update implementation of k10mwait.  consider renaming monmwait,
and getting rid of pipefitting in port.

Reference: /n/atom/patch/applied2013/nixmonupd
Date: Sat Sep 14 22:14:46 CES 2013
Signed-off-by: quanstro@quanstro.net

--- /sys/src/nix/k10/l64v.s	Sat Sep 14 22:14:16 2013
+++ /sys/src/nix/k10/l64v.s	Sat Sep 14 22:14:18 2013
@@ -363,40 +363,34 @@
 	HLT
 	RET
 
-TEXT _monitor(SB), 1, $-4			/* void monitor(void*); */
-	MOVQ	RARG, AX			/* linear address to monitor */
-	XORQ	CX, CX				/* no optional extensions yet */
-	XORQ	DX, DX				/* no optional hints yet */
-	BYTE $0x0f; BYTE $0x01; BYTE $0xc8	/* MONITOR */
-	RET
+#define MONITOR	BYTE $0x0f; BYTE $0x01; BYTE $0xc8
+#define MWAIT		BYTE $0x0f; BYTE $0x01; BYTE $0xc9
 
-TEXT _mwait(SB), 1, $-4				/* void mwait(u32int); */
-	MOVLQZX	RARG, CX			/* optional extensions */
-	BYTE $0x0f; BYTE $0x01; BYTE $0xc9	/* MWAIT */
-	RET
+/*
+ * uintptr monmwait(void*, uintptr)
+ */
+TEXT	k10mwait(SB),1,$16
+	MOVQ	val+8(FP), BX
+
+_mmstart:
+	CMPQ	(BP), BX			/* changed yet? */
+	JNE	_mmdone
 
-TEXT	k10mwait+0(SB),0,$16
-k10mwloop:
-	MOVQ	RARG, CX
-	MOVQ	val+8(FP), DX
-	MOVQ	(CX), AX
-	CMPQ	AX, DX
-	JNE		k10mwdone
-	MOVQ	RARG, AX			/* linear address to monitor */
-	XORQ	CX, CX				/* no optional extensions yet */
-	XORQ	DX, DX				/* no optional hints yet */
-	BYTE $0x0f; BYTE $0x01; BYTE $0xc8	/* MONITOR */
-	MOVQ	RARG, CX
-	MOVQ	(CX),AX
-	MOVQ	val+8(FP), DX
-	CMPQ	AX, DX
-	JNE		k10mwdone
-	XORQ	DX, DX
-	XORQ CX, CX			/* optional extensions */
-	BYTE $0x0f; BYTE $0x01; BYTE $0xc9	/* MWAIT */
-	JMP		k10mwloop
-k10mwdone:
-	RET	,
+	MOVQ	BP, AX				/* linear address to monitor */
+	XORQ	CX, CX				/* extensions */
+	XORQ	DX, DX				/* hints */
+	MONITOR
+
+	CMPQ	(BP), BX			/* changed yet? */
+	JNE	_mmdone
+
+	/*XORQ	CX, CX*/			/* extensions (different from monitor) */
+	XORQ	AX, AX				/* hints */
+	MWAIT
+						/* questionable: pop out on irq */
+_mmdone:
+	MOVQ	(BP), AX
+	RET
 
 TEXT mul64fract(SB), 1, $-4
 	MOVQ	a+8(FP), AX