*R10
	*R11
	*R12
}

type R6 R5
type R7 R5
type R8 R5

type R9 struct {
	*R13
	*R14
	*R15
	*R16
}

type R10 R9
type R11 R9
type R12 R9

type R13 struct {
	*R17
	*R18
	*R19
	*R20
}

type R14 R13
type R15 R13
type R16 R13

type R17 struct {
	*R21
	*R22
	*R23
	*R24
}

type R18 R17
type R19 R17

	MOVV	(g_sched+gobuf_sp)(g), R3

	// Now on a scheduling stack (a pthread-created stack).
g0:
	// Save room for two of our pointers.
	ADDV	$-16, R3
	MOVV	R13, 0(R3)	// save old g on stack
	MOVV	(g_stack+stack_hi)(R13), R13
	SUBVU	R12, R13
	MOVV	R13, 8(R3)	// save depth in old g stack (can't just save SP, as stack might be copied during a callback)
	JAL	(R25)

	// Restore g, stack pointer. R4 is return value.

	{"-12(R4)", "-12(R4)"},
	{"0(PC)", "0(PC)"},
	{"1024", "1024"},
	{"12(R(1))", "12(R1)"},
	{"12(R13)", "12(R13)"},
	{"R0", "R0"},
	{"R0->(32-1)", "R0->31"},
	{"R0<<R1", "R0<<R1"},
	{"R0>>R(1)", "R0>>R1"},
	{"R0@>(32-1)", "R0@>31"},
	{"R1", "R1"},
	{"R11", "R11"},
	{"R12", "R12"},
	{"R13", "R13"},
	{"R14", "R14"},
	{"R15", "R15"},
	{"R1<<2(R3)", "R1<<2(R3)"},
	{"R(1)<<2(R(3))", "R1<<2(R3)"},

	VPGATHERQQ Y2, (BP)(Y7*2), Y1           // c4e2ed914c7d00
	VPGATHERQQ X12, (R13)(X14*2), X11       // c40299915c7500
	VPGATHERQQ Y12, (R13)(Y14*2), Y11       // c4029d915c7500
	VPGATHERQQ X2, (BP)(X7*2), X1           // c4e2e9914c7d00
	VPGATHERQQ Y2, (BP)(Y7*2), Y1           // c4e2ed914c7d00
	VPGATHERQQ X12, (R13)(X14*2), X11       // c40299915c7500
	VPGATHERQQ Y12, (R13)(Y14*2), Y11       // c4029d915c7500

	MOVQ	hookId+8(FP), RARG0
	MOVQ	s1+16(FP), RARG1
	MOVQ	s2+24(FP), RARG2
	MOVQ	result+32(FP), RARG3

	get_tls(R12)
	MOVQ	g(R12), R14
	MOVQ	g_m(R14), R13

	// Switch to g0 stack.
	MOVQ	SP, R12		// callee-saved, preserved across the CALL
	MOVQ	m_g0(R13), R10
	CMPQ	R10, R14
	JE	call	// already on g0
	MOVQ	(g_sched+gobuf_sp)(R10), SP
call:
	ANDQ	$~15, SP	// alignment for gcc ABI
	CALL	AX
	MOVQ	R12, SP
	RET

	MOVQ	g(R12), R14
	MOVQ	g_m(R14), R13
	MOVQ	m_g0(R13), R15
	CMPQ	R13, R15
	JEQ	noswitch	// branch if already on g0
	MOVQ	R15, g(R12)	// g = m->g0
	MOVQ	R15, R14	// set g register
	PUSHQ	RARG1	// func arg
	PUSHQ	RARG0	// func arg
	CALL	runtime·racecallback(SB)
	POPQ	R12
	POPQ	R12
	// All registers are smashed after Go code, reload.
	get_tls(R12)
	MOVQ	g(R12), R13
	MOVQ	g_m(R13), R13

	// will clobber the test pattern created by the caller
	ADD	$(1024 * 8), R13

	// Ask signaller to setgid
	MOVW	$·Baton(SB), R2
storeloop:
	MOVW	0(R2), R0
	MOVW	$1, R1
	BL	cas<>(SB)
	BCC	storeloop

	// Wait for setgid completion
loop:
	MOVW	$0, R0
	MOVW	$0, R1
	BL	cas<>(SB)
	BCC	loop

	// Restore stack
	SUB	$(1024 * 8), R13

	MOVW	R4, R14

	}
	// Add flag register.
	l.addSpecial(
		"MOVW CPSR, R0\nMOVW R0, %d(R13)",
		"MOVW %d(R13), R0\nMOVW R0, CPSR",
		4)

	// Add floating point registers F0-F15 and flag register.
	var lfp = layout{stack: l.stack, sp: "R13"}
	lfp.addSpecial(
		"MOVW FPCR, R0\nMOVW R0, %d(R13)",
		"MOVW %d(R13), R0\nMOVW R0, FPCR",
		4)
	for i := 0; i <= 15; i++ {
		reg := fmt.Sprintf("F%d", i)
		lfp.add("MOVD", reg, 8)

// of the args of the ANDQ needs to be saved so it can be used as the arg to TESTQ.
func andWithUse(x, y int) int {
	z := x & y
	// amd64:`TESTQ\s(AX, AX|BX, BX|CX, CX|DX, DX|SI, SI|DI, DI|R8, R8|R9, R9|R10, R10|R11, R11|R12, R12|R13, R13|R15, R15)`
	if z == 0 {
		return 77
	}
	// use z by returning it
	return z
}

// Verify (OR x (NOT y)) rewrites to (ORN x y) where supported
func ornot(x, y int) int {
	// ppc64x:"ORN"
	z := x | ^y

	print("r8      ", hex(r.r8), "\n")
	print("r9      ", hex(r.r9), "\n")
	print("r10     ", hex(r.r10), "\n")
	print("r11     ", hex(r.r11), "\n")
	print("r12     ", hex(r.r12), "\n")
	print("r13     ", hex(r.r13), "\n")
	print("r14     ", hex(r.r14), "\n")
	print("r15     ", hex(r.r15), "\n")
	print("rip     ", hex(r.rip), "\n")
	print("rflags  ", hex(r.eflags), "\n")
	print("cs      ", hex(r.segcs), "\n")