CMPQ	R8, $8
	JB	small

	CMPQ	R8, $63
	JBE	loop
#ifndef hasAVX2
	CMPB	internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JEQ     big_loop_avx2
	JMP	big_loop
#else
	JMP	big_loop_avx2
#endif
loop:
	CMPQ	R8, $16
	JBE	_0through16
	MOVOU	(SI), X0
	MOVOU	(DI), X1
	PCMPEQB X0, X1
	PMOVMSKB X1, AX
	XORQ	$0xffff, AX	// convert EQ to NE
	JNE	diff16	// branch if at least one byte is not equal
	ADDQ	$16, SI

	TESTL	BX, BX
	JEQ	_0
	CMPL	BX, $2
	JBE	_1or2
	CMPL	BX, $4
	JB	_3
	JE	_4
	CMPL	BX, $8
	JBE	_5through8
	CMPL	BX, $16
	JBE	_9through16
#ifdef GO386_softfloat
	JMP	nosse2
#endif
	PXOR	X0, X0
	CMPL	BX, $32
	JBE	_17through32
	CMPL	BX, $64
	JBE	_33through64
	CMPL	BX, $128
	JBE	_65through128
	CMPL	BX, $256
	JBE	_129through256

loop:
	MOVOU	X0, 0(DI)
	MOVOU	X0, 16(DI)

	MOVL	ptr+0(FP), DI
	MOVL	n+4(FP), BX
	XORL	AX, AX

tail:
	TESTL	BX, BX
	JEQ	_0
	CMPL	BX, $2
	JBE	_1or2
	CMPL	BX, $4
	JB	_3
	JE	_4
	CMPL	BX, $8
	JBE	_5through8
	CMPL	BX, $16
	JBE	_9through16
	MOVL	BX, CX
	SHRL	$2, CX
	REP
	STOSL
	ANDL	$3, BX
	JNE	tail
	RET

_1or2:
	MOVB	AX, (DI)
	MOVB	AX, -1(DI)(BX*1)
	RET
_0:
	RET
_3:
	MOVW	AX, (DI)

	TESTQ	BX, BX
	JEQ	_0
	CMPQ	BX, $2
	JBE	_1or2
	CMPQ	BX, $4
	JBE	_3or4
	CMPQ	BX, $8
	JB	_5through7
	JE	_8
	CMPQ	BX, $16
	JBE	_9through16
	CMPQ	BX, $32
	JBE	_17through32
	CMPQ	BX, $64
	JBE	_33through64
	CMPQ	BX, $128
	JBE	_65through128
	CMPQ	BX, $256
	JBE	_129through256

	CMPB	internal∕cpu·X86+const_offsetX86HasERMS(SB), $1 // enhanced REP MOVSB/STOSB
	JNE	skip_erms

	TESTL	BX, BX
	JEQ	move_0
	CMPL	BX, $2
	JBE	move_1or2
	CMPL	BX, $4
	JB	move_3
	JE	move_4
	CMPL	BX, $8
	JBE	move_5through8
	CMPL	BX, $16
	JBE	move_9through16
#ifdef GO386_softfloat
	JMP	nosse2
#endif
	CMPL	BX, $32
	JBE	move_17through32
	CMPL	BX, $64
	JBE	move_33through64
	CMPL	BX, $128
	JBE	move_65through128

nosse2:
/*
 * check and set for backwards
 */

	XORL	$0xffff, BX	// convert EQ to NE
	JNE	diff16	// branch if at least one byte is not equal
	ADDL	$16, SI
	ADDL	$16, DI
	SUBL	$16, BP
	JMP	largeloop

diff16:
	BSFL	BX, BX	// index of first byte that differs
	XORL	DX, DX
	MOVB	(SI)(BX*1), CX
	CMPB	CX, (DI)(BX*1)
	SETHI	DX
	LEAL	-1(DX*2), DX	// convert 1/0 to +1/-1
	MOVL	DX, (AX)
	RET

mediumloop:
	CMPL	BP, $4
	JBE	_0through4
	MOVL	(SI), BX
	MOVL	(DI), CX

	// CX = size (want in BX)
	CMPQ	AX, BX
	JNE	neq
	MOVQ	$1, AX	// return 1
	RET
neq:
	MOVQ	AX, SI
	MOVQ	BX, DI
	MOVQ	CX, BX
	JMP	memeqbody<>(SB)

// memequal_varlen(a, b unsafe.Pointer) bool
TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT,$0-17
	// AX = a       (want in SI)
	// BX = b       (want in DI)
	// 8(DX) = size (want in BX)
	CMPQ	AX, BX
	JNE	neq
	MOVQ	$1, AX	// return 1
	RET
neq:

	PMOVMSKB X1, DX
	// Count number of matching bytes
	POPCNTL DX, DX
	// Accumulate into R12
	ADDQ DX, R12
	// Advance to next block.
	ADDQ	$16, DI
sseloopentry:
	CMPQ	DI, AX
	JBE	sseloop

	// Get the number of bytes to consider in the last 16 bytes
	ANDQ $15, BX
	JZ end

	// Create mask to ignore overlap between previous 16 byte block
	// and the next.
	MOVQ $16,CX
	SUBQ BX, CX

	JB loop33to63
fail_avx2:
	VZEROUPPER
fail:
	MOVQ $-1, (R11)
	RET
success_avx2:
	VZEROUPPER
	JMP success
sse42:
#ifndef hasSSE42
	CMPB internal∕cpu·X86+const_offsetX86HasSSE42(SB), $1
	JNE no_sse42
#endif
	CMPQ AX, $12
	// PCMPESTRI is slower than normal compare,
	// so using it makes sense only if we advance 4+ bytes per compare
	// This value was determined experimentally and is the ~same

	// for large sizes. The cutover is approximately 1K.
tail:
	TESTQ	BX, BX
	JEQ	move_0
	CMPQ	BX, $2
	JBE	move_1or2
	CMPQ	BX, $4
	JBE	move_3or4
	CMPQ	BX, $8
	JB	move_5through7
	JE	move_8
	CMPQ	BX, $16
	JBE	move_9through16

/*
 * check and set for backwards
 */
	CMPQ	SI, DI
	JLS	back

/*
 * forward copy loop
 */
forward:
	MOVQ	BX, CX