Merge branch 'dev'

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 2.6...3.0.2)
 
-project(xbyak LANGUAGES CXX VERSION 6.72)
+project(xbyak LANGUAGES CXX VERSION 6.73)
 
 file(GLOB headers xbyak/*.h)
 

doc/changelog.md

@@ -1,5 +1,6 @@
 # History
 
+* 2023/Aug/07 ver 6.73 add sha512/sm3/sm4/avx-vnni-int16
 * 2023/Aug/02 ver 6.72 add xbegin/xabort/xend
 * 2023/Jul/27 ver 6.71 Allocator supports huge page
 * 2023/Jul/05 ver 6.70 add alias of vclmulqdq, correct alias of pclmulqdq

gen/gen_code.cpp

@@ -208,6 +208,11 @@ void putX_X_XM(bool omitOnly)
 			{ 0xCF, "gf2p8affineinvqb", T_66 | T_0F3A | T_W1 | T_EVEX | T_YMM | T_EW1 | T_SAE_Z | T_B64, true, false, 3 },
 			{ 0xCE, "gf2p8affineqb", T_66 | T_0F3A | T_W1 | T_EVEX | T_YMM | T_EW1 | T_SAE_Z | T_B64, true, false, 3 },
 			{ 0xCF, "gf2p8mulb", T_66 | T_0F38 | T_W0 | T_EVEX | T_YMM | T_EW0 | T_SAE_Z, false, false, 3 },
+			{ 0xDA, "sm3msg1", T_0F38 | T_W0 | T_EVEX | T_EW0, false, false, 2 },
+			{ 0xDA, "sm3msg2", T_66 | T_0F38 | T_W0 | T_EVEX | T_EW0, false, false, 2 },
+			{ 0xDE, "sm3rnds2", T_66 | T_0F3A | T_W0 | T_EVEX | T_EW0, true, false, 2 },
+			{ 0xDA, "sm4key4", T_F3 | T_0F38 | T_W0 | T_EVEX | T_EW0, false, false, 2 },
+			{ 0xDA, "sm4rnds4", T_F2 | T_0F38 | T_W0 | T_EVEX | T_EW0, false, false, 2 },
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
@@ -1116,6 +1121,10 @@ void put()
 		puts("void cldemote(const Address& addr) { opMIB(addr, eax, 0x0F, 0x1C); }");
 		puts("void xabort(uint8_t imm) { db(0xC6); db(0xF8); db(imm); }");
 		puts("void xbegin(uint32_t rel) { db(0xC7); db(0xF8); dd(rel); }");
+
+		puts("void vsha512msg1(const Ymm& y, const Xmm& x) { if (!(y.isYMM() && x.isXMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y, 0, x, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCC); }");
+		puts("void vsha512msg2(const Ymm& y1, const Ymm& y2) { if (!(y1.isYMM() && y2.isYMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y1, 0, y2, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCD); }");
+		puts("void vsha512rnds2(const Ymm& y1, const Ymm& y2, const Xmm& x) { if (!(y1.isYMM() && y2.isYMM() && x.isXMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y1, &y2, x, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCB); }");
 	}
 	{
 		const struct Tbl {
@@ -1808,6 +1817,7 @@ void put()
 		}
 	}
 	// avx-vnni-int8
+	// avx-vnni-int16
 	{
 		const struct Tbl {
 			uint8_t code;
@@ -1820,6 +1830,13 @@ void put()
 			{ 0x51, "vpdpbsuds", T_F3 | T_0F38 | T_W0 | T_YMM },
 			{ 0x50, "vpdpbuud", T_0F38 | T_W0 | T_YMM },
 			{ 0x51, "vpdpbuuds", T_0F38 | T_W0 | T_YMM },
+
+			{ 0xD2, "vpdpwsud", T_F3 | T_0F38 | T_W0 | T_YMM },
+			{ 0xD3, "vpdpwsuds", T_F3 | T_0F38 | T_W0 | T_YMM },
+			{ 0xD2, "vpdpwusd", T_66 | T_0F38 | T_W0 | T_YMM },
+			{ 0xD3, "vpdpwusds", T_66 | T_0F38 | T_W0 | T_YMM },
+			{ 0xD2, "vpdpwuud", T_0F38 | T_W0 | T_YMM },
+			{ 0xD3, "vpdpwuuds", T_0F38 | T_W0 | T_YMM },
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];

meson.build

@@ -5,7 +5,7 @@
 project(
 	'xbyak',
 	'cpp',
-	version: '6.72',
+	version: '6.73',
 	license: 'BSD-3-Clause',
 	default_options: 'b_ndebug=if-release'
 )

readme.md

@@ -1,5 +1,5 @@
 
-# Xbyak 6.72 [![Badge Build]][Build Status]
+# Xbyak 6.73 [![Badge Build]][Build Status]
 
 *A C++ JIT assembler for x86 (IA32), x64 (AMD64, x86-64)*
 

readme.txt

@@ -1,5 +1,5 @@
 
-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 6.72
+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 6.73
 
 -----------------------------------------------------------------------------
 ◎概要
@@ -402,6 +402,7 @@ sample/{echo,hello}.bfは http://www.kmonos.net/alang/etc/brainfuck.php から
 -----------------------------------------------------------------------------
 ◎履歴
 
+2023/08/07 ver 6.73 sha512/sm3/sm4/avx-vnni-int16追加
 2023/08/02 ver 6.72 xabort, xbegin, xend追加
 2023/07/27 ver 6.71 Allocatorでhuge pageを考慮する。
 2023/07/05 ver 6.70 vpclmulqdqのailas追加

sample/Makefile

@@ -122,7 +122,7 @@ toyvm : toyvm.cpp $(XBYAK_INC)
 static_buf: static_buf.cpp $(XBYAK_INC)
 static_buf64: static_buf.cpp $(XBYAK_INC)
 test_util : test_util.cpp $(XBYAK_INC) ../xbyak/xbyak_util.h
-test_util2 : test_util.cpp $(XBYAK_INC) ../xbyak/xbyak_util.h
+test_util64 : 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)

sample/cpuid/arl.txt

@@ -0,0 +1,2 @@
+vendor intel
+ mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b uintr serialize avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd sha512 sm3 sm4 avx_vnni_int16

sample/cpuid/cpuid.sh

@@ -1,12 +1,25 @@
 #!/bin/bash
 
+UPDATE=0
+if [ $# -eq 1 ]; then
+UPDATE=1
+fi
+
+if [ $UPDATE == 1 ]; then
+  echo "UPDATE"
+fi
+
 make -C ../ test_util64
 
-cpus=(p4p mrm pnr nhm wsm snb ivb hsw bdw slt slm glm glp tnt skl cnl icl skx clx cpx icx knl knm tgl adl mtl rpl spr gnr srf grr)
+cpus=(p4p mrm pnr nhm wsm snb ivb hsw bdw slt slm glm glp tnt skl cnl icl skx clx cpx icx knl knm tgl adl mtl rpl spr gnr srf grr arl lnl)
 
 for cpu in ${cpus[@]} ; do
   echo $cpu
-  ~/bin/sde -$cpu -- ../test_util64 -cpuid > tmp.txt
-  diff $cpu.txt tmp.txt
+  if [ $UPDATE == 1 ]; then
+    ~/bin/sde -$cpu -- ../test_util64 -cpuid > $cpu.txt
+  else
+    ~/bin/sde -$cpu -- ../test_util64 -cpuid > tmp.txt
+    diff $cpu.txt tmp.txt
+  fi
 done
 

sample/cpuid/grr.txt

@@ -1,2 +1,2 @@
 vendor intel
- mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b serialize avx_vnni_int8 avx_ne_convert avx_ifma rao-int cmpccxadd
+ mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b uintr serialize avx_vnni_int8 avx_ne_convert avx_ifma rao-int cmpccxadd

sample/cpuid/lnl.txt

@@ -0,0 +1,2 @@
+vendor intel
+ mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt prefetchw enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b uintr serialize avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd sha512 sm3 sm4 avx_vnni_int16

sample/cpuid/srf.txt

@@ -1,2 +1,2 @@
 vendor intel
- mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b serialize avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd
+ mmx mmx2 cmov sse sse2 sse3 ssse3 sse41 sse42 popcnt aesni rdtscp xsave(xgetvb) osxsave pclmulqdq avx fma avx2 bmi1 bmi2 lzcnt enh_rep rdrand adx rdseed smap sha f16c movbe gfni vaes vpclmulqdq avx_vnni waitpkg clflushopt cldemote movdiri movdir64b uintr serialize avx_vnni_int8 avx_ne_convert avx_ifma cmpccxadd

sample/test_util.cpp

@@ -99,6 +99,10 @@ void putCPUinfo(bool onlyCpuidFeature)
 		{ Cpu::tRAO_INT, "rao-int" },
 		{ Cpu::tCMPCCXADD, "cmpccxadd" },
 		{ Cpu::tPREFETCHITI, "prefetchiti" },
+		{ Cpu::tSHA512, "sha512" },
+		{ Cpu::tSM3, "sm3" },
+		{ Cpu::tSM4, "sm4" },
+		{ Cpu::tAVX_VNNI_INT16, "avx_vnni_int16" },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		if (cpu.has(tbl[i].type)) printf(" %s", tbl[i].str);

test/misc.cpp

@@ -2170,4 +2170,115 @@ CYBOZU_TEST_AUTO(prefetchiti)
 	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
 }
 
+CYBOZU_TEST_AUTO(crypto)
+{
+	struct Code : Xbyak::CodeGenerator {
+		Code()
+		{
+			vsha512msg1(ymm3, xmm5);
+			vsha512msg2(ymm9, ymm10);
+			vsha512rnds2(ymm1, ymm3, xmm2);
+
+			vsm3msg1(xmm1, xmm2, xmm3);
+			vsm3msg1(xmm1, xmm2, ptr [rax]);
+			vsm3msg2(xmm5, xmm7, xmm3);
+			vsm3msg2(xmm5, xmm6, ptr [rax]);
+			vsm3rnds2(xmm5, xmm7, xmm3, 0x12);
+			vsm3rnds2(xmm5, xmm7, ptr [rcx], 0x34);
+
+			vsm4key4(xmm1, xmm2, xmm3);
+			vsm4key4(xmm1, xmm2, ptr [rdx]);
+			vsm4rnds4(xmm1, xmm2, xmm3);
+			vsm4rnds4(xmm5, xmm6, ptr [rcx+rax*4]);
+		}
+	} c;
+	const uint8_t tbl[] = {
+		// sha512
+		0xc4, 0xe2, 0x7f, 0xcc, 0xdd,
+		0xc4, 0x42, 0x7f, 0xcd, 0xca,
+		0xc4, 0xe2, 0x67, 0xcb, 0xca,
+
+		// sm3
+		0xC4, 0xE2, 0x68, 0xDA, 0xCB,
+		0xC4, 0xE2, 0x68, 0xDA, 0x08,
+		0xC4, 0xE2, 0x41, 0xDA, 0xEB,
+		0xC4, 0xE2, 0x49, 0xDA, 0x28,
+		0xC4, 0xE3, 0x41, 0xDE, 0xEB, 0x12,
+		0xC4, 0xE3, 0x41, 0xDE, 0x29, 0x34,
+
+		// sm4
+		0xc4, 0xe2, 0x6a, 0xda, 0xcb,
+		0xc4, 0xe2, 0x6a, 0xda, 0x0a,
+		0xc4, 0xe2, 0x6b, 0xda, 0xcb,
+		0xc4, 0xe2, 0x4b, 0xda, 0x2c, 0x81,
+	};
+	const size_t n = sizeof(tbl) / sizeof(tbl[0]);
+	CYBOZU_TEST_EQUAL(c.getSize(), n);
+	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+}
+
+CYBOZU_TEST_AUTO(avx_vnni_int)
+{
+	struct Code : Xbyak::CodeGenerator {
+		Code()
+		{
+			vpdpbssd(xmm1, xmm2, xmm3);
+			vpdpbssd(ymm1, ymm2, ptr [rax]);
+			vpdpbssds(xmm1, xmm2, xmm3);
+			vpdpbssds(ymm1, ymm2, ptr [rax]);
+			vpdpbsud(xmm1, xmm2, xmm3);
+			vpdpbsud(ymm1, ymm2, ptr [rax]);
+			vpdpbsuds(xmm1, xmm2, xmm3);
+			vpdpbsuds(ymm1, ymm2, ptr [rax]);
+			vpdpbuud(xmm1, xmm2, xmm3);
+			vpdpbuud(ymm1, ymm2, ptr [rax]);
+			vpdpbuuds(xmm1, xmm2, xmm3);
+			vpdpbuuds(ymm1, ymm2, ptr [rax]);
+
+			vpdpwsud(xmm1, xmm2, xmm3);
+			vpdpwsud(ymm1, ymm2, ptr [rax]);
+			vpdpwsuds(xmm1, xmm2, xmm3);
+			vpdpwsuds(ymm1, ymm2, ptr [rax]);
+			vpdpwusd(xmm1, xmm2, xmm3);
+			vpdpwusd(ymm1, ymm2, ptr [rax]);
+			vpdpwusds(xmm1, xmm2, xmm3);
+			vpdpwusds(ymm1, ymm2, ptr [rax]);
+			vpdpwuud(xmm1, xmm2, xmm3);
+			vpdpwuud(ymm1, ymm2, ptr [rax]);
+			vpdpwuuds(xmm1, xmm2, xmm3);
+			vpdpwuuds(ymm1, ymm2, ptr [rax]);
+		}
+	} c;
+	const uint8_t tbl[] = {
+		0xc4, 0xe2, 0x6b, 0x50, 0xcb,
+		0xc4, 0xe2, 0x6f, 0x50, 0x08,
+		0xc4, 0xe2, 0x6b, 0x51, 0xcb,
+		0xc4, 0xe2, 0x6f, 0x51, 0x08,
+		0xc4, 0xe2, 0x6a, 0x50, 0xcb,
+		0xc4, 0xe2, 0x6e, 0x50, 0x08,
+		0xc4, 0xe2, 0x6a, 0x51, 0xcb,
+		0xc4, 0xe2, 0x6e, 0x51, 0x08,
+		0xc4, 0xe2, 0x68, 0x50, 0xcb,
+		0xc4, 0xe2, 0x6c, 0x50, 0x08,
+		0xc4, 0xe2, 0x68, 0x51, 0xcb,
+		0xc4, 0xe2, 0x6c, 0x51, 0x08,
+		0xc4, 0xe2, 0x6a, 0xd2, 0xcb,
+		0xc4, 0xe2, 0x6e, 0xd2, 0x08,
+		0xc4, 0xe2, 0x6a, 0xd3, 0xcb,
+		0xc4, 0xe2, 0x6e, 0xd3, 0x08,
+		0xc4, 0xe2, 0x69, 0xd2, 0xcb,
+		0xc4, 0xe2, 0x6d, 0xd2, 0x08,
+		0xc4, 0xe2, 0x69, 0xd3, 0xcb,
+		0xc4, 0xe2, 0x6d, 0xd3, 0x08,
+		0xc4, 0xe2, 0x68, 0xd2, 0xcb,
+		0xc4, 0xe2, 0x6c, 0xd2, 0x08,
+		0xc4, 0xe2, 0x68, 0xd3, 0xcb,
+		0xc4, 0xe2, 0x6c, 0xd3, 0x08,
+	};
+	const size_t n = sizeof(tbl) / sizeof(tbl[0]);
+	CYBOZU_TEST_EQUAL(c.getSize(), n);
+	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+}
+
+
 #endif

xbyak/xbyak.h

@@ -155,7 +155,7 @@ namespace Xbyak {
 
 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x6720 /* 0xABCD = A.BC(.D) */
+	VERSION = 0x6730 /* 0xABCD = A.BC(.D) */
 };
 
 #ifndef MIE_INTEGER_TYPE_DEFINED

xbyak/xbyak_mnemonic.h

@@ -1,4 +1,4 @@
-const char *getVersionString() const { return "6.72"; }
+const char *getVersionString() const { return "6.73"; }
 void aadd(const Address& addr, const Reg32e &reg) { opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
 void aand(const Address& addr, const Reg32e &reg) { db(0x66); opModM(addr, reg, 0x0F, 0x38, 0x0FC); }
 void adc(const Operand& op, uint32_t imm) { opRM_I(op, imm, 0x10, 2); }
@@ -1219,6 +1219,12 @@ void vpdpbuud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1
 void vpdpbuuds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_0F38 | T_W0 | T_YMM, 0x51); }
 void vpdpwssd(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x52, encoding); }
 void vpdpwssds(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 0x53, encoding); }
+void vpdpwsud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_0F38 | T_W0 | T_YMM, 0xD2); }
+void vpdpwsuds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_0F38 | T_W0 | T_YMM, 0xD3); }
+void vpdpwusd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_YMM, 0xD2); }
+void vpdpwusds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_YMM, 0xD3); }
+void vpdpwuud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_0F38 | T_W0 | T_YMM, 0xD2); }
+void vpdpwuuds(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_0F38 | T_W0 | T_YMM, 0xD3); }
 void vperm2f128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x06, imm); }
 void vperm2i128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8_t imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x46, imm); }
 void vpermd(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x36); }
@@ -1348,8 +1354,16 @@ void vroundsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { op
 void vroundss(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, 0x0A, imm); }
 void vrsqrtps(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_0F | T_YMM, 0x52); }
 void vrsqrtss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_0F, 0x52); }
+void vsha512msg1(const Ymm& y, const Xmm& x) { if (!(y.isYMM() && x.isXMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y, 0, x, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCC); }
+void vsha512msg2(const Ymm& y1, const Ymm& y2) { if (!(y1.isYMM() && y2.isYMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y1, 0, y2, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCD); }
+void vsha512rnds2(const Ymm& y1, const Ymm& y2, const Xmm& x) { if (!(y1.isYMM() && y2.isYMM() && x.isXMM())) XBYAK_THROW(ERR_BAD_PARAMETER) opVex(y1, &y2, x, T_F2 | T_0F38 | T_W0 | T_YMM, 0xCB); }
 void vshufpd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX | T_B64, 0xC6, imm); }
 void vshufps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) { opAVX_X_X_XM(x1, x2, op, T_0F | T_EW0 | T_YMM | T_EVEX | T_B32, 0xC6, imm); }
+void vsm3msg1(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_0F38 | T_W0 | T_EW0 | T_EVEX, 0xDA); }
+void vsm3msg2(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_EVEX, 0xDA); }
+void vsm3rnds2(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_EW0 | T_EVEX, 0xDE, imm); }
+void vsm4key4(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F3 | T_0F38 | T_W0 | T_EW0 | T_EVEX, 0xDA); }
+void vsm4rnds4(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_F2 | T_0F38 | T_W0 | T_EW0 | T_EVEX, 0xDA); }
 void vsqrtpd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX | T_ER_Z | T_B64, 0x51); }
 void vsqrtps(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_0F | T_EW0 | T_YMM | T_EVEX | T_ER_Z | T_B32, 0x51); }
 void vsqrtsd(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, 0x51); }

xbyak/xbyak_util.h

@@ -466,6 +466,10 @@ class Cpu {
 	XBYAK_DEFINE_TYPE(75, tSERIALIZE);
 	XBYAK_DEFINE_TYPE(76, tUINTR);
 	XBYAK_DEFINE_TYPE(77, tXSAVE);
+	XBYAK_DEFINE_TYPE(78, tSHA512);
+	XBYAK_DEFINE_TYPE(79, tSM3);
+	XBYAK_DEFINE_TYPE(80, tSM4);
+	XBYAK_DEFINE_TYPE(81, tAVX_VNNI_INT16);
 
 #undef XBYAK_SPLIT_ID
 #undef XBYAK_DEFINE_TYPE
@@ -608,6 +612,9 @@ class Cpu {
 			if (EDX & (1U << 25)) type_ |= tAMX_INT8;
 			if (maxNumSubLeaves >= 1) {
 				getCpuidEx(7, 1, data);
+				if (EAX & (1U << 0)) type_ |= tSHA512;
+				if (EAX & (1U << 1)) type_ |= tSM3;
+				if (EAX & (1U << 2)) type_ |= tSM4;
 				if (EAX & (1U << 3)) type_ |= tRAO_INT;
 				if (EAX & (1U << 4)) type_ |= tAVX_VNNI;
 				if (type_ & tAVX512F) {
@@ -618,6 +625,7 @@ class Cpu {
 				if (EAX & (1U << 23)) type_ |= tAVX_IFMA;
 				if (EDX & (1U << 4)) type_ |= tAVX_VNNI_INT8;
 				if (EDX & (1U << 5)) type_ |= tAVX_NE_CONVERT;
+				if (EDX & (1U << 10)) type_ |= tAVX_VNNI_INT16;
 				if (EDX & (1U << 14)) type_ |= tPREFETCHITI;
 			}
 		}