ROUND2(R22,R3,R4,R5,M01,0xf61e2562, 5);
	ROUND2(R5,R22,R3,R4,M06,0xc040b340, 9);
	ROUND2(R4,R5,R22,R3,M11,0x265e5a51,14);
	ROUND2(R3,R4,R5,R22,M00,0xe9b6c7aa,20);
	ROUND2(R22,R3,R4,R5,M05,0xd62f105d, 5);
	ROUND2(R5,R22,R3,R4,M10, 0x2441453, 9);
	ROUND2(R4,R5,R22,R3,M15,0xd8a1e681,14);
	ROUND2(R3,R4,R5,R22,M04,0xe7d3fbc8,20);
	ROUND2(R22,R3,R4,R5,M09,0x21e1cde6, 5);
	ROUND2(R5,R22,R3,R4,M14,0xc33707d6, 9);

	SUB  R21, R5, R21

	// Greater than 8 -- load the rightmost remaining bytes in msg
	// and put into R17 (h1)
	MOVD (R4)(R21), R17
	MOVD $16, R22

	// Find the offset to those bytes
	SUB R5, R22, R22
	SLD $3, R22

	// Shift to get only the bytes in msg
	SRD R22, R17, R17

	// Put 1 at high end
	MOVD $1, R23
	SLD  $3, R21
	SLD  R21, R23, R23
	OR   R23, R17, R17

	// Remainder is 8

				{1, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
			},
			outputs: []outputInfo{
				{1, 0},
				{0, 670826495}, // R0 R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R19 R20 R21 R22 R23 R24 R25 R26 R30
			},
		},
	},
	{

	// We won't get a nested signal.
	MOVBZ	runtime·iscgo(SB), R22
	CMP	R22, $0
	BNE	nosaveg
	MOVD	m_gsignal(R21), R22	// g.m.gsignal
	CMP	R22, $0
	BEQ	nosaveg

	CMP	g, R22
	BEQ	nosaveg
	MOVD	(g_stack+stack_lo)(R22), R22 // g.m.gsignal.stack.lo
	MOVD	g, (R22)

	BL	(CTR)	// Call from VDSO

	MOVD	$0, (R22)	// clear g slot, R22 is unchanged by C code

	JMP	finish

nosaveg:

	MOVD    16(R8), R21       // R21 = x[i+1]
	MOVD    24(R8), R22       // R22 = x[i+2]
	MOVDU   32(R8), R23       // R23 = x[i+3]
	MULLD   R9, R20, R24      // R24 = z0[i]
	MULHDU  R9, R20, R20      // R20 = z1[i]
	ADDC    R4, R24           // R24 = z0[i] + c
	MULLD   R9, R21, R25
	MULHDU  R9, R21, R21
	ADDE    R20, R25
	MULLD   R9, R22, R26
	MULHDU  R9, R22, R22
	MULLD   R9, R23, R27
	MULHDU  R9, R23, R23

	AND $0x7ffffffffffff, R4, R14

	ADD R0>>51, R11, R11
	ADD R1>>51, R12, R12
	ADD R2>>51, R13, R13
	ADD R3>>51, R14, R14
	// R4>>51 * 19 + R10 -> R10
	LSR $51, R4, R21
	MOVD $19, R22
	MADD R22, R10, R21, R10

	STP (R10, R11), 0(R20)
	STP (R12, R13), 16(R20)
	MOVD R14, 32(R20)

	MOVD	R21, 216(R1)
	MOVD	R22, 224(R1)
	MOVD	R23, 232(R1)
	MOVD	R24, 240(R1)
	MOVD	R25, 248(R1)
	MOVD	R26, 256(R1)
	MOVD	R27, 264(R1)
	MOVD	R28, 272(R1)
	MOVD	R29, 280(R1)
	MOVD	g, 288(R1)
	MOVD	LR, R31
	MOVD	R31, 32(R1)
	CALL	runtime·debugCallCheck(SB)
	MOVD	40(R1), R22
	XOR	R0, R0
	CMP	R22, $0
	BEQ	good
	MOVD	48(R1), R22
	MOVD	$8, R20
	TW	$31, R0, R0

	SLD	$2,R18,R18	// p[6]*4
	MOVWZ	(R10)(R17),R21	// tab[0][p[7]]
	ADD	$1024,R10,R10	// tab[1]
	RLDICL	$56,R9,$56,R19	// p[5]
	SLD	$2,R19,R19	// p[5]*4:1
	MOVWZ	(R10)(R18),R22	// tab[1][p[6]]
	ADD	$1024,R10,R10	// tab[2]
	XOR	R21,R22,R21	// xor done R22
	CLRLSLDI $56,R9,$2,R20
	MOVWZ	(R10)(R19),R23	// tab[2][p[5]]
	ADD	$1024,R10,R10	// &tab[3]
	XOR	R21,R23,R21	// xor done R23
	MOVWZ	(R10)(R20),R24	// tab[3][p[4]]

	LDP	1*16(a_ptr), (R6, R7)
	LDP	2*16(a_ptr), (R8, R9)
	LDP	0*16(b_ptr), (R16, R17)
	LDP	1*16(b_ptr), (R19, R20)
	LDP	2*16(b_ptr), (R21, R22)
	CSEL	EQ, R16, R4, R4
	CSEL	EQ, R17, R5, R5
	CSEL	EQ, R19, R6, R6
	CSEL	EQ, R20, R7, R7
	CSEL	EQ, R21, R8, R8
	CSEL	EQ, R22, R9, R9
	STP	(R4, R5), 0*16(res_ptr)
	STP	(R6, R7), 1*16(res_ptr)
	STP	(R8, R9), 2*16(res_ptr)

	LDP	3*16(a_ptr), (R4, R5)

				// if(g->panic != nil && g->panic->argp == FP) g->panic->argp = bottom-of-frame
				//
				//	MOVD g_panic(g), R22
				//	CMP R22, $0
				//	BEQ end
				//	MOVD panic_argp(R22), R23
				//	ADD $(autosize+8), R1, R24
				//	CMP R23, R24
				//	BNE end
				//	ADD $8, R1, R25
				//	MOVD R25, panic_argp(R22)
				// end:
				//	NOP
				//
				// The NOP is needed to give the jumps somewhere to land.