case AORN:
		return OPVCC(31, 412, 0, 0)
	case AORNCC:
		return OPVCC(31, 412, 0, 1)

	case APOPCNTD:
		return OPVCC(31, 506, 0, 0) /* popcntd - v2.06 */
	case APOPCNTW:
		return OPVCC(31, 378, 0, 0) /* popcntw - v2.06 */
	case APOPCNTB:
		return OPVCC(31, 122, 0, 0) /* popcntb - v2.02 */
	case ACNTTZW:
		return OPVCC(31, 538, 0, 0) /* cnttzw - v3.00 */
	case ACNTTZWCC:

		// POPCNTx counts the number of set bits in the low-order (L=32,Q=64) bits of arg0.
		// POPCNTx instructions are only guaranteed to be available if GOAMD64>=v2.
		// For GOAMD64<v2, any use must be preceded by a successful runtime check of runtime.x86HasPOPCNT.
		{name: "POPCNTQ", argLength: 1, reg: gp11, asm: "POPCNTQ", clobberFlags: true},

    auto output_type = op.getType().clone(
        rewriter.getIntegerType(/*width=*/8, /*isSigned=*/false));
    auto pop_cnt = rewriter.create<TF::PopulationCountOp>(
        op.getLoc(), output_type, op.getOperand());
    auto cast_or_pop_cnt =
        rewriter.createOrFold<TF::CastOp>(op.getLoc(), op.getType(), pop_cnt);
    rewriter.replaceOp(op, {cast_or_pop_cnt});
    return success();
  }
};

	{APMULLW, ymm, Py1, opBytes{0xd5, Pe, 0xd5}},
	{APMULULQ, ymm, Py1, opBytes{0xf4, Pe, 0xf4}},
	{APOPAL, ynone, P32, opBytes{0x61}},
	{APOPAW, ynone, Pe, opBytes{0x61}},
	{APOPCNTW, yml_rl, Pef3, opBytes{0xb8}},
	{APOPCNTL, yml_rl, Pf3, opBytes{0xb8}},
	{APOPCNTQ, yml_rl, Pfw, opBytes{0xb8}},
	{APOPFL, ynone, P32, opBytes{0x9d}},
	{APOPFQ, ynone, Py, opBytes{0x9d}},
	{APOPFW, ynone, Pe, opBytes{0x9d}},

	// load on condition
	{i: 25, as: ALOCGR, a1: C_SCON, a2: C_REG, a6: C_REG},

	// find leftmost one
	{i: 8, as: AFLOGR, a1: C_REG, a6: C_REG},

	// population count
	{i: 9, as: APOPCNT, a1: C_REG, a6: C_REG},

	// compare
	{i: 70, as: ACMP, a1: C_REG, a6: C_REG},
	{i: 71, as: ACMP, a1: C_REG, a6: C_LCON},
	{i: 70, as: ACMPU, a1: C_REG, a6: C_REG},
	{i: 71, as: ACMPU, a1: C_REG, a6: C_LCON},