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)
	SHA256ROUND0(9, 0x12835b01, R15, R8, R9, R10, R11, R12, R13, R14)
	SHA256ROUND0(10, 0x243185be, R14, R15, R8, R9, R10, R11, R12, R13)

	SHA512ROUND1(23, 0x76f988da831153b5, R9, R10, R11, R12, R13, R14, R15, R8)
	SHA512ROUND1(24, 0x983e5152ee66dfab, R8, R9, R10, R11, R12, R13, R14, R15)
	SHA512ROUND1(25, 0xa831c66d2db43210, R15, R8, R9, R10, R11, R12, R13, R14)
	SHA512ROUND1(26, 0xb00327c898fb213f, R14, R15, R8, R9, R10, R11, R12, R13)
	SHA512ROUND1(27, 0xbf597fc7beef0ee4, R13, R14, R15, R8, R9, R10, R11, R12)

	MOVW	R4, 12(R13)
	MOVW	R5, 16(R13)
	MOVW	R6, 20(R13)
	MOVW	R7, 24(R13)
	MOVW	R8, 28(R13)
	MOVW	g, 32(R13)
	MOVW	R11, 36(R13)

	// Skip floating point registers on goarmsoftfp != 0.
	MOVB    runtime·goarmsoftfp(SB), R11
	CMP	$0, R11
	BNE     skipfpsave
	MOVD	F8, (40+8*0)(R13)
	MOVD	F9, (40+8*1)(R13)
	MOVD	F10, (40+8*2)(R13)
	MOVD	F11, (40+8*3)(R13)
	MOVD	F12, (40+8*4)(R13)

	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)
	SHA256ROUND0(9, 0x12835b01, R15, R8, R9, R10, R11, R12, R13, R14)
	SHA256ROUND0(10, 0x243185be, R14, R15, R8, R9, R10, R11, R12, R13)

	ORR	R7, R9
	LSL	R3, R6, R8
	SUB	$2, R1
	STP.W	(R9, R10), -16(R0)
loop:
	CBZ	R1, done
	LDP.W	-32(R2), (R10, R11)
	LDP	16(R2), (R12, R13)
	LSR	R4, R13, R23
	ORR	R8, R23		// z[i] = (x[i] << s) | (x[i-1] >> (64 - s))
	LSL	R3, R13
	LSR	R4, R12, R22
	ORR	R13, R22
	LSL	R3, R12
	LSR	R4, R11, R21
	ORR	R12, R21
	LSL	R3, R11
	LSR	R4, R10, R20
	ORR	R11, R20
	LSL	R3, R10, R8

	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)"},

	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

	}
	// 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)

	// load_g will clobber R0, Save R0
	MOVD	R0, R13
	load_g
	// restore R0
	MOVD	R13, R0
	MOVD	g_m(g), R13
	MOVD	m_g0(R13), R14
	CMP	R14, g
	BEQ	noswitch	// branch if already on g0
	MOVD	R14, g

	MOVD	R0, 8(RSP)	// func arg
	MOVD	R1, 16(RSP)	// func arg
	BL	runtime·racecallback(SB)

	// All registers are smashed after Go code, reload.
	MOVD	g_m(g), R13
	MOVD	m_curg(R13), g	// g = m->curg
ret: