CINVW	LS, R7, R13	// ed80875a
	CNEG	LS, R13, R7	// a7858dda
	CNEGW	EQ, R8, R13	// 0d15885a

// atomic ops
	LDARB	(R25), R2                            // 22ffdf08
	LDARH	(R5), R7                             // a7fcdf48
	LDAXPW	(R10), (R20, R16)                    // 54c17f88
	LDAXP	(R25), (R30, R11)                    // 3eaf7fc8
	LDAXRW	(R0), R2                             // 02fc5f88

		// arg1 = address of src memory (in R16 aka arm64.REGRT1, changed as side effect)
		// arg2 = address of the last element of src
		// arg3 = mem
		// returns mem
		//	LDP.P	16(R16), (R25, Rtmp)
		//	STP.P	(R25, Rtmp), 16(R17)
		//	CMP	Rarg2, R16
		//	BLE	-3(PC)
		// Note: the-end-of-src may be not a valid pointer. it's a problem if it is spilled.
		// the-end-of-src - 16 is within the area to copy, ok to spill.
		{

	MOVD $16, R16
	MOVD $32, R17
	MOVD $48, R18
	MOVD $64, R19
	MOVD $80, R20
	MOVD $96, R21
	MOVD $112, R22
	MOVD $128, R23
	MOVD $144, R24
	MOVD $160, R25
	MOVD $104, R26 // offset of sign+24(FP)

	LXVD2X (R16)(CPOOL), PH
	LXVD2X (R0)(CPOOL), PL

	LXVD2X (R17)(P2ptr), Y2L
	LXVD2X (R18)(P2ptr), Y2H
	XXPERMDI Y2H, Y2H, $2, Y2H