MOVQ	val+4(FP), M0
	MOVQ	M0, (AX)
	EMMS
	// This is essentially a no-op, but it provides required memory fencing.
	// It can be replaced with MFENCE, but MFENCE was introduced only on the Pentium4 (SSE2).
	XORL	AX, AX
	LOCK
	XADDL	AX, (SP)
	RET

// void	·Or8(byte volatile*, byte);
TEXT ·Or8(SB), NOSPLIT, $0-5
	MOVL	ptr+0(FP), AX
	MOVB	val+4(FP), BX
	LOCK
	ORB	BX, (AX)
	RET

	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
	CALL	runtime·mstart(SB)

	POP_REGS_HOST_TO_ABI0()

	XORL	AX, AX			// return 0 == success
	RET

// set tls base to DI
TEXT runtime·settls(SB),NOSPLIT,$0
	MOVQ	runtime·tls_g(SB), CX
	MOVQ	DI, 0(CX)(GS)
	RET

TEXT runtime·nanotime1(SB),NOSPLIT,$0-8