MOVD	0(R1), R3 // argc
	ADD	$8, R1, R4 // argv

	// Did the TLS pointer get set? If so, don't change it (e.g musl).
	CMP	R13, $0
	BNE	tls_and_argcv_in_reg

	MOVD	$runtime·m0+m_tls(SB), R13 // TLS
	ADD	$0x7000, R13

tls_and_argcv_in_reg:
	BR	main(SB)

TEXT main(SB),NOSPLIT,$-8
	MOVD	$runtime·rt0_go(SB), R12
	MOVD	R12, CTR

// license that can be found in the LICENSE file.

#include "textflag.h"

//in plan 9 argc is at top of stack followed by ptrs to arguments

TEXT _rt0_arm_plan9(SB),NOSPLIT|NOFRAME,$0
	MOVW	R0, _tos(SB)
	MOVW	0(R13), R0
	MOVW	$4(R13), R1
	B	runtime·rt0_go(SB)

#include "go_asm.h"
#include "go_tls.h"
#include "funcdata.h"
#include "textflag.h"

// We have to resort to TLS variable to save g (R13).
// One reason is that external code might trigger
// SIGSEGV, and our runtime.sigtramp don't even know we
// are in external code, and will continue to use R13,
// this might well result in another SIGSEGV.

// save_g saves the g register into pthread-provided

	JMP	bar<>+4(SB)(R11*4)
	JCS	2(PC)
	JMP	*4(SP) // JMP 4(SP)
	JCS	2(PC)
	JMP	*(R12) // JMP (R12)
	JCS	2(PC)
//	JMP	*(R12*4) // TODO: This line is silently dropped on the floor!
	JCS	2(PC)
	JMP	*(R12)(R13*4) // JMP (R12)(R13*4)
	JCS	2(PC)
	JMP	*(AX) // JMP (AX)
	JCS	2(PC)
	JMP	*(SP) // JMP (SP)
	JCS	2(PC)
//	JMP	*(AX*4) // TODO: This line is silently dropped on the floor!
	JCS	2(PC)
	JMP	*(AX)(AX*4) // JMP (AX)(AX*4)

	//TODO LDTR -0x1e(R3), R4                  // 64285eb8
	//TODO LDTR -0xe5(R3), R10                 // 6ab851f8
	//TODO LDTRB 0xf0(R13), R10                // aa094f38
	//TODO LDTRH 0xe8(R13), R23                // b7894e78
	//TODO LDTRSB -0x24(R20), R5               // 85cadd38
	//TODO LDTRSB -0x75(R9), R13               // 2db99838
	//TODO LDTRSH 0xef(R3), LR                 // 7ef8ce78
	//TODO LDTRSH 0x96(R19), R24               // 786a8978

	// Execute call on m->g0.
	get_tls(R15)
	CMPQ	R15, $0
	JE	noswitch

	MOVQ	g(R15), R13
	CMPQ	R13, $0
	JE	noswitch
	MOVQ	g_m(R13), R13
	CMPQ	R13, $0
	JE	noswitch
	// TODO(aram): do something about the cpu profiler here.

	MOVQ	m_g0(R13), R14
	CMPQ	g(R15), R14
	JNE	switch
	// executing on m->g0 already
	CALL	AX
	RET

switch:

	*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)"},

	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