// Load 4 vectors of a and b
	// XOR the corresponding vectors
	// from a and b and store the result
loop64:
	LXVD2X	(R4)(R8), VS32
	LXVD2X	(R4)(R10), VS34
	LXVD2X	(R4)(R14), VS36
	LXVD2X	(R4)(R15), VS38
	LXVD2X	(R5)(R8), VS33
	LXVD2X	(R5)(R10), VS35
	LXVD2X	(R5)(R14), VS37
	LXVD2X	(R5)(R15), VS39
	XXLXOR	VS32, VS33, VS32
	XXLXOR	VS34, VS35, VS34
	XXLXOR	VS36, VS37, VS36
	XXLXOR	VS38, VS39, VS38

	MOVD	$128, R20 \
	MOVD	$144, R21 \
	LXVD2X	(R0+Rkeyp), V6 \
	ADD	$16, Rkeyp \
	BEQ	CR1, L_start10 \
	BEQ	CR2, L_start12 \
	LXVD2X	(R0+Rkeyp), V7 \
	LXVD2X	(R12+Rkeyp), V8 \
	ADD	$32, Rkeyp \
	L_start12: \
	LXVD2X	(R0+Rkeyp), V9 \
	LXVD2X	(R12+Rkeyp), V10 \
	ADD	$32, Rkeyp \
	L_start10: \
	LXVD2X	(R0+Rkeyp), V11 \
	LXVD2X	(R12+Rkeyp), V12 \
	LXVD2X	(R14+Rkeyp), V13 \
	LXVD2X	(R15+Rkeyp), V14 \

	VSPLTB  $7, V6, V4
	MTVSRD  R5, VS39        // ŝ
	VSPLTB  $7, V7, V2
	ADD     $-2, R4, R16
	PCALIGN $16
loopback:
	ADD     $-1, R8, R10
	SLD     $3, R10
	LXVD2X  (R6)(R10), VS32 // load x[i-1], x[i]
	SLD     $3, R8, R12
	LXVD2X  (R6)(R12), VS33 // load x[i], x[i+1]

	VSRD    V0, V4, V3      // x[i-1]>>s, x[i]>>s
	VSLD    V1, V2, V5      // x[i]<<ŝ, x[i+1]<<ŝ

	VSHASIGMAW $1, $15, V1, V2      // 10418e82
	VSHASIGMAD $2, V1, $15, V2      // 104196c2
	VSHASIGMAD $2, $15, V1, V2      // 104196c2

	LXVD2X (R3)(R4), VS1            // 7c241e98
	LXVD2X (R3)(R0), VS1            // 7c201e98
	LXVD2X (R3), VS1                // 7c201e98
	LXVDSX (R3)(R4), VS1            // 7c241a98
	LXVDSX (R3)(R0), VS1            // 7c201a98
	LXVDSX (R3), VS1                // 7c201a98

			unsafePoint:    true,
		},

		// R31 is temp register
		// Loop code:
		//	MOVD len/32,R31		set up loop ctr
		//	MOVD R31,CTR
		//	MOVD $16,R31		index register
		// loop:
		//	LXVD2X (R0)(R4),VS32
		//	LXVD2X (R31)(R4),VS33
		//	ADD  R4,$32          increment src
		//	STXVD2X VS32,(R0)(R3)
		//	STXVD2X VS33,(R31)(R3)
		//	ADD  R3,$32          increment dst
		//	BC 16,0,loop         branch ctr

		// number of moves are generated based on the
		// size.
		// When moving >= 64 bytes a loop is used
		//	MOVD len/32,REG_TMP
		//	MOVD REG_TMP,CTR
		//	MOVD $16,REG_TMP
		// top:
		//	LXVD2X (R0)(R21),VS32
		//	LXVD2X (R31)(R21),VS33
		//	ADD $32,R21
		//	STXVD2X VS32,(R0)(R20)
		//	STXVD2X VS33,(R31)(R20)
		//	ADD $32,R20
		//	BC 16,0,top
		// Bytes not moved by this loop are moved

		case AVSBOX: /* vsbox */
			opset(AVSBOX, r0)

		case AVSHASIGMA: /* vshasigmaw, vshasigmad */
			opset(AVSHASIGMAW, r0)
			opset(AVSHASIGMAD, r0)

		case ALXVD2X: /* lxvd2x, lxvdsx, lxvw4x, lxvh8x, lxvb16x */
			opset(ALXVDSX, r0)
			opset(ALXVW4X, r0)
			opset(ALXVH8X, r0)
			opset(ALXVB16X, r0)

		case ALXV: /* lxv */
			opset(ALXV, r0)