | T_rd_sae, zmm4, zmm2); // the position of `|` is arbitrary.
+vcmppd k4{k3}, zmm1, zmm2, {sae}, 5     --> vcmppd(k4 | k3, zmm1, zmm2 | T_sae, 5);
+
+vaddpd xmm1, xmm2, [rax+256]            --> vaddpd(xmm1, xmm2, ptr [rax+256]);
+vaddpd xmm1, xmm2, [rax+256]{1to2}      --> vaddpd(xmm1, xmm2, ptr_b [rax+256]);
+vaddpd ymm1, ymm2, [rax+256]{1to4}      --> vaddpd(ymm1, ymm2, ptr_b [rax+256]);
+vaddpd zmm1, zmm2, [rax+256]{1to8}      --> vaddpd(zmm1, zmm2, ptr_b [rax+256]);
+vaddps zmm1, zmm2, [rax+rcx*8+8]{1to16} --> vaddps(zmm1, zmm2, ptr_b [rax+rcx*8+8]);
+vmovsd [rax]{k1}, xmm4                  --> vmovsd(ptr [rax] | k1, xmm4);
+
+vcvtpd2dq xmm16, oword [eax+33]         --> vcvtpd2dq(xmm16, xword [eax+33]); // use xword for m128 instead of oword
+                                            vcvtpd2dq(xmm16, ptr [eax+33]); // default xword
+vcvtpd2dq xmm21, [eax+32]{1to2}         --> vcvtpd2dq(xmm21, ptr_b [eax+32]);
+vcvtpd2dq xmm0, yword [eax+33]          --> vcvtpd2dq(xmm0, yword [eax+33]); // use yword for m256
+vcvtpd2dq xmm19, [eax+32]{1to4}         --> vcvtpd2dq(xmm19, yword_b [eax+32]); // use yword_b to broadcast
+
+vfpclassps k5{k3}, zword [rax+64], 5    --> vfpclassps(k5|k3, zword [rax+64], 5); // specify m512
+vfpclasspd k5{k3}, [rax+64]{1to2}, 5    --> vfpclasspd(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 128-bit
+vfpclassps k5{k3}, [rax+64]{1to4}, 5    --> vfpclassps(k5|k3, yword_b [rax+64], 5); // broadcast 64-bit to 256-bit
+
+vpdpbusd(xm0, xm1, xm2); // default encoding is EVEX
+vpdpbusd(xm0, xm1, xm2, EvexEncoding); // same as the above
+vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX encoding
+```
+### Remark
+* `k1`, ..., `k7` are opmask registers.
+  - `k0` is dealt as no mask.
+  - e.g. `vmovaps(zmm0|k0, ptr[rax]);` and `vmovaps(zmm0|T_z, ptr[rax]);` are same to `vmovaps(zmm0, ptr[rax]);`.
+* use `| T_z`, `| T_sae`, `| T_rn_sae`, `| T_rd_sae`, `| T_ru_sae`, `| T_rz_sae` instead of `,{z}`, `,{sae}`, `,{rn-sae}`, `,{rd-sae}`, `,{ru-sae}`, `,{rz-sae}` respectively.
+* `k4 | k3` is different from `k3 | k4`.
+* use `ptr_b` for broadcast `{1toX}`. X is automatically determined.
+* specify `xword`/`yword`/`zword(_b)` for m128/m256/m512 if necessary.
+
+## Label
+Two kinds of Label are supported. (String literal and Label class).
+
+### String literal
+```
+L("L1");
+  jmp("L1");
+
+  jmp("L2");
+  ...
+  a few mnemonics (8-bit displacement jmp)
+  ...
+L("L2");
+
+  jmp("L3", T_NEAR);
+  ...
+  a lot of mnemonics (32-bit displacement jmp)
+  ...
+L("L3");
+```
+
+* Call `hasUndefinedLabel()` to verify your code has no undefined label.
+* you can use a label for immediate value of mov like as `mov(eax, "L2")`.
+
+### Support `@@`, `@f`, `@b` like MASM
+
+```
+L("@@"); // <A>
+  jmp("@b"); // jmp to <A>
+  jmp("@f"); // jmp to <B>
+L("@@"); // <B>
+  jmp("@b"); // jmp to <B>
+  mov(eax, "@b");
+  jmp(eax); // jmp to <B>
+```
+
+### Local label
+
+Label symbols beginning with a period between `inLocalLabel()` and `outLocalLabel()`
+are treated as a local label.
+`inLocalLabel()` and `outLocalLabel()` can be nested.
+
+```
+void func1()
+{
+    inLocalLabel();
+  L(".lp"); // <A> ; local label
+    ...
+    jmp(".lp"); // jmp to <A>
+  L("aaa"); // global label <C>
+    outLocalLabel();
+
+    inLocalLabel();
+  L(".lp"); // <B> ; local label
+    func1();
+    jmp(".lp"); // jmp to <B>
+    inLocalLabel();
+    jmp("aaa"); // jmp to <C>
+}
+```
+
+### short and long jump
+Xbyak deals with jump mnemonics of an undefined label as short jump if no type is specified.
+So if the size between jmp and label is larger than 127 byte, then xbyak will cause an error.
+
+```
+jmp("short-jmp"); // short jmp
+// small code
+L("short-jmp");
+
+jmp("long-jmp");
+// long code
+L("long-jmp"); // throw exception
+```
+Then specify T_NEAR for jmp.
+```
+jmp("long-jmp", T_NEAR); // long jmp
+// long code
+L("long-jmp");
+```
+Or call `setDefaultJmpNEAR(true);` once, then the default type is set to T_NEAR.
+```
+jmp("long-jmp"); // long jmp
+// long code
+L("long-jmp");
+```
+
+### Label class
+
+`L()` and `jxx()` support Label class.
+
+```
+  Xbyak::Label label1, label2;
+L(label1);
+  ...
+  jmp(label1);
+  ...
+  jmp(label2);
+  ...
+L(label2);
+```
+
+Use `putL` for jmp table
+```
+    Label labelTbl, L0, L1, L2;
+    mov(rax, labelTbl);
+    // rdx is an index of jump table
+    jmp(ptr [rax + rdx * sizeof(void*)]);
+L(labelTbl);
+    putL(L0);
+    putL(L1);
+    putL(L2);
+L(L0);
+    ....
+L(L1);
+    ....
+```
+
+`assignL(dstLabel, srcLabel)` binds dstLabel with srcLabel.
+
+```
+  Label label2;
+  Label label1 = L(); // make label1 ; same to Label label1; L(label1);
+  ...
+  jmp(label2); // label2 is not determined here
+  ...
+  assignL(label2, label1); // label2 <- label1
+```
+The `jmp` in the above code jumps to label1 assigned by `assignL`.
+
+**Note**:
+* srcLabel must be used in `L()`.
+* dstLabel must not be used in `L()`.
+
+`Label::getAddress()` returns the address specified by the label instance and 0 if not specified.
+```
+// not AutoGrow mode
+Label  label;
+assert(label.getAddress() == 0);
+L(label);
+assert(label.getAddress() == getCurr());
+```
+
+### Rip ; relative addressing
+```
+Label label;
+mov(eax, ptr [rip + label]); // eax = 4
+...
+
+L(label);
+dd(4);
+```
+```
+int x;
+...
+  mov(eax, ptr[rip + &x]); // throw exception if the difference between &x and current position is larger than 2GiB
+```
+
+## Far jump
+
+Use `word|dword|qword` instead of `ptr` to specify the address size.
+
+### 32 bit mode
+```
+jmp(word[eax], T_FAR);  // jmp m16:16(FF /5)
+jmp(dword[eax], T_FAR); // jmp m16:32(FF /5)
+```
+
+### 64 bit mode
+```
+jmp(word[rax], T_FAR);  // jmp m16:16(FF /5)
+jmp(dword[rax], T_FAR); // jmp m16:32(FF /5)
+jmp(qword[rax], T_FAR); // jmp m16:64(REX.W FF /5)
+```
+The same applies to `call`.
+
+## Code size
+The default max code size is 4096 bytes.
+Specify the size in constructor of `CodeGenerator()` if necessary.
+
+```
+class Quantize : public Xbyak::CodeGenerator {
+public:
+  Quantize()
+    : CodeGenerator(8192)
+  {
+  }
+  ...
+};
+```
+
+## User allocated memory
+
+You can make jit code on prepared memory.
+
+Call `setProtectModeRE` yourself to change memory mode if using the prepared memory.
+
+```
+uint8_t alignas(4096) buf[8192]; // C++11 or later
+
+struct Code : Xbyak::CodeGenerator {
+    Code() : Xbyak::CodeGenerator(sizeof(buf), buf)
+    {
+        mov(rax, 123);
+        ret();
+    }
+};
+
+int main()
+{
+    Code c;
+    c.setProtectModeRE(); // set memory to Read/Exec
+    printf("%d\n", c.getCode<int(*)()>()());
+}
+```
+
+**Note**: See [../sample/test0.cpp](../sample/test0.cpp).
+
+### AutoGrow
+
+The memory region for jit is automatically extended if necessary when `AutoGrow` is specified in a constructor of `CodeGenerator`.
+
+Call `ready()` or `readyRE()` before calling `getCode()` to fix jump address.
+```
+struct Code : Xbyak::CodeGenerator {
+  Code()
+    : Xbyak::CodeGenerator(<default memory size>, Xbyak::AutoGrow)
+  {
+     ...
+  }
+};
+Code c;
+// generate code for jit
+c.ready(); // mode = Read/Write/Exec
+```
+
+**Note**:
+* Don't use the address returned by `getCurr()` before calling `ready()` because it may be invalid address.
+
+### Read/Exec mode
+Xbyak set Read/Write/Exec mode to memory to run jit code.
+If you want to use Read/Exec mode for security, then specify `DontSetProtectRWE` for `CodeGenerator` and
+call `setProtectModeRE()` after generating jit code.
+
+```
+struct Code : Xbyak::CodeGenerator {
+    Code()
+        : Xbyak::CodeGenerator(4096, Xbyak::DontSetProtectRWE)
+    {
+        mov(eax, 123);
+        ret();
+    }
+};
+
+Code c;
+c.setProtectModeRE();
+...
+
+```
+Call `readyRE()` instead of `ready()` when using `AutoGrow` mode.
+See [protect-re.cpp](../sample/protect-re.cpp).
+
+## Exception-less mode
+If `XBYAK_NO_EXCEPTION` is defined, then gcc/clang can compile xbyak with `-fno-exceptions`.
+In stead of throwing an exception, `Xbyak::GetError()` returns non-zero value (e.g. `ERR_BAD_ADDRESSING`) if there is something wrong.
+The status will not be changed automatically, then you should reset it by `Xbyak::ClearError()`.
+`CodeGenerator::reset()` calls `ClearError()`.
+
+## Macro
+
+* **XBYAK32** is defined on 32bit.
+* **XBYAK64** is defined on 64bit.
+* **XBYAK64_WIN** is defined on 64bit Windows(VC).
+* **XBYAK64_GCC** is defined on 64bit gcc, cygwin.
+* define **XBYAK_USE_OP_NAMES** on gcc with `-fno-operator-names` if you want to use `and()`, ....
+* define **XBYAK_ENABLE_OMITTED_OPERAND** if you use omitted destination such as `vaddps(xmm2, xmm3);`(deprecated in the future).
+* define **XBYAK_UNDEF_JNL** if Bessel function jnl is defined as macro.
+* define **XBYAK_NO_EXCEPTION** for a compiler option `-fno-exceptions`.
+* define **XBYAK_USE_MEMFD** on Linux then /proc/self/maps shows the area used by xbyak.
+* define **XBYAK_OLD_DISP_CHECK** if the old disp check is necessary (deprecated in the future).
+
+## Sample
+
+* [test0.cpp](../sample/test0.cpp) ; tiny sample (x86, x64)
+* [quantize.cpp](../sample/quantize.cpp) ; JIT optimized quantization by fast division (x86 only)
+* [calc.cpp](../sample/calc.cpp) ; assemble and estimate a given polynomial (x86, x64)
+* [bf.cpp](../sample/bf.cpp) ; JIT brainfuck (x86, x64)
--- a/gen/Makefile
+++ b/gen/Makefile
@ -1,7 +1,7 @@
 TARGET=../xbyak/xbyak_mnemonic.h
 BIN=sortline gen_code gen_avx512
-CFLAGS=-I../ -O2 -DXBYAK_NO_OP_NAMES -Wall -Wextra -Wno-missing-field-initializers
-all: $(TARGET)
+CFLAGS=-I../ -O2 -DXBYAK_NO_OP_NAMES -Wall -Wextra -Wno-missing-field-initializers $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS)
+all: $(TARGET) ../CMakeLists.txt ../meson.build ../readme.md ../readme.txt
 sortline: sortline.cpp
 	$(CXX) $(CFLAGS) $< -o $@
 gen_code: gen_code.cpp ../xbyak/xbyak.h avx_type.hpp
@ -22,5 +22,18 @@ $(TARGET): $(BIN)
 	echo "#endif" >> $@
 	echo "#endif" >> $@

+VER=$(shell head -n 1 ../xbyak/xbyak_mnemonic.h|grep -o "[0-9.]*")
+../CMakeLists.txt: $(TARGET)
+	sed -i -e 's/CXX VERSION [0-9.]*/CXX VERSION $(VER)/' $@
+
+../meson.build: $(TARGET)
+	sed -i -e "s/version: '[0-9.]*',/version: '$(VER)',/" $@
+
+../readme.md: $(TARGET)
+	sed -l 2 -i -e "s/# Xbyak [0-9.]*/# Xbyak $(VER)/" $@
+
+../readme.txt: $(TARGET)
+	sed -l 2 -i -e "s/Xbyak [0-9.]*/Xbyak $(VER)/" $@
+
 clean:
 	$(RM) $(BIN) $(TARGET)
--- a/gen/avx_type.hpp
+++ b/gen/avx_type.hpp
@ -12,9 +12,10 @@
 		//
 		T_N_VL = 1 << 3, // N * (1, 2, 4) for VL
 		T_DUP = 1 << 4, // N = (8, 32, 64)
-		T_66 = 1 << 5,
-		T_F3 = 1 << 6,
-		T_F2 = 1 << 7,
+		T_66 = 1 << 5, // pp = 1
+		T_F3 = 1 << 6, // pp = 2
+		T_F2 = T_66 | T_F3, // pp = 3
+		T_ER_R = 1 << 7, // reg{er}
 		T_0F = 1 << 8,
 		T_0F38 = 1 << 9,
 		T_0F3A = 1 << 10,
@ -35,11 +36,18 @@
 		T_MUST_EVEX = 1 << 25, // contains T_EVEX
 		T_B32 = 1 << 26, // m32bcst
 		T_B64 = 1 << 27, // m64bcst
+		T_B16 = T_B32 | T_B64, // m16bcst
 		T_M_K = 1 << 28, // mem{k}
 		T_VSIB = 1 << 29,
 		T_MEM_EVEX = 1 << 30, // use evex if mem
+		T_FP16 = 1 << 31,
+		T_MAP5 = T_FP16 | T_0F,
+		T_MAP6 = T_FP16 | T_0F38,
 		T_XXX
 	};
+	// T_66 = 1, T_F3 = 2, T_F2 = 3
+	uint32_t getPP(int type) { return (type >> 5) & 3; }
+

 const int NONE = 256; // same as Xbyak::CodeGenerator::NONE

@ -62,25 +70,30 @@ std::string type2String(int type)
 		if (!str.empty()) str += " | ";
 		str += "T_DUP";
 	}
-	if (type & T_66) {
-		if (!str.empty()) str += " | ";
-		str += "T_66";
-	}
-	if (type & T_F3) {
-		if (!str.empty()) str += " | ";
-		str += "T_F3";
-	}
 	if (type & T_F2) {
 		if (!str.empty()) str += " | ";
-		str += "T_F2";
+		switch (type & T_F2) {
+		case T_66: str += "T_66"; break;
+		case T_F3: str += "T_F3"; break;
+		case T_F2: str += "T_F2"; break;
+		default: break;
+		}
 	}
 	if (type & T_0F) {
 		if (!str.empty()) str += " | ";
-		str += "T_0F";
+		if (type & T_FP16) {
+			str += "T_MAP5";
+		} else {
+			str += "T_0F";
+		}
 	}
 	if (type & T_0F38) {
 		if (!str.empty()) str += " | ";
-		str += "T_0F38";
+		if (type & T_FP16) {
+			str += "T_MAP6";
+		} else {
+			str += "T_0F38";
+		}
 	}
 	if (type & T_0F3A) {
 		if (!str.empty()) str += " | ";
@ -130,6 +143,10 @@ std::string type2String(int type)
 		if (!str.empty()) str += " | ";
 		str += "T_ER_Z";
 	}
+	if (type & T_ER_R) {
+		if (!str.empty()) str += " | ";
+		str += "T_ER_R";
+	}
 	if (type & T_SAE_X) {
 		if (!str.empty()) str += " | ";
 		str += "T_SAE_X";
@ -148,9 +165,12 @@ std::string type2String(int type)
 	}
 	if (type & T_B32) {
 		if (!str.empty()) str += " | ";
-		str += "T_B32";
-	}
-	if (type & T_B64) {
+		if (type & T_B64) {
+			str += "T_B16"; // T_B16 = T_B32 | T_B64
+		} else {
+			str += "T_B32";
+		}
+	} else if (type & T_B64) {
 		if (!str.empty()) str += " | ";
 		str += "T_B64";
 	}
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@ -107,6 +107,8 @@ void putVcmp()
 		{ 0xC2, "vcmpps", T_0F | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_YMM | T_B32, true },
 		{ 0xC2, "vcmpsd", T_0F | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_F2 | T_N8, true },
 		{ 0xC2, "vcmpss", T_0F | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_F3 | T_N4, true },
+		{ 0xC2, "vcmpph", T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_YMM | T_B16, true },
+		{ 0xC2, "vcmpsh", T_F3 | T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, true },

 		{ 0x74, "vpcmpeqb", T_66 | T_0F | T_MUST_EVEX | T_YMM, false },
 		{ 0x75, "vpcmpeqw", T_66 | T_0F | T_MUST_EVEX | T_YMM, false },
@ -144,6 +146,25 @@ void putVcmp()
 		printf("void %s(const Opmask& k, const Xmm& x, const Operand& op%s) { opAVX_K_X_XM(k, x, op, %s, 0x%02X%s); }\n"
 			, p->name, p->hasIMM ? ", uint8_t imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
 	}
+	puts("void vcomish(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, 0x2F); }");
+	puts("void vucomish(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, 0x2E); }");
+}
+
+void putVcmpAlias()
+{
+	const char pred[32][16] = {
+		"eq", "lt", "le", "unord", "neq", "nlt", "nle", "ord",
+		"eq_uq", "nge", "ngt", "false", "neq_oq", "ge", "gt",
+		"true", "eq_os", "lt_oq", "le_oq", "unord_s", "neq_us", "nlt_uq", "nle_uq", "ord_s",
+		"eq_us", "nge_uq", "ngt_uq", "false_os", "neq_os", "ge_oq", "gt_oq", "true_us"
+	};
+	const char suf[][4] = { "pd", "ps", "sd", "ss" };
+	for (int i = 0; i < 4; i++) {
+		const char *s = suf[i];
+		for (int j = 0; j < 32; j++) {
+			printf("void vcmp%s%s(const Opmask& k, const Xmm& x, const Operand& op) { vcmp%s(k, x, op, %d); }\n", pred[j], s, s, j);
+		}
+	}
 }

 // XM_X
@ -178,6 +199,14 @@ void putX_XM()
 		{ 0x89, "vpexpandq", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_N8 },
 		{ 0x42, "vgetexppd", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_B64 | T_SAE_Z },
 		{ 0x42, "vgetexpps", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z },
+		{ 0x42, "vgetexpph", T_66 | T_MAP6 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_SAE_Z },
+
+		{ 0x7D, "vcvtph2uw", T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_ER_Z },
+		{ 0x7D, "vcvtph2w", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_ER_Z },
+		{ 0x7C, "vcvttph2uw", T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_SAE_Z },
+		{ 0x7C, "vcvttph2w", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_SAE_Z },
+		{ 0x7D, "vcvtuw2ph", T_F2 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_ER_Z },
+		{ 0x7D, "vcvtw2ph", T_F3 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_ER_Z },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@ -208,6 +237,8 @@ void putM_X()
 		{ 0x7F, "vmovdqu16", T_F2 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
 		{ 0x7F, "vmovdqu32", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW0 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
 		{ 0x7F, "vmovdqu64", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
+		{ 0x11, "vmovsh", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2 | T_M_K },
+		{ 0x7E, "vmovw", T_66 | T_MAP5 | T_MUST_EVEX | T_N2 },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@ -299,8 +330,10 @@ void putX_X_XM_IMM()

 		{ 0x43, "vgetexpsd", T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_SAE_X | T_N8, false },
 		{ 0x43, "vgetexpss", T_66 | T_0F38 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N4, false },
+		{ 0x43, "vgetexpsh", T_66 | T_MAP6 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, false },
 		{ 0x27, "vgetmantsd", T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_SAE_X | T_N8, true },
 		{ 0x27, "vgetmantss", T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N4, true },
+		{ 0x27, "vgetmantsh", T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, true },

 		{ 0x54, "vfixupimmpd", T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64 | T_SAE_Z, true },
 		{ 0x54, "vfixupimmps", T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, true },
@ -310,17 +343,26 @@ void putX_X_XM_IMM()
 		{ 0x4D, "vrcp14sd", T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_N8, false },
 		{ 0x4D, "vrcp14ss", T_66 | T_0F38 | T_MUST_EVEX | T_EW0 | T_N4, false },

+		{ 0x4D, "vrcpsh", T_66 | T_MAP6 | T_MUST_EVEX | T_EW0 | T_N2, false },
+
 		{ 0x4F, "vrsqrt14sd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_N8, false },
 		{ 0x4F, "vrsqrt14ss", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_N4, false },

-		{ 0x0B, "vrndscalesd", T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, true },
-		{ 0x0A, "vrndscaless", T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, true },
+		{ 0x4F, "vrsqrtsh", T_66 | T_MAP6 | T_MUST_EVEX | T_EW0 | T_N2, false },
+		{ 0x51, "vsqrtsh", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_ER_X | T_N2, false },
+
+		{ 0x0B, "vrndscalesd", T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8 | T_SAE_X, true },
+		{ 0x0A, "vrndscaless", T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4 | T_SAE_X, true },
+		{ 0x0A, "vrndscalesh", T_0F3A | T_MUST_EVEX | T_EW0 | T_N2 | T_SAE_X, true },

 		{ 0x2C, "vscalefpd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, false },
 		{ 0x2C, "vscalefps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_ER_Z, false },
 		{ 0x2D, "vscalefsd", T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_ER_X | T_N8, false },
 		{ 0x2D, "vscalefss", T_66 | T_0F38 | T_MUST_EVEX | T_EW0 | T_ER_X | T_N4, false },

+		{ 0x2C, "vscalefph", T_66 | T_MAP6 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_ER_Z, false },
+		{ 0x2D, "vscalefsh", T_66 | T_MAP6 | T_MUST_EVEX | T_EW0 | T_ER_X | T_N2, false },
+
 		{ 0x42, "vdbpsadbw", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0, true },
 		{ 0x83, "vpmultishiftqb", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },

@ -343,6 +385,7 @@ void putX_X_XM_IMM()

 		{ 0x57, "vreducesd", T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_SAE_X | T_N8, true },
 		{ 0x57, "vreducess", T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N4, true },
+		{ 0x57, "vreducesh", T_0F3A | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, true },

 		{ 0xB4, "vpmadd52luq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
 		{ 0xB5, "vpmadd52huq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
@ -365,6 +408,11 @@ void putX_X_XM_IMM()

 		{ 0x72, "vcvtne2ps2bf16", T_F2 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
 		{ 0x52, "vdpbf16ps", T_F3 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
+
+		{ 0x5A, "vcvtsd2sh", T_F2 | T_MAP5 | T_MUST_EVEX | T_EW1 | T_ER_X | T_N8, false },
+		{ 0x5A, "vcvtsh2sd", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, false },
+		{ 0x13, "vcvtsh2ss", T_MAP6 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, false },
+		{ 0x1D, "vcvtss2sh", T_MAP5 | T_MUST_EVEX | T_EW0 | T_ER_X | T_N4, false },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@ -488,20 +536,81 @@ void putBroadcast(bool only64bit)

 void putCvt()
 {
-	puts("void vcvtpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x79); }");
-	puts("void vcvtps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 0x7B); }");
-	puts("void vcvtps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 0x79); }");
-	puts("void vcvtqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x5B); }");
-	puts("void vcvttpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, 0x78); }");
-	puts("void vcvttps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 0x7A); }");
-	puts("void vcvttps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 0x78); }");
-	puts("void vcvtudq2pd(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_F3 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL, 0x7A); }");
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		int type;
+		int ptn;
+	} tbl[] = {
+		{ 0x79, "vcvtsd2usi", T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_ER_X, 0 },
+		{ 0x79, "vcvtss2usi", T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_ER_X, 0 },
+		{ 0x78, "vcvttsd2usi", T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_SAE_X, 0 },
+		{ 0x78, "vcvttss2usi", T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_SAE_X, 0 },
+		{ 0x2D, "vcvtsh2si", T_F3 | T_MAP5 | T_MUST_EVEX | T_N2 | T_ER_X, 0 },
+		{ 0x79, "vcvtsh2usi", T_F3 | T_MAP5 | T_MUST_EVEX | T_N2 | T_ER_X, 0 },
+		{ 0x2C, "vcvttsh2si", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2 | T_SAE_X, 0 },
+		{ 0x78, "vcvttsh2usi", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2 | T_SAE_X, 0 },

-	puts("void vcvtsd2usi(const Reg32e& r, const Operand& op) { int type = (T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_ER_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x79); }");
-	puts("void vcvtss2usi(const Reg32e& r, const Operand& op) { int type = (T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_ER_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x79); }");
-	puts("void vcvttsd2usi(const Reg32e& r, const Operand& op) { int type = (T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_SAE_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x78); }");
-	puts("void vcvttss2usi(const Reg32e& r, const Operand& op) { int type = (T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_SAE_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x78); }");
-	puts("void vcvtuqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F2 | T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x7A); }");
+		{ 0x7B, "vcvtps2qq", T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 1 },
+		{ 0x79, "vcvtps2uqq", T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 1 },
+		{ 0x7A, "vcvttps2qq", T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 1 },
+		{ 0x78, "vcvttps2uqq", T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 1 },
+		{ 0x7A, "vcvtudq2pd", T_F3 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL, 1 },
+		{ 0x5B, "vcvtph2dq", T_66 | T_MAP5 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_ER_Y | T_N8 | T_N_VL, 1 },
+		{ 0x13, "vcvtph2psx", T_66 | T_MAP6 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Y | T_N8 | T_N_VL, 1 },
+		{ 0x79, "vcvtph2udq", T_MAP5 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_ER_Y | T_N8 | T_N_VL, 1 },
+		{ 0x5B, "vcvttph2dq", T_F3 | T_MAP5 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Y | T_N8 | T_N_VL, 1 },
+		{ 0x78, "vcvttph2udq", T_MAP5 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Y | T_N8 | T_N_VL, 1 },
+
+		{ 0x79, "vcvtpd2udq", T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 2 },
+		{ 0x5B, "vcvtqq2ps", T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 2 },
+		{ 0x78, "vcvttpd2udq", T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, 2 },
+		{ 0x7A, "vcvtuqq2ps", T_F2 | T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 2 },
+
+		{ 0x5A, "vcvtph2pd", T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_N4 | T_N_VL | T_SAE_X, 3 },
+		{ 0x7B, "vcvtph2qq", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_N4 | T_N_VL | T_ER_X, 3 },
+		{ 0x79, "vcvtph2uqq", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_N4 | T_N_VL | T_ER_X, 3 },
+		{ 0x78, "vcvttph2uqq", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_N4 | T_N_VL | T_SAE_X, 3 },
+		{ 0x7A, "vcvttph2qq", T_66 | T_MAP5 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_N4 | T_N_VL | T_SAE_X, 3 },
+
+		{ 0x5B, "vcvtdq2ph", T_MAP5 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_ER_Z | T_N16 | T_N_VL, 4 },
+		{ 0x1D, "vcvtps2phx", T_66 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_B32 | T_ER_Z | T_N16 | T_N_VL, 4 },
+		{ 0x7A, "vcvtudq2ph", T_F2 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_B32 | T_ER_Z | T_N16 | T_N_VL, 4 },
+
+		{ 0x5A, "vcvtpd2ph", T_66 | T_MAP5 | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z | T_N16 | T_N_VL, 5 },
+		{ 0x5B, "vcvtqq2ph", T_MAP5 | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z | T_N16 | T_N_VL, 5 },
+		{ 0x7A, "vcvtuqq2ph", T_F2 | T_MAP5 | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z | T_N16 | T_N_VL, 5 },
+
+		{ 0x2A, "vcvtsi2sh", T_F3 | T_MAP5 | T_MUST_EVEX | T_ER_R | T_M_K, 6 },
+		{ 0x7B, "vcvtusi2sh", T_F3 | T_MAP5 | T_MUST_EVEX | T_ER_R | T_M_K, 6 },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		const Tbl& p = tbl[i];
+		std::string type = type2String(p.type);
+		switch (p.ptn) {
+		case 0:
+			printf("void %s(const Reg32e& r, const Operand& op) { int type = (%s) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 1:
+			printf("void %s(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 2:
+			printf("void %s(const Xmm& x, const Operand& op) { opCvt2(x, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 3:
+			printf("void %s(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 4:
+			printf("void %s(const Xmm& x, const Operand& op) { checkCvt4(x, op); opCvt(x, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 5:
+			printf("void %s(const Xmm& x, const Operand& op) { opCvt5(x, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		case 6:
+			printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op) { if (!(x1.isXMM() && x2.isXMM() && op.isBit(32|64))) XBYAK_THROW(ERR_BAD_COMBINATION) int type = (%s) | (op.isBit(32) ? (T_EW0 | T_N4) : (T_EW1 | T_N8)); opVex(x1, &x2, op, type, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			break;
+		}
+	}
 	puts("void vcvtusi2sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F2 | T_0F | T_MUST_EVEX, T_W1 | T_EW1 | T_ER_X | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }");
 	puts("void vcvtusi2ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F3 | T_0F | T_MUST_EVEX | T_ER_X, T_W1 | T_EW1 | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }");
 }
@ -628,14 +737,21 @@ void putX_XM_IMM()
 	} tbl[] = {
 		{ 0x26, "vgetmantpd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, true },
 		{ 0x26, "vgetmantps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_SAE_Z, true },
+		{ 0x26, "vgetmantph", T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Z, true },
 		{ 0x4C, "vrcp14pd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
 		{ 0x4C, "vrcp14ps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false },

+		{ 0x4C, "vrcpph", T_66 | T_MAP6 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16, false },
+
 		{ 0x4E, "vrsqrt14pd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
 		{ 0x4E, "vrsqrt14ps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false },

-		{ 0x09, "vrndscalepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, true },
-		{ 0x08, "vrndscaleps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, true },
+		{ 0x4E, "vrsqrtph", T_66 | T_MAP6 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16, false },
+		{ 0x51, "vsqrtph", T_MAP5| T_YMM | T_MUST_EVEX  | T_EW0 | T_ER_Z | T_B16, false },
+
+		{ 0x09, "vrndscalepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, true },
+		{ 0x08, "vrndscaleps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_SAE_Z, true },
+		{ 0x08, "vrndscaleph", T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Z, true },

 		{ 0xC4, "vpconflictd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false },
 		{ 0xC4, "vpconflictq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
@ -645,6 +761,7 @@ void putX_XM_IMM()

 		{ 0x56, "vreducepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, true },
 		{ 0x56, "vreduceps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_SAE_Z, true },
+		{ 0x56, "vreduceph", T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B16 | T_SAE_Z, true },

 		{ 0x54, "vpopcntb", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z, false },
 		{ 0x54, "vpopcntw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, false },
@ -704,8 +821,10 @@ void putMisc()

 	puts("void vfpclasspd(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64, 0x66, imm); }");
 	puts("void vfpclassps(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }");
+	puts("void vfpclassph(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B16, 0x66, imm); }");
 	puts("void vfpclasssd(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); }");
 	puts("void vfpclassss(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, 0x67, imm); }");
+	puts("void vfpclasssh(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_0F3A | T_MUST_EVEX | T_EW0 | T_N2, 0x67, imm); }");

 	puts("void vpshufbitqmb(const Opmask& k, const Xmm& x, const Operand& op) { opVex(k, &x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x8F); }");
 	puts("void vcvtneps2bf16(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x72); }");
@ -724,6 +843,126 @@ void putV4FMA()
 	puts("void vp4dpwssds(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x53); }");
 }

+void putFP16_1()
+{
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+	} tbl[] = {
+		{ 0x58, "add" },
+		{ 0x5C, "sub" },
+		{ 0x59, "mul" },
+		{ 0x5E, "div" },
+		{ 0x5F, "max" },
+		{ 0x5D, "min" },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		const Tbl *p = &tbl[i];
+		printf("void v%sph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x%02X); }\n", p->name, p->code);
+		printf("void v%ssh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x%02X); }\n", p->name, p->code);
+	}
+}
+
+void putFP16_FMA()
+{
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		bool isPH;
+	} tbl[] = {
+		{ 0x06, "vfmaddsub", true },
+		{ 0x07, "vfmsubadd", true },
+		{ 0x08, "vfmadd", true },
+		{ 0x0C, "vfnmadd", true },
+		{ 0x0A, "vfmsub", true },
+		{ 0x0E, "vfnmsub", true },
+		{ 0x09, "vfmadd", false },
+		{ 0x0D, "vfnmadd", false },
+		{ 0x0B, "vfmsub", false },
+		{ 0x0F, "vfnmsub", false },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		for (int k = 0; k < 3; k++) {
+			const struct Ord {
+				const char *str;
+				uint8_t code;
+			} ord[] = {
+				{ "132", 0x90 },
+				{ "213", 0xA0 },
+				{ "231", 0xB0 },
+			};
+			int t = T_66 | T_MAP6 | T_EW0 | T_MUST_EVEX;
+			const char *suf = 0;
+			if (tbl[i].isPH) {
+				t |= T_ER_Z | T_YMM | T_B16;
+				suf = "ph";
+			} else {
+				t |= T_ER_X | T_N2;
+				suf = "sh";
+			}
+			std::string type = type2String(t);
+			printf("void %s%s%s(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X); }\n"
+				, tbl[i].name, ord[k].str, suf, type.c_str(), tbl[i].code | ord[k].code);
+		}
+	}
+}
+void putFP16_FMA2()
+{
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		bool isPH;
+	} tbl[] = {
+		{ 0x56, "maddc", true },
+		{ 0xD6, "mulc", true },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		for (int j = 0; j < 2; j++) {
+			int t = T_MAP6 | T_EW0 | T_MUST_EVEX;
+			if (j == 0) {
+				t |= T_F2;
+			} else {
+				t |= T_F3;
+			}
+			const char *suf = 0;
+			if (tbl[i].isPH) {
+				t |= T_ER_Z | T_YMM | T_B32;
+				suf = "ph";
+			} else {
+				t |= T_ER_X | T_N2;
+				suf = "sh";
+			}
+			std::string type = type2String(t);
+			printf("void vf%s%s%s(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X); }\n"
+				, j == 0 ? "c" : "", tbl[i].name, suf, type.c_str(), tbl[i].code);
+		}
+	}
+}
+
+void putFP16_2()
+{
+	{
+		int t = T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2;
+		std::string type = type2String(t);
+		printf("void vmovsh(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, %s, 0x10); }\n", type.c_str());
+		printf("void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2, x3, %s, 0x10); }\n", type.c_str());
+	}
+	{
+		int t = T_66 | T_MAP5 | T_MUST_EVEX | T_N2;
+		std::string type = type2String(t);
+		printf("void vmovw(const Xmm& x, const Operand& op) { if (!op.isREG(32|64) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, xm0, op, %s, 0x6E); }\n", type.c_str());
+		printf("void vmovw(const Reg32e& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, r, %s, 0x7E); }\n", type.c_str());
+	}
+}
+
+void putFP16()
+{
+	putFP16_1();
+	putFP16_FMA();
+	putFP16_FMA2();
+	putFP16_2();
+}
+
 int main(int argc, char *[])
 {
 	bool only64bit = argc == 2;
@ -733,6 +972,7 @@ int main(int argc, char *[])
 		return 0;
 	}
 	putVcmp();
+	putVcmpAlias();
 	putX_XM();
 	putM_X();
 	putXM_X();
@ -747,4 +987,5 @@ int main(int argc, char *[])
 	putMisc();
 	putScatter();
 	putV4FMA();
+	putFP16();
 }
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@ -26,6 +26,7 @@ struct GenericTbl {
 	uint8_t code1;
 	uint8_t code2;
 	uint8_t code3;
+	uint8_t code4;
 };

 void putGeneric(const GenericTbl *p, size_t n)
@ -34,6 +35,7 @@ void putGeneric(const GenericTbl *p, size_t n)
 		printf("void %s() { db(0x%02X); ", p->name, p->code1);
 		if (p->code2) printf("db(0x%02X); ", p->code2);
 		if (p->code3) printf("db(0x%02X); ", p->code3);
+		if (p->code4) printf("db(0x%02X); ", p->code4);
 		printf("}\n");
 		p++;
 	}
@ -661,6 +663,9 @@ void put()
 			{ "cmpsb", 0xA6 },
 			{ "cmpsw", 0x66, 0xA7 },
 			{ "cmpsd", 0xA7 },
+			{ "endbr32", 0xF3, 0x0F, 0x1E, 0xFB },
+			{ "endbr64", 0xF3, 0x0F, 0x1E, 0xFA },
+			{ "hlt", 0xF4 },
 			{ "int3", 0xCC },
 			{ "scasb", 0xAE },
 			{ "scasw", 0x66, 0xAF },
@ -1040,11 +1045,14 @@ void put()
 		puts("void lea(const Reg& reg, const Address& addr) { if (!reg.isBit(16 | i32e)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModM(addr, reg, 0x8D); }");
 		puts("void bswap(const Reg32e& reg) { opModR(Reg32(1), reg, 0x0F); }");
 		puts("void ret(int imm = 0) { if (imm) { db(0xC2); dw(imm); } else { db(0xC3); } }");
+		puts("void retf(int imm = 0) { if (imm) { db(0xCA); dw(imm); } else { db(0xCB); } }");

 		puts("void xadd(const Operand& op, const Reg& reg) { opModRM(reg, op, (op.isREG() && reg.isREG() && op.getBit() == reg.getBit()), op.isMEM(), 0x0F, 0xC0 | (reg.isBit(8) ? 0 : 1)); }");
 		puts("void cmpxchg(const Operand& op, const Reg& reg) { opModRM(reg, op, (op.isREG() && reg.isREG() && op.getBit() == reg.getBit()), op.isMEM(), 0x0F, 0xB0 | (reg.isBit(8) ? 0 : 1)); }");
 		puts("void movbe(const Reg& reg, const Address& addr) { opModM(addr, reg, 0x0F, 0x38, 0xF0); }");
 		puts("void movbe(const Address& addr, const Reg& reg) { opModM(addr, reg, 0x0F, 0x38, 0xF1); }");
+		puts("void movdiri(const Address& addr, const Reg32e& reg) { opModM(addr, reg, 0x0F, 0x38, 0xF9); }");
+		puts("void movdir64b(const Reg& reg, const Address& addr) { db(0x66); opModM(addr, reg.cvt32(), 0x0F, 0x38, 0xF8); }");
 		puts("void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); }");
 		puts("void adox(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0xF3, isREG32_REG32orMEM, NONE, 0x38); }");
 		puts("void cmpxchg8b(const Address& addr) { opModM(addr, Reg32(1), 0x0F, 0xC7); }");
@ -1079,6 +1087,11 @@ void put()
 		puts("void rdrand(const Reg& r) { if (r.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModR(Reg(6, Operand::REG, r.getBit()), r, 0x0F, 0xC7); }");
 		puts("void rdseed(const Reg& r) { if (r.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModR(Reg(7, Operand::REG, r.getBit()), r, 0x0F, 0xC7); }");
 		puts("void crc32(const Reg32e& reg, const Operand& op) { if (reg.isBit(32) && op.isBit(16)) db(0x66); db(0xF2); opModRM(reg, op, op.isREG(), op.isMEM(), 0x0F, 0x38, 0xF0 | (op.isBit(8) ? 0 : 1)); }");
+		puts("void tpause(const Reg32& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) db(0x66); db(0x0F); db(0xAE); setModRM(3, 6, idx); }");
+		puts("void umonitor(const Reg& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) int bit = r.getBit(); if (BIT != bit) { if ((BIT == 32 && bit == 16) || (BIT == 64 && bit == 32)) { db(0x67); } else { XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) } } db(0xF3); db(0x0F); db(0xAE); setModRM(3, 6, idx); }");
+		puts("void umwait(const Reg32& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) db(0xF2); db(0x0F); db(0xAE); setModRM(3, 6, idx); }");
+		puts("void clwb(const Address& addr) { db(0x66); opMIB(addr, esi, 0x0F, 0xAE); }");
+		puts("void cldemote(const Address& addr) { opMIB(addr, eax, 0x0F, 0x1C); }");
 	}
 	{
 		const struct Tbl {
@ -1207,8 +1220,8 @@ void put()
 			printf("void v%spd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x%02X); }\n", p->name, p->code);
 			printf("void v%sps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x%02X); }\n", p->name, p->code);
 			if (p->only_pd_ps) continue;
-			printf("void v%ssd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x%02X); }\n", p->name, p->code);
-			printf("void v%sss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x%02X); }\n", p->name, p->code);
+			printf("void v%ssd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x%02X); }\n", p->name, p->code);
+			printf("void v%sss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x%02X); }\n", p->name, p->code);
 		}
 	}
 	putX_X_XM(false);
@ -1299,7 +1312,8 @@ void put()
 			if (p->mode & 1) {
 				const char *immS1 = p->hasIMM ? ", uint8_t imm" : "";
 				const char *immS2 = p->hasIMM ? ", imm" : ", NONE";
-				const char *pref = p->type & T_66 ? "0x66" : p->type & T_F2 ? "0xF2" : p->type & T_F3 ? "0xF3" : "NONE";
+				const char *prefTbl[5] = { "NONE", "0x66", "0xF3", "0xF2" };
+				const char *pref = prefTbl[getPP(p->type)];
 				const char *suf = p->type & T_0F38 ? "0x38" : p->type & T_0F3A ? "0x3A" : "NONE";
 				printf("void %s(const Xmm& xmm, const Operand& op%s) { opGen(xmm, op, 0x%02X, %s, isXMM_XMMorMEM%s, %s); }\n", p->name, immS1, p->code, pref, immS2, suf);
 			}
@ -1350,11 +1364,12 @@ void put()
 			{ 0xDE, "aesdec", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
 			{ 0xDF, "aesdeclast", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
 		};
+		const uint8_t ppTbl[] = { 0, 0x66, 0xf3, 0xf2 };
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
 			std::string type = type2String(p->type);
 			if (p->mode & 1) {
-				uint8_t pref = p->type & T_66 ? 0x66 : p->type & T_F2 ? 0xF2 : p->type & T_F3 ? 0xF3 : 0;
+				uint8_t pref = ppTbl[getPP(p->type)];
 				printf("void %s(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x%02X, 0x%02X, isXMM_XMMorMEM%s); }\n", p->name, p->code, pref, p->type & T_0F38 ? ", NONE, 0x38" : "");
 			}
 			if (p->mode & 2) {
@ -1648,7 +1663,7 @@ void put()
 		puts("void vcvttpd2dq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_66 | T_0F | T_YMM | T_EVEX |T_EW1 | T_B64 | T_ER_Z, 0xE6); }");

 		puts("void vcvtph2ps(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_0F38 | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y, 0x13); }");
-		puts("void vcvtps2ph(const Operand& op, const Xmm& x, uint8_t imm) { checkCvt1(x, op); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y, 0x1D, imm); }");
+		puts("void vcvtps2ph(const Operand& op, const Xmm& x, uint8_t imm) { checkCvt1(x, op); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y | T_M_K, 0x1D, imm); }");

 	}
 	// haswell gpr(reg, reg, r/m)
--- a/meson.build
+++ b/meson.build
@ -0,0 +1,45 @@
+# SPDX-FileCopyrightText: 2021 Andrea Pappacoda
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+project(
+	'xbyak',
+	'cpp',
+	version: '6.06',
+	license: 'BSD-3-Clause',
+	default_options: 'b_ndebug=if-release'
+)
+
+install_subdir('xbyak', install_dir: get_option('includedir'))
+
+xbyak_dep = declare_dependency(include_directories: include_directories('.'))
+
+if meson.version().version_compare('>=0.54.0')
+	meson.override_dependency(meson.project_name(), xbyak_dep)
+endif
+
+import('pkgconfig').generate(
+	name: meson.project_name(),
+	description: 'JIT assembler for x86(IA32), x64(AMD64, x86-64)',
+	version: meson.project_version(),
+	url: 'https://github.com/herumi/xbyak'
+)
+
+if meson.version().version_compare('>=0.50.0')
+	cmake = import('cmake')
+
+	cmake.write_basic_package_version_file(
+		name: meson.project_name(),
+		version: meson.project_version()
+	)
+
+	cmake_conf = configuration_data()
+	cmake_conf.set('TARGET_NAME', meson.project_name() + '::' + meson.project_name())
+	cmake_conf.set('ABSOLUTE_INCLUDE_DIR', get_option('prefix')/get_option('includedir'))
+
+	cmake.configure_package_config_file(
+		name: meson.project_name(),
+		input: 'cmake'/'meson-config.cmake.in',
+		configuration: cmake_conf
+	)
+endif
--- a/readme.md
+++ b/readme.md
@ -1,6 +1,13 @@
-[![Build Status](https://github.com/herumi/xbyak/actions/workflows/main.yml/badge.svg)](https://github.com/herumi/xbyak/actions/workflows/main.yml)

-# Xbyak 5.991 ; JIT assembler for x86(IA32), x64(AMD64, x86-64) by C++
+# Xbyak 6.06 [![Badge Build]][Build Status]
+
+*A C++ JIT assembler for x86 (IA32), x64 (AMD64, x86-64)*
+
+## Menu
+
+- [Install]
+- [Usage]
+- [Changelog]

 ## Abstract

@ -10,15 +17,23 @@ The pronunciation of Xbyak is `kəi-bja-k`.
 It is named from a Japanese word [開闢](https://translate.google.com/?hl=ja&sl=ja&tl=en&text=%E9%96%8B%E9%97%A2&op=translate), which means the beginning of the world.

 ## Feature
-* header file only
-* Intel/MASM like syntax
-* fully support AVX-512
+
+- header file only
+- Intel/MASM like syntax
+- fully support AVX-512

 **Note**:
 Use `and_()`, `or_()`, ... instead of `and()`, `or()`.
 If you want to use them, then specify `-fno-operator-names` option to gcc/clang.

 ### News
+
+- add movdiri, movdir64b, clwb, cldemote
+- WAITPKG instructions (tpause, umonitor, umwait) are supported.
+- MmapAllocator supports memfd with user-defined strings. see sample/memfd.cpp
+- strictly check address offset disp32 in a signed 32-bit integer. e.g., `ptr[(void*)0xffffffff]` causes an error.
+  - define `XBYAK_OLD_DISP_CHECK` if you need an old check, but the option will be remoevd.
+- add `jmp(mem, T_FAR)`, `call(mem, T_FAR)` `retf()` for far absolute indirect jump.
 - vnni instructions such as vpdpbusd supports vex encoding.
 - (break backward compatibility) `push(byte, imm)` (resp. `push(word, imm)`) forces to cast `imm` to 8(resp. 16) bit.
 - (Windows) `#include <winsock2.h>` has been removed from xbyak.h, so add it explicitly if you need it.
@ -27,590 +42,34 @@ If you want to use them, then specify `-fno-operator-names` option to gcc/clang.

 ### Supported OS

-* Windows Xp, Vista, Windows 7, Windows 10(32bit, 64bit)
-* Linux(32bit, 64bit)
-* Intel macOS
+- Windows (Xp, Vista, 7, 10, 11) (32 / 64 bit)
+- Linux (32 / 64 bit)
+- macOS (Intel CPU)

 ### Supported Compilers

 Almost C++03 or later compilers for x86/x64 such as Visual Studio, g++, clang++, Intel C++ compiler and g++ on mingw/cygwin.

-## Install
-
-The following files are necessary. Please add the path to your compile directory.
-
-* xbyak.h
-* xbyak_mnemonic.h
-* xbyak_util.h
-
-Linux:
-```
-make install
-```
-
-These files are copied into `/usr/local/include/xbyak`.
-
-## How to use it
-
-Inherit `Xbyak::CodeGenerator` class and make the class method.
-```
-#include <xbyak/xbyak.h>
-
-struct Code : Xbyak::CodeGenerator {
-    Code(int x)
-    {
-        mov(eax, x);
-        ret();
-    }
-};
-```
-Or you can pass the instance of CodeGenerator without inheriting.
-```
-void genCode(Xbyak::CodeGenerator& code, int x) {
-    using namespace Xbyak::util;
-    code.mov(eax, x);
-    code.ret();
-}
-```
-
-Make an instance of the class and get the function
-pointer by calling `getCode()` and call it.
-```
-Code c(5);
-int (*f)() = c.getCode<int (*)()>();
-printf("ret=%d\n", f()); // ret = 5
-```
-
-## Syntax
-Similar to MASM/NASM syntax with parentheses.
-
-```
-NASM              Xbyak
-mov eax, ebx  --> mov(eax, ebx);
-inc ecx           inc(ecx);
-ret           --> ret();
-```
-
-## Addressing
-Use `qword`, `dword`, `word` and `byte` if it is necessary to specify the size of memory,
-otherwise use `ptr`.
-
-```
-(ptr|qword|dword|word|byte) [base + index * (1|2|4|8) + displacement]
-                            [rip + 32bit disp] ; x64 only
-
-NASM                   Xbyak
-mov eax, [ebx+ecx] --> mov(eax, ptr [ebx+ecx]);
-mov al, [ebx+ecx]  --> mov(al, ptr [ebx + ecx]);
-test byte [esp], 4 --> test(byte [esp], 4);
-inc qword [rax]    --> inc(qword [rax]);
-```
-**Note**: `qword`, ... are member variables, then don't use `dword` as unsigned int type.
-
-### How to use Selector (Segment Register)
-```
-mov eax, [fs:eax] --> putSeg(fs);
-                      mov(eax, ptr [eax]);
-mov ax, cs        --> mov(ax, cs);
-```
-**Note**: Segment class is not derived from `Operand`.
-
-## AVX
-
-```
-vaddps(xmm1, xmm2, xmm3); // xmm1 <- xmm2 + xmm3
-vaddps(xmm2, xmm3, ptr [rax]); // use ptr to access memory
-vgatherdpd(xmm1, ptr [ebp + 256 + xmm2*4], xmm3);
-```
-
-**Note**:
-If `XBYAK_ENABLE_OMITTED_OPERAND` is defined, then you can use two operand version for backward compatibility.
-But the newer version will not support it.
-```
-vaddps(xmm2, xmm3); // xmm2 <- xmm2 + xmm3
-```
-
-## AVX-512
-
-```
-vaddpd zmm2, zmm5, zmm30                --> vaddpd(zmm2, zmm5, zmm30);
-vaddpd xmm30, xmm20, [rax]              --> vaddpd(xmm30, xmm20, ptr [rax]);
-vaddps xmm30, xmm20, [rax]              --> vaddps(xmm30, xmm20, ptr [rax]);
-vaddpd zmm2{k5}, zmm4, zmm2             --> vaddpd(zmm2 | k5, zmm4, zmm2);
-vaddpd zmm2{k5}{z}, zmm4, zmm2          --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2);
-vaddpd zmm2{k5}{z}, zmm4, zmm2,{rd-sae} --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2 | T_rd_sae);
-                                            vaddpd(zmm2 | k5 | T_z | T_rd_sae, zmm4, zmm2); // the position of `|` is arbitrary.
-vcmppd k4{k3}, zmm1, zmm2, {sae}, 5     --> vcmppd(k4 | k3, zmm1, zmm2 | T_sae, 5);
-
-vaddpd xmm1, xmm2, [rax+256]            --> vaddpd(xmm1, xmm2, ptr [rax+256]);
-vaddpd xmm1, xmm2, [rax+256]{1to2}      --> vaddpd(xmm1, xmm2, ptr_b [rax+256]);
-vaddpd ymm1, ymm2, [rax+256]{1to4}      --> vaddpd(ymm1, ymm2, ptr_b [rax+256]);
-vaddpd zmm1, zmm2, [rax+256]{1to8}      --> vaddpd(zmm1, zmm2, ptr_b [rax+256]);
-vaddps zmm1, zmm2, [rax+rcx*8+8]{1to16} --> vaddps(zmm1, zmm2, ptr_b [rax+rcx*8+8]);
-vmovsd [rax]{k1}, xmm4                  --> vmovsd(ptr [rax] | k1, xmm4);
-
-vcvtpd2dq xmm16, oword [eax+33]         --> vcvtpd2dq(xmm16, xword [eax+33]); // use xword for m128 instead of oword
-                                            vcvtpd2dq(xmm16, ptr [eax+33]); // default xword
-vcvtpd2dq xmm21, [eax+32]{1to2}         --> vcvtpd2dq(xmm21, ptr_b [eax+32]);
-vcvtpd2dq xmm0, yword [eax+33]          --> vcvtpd2dq(xmm0, yword [eax+33]); // use yword for m256
-vcvtpd2dq xmm19, [eax+32]{1to4}         --> vcvtpd2dq(xmm19, yword_b [eax+32]); // use yword_b to broadcast
-
-vfpclassps k5{k3}, zword [rax+64], 5    --> vfpclassps(k5|k3, zword [rax+64], 5); // specify m512
-vfpclasspd k5{k3}, [rax+64]{1to2}, 5    --> vfpclasspd(k5|k3, xword_b [rax+64], 5); // broadcast 64-bit to 128-bit
-vfpclassps k5{k3}, [rax+64]{1to4}, 5    --> vfpclassps(k5|k3, yword_b [rax+64], 5); // broadcast 64-bit to 256-bit
-
-vpdpbusd(xm0, xm1, xm2); // default encoding is EVEX
-vpdpbusd(xm0, xm1, xm2, EvexEncoding); // same as the above
-vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX encoding
-```
-### Remark
-* `k1`, ..., `k7` are opmask registers.
-  - `k0` is dealt as no mask.
-  - e.g. `vmovaps(zmm0|k0, ptr[rax]);` and `vmovaps(zmm0|T_z, ptr[rax]);` are same to `vmovaps(zmm0, ptr[rax]);`.
-* use `| T_z`, `| T_sae`, `| T_rn_sae`, `| T_rd_sae`, `| T_ru_sae`, `| T_rz_sae` instead of `,{z}`, `,{sae}`, `,{rn-sae}`, `,{rd-sae}`, `,{ru-sae}`, `,{rz-sae}` respectively.
-* `k4 | k3` is different from `k3 | k4`.
-* use `ptr_b` for broadcast `{1toX}`. X is automatically determined.
-* specify `xword`/`yword`/`zword(_b)` for m128/m256/m512 if necessary.
-
-## Label
-Two kinds of Label are supported. (String literal and Label class).
-
-### String literal
-```
-L("L1");
-  jmp("L1");
-
-  jmp("L2");
-  ...
-  a few mnemonics (8-bit displacement jmp)
-  ...
-L("L2");
-
-  jmp("L3", T_NEAR);
-  ...
-  a lot of mnemonics (32-bit displacement jmp)
-  ...
-L("L3");
-```
-
-* Call `hasUndefinedLabel()` to verify your code has no undefined label.
-* you can use a label for immediate value of mov like as `mov(eax, "L2")`.
-
-### Support `@@`, `@f`, `@b` like MASM
-
-```
-L("@@"); // <A>
-  jmp("@b"); // jmp to <A>
-  jmp("@f"); // jmp to <B>
-L("@@"); // <B>
-  jmp("@b"); // jmp to <B>
-  mov(eax, "@b");
-  jmp(eax); // jmp to <B>
-```
-
-### Local label
-
-Label symbols beginning with a period between `inLocalLabel()` and `outLocalLabel()`
-are treated as a local label.
-`inLocalLabel()` and `outLocalLabel()` can be nested.
-
-```
-void func1()
-{
-    inLocalLabel();
-  L(".lp"); // <A> ; local label
-    ...
-    jmp(".lp"); // jmp to <A>
-  L("aaa"); // global label <C>
-    outLocalLabel();
-
-    inLocalLabel();
-  L(".lp"); // <B> ; local label
-    func1();
-    jmp(".lp"); // jmp to <B>
-    inLocalLabel();
-    jmp("aaa"); // jmp to <C>
-}
-```
-
-### short and long jump
-Xbyak deals with jump mnemonics of an undefined label as short jump if no type is specified.
-So if the size between jmp and label is larger than 127 byte, then xbyak will cause an error.
-
-```
-jmp("short-jmp"); // short jmp
-// small code
-L("short-jmp");
-
-jmp("long-jmp");
-// long code
-L("long-jmp"); // throw exception
-```
-Then specify T_NEAR for jmp.
-```
-jmp("long-jmp", T_NEAR); // long jmp
-// long code
-L("long-jmp");
-```
-Or call `setDefaultJmpNEAR(true);` once, then the default type is set to T_NEAR.
-```
-jmp("long-jmp"); // long jmp
-// long code
-L("long-jmp");
-```
-
-### Label class
-
-`L()` and `jxx()` support Label class.
-
-```
-  Xbyak::Label label1, label2;
-L(label1);
-  ...
-  jmp(label1);
-  ...
-  jmp(label2);
-  ...
-L(label2);
-```
-
-Use `putL` for jmp table
-```
-    Label labelTbl, L0, L1, L2;
-    mov(rax, labelTbl);
-    // rdx is an index of jump table
-    jmp(ptr [rax + rdx * sizeof(void*)]);
-L(labelTbl);
-    putL(L0);
-    putL(L1);
-    putL(L2);
-L(L0);
-    ....
-L(L1);
-    ....
-```
-
-`assignL(dstLabel, srcLabel)` binds dstLabel with srcLabel.
-
-```
-  Label label2;
-  Label label1 = L(); // make label1 ; same to Label label1; L(label1);
-  ...
-  jmp(label2); // label2 is not determined here
-  ...
-  assignL(label2, label1); // label2 <- label1
-```
-The `jmp` in the above code jumps to label1 assigned by `assignL`.
-
-**Note**:
-* srcLabel must be used in `L()`.
-* dstLabel must not be used in `L()`.
-
-`Label::getAddress()` returns the address specified by the label instance and 0 if not specified.
-```
-// not AutoGrow mode
-Label  label;
-assert(label.getAddress() == 0);
-L(label);
-assert(label.getAddress() == getCurr());
-```
-
-### Rip ; relative addressing
-```
-Label label;
-mov(eax, ptr [rip + label]); // eax = 4
-...
-
-L(label);
-dd(4);
-```
-```
-int x;
-...
-  mov(eax, ptr[rip + &x]); // throw exception if the difference between &x and current position is larger than 2GiB
-```
-
-## Code size
-The default max code size is 4096 bytes.
-Specify the size in constructor of `CodeGenerator()` if necessary.
-
-```
-class Quantize : public Xbyak::CodeGenerator {
-public:
-  Quantize()
-    : CodeGenerator(8192)
-  {
-  }
-  ...
-};
-```
-
-## User allocated memory
-
-You can make jit code on prepared memory.
-
-Call `setProtectModeRE` yourself to change memory mode if using the prepared memory.
-
-```
-uint8_t alignas(4096) buf[8192]; // C++11 or later
-
-struct Code : Xbyak::CodeGenerator {
-    Code() : Xbyak::CodeGenerator(sizeof(buf), buf)
-    {
-        mov(rax, 123);
-        ret();
-    }
-};
-
-int main()
-{
-    Code c;
-    c.setProtectModeRE(); // set memory to Read/Exec
-    printf("%d\n", c.getCode<int(*)()>()());
-}
-```
-
-**Note**: See [sample/test0.cpp](sample/test0.cpp).
-
-### AutoGrow
-
-The memory region for jit is automatically extended if necessary when `AutoGrow` is specified in a constructor of `CodeGenerator`.
-
-Call `ready()` or `readyRE()` before calling `getCode()` to fix jump address.
-```
-struct Code : Xbyak::CodeGenerator {
-  Code()
-    : Xbyak::CodeGenerator(<default memory size>, Xbyak::AutoGrow)
-  {
-     ...
-  }
-};
-Code c;
-// generate code for jit
-c.ready(); // mode = Read/Write/Exec
-```
-
-**Note**:
-* Don't use the address returned by `getCurr()` before calling `ready()` because it may be invalid address.
-
-### Read/Exec mode
-Xbyak set Read/Write/Exec mode to memory to run jit code.
-If you want to use Read/Exec mode for security, then specify `DontSetProtectRWE` for `CodeGenerator` and
-call `setProtectModeRE()` after generating jit code.
-
-```
-struct Code : Xbyak::CodeGenerator {
-    Code()
-        : Xbyak::CodeGenerator(4096, Xbyak::DontSetProtectRWE)
-    {
-        mov(eax, 123);
-        ret();
-    }
-};
-
-Code c;
-c.setProtectModeRE();
-...
-
-```
-Call `readyRE()` instead of `ready()` when using `AutoGrow` mode.
-See [protect-re.cpp](sample/protect-re.cpp).
-
-## Exception-less mode
-If `XBYAK_NO_EXCEPTION` is defined, then gcc/clang can compile xbyak with `-fno-exceptions`.
-In stead of throwing an exception, `Xbyak::GetError()` returns non-zero value (e.g. `ERR_BAD_ADDRESSING`) if there is something wrong.
-The status will not be changed automatically, then you should reset it by `Xbyak::ClearError()`.
-`CodeGenerator::reset()` calls `ClearError()`.
-
-## Macro
-
-* **XBYAK32** is defined on 32bit.
-* **XBYAK64** is defined on 64bit.
-* **XBYAK64_WIN** is defined on 64bit Windows(VC).
-* **XBYAK64_GCC** is defined on 64bit gcc, cygwin.
-* define **XBYAK_USE_OP_NAMES** on gcc with `-fno-operator-names` if you want to use `and()`, ....
-* define **XBYAK_ENABLE_OMITTED_OPERAND** if you use omitted destination such as `vaddps(xmm2, xmm3);`(deprecated in the future).
-* define **XBYAK_UNDEF_JNL** if Bessel function jnl is defined as macro.
-* define **XBYAK_NO_EXCEPTION** for a compiler option `-fno-exceptions`.
-
-## Sample
-
-* [test0.cpp](sample/test0.cpp) ; tiny sample (x86, x64)
-* [quantize.cpp](sample/quantize.cpp) ; JIT optimized quantization by fast division (x86 only)
-* [calc.cpp](sample/calc.cpp) ; assemble and estimate a given polynomial (x86, x64)
-* [bf.cpp](sample/bf.cpp) ; JIT brainfuck (x86, x64)
-
 ## License

-modified new BSD License
-http://opensource.org/licenses/BSD-3-Clause
-
-## History
-* 2020/Nov/16 ver 5.991 disable constexpr for gcc-5 with -std=c++-14
-* 2020/Oct/19 ver 5.99 support VNNI instructions(Thanks to akharito)
-* 2020/Oct/17 ver 5.98 support the form of [scale * reg]
-* 2020/Sep/08 ver 5.97 replace uint32 with uint32_t etc.
-* 2020/Aug/28 ver 5.95 some constructors of register classes support constexpr if C++14 or later
-* 2020/Aug/04 ver 5.941 `CodeGenerator::reset()` calls `ClearError()`.
-* 2020/Jul/28 ver 5.94 remove #include <winsock2.h> (only windows)
-* 2020/Jul/21 ver 5.93 support exception-less mode
-* 2020/Jun/30 ver 5.92 support Intel AMX instruction set (Thanks to nshustrov)
-* 2020/Jun/22 ver 5.913 fix mov(r64, imm64) on 32-bit env with XBYAK64
-* 2020/Jun/19 ver 5.912 define MAP_JIT on macOS regardless of Xcode version (Thanks to rsdubtso)
-* 2020/May/10 ver 5.911 XBYAK_USE_MMAP_ALLOCATOR is defined unless XBYAK_DONT_USE_MMAP_ALLOCATOR is defined.
-* 2020/Apr/20 ver 5.91 accept mask register k0 (it means no mask)
-* 2020/Apr/09 ver 5.90 kmov{b,d,w,q} throws exception for an unsupported register
-* 2020/Feb/26 ver 5.891 fix typo of type
-* 2020/Jan/03 ver 5.89 fix error of vfpclasspd
-* 2019/Dec/20 ver 5.88 fix compile error on Windows
-* 2019/Dec/19 ver 5.87 add setDefaultJmpNEAR(), which deals with `jmp` of an undefined label as T_NEAR if no type is specified.
-* 2019/Dec/13 ver 5.86 [changed] revert to the behavior before v5.84 if -fno-operator-names is defined (and() is available)
-* 2019/Dec/07 ver 5.85 append MAP_JIT flag to mmap for macOS mojave or later
-* 2019/Nov/29 ver 5.84 [changed] XBYAK_NO_OP_NAMES is defined unless XBYAK_USE_OP_NAMES is defined
-* 2019/Oct/12 ver 5.83 exit(1) was removed
-* 2019/Sep/23 ver 5.82 support monitorx, mwaitx, clzero (thanks to @MagurosanTeam)
-* 2019/Sep/14 ver 5.81 support some generic mnemonics.
-* 2019/Aug/01 ver 5.802 fix detection of AVX512_BF16 (thanks to vpirogov)
-* 2019/May/27 support vp2intersectd, vp2intersectq (not tested)
-* 2019/May/26 ver 5.80 support vcvtne2ps2bf16, vcvtneps2bf16, vdpbf16ps
-* 2019/Apr/27 ver 5.79 vcmppd/vcmpps supports ptr_b(thanks to jkopinsky)
-* 2019/Apr/15 ver 5.78 rewrite Reg::changeBit() (thanks to MerryMage)
-* 2019/Mar/06 ver 5.77 fix number of cores that share LLC cache by densamoilov
-* 2019/Jan/17 ver 5.76 add Cpu::getNumCores() by shelleygoel
-* 2018/Oct/31 ver 5.751 recover Xbyak::CastTo for compatibility
-* 2018/Oct/29 ver 5.75 unlink LabelManager from Label when msg is destroyed
-* 2018/Oct/21 ver 5.74 support RegRip +/- int. Xbyak::CastTo is removed
-* 2018/Oct/15 util::AddressFrame uses push/pop instead of mov
-* 2018/Sep/19 ver 5.73 fix evex encoding of vpslld, vpslldq, vpsllw, etc for (reg, mem, imm8)
-* 2018/Sep/19 ver 5.72 fix the encoding of vinsertps for disp8N(Thanks to petercaday)
-* 2018/Sep/04 ver 5.71 L() returns a new label instance
-* 2018/Aug/27 ver 5.70 support setProtectMode() and DontUseProtect for read/exec setting
-* 2018/Aug/24 ver 5.68 fix wrong VSIB encoding with vector index >= 16(thanks to petercaday)
-* 2018/Aug/14 ver 5.67 remove mutable in Address ; fix setCacheHierarchy for cloud vm
-* 2018/Jul/26 ver 5.661 support mingw64
-* 2018/Jul/24 ver 5.66 add CodeArray::PROTECT_RE to mode of protect()
-* 2018/Jun/26 ver 5.65 fix push(qword [mem])
-* 2018/Mar/07 ver 5.64 fix zero division in Cpu() on some cpu
-* 2018/Feb/14 ver 5.63 fix Cpu::setCacheHierarchy() and fix EvexModifierZero for clang<3.9(thanks to mgouicem)
-* 2018/Feb/13 ver 5.62 Cpu::setCacheHierarchy() by mgouicem and rsdubtso
-* 2018/Feb/07 ver 5.61 vmov* supports mem{k}{z}(I forgot it)
-* 2018/Jan/24 ver 5.601 add xword, yword, etc. into Xbyak::util namespace
-* 2018/Jan/05 ver 5.60 support AVX-512 for Ice lake(319433-030.pdf)
-* 2017/Aug/22 ver 5.53 fix mpx encoding, add bnd() prefix
-* 2017/Aug/18 ver 5.52 fix align (thanks to MerryMage)
-* 2017/Aug/17 ver 5.51 add multi-byte nop and align() uses it(thanks to inolen)
-* 2017/Aug/08 ver 5.50 add mpx(thanks to magurosan)
-* 2017/Aug/08 ver 5.45 add sha(thanks to magurosan)
-* 2017/Aug/08 ver 5.44 add prefetchw(thanks to rsdubtso)
-* 2017/Jul/12 ver 5.432 reduce warnings of PVS studio
-* 2017/Jul/09 ver 5.431 fix hasRex() (no affect) (thanks to drillsar)
-* 2017/May/14 ver 5.43 fix CodeGenerator::resetSize() (thanks to gibbed)
-* 2017/May/13 ver 5.42 add movs{b,w,d,q}
-* 2017/Jan/26 ver 5.41 add prefetchwt1 and support for scale == 0(thanks to rsdubtso)
-* 2016/Dec/14 ver 5.40 add Label::getAddress() method to get the pointer specified by the label
-* 2016/Dec/09 ver 5.34 fix handling of negative offsets when encoding disp8N(thanks to rsdubtso)
-* 2016/Dec/08 ver 5.33 fix encoding of vpbroadcast{b,w,d,q}, vpinsr{b,w}, vpextr{b,w} for disp8N
-* 2016/Dec/01 ver 5.32 rename __xgetbv() to _xgetbv() to support clang for Visual Studio(thanks to freiro)
-* 2016/Nov/27 ver 5.31 rename AVX512_4VNNI to AVX512_4VNNIW
-* 2016/Nov/27 ver 5.30 add AVX512_4VNNI, AVX512_4FMAPS instructions(thanks to rsdubtso)
-* 2016/Nov/26 ver 5.20 add detection of AVX512_4VNNI and AVX512_4FMAPS(thanks to rsdubtso)
-* 2016/Nov/20 ver 5.11 lost vptest for ymm(thanks to gregory38)
-* 2016/Nov/20 ver 5.10 add addressing [rip+&var]
-* 2016/Sep/29 ver 5.03 fix detection ERR_INVALID_OPMASK_WITH_MEMORY(thanks to PVS-Studio)
-* 2016/Aug/15 ver 5.02 xbyak does not include xbyak_bin2hex.h
-* 2016/Aug/15 ver 5.011 fix detection of version of gcc 5.4
-* 2016/Aug/03 ver 5.01 disable omitted operand
-* 2016/Jun/24 ver 5.00 support avx-512 instruction set
-* 2016/Jun/13 avx-512 add mask instructions
-* 2016/May/05 ver 4.91 add detection of AVX-512 to Xbyak::util::Cpu
-* 2016/Mar/14 ver 4.901 comment to ready() function(thanks to skmp)
-* 2016/Feb/04 ver 4.90 add jcc(const void *addr);
-* 2016/Jan/30 ver 4.89 vpblendvb supports ymm reg(thanks to John Funnell)
-* 2016/Jan/24 ver 4.88 lea, cmov supports 16-bit register(thanks to whyisthisfieldhere)
-* 2015/Oct/05 ver 4.87 support segment selectors
-* 2015/Aug/18 ver 4.86 fix [rip + label] addressing with immediate value(thanks to whyisthisfieldhere)
-* 2015/Aug/10 ver 4.85 Address::operator==() is not correct(thanks to inolen)
-* 2015/Jun/22 ver 4.84 call() support variadic template if available(thanks to randomstuff)
-* 2015/Jun/16 ver 4.83 support movbe(thanks to benvanik)
-* 2015/May/24 ver 4.82 support detection of F16C
-* 2015/Apr/25 ver 4.81 fix the condition to throw exception for setSize(thanks to whyisthisfieldhere)
-* 2015/Apr/22 ver 4.80 rip supports label(thanks to whyisthisfieldhere)
-* 2015/Jar/28 ver 4.71 support adcx, adox, cmpxchg, rdseed, stac
-* 2014/Oct/14 ver 4.70 support MmapAllocator
-* 2014/Jun/13 ver 4.62 disable warning of VC2014
-* 2014/May/30 ver 4.61 support bt, bts, btr, btc
-* 2014/May/28 ver 4.60 support vcvtph2ps, vcvtps2ph
-* 2014/Apr/11 ver 4.52 add detection of rdrand
-* 2014/Mar/25 ver 4.51 remove state information of unreferenced labels
-* 2014/Mar/16 ver 4.50 support new Label
-* 2014/Mar/05 ver 4.40 fix wrong detection of BMI/enhanced rep on VirtualBox
-* 2013/Dec/03 ver 4.30 support Reg::cvt8(), cvt16(), cvt32(), cvt64()
-* 2013/Oct/16 ver 4.21 label support std::string
-* 2013/Jul/30 ver 4.20 [break backward compatibility] split Reg32e class into RegExp(base+index*scale+disp) and Reg32e(means Reg32 or Reg64)
-* 2013/Jul/04 ver 4.10 [break backward compatibility] change the type of Xbyak::Error from enum to a class
-* 2013/Jun/21 ver 4.02 add putL(LABEL) function to put the address of the label
-* 2013/Jun/21 ver 4.01 vpsllw, vpslld, vpsllq, vpsraw, vpsrad, vpsrlw, vpsrld, vpsrlq support (ymm, ymm, xmm). support vpbroadcastb, vpbroadcastw, vpbroadcastd, vpbroadcastq(thanks to Gabest).
-* 2013/May/30 ver 4.00 support AVX2, VEX-encoded GPR-instructions
-* 2013/Mar/27 ver 3.80 support mov(reg, "label");
-* 2013/Mar/13 ver 3.76 add cqo(), jcxz(), jecxz(), jrcxz()
-* 2013/Jan/15 ver 3.75 add setSize() to modify generated code
-* 2013/Jan/12 ver 3.74 add CodeGenerator::reset() ; add Allocator::useProtect()
-* 2013/Jan/06 ver 3.73 use unordered_map if possible
-* 2012/Dec/04 ver 3.72 eax, ebx, ... are member variables of CodeGenerator(revert), Xbyak::util::eax, ... are static const.
-* 2012/Nov/17 ver 3.71 and_(), or_(), xor_(), not_() are available if XBYAK_NO_OP_NAMES is not defined.
-* 2012/Nov/17 change eax, ebx, ptr and so on in CodeGenerator as static member and alias of them are defined in Xbyak::util.
-* 2012/Nov/09 ver 3.70 XBYAK_NO_OP_NAMES macro is added to use and_() instead of and() (thanks to Mattias)
-* 2012/Nov/01 ver 3.62 add fwait/fnwait/finit/fninit
-* 2012/Nov/01 ver 3.61 add fldcw/fstcw
-* 2012/May/03 ver 3.60 change interface of Allocator
-* 2012/Mar/23 ver 3.51 fix userPtr mode
-* 2012/Mar/19 ver 3.50 support AutoGrow mode
-* 2011/Nov/09 ver 3.05 fix bit property of rip addresing / support movsxd
-* 2011/Aug/15 ver 3.04 fix dealing with imm8 such as add(dword [ebp-8], 0xda); (thanks to lolcat)
-* 2011/Jun/16 ver 3.03 fix __GNUC_PREREQ macro for Mac gcc(thanks to t_teruya)
-* 2011/Apr/28 ver 3.02 do not use xgetbv on Mac gcc
-* 2011/May/24 ver 3.01 fix typo of OSXSAVE
-* 2011/May/23 ver 3.00 add vcmpeqps and so on
-* 2011/Feb/16 ver 2.994 beta add vmovq for 32-bit mode(I forgot it)
-* 2011/Feb/16 ver 2.993 beta remove cvtReg to avoid thread unsafe
-* 2011/Feb/10 ver 2.992 beta support one argument syntax for fadd like nasm
-* 2011/Feb/07 ver 2.991 beta fix pextrw reg, xmm, imm(Thanks to Gabest)
-* 2011/Feb/04 ver 2.99 beta support AVX
-* 2010/Dec/08 ver 2.31 fix ptr [rip + 32bit offset], support rdtscp
-* 2010/Oct/19 ver 2.30 support pclmulqdq, aesdec, aesdeclast, aesenc, aesenclast, aesimc, aeskeygenassist
-* 2010/Jun/07 ver 2.29 fix call(<label>)
-* 2010/Jun/17 ver 2.28 move some member functions to public
-* 2010/Jun/01 ver 2.27 support encoding of mov(reg64, imm) like yasm(not nasm)
-* 2010/May/24 ver 2.26 fix sub(rsp, 1000)
-* 2010/Apr/26 ver 2.25 add jc/jnc(I forgot to implement them...)
-* 2010/Apr/16 ver 2.24 change the prototype of rewrite() method
-* 2010/Apr/15 ver 2.23 fix align() and xbyak_util.h for Mac
-* 2010/Feb/16 ver 2.22 fix inLocalLabel()/outLocalLabel()
-* 2009/Dec/09 ver 2.21 support cygwin(gcc 4.3.2)
-* 2009/Nov/28 support a part of FPU
-* 2009/Jun/25 fix mov(qword[rax], imm); (thanks to Martin)
-* 2009/Mar/10 fix redundant REX.W prefix on jmp/call reg64
-* 2009/Feb/24 add movq reg64, mmx/xmm; movq mmx/xmm, reg64
-* 2009/Feb/13 movd(xmm7, dword[eax]) drops 0x66 prefix (thanks to Gabest)
-* 2008/Dec/30 fix call in short relative address(thanks to kato san)
-* 2008/Sep/18 support @@, @f, @b and localization of label(thanks to nobu-q san)
-* 2008/Sep/18 support (ptr[rip + 32bit offset]) (thanks to Dango-Chu san)
-* 2008/Jun/03 fix align(). mov(ptr[eax],1) throws ERR_MEM_SIZE_IS_NOT_SPECIFIED.
-* 2008/Jun/02 support memory interface allocated by user
-* 2008/May/26 fix protect() to avoid invalid setting(thanks to shinichiro_h san)
-* 2008/Apr/30 add cmpxchg16b, cdqe
-* 2008/Apr/29 support x64
-* 2008/Apr/14 code refactoring
-* 2008/Mar/12 add bsr/bsf
-* 2008/Feb/14 fix output of sub eax, 1234 (thanks to Robert)
-* 2007/Nov/5  support lock, xadd, xchg
-* 2007/Nov/2  support SSSE3/SSE4 (thanks to Dango-Chu san)
-* 2007/Feb/4  fix the bug that exception doesn't occur under the condition which the offset of jmp mnemonic without T_NEAR is over 127.
-* 2007/Jan/21 fix the bug to create address like [disp] select smaller representation for mov (eax|ax|al, [disp])
-* 2007/Jan/4  first version
+[BSD-3-Clause License](http://opensource.org/licenses/BSD-3-Clause)

 ## Author
-MITSUNARI Shigeo(herumi@nifty.com)
+
+#### 光成滋生 Mitsunari Shigeo
+ [GitHub](https://github.com/herumi) | [Website (Japanese)](http://herumi.in.coocan.jp/) | [herumi@nifty.com](mailto:herumi@nifty.com)

 ## Sponsors welcome
 [GitHub Sponsor](https://github.com/sponsors/herumi)
+
+<!----------------------------------------------------------------------------->
+
+[Badge Build]: https://github.com/herumi/xbyak/actions/workflows/main.yml/badge.svg
+[Build Status]: https://github.com/herumi/xbyak/actions/workflows/main.yml
+
+[License]: COPYRIGHT
+
+[Changelog]: doc/changelog.md
+[Install]: doc/install.md
+[Usage]: doc/usage.md
+
--- a/readme.txt
+++ b/readme.txt
@ -1,5 +1,5 @@

-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 5.991
+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 6.06

 -----------------------------------------------------------------------------
 ◎概要
@ -277,6 +277,24 @@ L(label);
 assert(label.getAddress(), getCurr());
 ```

+4. farジャンプ
+
+`jmp(mem, T_FAR)`, `call(mem, T_FAR)`, `retf()`をサポートします。
+サイズを明示するために`ptr`の代わりに`word|dword|qword`を利用してください。
+
+32bit
+```
+jmp(word[eax], T_FAR);  // jmp m16:16(FF /5)
+jmp(dword[eax], T_FAR); // jmp m16:32(FF /5)
+```
+
+64bit
+```
+jmp(word[rax], T_FAR);  // jmp m16:16(FF /5)
+jmp(dword[rax], T_FAR); // jmp m16:32(FF /5)
+jmp(qword[rax], T_FAR); // jmp m16:64(REX.W FF /5)
+```
+
 ・Xbyak::CodeGenerator()コンストラクタインタフェース

@param maxSize [in] コード生成最大サイズ(デフォルト4096byte)
@ -382,6 +400,21 @@ sample/{echo,hello}.bfは http://www.kmonos.net/alang/etc/brainfuck.php から
 -----------------------------------------------------------------------------
 ◎履歴

+2022/06/01 ver 6.06 Cpu::TypeクラスのリファクタリングとXBYAK_USE_MEMFDが定義されたときのMmapAllocatorの改善
+2022/05/20 ver 6.052 Cpu::operator==()を正しく定義
+2022/05/13 ver 6.051 XYBAK_NO_EXCEPTIONを定義したときのCpuクラスのコンパイルエラー修正
+2022/05/12 ver 6.05 movdiri, movdir64b, clwb, cldemoteを追加
+2022/04/05 ver 6.04 tpause, umonitor, umwaitを追加
+2022/03/08 ver 6.03 MmapAllocatorがmemfd用のユーザ定義文字列をサポート
+2022/01/28 ver 6.02 dispacementの32bit範囲チェックの厳密化
+2021/12/14 ver 6.01 T_FAR jump/callとretfをサポート
+2021/09/14 ver 6.00 AVX512-FP16を完全サポート
+2021/09/09 ver 5.997 vrndscale*を{sae}をサポートするよう修正
+2021/09/03 ver 5.996 v{add,sub,mul,div,max,min}{sd,ss}をT_rd_saeなどをサポートするよう修正
+2021/08/15 ver 5.995 Linux上でXBYAK_USE_MEMFDが定義されたなら/proc/self/mapsにラベル追加
+2021/06/17 ver 5.994 マスクレジスタ用のvcmpXX{ps,pd,ss,sd}のalias追加
+2021/06/06 ver 5.993 gather/scatterのレジスタの組み合わせの厳密なチェック
+2021/05/09 ver 5.992 endbr32とendbr64のサポート
 2020/11/16 ver 5.991 g++-5のC++14でconstexpr機能の抑制
 2020/10/19 ver 5.99 VNNI命令サポート(Thanks to akharito)
 2020/10/17 ver 5.98 [scale * reg]のサポート
--- a/sample/Makefile
+++ b/sample/Makefile
@ -37,6 +37,7 @@ endif

 ifneq ($(OS),mac)
 TARGET += static_buf64
+TARGET += memfd
 endif


@ -51,7 +52,7 @@ all: $(TARGET)

 CFLAGS_WARN=-Wall -Wextra -Wformat=2 -Wcast-qual -Wcast-align -Wwrite-strings -Wfloat-equal -Wpointer-arith #-pedantic

-CFLAGS=-g -O2 -fomit-frame-pointer -Wall -I../ $(CFLAGS_WARN)
+CFLAGS=-g -O2 -fomit-frame-pointer -Wall -I../ $(CFLAGS_WARN) $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS)

 test:
 	$(CXX) $(CFLAGS) test0.cpp -o $@ -m32
@ -95,6 +96,8 @@ jmp_table:
 	$(CXX) $(CFLAGS) jmp_table.cpp -o $@ -m32
 jmp_table64:
 	$(CXX) $(CFLAGS) jmp_table.cpp -o $@ -m64
+memfd:
+	$(CXX) $(CFLAGS) memfd.cpp -o $@ -m64
 profiler: profiler.cpp ../xbyak/xbyak_util.h
 	$(CXX) $(CFLAGS) profiler.cpp -o $@
 profiler-vtune: profiler.cpp ../xbyak/xbyak_util.h
@ -121,3 +124,4 @@ test_util : test_util.cpp $(XBYAK_INC) ../xbyak/xbyak_util.h
 test_util2 : test_util.cpp $(XBYAK_INC) ../xbyak/xbyak_util.h
 jmp_table: jmp_table.cpp $(XBYAK_INC)
 jmp_table64: jmp_table.cpp $(XBYAK_INC)
+memfd: memfd.cpp $(XBYAK_INC)
--- a/sample/memfd.cpp
+++ b/sample/memfd.cpp
@ -0,0 +1,39 @@
+/*
+	a sample to use MmapAllocator with an user-defined name
+	cat /proc/`psidof ./memfd`/maps
+
+7fca70b44000-7fca70b4a000 rw-p 00000000 00:00 0
+7fca70b67000-7fca70b68000 rwxs 00000000 00:05 19960170                   /memfd:xyz (deleted)
+7fca70b68000-7fca70b69000 rwxs 00000000 00:05 19960169                   /memfd:abc (deleted)
+7fca70b69000-7fca70b6a000 r--p 00029000 103:03 19136541                  /lib/x86_64-linux-gnu/ld-2.27.so
+7fca70b6a000-7fca70b6b000 rw-p 0002a000 103:03 19136541                  /lib/x86_64-linux-gnu/ld-2.27.so
+*/
+#define XBYAK_USE_MEMFD
+#include <xbyak/xbyak.h>
+#include <fstream>
+
+class Code : Xbyak::MmapAllocator, public Xbyak::CodeGenerator {
+public:
+    Code(const char *name, int v)
+        : Xbyak::MmapAllocator(name)
+        , Xbyak::CodeGenerator(4096, nullptr, this /* specify external MmapAllocator */)
+    {
+        mov(eax, v);
+        ret();
+    }
+};
+
+int main()
+{
+    Code c1("Xbyak::abc", 123);
+    Code c2("Xbyak::xyz", 456);
+    printf("c1 %d\n", c1.getCode<int (*)()>()());
+    printf("c2 %d\n", c2.getCode<int (*)()>()());
+	std::ifstream ifs("/proc/self/maps", std::ios::binary);
+	if (ifs) {
+		std::string line;
+		while (std::getline(ifs, line)) {
+			printf("%s\n", line.c_str());
+		}
+	}
+}
--- a/sample/test_util.cpp
+++ b/sample/test_util.cpp
@ -7,14 +7,13 @@ struct PopCountTest : public Xbyak::CodeGenerator {
 	PopCountTest(int n)
 		: Xbyak::CodeGenerator(4096, Xbyak::DontSetProtectRWE)
 	{
-ret();
 		mov(eax, n);
 		popcnt(eax, eax);
 		ret();
 	}
 };

-void putCPUinfo()
+void putCPUinfo(bool onlyCpuidFeature)
 {
 	using namespace Xbyak::util;
 	Cpu cpu;
@ -35,8 +34,6 @@ void putCPUinfo()
 		{ Cpu::tPOPCNT, "popcnt" },
 		{ Cpu::t3DN, "3dn" },
 		{ Cpu::tE3DN, "e3dn" },
-		{ Cpu::tSSE4a, "sse4a" },
-		{ Cpu::tSSE5, "sse5" },
 		{ Cpu::tAESNI, "aesni" },
 		{ Cpu::tRDTSCP, "rdtscp" },
 		{ Cpu::tOSXSAVE, "osxsave(xgetvb)" },
@ -85,11 +82,19 @@ void putCPUinfo()
 		{ Cpu::tAMX_INT8, "amx(int8)" },
 		{ Cpu::tAMX_BF16, "amx(bf16)" },
 		{ Cpu::tAVX_VNNI, "avx_vnni" },
+		{ Cpu::tAVX512_FP16, "avx512_fp16" },
+		{ Cpu::tWAITPKG, "waitpkg" },
+		{ Cpu::tCLFLUSHOPT, "clflushopt" },
+		{ Cpu::tCLDEMOTE, "cldemote" },
+		{ Cpu::tMOVDIRI, "movdiri" },
+		{ Cpu::tMOVDIR64B, "movdir64b" },
+		{ Cpu::tCLZERO, "clzero" },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		if (cpu.has(tbl[i].type)) printf(" %s", tbl[i].str);
 	}
 	printf("\n");
+	if (onlyCpuidFeature) return;
 	if (cpu.has(Cpu::tPOPCNT)) {
 		const int n = 0x12345678; // bitcount = 13
 		const int ok = 13;
@ -123,12 +128,15 @@ void putCPUinfo()
 	printf("CoreLevel=%u\n", cpu.getNumCores(Xbyak::util::CoreLevel));
 }

-int main()
+int main(int argc, char *argv[])
 {
+	bool onlyCpuidFeature = argc == 2 && strcmp(argv[1], "-cpuid") == 0;
+	if (!onlyCpuidFeature) {
 #ifdef XBYAK32
-	puts("32bit");
+		puts("32bit");
 #else
-	puts("64bit");
+		puts("64bit");
 #endif
-	putCPUinfo();
+	}
+	putCPUinfo(onlyCpuidFeature);
 }
--- a/test/Makefile
+++ b/test/Makefile
@ -1,4 +1,4 @@
-TARGET = make_nm normalize_prefix bad_address misc cvt_test cvt_test32 noexception
+TARGET = make_nm normalize_prefix bad_address misc cvt_test cvt_test32 noexception misc32
 XBYAK_INC=../xbyak/xbyak.h
 UNAME_S=$(shell uname -s)
 BIT=32
@ -22,7 +22,7 @@ all: $(TARGET)

 CFLAGS_WARN=-Wall -Wextra -Wformat=2 -Wcast-qual -Wcast-align -Wwrite-strings -Wfloat-equal -Wpointer-arith

-CFLAGS=-O2 -fomit-frame-pointer -Wall -fno-operator-names -I../ -I./ $(CFLAGS_WARN) #-std=c++0x
+CFLAGS=-O2 -fomit-frame-pointer -Wall -fno-operator-names -I../ -I./ $(CFLAGS_WARN) $(CXXFLAGS) $(CPPFLAGS) $(LDFLAGS) #-std=c++0x
 make_nm:
 	$(CXX) $(CFLAGS) make_nm.cpp -o $@
 normalize_prefix: normalize_prefix.cpp ../xbyak/xbyak.h
@ -41,6 +41,8 @@ bad_address: bad_address.cpp ../xbyak/xbyak.h
 	$(CXX) $(CFLAGS) bad_address.cpp -o $@
 misc: misc.cpp ../xbyak/xbyak.h
 	$(CXX) $(CFLAGS) misc.cpp -o $@
+misc32: misc.cpp ../xbyak/xbyak.h
+	$(CXX) $(CFLAGS) misc.cpp -o $@ -DXBYAK32
 cvt_test: cvt_test.cpp ../xbyak/xbyak.h
 	$(CXX) $(CFLAGS) $< -o $@
 cvt_test32: cvt_test.cpp ../xbyak/xbyak.h
@ -62,6 +64,7 @@ ifneq ($(ONLY_64BIT),1)
 endif
 	./bad_address
 	./misc
+	./misc32
 	./cvt_test
 ifeq ($(BIT),64)
 	./test_address.sh 64
@ -95,7 +98,7 @@ test:
 	$(MAKE) test_avx512

 clean:
-	rm -rf *.o $(TARGET) lib_run nm.cpp nm_frame make_512
+	$(RM) a.asm *.lst *.obj *.o $(TARGET) lib_run nm.cpp nm_frame make_512

 lib_run: lib_test.cpp lib_run.cpp lib.h
 	$(CXX) $(CFLAGS) lib_run.cpp lib_test.cpp -o lib_run
--- a/test/address.cpp
+++ b/test/address.cpp
@ -31,7 +31,7 @@ void genVsib(bool isJIT)
 		"xmm3 * 8 + edi + 123",
 		"xmm2 * 2 + 5",
 		"eax + xmm0",
-		"esp + xmm4",
+		"esp + xmm2",
 	};
 	const char *vm32yTbl[] = {
 		"ymm0",
@ -42,7 +42,7 @@ void genVsib(bool isJIT)
 		"ymm3 * 8 + edi + 123",
 		"ymm2 * 2 + 5",
 		"eax + ymm0",
-		"esp + ymm4",
+		"esp + ymm2",
 	};
 	genVsibSub(isJIT, "vgatherdpd", vm32xTbl, NUM_OF_ARRAY(vm32xTbl));
 	genVsibSub(isJIT, "vgatherqpd", vm32yTbl, NUM_OF_ARRAY(vm32yTbl));
@ -93,7 +93,7 @@ void genAddress(bool isJIT, const char regTbl[][5], size_t regTblNum)
 					}
 					if (isFirst) {
 						if (isJIT) printf("(void*)");
-						printf("0x%08X", disp);
+						printf("%d", disp);
 					} else {
 						if (disp >= 0) {
 							putchar('+');
--- a/test/jmp.cpp
+++ b/test/jmp.cpp
@ -1383,3 +1383,18 @@ CYBOZU_TEST_AUTO(setDefaultJmpNEAR)
 		}
 	}
 }
+
+CYBOZU_TEST_AUTO(ambiguousFarJmp)
+{
+	struct Code : Xbyak::CodeGenerator {
+#ifdef XBYAK32
+		void genJmp() { jmp(ptr[eax], T_FAR); }
+		void genCall() { call(ptr[eax], T_FAR); }
+#else
+		void genJmp() { jmp(ptr[rax], T_FAR); }
+		void genCall() { call(ptr[rax], T_FAR); }
+#endif
+	} code;
+	CYBOZU_TEST_EXCEPTION(code.genJmp(), std::exception);
+	CYBOZU_TEST_EXCEPTION(code.genCall(), std::exception);
+}
--- a/test/make_512.cpp
+++ b/test/make_512.cpp
@ -1366,6 +1366,8 @@ public:
 				put(p, _ZMM, _ZMM, mem | _MEM);
 			}
 		}
+		put("vaddss", XMM, _XMM, XMM_ER);
+		put("vaddsd", XMM, _XMM, XMM_ER);
 #endif
 	}
 	void putAVX1()
@ -1949,14 +1951,16 @@ public:
 		put("vrndscalepd", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
 		put("vrndscalepd", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
 		put("vrndscalepd", ZMM_KZ, _ZMM | _MEM | M_1to8, IMM8);
+		put("vrndscalepd", ZMM_KZ, _ZMM | ZMM_SAE, IMM8);

 		put("vrndscaleps", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
 		put("vrndscaleps", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
 		put("vrndscaleps", ZMM_KZ, _ZMM | _MEM | M_1to16, IMM8);
+		put("vrndscaleps", ZMM_KZ, _ZMM | ZMM_SAE, IMM8);

-		put("vrndscalesd", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
+		put("vrndscalesd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);

-		put("vrndscaless", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
+		put("vrndscaless", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);

 		put("vscalefpd", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
 		put("vscalefpd", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
--- a/test/make_nm.cpp
+++ b/test/make_nm.cpp
@ -179,6 +179,19 @@ class Test {
 			printf("\n");
 		}
 	}
+	void put(const char *nm, const char *para1, uint64_t op2, const char *para3) const
+	{
+		for (int j = 0; j < bitEnd; j++) {
+			if ((op2 & (1ULL << j)) == 0) continue;
+			printf("%s ", nm);
+			if (isXbyak_) printf("(");
+			printf("%s", para1);
+			if (!(op2 & NOPARA)) printf(", %s", get(1ULL << j));
+			printf(", %s", para3);
+			if (isXbyak_) printf("); dump();");
+			printf("\n");
+		}
+	}
 	const char *get(uint64_t type) const
 	{
 		int idx = (rand() / 31) & 7;
@ -499,6 +512,7 @@ class Test {
 			"cmpsb",
 			"cmpsw",
 			"cmpsd",
+			"hlt",
 			"int3",
 			"leave",
 			"lodsb",
@ -623,6 +637,7 @@ class Test {
 				"fstsw",
 				"fnstsw",
 				"fxrstor",
+				"clwb",
 			};
 			for (size_t i = 0; i < NUM_OF_ARRAY(memTbl); i++) {
 				put(memTbl[i], MEM);
@ -685,6 +700,24 @@ class Test {
 			puts("pshufb xmm14, [rel label0]");
 		}
 #endif
+#endif
+	}
+	void putFarJmp() const
+	{
+#ifdef XBYAK64
+		put("jmp", "word[rax],T_FAR", "far word [rax]");
+		put("jmp", "dword[rax],T_FAR", "far dword [rax]");
+		put("jmp", "qword[rax],T_FAR", "far qword [rax]");
+
+		put("call", "word[rax],T_FAR", "far word [rax]");
+		put("call", "dword[rax],T_FAR", "far dword [rax]");
+		put("call", "qword[rax],T_FAR", "far qword [rax]");
+#else
+		put("jmp", "dword[eax],T_FAR", "far dword [eax]");
+		put("jmp", "word[eax],T_FAR", "far word [eax]");
+
+		put("call", "dword[eax],T_FAR", "far dword [eax]");
+		put("call", "word[eax],T_FAR", "far word [eax]");
 #endif
 	}
 	void putMMX1() const
@ -1237,6 +1270,10 @@ class Test {
 		put("mov", REG64, "0x12345678", "0x12345678");
 		put("mov", REG64, "0xffffffff12345678LL", "0xffffffff12345678");
 		put("mov", REG32e|REG16|REG8|RAX|EAX|AX|AL, IMM);
+
+		put("mov", EAX, "ptr[(void*)-1]", "[-1]");
+		put("mov", EAX, "ptr[(void*)0x7fffffff]", "[0x7fffffff]");
+		put("mov", EAX, "ptr[(void*)0xffffffffffffffff]", "[0xffffffffffffffff]");
 	}
 	void putEtc() const
 	{
@ -1244,6 +1281,9 @@ class Test {
 			const char *p = "ret";
 			put(p);
 			put(p, IMM);
+			p = "retf";
+			put(p);
+			put(p, IMM);
 			p = "mov";
 			put(p, EAX|REG32|MEM|MEM_ONLY_DISP, REG32|EAX);
 			put(p, REG64|MEM|MEM_ONLY_DISP, REG64|RAX);
@ -1480,6 +1520,7 @@ class Test {
 		put("pextrq", REG64|MEM, XMM, IMM);
 		put("pinsrq", XMM, REG64|MEM, IMM);
 #endif
+
 	}
 	void putSHA() const
 	{
@ -2361,16 +2402,16 @@ public:
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
 			const char *name = p.name;
-			put(name, XMM, VM32X, XMM);
+			put(name, "xmm3", VM32X, "xmm5");
 			switch (p.mode) {
 			case y_vx_y:
-				put(name, YMM, VM32X, YMM);
+				put(name, "ymm3", VM32X, "ymm5");
 				break;
 			case y_vy_y:
-				put(name, YMM, VM32Y, YMM);
+				put(name, "ymm3", VM32Y, "ymm5");
 				break;
 			case x_vy_x:
-				put(name, XMM, VM32Y, XMM);
+				put(name, "xmm3", VM32Y, "xmm5");
 				break;
 			default:
 				printf("ERR mode=%d\n", p.mode);
@ -2516,6 +2557,7 @@ public:
 #else // USE_AVX

 		putJmp();
+		putFarJmp();

 #ifdef USE_YASM

--- a/test/misc.cpp
+++ b/test/misc.cpp
--- a/test/sf_test.cpp
+++ b/test/sf_test.cpp
@ -218,7 +218,7 @@ void check(int x, int y)
 	}
 }

-void verify(const Xbyak::uint8_t *f, int pNum)
+void verify(const uint8_t *f, int pNum)
 {
 	switch (pNum) {
 	case 0:
@ -264,7 +264,7 @@ void testAll()
 				}
 				for (int tNum = 0; tNum < maxNum; tNum++) {
 //					printf("pNum=%d, tNum=%d, stackSize=%d\n", pNum, tNum | opt, stackSize);
-					const Xbyak::uint8_t *f = code.getCurr();
+					const uint8_t *f = code.getCurr();
 					code.gen(pNum, tNum | opt, stackSize);
 					verify(f, pNum);
 					/*
--- a/xbyak/xbyak.h
+++ b/xbyak/xbyak.h
@ -95,6 +95,12 @@
 	#include <stdint.h>
 #endif

+// MFD_CLOEXEC defined only linux 3.17 or later.
+// Android wraps the memfd_create syscall from API version 30.
+#if !defined(MFD_CLOEXEC) || (defined(__ANDROID__) && __ANDROID_API__ < 30)
+	#undef XBYAK_USE_MEMFD
+#endif
+
 #if defined(_WIN64) || defined(__MINGW64__) || (defined(__CYGWIN__) && defined(__x86_64__))
 	#define XBYAK64_WIN
 #elif defined(__x86_64__)
@ -138,7 +144,7 @@ namespace Xbyak {

 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x5991 /* 0xABCD = A.BC(D) */
+	VERSION = 0x6060 /* 0xABCD = A.BC(D) */
 };

 #ifndef MIE_INTEGER_TYPE_DEFINED
@ -209,6 +215,7 @@ enum {
 	ERR_INVALID_MIB_ADDRESS,
 	ERR_X2APIC_IS_NOT_SUPPORTED,
 	ERR_NOT_SUPPORTED,
+	ERR_SAME_REGS_ARE_INVALID,
 	ERR_INTERNAL // Put it at last.
 };

@ -261,6 +268,7 @@ inline const char *ConvertErrorToString(int err)
 		"invalid mib address",
 		"x2APIC is not supported",
 		"not supported",
+		"same regs are invalid",
 		"internal error"
 	};
 	assert(ERR_INTERNAL + 1 == sizeof(errTbl) / sizeof(*errTbl));
@ -285,10 +293,10 @@ inline void SetError(int err) {
 inline void ClearError() {
 	local::GetErrorRef() = 0;
 }
-inline int GetError() { return local::GetErrorRef(); }
+inline int GetError() { return Xbyak::local::GetErrorRef(); }

-#define XBYAK_THROW(err) { local::SetError(err); return; }
-#define XBYAK_THROW_RET(err, r) { local::SetError(err); return r; }
+#define XBYAK_THROW(err) { Xbyak::local::SetError(err); return; }
+#define XBYAK_THROW_RET(err, r) { Xbyak::local::SetError(err); return r; }

 #else
 class Error : public std::exception {
@ -377,6 +385,7 @@ enum LabelMode {
 	custom allocator
 */
 struct Allocator {
+	explicit Allocator(const std::string& = "") {} // same interface with MmapAllocator
 	virtual uint8_t *alloc(size_t size) { return reinterpret_cast<uint8_t*>(AlignedMalloc(size, inner::ALIGN_PAGE_SIZE)); }
 	virtual void free(uint8_t *p) { AlignedFree(p); }
 	virtual ~Allocator() {}
@ -408,10 +417,21 @@ inline int getMacOsVersion()

 } // util
 #endif
-class MmapAllocator : Allocator {
-	typedef XBYAK_STD_UNORDERED_MAP<uintptr_t, size_t> SizeList;
-	SizeList sizeList_;
+class MmapAllocator : public Allocator {
+	struct Allocation {
+		size_t size;
+#if defined(XBYAK_USE_MEMFD)
+		// fd_ is only used with XBYAK_USE_MEMFD. We keep the file open
+		// during the lifetime of each allocation in order to support
+		// checkpoint/restore by unprivileged users.
+		int fd;
+#endif
+	};
+	const std::string name_; // only used with XBYAK_USE_MEMFD
+	typedef XBYAK_STD_UNORDERED_MAP<uintptr_t, Allocation> AllocationList;
+	AllocationList allocList_;
 public:
+	explicit MmapAllocator(const std::string& name = "xbyak") : name_(name) {}
 	uint8_t *alloc(size_t size)
 	{
 		const size_t alignedSizeM1 = inner::ALIGN_PAGE_SIZE - 1;
@ -427,21 +447,44 @@ public:
 		const int mojaveVersion = 18;
 		if (util::getMacOsVersion() >= mojaveVersion) mode |= MAP_JIT;
 #endif
-		void *p = mmap(NULL, size, PROT_READ | PROT_WRITE, mode, -1, 0);
-		if (p == MAP_FAILED) XBYAK_THROW_RET(ERR_CANT_ALLOC, 0)
+		int fd = -1;
+#if defined(XBYAK_USE_MEMFD)
+		fd = memfd_create(name_.c_str(), MFD_CLOEXEC);
+		if (fd != -1) {
+			mode = MAP_SHARED;
+			if (ftruncate(fd, size) != 0) {
+				close(fd);
+				XBYAK_THROW_RET(ERR_CANT_ALLOC, 0)
+			}
+		}
+#endif
+		void *p = mmap(NULL, size, PROT_READ | PROT_WRITE, mode, fd, 0);
+		if (p == MAP_FAILED) {
+			if (fd != -1) close(fd);
+			XBYAK_THROW_RET(ERR_CANT_ALLOC, 0)
+		}
 		assert(p);
-		sizeList_[(uintptr_t)p] = size;
+		Allocation &alloc = allocList_[(uintptr_t)p];
+		alloc.size = size;
+#if defined(XBYAK_USE_MEMFD)
+		alloc.fd = fd;
+#endif
 		return (uint8_t*)p;
 	}
 	void free(uint8_t *p)
 	{
 		if (p == 0) return;
-		SizeList::iterator i = sizeList_.find((uintptr_t)p);
-		if (i == sizeList_.end()) XBYAK_THROW(ERR_BAD_PARAMETER)
-		if (munmap((void*)i->first, i->second) < 0) XBYAK_THROW(ERR_MUNMAP)
-		sizeList_.erase(i);
+		AllocationList::iterator i = allocList_.find((uintptr_t)p);
+		if (i == allocList_.end()) XBYAK_THROW(ERR_BAD_PARAMETER)
+		if (munmap((void*)i->first, i->second.size) < 0) XBYAK_THROW(ERR_MUNMAP)
+#if defined(XBYAK_USE_MEMFD)
+		if (i->second.fd != -1) close(i->second.fd);
+#endif
+		allocList_.erase(i);
 	}
 };
+#else
+typedef Allocator MmapAllocator;
 #endif

 class Address;
@ -1557,6 +1600,7 @@ public:
 	enum LabelType {
 		T_SHORT,
 		T_NEAR,
+		T_FAR, // far jump
 		T_AUTO // T_SHORT if possible
 	};
 private:
@ -1605,6 +1649,11 @@ private:
 	{
 		return op1.isREG(i32e) && ((op2.isREG(i32e) && op1.getBit() == op2.getBit()) || op2.isMEM());
 	}
+	static inline bool isValidSSE(const Operand& op1)
+	{
+		// SSE instructions do not support XMM16 - XMM31
+		return !(op1.isXMM() && op1.getIdx() >= 16);
+	}
 	void rex(const Operand& op1, const Operand& op2 = Operand())
 	{
 		uint8_t rex = 0;
@ -1635,9 +1684,10 @@ private:
 		//
 		T_N_VL = 1 << 3, // N * (1, 2, 4) for VL
 		T_DUP = 1 << 4, // N = (8, 32, 64)
-		T_66 = 1 << 5,
-		T_F3 = 1 << 6,
-		T_F2 = 1 << 7,
+		T_66 = 1 << 5, // pp = 1
+		T_F3 = 1 << 6, // pp = 2
+		T_F2 = T_66 | T_F3, // pp = 3
+		T_ER_R = 1 << 7, // reg{er}
 		T_0F = 1 << 8,
 		T_0F38 = 1 << 9,
 		T_0F3A = 1 << 10,
@ -1658,11 +1708,17 @@ private:
 		T_MUST_EVEX = 1 << 25, // contains T_EVEX
 		T_B32 = 1 << 26, // m32bcst
 		T_B64 = 1 << 27, // m64bcst
+		T_B16 = T_B32 | T_B64, // m16bcst (Be careful)
 		T_M_K = 1 << 28, // mem{k}
 		T_VSIB = 1 << 29,
 		T_MEM_EVEX = 1 << 30, // use evex if mem
+		T_FP16 = 1 << 31, // avx512-fp16
+		T_MAP5 = T_FP16 | T_0F,
+		T_MAP6 = T_FP16 | T_0F38,
 		T_XXX
 	};
+	// T_66 = 1, T_F3 = 2, T_F2 = 3
+	uint32_t getPP(int type) const { return (type >> 5) & 3; }
 	void vex(const Reg& reg, const Reg& base, const Operand *v, int type, int code, bool x = false)
 	{
 		int w = (type & T_W1) ? 1 : 0;
@ -1671,7 +1727,7 @@ private:
 		bool b = base.isExtIdx();
 		int idx = v ? v->getIdx() : 0;
 		if ((idx | reg.getIdx() | base.getIdx()) >= 16) XBYAK_THROW(ERR_BAD_COMBINATION)
-		uint32_t pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0;
+		uint32_t pp = getPP(type);
 		uint32_t vvvv = (((~idx) & 15) << 3) | (is256 ? 4 : 0) | pp;
 		if (!b && !x && !w && (type & T_0F)) {
 			db(0xC5); db((r ? 0 : 0x80) | vvvv);
@ -1688,6 +1744,7 @@ private:
 	}
 	void verifyER(const Reg& r, int type) const
 	{
+		if ((type & T_ER_R) && r.isREG(32|64)) return;
 		if (((type & T_ER_X) && r.isXMM()) || ((type & T_ER_Y) && r.isYMM()) || ((type & T_ER_Z) && r.isZMM())) return;
 		XBYAK_THROW(ERR_ER_IS_INVALID)
 	}
@ -1702,9 +1759,9 @@ private:
 	{
 		if (!(type & (T_EVEX | T_MUST_EVEX))) XBYAK_THROW_RET(ERR_EVEX_IS_INVALID, 0)
 		int w = (type & T_EW1) ? 1 : 0;
-		uint32_t mm = (type & T_0F) ? 1 : (type & T_0F38) ? 2 : (type & T_0F3A) ? 3 : 0;
-		uint32_t pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0;
-
+		uint32_t mmm = (type & T_0F) ? 1 : (type & T_0F38) ? 2 : (type & T_0F3A) ? 3 : 0;
+		if (type & T_FP16) mmm |= 4;
+		uint32_t pp = getPP(type);
 		int idx = v ? v->getIdx() : 0;
 		uint32_t vvvv = ~idx;

@ -1727,7 +1784,7 @@ private:
 			VL = (std::max)((std::max)(reg.getBit(), base.getBit()), VL);
 			LL = (VL == 512) ? 2 : (VL == 256) ? 1 : 0;
 			if (b) {
-				disp8N = (type & T_B32) ? 4 : 8;
+				disp8N = ((type & T_B16) == T_B16) ? 2 : (type & T_B32) ? 4 : 8;
 			} else if (type & T_DUP) {
 				disp8N = VL == 128 ? 8 : VL == 256 ? 32 : 64;
 			} else {
@ -1746,7 +1803,7 @@ private:
 		if (aaa == 0) aaa = verifyDuplicate(base.getOpmaskIdx(), reg.getOpmaskIdx(), (v ? v->getOpmaskIdx() : 0), ERR_OPMASK_IS_ALREADY_SET);
 		if (aaa == 0) z = 0; // clear T_z if mask is not set
 		db(0x62);
-		db((R ? 0x80 : 0) | (X ? 0x40 : 0) | (B ? 0x20 : 0) | (Rp ? 0x10 : 0) | (mm & 3));
+		db((R ? 0x80 : 0) | (X ? 0x40 : 0) | (B ? 0x20 : 0) | (Rp ? 0x10 : 0) | mmm);
 		db((w == 1 ? 0x80 : 0) | ((vvvv & 15) << 3) | 4 | (pp & 3));
 		db((z ? 0x80 : 0) | ((LL & 3) << 5) | (b ? 0x10 : 0) | (Vp ? 8 : 0) | (aaa & 7));
 		db(code);
@ -1760,8 +1817,15 @@ private:
 	{
 		uint64_t disp64 = e.getDisp();
 #ifdef XBYAK64
+#ifdef XBYAK_OLD_DISP_CHECK
+		// treat 0xffffffff as 0xffffffffffffffff
 		uint64_t high = disp64 >> 32;
 		if (high != 0 && high != 0xFFFFFFFF) XBYAK_THROW(ERR_OFFSET_IS_TOO_BIG)
+#else
+		// displacement should be a signed 32-bit value, so also check sign bit
+		uint64_t high = disp64 >> 31;
+		if (high != 0 && high != 0x1FFFFFFFF) XBYAK_THROW(ERR_OFFSET_IS_TOO_BIG)
+#endif
 #endif
 		uint32_t disp = static_cast<uint32_t>(disp64);
 		const Reg& base = e.getBase();
@ -1862,6 +1926,7 @@ private:
 	template<class T>
 	void opJmp(T& label, LabelType type, uint8_t shortCode, uint8_t longCode, uint8_t longPref)
 	{
+		if (type == T_FAR) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (isAutoGrow() && size_ + 16 >= maxSize_) growMemory(); /* avoid splitting code of jmp */
 		size_t offset = 0;
 		if (labelMgr_.getOffset(&offset, label)) { /* label exists */
@ -1882,6 +1947,7 @@ private:
 	}
 	void opJmpAbs(const void *addr, LabelType type, uint8_t shortCode, uint8_t longCode, uint8_t longPref = 0)
 	{
+		if (type == T_FAR) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (isAutoGrow()) {
 			if (!isNEAR(type)) XBYAK_THROW(ERR_ONLY_T_NEAR_IS_SUPPORTED_IN_AUTO_GROW)
 			if (size_ + 16 >= maxSize_) growMemory();
@ -1894,6 +1960,16 @@ private:
 		}

 	}
+	void opJmpOp(const Operand& op, LabelType type, int ext)
+	{
+		const int bit = 16|i32e;
+		if (type == T_FAR) {
+			if (!op.isMEM(bit)) XBYAK_THROW(ERR_NOT_SUPPORTED)
+			opR_ModM(op, bit, ext + 1, 0xFF, NONE, NONE, false);
+		} else {
+			opR_ModM(op, bit, ext, 0xFF, NONE, NONE, true);
+		}
+	}
 	// reg is reg field of ModRM
 	// immSize is the size for immediate value
 	// disp8N = 0(normal), disp8N = 1(force disp32), disp8N = {2, 4, 8} ; compressed displacement
@ -1920,6 +1996,7 @@ private:
 	void opGen(const Operand& reg, const Operand& op, int code, int pref, bool isValid(const Operand&, const Operand&), int imm8 = NONE, int preCode = NONE)
 	{
 		if (isValid && !isValid(reg, op)) XBYAK_THROW(ERR_BAD_COMBINATION)
+		if (!isValidSSE(reg) || !isValidSSE(op)) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (pref != NONE) db(pref);
 		if (op.isMEM()) {
 			opModM(op.getAddress(), reg.getReg(), 0x0F, preCode, code, (imm8 != NONE) ? 1 : 0);
@ -1930,6 +2007,7 @@ private:
 	}
 	void opMMX_IMM(const Mmx& mmx, int imm8, int code, int ext)
 	{
+		if (!isValidSSE(mmx)) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (mmx.isXMM()) db(0x66);
 		opModR(Reg32(ext), mmx, 0x0F, code);
 		db(imm8);
@ -1940,6 +2018,7 @@ private:
 	}
 	void opMovXMM(const Operand& op1, const Operand& op2, int code, int pref)
 	{
+		if (!isValidSSE(op1) || !isValidSSE(op2)) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (pref != NONE) db(pref);
 		if (op1.isXMM() && op2.isMEM()) {
 			opModM(op2.getAddress(), op1.getReg(), 0x0F, code);
@ -1951,6 +2030,7 @@ private:
 	}
 	void opExt(const Operand& op, const Mmx& mmx, int code, int imm, bool hasMMX2 = false)
 	{
+		if (!isValidSSE(op) || !isValidSSE(mmx)) XBYAK_THROW(ERR_NOT_SUPPORTED)
 		if (hasMMX2 && op.isREG(i32e)) { /* pextrw is special */
 			if (mmx.isXMM()) db(0x66);
 			opModR(op.getReg(), mmx, 0x0F, 0xC5); db(imm);
@ -2211,11 +2291,15 @@ private:
 	{
 		if (!(x.isXMM() && op.is(Operand::XMM | Operand::YMM | Operand::MEM)) && !(x.isYMM() && op.is(Operand::ZMM | Operand::MEM))) XBYAK_THROW(ERR_BAD_COMBINATION)
 	}
+	void opCvt(const Xmm& x, const Operand& op, int type, int code)
+	{
+		Operand::Kind kind = x.isXMM() ? (op.isBit(256) ? Operand::YMM : Operand::XMM) : Operand::ZMM;
+		opVex(x.copyAndSetKind(kind), &xm0, op, type, code);
+	}
 	void opCvt2(const Xmm& x, const Operand& op, int type, int code)
 	{
 		checkCvt2(x, op);
-		Operand::Kind kind = x.isXMM() ? (op.isBit(256) ? Operand::YMM : Operand::XMM) : Operand::ZMM;
-		opVex(x.copyAndSetKind(kind), &xm0, op, type, code);
+		opCvt(x, op, type, code);
 	}
 	void opCvt3(const Xmm& x1, const Xmm& x2, const Operand& op, int type, int type64, int type32, uint8_t code)
 	{
@ -2224,6 +2308,18 @@ private:
 		const Operand *p = op.isREG() ? &x : &op;
 		opVex(x1, &x2, *p, type | (op.isBit(64) ? type64 : type32), code);
 	}
+	// (x, x/y/xword/yword), (y, z/m)
+	void checkCvt4(const Xmm& x, const Operand& op) const
+	{
+		if (!(x.isXMM() && op.is(Operand::XMM | Operand::YMM | Operand::MEM) && op.isBit(128|256)) && !(x.isYMM() && op.is(Operand::ZMM | Operand::MEM))) XBYAK_THROW(ERR_BAD_COMBINATION)
+	}
+	// (x, x/y/z/xword/yword/zword)
+	void opCvt5(const Xmm& x, const Operand& op, int type, int code)
+	{
+		if (!(x.isXMM() && op.isBit(128|256|512))) XBYAK_THROW(ERR_BAD_COMBINATION)
+		Operand::Kind kind = op.isBit(128) ? Operand::XMM : op.isBit(256) ? Operand::YMM : Operand::ZMM;
+		opVex(x.copyAndSetKind(kind), &xm0, op, type, code);
+	}
 	const Xmm& cvtIdx0(const Operand& x) const
 	{
 		return x.isZMM() ? zm0 : x.isYMM() ? ym0 : xm0;
@ -2261,7 +2357,11 @@ private:
 			}
 			if (!isOK) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
 		}
-		opAVX_X_X_XM(isAddrYMM ? Ymm(x1.getIdx()) : x1, isAddrYMM ? Ymm(x2.getIdx()) : x2, addr, type, code);
+		int i1 = x1.getIdx();
+		int i2 = regExp.getIndex().getIdx();
+		int i3 = x2.getIdx();
+		if (i1 == i2 || i1 == i3 || i2 == i3) XBYAK_THROW(ERR_SAME_REGS_ARE_INVALID);
+		opAVX_X_X_XM(isAddrYMM ? Ymm(i1) : x1, isAddrYMM ? Ymm(i3) : x2, addr, type, code);
 	}
 	enum {
 		xx_yy_zz = 0,
@ -2284,7 +2384,12 @@ private:
 	void opGather2(const Xmm& x, const Address& addr, int type, uint8_t code, int mode)
 	{
 		if (x.hasZero()) XBYAK_THROW(ERR_INVALID_ZERO)
-		checkGather2(x, addr.getRegExp().getIndex(), mode);
+		const RegExp& regExp = addr.getRegExp();
+		checkGather2(x, regExp.getIndex(), mode);
+		int maskIdx = x.getOpmaskIdx();
+		if ((type & T_M_K) && addr.getOpmaskIdx()) maskIdx = addr.getOpmaskIdx();
+		if (maskIdx == 0) XBYAK_THROW(ERR_K0_IS_INVALID);
+		if (!(type & T_M_K) && x.getIdx() == regExp.getIndex().getIdx()) XBYAK_THROW(ERR_SAME_REGS_ARE_INVALID);
 		opVex(x, 0, addr, type, code);
 	}
 	/*
@ -2424,13 +2529,13 @@ public:

 	// set default type of `jmp` of undefined label to T_NEAR
 	void setDefaultJmpNEAR(bool isNear) { isDefaultJmpNEAR_ = isNear; }
-	void jmp(const Operand& op) { opR_ModM(op, BIT, 4, 0xFF, NONE, NONE, true); }
+	void jmp(const Operand& op, LabelType type = T_AUTO) { opJmpOp(op, type, 4); }
 	void jmp(std::string label, LabelType type = T_AUTO) { opJmp(label, type, 0xEB, 0xE9, 0); }
 	void jmp(const char *label, LabelType type = T_AUTO) { jmp(std::string(label), type); }
 	void jmp(const Label& label, LabelType type = T_AUTO) { opJmp(label, type, 0xEB, 0xE9, 0); }
 	void jmp(const void *addr, LabelType type = T_AUTO) { opJmpAbs(addr, type, 0xEB, 0xE9); }

-	void call(const Operand& op) { opR_ModM(op, 16 | i32e, 2, 0xFF, NONE, NONE, true); }
+	void call(const Operand& op, LabelType type = T_AUTO) { opJmpOp(op, type, 2); }
 	// call(string label), not const std::string&
 	void call(std::string label) { opJmp(label, T_NEAR, 0, 0xE8, 0); }
 	void call(const char *label) { call(std::string(label)); }
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@ -1,4 +1,4 @@
-const char *getVersionString() const { return "5.991"; }
+const char *getVersionString() const { return "6.06"; }
 void adc(const Operand& op, uint32_t imm) { opRM_I(op, imm, 0x10, 2); }
 void adc(const Operand& op1, const Operand& op2) { opRM_RM(op1, op2, 0x10); }
 void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); }
@ -57,9 +57,11 @@ void cbw() { db(0x66); db(0x98); }
 void cdq() { db(0x99); }
 void clc() { db(0xF8); }
 void cld() { db(0xFC); }
+void cldemote(const Address& addr) { opMIB(addr, eax, 0x0F, 0x1C); }
 void clflush(const Address& addr) { opModM(addr, Reg32(7), 0x0F, 0xAE); }
 void clflushopt(const Address& addr) { db(0x66); opModM(addr, Reg32(7), 0x0F, 0xAE); }
 void cli() { db(0xFA); }
+void clwb(const Address& addr) { db(0x66); opMIB(addr, esi, 0x0F, 0xAE); }
 void clzero() { db(0x0F); db(0x01); db(0xFC); }
 void cmc() { db(0xF5); }
 void cmova(const Reg& reg, const Operand& op) { opModRM(reg, op, op.isREG(16 | i32e), op.isMEM(), 0x0F, 0x40 | 7); }//-V524
@ -172,6 +174,8 @@ void divss(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5E, 0xF3, isXMM
 void dppd(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x41, 0x66, isXMM_XMMorMEM, static_cast<uint8_t>(imm), 0x3A); }
 void dpps(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op, 0x40, 0x66, isXMM_XMMorMEM, static_cast<uint8_t>(imm), 0x3A); }
 void emms() { db(0x0F); db(0x77); }
+void endbr32() { db(0xF3); db(0x0F); db(0x1E); db(0xFB); }
+void endbr64() { db(0xF3); db(0x0F); db(0x1E); db(0xFA); }
 void enter(uint16_t x, uint8_t y) { db(0xC8); dw(x); db(y); }
 void extractps(const Operand& op, const Xmm& xmm, uint8_t imm) { opExt(op, xmm, 0x17, imm); }
 void f2xm1() { db(0xD9); db(0xF0); }
@ -321,6 +325,7 @@ void gf2p8affineqb(const Xmm& xmm, const Operand& op, int imm) { opGen(xmm, op,
 void gf2p8mulb(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0xCF, 0x66, isXMM_XMMorMEM, NONE, 0x38); }
 void haddpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x7C, 0x66, isXMM_XMMorMEM); }
 void haddps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x7C, 0xF2, isXMM_XMMorMEM); }
+void hlt() { db(0xF4); }
 void hsubpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x7D, 0x66, isXMM_XMMorMEM); }
 void hsubps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x7D, 0xF2, isXMM_XMMorMEM); }
 void idiv(const Operand& op) { opR_ModM(op, 0, 7, 0xF6); }
@ -498,6 +503,8 @@ void movd(const Mmx& mmx, const Address& addr) { if (mmx.isXMM()) db(0x66); opMo
 void movd(const Mmx& mmx, const Reg32& reg) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x6E); }
 void movd(const Reg32& reg, const Mmx& mmx) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x7E); }
 void movddup(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x12, 0xF2, isXMM_XMMorMEM, NONE, NONE); }
+void movdir64b(const Reg& reg, const Address& addr) { db(0x66); opModM(addr, reg.cvt32(), 0x0F, 0x38, 0xF8); }
+void movdiri(const Address& addr, const Reg32e& reg) { opModM(addr, reg, 0x0F, 0x38, 0xF9); }
 void movdq2q(const Mmx& mmx, const Xmm& xmm) { db(0xF2); opModR(mmx, xmm, 0x0F, 0xD6); }
 void movdqa(const Address& addr, const Xmm& xmm) { db(0x66); opModM(addr, xmm, 0x0F, 0x7F); }
 void movdqa(const Xmm& xmm, const Operand& op) { opMMX(xmm, op, 0x6F, 0x66); }
@ -717,6 +724,7 @@ void repne() { db(0xF2); }
 void repnz() { db(0xF2); }
 void repz() { db(0xF3); }
 void ret(int imm = 0) { if (imm) { db(0xC2); dw(imm); } else { db(0xC3); } }
+void retf(int imm = 0) { if (imm) { db(0xCA); dw(imm); } else { db(0xCB); } }
 void rol(const Operand& op, const Reg8& _cl) { opShift(op, _cl, 0); }
 void rol(const Operand& op, int imm) { opShift(op, imm, 0); }
 void ror(const Operand& op, const Reg8& _cl) { opShift(op, _cl, 1); }
@ -809,18 +817,21 @@ void subsd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5C, 0xF2, isXMM
 void subss(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5C, 0xF3, isXMM_XMMorMEM); }
 void sysenter() { db(0x0F); db(0x34); }
 void sysexit() { db(0x0F); db(0x35); }
+void tpause(const Reg32& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) db(0x66); db(0x0F); db(0xAE); setModRM(3, 6, idx); }
 void tzcnt(const Reg&reg, const Operand& op) { opSp1(reg, op, 0xF3, 0x0F, 0xBC); }
 void ucomisd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x2E, 0x66, isXMM_XMMorMEM); }
 void ucomiss(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x2E, 0x100, isXMM_XMMorMEM); }
 void ud2() { db(0x0F); db(0x0B); }
+void umonitor(const Reg& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) int bit = r.getBit(); if (BIT != bit) { if ((BIT == 32 && bit == 16) || (BIT == 64 && bit == 32)) { db(0x67); } else { XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) } } db(0xF3); db(0x0F); db(0xAE); setModRM(3, 6, idx); }
+void umwait(const Reg32& r) { int idx = r.getIdx(); if (idx > 7) XBYAK_THROW(ERR_BAD_PARAMETER) db(0xF2); db(0x0F); db(0xAE); setModRM(3, 6, idx); }
 void unpckhpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x15, 0x66, isXMM_XMMorMEM); }
 void unpckhps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x15, 0x100, isXMM_XMMorMEM); }
 void unpcklpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x14, 0x66, isXMM_XMMorMEM); }
 void unpcklps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x14, 0x100, isXMM_XMMorMEM); }
 void vaddpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x58); }
 void vaddps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x58); }
-void vaddsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x58); }
-void vaddss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x58); }
+void vaddsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x58); }
+void vaddss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x58); }
 void vaddsubpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_66 | T_0F | T_YMM, 0xD0); }
 void vaddsubps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_F2 | T_0F | T_YMM, 0xD0); }
 void vaesdec(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_66 | T_0F38 | T_YMM | T_EVEX, 0xDE); }
@ -982,7 +993,7 @@ void vcvtpd2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_66 | T_
 void vcvtph2ps(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_0F38 | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y, 0x13); }
 void vcvtps2dq(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x5B); }
 void vcvtps2pd(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_0F | T_YMM | T_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 0x5A); }
-void vcvtps2ph(const Operand& op, const Xmm& x, uint8_t imm) { checkCvt1(x, op); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y, 0x1D, imm); }
+void vcvtps2ph(const Operand& op, const Xmm& x, uint8_t imm) { checkCvt1(x, op); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N8 | T_N_VL | T_SAE_Y | T_M_K, 0x1D, imm); }
 void vcvtsd2si(const Reg32& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F2 | T_W0 | T_EVEX | T_EW0 | T_N4 | T_ER_X, 0x2D); }
 void vcvtsd2ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX | T_ER_X, 0x5A); }
 void vcvtsi2sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_0F | T_F2 | T_EVEX, T_W1 | T_EW1 | T_ER_X | T_N8, T_W0 | T_EW0 | T_N4, 0x2A); }
@ -995,8 +1006,8 @@ void vcvttsd2si(const Reg32& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()
 void vcvttss2si(const Reg32& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F3 | T_W0 | T_EVEX | T_EW0 | T_SAE_X | T_N8, 0x2C); }
 void vdivpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x5E); }
 void vdivps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x5E); }
-void vdivsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x5E); }
-void vdivss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5E); }
+void vdivsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x5E); }
+void vdivss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x5E); }
 void vdppd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0, 0x41, imm); }
 void vdpps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x40, imm); }
 void vextractf128(const Operand& op, const Ymm& y, uint8_t imm) { if (!(op.isXMEM() && y.isYMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y, 0, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x19, imm); }
@ -1085,12 +1096,12 @@ void vmaskmovps(const Address& addr, const Xmm& x1, const Xmm& x2) { opAVX_X_X_X
 void vmaskmovps(const Xmm& x1, const Xmm& x2, const Address& addr) { opAVX_X_X_XM(x1, x2, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x2C); }
 void vmaxpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x5F); }
 void vmaxps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x5F); }
-void vmaxsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x5F); }
-void vmaxss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5F); }
+void vmaxsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x5F); }
+void vmaxss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x5F); }
 void vminpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x5D); }
 void vminps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x5D); }
-void vminsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x5D); }
-void vminss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5D); }
+void vminsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x5D); }
+void vminss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x5D); }
 void vmovapd(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX | T_M_K, 0x29); }
 void vmovapd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX, 0x28); }
 void vmovaps(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_0F | T_EW0 | T_YMM | T_EVEX | T_M_K, 0x29); }
@ -1136,8 +1147,8 @@ void vmovups(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_0F | T
 void vmpsadbw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x42, imm); }
 void vmulpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x59); }
 void vmulps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x59); }
-void vmulsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x59); }
-void vmulss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x59); }
+void vmulsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x59); }
+void vmulss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x59); }
 void vorpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x56); }
 void vorps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x56); }
 void vpabsb(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F38 | T_YMM | T_EVEX, 0x1C); }
@ -1320,8 +1331,8 @@ void vsqrtss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1,
 void vstmxcsr(const Address& addr) { opAVX_X_X_XM(xm3, xm0, addr, T_0F, 0xAE); }
 void vsubpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x5C); }
 void vsubps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x5C); }
-void vsubsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_Z | T_N8, 0x5C); }
-void vsubss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5C); }
+void vsubsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F2 | T_EW1 | T_EVEX | T_ER_X | T_N8, 0x5C); }
+void vsubss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_X | T_N4, 0x5C); }
 void vtestpd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F38 | T_YMM, 0x0F); }
 void vtestps(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F38 | T_YMM, 0x0E); }
 void vucomisd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N8 | T_66 | T_0F | T_EW1 | T_EVEX | T_SAE_X, 0x2E); }
@ -1739,6 +1750,8 @@ void v4fmaddps(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM
 void v4fmaddss(const Xmm& x1, const Xmm& x2, const Address& addr) { opAVX_X_X_XM(x1, x2, addr, T_0F38 | T_F2 | T_EW0 | T_MUST_EVEX | T_N16, 0x9B); }
 void v4fnmaddps(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0xAA); }
 void v4fnmaddss(const Xmm& x1, const Xmm& x2, const Address& addr) { opAVX_X_X_XM(x1, x2, addr, T_0F38 | T_F2 | T_EW0 | T_MUST_EVEX | T_N16, 0xAB); }
+void vaddph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x58); }
+void vaddsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x58); }
 void valignd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x03, imm); }
 void valignq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x03, imm); }
 void vblendmpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x65); }
@ -1753,41 +1766,206 @@ void vbroadcasti32x4(const Ymm& y, const Operand& op) { opAVX_X_XM_IMM(y, op, T_
 void vbroadcasti32x8(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_N32, 0x5B); }
 void vbroadcasti64x2(const Ymm& y, const Operand& op) { opAVX_X_XM_IMM(y, op, T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_N16, 0x5A); }
 void vbroadcasti64x4(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_N32, 0x5B); }
+void vcmpeq_ospd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 16); }
+void vcmpeq_osps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 16); }
+void vcmpeq_ossd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 16); }
+void vcmpeq_osss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 16); }
+void vcmpeq_uqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 8); }
+void vcmpeq_uqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 8); }
+void vcmpeq_uqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 8); }
+void vcmpeq_uqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 8); }
+void vcmpeq_uspd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 24); }
+void vcmpeq_usps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 24); }
+void vcmpeq_ussd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 24); }
+void vcmpeq_usss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 24); }
+void vcmpeqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 0); }
+void vcmpeqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 0); }
+void vcmpeqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 0); }
+void vcmpeqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 0); }
+void vcmpfalse_ospd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 27); }
+void vcmpfalse_osps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 27); }
+void vcmpfalse_ossd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 27); }
+void vcmpfalse_osss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 27); }
+void vcmpfalsepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 11); }
+void vcmpfalseps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 11); }
+void vcmpfalsesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 11); }
+void vcmpfalsess(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 11); }
+void vcmpge_oqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 29); }
+void vcmpge_oqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 29); }
+void vcmpge_oqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 29); }
+void vcmpge_oqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 29); }
+void vcmpgepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 13); }
+void vcmpgeps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 13); }
+void vcmpgesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 13); }
+void vcmpgess(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 13); }
+void vcmpgt_oqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 30); }
+void vcmpgt_oqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 30); }
+void vcmpgt_oqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 30); }
+void vcmpgt_oqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 30); }
+void vcmpgtpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 14); }
+void vcmpgtps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 14); }
+void vcmpgtsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 14); }
+void vcmpgtss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 14); }
+void vcmple_oqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 18); }
+void vcmple_oqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 18); }
+void vcmple_oqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 18); }
+void vcmple_oqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 18); }
+void vcmplepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 2); }
+void vcmpleps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 2); }
+void vcmplesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 2); }
+void vcmpless(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 2); }
+void vcmplt_oqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 17); }
+void vcmplt_oqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 17); }
+void vcmplt_oqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 17); }
+void vcmplt_oqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 17); }
+void vcmpltpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 1); }
+void vcmpltps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 1); }
+void vcmpltsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 1); }
+void vcmpltss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 1); }
+void vcmpneq_oqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 12); }
+void vcmpneq_oqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 12); }
+void vcmpneq_oqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 12); }
+void vcmpneq_oqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 12); }
+void vcmpneq_ospd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 28); }
+void vcmpneq_osps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 28); }
+void vcmpneq_ossd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 28); }
+void vcmpneq_osss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 28); }
+void vcmpneq_uspd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 20); }
+void vcmpneq_usps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 20); }
+void vcmpneq_ussd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 20); }
+void vcmpneq_usss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 20); }
+void vcmpneqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 4); }
+void vcmpneqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 4); }
+void vcmpneqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 4); }
+void vcmpneqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 4); }
+void vcmpnge_uqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 25); }
+void vcmpnge_uqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 25); }
+void vcmpnge_uqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 25); }
+void vcmpnge_uqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 25); }
+void vcmpngepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 9); }
+void vcmpngeps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 9); }
+void vcmpngesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 9); }
+void vcmpngess(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 9); }
+void vcmpngt_uqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 26); }
+void vcmpngt_uqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 26); }
+void vcmpngt_uqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 26); }
+void vcmpngt_uqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 26); }
+void vcmpngtpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 10); }
+void vcmpngtps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 10); }
+void vcmpngtsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 10); }
+void vcmpngtss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 10); }
+void vcmpnle_uqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 22); }
+void vcmpnle_uqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 22); }
+void vcmpnle_uqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 22); }
+void vcmpnle_uqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 22); }
+void vcmpnlepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 6); }
+void vcmpnleps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 6); }
+void vcmpnlesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 6); }
+void vcmpnless(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 6); }
+void vcmpnlt_uqpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 21); }
+void vcmpnlt_uqps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 21); }
+void vcmpnlt_uqsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 21); }
+void vcmpnlt_uqss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 21); }
+void vcmpnltpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 5); }
+void vcmpnltps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 5); }
+void vcmpnltsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 5); }
+void vcmpnltss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 5); }
+void vcmpord_spd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 23); }
+void vcmpord_sps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 23); }
+void vcmpord_ssd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 23); }
+void vcmpord_sss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 23); }
+void vcmpordpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 7); }
+void vcmpordps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 7); }
+void vcmpordsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 7); }
+void vcmpordss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 7); }
 void vcmppd(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_66 | T_0F | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0xC2, imm); }
+void vcmpph(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0xC2, imm); }
 void vcmpps(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_0F | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0xC2, imm); }
 void vcmpsd(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_N8 | T_F2 | T_0F | T_EW1 | T_SAE_Z | T_MUST_EVEX, 0xC2, imm); }
+void vcmpsh(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_N2 | T_F3 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0xC2, imm); }
 void vcmpss(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opAVX_K_X_XM(k, x, op, T_N4 | T_F3 | T_0F | T_EW0 | T_SAE_Z | T_MUST_EVEX, 0xC2, imm); }
+void vcmptrue_uspd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 31); }
+void vcmptrue_usps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 31); }
+void vcmptrue_ussd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 31); }
+void vcmptrue_usss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 31); }
+void vcmptruepd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 15); }
+void vcmptrueps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 15); }
+void vcmptruesd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 15); }
+void vcmptruess(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 15); }
+void vcmpunord_spd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 19); }
+void vcmpunord_sps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 19); }
+void vcmpunord_ssd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 19); }
+void vcmpunord_sss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 19); }
+void vcmpunordpd(const Opmask& k, const Xmm& x, const Operand& op) { vcmppd(k, x, op, 3); }
+void vcmpunordps(const Opmask& k, const Xmm& x, const Operand& op) { vcmpps(k, x, op, 3); }
+void vcmpunordsd(const Opmask& k, const Xmm& x, const Operand& op) { vcmpsd(k, x, op, 3); }
+void vcmpunordss(const Opmask& k, const Xmm& x, const Operand& op) { vcmpss(k, x, op, 3); }
+void vcomish(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, 0x2F); }
 void vcompressb(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N1 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x63); }
 void vcompresspd(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x8A); }
 void vcompressps(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x8A); }
 void vcompressw(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, T_N2 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x63); }
+void vcvtdq2ph(const Xmm& x, const Operand& op) { checkCvt4(x, op); opCvt(x, op, T_N16 | T_N_VL | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x5B); }
 void vcvtne2ps2bf16(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x72); }
 void vcvtneps2bf16(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x72); }
+void vcvtpd2ph(const Xmm& x, const Operand& op) { opCvt5(x, op, T_N16 | T_N_VL | T_66 | T_MAP5 | T_EW1 | T_ER_Z | T_MUST_EVEX | T_B64, 0x5A); }
 void vcvtpd2qq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x7B); }
-void vcvtpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x79); }
+void vcvtpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x79); }
 void vcvtpd2uqq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x79); }
-void vcvtps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 0x7B); }
+void vcvtph2dq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_YMM | T_ER_Y | T_MUST_EVEX | T_B16, 0x5B); }
+void vcvtph2pd(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, T_N4 | T_N_VL | T_MAP5 | T_EW0 | T_YMM | T_SAE_X | T_MUST_EVEX | T_B16, 0x5A); }
+void vcvtph2psx(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_MAP6 | T_EW0 | T_YMM | T_SAE_Y | T_MUST_EVEX | T_B16, 0x13); }
+void vcvtph2qq(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, T_N4 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_YMM | T_ER_X | T_MUST_EVEX | T_B16, 0x7B); }
+void vcvtph2udq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_MAP5 | T_EW0 | T_YMM | T_ER_Y | T_MUST_EVEX | T_B16, 0x79); }
+void vcvtph2uqq(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, T_N4 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_YMM | T_ER_X | T_MUST_EVEX | T_B16, 0x79); }
+void vcvtph2uw(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }
+void vcvtph2w(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }
+void vcvtps2phx(const Xmm& x, const Operand& op) { checkCvt4(x, op); opCvt(x, op, T_N16 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_ER_Z | T_MUST_EVEX | T_B32, 0x1D); }
+void vcvtps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_0F | T_EW0 | T_YMM | T_ER_Y | T_MUST_EVEX | T_B32, 0x7B); }
 void vcvtps2udq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_0F | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x79); }
-void vcvtps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_ER_Y, 0x79); }
+void vcvtps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_0F | T_EW0 | T_YMM | T_ER_Y | T_MUST_EVEX | T_B32, 0x79); }
 void vcvtqq2pd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0xE6); }
-void vcvtqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x5B); }
-void vcvtsd2usi(const Reg32e& r, const Operand& op) { int type = (T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_ER_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x79); }
-void vcvtss2usi(const Reg32e& r, const Operand& op) { int type = (T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_ER_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x79); }
+void vcvtqq2ph(const Xmm& x, const Operand& op) { opCvt5(x, op, T_N16 | T_N_VL | T_MAP5 | T_EW1 | T_ER_Z | T_MUST_EVEX | T_B64, 0x5B); }
+void vcvtqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x5B); }
+void vcvtsd2sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_F2 | T_MAP5 | T_EW1 | T_ER_X | T_MUST_EVEX, 0x5A); }
+void vcvtsd2usi(const Reg32e& r, const Operand& op) { int type = (T_N8 | T_F2 | T_0F | T_ER_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x79); }
+void vcvtsh2sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x5A); }
+void vcvtsh2si(const Reg32e& r, const Operand& op) { int type = (T_N2 | T_F3 | T_MAP5 | T_ER_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x2D); }
+void vcvtsh2ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_MAP6 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x13); }
+void vcvtsh2usi(const Reg32e& r, const Operand& op) { int type = (T_N2 | T_F3 | T_MAP5 | T_ER_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x79); }
+void vcvtsi2sh(const Xmm& x1, const Xmm& x2, const Operand& op) { if (!(x1.isXMM() && x2.isXMM() && op.isBit(32|64))) XBYAK_THROW(ERR_BAD_COMBINATION) int type = (T_F3 | T_MAP5 | T_ER_R | T_MUST_EVEX | T_M_K) | (op.isBit(32) ? (T_EW0 | T_N4) : (T_EW1 | T_N8)); opVex(x1, &x2, op, type, 0x2A); }
+void vcvtss2sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_MAP5 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x1D); }
+void vcvtss2usi(const Reg32e& r, const Operand& op) { int type = (T_N4 | T_F3 | T_0F | T_ER_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x79); }
 void vcvttpd2qq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x7A); }
-void vcvttpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, 0x78); }
+void vcvttpd2udq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x78); }
 void vcvttpd2uqq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x78); }
-void vcvttps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 0x7A); }
+void vcvttph2dq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_F3 | T_MAP5 | T_EW0 | T_YMM | T_SAE_Y | T_MUST_EVEX | T_B16, 0x5B); }
+void vcvttph2qq(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, T_N4 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_YMM | T_SAE_X | T_MUST_EVEX | T_B16, 0x7A); }
+void vcvttph2udq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_MAP5 | T_EW0 | T_YMM | T_SAE_Y | T_MUST_EVEX | T_B16, 0x78); }
+void vcvttph2uqq(const Xmm& x, const Operand& op) { if (!op.isXMM() && !op.isMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(x, 0, op, T_N4 | T_N_VL | T_66 | T_MAP5 | T_EW0 | T_YMM | T_SAE_X | T_MUST_EVEX | T_B16, 0x78); }
+void vcvttph2uw(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x7C); }
+void vcvttph2w(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP5 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x7C); }
+void vcvttps2qq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_0F | T_EW0 | T_YMM | T_SAE_Y | T_MUST_EVEX | T_B32, 0x7A); }
 void vcvttps2udq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_0F | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x78); }
-void vcvttps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_66 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL | T_SAE_Y, 0x78); }
-void vcvttsd2usi(const Reg32e& r, const Operand& op) { int type = (T_F2 | T_0F | T_MUST_EVEX | T_N8 | T_SAE_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x78); }
-void vcvttss2usi(const Reg32e& r, const Operand& op) { int type = (T_F3 | T_0F | T_MUST_EVEX | T_N4 | T_SAE_X) | (r.isREG(64) ? T_EW1 : T_EW0); opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, type, 0x78); }
-void vcvtudq2pd(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_F3 | T_0F | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_N8 | T_N_VL, 0x7A); }
+void vcvttps2uqq(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_66 | T_0F | T_EW0 | T_YMM | T_SAE_Y | T_MUST_EVEX | T_B32, 0x78); }
+void vcvttsd2usi(const Reg32e& r, const Operand& op) { int type = (T_N8 | T_F2 | T_0F | T_SAE_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x78); }
+void vcvttsh2si(const Reg32e& r, const Operand& op) { int type = (T_N2 | T_F3 | T_MAP5 | T_EW0 | T_SAE_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x2C); }
+void vcvttsh2usi(const Reg32e& r, const Operand& op) { int type = (T_N2 | T_F3 | T_MAP5 | T_EW0 | T_SAE_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x78); }
+void vcvttss2usi(const Reg32e& r, const Operand& op) { int type = (T_N4 | T_F3 | T_0F | T_SAE_X | T_MUST_EVEX) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x78); }
+void vcvtudq2pd(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_N8 | T_N_VL | T_F3 | T_0F | T_EW0 | T_YMM | T_MUST_EVEX | T_B32, 0x7A); }
+void vcvtudq2ph(const Xmm& x, const Operand& op) { checkCvt4(x, op); opCvt(x, op, T_N16 | T_N_VL | T_F2 | T_MAP5 | T_EW0 | T_ER_Z | T_MUST_EVEX | T_B32, 0x7A); }
 void vcvtudq2ps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x7A); }
 void vcvtuqq2pd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x7A); }
-void vcvtuqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F2 | T_0F | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x7A); }
+void vcvtuqq2ph(const Xmm& x, const Operand& op) { opCvt5(x, op, T_N16 | T_N_VL | T_F2 | T_MAP5 | T_EW1 | T_ER_Z | T_MUST_EVEX | T_B64, 0x7A); }
+void vcvtuqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F2 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x7A); }
 void vcvtusi2sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F2 | T_0F | T_MUST_EVEX, T_W1 | T_EW1 | T_ER_X | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }
+void vcvtusi2sh(const Xmm& x1, const Xmm& x2, const Operand& op) { if (!(x1.isXMM() && x2.isXMM() && op.isBit(32|64))) XBYAK_THROW(ERR_BAD_COMBINATION) int type = (T_F3 | T_MAP5 | T_ER_R | T_MUST_EVEX | T_M_K) | (op.isBit(32) ? (T_EW0 | T_N4) : (T_EW1 | T_N8)); opVex(x1, &x2, op, type, 0x7B); }
 void vcvtusi2ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F3 | T_0F | T_MUST_EVEX | T_ER_X, T_W1 | T_EW1 | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }
+void vcvtuw2ph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }
+void vcvtw2ph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }
 void vdbpsadbw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x42, imm); }
+void vdivph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x5E); }
+void vdivsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x5E); }
 void vdpbf16ps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x52); }
 void vexp2pd(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_B64 | T_SAE_Z, 0xC8); }
 void vexp2ps(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, 0xC8); }
@ -1801,13 +1979,49 @@ void vextracti32x4(const Operand& op, const Ymm& r, uint8_t imm) { if (!op.is(Op
 void vextracti32x8(const Operand& op, const Zmm& r, uint8_t imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3B, imm); }
 void vextracti64x2(const Operand& op, const Ymm& r, uint8_t imm) { if (!op.is(Operand::MEM | Operand::XMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x39, imm); }
 void vextracti64x4(const Operand& op, const Zmm& r, uint8_t imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3B, imm); }
+void vfcmaddcph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x56); }
+void vfcmulcph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0xD6); }
 void vfixupimmpd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x54, imm); }
 void vfixupimmps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x54, imm); }
 void vfixupimmsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); }
 void vfixupimmss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); }
+void vfmadd132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x98); }
+void vfmadd132sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x99); }
+void vfmadd213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xA8); }
+void vfmadd213sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xA9); }
+void vfmadd231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xB8); }
+void vfmadd231sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xB9); }
+void vfmaddcph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x56); }
+void vfmaddsub132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x96); }
+void vfmaddsub213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xA6); }
+void vfmaddsub231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xB6); }
+void vfmsub132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x9A); }
+void vfmsub132sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x9B); }
+void vfmsub213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xAA); }
+void vfmsub213sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xAB); }
+void vfmsub231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xBA); }
+void vfmsub231sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xBB); }
+void vfmsubadd132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x97); }
+void vfmsubadd213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xA7); }
+void vfmsubadd231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xB7); }
+void vfmulcph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0xD6); }
+void vfnmadd132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x9C); }
+void vfnmadd132sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x9D); }
+void vfnmadd213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xAC); }
+void vfnmadd213sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xAD); }
+void vfnmadd231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xBC); }
+void vfnmadd231sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xBD); }
+void vfnmsub132ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x9E); }
+void vfnmsub132sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x9F); }
+void vfnmsub213ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xAE); }
+void vfnmsub213sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xAF); }
+void vfnmsub231ph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0xBE); }
+void vfnmsub231sh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0xBF); }
 void vfpclasspd(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64, 0x66, imm); }
+void vfpclassph(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B16, 0x66, imm); }
 void vfpclassps(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }
 void vfpclasssd(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); }
+void vfpclasssh(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_0F3A | T_MUST_EVEX | T_EW0 | T_N2, 0x67, imm); }
 void vfpclassss(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, 0x67, imm); }
 void vgatherdpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 1); }
 void vgatherdps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 0); }
@ -1822,12 +2036,16 @@ void vgatherpf1qps(const Address& addr) { opGatherFetch(addr, zm2, T_N4 | T_66 |
 void vgatherqpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x93, 0); }
 void vgatherqps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x93, 2); }
 void vgetexppd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x42); }
+void vgetexpph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x42); }
 void vgetexpps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x42); }
 void vgetexpsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x43); }
+void vgetexpsh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x43); }
 void vgetexpss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x43); }
 void vgetmantpd(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x26, imm); }
+void vgetmantph(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x26, imm); }
 void vgetmantps(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x26, imm); }
 void vgetmantsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x27, imm); }
+void vgetmantsh(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x27, imm); }
 void vgetmantss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x27, imm); }
 void vinsertf32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8_t imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x18, imm); }
 void vinsertf32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8_t imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x1A, imm); }
@ -1837,6 +2055,10 @@ void vinserti32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8_t imm)
 void vinserti32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8_t imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3A, imm); }
 void vinserti64x2(const Ymm& r1, const Ymm& r2, const Operand& op, uint8_t imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x38, imm); }
 void vinserti64x4(const Zmm& r1, const Zmm& r2, const Operand& op, uint8_t imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3A, imm); }
+void vmaxph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x5F); }
+void vmaxsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x5F); }
+void vminph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x5D); }
+void vminsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x5D); }
 void vmovdqa32(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
 void vmovdqa32(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
 void vmovdqa64(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
@ -1849,6 +2071,14 @@ void vmovdqu64(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_F3
 void vmovdqu64(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
 void vmovdqu8(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
 void vmovdqu8(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
+void vmovsh(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX | T_M_K, 0x11); }
+void vmovsh(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX, 0x10); }
+void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2, x3, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX, 0x10); }
+void vmovw(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_N2 | T_66 | T_MAP5 | T_MUST_EVEX, 0x7E); }
+void vmovw(const Reg32e& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, r, T_N2 | T_66 | T_MAP5 | T_MUST_EVEX, 0x7E); }
+void vmovw(const Xmm& x, const Operand& op) { if (!op.isREG(32|64) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, xm0, op, T_N2 | T_66 | T_MAP5 | T_MUST_EVEX, 0x6E); }
+void vmulph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x59); }
+void vmulsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x59); }
 void vp2intersectd(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) XBYAK_THROW(ERR_OPMASK_IS_ALREADY_SET) opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW0 | T_B32, 0x68); }
 void vp2intersectq(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) XBYAK_THROW(ERR_OPMASK_IS_ALREADY_SET) opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW1 | T_B64, 0x68); }
 void vp4dpwssd(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x52); }
@ -2006,14 +2236,20 @@ void vrcp28pd(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_
 void vrcp28ps(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, 0xCA); }
 void vrcp28sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0xCB); }
 void vrcp28ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0xCB); }
+void vrcpph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_MUST_EVEX | T_B16, 0x4C); }
+void vrcpsh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_MUST_EVEX, 0x4D); }
 void vreducepd(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x56, imm); }
+void vreduceph(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x56, imm); }
 void vreduceps(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x56, imm); }
 void vreducesd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x57, imm); }
+void vreducesh(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x57, imm); }
 void vreducess(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x57, imm); }
-void vrndscalepd(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x09, imm); }
-void vrndscaleps(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX | T_B32, 0x08, imm); }
-void vrndscalesd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_MUST_EVEX, 0x0B, imm); }
-void vrndscaless(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_MUST_EVEX, 0x0A, imm); }
+void vrndscalepd(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x09, imm); }
+void vrndscaleph(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B16, 0x08, imm); }
+void vrndscaleps(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x08, imm); }
+void vrndscalesd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x0B, imm); }
+void vrndscalesh(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x0A, imm); }
+void vrndscaless(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x0A, imm); }
 void vrsqrt14pd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_B64, 0x4E); }
 void vrsqrt14ps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_B32, 0x4E); }
 void vrsqrt14sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x4F); }
@ -2022,9 +2258,13 @@ void vrsqrt28pd(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 |
 void vrsqrt28ps(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, 0xCC); }
 void vrsqrt28sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0xCD); }
 void vrsqrt28ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0xCD); }
+void vrsqrtph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_MUST_EVEX | T_B16, 0x4E); }
+void vrsqrtsh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_MUST_EVEX, 0x4F); }
 void vscalefpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x2C); }
+void vscalefph(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x2C); }
 void vscalefps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B32, 0x2C); }
 void vscalefsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_ER_X | T_MUST_EVEX, 0x2D); }
+void vscalefsh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_66 | T_MAP6 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x2D); }
 void vscalefss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x2D); }
 void vscatterdpd(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA2, 1); }
 void vscatterdps(const Address& addr, const Xmm& x) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K | T_VSIB, 0xA2, 0); }
@ -2042,6 +2282,11 @@ void vshuff32x4(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) {
 void vshuff64x2(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, 0x23, imm); }
 void vshufi32x4(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, 0x43, imm); }
 void vshufi64x2(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, 0x43, imm); }
+void vsqrtph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x51); }
+void vsqrtsh(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_ER_X | T_MUST_EVEX, 0x51); }
+void vsubph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x5C); }
+void vsubsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x5C); }
+void vucomish(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_MAP5 | T_MUST_EVEX | T_EW0 | T_SAE_X | T_N2, 0x2E); }
 #ifdef XBYAK64
 void kmovq(const Opmask& k, const Reg64& r) { opVex(k, 0, r, T_L0 | T_0F | T_F2 | T_W1, 0x92); }
 void kmovq(const Reg64& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_F2 | T_W1, 0x93); }
--- a/xbyak/xbyak_util.h
+++ b/xbyak/xbyak_util.h
@ -17,7 +17,6 @@
 	utility class and functions for Xbyak
 	Xbyak::util::Clock ; rdtsc timer
 	Xbyak::util::Cpu ; detect CPU
-	@note this header is UNDER CONSTRUCTION!
 */
 #include "xbyak.h"
 #endif // XBYAK_ONLY_CLASS_CPU
@ -27,8 +26,8 @@
 #endif

 #ifdef XBYAK_INTEL_CPU_SPECIFIC
-#ifdef _MSC_VER
-	#if (_MSC_VER < 1400) && defined(XBYAK32)
+#ifdef _WIN32
+	#if defined(_MSC_VER) && (_MSC_VER < 1400) && defined(XBYAK32)
 		static inline __declspec(naked) void __cpuid(int[4], int)
 		{
 			__asm {
@ -88,32 +87,64 @@ typedef enum {
   CoreLevel = 2
 } IntelCpuTopologyLevel;

+namespace local {
+
+template<uint64_t L, uint64_t H = 0>
+struct TypeT {
+};
+
+template<uint64_t L1, uint64_t H1, uint64_t L2, uint64_t H2>
+TypeT<L1 | L2, H1 | H2> operator|(TypeT<L1, H1>, TypeT<L2, H2>) { return TypeT<L1 | L2, H1 | H2>(); }
+
+} // local
+
 /**
 	CPU detection class
+	@note static inline const member is supported by c++17 or later, so use template hack
 */
 class Cpu {
-	uint64_t type_;
+public:
+	class Type {
+		uint64_t L;
+		uint64_t H;
+	public:
+		Type(uint64_t L = 0, uint64_t H = 0) : L(L), H(H) { }
+		template<uint64_t L_, uint64_t H_>
+		Type(local::TypeT<L_, H_>) : L(L_), H(H_) {}
+		Type& operator&=(const Type& rhs) { L &= rhs.L; H &= rhs.H; return *this; }
+		Type& operator|=(const Type& rhs) { L |= rhs.L; H |= rhs.H; return *this; }
+		Type operator&(const Type& rhs) const { Type t = *this; t &= rhs; return t; }
+		Type operator|(const Type& rhs) const { Type t = *this; t |= rhs; return t; }
+		bool operator==(const Type& rhs) const { return H == rhs.H && L == rhs.L; }
+		bool operator!=(const Type& rhs) const { return !operator==(rhs); }
+		// without explicit because backward compatilibity
+		operator bool() const { return (H | L) != 0; }
+		uint64_t getL() const { return L; }
+		uint64_t getH() const { return H; }
+	};
+private:
+	Type type_;
 	//system topology
 	bool x2APIC_supported_;
 	static const size_t maxTopologyLevels = 2;
-	unsigned int numCores_[maxTopologyLevels];
+	uint32_t numCores_[maxTopologyLevels];

-	static const unsigned int maxNumberCacheLevels = 10;
-	unsigned int dataCacheSize_[maxNumberCacheLevels];
-	unsigned int coresSharignDataCache_[maxNumberCacheLevels];
-	unsigned int dataCacheLevels_;
+	static const uint32_t maxNumberCacheLevels = 10;
+	uint32_t dataCacheSize_[maxNumberCacheLevels];
+	uint32_t coresSharignDataCache_[maxNumberCacheLevels];
+	uint32_t dataCacheLevels_;

-	unsigned int get32bitAsBE(const char *x) const
+	uint32_t get32bitAsBE(const char *x) const
 	{
 		return x[0] | (x[1] << 8) | (x[2] << 16) | (x[3] << 24);
 	}
-	unsigned int mask(int n) const
+	uint32_t mask(int n) const
 	{
 		return (1U << n) - 1;
 	}
 	void setFamily()
 	{
-		unsigned int data[4] = {};
+		uint32_t data[4] = {};
 		getCpuid(1, data);
 		stepping = data[0] & mask(4);
 		model = (data[0] >> 4) & mask(4);
@ -132,15 +163,15 @@ class Cpu {
 			displayModel = model;
 		}
 	}
-	unsigned int extractBit(unsigned int val, unsigned int base, unsigned int end)
+	uint32_t extractBit(uint32_t val, uint32_t base, uint32_t end)
 	{
 		return (val >> base) & ((1u << (end - base)) - 1);
 	}
 	void setNumCores()
 	{
-		if ((type_ & tINTEL) == 0) return;
+		if (!has(tINTEL)) return;

-		unsigned int data[4] = {};
+		uint32_t data[4] = {};

 		 /* CAUTION: These numbers are configuration as shipped by Intel. */
 		getCpuidEx(0x0, 0, data);
@ -152,7 +183,7 @@ class Cpu {
 				leaf 0xB can be zeroed-out by a hypervisor
 			*/
 			x2APIC_supported_ = true;
-			for (unsigned int i = 0; i < maxTopologyLevels; i++) {
+			for (uint32_t i = 0; i < maxTopologyLevels; i++) {
 				getCpuidEx(0xB, i, data);
 				IntelCpuTopologyLevel level = (IntelCpuTopologyLevel)extractBit(data[2], 8, 15);
 				if (level == SmtLevel || level == CoreLevel) {
@ -176,14 +207,14 @@ class Cpu {
 	}
 	void setCacheHierarchy()
 	{
-		if ((type_ & tINTEL) == 0) return;
-		const unsigned int NO_CACHE = 0;
-		const unsigned int DATA_CACHE = 1;
-//		const unsigned int INSTRUCTION_CACHE = 2;
-		const unsigned int UNIFIED_CACHE = 3;
-		unsigned int smt_width = 0;
-		unsigned int logical_cores = 0;
-		unsigned int data[4] = {};
+		if (!has(tINTEL)) return;
+		const uint32_t NO_CACHE = 0;
+		const uint32_t DATA_CACHE = 1;
+//		const uint32_t INSTRUCTION_CACHE = 2;
+		const uint32_t UNIFIED_CACHE = 3;
+		uint32_t smt_width = 0;
+		uint32_t logical_cores = 0;
+		uint32_t data[4] = {};

 		if (x2APIC_supported_) {
 			smt_width = numCores_[0];
@ -201,10 +232,10 @@ class Cpu {
 		*/
 		for (int i = 0; dataCacheLevels_ < maxNumberCacheLevels; i++) {
 			getCpuidEx(0x4, i, data);
-			unsigned int cacheType = extractBit(data[0], 0, 4);
+			uint32_t cacheType = extractBit(data[0], 0, 4);
 			if (cacheType == NO_CACHE) break;
 			if (cacheType == DATA_CACHE || cacheType == UNIFIED_CACHE) {
-				unsigned int actual_logical_cores = extractBit(data[0], 14, 25) + 1;
+				uint32_t actual_logical_cores = extractBit(data[0], 14, 25) + 1;
 				if (logical_cores != 0) { // true only if leaf 0xB is supported and valid
 					actual_logical_cores = (std::min)(actual_logical_cores, logical_cores);
 				}
@ -231,7 +262,7 @@ public:
 	int displayFamily; // family + extFamily
 	int displayModel; // model + extModel

-	unsigned int getNumCores(IntelCpuTopologyLevel level) const {
+	uint32_t getNumCores(IntelCpuTopologyLevel level) const {
 		if (!x2APIC_supported_) XBYAK_THROW_RET(ERR_X2APIC_IS_NOT_SUPPORTED, 0)
 		switch (level) {
 		case SmtLevel: return numCores_[level - 1];
@ -240,13 +271,13 @@ public:
 		}
 	}

-	unsigned int getDataCacheLevels() const { return dataCacheLevels_; }
-	unsigned int getCoresSharingDataCache(unsigned int i) const
+	uint32_t getDataCacheLevels() const { return dataCacheLevels_; }
+	uint32_t getCoresSharingDataCache(uint32_t i) const
 	{
 		if (i >= dataCacheLevels_) XBYAK_THROW_RET(ERR_BAD_PARAMETER, 0)
 		return coresSharignDataCache_[i];
 	}
-	unsigned int getDataCacheSize(unsigned int i) const
+	uint32_t getDataCacheSize(uint32_t i) const
 	{
 		if (i >= dataCacheLevels_) XBYAK_THROW_RET(ERR_BAD_PARAMETER, 0)
 		return dataCacheSize_[i];
@ -255,10 +286,10 @@ public:
 	/*
 		data[] = { eax, ebx, ecx, edx }
 	*/
-	static inline void getCpuid(unsigned int eaxIn, unsigned int data[4])
+	static inline void getCpuid(uint32_t eaxIn, uint32_t data[4])
 	{
 #ifdef XBYAK_INTEL_CPU_SPECIFIC
-	#ifdef _MSC_VER
+	#ifdef _WIN32
 		__cpuid(reinterpret_cast<int*>(data), eaxIn);
 	#else
 		__cpuid(eaxIn, data[0], data[1], data[2], data[3]);
@ -268,7 +299,7 @@ public:
 		(void)data;
 #endif
 	}
-	static inline void getCpuidEx(unsigned int eaxIn, unsigned int ecxIn, unsigned int data[4])
+	static inline void getCpuidEx(uint32_t eaxIn, uint32_t ecxIn, uint32_t data[4])
 	{
 #ifdef XBYAK_INTEL_CPU_SPECIFIC
 	#ifdef _MSC_VER
@ -288,7 +319,7 @@ public:
 	#ifdef _MSC_VER
 		return _xgetbv(0);
 	#else
-		unsigned int eax, edx;
+		uint32_t eax, edx;
 		// xgetvb is not support on gcc 4.2
 //		__asm__ volatile("xgetbv" : "=a"(eax), "=d"(edx) : "c"(0));
 		__asm__ volatile(".byte 0x0f, 0x01, 0xd0" : "=a"(eax), "=d"(edx) : "c"(0));
@ -298,77 +329,78 @@ public:
 		return 0;
 #endif
 	}
-	typedef uint64_t Type;

-	static const Type NONE = 0;
-	static const Type tMMX = 1 << 0;
-	static const Type tMMX2 = 1 << 1;
-	static const Type tCMOV = 1 << 2;
-	static const Type tSSE = 1 << 3;
-	static const Type tSSE2 = 1 << 4;
-	static const Type tSSE3 = 1 << 5;
-	static const Type tSSSE3 = 1 << 6;
-	static const Type tSSE41 = 1 << 7;
-	static const Type tSSE42 = 1 << 8;
-	static const Type tPOPCNT = 1 << 9;
-	static const Type tAESNI = 1 << 10;
-	static const Type tSSE5 = 1 << 11;
-	static const Type tOSXSAVE = 1 << 12;
-	static const Type tPCLMULQDQ = 1 << 13;
-	static const Type tAVX = 1 << 14;
-	static const Type tFMA = 1 << 15;
-
-	static const Type t3DN = 1 << 16;
-	static const Type tE3DN = 1 << 17;
-	static const Type tSSE4a = 1 << 18;
-	static const Type tRDTSCP = 1 << 19;
-	static const Type tAVX2 = 1 << 20;
-	static const Type tBMI1 = 1 << 21; // andn, bextr, blsi, blsmsk, blsr, tzcnt
-	static const Type tBMI2 = 1 << 22; // bzhi, mulx, pdep, pext, rorx, sarx, shlx, shrx
-	static const Type tLZCNT = 1 << 23;
-
-	static const Type tINTEL = 1 << 24;
-	static const Type tAMD = 1 << 25;
-
-	static const Type tENHANCED_REP = 1 << 26; // enhanced rep movsb/stosb
-	static const Type tRDRAND = 1 << 27;
-	static const Type tADX = 1 << 28; // adcx, adox
-	static const Type tRDSEED = 1 << 29; // rdseed
-	static const Type tSMAP = 1 << 30; // stac
-	static const Type tHLE = uint64_t(1) << 31; // xacquire, xrelease, xtest
-	static const Type tRTM = uint64_t(1) << 32; // xbegin, xend, xabort
-	static const Type tF16C = uint64_t(1) << 33; // vcvtph2ps, vcvtps2ph
-	static const Type tMOVBE = uint64_t(1) << 34; // mobve
-	static const Type tAVX512F = uint64_t(1) << 35;
-	static const Type tAVX512DQ = uint64_t(1) << 36;
-	static const Type tAVX512_IFMA = uint64_t(1) << 37;
-	static const Type tAVX512IFMA = tAVX512_IFMA;
-	static const Type tAVX512PF = uint64_t(1) << 38;
-	static const Type tAVX512ER = uint64_t(1) << 39;
-	static const Type tAVX512CD = uint64_t(1) << 40;
-	static const Type tAVX512BW = uint64_t(1) << 41;
-	static const Type tAVX512VL = uint64_t(1) << 42;
-	static const Type tAVX512_VBMI = uint64_t(1) << 43;
-	static const Type tAVX512VBMI = tAVX512_VBMI; // changed by Intel's manual
-	static const Type tAVX512_4VNNIW = uint64_t(1) << 44;
-	static const Type tAVX512_4FMAPS = uint64_t(1) << 45;
-	static const Type tPREFETCHWT1 = uint64_t(1) << 46;
-	static const Type tPREFETCHW = uint64_t(1) << 47;
-	static const Type tSHA = uint64_t(1) << 48;
-	static const Type tMPX = uint64_t(1) << 49;
-	static const Type tAVX512_VBMI2 = uint64_t(1) << 50;
-	static const Type tGFNI = uint64_t(1) << 51;
-	static const Type tVAES = uint64_t(1) << 52;
-	static const Type tVPCLMULQDQ = uint64_t(1) << 53;
-	static const Type tAVX512_VNNI = uint64_t(1) << 54;
-	static const Type tAVX512_BITALG = uint64_t(1) << 55;
-	static const Type tAVX512_VPOPCNTDQ = uint64_t(1) << 56;
-	static const Type tAVX512_BF16 = uint64_t(1) << 57;
-	static const Type tAVX512_VP2INTERSECT = uint64_t(1) << 58;
-	static const Type tAMX_TILE = uint64_t(1) << 59;
-	static const Type tAMX_INT8 = uint64_t(1) << 60;
-	static const Type tAMX_BF16 = uint64_t(1) << 61;
-	static const Type tAVX_VNNI = uint64_t(1) << 62;
+	static const local::TypeT<0> NONE;
+	static const local::TypeT<1 << 0> tMMX;
+	static const local::TypeT<1 << 1> tMMX2;
+	static const local::TypeT<1 << 2> tCMOV;
+	static const local::TypeT<1 << 3> tSSE;
+	static const local::TypeT<1 << 4> tSSE2;
+	static const local::TypeT<1 << 5> tSSE3;
+	static const local::TypeT<1 << 6> tSSSE3;
+	static const local::TypeT<1 << 7> tSSE41;
+	static const local::TypeT<1 << 8> tSSE42;
+	static const local::TypeT<1 << 9> tPOPCNT;
+	static const local::TypeT<1 << 10> tAESNI;
+	static const local::TypeT<1 << 11> tAVX512_FP16;
+	static const local::TypeT<1 << 12> tOSXSAVE;
+	static const local::TypeT<1 << 13> tPCLMULQDQ;
+	static const local::TypeT<1 << 14> tAVX;
+	static const local::TypeT<1 << 15> tFMA;
+	static const local::TypeT<1 << 16> t3DN;
+	static const local::TypeT<1 << 17> tE3DN;
+	static const local::TypeT<1 << 18> tWAITPKG;
+	static const local::TypeT<1 << 19> tRDTSCP;
+	static const local::TypeT<1 << 20> tAVX2;
+	static const local::TypeT<1 << 21> tBMI1; // andn, bextr, blsi, blsmsk, blsr, tzcnt
+	static const local::TypeT<1 << 22> tBMI2; // bzhi, mulx, pdep, pext, rorx, sarx, shlx, shrx
+	static const local::TypeT<1 << 23> tLZCNT;
+	static const local::TypeT<1 << 24> tINTEL;
+	static const local::TypeT<1 << 25> tAMD;
+	static const local::TypeT<1 << 26> tENHANCED_REP; // enhanced rep movsb/stosb
+	static const local::TypeT<1 << 27> tRDRAND;
+	static const local::TypeT<1 << 28> tADX; // adcx, adox
+	static const local::TypeT<1 << 29> tRDSEED; // rdseed
+	static const local::TypeT<1 << 30> tSMAP; // stac
+	static const local::TypeT<uint64_t(1) << 31> tHLE; // xacquire, xrelease, xtest
+	static const local::TypeT<uint64_t(1) << 32> tRTM; // xbegin, xend, xabort
+	static const local::TypeT<uint64_t(1) << 33> tF16C; // vcvtph2ps, vcvtps2ph
+	static const local::TypeT<uint64_t(1) << 34> tMOVBE; // mobve
+	static const local::TypeT<uint64_t(1) << 35> tAVX512F;
+	static const local::TypeT<uint64_t(1) << 36> tAVX512DQ;
+	static const local::TypeT<uint64_t(1) << 37> tAVX512_IFMA;
+	static const local::TypeT<uint64_t(1) << 37> tAVX512IFMA;// = tAVX512_IFMA;
+	static const local::TypeT<uint64_t(1) << 38> tAVX512PF;
+	static const local::TypeT<uint64_t(1) << 39> tAVX512ER;
+	static const local::TypeT<uint64_t(1) << 40> tAVX512CD;
+	static const local::TypeT<uint64_t(1) << 41> tAVX512BW;
+	static const local::TypeT<uint64_t(1) << 42> tAVX512VL;
+	static const local::TypeT<uint64_t(1) << 43> tAVX512_VBMI;
+	static const local::TypeT<uint64_t(1) << 43> tAVX512VBMI; // = tAVX512_VBMI; // changed by Intel's manual
+	static const local::TypeT<uint64_t(1) << 44> tAVX512_4VNNIW;
+	static const local::TypeT<uint64_t(1) << 45> tAVX512_4FMAPS;
+	static const local::TypeT<uint64_t(1) << 46> tPREFETCHWT1;
+	static const local::TypeT<uint64_t(1) << 47> tPREFETCHW;
+	static const local::TypeT<uint64_t(1) << 48> tSHA;
+	static const local::TypeT<uint64_t(1) << 49> tMPX;
+	static const local::TypeT<uint64_t(1) << 50> tAVX512_VBMI2;
+	static const local::TypeT<uint64_t(1) << 51> tGFNI;
+	static const local::TypeT<uint64_t(1) << 52> tVAES;
+	static const local::TypeT<uint64_t(1) << 53> tVPCLMULQDQ;
+	static const local::TypeT<uint64_t(1) << 54> tAVX512_VNNI;
+	static const local::TypeT<uint64_t(1) << 55> tAVX512_BITALG;
+	static const local::TypeT<uint64_t(1) << 56> tAVX512_VPOPCNTDQ;
+	static const local::TypeT<uint64_t(1) << 57> tAVX512_BF16;
+	static const local::TypeT<uint64_t(1) << 58> tAVX512_VP2INTERSECT;
+	static const local::TypeT<uint64_t(1) << 59> tAMX_TILE;
+	static const local::TypeT<uint64_t(1) << 60> tAMX_INT8;
+	static const local::TypeT<uint64_t(1) << 61> tAMX_BF16;
+	static const local::TypeT<uint64_t(1) << 62> tAVX_VNNI;
+	static const local::TypeT<uint64_t(1) << 63> tCLFLUSHOPT;
+	static const local::TypeT<0, 1 << 0> tCLDEMOTE;
+	static const local::TypeT<0, 1 << 1> tMOVDIRI;
+	static const local::TypeT<0, 1 << 2> tMOVDIR64B;
+	static const local::TypeT<0, 1 << 3> tCLZERO; // AMD Zen

 	Cpu()
 		: type_(NONE)
@ -378,13 +410,13 @@ public:
 		, coresSharignDataCache_()
 		, dataCacheLevels_(0)
 	{
-		unsigned int data[4] = {};
-		const unsigned int& EAX = data[0];
-		const unsigned int& EBX = data[1];
-		const unsigned int& ECX = data[2];
-		const unsigned int& EDX = data[3];
+		uint32_t data[4] = {};
+		const uint32_t& EAX = data[0];
+		const uint32_t& EBX = data[1];
+		const uint32_t& ECX = data[2];
+		const uint32_t& EDX = data[3];
 		getCpuid(0, data);
-		const unsigned int maxNum = EAX;
+		const uint32_t maxNum = EAX;
 		static const char intel[] = "ntel";
 		static const char amd[] = "cAMD";
 		if (ECX == get32bitAsBE(amd)) {
@ -407,7 +439,8 @@ public:

 		// Extended flags information
 		getCpuid(0x80000000, data);
-		if (EAX >= 0x80000001) {
+		const uint32_t maxExtendedNum = EAX;
+		if (maxExtendedNum >= 0x80000001) {
 			getCpuid(0x80000001, data);

 			if (EDX & (1U << 31)) type_ |= t3DN;
@ -419,6 +452,11 @@ public:
 			if (ECX & (1U << 8)) type_ |= tPREFETCHW;
 		}

+		if (maxExtendedNum >= 0x80000008) {
+			getCpuid(0x80000008, data);
+			if (EBX & (1U << 0)) type_ |= tCLZERO;
+		}
+
 		getCpuid(1, data);
 		if (ECX & (1U << 0)) type_ |= tSSE3;
 		if (ECX & (1U << 9)) type_ |= tSSSE3;
@ -469,6 +507,7 @@ public:
 						if (EDX & (1U << 2)) type_ |= tAVX512_4VNNIW;
 						if (EDX & (1U << 3)) type_ |= tAVX512_4FMAPS;
 						if (EDX & (1U << 8)) type_ |= tAVX512_VP2INTERSECT;
+						if ((type_ & tAVX512BW) && (EDX & (1U << 23))) type_ |= tAVX512_FP16;
 					}
 				}
 			}
@ -483,11 +522,16 @@ public:
 			if (EBX & (1U << 18)) type_ |= tRDSEED;
 			if (EBX & (1U << 19)) type_ |= tADX;
 			if (EBX & (1U << 20)) type_ |= tSMAP;
+			if (EBX & (1U << 23)) type_ |= tCLFLUSHOPT;
 			if (EBX & (1U << 4)) type_ |= tHLE;
 			if (EBX & (1U << 11)) type_ |= tRTM;
 			if (EBX & (1U << 14)) type_ |= tMPX;
 			if (EBX & (1U << 29)) type_ |= tSHA;
 			if (ECX & (1U << 0)) type_ |= tPREFETCHWT1;
+			if (ECX & (1U << 5)) type_ |= tWAITPKG;
+			if (ECX & (1U << 25)) type_ |= tCLDEMOTE;
+			if (ECX & (1U << 27)) type_ |= tMOVDIRI;
+			if (ECX & (1U << 28)) type_ |= tMOVDIR64B;
 			if (EDX & (1U << 24)) type_ |= tAMX_TILE;
 			if (EDX & (1U << 25)) type_ |= tAMX_INT8;
 			if (EDX & (1U << 22)) type_ |= tAMX_BF16;
@ -511,9 +555,9 @@ public:
 		printf("display:family=%X, model=%X\n", displayFamily, displayModel);
 #endif
 	}
-	bool has(Type type) const
+	bool has(const Type& type) const
 	{
-		return (type & type_) != 0;
+		return (type & type_) == type;
 	}
 };

@ -526,7 +570,7 @@ public:
 	#ifdef _MSC_VER
 		return __rdtsc();
 	#else
-		unsigned int eax, edx;
+		uint32_t eax, edx;
 		__asm__ volatile("rdtsc" : "=a"(eax), "=d"(edx));
 		return ((uint64_t)edx << 32) | eax;
 	#endif
@ -563,7 +607,7 @@ const int UseRDX = 1 << 7;

 class Pack {
 	static const size_t maxTblNum = 15;
-	const Xbyak::Reg64 *tbl_[maxTblNum];
+	Xbyak::Reg64 tbl_[maxTblNum];
 	size_t n_;
 public:
 	Pack() : tbl_(), n_(0) {}
@ -580,32 +624,36 @@ public:
 		return *this;
 	}
 	Pack(const Xbyak::Reg64& t0)
-	{ n_ = 1; tbl_[0] = &t0; }
+	{ n_ = 1; tbl_[0] = t0; }
 	Pack(const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 2; tbl_[0] = &t0; tbl_[1] = &t1; }
+	{ n_ = 2; tbl_[0] = t0; tbl_[1] = t1; }
 	Pack(const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 3; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; }
+	{ n_ = 3; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; }
 	Pack(const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 4; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; }
+	{ n_ = 4; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; }
 	Pack(const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 5; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; }
+	{ n_ = 5; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; }
 	Pack(const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 6; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; tbl_[5] = &t5; }
+	{ n_ = 6; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; }
 	Pack(const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 7; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; tbl_[5] = &t5; tbl_[6] = &t6; }
+	{ n_ = 7; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; }
 	Pack(const Xbyak::Reg64& t7, const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 8; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; tbl_[5] = &t5; tbl_[6] = &t6; tbl_[7] = &t7; }
+	{ n_ = 8; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; tbl_[7] = t7; }
 	Pack(const Xbyak::Reg64& t8, const Xbyak::Reg64& t7, const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 9; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; tbl_[5] = &t5; tbl_[6] = &t6; tbl_[7] = &t7; tbl_[8] = &t8; }
+	{ n_ = 9; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; tbl_[7] = t7; tbl_[8] = t8; }
 	Pack(const Xbyak::Reg64& t9, const Xbyak::Reg64& t8, const Xbyak::Reg64& t7, const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
-	{ n_ = 10; tbl_[0] = &t0; tbl_[1] = &t1; tbl_[2] = &t2; tbl_[3] = &t3; tbl_[4] = &t4; tbl_[5] = &t5; tbl_[6] = &t6; tbl_[7] = &t7; tbl_[8] = &t8; tbl_[9] = &t9; }
+	{ n_ = 10; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; tbl_[7] = t7; tbl_[8] = t8; tbl_[9] = t9; }
+	Pack(const Xbyak::Reg64& ta, const Xbyak::Reg64& t9, const Xbyak::Reg64& t8, const Xbyak::Reg64& t7, const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
+	{ n_ = 11; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; tbl_[7] = t7; tbl_[8] = t8; tbl_[9] = t9; tbl_[10] = ta; }
+	Pack(const Xbyak::Reg64& tb, const Xbyak::Reg64& ta, const Xbyak::Reg64& t9, const Xbyak::Reg64& t8, const Xbyak::Reg64& t7, const Xbyak::Reg64& t6, const Xbyak::Reg64& t5, const Xbyak::Reg64& t4, const Xbyak::Reg64& t3, const Xbyak::Reg64& t2, const Xbyak::Reg64& t1, const Xbyak::Reg64& t0)
+	{ n_ = 12; tbl_[0] = t0; tbl_[1] = t1; tbl_[2] = t2; tbl_[3] = t3; tbl_[4] = t4; tbl_[5] = t5; tbl_[6] = t6; tbl_[7] = t7; tbl_[8] = t8; tbl_[9] = t9; tbl_[10] = ta; tbl_[11] = tb; }
 	Pack& append(const Xbyak::Reg64& t)
 	{
 		if (n_ == maxTblNum) {
 			fprintf(stderr, "ERR Pack::can't append\n");
 			XBYAK_THROW_RET(ERR_BAD_PARAMETER, *this)
 		}
-		tbl_[n_++] = &t;
+		tbl_[n_++] = t;
 		return *this;
 	}
 	void init(const Xbyak::Reg64 *tbl, size_t n)
@ -616,7 +664,7 @@ public:
 		}
 		n_ = n;
 		for (size_t i = 0; i < n; i++) {
-			tbl_[i] = &tbl[i];
+			tbl_[i] = tbl[i];
 		}
 	}
 	const Xbyak::Reg64& operator[](size_t n) const
@ -625,7 +673,7 @@ public:
 			fprintf(stderr, "ERR Pack bad n=%d(%d)\n", (int)n, (int)n_);
 			XBYAK_THROW_RET(ERR_BAD_PARAMETER, rax)
 		}
-		return *tbl_[n];
+		return tbl_[n];
 	}
 	size_t size() const { return n_; }
 	/*
@ -648,7 +696,7 @@ public:
 	void put() const
 	{
 		for (size_t i = 0; i < n_; i++) {
-			printf("%s ", tbl_[i]->toString());
+			printf("%s ", tbl_[i].toString());
 		}
 		printf("\n");
 	}