"VPMSUMW",
	"VPMSUMD",
	"VMSUMUDM",
	"VR",
	"VRLB",
	"VRLH",
	"VRLW",
	"VRLD",
	"VS",
	"VSLB",
	"VSLH",
	"VSLW",
	"VSL",
	"VSLO",
	"VSRB",
	"VSRH",
	"VSRW",
	"VSR",
	"VSRO",
	"VSLD",
	"VSRD",
	"VSA",
	"VSRAB",
	"VSRAH",
	"VSRAW",
	"VSRAD",
	"VSOI",
	"VSLDOI",
	"VCLZ",
	"VCLZB",
	"VCLZH",
	"VCLZW",
	"VCLZD",
	"VPOPCNT",

	CMP	R5, $16		// Use GPR checks for check for len <= 16
	BLE	check0_16
	MOVD	$0, R3		// Assume no-match in case BGELR CR6 returns
	CMP	R5, $32		// Use overlapping VSX loads for len <= 32
	BLE	check17_32	// Do a pair of overlapping VSR compares
	CMP	R5, $64
	BLE	check33_64	// Hybrid check + overlap compare.

setup64:
	SRD	$6, R5, R6	// number of 64 byte chunks to compare
	MOVD	R6, CTR
	MOVD	$16, R14	// index for VSX loads and stores
	MOVD	$32, R15

	LXVD2X	(R0)(R9),V2
	VCMPEQUBCC	V2,V1,V6
	BNE	CR6,foundat0		// Match found at R8+48 bytes, jump out

	BR	notfound


cmp8:	// Length 8 - 15
#ifdef GOPPC64_power10
	// Load all the bytes into a single VSR in BE order.
	SLD	$56,R4,R5
	LXVLL	R3,R5,V2
	// Compare and count the number which don't match.
	VCMPEQUB	V2,V1,V6
	VCLZLSBB	V6,R3
	// If count is the number of bytes, or more. No matches are found.
	CMPU	R3,R4