diff --git a/gen/gen_code.cpp b/gen/gen_code.cpp
index 312b2177..44d1f361 100644
--- a/gen/gen_code.cpp
+++ b/gen/gen_code.cpp
@@ -1941,6 +1941,10 @@ void put64()
 	puts("void vmovq(const Xmm& x, const Reg64& r) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x6E); }");
 	puts("void vmovq(const Reg64& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x7E); }");
 	puts("void jmpabs(uint64_t addr) { db(0xD5); db(0x00); db(0xA1); dq(addr); }");
+	puts("void push2(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(6), T_VEX|T_ND1|T_W0, 0xFF); }");
+	puts("void push2p(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(6), T_VEX|T_ND1|T_W1, 0xFF); }");
+	puts("void pop2(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(0), T_VEX|T_ND1|T_W0, 0x8F); }");
+	puts("void pop2p(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(0), T_VEX|T_ND1|T_W1, 0x8F); }");
 	// CMPccXADD
 	{
 		const struct Tbl {
diff --git a/test/apx.cpp b/test/apx.cpp
index ba767e23..3e0b0afb 100644
--- a/test/apx.cpp
+++ b/test/apx.cpp
@@ -1397,3 +1397,38 @@ CYBOZU_TEST_AUTO(shift_3op)
 	CYBOZU_TEST_EQUAL(c.getSize(), n);
 	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
 }
+
+CYBOZU_TEST_AUTO(push2_pop2)
+{
+	struct Code : Xbyak::CodeGenerator {
+		Code()
+		{
+			push2(r20, r30);
+			push2(rax, rcx);
+			push2p(r20, r30);
+			push2p(rdx, r8);
+
+			pop2(rax, rcx);
+			pop2(r20, r30);
+			pop2p(rax, rcx);
+			pop2p(r20, r30);
+		}
+	} c;
+	const uint8_t tbl[] = {
+		// push2
+		0x62, 0xdc, 0x5c, 0x10, 0xff, 0xf6,
+		0x62, 0xf4, 0x7c, 0x18, 0xff, 0xf1,
+		// push2p (What is this?)
+		0x62, 0xdc, 0xdc, 0x10, 0xff, 0xf6,
+		0x62, 0xd4, 0xec, 0x18, 0xff, 0xf0,
+		// pop2
+		0x62, 0xf4, 0x7c, 0x18, 0x8f, 0xc1,
+		0x62, 0xdc, 0x5c, 0x10, 0x8f, 0xc6,
+		// pop2p
+		0x62, 0xf4, 0xfc, 0x18, 0x8f, 0xc1,
+		0x62, 0xdc, 0xdc, 0x10, 0x8f, 0xc6,
+	};
+	const size_t n = sizeof(tbl);
+	CYBOZU_TEST_EQUAL(c.getSize(), n);
+	CYBOZU_TEST_EQUAL_ARRAY(c.getCode(), tbl, n);
+}
diff --git a/xbyak/xbyak.h b/xbyak/xbyak.h
index d4c5b14c..e092b05b 100644
--- a/xbyak/xbyak.h
+++ b/xbyak/xbyak.h
@@ -1927,7 +1927,7 @@ class CodeGenerator : public CodeArray {
 		int B3 = b.isExtIdx() ? 0 : 0x20;
 		int R4 = r.isExtIdx2() ? 0 : 0x10;
 		int B4 = b.isExtIdx2() ? 0x08 : 0;
-		int w = r.isBit(64) || v.isBit(64) || (type & T_W1);
+		int w = (type & T_W0) ? 0 : (r.isBit(64) || v.isBit(64) || (type & T_W1));
 		int V = (~v.getIdx() & 15) << 3;
 		int X4 = x.isExtIdx2() ? 0 : 0x04;
 		int pp = (type & (T_F2|T_F3|T_66)) ? getPP(type) : (r.isBit(16) || v.isBit(16));
diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h
index 54623d5d..b5a29332 100644
--- a/xbyak/xbyak_mnemonic.h
+++ b/xbyak/xbyak_mnemonic.h
@@ -1732,6 +1732,10 @@ void vcvttsd2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()
 void vmovq(const Xmm& x, const Reg64& r) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x6E); }
 void vmovq(const Reg64& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x7E); }
 void jmpabs(uint64_t addr) { db(0xD5); db(0x00); db(0xA1); dq(addr); }
+void push2(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(6), T_VEX|T_ND1|T_W0, 0xFF); }
+void push2p(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(6), T_VEX|T_ND1|T_W1, 0xFF); }
+void pop2(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(0), T_VEX|T_ND1|T_W0, 0x8F); }
+void pop2p(const Reg64& r1, const Reg64& r2) { opROO(r1, r2, Reg64(0), T_VEX|T_ND1|T_W1, 0x8F); }
 void cmpbexadd(const Address& addr, const Reg32e& r1, const Reg32e& r2) { opRRO(r1, r2, addr, T_66 | T_0F38, 0xE6); }
 void cmpbxadd(const Address& addr, const Reg32e& r1, const Reg32e& r2) { opRRO(r1, r2, addr, T_66 | T_0F38, 0xE2); }
 void cmplexadd(const Address& addr, const Reg32e& r1, const Reg32e& r2) { opRRO(r1, r2, addr, T_66 | T_0F38, 0xEE); }