VPCMPEQB	Y2, Y3, Y5
	VPAND	Y4, Y5, Y6
	VPMOVMSKB Y6, DX
	ADDQ	$64, SI
	ADDQ	$64, DI
	SUBQ	$64, BX
	CMPL	DX, $0xffffffff
	JEQ	hugeloop_avx2
	VZEROUPPER
	XORQ	AX, AX	// return 0
	RET

bigloop_avx2:
	VZEROUPPER

	// 8 bytes at a time using 64-bit register
	PCALIGN $16
bigloop:
	CMPQ	BX, $8
	JBE	leftover
	MOVQ	(SI), CX
	MOVQ	(DI), DX
	ADDQ	$8, SI
	ADDQ	$8, DI
	SUBQ	$8, BX

				{0, 239}, // AX CX DX BX BP SI DI
				{1, 239}, // AX CX DX BX BP SI DI
			},
			outputs: []outputInfo{
				{0, 239}, // AX CX DX BX BP SI DI
			},
		},
	},
	{
		name:         "ADCLconst",
		auxType:      auxInt32,
		argLen:       2,
		resultInArg0: true,
		clobberFlags: true,
		asm:          x86.AADCL,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 239}, // AX CX DX BX BP SI DI

	*r = r.Intersect(src.Bounds().Add(orig.Sub(*sp)))
	if mask != nil {
		*r = r.Intersect(mask.Bounds().Add(orig.Sub(*mp)))
	}
	dx := r.Min.X - orig.X
	dy := r.Min.Y - orig.Y
	if dx == 0 && dy == 0 {
		return
	}
	sp.X += dx
	sp.Y += dy
	if mp != nil {
		mp.X += dx
		mp.Y += dy
	}
}

func processBackward(dst image.Image, r image.Rectangle, src image.Image, sp image.Point) bool {

	MOVL y+8(FP), BP
	LEAL (DI)(BX*4), DI
	LEAL (SI)(BX*4), SI
	NEGL BX			// i = -n
	MOVL $0, CX		// c = 0
	JMP E6

L6:	MOVL (SI)(BX*4), AX
	MULL BP
	ADDL CX, AX
	ADCL $0, DX
	ADDL AX, (DI)(BX*4)
	ADCL $0, DX
	MOVL DX, CX
	ADDL $1, BX		// i++

E6:	CMPL BX, $0		// i < 0
	JL L6

	MOVL CX, c+12(FP)

	JMP	syscall·Syscall6(SB)

TEXT ·SyscallNoError(SB),NOSPLIT,$0-48
	CALL	runtime·entersyscall(SB)
	MOVQ	a1+8(FP), DI
	MOVQ	a2+16(FP), SI
	MOVQ	a3+24(FP), DX
	MOVQ	$0, R10
	MOVQ	$0, R8
	MOVQ	$0, R9
	MOVQ	trap+0(FP), AX	// syscall entry
	SYSCALL
	MOVQ	AX, r1+32(FP)
	MOVQ	DX, r2+40(FP)
	CALL	runtime·exitsyscall(SB)
	RET

TEXT ·RawSyscall(SB),NOSPLIT,$0-56
	JMP	syscall·RawSyscall(SB)

TEXT ·RawSyscall6(SB),NOSPLIT,$0-80

TEXT ·rawSyscallNoError(SB),NOSPLIT,$0-24
	MOVL	trap+0(FP), AX	// syscall entry
	MOVL	a1+4(FP), BX
	MOVL	a2+8(FP), CX
	MOVL	a3+12(FP), DX
	MOVL	$0, SI
	MOVL	$0, DI
	INVOKE_SYSCALL
	MOVL	AX, r1+16(FP)
	MOVL	DX, r2+20(FP)
	RET

#define SYS_SOCKETCALL 102	/* from zsysnum_linux_386.go */

// func socketcall(call int, a0, a1, a2, a3, a4, a5 uintptr) (n int, err int)

TEXT ·SyscallNoError(SB),NOSPLIT,$0-24
	CALL	runtime·entersyscall(SB)
	MOVL	trap+0(FP), AX  // syscall entry
	MOVL	a1+4(FP), BX
	MOVL	a2+8(FP), CX
	MOVL	a3+12(FP), DX
	MOVL	$0, SI
	MOVL	$0, DI
	INVOKE_SYSCALL
	MOVL	AX, r1+16(FP)
	MOVL	DX, r2+20(FP)
	CALL	runtime·exitsyscall(SB)
	RET

TEXT ·RawSyscall(SB),NOSPLIT,$0-28
	JMP	syscall·RawSyscall(SB)

TEXT ·RawSyscall6(SB),NOSPLIT,$0-40

TEXT runtime·gogo(SB), NOSPLIT, $0-8
	MOVQ	buf+0(FP), BX		// gobuf
	MOVQ	gobuf_g(BX), DX
	MOVQ	0(DX), CX		// make sure g != nil
	JMP	gogo<>(SB)

TEXT gogo<>(SB), NOSPLIT, $0
	get_tls(CX)
	MOVQ	DX, g(CX)
	MOVQ	DX, R14		// set the g register
	MOVQ	gobuf_sp(BX), SP	// restore SP
	MOVQ	gobuf_ret(BX), AX
	MOVQ	gobuf_ctxt(BX), DX
	MOVQ	gobuf_bp(BX), BP
	MOVQ	$0, gobuf_sp(BX)	// clear to help garbage collector

	// some values to and from memory to provide a temporary.
	// The initial temporary is X15, so zero its memory instead
	// of X15 itself.
	MOVL $0, DX
	MOVQ DX, X13
	MOVQ DX, X14
	MOVOU X14, (15*16)(BX)

	// 4 iterations. Each iteration is 8 quarter-rounds.
	MOVL $4, DX
loop:
	QR(X0, X4, X8, X12, X15)
	MOVOU X4, (4*16)(BX) // save X4
	QR(X1, X5, X9, X13, X15)
	MOVOU (15*16)(BX), X15 // reload X15; temp now X4

	ADDL -2147483648(AX), AX // 038000000080
	// Make sure MOV CR/DR continues to work after changing its movtabs.
	MOVL CR0, AX // 0f20c0
	MOVL CR0, DX // 0f20c2
	MOVL CR4, DI // 0f20e7
	MOVL AX, CR0 // 0f22c0
	MOVL DX, CR0 // 0f22c2
	MOVL DI, CR4 // 0f22e7
	MOVL DR0, AX // 0f21c0
	MOVL DR6, DX // 0f21f2
	MOVL DR7, SI // 0f21fe
	// Test other movtab entries.
	PUSHL SS // 16
	PUSHL FS // 0fa0
	POPL FS  // 0fa1
	POPL SS  // 17