> I want to do some work on it to make use of the process ID register on
> the SA11x0 and XScale processors, and need to take the time to see how
> it works.


On Xscale (or any ARMv5 chip, for that matter) ffs() is easy:

	ffs:
		rsb	r1, r0, #0
		and	r0, r0, r1	@ Find least significant bit
		clz	r0, r0		@ Count leading zeros
		rsb	r0, r0, #32	@ Normalize to ffs form (from LSB)
		mov	pc, lr

No loops, 4 cycles (plus any stalls I haven't thought about :-)

R.