MOVW	R0, R8

	MOVW	(0*4)(R8), R7 // fn
	MOVW	(1*4)(R8), R0 // a1
	MOVW	(2*4)(R8), R1 // a2
	MOVW	(3*4)(R8), R2 // a3
	MOVW	(4*4)(R8), R3 // a4
	MOVW	(5*4)(R8), R4 // a5
	MOVW	R4, 0(R13)
	MOVW	(6*4)(R8), R5 // a6
	MOVW	R5, 4(R13)
	MOVW	(7*4)(R8), R6 // a7
	MOVW	R6, 8(R13)
	MOVW	(8*4)(R8), R4 // a8
	MOVW	R4, 12(R13)
	MOVW	(9*4)(R8), R5 // a9
	MOVW	R5, 16(R13)

	SHA256ROUND0(3, 0xe9b5dba5, R13, R14, R15, R8, R9, R10, R11, R12)
	SHA256ROUND0(4, 0x3956c25b, R12, R13, R14, R15, R8, R9, R10, R11)
	SHA256ROUND0(5, 0x59f111f1, R11, R12, R13, R14, R15, R8, R9, R10)
	SHA256ROUND0(6, 0x923f82a4, R10, R11, R12, R13, R14, R15, R8, R9)
	SHA256ROUND0(7, 0xab1c5ed5, R9, R10, R11, R12, R13, R14, R15, R8)
	SHA256ROUND0(8, 0xd807aa98, R8, R9, R10, R11, R12, R13, R14, R15)

	MOVQ (CX), AX
	MULQ 8(BX)
	MOVQ AX, R9
	MOVQ DX, R8

	// r1 += a1×b0
	MOVQ 8(CX), AX
	MULQ (BX)
	ADDQ AX, R9
	ADCQ DX, R8

	// r1 += 19×a2×b4
	MOVQ   16(CX), AX
	IMUL3Q $0x13, AX, AX
	MULQ   32(BX)
	ADDQ   AX, R9
	ADCQ   DX, R8

	// r1 += 19×a3×b3
	MOVQ   24(CX), AX
	IMUL3Q $0x13, AX, AX
	MULQ   24(BX)
	ADDQ   AX, R9
	ADCQ   DX, R8

	// r1 += 19×a4×b2
	MOVQ   32(CX), AX

	CMPQ DI, $32
	JG large
	MOVQ x+24(FP), R8
	MOVQ y+48(FP), CX	// c = y
	MOVQ z+0(FP), R10

	MOVQ $0, SI		// i = 0

	// s/JL/JMP/ below to disable the unrolled loop
	SUBQ $4, DI		// n -= 4
	JL V3			// if n < 4 goto V3

U3:	// n >= 0
	// regular loop body unrolled 4x
	MOVQ 0(R8)(SI*8), R11
	MOVQ 8(R8)(SI*8), R12
	MOVQ 16(R8)(SI*8), R13
	MOVQ 24(R8)(SI*8), R14
	ADDQ CX, R11
	ADCQ $0, R12

	VPMADD52HUQ -7(DI)(R8*4), X11, K1, X18             // 62a2a509b59487f9ffffff
	VPMADD52HUQ X7, X31, K1, X18                       // 62e28501b5d7
	VPMADD52HUQ X0, X31, K1, X18                       // 62e28501b5d0
	VPMADD52HUQ 17(SP)(BP*2), X31, K1, X18             // 62e28501b5946c11000000
	VPMADD52HUQ -7(DI)(R8*4), X31, K1, X18             // 62a28501b59487f9ffffff

	CMPQ	SI, DI
	JEQ	allsame
	CMPQ	BX, DX
	MOVQ	DX, R8
	CMOVQLT	BX, R8 // R8 = min(alen, blen) = # of bytes to compare
	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

	LSR	R4, R6, R5	// return value
	LSL	R3, R6, R8	// x[i] << s
	SUB	$1, R1
one:	TBZ	$0, R1, two
	MOVD.W	-8(R2), R6
	LSR	R4, R6, R7
	ORR	R8, R7
	LSL	R3, R6, R8
	SUB	$1, R1
	MOVD.W	R7, -8(R0)
two:
	TBZ	$1, R1, loop
	LDP.W	-16(R2), (R6, R7)
	LSR	R4, R7, R10
	ORR	R8, R10
	LSL	R3, R7
	LSR	R4, R6, R9
	ORR	R7, R9
	LSL	R3, R6, R8
	SUB	$2, R1
	STP.W	(R9, R10), -16(R0)
loop:

	VP4DPWSSDS -7(DI)(R8*1), [Z4-Z7], K1, Z31          // 62225f4953bc07f9ffffff
	VP4DPWSSDS (SP), [Z4-Z7], K1, Z31                  // 62625f49533c24
	VP4DPWSSDS -7(DI)(R8*1), [Z14-Z17], K1, Z31        // 62220f4953bc07f9ffffff
	VP4DPWSSDS (SP), [Z14-Z17], K1, Z31                // 62620f49533c24
	VP4DPWSSDS -7(DI)(R8*1), [Z24-Z27], K1, Z31        // 62223f4153bc07f9ffffff

	SUBV	$8, R3
	SUBV	$8, R4

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)

	MOVD $·Fgetxattr(SB), R8
	MOVD R8, ret+0(FP)
	RET

// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT

TEXT ·get_FsetxattrAddr(SB), NOSPLIT|NOFRAME, $0-8
	MOVD $·Fsetxattr(SB), R8
	MOVD R8, ret+0(FP)
	RET

// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT

TEXT ·get_accept4Addr(SB), NOSPLIT|NOFRAME, $0-8
	MOVD $·accept4(SB), R8
	MOVD R8, ret+0(FP)
	RET