diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 8929b7a..51446ae 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -772,10 +772,9 @@ void putFP16_FMA()
 		bool isPH;
 	} tbl[] = {
 		{ 0x06, "vfmaddsub", true },
+		{ 0x07, "vfmsubadd", true },
 /*
 		{ 0x, "vfmadd", false },
-		{ 0x06, "vfmaddsub", true },
-		{ 0x07, "vfmsubadd", true },
 		{ 0x0A, "vfmsub", true },
 		{ 0x0B, "vfmsub", false },
 		{ 0x0C, "vfnmadd", true },
diff --git a/test/misc.cpp b/test/misc.cpp
index db21de1..27ecfef 100644
--- a/test/misc.cpp
+++ b/test/misc.cpp
@@ -904,6 +904,14 @@ CYBOZU_TEST_AUTO(vaddph)
 			vfmaddsub213ph(zmm1, zmm2, ptr [rax+0x40]);
 			vfmaddsub213ph(zmm1, zmm2, ptr_b [rax+0x40]);
 			vfmaddsub213ph(zmm1|T_ru_sae, zmm2, zmm5);
+
+			vfmsubadd132ph(xmm1, xmm2, ptr [rax+0x40]);
+			vfmsubadd132ph(xmm1, xmm2, ptr_b [rax+0x40]);
+			vfmsubadd132ph(ymm1, ymm2, ptr [rax+0x40]);
+			vfmsubadd132ph(ymm1, ymm2, ptr_b [rax+0x40]);
+			vfmsubadd132ph(zmm1, zmm2, ptr [rax+0x40]);
+			vfmsubadd132ph(zmm1, zmm2, ptr_b [rax+0x40]);
+			vfmsubadd132ph(zmm1|T_ru_sae, zmm2, zmm5);
 		}
 	} c;
 	const uint8_t tbl[] = {
@@ -946,6 +954,15 @@ CYBOZU_TEST_AUTO(vaddph)
 		0x62, 0xf6, 0x6d, 0x48, 0xa6, 0x48, 0x01,
 		0x62, 0xf6, 0x6d, 0x58, 0xa6, 0x48, 0x20,
 		0x62, 0xf6, 0x6d, 0x58, 0xa6, 0xcd,
+
+		// vfmsubadd132ph
+		0x62, 0xf6, 0x6d, 0x08, 0x97, 0x48, 0x04,
+		0x62, 0xf6, 0x6d, 0x18, 0x97, 0x48, 0x20,
+		0x62, 0xf6, 0x6d, 0x28, 0x97, 0x48, 0x02,
+		0x62, 0xf6, 0x6d, 0x38, 0x97, 0x48, 0x20,
+		0x62, 0xf6, 0x6d, 0x48, 0x97, 0x48, 0x01,
+		0x62, 0xf6, 0x6d, 0x58, 0x97, 0x48, 0x20,
+		0x62, 0xf6, 0x6d, 0x58, 0x97, 0xcd,
 	};
 	const size_t n = sizeof(tbl) / sizeof(tbl[0]);
 	CYBOZU_TEST_EQUAL(c.getSize(), n);
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index ceac1a1..7b45a56 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1945,6 +1945,9 @@ void vfixupimmss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8_t imm) {
 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 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 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 vfpclassps(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }
 void vfpclasssd(const Opmask& k, const Operand& op, uint8_t imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); }