Re: Alternative implementation of the generic __ffs

[dean gaudet]

On Fri, 18 Apr 2008, Harvey Harrison wrote:

> On Fri, 2008-04-18 at 17:58 -0700, Joe Perches wrote:
> > On Fri, 2008-04-18 at 17:20 -0700, dean gaudet wrote:
> > > any reasonable compiler should figure out the two are the same... but i 
> > > really prefer spelling out the lack of dependencies of the computations by 
> > > breaking it out per-bit.
> > 
> > It seems gcc 4.3 (-Os or -O2) isn't a reasonable compiler.
> > 
> > I think this might be best:
> > 
> > int ffs32(unsigned int value)
> > {
> > 	int x;
> > 
> > 	value &= -value;
> > 	if (!(value & 0x55555555))
> > 		x = 1;
> > 	else
> > 		x = 0;
> > 	if (!(value & 0x33333333))
> > 		x |= 2;
> > 	if (!(value & 0x0f0f0f0f))
> > 		x |= 4;
> > 	if (!(value & 0x00ff00ff))
> > 		x |= 8;
> > 	if (!(value & 0x0000ffff))
> > 		x |= 16;
> > 
> > 	return x;
> > }
> > 
> 
> That produces the shortest assembly for me, also uses the fewest
> registers.

unfortunately it kind of defeats the purpose of the original code... which 
is high parallelism / no-dependencies.

have you benchmarked it?

-dean

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/