reg := fmt.Sprintf("R%d", i)
		l.add("MOVD", reg, 8)
	}
	l.addSpecial(
		"MOVW CR, R31\nMOVW R31, %d(R1)",
		"MOVW %d(R1), R31\nMOVFL R31, $0xff", // this is MOVW R31, CR
		8)                                    // CR is 4-byte wide, but just keep the alignment
	l.addSpecial(
		"MOVD XER, R31\nMOVD R31, %d(R1)",
		"MOVD %d(R1), R31\nMOVD R31, XER",
		8)
	// Add floating point registers F0-F31.
	for i := 0; i <= 31; i++ {

	// save scratch register R31 first
	MOVD	R31, -184(R1)
	MOVD	0(R1), R31
	// save caller LR
	MOVD	R31, -304(R1)
	MOVD	-32(R1), R31
	// save argument frame size
	MOVD	R31, -192(R1)
	MOVD	LR, R31
	MOVD	R31, -320(R1)
	ADD	$-320, R1
	// save all registers that can contain pointers
	// and the CR register
	MOVW	CR, R31
	MOVD	R31, 8(R1)
	MOVD	R2, 24(R1)
	MOVD	R3, 56(R1)

			faultOnNilArg0: true,
			unsafePoint:    true,
		},

		// R31 is temp register
		// Loop code:
		//	MOVD len/32,R31		set up loop ctr
		//	MOVD R31,CTR
		//	MOVD $16,R31		index register
		// loop:
		//	LXVD2X (R0)(R4),VS32
		//	LXVD2X (R31)(R4),VS33
		//	ADD  R4,$32          increment src
		//	STXVD2X VS32,(R0)(R3)
		//	STXVD2X VS33,(R31)(R3)
		//	ADD  R3,$32          increment dst

	{"R20", "R20"},
	{"R21", "R21"},
	{"R22", "R22"},
	{"R23", "R23"},
	{"R24", "R24"},
	{"R25", "R25"},
	{"R26", "R26"},
	{"R27", "R27"},
	{"R28", "R28"},
	{"R29", "R29"},
	{"R3", "R3"},
	{"R31", "R31"},
	{"R4", "R4"},
	{"R5", "R5"},
	{"R6", "R6"},
	{"R7", "R7"},
	{"R8", "R8"},
	{"R9", "R9"},
	{"SPR(269)", "SPR(269)"},
	{"a(FP)", "a(FP)"},
	{"g", "g"},
	{"ret+8(FP)", "ret+8(FP)"},

		// It saves all GP registers if necessary,
		// but clobbers R31 (LR) because it's a call
		// and R23 (REGTMP).
		// Returns a pointer to a write barrier buffer in R25.
		{name: "LoweredWB", argLength: 1, reg: regInfo{clobbers: (callerSave &^ gpg) | buildReg("R31"), outputs: []regMask{buildReg("R25")}}, clobberFlags: true, aux: "Int64"},

	MOVV	0(R4), R0	// make sure g != nil
	JMP	gogo<>(SB)

TEXT gogo<>(SB), NOSPLIT|NOFRAME, $0
	MOVV	R4, g
	JAL	runtime·save_g(SB)

	MOVV	0(g), R2
	MOVV	gobuf_sp(R3), R29
	MOVV	gobuf_lr(R3), R31
	MOVV	gobuf_ret(R3), R1
	MOVV	gobuf_ctxt(R3), REGCTXT
	MOVV	R0, gobuf_sp(R3)
	MOVV	R0, gobuf_ret(R3)
	MOVV	R0, gobuf_lr(R3)
	MOVV	R0, gobuf_ctxt(R3)
	MOVV	gobuf_pc(R3), R4
	JMP	(R4)

	// are callee-save in the gcc calling convention, so save them.
	MOVW	R23, R8
	MOVW	g, R9
	MOVW	R31, R10 // this call frame does not save LR

	JAL	runtime·load_g(SB)
	MOVW	g_m(g), R1
	MOVW	m_curg(R1), R1
	MOVW	(g_stack+stack_hi)(R1), R2 // return value in R2

	MOVW	R8, R23
	MOVW	R9, g
	MOVW	R10, R31

	RET

// The top-most function running on a goroutine
// returns to goexit+PCQuantum.

		// It saves all GP registers if necessary,
		// but clobbers R31 (LR) because it's a call
		// and R23 (REGTMP).
		// Returns a pointer to a write barrier buffer in R25.
		{name: "LoweredWB", argLength: 1, reg: regInfo{clobbers: (callerSave &^ gpg) | buildReg("R31"), outputs: []regMask{buildReg("R25")}}, clobberFlags: true, aux: "Int64"},

	                             ld    r3, r3, ...

	MOVD 131072(R3), R4          addis r31, r4, 2           pld r4, 131072(r3)
	                             ld    r4, 0(R3)

	ADD $131073, R3              lis  r31, 2                paddi r3, r3, 131073
	                             addi r31, 1
	                             add  r3,r31,r3

	MOVD $131073, R3             lis  r3, 2                 pli r3, 131073

	// so this function is reentrant.
	MOVV	m_vdsoPC(R17), R2
	MOVV	m_vdsoSP(R17), R3
	MOVV	R2, 8(R29)
	MOVV	R3, 16(R29)

	MOVV	$ret-8(FP), R2 // caller's SP
	MOVV	R31, m_vdsoPC(R17)
	MOVV	R2, m_vdsoSP(R17)

	MOVV	m_curg(R17), R4
	MOVV	g, R5
	BNE	R4, R5, noswitch

	MOVV	m_g0(R17), R4
	MOVV	(g_sched+gobuf_sp)(R4), R1	// Set SP to g0 stack

noswitch: