putle on big endian machines failed because it started at the wrong
end of the number.  the following version is temptingly cleaner, but
is a miserable performer on mips (10x the run time), and shows poorly
on arm as well.

/* slow as molassas on risc-style machines */
void
putle(uchar *t, uvlong r, int w)
{
	uint i;

	for(i = 0; i < w; i++)
		t[i] = r>>i*8;
}

Reference: /n/atom/patch/applied2013/putleonbe
Date: Sat Oct 12 08:58:45 CES 2013
Signed-off-by: quanstro@quanstro.net

--- /sys/src/libc/port/getbe.c	Sat Oct 12 08:44:14 2013
+++ /sys/src/libc/port/getbe.c	Sat Oct 12 08:44:15 2013
@@ -20,10 +20,13 @@
 putle(uchar *t, uvlong r, int w)
 {
 	uchar *o, *f;
-	uint i;
+	uint i, k;
 
 	f = (uchar*)&r;
 	o = (uchar*)&lorder;
+	k = (o[0]!=0)*(8-w);		/* if big endian, use tail not head */
+	o += k;
+	f += k;
 	for(i = 0; i < w; i++)
 		t[o[i]] = f[i];
 }