MOVBU	(R3), R6
	ADDU	$1, R3
	MOVBU	(R4), R7
	ADDU	$1, R4
	BEQ	R6, R7 , loop
	// bytes differed
	SGTU	R6, R7, R8
	MOVW	$-1, R6
	CMOVZ	R8, R6, R8
	JMP	cmp_ret
samebytes:
	SGTU	R1, R2, R6
	SGTU	R2, R1, R7
	SUBU	R7, R6, R8
cmp_ret:
	MOVW	R8, ret+16(FP)

	ADDV	R1, R2, R4

	// if less than 16 bytes or no MSA, do words check
	SGTU	$16, R2, R3
	BNE	R3, no_msa
	MOVBU	internal∕cpu·MIPS64X+const_offsetMIPS64XHasMSA(SB), R3
	BEQ	R3, R0, no_msa

	VMOVB	$0, W0

	SGTU	$128, R2, R3
	BEQ	R3, msa_large

	AND	$15, R2, R5
	XOR	R2, R5, R6
	ADDVU	R1, R6

msa_small:
	VMOVB	W0, (R1)
	ADDVU	$16, R1
	SGTU	R6, R1, R3
	BNE	R3, R0, msa_small
	BEQ	R5, R0, done
	VMOVB	W0, -16(R4)

TEXT runtime·memmove<ABIInternal>(SB), NOSPLIT|NOFRAME, $0-24
	BNE	R6, check
	RET

check:
	SGTU	R4, R5, R7
	BNE	R7, backward

	ADDV	R4, R6, R9 // end pointer

	// if the two pointers are not of same alignments, do byte copying
	SUBVU	R5, R4, R7
	AND	$7, R7
	BNE	R7, out

	// if less than 8 bytes, do byte copying
	SGTU	$8, R6, R7
	BNE	R7, out

	// do one byte at a time until 8-aligned
	AND	$7, R4, R8

	MOVV	from+8(FP), R2
	MOVV	n+16(FP), R3
	BNE	R3, check
	RET

check:
	SGTU	R1, R2, R4
	BNE	R4, backward

	ADDV	R1, R3, R6 // end pointer

	// if the two pointers are not of same alignments, do byte copying
	SUBVU	R2, R1, R4
	AND	$7, R4
	BNE	R4, out

	// if less than 8 bytes, do byte copying
	SGTU	$8, R3, R4
	BNE	R4, out

	// do one byte at a time until 8-aligned
	AND	$7, R1, R5

byte_loop:
	BEQ	R3, R8, samebytes
	MOVBU	(R3), R6
	ADDVU	$1, R3
	MOVBU	(R4), R7
	ADDVU	$1, R4
	BEQ	R6, R7, byte_loop

byte_cmp:
	SGTU	R6, R7, R8 // R8 = 1 if (R6 > R7)
	BNE	R0, R8, ret
	MOVV	$-1, R8
	JMP	ret

samebytes:
	SGTU	R1, R2, R6
	SGTU	R2, R1, R7
	SUBV	R7, R6, R8

ret:
	MOVV	R8, (R9)

byte_loop:
	BEQ	R4, R12, samebytes
	MOVBU	(R4), R8
	ADDVU	$1, R4
	MOVBU	(R6), R9
	ADDVU	$1, R6
	BEQ	R8, R9, byte_loop

byte_cmp:
	SGTU	R8, R9, R4 // R12 = 1 if (R8 > R9)
	BNE	R0, R4, ret
	MOVV	$-1, R4
	JMP	ret

samebytes:
	SGTU	R5, R7, R8
	SGTU	R7, R5, R9
	SUBV	R9, R8, R4

ret:

	SGT	$65536, R4, R5		// 1e02001485781200
	SGT	$4096, R4, R5		// 3e00001485781200
	SGT	$65536, R4		// 1e02001484781200
	SGT	$4096, R4		// 3e00001484781200
	SGTU	$65536, R4, R5		// 1e02001485f81200
	SGTU	$4096, R4, R5		// 3e00001485f81200
	SGTU	$65536, R4		// 1e02001484f81200
	SGTU	$4096, R4		// 3e00001484f81200
	ADDU	$65536, R4, R5		// 1e02001485781000
	ADDU	$4096, R4, R5		// 3e00001485781000
	ADDU	$65536, R4		// 1e02001484781000

	ADDV	R4, R5, R6

	// if less than 8 bytes, do one byte at a time
	SGTU	$8, R5, R8
	BNE	R8, out

	// 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:

	SGT	$74565, R4, R5 			// 5e020014de178d0385781200
	SGT	$4097, R4 			// 3e000014de07800384781200
	SGT	$4097, R4, R5 			// 3e000014de07800385781200
	SGTU	$74565, R4 			// 5e020014de178d0384f81200
	SGTU	$74565, R4, R5 			// 5e020014de178d0385f81200
	SGTU	$4097, R4 			// 3e000014de07800384f81200
	SGTU	$4097, R4, R5 			// 3e000014de07800385f81200
	ADDU	$74565, R4 			// 5e020014de178d0384781000
	ADDU	$74565, R4, R5 			// 5e020014de178d0385781000

// memequal(a, b unsafe.Pointer, size uintptr) bool
TEXT runtime·memequal(SB),NOSPLIT|NOFRAME,$0-25
	MOVV	a+0(FP), R1
	MOVV	b+8(FP), R2
	BEQ	R1, R2, eq
	MOVV	size+16(FP), R3
	ADDV	R1, R3, R4

	// chunk size is 16
	SGTU	$16, R3, R8
	BEQ	R0, R8, chunk_entry

byte_loop:
	BNE	R1, R4, byte_test
	MOVV	$1, R1
	MOVB	R1, ret+24(FP)
	RET
byte_test:
	MOVBU	(R1), R6
	ADDV	$1, R1
	MOVBU	(R2), R7
	ADDV	$1, R2