MOVD $16, R20
	// V16
	LXVW4X (CONSTBASE)(R0), VS48
	ADD $80,CONSTBASE

	// Load key into V17,V18
	LXVW4X (KEY)(R0), VS49
	LXVW4X (KEY)(R8), VS50

	// Load CNT, NONCE into V19
	LXVW4X (CNT)(R0), VS51

	// Clear V27
	VXOR V27, V27, V27

	// V28
	LXVW4X (CONSTBASE)(R11), VS60

	// Load mask constants for VPERMXOR
	LXVW4X (MASK)(R0), V20
	LXVW4X (MASK)(R20), V21

	ANDCC	$0x3F, R5, R5	// len%64==0?

	PCALIGN $16
loop64:
	LXVD2X	(R8+R0), V0
	LXVD2X	(R4+R0), V1
	VCMPEQUBCC V0, V1, V2	// compare, setting CR6
	BGELR_CR6
	LXVD2X	(R8+R14), V0
	LXVD2X	(R4+R14), V1
	VCMPEQUBCC	V0, V1, V2
	BGELR_CR6
	LXVD2X	(R8+R15), V0
	LXVD2X	(R4+R15), V1
	VCMPEQUBCC	V0, V1, V2
	BGELR_CR6
	LXVD2X	(R8+R16), V0
	LXVD2X	(R4+R16), V1
	VCMPEQUBCC	V0, V1, V2
	BGELR_CR6

	// 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

func moveDisjointStack32() {
	var s [32]byte
	// ppc64x:-".*memmove"
	// ppc64x/power8:"LXVD2X",-"ADD",-"BC"
	// ppc64x/power9:"LXV",-"LXVD2X",-"ADD",-"BC"
	copy(s[:], x[:32])
	runtime.KeepAlive(&s)
}

func moveDisjointStack64() {
	var s [96]byte
	// ppc64x:-".*memmove"
	// ppc64x/power8:"LXVD2X","ADD","BC"
	// ppc64x/power9:"LXV",-"LXVD2X",-"ADD",-"BC"
	copy(s[:], x[:96])
	runtime.KeepAlive(&s)
}

	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 \

	MOVD	$48, R18; \
	MOVD	$64, R19; \
	LXVD2X (blk_key)(R0), VS0; \
	LXVD2X (blk_key)(R16), VS1; \
	LXVD2X (blk_key)(R17), VS2; \
	LXVD2X (blk_key)(R18), VS3; \
	LXVD2X (blk_key)(R19), VS4; \
	ADD $64, R16; \
	ADD $64, R17; \
	ADD $64, R18; \
	ADD $64, R19; \
	LXVD2X (blk_key)(R16), VS5; \
	LXVD2X (blk_key)(R17), VS6; \
	LXVD2X (blk_key)(R18), VS7; \
	LXVD2X (blk_key)(R19), VS8; \
	ADD $64, R16; \

	DCBTST	(TGT)			// prepare data cache
	DCBT	(SRC)
	MOVD	OCTWORDS, CTR		// Number of 64 byte chunks
	MOVD	$32, IDX32
	MOVD	$48, IDX48
	PCALIGN	$16

forward64:
	LXVD2X	(R0)(SRC), VS32		// load 64 bytes
	LXVD2X	(IDX16)(SRC), VS33
	LXVD2X	(IDX32)(SRC), VS34
	LXVD2X	(IDX48)(SRC), VS35
	ADD	$64, SRC
	STXVD2X	VS32, (R0)(TGT)		// store 64 bytes
	STXVD2X	VS33, (IDX16)(TGT)
	STXVD2X	VS34, (IDX32)(TGT)
	STXVD2X VS35, (IDX48)(TGT)

	PCALIGN	$16
loop64:
	LXVD2X	(R0)(R8),V2	// Scan 64 bytes at a time, starting at &s[0]
	VCMPEQUBCC	V2,V1,V6
	BNE	CR6,foundat0	// Match found at R8, jump out

	LXVD2X	(R11)(R8),V2
	VCMPEQUBCC	V2,V1,V6
	BNE	CR6,foundat1	// Match found at R8+16 bytes, jump out

	LXVD2X	(R12)(R8),V2
	VCMPEQUBCC	V2,V1,V6
	BNE	CR6,foundat2	// Match found at R8+32 bytes, jump out

	LXVD2X	(R6)(R8),V2
	VCMPEQUBCC	V2,V1,V6

	MOVWZ	$8, TEMP
	LVSL	(TEMP)(R0), LEMASK
	VSPLTISB	$0x0F, KI
	VXOR	KI, LEMASK, LEMASK
#endif

	LXVD2X	(CTX)(R_x000), VS32	// v0 = vs32
	LXVD2X	(CTX)(R_x010), VS34	// v2 = vs34
	LXVD2X	(CTX)(R_x020), VS36	// v4 = vs36

	// unpack the input values into vector registers
	VSLDOI	$8, V0, V0, V1
	LXVD2X	(CTX)(R_x030), VS38	// v6 = vs38
	VSLDOI	$8, V2, V2, V3
	VSLDOI	$8, V4, V4, V5
	VSLDOI	$8, V6, V6, V7

	BLT	tail		// Jump to the small string case
	SRD	$5, R4, R20
	MOVD	R20, CTR
	MOVD	$16, R21
	XXLXOR	V4, V4, V4
	XXLXOR	V5, V5, V5

	PCALIGN	$16
cmploop:
	LXVD2X	(R0)(R3), V0	// Count 32B per loop with two vector accumulators.
	LXVD2X	(R21)(R3), V2
	VCMPEQUB V2, V1, V2
	VCMPEQUB V0, V1, V0
	VPOPCNTD V2, V2		// A match is 0xFF or 0. Count the bits into doubleword buckets.
	VPOPCNTD V0, V0