Merge branch 'dev'

.github/workflows/main.yml

@@ -19,6 +19,11 @@ jobs:
     steps:
     - uses: actions/checkout@v4
     - run: apt -y update
-    - run: apt -y install g++-multilib libboost-dev make nasm yasm
+    - run: apt -y install g++-multilib libboost-dev make nasm yasm wget xz-utils python3
     - run: make test
     - run: make -C sample CXXFLAGS="-DXBYAK_NO_EXCEPTION"
+    - run: |
+        cd test
+        wget https://downloadmirror.intel.com/831748/sde-external-9.44.0-2024-08-22-lin.tar.xz
+        tar xvf sde-external-9.44.0-2024-08-22-lin.tar.xz
+        env XED=sde-external-9.44.0-2024-08-22-lin/xed64 make xed_test

.gitignore

@@ -1 +1,2 @@
 /build* # cmake
+*CVS

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.5)
 
-project(xbyak LANGUAGES CXX VERSION 7.09.1)
+project(xbyak LANGUAGES CXX VERSION 7.10)
 
 file(GLOB headers xbyak/*.h)
 

doc/changelog.md

@@ -1,5 +1,6 @@
 # History
 
+* 2024/Oct/13 ver 7.10 support AVX10 integer and fp16 vnni, media new instructions. setDefaultEncoding is extended.
 * 2024/Oct/10 ver 7.09.1 fix the names of vpcompressb and vpcompressw
 * 2024/Oct/08 ver 7.09 support YMM embedded rounding of AVX10.2 and fix some mnemonics with {sae}/{er}.
 * 2024/Oct/07 ver 7.08 support rdfsbase etc.

doc/usage.md

@@ -106,18 +106,37 @@ vcvtpd2dq xmm19, [eax+32]{1to4}         --> vcvtpd2dq(xmm19, yword_b [eax+32]);
 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
+```
+
+## Selecting AVX512-VNNI, AVX-VNNI, AVX-VNNI-INT8 etc.
+Some mnemonics have two types of encodings: VEX and EVEX.
+The functions for these mnemonics include an optional parameter as the last argument to specify the encoding.
+The default behavior depends on the order in which the instruction was introduced (whether VEX or EVEX came first),
+and can be specified using setDefaultEncoding.
 
-vpdpbusd(xm0, xm1, xm2); // default encoding is EVEX
+```
+vpdpbusd(xm0, xm1, xm2); // default encoding: EVEX (AVX512-VNNI)
 vpdpbusd(xm0, xm1, xm2, EvexEncoding); // same as the above
-vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX encoding
+vpdpbusd(xm0, xm1, xm2, VexEncoding); // VEX (AVX-VNNI)
 setDefaultEncoding(VexEncoding); // default encoding is VEX
-vpdpbusd(xm0, xm1, xm2); // VEX encoding
+vpdpbusd(xm0, xm1, xm2); // VEX
+
+vmpsadbw(xm1, xm3, xm15, 3); // default encoding: VEX (AVX-VNNI)
+vmpsadbw(xm1, xm3, xm15, 3, VexEncoding); // same as the above
+vmpsadbw(xm1, xm3, xm15, 3, EvexEncoding); // EVEX (AVX10.2)
+setDefaultEncoding(VexEncoding, EvexEncoding); // use 2nd argument.
+vmpsadbw(xm1, xm3, xm15, 3); // EVEX
 ```
 
-- setDefaultEncoding(PreferredEncoding encoding);
-  - Set the default encoding to select EVEX or VEX.
-  - The default value is EvexEncoding.
-  - This function affects only an instruction that has a PreferredEncoding argument such as vpdpbusd.
+- `setDefaultEncoding(PreferredEncoding vnniEnc = EvexEncoding, PreferredEncoding avx10Enc = VexEncoding)`
+Control the default encoding of mnemonics with `Xbyak::PreferredEncoding` param.
+
+param|vnniEnc|avx10Enc
+-|-|-
+EvexEncoding|AVX512-VNNI|AVX10.2
+VexEncoding|AVX-VNNI|AVX-VNNI-INT8
+default|EvexEncoding|VexEncoding
+mnemonic|vpdpbusd, vpdpbusds, vpdpwssd, vpdpwssds|vmpsadbw, vpdpbssd, vpdpbssds, vpdpbsud, vpdpbsuds, vpdpbuud, vpdpbuuds, vpdpwsud vpdpwsuds vpdpwusd vpdpwusds vpdpwuud, vpdpwuuds
 
 ### Remark
 * `k1`, ..., `k7` are opmask registers.

gen/gen_avx512.cpp

@@ -137,8 +137,6 @@ 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" : "", s.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()
@@ -198,6 +196,19 @@ void putX_XM()
 		{ 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 },
+
+		{ 0x51, "vsqrtnepbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16 },
+
+		{ 0x2F, "vcomish", T_MUST_EVEX | T_MAP5 | T_EW0 | T_SAE_X | T_N2 },
+		{ 0x2E, "vucomish", T_MUST_EVEX | T_MAP5 | T_EW0 | T_SAE_X | T_N2 },
+
+		{ 0x2F, "vcomxsd", T_MUST_EVEX | T_F3 | T_0F | T_EW1 | T_SAE_X | T_N8 },
+		{ 0x2F, "vcomxsh", T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_SAE_X | T_N2 },
+		{ 0x2F, "vcomxss", T_MUST_EVEX | T_F2 | T_0F | T_EW0  | T_SAE_X | T_N4 },
+
+		{ 0x2E, "vucomxsd", T_MUST_EVEX | T_F3 | T_0F | T_EW1 | T_SAE_X | T_N8 },
+		{ 0x2E, "vucomxsh", T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_SAE_X | T_N2 },
+		{ 0x2E, "vucomxss", T_MUST_EVEX | T_F2 | T_0F | T_EW0  | T_SAE_X | T_N4 },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@@ -316,6 +327,9 @@ void putX_X_XM_IMM()
 		{ 0x77, "vpermi2ps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false },
 		{ 0x77, "vpermi2pd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
 
+		{ 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 },
+
 		{ 0x25, "vpternlogd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, true },
 		{ 0x25, "vpternlogq", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, true },
 
@@ -401,6 +415,38 @@ void putX_X_XM_IMM()
 		{ 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 },
+
+		{ 0x58, "vaddnepbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x5E, "vdivnepbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x5F, "vmaxpbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x5D, "vminpbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x59, "vmulnepbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x2C, "vscalefpbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16,false },
+		{ 0x5C, "vsubnepbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+
+		{ 0x98, "vfmadd132nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xA8, "vfmadd213nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xB8, "vfmadd231nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+
+		{ 0x9C, "vfnmadd132nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xAC, "vfnmadd213nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xBC, "vfnmadd231nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+
+		{ 0x9A, "vfmsub132nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xAA, "vfmsub213nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xBA, "vfmsub231nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+
+		{ 0x9E, "vfnmsub132nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xAE, "vfnmsub213nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0xBE, "vfnmsub231nepbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+
+		{ 0x67, "vcvt2ps2phx", T_MUST_EVEX | T_66 | T_0F38 | T_EW0 | T_YMM | T_B32 | T_ER_Y | T_ER_Z, false },
+		{ 0x74, "vcvtne2ph2bf8", T_MUST_EVEX | T_F2 | T_0F38 | T_EW0 | T_YMM | T_B16 | T_N1, false },
+		{ 0x74, "vcvtne2ph2bf8s", T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_YMM | T_B16 | T_N1, false },
+		{ 0x18, "vcvtne2ph2hf8", T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_YMM | T_B16 | T_N1, false },
+		{ 0x1B, "vcvtne2ph2hf8s", T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_YMM | T_B16 | T_N1, false },
+
+		{ 0x52, "vdpphps", T_MUST_EVEX | T_0F38 | T_EW0 | T_YMM | T_B32, false },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@@ -410,6 +456,45 @@ void putX_X_XM_IMM()
 	}
 }
 
+void putX_X_XM_IMM_AVX10()
+{
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		uint64_t type;
+		uint64_t typeVex;
+		uint64_t typeEvex;
+		int sel;
+		bool hasIMM;
+	} tbl[] = {
+		// vpdpb[su,uu,ss]d[,s]
+		{ 0x50, "vpdpbssd", T_F2|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0x51, "vpdpbssds", T_F2|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0x50, "vpdpbsud", T_F3|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0x51, "vpdpbsuds", T_F3|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0x50, "vpdpbuud", T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0x51, "vpdpbuuds", T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+
+		// vpdpw[su,us,uu]d[,s]
+		{ 0xD2, "vpdpwsud", T_F3|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0xD3, "vpdpwsuds", T_F3|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0xD2, "vpdpwusd", T_66|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0xD3, "vpdpwusds", T_66|T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0xD2, "vpdpwuud", T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+		{ 0xD3, "vpdpwuuds", T_0F38|T_YMM, T_W0, T_EW0|T_B32, 1, false },
+
+		{ 0x42, "vmpsadbw", T_0F3A|T_YMM, T_66|T_W0|T_YMM, T_F3|T_0F3A|T_EW0|T_B32, 1, true },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
+		const Tbl *p = &tbl[i];
+		std::string s = type2String(p->type);
+		std::string sVex = type2String(p->typeVex);
+		std::string sEvex = type2String(p->typeEvex);
+		printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op%s, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, %s, 0x%02X, encoding, %s, %s, %s, %d); }\n"
+			, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? "imm" : "NONE", sVex.c_str(), sEvex.c_str(), p->sel);
+	}
+}
+
 void putShift()
 {
 	const struct Tbl {
@@ -571,6 +656,8 @@ void putCvt()
 
 		{ 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 },
+
+		{ 0x72, "vcvtneps2bf16", T_MUST_EVEX | T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32, 2 },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
@@ -758,6 +845,15 @@ void putX_XM_IMM()
 
 		{ 0x62, "vpexpandb", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_N1, false },
 		{ 0x62, "vpexpandw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_N2, false },
+
+		{ 0x2F, "vcomsbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_N2, false },
+		{ 0x42, "vgetexppbf16", T_MUST_EVEX | T_66 | T_MAP5 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x26, "vgetmantpbf16", T_MUST_EVEX | T_F2 | T_0F3A | T_EW0 | T_YMM | T_B16, true },
+		{ 0x4C, "vrcppbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x56, "vreducenepbf16", T_MUST_EVEX | T_F2 | T_0F3A | T_EW0 | T_YMM | T_B16, true },
+		{ 0x08, "vrndscalenepbf16", T_MUST_EVEX | T_F2 | T_0F3A | T_EW0 | T_YMM | T_B16, true },
+		{ 0x4E, "vrsqrtpbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
+		{ 0x2C, "vscalefpbf16", T_MUST_EVEX | T_MAP6 | T_EW0 | T_YMM | T_B16, false },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
@@ -952,6 +1048,41 @@ void putFP16()
 	putFP16_2();
 }
 
+void putAVX10_2()
+{
+	puts("void vcmppbf16(const Opmask& k, const Xmm& x, const Operand& op, uint8_t imm) { opVex(k, &x, op, T_MUST_EVEX|T_F2|T_0F3A|T_EW0|T_YMM|T_B16, 0xC2, imm); }");
+	puts("void vfpclasspbf16(const Opmask& k, const Operand& op, uint8_t imm) { opVex(k.changeBit(op.getBit()), 0, op, T_MUST_EVEX|T_F2|T_0F3A|T_EW0|T_YMM|T_B16, 0x66, imm); }");
+
+	const struct Tbl {
+		uint8_t code;
+		const char *name;
+		uint64_t type;
+	} tbl1[] = {
+		{ 0x74, "vcvtbiasph2bf8", T_MUST_EVEX | T_0F38 | T_EW0 |T_YMM | T_B16 },
+		{ 0x74, "vcvtbiasph2bf8s", T_MUST_EVEX | T_MAP5 | T_EW0 |T_YMM | T_B16 },
+		{ 0x18, "vcvtbiasph2hf8", T_MUST_EVEX | T_MAP5 | T_EW0 | T_YMM | T_B16 },
+		{ 0x1B, "vcvtbiasph2hf8s", T_MUST_EVEX | T_MAP5 | T_EW0 | T_YMM | T_B16 },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl1); i++) {
+		const Tbl *p = &tbl1[i];
+		std::string s = type2String(p->type);
+		printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt6(x1, x2, op, %s, 0x%02X); }\n" , p->name, s.c_str(), p->code);
+	}
+	puts("void vcvthf82ph(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, T_MUST_EVEX | T_F2 | T_MAP5 | T_EW0 | T_YMM | T_N1, 0x1E); }");
+
+	const Tbl tbl2[] = {
+		{ 0x74, "vcvtneph2bf8", T_MUST_EVEX | T_F3 | T_0F38 | T_EW0 | T_YMM | T_B16 },
+		{ 0x74, "vcvtneph2bf8s", T_MUST_EVEX | T_F3 |T_MAP5 | T_EW0 | T_YMM | T_B16 },
+		{ 0x18, "vcvtneph2hf8", T_MUST_EVEX | T_F3 |T_MAP5 | T_EW0 | T_YMM | T_B16 },
+		{ 0x1B, "vcvtneph2hf8s", T_MUST_EVEX | T_F3 |T_MAP5 | T_EW0 | T_YMM | T_B16 },
+	};
+	for (size_t i = 0; i < NUM_OF_ARRAY(tbl2); i++) {
+		const Tbl *p = &tbl2[i];
+		std::string s = type2String(p->type);
+		printf("void %s(const Xmm& x, const Operand& op) { opCvt2(x, op, %s, 0x%02X); }\n" , p->name, s.c_str(), p->code);
+	}
+}
+
 int main(int argc, char *[])
 {
 	bool only64bit = argc == 2;
@@ -966,6 +1097,7 @@ int main(int argc, char *[])
 	putM_X();
 	putXM_X();
 	putX_X_XM_IMM();
+	putX_X_XM_IMM_AVX10();
 	putShift();
 	putExtractInsert();
 	putCvt();
@@ -977,4 +1109,5 @@ int main(int argc, char *[])
 	putScatter();
 	putV4FMA();
 	putFP16();
+	putAVX10_2();
 }

gen/gen_code.cpp

@@ -57,7 +57,7 @@ void putX_X_XM(bool omitOnly)
 			{ 0x0C, "blendps", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 3 },
 			{ 0x41, "dppd", T_0F3A | T_66 | T_W0, true, true, 3 },
 			{ 0x40, "dpps", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 3 },
-			{ 0x42, "mpsadbw", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 3 },
+			{ 0x42, "mpsadbw", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 1 },
 			{ 0x0E, "pblendw", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 3 },
 			{ 0x02, "pblendd", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 2 },
 			{ 0x0B, "roundsd", T_0F3A | T_66 | T_W0, true, true, 3 },
@@ -1802,7 +1802,6 @@ void put()
 			const Tbl& p = tbl[i];
 			printf("void %s(const Xmm& x, const Address& addr) { opVex(x, 0, addr, %s, 0x%02X); }\n", p.name, type2String(p.type).c_str(), p.code);
 		}
-		printf("void vcvtneps2bf16(const Xmm& x, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opCvt2(x, op, %s|orEvexIf(encoding), 0x72); }\n", type2String(T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32).c_str());
 	}
 	// haswell gpr(reg, reg, r/m)
 	{
@@ -1893,8 +1892,6 @@ void put()
 			{ 0x51, "vpdpbusds", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
 			{ 0x52, "vpdpwssd", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
 			{ 0x53, "vpdpwssds", T_66 | T_0F38 | T_YMM | T_EW0 | T_SAE_Z | T_B32},
-			{ 0xB4, "vpmadd52luq", T_66 | T_0F38 | T_YMM | T_EW1 | T_B64 },
-			{ 0xB5, "vpmadd52huq", T_66 | T_0F38 | T_YMM | T_EW1 | T_B64 },
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
@@ -1904,32 +1901,34 @@ void put()
 	}
 	// avx-vnni-int8
 	// avx-vnni-int16
+#if 0
 	{
 		const struct Tbl {
 			uint8_t code;
 			const char *name;
 			uint64_t type;
 		} tbl[] = {
-			{ 0x50, "vpdpbssd", T_F2 | T_0F38 | T_W0 | T_YMM },
-			{ 0x51, "vpdpbssds", T_F2 | T_0F38 | T_W0 | T_YMM },
-			{ 0x50, "vpdpbsud", T_F3 | T_0F38 | T_W0 | T_YMM },
-			{ 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 },
+//			{ 0x50, "vpdpbssd", T_F2 | T_0F38 | T_W0 | T_YMM },
+//			{ 0x51, "vpdpbssds", T_F2 | T_0F38 | T_W0 | T_YMM },
+//			{ 0x50, "vpdpbsud", T_F3 | T_0F38 | T_W0 | T_YMM },
+//			{ 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];
 			std::string s = type2String(p->type);
 			printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 		}
 	}
+#endif
 }
 
 void put32()

meson.build

@@ -5,7 +5,7 @@
 project(
 	'xbyak',
 	'cpp',
-	version: '7.09.1',
+	version: '7.10',
 	license: 'BSD-3-Clause',
 	default_options: 'b_ndebug=if-release'
 )

readme.md

@@ -1,5 +1,5 @@
 
-# Xbyak 7.09.1 [![Badge Build]][Build Status]
+# Xbyak 7.10 [![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 7.09.1
+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 7.10
 
 -----------------------------------------------------------------------------
 ◎概要

test/Makefile

@@ -60,9 +60,9 @@ apx: apx.cpp $(XBYAK_INC)
 avx10_test: avx10_test.cpp $(XBYAK_INC)
 	$(CXX) $(CFLAGS) avx10_test.cpp -o $@ -DXBYAK64
 
-TEST_FILES=avx10.txt misc.txt
+TEST_FILES=old.txt new-ymm.txt bf16.txt comp.txt convert.txt
 xed_test:
-	@for target in $(addprefix target/, $(TEST_FILES)); do ./test_by_xed.sh $$target; done
+	@for target in $(addprefix avx10/, $(TEST_FILES)); do ./test_by_xed.sh $$target; done
 
 test_nm: normalize_prefix $(TARGET)
 	$(MAKE) -C ../gen