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)

	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.

	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

	// 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:

	MOVQ	p_len+16(FP),	DX
	SHRQ	$6,		DX
	SHLQ	$6,		DX

	MOVQ	$K_XMM_AR<>(SB), R8

	MOVQ	DI, R9
	MOVQ	SI, R10
	LEAQ	64(SI), R13

	ADDQ	SI, DX
	ADDQ	$64, DX
	MOVQ	DX, R11

	CMPQ	R13, R11
	CMOVQCC	R8, R13

	VMOVDQU	BSWAP_SHUFB_CTL<>(SB), Y10

	CALC // RET is inside macros

DATA K_XMM_AR<>+0x00(SB)/4,$0x5a827999
DATA K_XMM_AR<>+0x04(SB)/4,$0x5a827999

	MOVD	R9, R0

	// Now on a scheduling stack (a pthread-created stack).
	// Save room for two of our pointers /*, plus 32 bytes of callee
	// save area that lives on the caller stack. */
	MOVD	RSP, R13
	SUB	$16, R13
	MOVD	R13, RSP
	MOVD	R4, 0(RSP)	// save old g on stack
	MOVD	(g_stack+stack_hi)(R4), R4
	SUB	R2, R4
	MOVD	R4, 8(RSP)	// save depth in old g stack (can't just save SP, as stack might be copied during a callback)

	// GOT slot is filled by the dynamic linker with the offset of the thread-local
	// symbol from the thread pointer (R13)).
	R_POWER_TLS_IE

	// R_POWER_TLS marks an X-form instruction such as "ADD R3,R13,R4" as completing
	// a sequence of GOT-relative relocations to compute a TLS address. This can be
	// used by the system linker to to rewrite the GOT-relative TLS relocation into a

// Must obey the gcc calling convention.
TEXT _cgo_topofstack(SB),NOSPLIT|NOFRAME,$0
	// g (R13), R10, R11 and LR (R14) are callee-save in the C ABI, so save them
	MOVD	g, R1
	MOVD	R10, R3
	MOVD	LR, R4
	MOVD	R11, R5

	BL	runtime·load_g(SB)	// clobbers g (R13), R10, R11
	MOVD	g_m(g), R2
	MOVD	m_curg(R2), R2
	MOVD	(g_stack+stack_hi)(R2), R2

	MOVD	R1, g
	MOVD	R3, R10