LEAL	ret+12(FP), AX
	JMP	memeqbody<>(SB)
eq:
	MOVB    $1, ret+12(FP)
	RET

// memequal_varlen(a, b unsafe.Pointer) bool
TEXT runtime·memequal_varlen(SB),NOSPLIT,$0-9
	MOVL    a+0(FP), SI
	MOVL    b+4(FP), DI
	CMPL    SI, DI
	JEQ     eq
	MOVL    4(DX), BX    // compiler stores size at offset 4 in the closure
	LEAL	ret+8(FP), AX
	JMP	memeqbody<>(SB)
eq:
	MOVB    $1, ret+8(FP)
	RET

// a in SI
// b in DI

	// R6 = b_cap (unused)
	// R7 = byte to find
	AND	$0xff, R7
	MOVV	R4, R6		// store base for later
	ADDV	R4, R5		// end
	ADDV	$-1, R4

	PCALIGN	$16
loop:
	ADDV	$1, R4
	BEQ	R4, R5, notfound
	MOVBU	(R4), R8
	BNE	R7, R8, loop

	SUBV	R6, R4		// remove base
	RET

notfound:
	MOVV	$-1, R4
	RET

TEXT ·IndexByteString<ABIInternal>(SB),NOSPLIT,$0-32
	// R4 = s_base
	// R5 = s_len

	MOVHU	3(R0), R3
	CMP	R6, R3
	BNE	loop_6
	B	found
len_4_5:
	TBZ	$0, R3, len_4
len_5:
	// R5 and R7 contain 5-byte of sep
	MOVWU	(R2), R5
	MOVBU	4(R2), R7
loop_5:
	CMP	R4, R0
	BHI	not_found
	MOVWU.P	1(R0), R3
	CMP	R5, R3
	BNE	loop_5
	MOVBU	3(R0), R3
	CMP	R7, R3
	BNE	loop_5
	B	found
len_4:
	// R5 contains 4-byte of sep
	MOVWU	(R2), R5
loop_4:
	CMP	R4, R0
	BHI	not_found

	BLT	tail
	CLC	$256, 0(R3), 0(R5)
	BNE	notequal
	SUB	$256, R6
	LA	256(R3), R3
	LA	256(R5), R5
	BR	loop
tail:
	SUB	$1, R6, R8
	EXRL	$memeqbodyclc<>(SB), R8
	BEQ	equal
notequal:
	MOVB	$0, 0(R7)
	RET
equal:
	MOVB	$1, 0(R7)
	RET
tiny:
	MOVD	$0, R2
	CMPBLT	R6, $16, lt16
	MOVD	0(R3), R8
	MOVD	0(R5), R9
	CMPBNE	R8, R9, notequal
	MOVD	8(R3), R8
	MOVD	8(R5), R9
	CMPBNE	R8, R9, notequal

less_than_8:
	TBZ	$2, R11, less_than_4

	MOVWU.P	4(R13), R10
	CRC32CW	R10, R9

less_than_4:
	TBZ	$1, R11, less_than_2

	MOVHU.P	2(R13), R10
	CRC32CH	R10, R9

less_than_2:
	TBZ	$0, R11, done

	MOVBU	(R13), R10
	CRC32CB	R10, R9

done:
	MOVWU	R9, ret+32(FP)
	RET

// ieeeUpdate updates the non-inverted crc with the given data.

// func ieeeUpdate(crc uint32, p []byte) uint32

	MOVBZ 0(SRC), TMP
	MOVBZ TMP, 0(TGT)
	RET

backward:
	// Copying backwards proceeds by copying R7 bytes then copying R6 double words.
	// R3 and R4 are advanced to the end of the destination/source buffers
	// respectively and moved back as we copy.

	ADD	LEN, SRC, SRC		// end of source
	ADD	TGT, LEN, TGT		// end of dest

	BEQ	nobackwardtail		// earlier condition

	MOVD	BYTES, CTR			// bytes to move

	"BGEZ",
	"BGT",
	"BGTU",
	"BGTZ",
	"BLE",
	"BLEU",
	"BLEZ",
	"BLTZ",
	"BNEZ",
	"FABSD",
	"FABSS",
	"FNEGD",
	"FNEGS",
	"FNED",
	"FNES",
	"MOV",
	"MOVB",
	"MOVBU",
	"MOVF",
	"MOVD",
	"MOVH",
	"MOVHU",
	"MOVW",
	"MOVWU",
	"NEG",
	"NEGW",
	"NOT",
	"SEQZ",
	"SNEZ",
	"LAST",

	// do one byte at a time until 8-aligned
	AND	$7, R4, R8
	BEQ	R8, words
	MOVB	R0, (R4)
	ADDV	$1, R4
	JMP	-4(PC)

words:
	// do 8 bytes at a time if there is room
	ADDV	$-7, R6, R5

	PCALIGN	$16
	SGTU	R5, R4, R8
	BEQ	R8, out
	MOVV	R0, (R4)
	ADDV	$8, R4
	JMP	-4(PC)

out:
	BEQ	R4, R6, done
	MOVB	R0, (R4)
	ADDV	$1, R4
	JMP	-3(PC)
done:

	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
	CMPL	BX, CX

Daniel Grimm <******@****.***> 1708485021 +0100