diff --git a/gen/avx_type.hpp b/gen/avx_type.hpp
index 0374c96..4436f28 100644
--- a/gen/avx_type.hpp
+++ b/gen/avx_type.hpp
@@ -15,15 +15,9 @@ std::string type2String(uint64_t type)
 		assert(low < int(sizeof(tbl) / sizeof(tbl[0])));
 		str = tbl[low - 1];
 	}
-	if (type & T_N_VL) {
-		str += "|T_N_VL";
-	}
-	if (type & T_VEX) {
-		str += "|T_VEX";
-	}
-	if ((type & T_NX_MASK) == T_DUP) {
-		str += "|T_DUP";
-	}
+	if (type & T_N_VL) str += "|T_N_VL";
+	if (type & T_VEX) str += "|T_VEX";
+	if ((type & T_NX_MASK) == T_DUP) str += "|T_DUP";
 	switch (type & T_F2) { // T_F2 = T_66|T_F3
 	case T_66: str += "|T_66"; break;
 	case T_F3: str += "|T_F3"; break;
@@ -44,75 +38,36 @@ std::string type2String(uint64_t type)
 			str += "|T_0F38";
 		}
 	}
-	if (type & T_0F3A) {
-		str += "|T_0F3A";
-	}
-	if (type & T_L0) {
-		str += "|T_L0";
-	}
-	if (type & T_L1) {
-		str += "|T_L1";
-	}
-	if (type & T_W0) {
-		str += "|T_W0";
-	}
-	if (type & T_W1) {
-		str += "|T_W1";
-	}
-	if (type & T_EW0) {
-		str += "|T_EW0";
-	}
-	if (type & T_EW1) {
-		str += "|T_EW1";
-	}
-	if (type & T_YMM) {
-		str += "|T_YMM";
-	}
-	if (type & T_EVEX) {
-		str += "|T_EVEX";
-	}
-	if (type & T_ER_X) {
-		str += "|T_ER_X";
-	}
-	if (type & T_ER_Y) {
-		str += "|T_ER_Y";
-	}
-	if (type & T_ER_Z) {
-		str += "|T_ER_Z";
-	}
-	if (type & T_ER_R) {
-		str += "|T_ER_R";
-	}
-	if (type & T_SAE_X) {
-		str += "|T_SAE_X";
-	}
-	if (type & T_SAE_Y) {
-		str += "|T_SAE_Y";
-	}
-	if (type & T_SAE_Z) {
-		str += "|T_SAE_Z";
-	}
-	if (type & T_MUST_EVEX) {
-		str += "|T_MUST_EVEX";
-	}
+	if (type & T_0F3A) str += "|T_0F3A";
+	if (type & T_L0) str += "|T_L0";
+	if (type & T_L1) str += "|T_L1";
+	if (type & T_W0) str += "|T_W0";
+	if (type & T_W1) str += "|T_W1";
+	if (type & T_EW0) str += "|T_EW0";
+	if (type & T_EW1) str += "|T_EW1";
+	if (type & T_YMM) str += "|T_YMM";
+	if (type & T_EVEX) str += "|T_EVEX";
+	if (type & T_ER_X) str += "|T_ER_X";
+	if (type & T_ER_Y) str += "|T_ER_Y";
+	if (type & T_ER_Z) str += "|T_ER_Z";
+	if (type & T_ER_R) str += "|T_ER_R";
+	if (type & T_SAE_X) str += "|T_SAE_X";
+	if (type & T_SAE_Y) str += "|T_SAE_Y";
+	if (type & T_SAE_Z) str += "|T_SAE_Z";
+	if (type & T_MUST_EVEX) str += "|T_MUST_EVEX";
+
 	switch (type & T_B16) { // T_B16 = T_B32 | T_B64
 	case T_B16: str += "|T_B16"; break;
 	case T_B32: str += "|T_B32"; break;
 	case T_B64: str += "|T_B64"; break;
 	default: break;
 	}
-	if (type & T_M_K) {
-		str += "|T_M_K";
-	}
-	if (type & T_VSIB) {
-		str += "|T_VSIB";
-	}
-	if (type & T_MEM_EVEX) {
-		str += "|T_MEM_EVEX";
-	}
-	if (type & T_NF) {
-		str += "|T_NF";
-	}
+	if (type & T_M_K) str += "|T_M_K";
+	if (type & T_VSIB) str += "|T_VSIB";
+	if (type & T_MEM_EVEX) str += "|T_MEM_EVEX";
+	if (type & T_NF) str += "|T_NF";
+	if (type & T_NO_OR1) str += "|T_NO_OR1";
+
 	if (str[0] == '|') str = str.substr(1);
 	return str;
 }
diff --git a/gen/avx_type_def.h b/gen/avx_type_def.h
index 094ef8f..5e74a93 100644
--- a/gen/avx_type_def.h
+++ b/gen/avx_type_def.h
@@ -1,51 +1,49 @@
 	// @@@begin of avx_type_def.h
-	enum AVXtype {
-		T_NONE = 0,
+	static const uint64_t T_NONE = 0ull;
 		// low 3 bit
-		T_N1 = 1,
-		T_N2 = 2,
-		T_N4 = 3,
-		T_N8 = 4,
-		T_N16 = 5,
-		T_N32 = 6,
-		T_NX_MASK = 7,
-		T_DUP = T_NX_MASK,//1 << 4, // N = (8, 32, 64)
-		T_N_VL = 1 << 3, // N * (1, 2, 4) for VL
-		T_VEX = 1 << 4,
-		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,
-		T_L0 = 1 << 11,
-		T_L1 = 1 << 12,
-		T_W0 = 1 << 13,
-		T_W1 = 1 << 14,
-		T_EW0 = 1 << 15,
-		T_EW1 = 1 << 16,
-		T_YMM = 1 << 17, // support YMM, ZMM
-		T_EVEX = 1 << 18,
-		T_ER_X = 1 << 19, // xmm{er}
-		T_ER_Y = 1 << 20, // ymm{er}
-		T_ER_Z = 1 << 21, // zmm{er}
-		T_SAE_X = 1 << 22, // xmm{sae}
-		T_SAE_Y = 1 << 23, // ymm{sae}
-		T_SAE_Z = 1 << 24, // zmm{sae}
-		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 = 1u << 30, // use evex if mem
-		T_FP16 = 1u << 31, // avx512-fp16
-		T_MAP5 = T_FP16 | T_0F,
-		T_MAP6 = T_FP16 | T_0F38,
-		T_NF = 1ull << 32, // T_nf
-		T_XXX
-	};
+	static const uint64_t T_N1 = 1ull;
+	static const uint64_t T_N2 = 2ull;
+	static const uint64_t T_N4 = 3ull;
+	static const uint64_t T_N8 = 4ull;
+	static const uint64_t T_N16 = 5ull;
+	static const uint64_t T_N32 = 6ull;
+	static const uint64_t T_NX_MASK = 7ull;
+	static const uint64_t T_DUP = T_NX_MASK;//1 << 4, // N = (8, 32, 64)
+	static const uint64_t T_N_VL = 1ull << 3; // N * (1, 2, 4) for VL
+	static const uint64_t T_VEX = 1ull << 4;
+	static const uint64_t T_66 = 1ull << 5; // pp = 1
+	static const uint64_t T_F3 = 1ull << 6; // pp = 2
+	static const uint64_t T_F2 = T_66 | T_F3; // pp = 3
+	static const uint64_t T_ER_R = 1ull << 7; // reg{er}
+	static const uint64_t T_0F = 1ull << 8;
+	static const uint64_t T_0F38 = 1ull << 9;
+	static const uint64_t T_0F3A = 1ull << 10;
+	static const uint64_t T_L0 = 1ull << 11;
+	static const uint64_t T_L1 = 1ull << 12;
+	static const uint64_t T_W0 = 1ull << 13;
+	static const uint64_t T_W1 = 1ull << 14;
+	static const uint64_t T_EW0 = 1ull << 15;
+	static const uint64_t T_EW1 = 1ull << 16;
+	static const uint64_t T_YMM = 1ull << 17; // support YMM, ZMM
+	static const uint64_t T_EVEX = 1ull << 18;
+	static const uint64_t T_ER_X = 1ull << 19; // xmm{er}
+	static const uint64_t T_ER_Y = 1ull << 20; // ymm{er}
+	static const uint64_t T_ER_Z = 1ull << 21; // zmm{er}
+	static const uint64_t T_SAE_X = 1ull << 22; // xmm{sae}
+	static const uint64_t T_SAE_Y = 1ull << 23; // ymm{sae}
+	static const uint64_t T_SAE_Z = 1ull << 24; // zmm{sae}
+	static const uint64_t T_MUST_EVEX = 1ull << 25; // contains T_EVEX
+	static const uint64_t T_B32 = 1ull << 26; // m32bcst
+	static const uint64_t T_B64 = 1ull << 27; // m64bcst
+	static const uint64_t T_B16 = T_B32 | T_B64; // m16bcst (Be careful)
+	static const uint64_t T_M_K = 1ull << 28; // mem{k}
+	static const uint64_t T_VSIB = 1ull << 29;
+	static const uint64_t T_MEM_EVEX = 1ull << 30; // use evex if mem
+	static const uint64_t T_FP16 = 1ull << 31; // avx512-fp16
+	static const uint64_t T_MAP5 = T_FP16 | T_0F;
+	static const uint64_t T_MAP6 = T_FP16 | T_0F38;
+	static const uint64_t T_NF = 1ull << 32; // T_nf
+	static const uint64_t T_NO_OR1 = 1ull << 33; // does not "code | 1"
 	// T_66 = 1, T_F3 = 2, T_F2 = 3
 	static inline uint32_t getPP(uint64_t type) { return (type >> 5) & 3; }
 	// @@@end of avx_type_def.h
diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 49d438a..7c4d9ba 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -142,9 +142,9 @@ void putVcmp()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
+		std::string s = type2String(p->type);
 		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" : "");
+			, 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); }");
@@ -210,8 +210,8 @@ void putX_XM()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
-		printf("void %s(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
+		std::string s = type2String(p->type);
+		printf("void %s(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 	}
 	puts("void vpabsq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_B64 | T_YMM, 0x1F); }");
 
@@ -242,8 +242,8 @@ void putM_X()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
-		printf("void %s(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
+		std::string s = type2String(p->type);
+		printf("void %s(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 	}
 }
 
@@ -265,8 +265,8 @@ void putXM_X()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
-		printf("void %s(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
+		std::string s = type2String(p->type);
+		printf("void %s(const Operand& op, const Xmm& x) { opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 	}
 }
 
@@ -413,9 +413,9 @@ void putX_X_XM_IMM()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
+		std::string s = type2String(p->type);
 		printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op%s) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X%s); }\n"
-			, p->name, p->hasIMM ? ", uint8_t imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
+			, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? ", imm" : "");
 	}
 }
 
@@ -435,8 +435,8 @@ void putShift()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type);
-		printf("void %s(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_X_XM(Xmm(x.getKind(), %d), x, op, %s, 0x%02X, imm); }\n", p.name, p.idx, type.c_str(), p.code);
+		std::string s = type2String(p.type);
+		printf("void %s(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_X_XM(Xmm(x.getKind(), %d), x, op, %s, 0x%02X, imm); }\n", p.name, p.idx, s.c_str(), p.code);
 	}
 }
 
@@ -461,9 +461,9 @@ void putExtractInsert()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
+			std::string s = type2String(p.type);
 			const char *kind = p.isZMM ? "Operand::MEM | Operand::YMM" : "Operand::MEM | Operand::XMM";
-			printf("void %s(const Operand& op, const %s& r, uint8_t imm) { if (!op.is(%s)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, %s, 0x%2X, imm); }\n", p.name, p.isZMM ? "Zmm" : "Ymm", kind, type.c_str(), p.code);
+			printf("void %s(const Operand& op, const %s& r, uint8_t imm) { if (!op.is(%s)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, %s, 0x%2X, imm); }\n", p.name, p.isZMM ? "Zmm" : "Ymm", kind, s.c_str(), p.code);
 		}
 	}
 	{
@@ -485,12 +485,12 @@ void putExtractInsert()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
+			std::string s = type2String(p.type);
 			const char *x = p.isZMM ? "Zmm" : "Ymm";
 			const char *cond = p.isZMM ? "op.is(Operand::MEM | Operand::YMM)" : "(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))";
 			printf("void %s(const %s& r1, const %s& r2, const Operand& op, uint8_t imm) {"
 				"if (!%s) XBYAK_THROW(ERR_BAD_COMBINATION) "
-				"opVex(r1, &r2, op, %s, 0x%2X, imm); }\n", p.name, x, x, cond, type.c_str(), p.code);
+				"opVex(r1, &r2, op, %s, 0x%2X, imm); }\n", p.name, x, x, cond, s.c_str(), p.code);
 		}
 	}
 }
@@ -511,9 +511,9 @@ void putBroadcast(bool only64bit)
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
+			std::string s = type2String(p.type);
 			if ((only64bit && p.reg == 64) || (!only64bit && p.reg != 64)) {
-				printf("void %s(const Xmm& x, const Reg%d& r) { opVex(x, 0, r, %s, 0x%02X); }\n", p.name, p.reg, type.c_str(), p.code);
+				printf("void %s(const Xmm& x, const Reg%d& r) { opVex(x, 0, r, %s, 0x%02X); }\n", p.name, p.reg, s.c_str(), p.code);
 			}
 		}
 	}
@@ -583,28 +583,28 @@ void putCvt()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type);
+		std::string s = type2String(p.type);
 		switch (p.ptn) {
 		case 0:
-			printf("void %s(const Reg32e& r, const Operand& op) { uint64_t type = (%s) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			printf("void %s(const Reg32e& r, const Operand& op) { uint64_t type = (%s) | (r.isREG(64) ? T_EW1 : T_EW0); opVex(r, &xm0, op, type, 0x%02X); }\n", p.name, s.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);
+			printf("void %s(const Xmm& x, const Operand& op) { checkCvt1(x, op); opVex(x, 0, op, %s, 0x%02X); }\n", p.name, s.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);
+			printf("void %s(const Xmm& x, const Operand& op) { opCvt2(x, op, %s, 0x%02X); }\n", p.name, s.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);
+			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, s.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);
+			printf("void %s(const Xmm& x, const Operand& op) { checkCvt4(x, op); opCvt(x, op, %s, 0x%02X); }\n", p.name, s.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);
+			printf("void %s(const Xmm& x, const Operand& op) { opCvt5(x, op, %s, 0x%02X); }\n", p.name, s.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) uint64_t 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);
+			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) uint64_t type = (%s) | (op.isBit(32) ? (T_EW0 | T_N4) : (T_EW1 | T_N8)); opVex(x1, &x2, op, type, 0x%02X); }\n", p.name, s.c_str(), p.code);
 			break;
 		}
 	}
@@ -636,8 +636,8 @@ void putGather()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type | T_VSIB);
-		printf("void %s(const Xmm& x, const Address& addr) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, type.c_str(), p.code, p.mode);
+		std::string s = type2String(p.type | T_VSIB);
+		printf("void %s(const Xmm& x, const Address& addr) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, s.c_str(), p.code, p.mode);
 	}
 }
 void putScatter()
@@ -660,8 +660,8 @@ void putScatter()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
-		std::string type = type2String(p.type | T_VSIB);
-		printf("void %s(const Address& addr, const Xmm& x) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, type.c_str(), p.code, p.mode);
+		std::string s = type2String(p.type | T_VSIB);
+		printf("void %s(const Address& addr, const Xmm& x) { opGather2(x, addr, %s, 0x%02X, %d); }\n", p.name, s.c_str(), p.code, p.mode);
 	}
 }
 
@@ -718,8 +718,8 @@ void putMov()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
-			printf("void %s(const Operand& op, const Xmm& x) { opVmov(op, x, %s, 0x%02X, %s); }\n", p.name, type.c_str(), p.code, p.mode ? "true" : "false");
+			std::string s = type2String(p.type);
+			printf("void %s(const Operand& op, const Xmm& x) { opVmov(op, x, %s, 0x%02X, %s); }\n", p.name, s.c_str(), p.code, p.mode ? "true" : "false");
 		}
 	}
 }
@@ -770,9 +770,9 @@ void putX_XM_IMM()
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl *p = &tbl[i];
-		std::string type = type2String(p->type);
+		std::string s = type2String(p->type);
 		printf("void %s(const Xmm& x, const Operand& op%s) { opAVX_X_XM_IMM(x, op, %s, 0x%02X%s); }\n"
-			, p->name, p->hasIMM ? ", uint8_t imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
+			, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? ", imm" : "");
 	}
 }
 
@@ -810,9 +810,9 @@ void putMisc()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type | T_66 | T_0F38 | T_MUST_EVEX | T_M_K | T_VSIB);
+			std::string s = type2String(p.type | T_66 | T_0F38 | T_MUST_EVEX | T_M_K | T_VSIB);
 			printf("void %s(const Address& addr) { opGatherFetch(addr, zm%d, %s, 0x%2X, Operand::%s); }\n"
-				, p.name, p.zm, type.c_str(), p.code, p.isZmm ? "ZMM" : "YMM");
+				, p.name, p.zm, s.c_str(), p.code, p.isZmm ? "ZMM" : "YMM");
 		}
 	}
 
@@ -887,18 +887,18 @@ void putFP16_FMA()
 				{ "213", 0xA0 },
 				{ "231", 0xB0 },
 			};
-			int t = T_66 | T_MAP6 | T_EW0 | T_MUST_EVEX;
+			uint64_t type = 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;
+				type |= T_ER_Z | T_YMM | T_B16;
 				suf = "ph";
 			} else {
-				t |= T_ER_X | T_N2;
+				type |= T_ER_X | T_N2;
 				suf = "sh";
 			}
-			std::string type = type2String(t);
+			std::string s = type2String(type);
 			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);
+				, tbl[i].name, ord[k].str, suf, s.c_str(), tbl[i].code | ord[k].code);
 		}
 	}
 }
@@ -928,9 +928,9 @@ void putFP16_FMA2()
 				t |= T_ER_X | T_N2;
 				suf = "sh";
 			}
-			std::string type = type2String(t);
+			std::string s = 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);
+				, j == 0 ? "c" : "", tbl[i].name, suf, s.c_str(), tbl[i].code);
 		}
 	}
 }
@@ -938,16 +938,16 @@ void putFP16_FMA2()
 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());
+		uint64_t type = T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2;
+		std::string s = type2String(type);
+		printf("void vmovsh(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, %s, 0x10); }\n", s.c_str());
+		printf("void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2, x3, %s, 0x10); }\n", s.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());
+		uint64_t type = T_66 | T_MAP5 | T_MUST_EVEX | T_N2;
+		std::string s = type2String(type);
+		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", s.c_str());
+		printf("void vmovw(const Reg32e& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, r, %s, 0x7E); }\n", s.c_str());
 	}
 }
 
diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
index 272184f..fd60773 100644
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@@ -216,7 +216,7 @@ void putX_X_XM(bool omitOnly)
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
+			std::string s = type2String(p->type);
 			if (omitOnly) {
 				if (p->enableOmit) {
 					printf("void v%s(const Xmm& x, const Operand& op%s) { v%s(x, x, op%s); }\n", p->name, p->hasIMM ? ", uint8_t imm" : "", p->name, p->hasIMM ? ", imm" : "");
@@ -231,7 +231,7 @@ void putX_X_XM(bool omitOnly)
 				}
 				if (p->mode & 2) {
 					printf("void v%s(const Xmm& x1, const Xmm& x2, const Operand& op%s) { opAVX_X_X_XM(x1, x2, op, %s, 0x%02X%s); }\n"
-					, p->name, p->hasIMM ? ", uint8_t imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
+					, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? ", imm" : "");
 				}
 			}
 		}
@@ -483,8 +483,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
-			printf("void %s(const Xmm& reg1, const Xmm& reg2) { opRR(reg1, reg2, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
+			std::string s = type2String(p->type);
+			printf("void %s(const Xmm& reg1, const Xmm& reg2) { opRR(reg1, reg2, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 		}
 	}
 	{
@@ -515,11 +515,10 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type | T_0F);
-			printf("void %s(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM); }\n", p->name, type.c_str(), p->code);
+			std::string s = type2String(p->type | T_0F);
+			printf("void %s(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM); }\n", p->name, s.c_str(), p->code);
 		}
 	}
-
 	{
 		// special type
 		const struct Tbl {
@@ -543,8 +542,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type | T_0F);
-			printf("void %s(const Reg& reg, const Operand& op) { opSSE(reg, op, %s, 0x%02X, %s); }\n", p->name, type.c_str(), p->code, p->cond);
+			std::string s = type2String(p->type | T_0F);
+			printf("void %s(const Reg& reg, const Operand& op) { opSSE(reg, op, %s, 0x%02X, %s); }\n", p->name, s.c_str(), p->code, p->cond);
 		}
 	}
 	{
@@ -581,8 +580,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
-			printf("void %s(const Operand& op1, const Operand& op2) { opMovXMM(op1, op2, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
+			std::string s = type2String(p->type);
+			printf("void %s(const Operand& op1, const Operand& op2) { opMovXMM(op1, op2, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
 		}
 	}
 	{
@@ -812,9 +811,9 @@ void put()
 			printf("void %s(const Operand& op1, const Operand& op2) { opRO_MR(op1, op2, 0x%02X); }\n", p->name, p->code);
 			printf("void %s(const Operand& op, uint32_t imm) { opOI(op, imm, 0x%02X, %d); }\n", p->name, p->code, p->ext);
 			if (!p->support3op) continue;
-			std::string type = type2String(0);//p->type);
-			printf("void %s(const Reg& d, const Operand& op1, const Operand& op2) { opROO(d, op1, op2, %s, 0x%02X); }\n", p->name, type.c_str(), p->code);
-			printf("void %s(const Reg& d, const Operand& op, uint32_t imm) { opROI(d, op, imm, %s, %d); }\n", p->name, type.c_str(), p->ext);
+			std::string s = type2String(0);//p->type);
+			printf("void %s(const Reg& d, const Operand& op1, const Operand& op2) { opROO(d, op1, op2, %s, 0x%02X); }\n", p->name, s.c_str(), p->code);
+			printf("void %s(const Reg& d, const Operand& op, uint32_t imm) { opROI(d, op, imm, %s, %d); }\n", p->name, s.c_str(), p->ext);
 		}
 	}
 	{
@@ -1354,15 +1353,15 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
+			std::string s = type2String(p->type);
 			if (p->mode & 1) {
 				const char *immS1 = p->hasIMM ? ", uint8_t imm" : "";
 				const char *immS2 = p->hasIMM ? ", imm" : ", NONE";
-				printf("void %s(const Xmm& xmm, const Operand& op%s) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM%s); }\n", p->name, immS1, type.c_str(), p->code, immS2);
+				printf("void %s(const Xmm& xmm, const Operand& op%s) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM%s); }\n", p->name, immS1, s.c_str(), p->code, immS2);
 			}
 			if (p->mode & 2) {
 				printf("void v%s(const Xmm& xm, const Operand& op%s) { opAVX_X_XM_IMM(xm, op, %s, 0x%02X%s); }\n"
-					, p->name, p->hasIMM ? ", uint8_t imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
+					, p->name, p->hasIMM ? ", uint8_t imm" : "", s.c_str(), p->code, p->hasIMM ? ", imm" : "");
 			}
 		}
 	}
@@ -1382,9 +1381,9 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
+			std::string s = type2String(p->type);
 			printf("void v%s(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, %s, 0x%02X); }\n"
-				, p->name, type.c_str(), p->code);
+				, p->name, s.c_str(), p->code);
 		}
 	}
 	// (x, x/m), (y, y/m), (x, x, x/m), (y, y, y/m)
@@ -1409,13 +1408,13 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
+			std::string s = type2String(p->type);
 			if (p->mode & 1) {
-				printf("void %s(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM); }\n", p->name, type.c_str(), p->code);
+				printf("void %s(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, %s, 0x%02X, isXMM_XMMorMEM); }\n", p->name, s.c_str(), p->code);
 			}
 			if (p->mode & 2) {
 				printf("void v%s(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, %s, 0x%02X); }\n"
-					, p->name, type.c_str(), p->code);
+					, p->name, s.c_str(), p->code);
 			}
 		}
 	}
@@ -1449,8 +1448,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
-			printf("void %s(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			std::string s = type2String(p.type);
+			printf("void %s(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, %s, 0x%02X); }\n", p.name, s.c_str(), p.code);
 		}
 	}
 	// vpermq, vpermpd
@@ -1465,8 +1464,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
-			printf("void %s(const Ymm& y, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(y, op, %s, 0x%02X, imm); }\n", p.name, type.c_str(), p.code);
+			std::string s = type2String(p.type);
+			printf("void %s(const Ymm& y, const Operand& op, uint8_t imm) { opAVX_X_XM_IMM(y, op, %s, 0x%02X, imm); }\n", p.name, s.c_str(), p.code);
 		}
 	}
 	// vcmpeqps
@@ -1504,11 +1503,11 @@ void put()
 			const Tbl& p = tbl[i];
 			char c = p.isH ? 'h' : 'l';
 			const char *suf = p.isPd ? "pd" : "ps";
-			const char *type = p.isPd ? "T_0F | T_66 | T_EVEX | T_EW1 | T_N8" : "T_0F | T_EVEX | T_EW0 | T_N8";
+			std::string s = type2String(p.isPd ? (T_0F | T_66 | T_EVEX | T_EW1 | T_N8) : (T_0F | T_EVEX | T_EW0 | T_N8));
 			printf("void vmov%c%s(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, %s, 0x%02X); }\n"
-				, c, suf, type, p.code);
+				, c, suf, s.c_str(), p.code);
 			printf("void vmov%c%s(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, %s, 0x%02X); }\n"
-				, c, suf, type, p.code + 1);
+				, c, suf, s.c_str(), p.code + 1);
 		}
 	}
 	// FMA
@@ -1557,9 +1556,9 @@ void put()
 					} else { // ss
 						t |= T_ER_X | T_N4;
 					}
-					std::string type = type2String(t);
+					std::string s = 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.c_str(), type.c_str(), tbl[i].code + ord[k].code);
+						, tbl[i].name, ord[k].str, suf.c_str(), s.c_str(), tbl[i].code + ord[k].code);
 				}
 			}
 		}
@@ -1583,8 +1582,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
-			printf("void %s(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p.name, type.c_str(), p.code);
+			std::string s = type2String(p.type);
+			printf("void %s(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, %s, 0x%02X); }\n", p.name, s.c_str(), p.code);
 		}
 
 		puts("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); }");
@@ -1634,8 +1633,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl& p = tbl[i];
-			std::string type = type2String(p.type);
-			printf("void v%s(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_X_XM(Xmm(x.getKind(), %d), x, op, %s, 0x%02X, imm); }\n", p.name, p.idx, type.c_str(), p.code);
+			std::string s = type2String(p.type);
+			printf("void v%s(const Xmm& x, const Operand& op, uint8_t imm) { opAVX_X_X_XM(Xmm(x.getKind(), %d), x, op, %s, 0x%02X, imm); }\n", p.name, p.idx, s.c_str(), p.code);
 		}
 	}
 	// 4-op
@@ -1724,7 +1723,7 @@ 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);
 		}
-		puts("void vcvtneps2bf16(const Xmm& x, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opCvt2(x, op, T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32 | orEvexIf(encoding), 0x72); }");
+		printf("void vcvtneps2bf16(const Xmm& x, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opCvt2(x, op, %s|orEvexIf(encoding), 0x72); }", type2String(T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32).c_str());
 	}
 	// haswell gpr(reg, reg, r/m)
 	{
@@ -1820,8 +1819,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = type2String(p->type);
-			printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, %s, 0x%02X, encoding); }\n", p->name, type.c_str(), p->code);
+			std::string s = type2String(p->type);
+			printf("void %s(const Xmm& x1, const Xmm& x2, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opEncoding(x1, x2, op, %s, 0x%02X, encoding); }\n", p->name, s.c_str(), p->code);
 		}
 	}
 	// avx-vnni-int8
@@ -1848,8 +1847,8 @@ void put()
 		};
 		for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 			const Tbl *p = &tbl[i];
-			std::string type = 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, type.c_str(), p->code);
+			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);
 		}
 	}
 }
diff --git a/xbyak/xbyak.h b/xbyak/xbyak.h
index 1560cd1..e0e9f49 100644
--- a/xbyak/xbyak.h
+++ b/xbyak/xbyak.h
@@ -1756,53 +1756,51 @@ private:
 		if (rex) db(rex);
 	}
 	// @@@begin of avx_type_def.h
-	enum AVXtype {
-		T_NONE = 0,
+	static const uint64_t T_NONE = 0ull;
 		// low 3 bit
-		T_N1 = 1,
-		T_N2 = 2,
-		T_N4 = 3,
-		T_N8 = 4,
-		T_N16 = 5,
-		T_N32 = 6,
-		T_NX_MASK = 7,
-		T_DUP = T_NX_MASK,//1 << 4, // N = (8, 32, 64)
-		T_N_VL = 1 << 3, // N * (1, 2, 4) for VL
-		T_VEX = 1 << 4,
-		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,
-		T_L0 = 1 << 11,
-		T_L1 = 1 << 12,
-		T_W0 = 1 << 13,
-		T_W1 = 1 << 14,
-		T_EW0 = 1 << 15,
-		T_EW1 = 1 << 16,
-		T_YMM = 1 << 17, // support YMM, ZMM
-		T_EVEX = 1 << 18,
-		T_ER_X = 1 << 19, // xmm{er}
-		T_ER_Y = 1 << 20, // ymm{er}
-		T_ER_Z = 1 << 21, // zmm{er}
-		T_SAE_X = 1 << 22, // xmm{sae}
-		T_SAE_Y = 1 << 23, // ymm{sae}
-		T_SAE_Z = 1 << 24, // zmm{sae}
-		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 = 1u << 30, // use evex if mem
-		T_FP16 = 1u << 31, // avx512-fp16
-		T_MAP5 = T_FP16 | T_0F,
-		T_MAP6 = T_FP16 | T_0F38,
-		T_NF = 1ull << 32, // T_nf
-		T_XXX
-	};
+	static const uint64_t T_N1 = 1ull;
+	static const uint64_t T_N2 = 2ull;
+	static const uint64_t T_N4 = 3ull;
+	static const uint64_t T_N8 = 4ull;
+	static const uint64_t T_N16 = 5ull;
+	static const uint64_t T_N32 = 6ull;
+	static const uint64_t T_NX_MASK = 7ull;
+	static const uint64_t T_DUP = T_NX_MASK;//1 << 4, // N = (8, 32, 64)
+	static const uint64_t T_N_VL = 1ull << 3; // N * (1, 2, 4) for VL
+	static const uint64_t T_VEX = 1ull << 4;
+	static const uint64_t T_66 = 1ull << 5; // pp = 1
+	static const uint64_t T_F3 = 1ull << 6; // pp = 2
+	static const uint64_t T_F2 = T_66 | T_F3; // pp = 3
+	static const uint64_t T_ER_R = 1ull << 7; // reg{er}
+	static const uint64_t T_0F = 1ull << 8;
+	static const uint64_t T_0F38 = 1ull << 9;
+	static const uint64_t T_0F3A = 1ull << 10;
+	static const uint64_t T_L0 = 1ull << 11;
+	static const uint64_t T_L1 = 1ull << 12;
+	static const uint64_t T_W0 = 1ull << 13;
+	static const uint64_t T_W1 = 1ull << 14;
+	static const uint64_t T_EW0 = 1ull << 15;
+	static const uint64_t T_EW1 = 1ull << 16;
+	static const uint64_t T_YMM = 1ull << 17; // support YMM, ZMM
+	static const uint64_t T_EVEX = 1ull << 18;
+	static const uint64_t T_ER_X = 1ull << 19; // xmm{er}
+	static const uint64_t T_ER_Y = 1ull << 20; // ymm{er}
+	static const uint64_t T_ER_Z = 1ull << 21; // zmm{er}
+	static const uint64_t T_SAE_X = 1ull << 22; // xmm{sae}
+	static const uint64_t T_SAE_Y = 1ull << 23; // ymm{sae}
+	static const uint64_t T_SAE_Z = 1ull << 24; // zmm{sae}
+	static const uint64_t T_MUST_EVEX = 1ull << 25; // contains T_EVEX
+	static const uint64_t T_B32 = 1ull << 26; // m32bcst
+	static const uint64_t T_B64 = 1ull << 27; // m64bcst
+	static const uint64_t T_B16 = T_B32 | T_B64; // m16bcst (Be careful)
+	static const uint64_t T_M_K = 1ull << 28; // mem{k}
+	static const uint64_t T_VSIB = 1ull << 29;
+	static const uint64_t T_MEM_EVEX = 1ull << 30; // use evex if mem
+	static const uint64_t T_FP16 = 1ull << 31; // avx512-fp16
+	static const uint64_t T_MAP5 = T_FP16 | T_0F;
+	static const uint64_t T_MAP6 = T_FP16 | T_0F38;
+	static const uint64_t T_NF = 1ull << 32; // T_nf
+	static const uint64_t T_NO_OR1 = 1ull << 33; // does not "code | 1"
 	// T_66 = 1, T_F3 = 2, T_F2 = 3
 	static inline uint32_t getPP(uint64_t type) { return (type >> 5) & 3; }
 	// @@@end of avx_type_def.h
@@ -1995,6 +1993,7 @@ private:
 		} else if (type & T_0F3A) {
 			db(0x0F); db(0x3A);
 		}
+//		db(code | (!(type & T_NO_OR1) && !r.isBit(8)));
 		db(code | (type == 0 && !r.isBit(8)));
 	}
 	void opRR(const Reg& reg1, const Reg& reg2, uint64_t type, int code)
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index 44afcff..62e49db 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -29,7 +29,6 @@ void and_(const Operand& op, uint32_t imm) { opOI(op, imm, 0x20, 4); }
 void and_(const Operand& op1, const Operand& op2) { opRO_MR(op1, op2, 0x20); }
 void and_(const Reg& d, const Operand& op, uint32_t imm) { opROI(d, op, imm, T_NONE, 4); }
 void and_(const Reg& d, const Operand& op1, const Operand& op2) { opROO(d, op1, op2, T_NONE, 0x20); }
-void andn(const Reg32e& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_0F38, 0xf2); }
 void andnpd(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, T_0F | T_66, 0x55, isXMM_XMMorMEM); }
 void andnps(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, T_0F, 0x55, isXMM_XMMorMEM); }
 void andpd(const Xmm& xmm, const Operand& op) { opSSE(xmm, op, T_0F | T_66, 0x54, isXMM_XMMorMEM); }
@@ -1015,7 +1014,7 @@ void vcvtneebf162ps(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_F3|
 void vcvtneeph2ps(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_66|T_0F38|T_W0|T_YMM, 0xB0); }
 void vcvtneobf162ps(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_F2|T_0F38|T_W0|T_YMM, 0xB0); }
 void vcvtneoph2ps(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_0F38|T_W0|T_YMM, 0xB0); }
-void vcvtneps2bf16(const Xmm& x, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opCvt2(x, op, T_F3 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_B32 | orEvexIf(encoding), 0x72); }
+void vcvtneps2bf16(const Xmm& x, const Operand& op, PreferredEncoding encoding = DefaultEncoding) { opCvt2(x, op, T_F3|T_0F38|T_EW0|T_YMM|T_SAE_Z|T_B32|orEvexIf(encoding), 0x72); }void andn(const Reg32e& r1, const Reg32e& r2, const Operand& op) { opRRO(r1, r2, op, T_0F38, 0xf2); }
 void vcvtpd2dq(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_F2 | T_YMM | T_EVEX | T_EW1 | T_B64 | T_ER_Z, 0xE6); }
 void vcvtpd2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_0F | T_66 | T_YMM | T_EVEX | T_EW1 | T_B64 | T_ER_Z, 0x5A); }
 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); }
@@ -1142,15 +1141,15 @@ void vmovdqa(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66|T_0
 void vmovdqu(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_F3|T_0F|T_YMM, 0x7F); }
 void vmovdqu(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3|T_0F|T_YMM, 0x6F); }
 void vmovhlps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x12); }
-void vmovhpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x17); }
-void vmovhpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x16); }
-void vmovhps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x17); }
-void vmovhps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x16); }
+void vmovhpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8|T_66|T_0F|T_EW1|T_EVEX, 0x17); }
+void vmovhpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_N8|T_66|T_0F|T_EW1|T_EVEX, 0x16); }
+void vmovhps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8|T_0F|T_EW0|T_EVEX, 0x17); }
+void vmovhps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_N8|T_0F|T_EW0|T_EVEX, 0x16); }
 void vmovlhps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x16); }
-void vmovlpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x13); }
-void vmovlpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x12); }
-void vmovlps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x13); }
-void vmovlps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x12); }
+void vmovlpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8|T_66|T_0F|T_EW1|T_EVEX, 0x13); }
+void vmovlpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_N8|T_66|T_0F|T_EW1|T_EVEX, 0x12); }
+void vmovlps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8|T_0F|T_EW0|T_EVEX, 0x13); }
+void vmovlps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_N8|T_0F|T_EW0|T_EVEX, 0x12); }
 void vmovmskpd(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_66 | T_W0 | T_YMM, 0x50); }
 void vmovmskps(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_W0 | T_YMM, 0x50); }
 void vmovntdq(const Address& addr, const Xmm& x) { opVex(x, 0, addr, T_0F | T_66 | T_YMM | T_EVEX | T_EW0, 0xE7); }
@@ -2059,44 +2058,44 @@ 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|T_NF, 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|T_NF, 0xD6); }
+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|T_NF|T_NO_OR1, 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|T_NF|T_NO_OR1, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 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|T_NF, 0xBF); }
+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|T_NF|T_NO_OR1, 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|T_NF|T_NO_OR1, 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); }
@@ -2152,11 +2151,11 @@ void vmovdqu64(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3|T_0
 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|T_NF, 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|T_NF, 0x10); }
+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|T_NF, 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|T_NF, 0x6E); }
+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); }