MOVL	m_g0(CX), DX

	// Layout new m scheduler stack on os stack.
	MOVL	SP, AX
	MOVL	AX, (g_stack+stack_hi)(DX)
	SUBL	$(64*1024), AX		// stack size
	MOVL	AX, (g_stack+stack_lo)(DX)
	MOVL	AX, g_stackguard0(DX)
	MOVL	AX, g_stackguard1(DX)

	// Initialize procid from TOS struct.
	MOVL	_tos(SB), AX
	MOVL	48(AX), AX
	MOVL	AX, m_procid(CX)	// save pid as m->procid

	// Finally, initialize g.
	get_tls(BX)

	MOVQ	m_g0(CX), DX

	// Layout new m scheduler stack on os stack.
	MOVQ	SP, AX
	MOVQ	AX, (g_stack+stack_hi)(DX)
	SUBQ	$(64*1024), AX		// stack size
	MOVQ	AX, (g_stack+stack_lo)(DX)
	MOVQ	AX, g_stackguard0(DX)
	MOVQ	AX, g_stackguard1(DX)

	// Initialize procid from TOS struct.
	MOVQ	_tos(SB), AX
	MOVL	64(AX), AX
	MOVQ	AX, m_procid(CX)	// save pid as m->procid

	// Finally, initialize g.
	get_tls(BX)

	MOVL	SP, AX
	MOVL	AX, (g_stack+stack_hi)(DX)
	SUBL	$(64*1024), AX		// initial stack size (adjusted later)
	MOVL	AX, (g_stack+stack_lo)(DX)
	ADDL	$const_stackGuard, AX
	MOVL	AX, g_stackguard0(DX)
	MOVL	AX, g_stackguard1(DX)

	// Set up tls.
	LEAL	m_tls(CX), DI
	MOVL	CX, g_m(DX)
	MOVL	DX, g(DI)
	MOVL	DI, 4(SP)
	CALL	runtime·setldt(SB) // clobbers CX and DX

	MOVQ	SP, AX
	MOVQ	AX, (g_stack+stack_hi)(DX)
	SUBQ	$(0x100000), AX		// stack size
	MOVQ	AX, (g_stack+stack_lo)(DX)
	ADDQ	$const_stackGuard, AX
	MOVQ	AX, g_stackguard0(DX)
	MOVQ	AX, g_stackguard1(DX)

	// Someday the convention will be D is always cleared.
	CLD

	CALL	runtime·stackcheck(SB)	// clobbers AX,CX
	CALL	runtime·mstart(SB)

	XORL	AX, AX			// return 0 == success

	MOVW	m_g0(R1), g

	// Layout new m scheduler stack on os stack.
	MOVW	R13, R0
	MOVW	R0, g_stack+stack_hi(g)
	SUB	$(64*1024), R0
	MOVW	R0, (g_stack+stack_lo)(g)
	MOVW	R0, g_stackguard0(g)
	MOVW	R0, g_stackguard1(g)

	// Initialize procid from TOS struct.
	MOVW	_tos(SB), R0
	MOVW	48(R0), R0
	MOVW	R0, m_procid(R1)	// save pid as m->procid

	BL	runtime·mstart(SB)

	// Exit the thread.

	MOVD	R0, g_m(g)
	BL	runtime·save_g(SB)

	// Set up stack guards for OS stack.
	MOVD	RSP, R0
	MOVD	R0, g_stack+stack_hi(g)
	SUB	$(64*1024), R0
	MOVD	R0, (g_stack+stack_lo)(g)
	MOVD	R0, g_stackguard0(g)
	MOVD	R0, g_stackguard1(g)

	BL	runtime·emptyfunc(SB)	// fault if stack check is wrong
	BL	runtime·mstart(SB)

	RESTORE_R19_TO_R28(8*3)

	// Exit the thread.
	MOVD	$0, R0
	RET

	BL	runtime·save_g(SB)

	// Layout new m scheduler stack on os stack.
	MOVW	R13, R0
	MOVW	R0, g_stack+stack_hi(g)
	SUB	$(64*1024), R0
	MOVW	R0, (g_stack+stack_lo)(g)
	MOVW	R0, g_stackguard0(g)
	MOVW	R0, g_stackguard1(g)

	BL	runtime·emptyfunc(SB)	// fault if stack check is wrong
	BL	runtime·mstart(SB)

	// Exit the thread.
	MOVW	$0, R0
	MOVM.IA.W (R13), [R4-R11, R15]		// pop {r4-r11, pc}

	MOVD	R1, R3
	MOVD	R3, (g_stack+stack_hi)(g)
	SUB	$(const_threadStackSize), R3		// stack size
	MOVD	R3, (g_stack+stack_lo)(g)
	ADD	$const_stackGuard, R3
	MOVD	R3, g_stackguard0(g)
	MOVD	R3, g_stackguard1(g)

	BL	runtime·mstart(SB)

	MOVD R0, R3
	RET


#define CSYSCALL()			\
	MOVD	0(R12), R12		\
	MOVD	R2, 40(R1)		\
	MOVD	0(R12), R0		\
	MOVD	8(R12), R2		\

	MOVQ	SP, AX
	MOVQ	AX, (g_stack+stack_hi)(DX)
	SUBQ	$(64*1024), AX		// initial stack size (adjusted later)
	MOVQ	AX, (g_stack+stack_lo)(DX)
	ADDQ	$const_stackGuard, AX
	MOVQ	AX, g_stackguard0(DX)
	MOVQ	AX, g_stackguard1(DX)

	// Set up tls.
	LEAQ	m_tls(CX), DI
	MOVQ	CX, g_m(DX)
	MOVQ	DX, g(DI)
	CALL	runtime·settls(SB) // clobbers CX

	CALL	runtime·stackcheck(SB)	// clobbers AX,CX

	// create istack out of the given (operating system) stack.
	// _cgo_init may update stackguard.
	MOVV	$runtime·g0(SB), g
	MOVV	$(-64*1024), R23
	ADDV	R23, R29, R1
	MOVV	R1, g_stackguard0(g)
	MOVV	R1, g_stackguard1(g)
	MOVV	R1, (g_stack+stack_lo)(g)
	MOVV	R29, (g_stack+stack_hi)(g)

	// if there is a _cgo_init, call it using the gcc ABI.
	MOVV	_cgo_init(SB), R25
	BEQ	R25, nocgo