add vcvtusi2sh
diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp
index 99a5f8d..35960bb 100644
--- a/gen/gen_avx512.cpp
+++ b/gen/gen_avx512.cpp
@@ -582,6 +582,7 @@
 		{ 0x7A, "vcvtuqq2ph", T_F2 | T_MAP5 | T_MUST_EVEX | T_EW1 | T_B64 | T_ER_Z | T_N16 | T_N_VL, 5 },
 
 		{ 0x2A, "vcvtsi2sh", T_F3 | T_MAP5 | T_MUST_EVEX | T_ER_R | T_M_K, 6 },
+		{ 0x7B, "vcvtusi2sh", T_F3 | T_MAP5 | T_MUST_EVEX | T_ER_R | T_M_K, 6 },
 	};
 	for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
 		const Tbl& p = tbl[i];
diff --git a/test/misc.cpp b/test/misc.cpp
index 1f920a8..3b10c99 100644
--- a/test/misc.cpp
+++ b/test/misc.cpp
@@ -1321,6 +1321,11 @@
 			vcvtsi2sh(xmm1, xmm2, dword [rax+0x40]);
 			vcvtsi2sh(xmm1|T_rd_sae, xmm2, r9);
 			vcvtsi2sh(xmm1, xmm2, qword [rax+0x40]);
+
+			vcvtusi2sh(xmm1|T_rd_sae, xmm2, eax);
+			vcvtusi2sh(xmm1, xmm2, dword [rax+0x40]);
+			vcvtusi2sh(xmm1|T_rd_sae, xmm2, r9);
+			vcvtusi2sh(xmm1, xmm2, qword [rax+0x40]);
 		}
 	} c;
 	const uint8_t tbl[] = {
@@ -1846,6 +1851,12 @@
 		0x62, 0xf5, 0x6e, 0x08, 0x2a, 0x48, 0x10,
 		0x62, 0xd5, 0xee, 0x38, 0x2a, 0xc9,
 		0x62, 0xf5, 0xee, 0x08, 0x2a, 0x48, 0x08,
+
+		// vcvtusi2sh
+		0x62, 0xf5, 0x6e, 0x38, 0x7b, 0xc8,
+		0x62, 0xf5, 0x6e, 0x08, 0x7b, 0x48, 0x10,
+		0x62, 0xd5, 0xee, 0x38, 0x7b, 0xc9,
+		0x62, 0xf5, 0xee, 0x08, 0x7b, 0x48, 0x08,
 	};
 	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 8c67a32..f76c2e7 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1950,6 +1950,7 @@
 void vcvtuqq2ph(const Xmm& x, const Operand& op) { opCvt5(x, op, T_N16 | T_N_VL | T_F2 | T_MAP5 | T_EW1 | T_ER_Z | T_MUST_EVEX | T_B64, 0x7A); }
 void vcvtuqq2ps(const Xmm& x, const Operand& op) { opCvt2(x, op, T_F2 | T_0F | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x7A); }
 void vcvtusi2sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F2 | T_0F | T_MUST_EVEX, T_W1 | T_EW1 | T_ER_X | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }
+void vcvtusi2sh(const Xmm& x1, const Xmm& x2, const Operand& op) { if (!(x1.isXMM() && x2.isXMM() && op.isBit(32|64))) XBYAK_THROW(ERR_BAD_COMBINATION) int type = (T_F3 | T_MAP5 | T_ER_R | T_MUST_EVEX | T_M_K) | (op.isBit(32) ? (T_EW0 | T_N4) : (T_EW1 | T_N8)); opVex(x1, &x2, op, type, 0x7B); }
 void vcvtusi2ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opCvt3(x1, x2, op, T_F3 | T_0F | T_MUST_EVEX | T_ER_X, T_W1 | T_EW1 | T_N8, T_W0 | T_EW0 | T_N4, 0x7B); }
 void vcvtuw2ph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }
 void vcvtw2ph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_MAP5 | T_EW0 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B16, 0x7D); }