Re: [KJ] powers of 2, and the boundary case of zero

[Arnaldo Carvalho de Melo <acme@mandriva.com>]

On Tue, Jan 09, 2007 at 01:57:19PM -0500, Robert P. J. Day wrote:
> On Tue, 9 Jan 2007, Randy Dunlap wrote:
> 
> > On Tue, 9 Jan 2007 11:16:50 -0500 (EST) Robert P. J. Day wrote:
> >
> > > On Wed, 10 Jan 2007, Darren Jenkins wrote:
> > >
> > > > > Robert P. J. Day wrote:
> > > >
> > > > > i would disagree.  as i wrote just a bit earlier, i think the
> > > > > fundamental definition of a power of 2 should *not* include zero,
> > > > > because of both mathematical and source code precedent.
> > > > >
> > > > > if anything, i would propose something like:
> > > > >
> > > > > #define IS_POWER_OF_2(x)                (x != 0 && ((x & (x - 1)) = 0))
> > > > > #define IS_POWER_OF_2_OR_ZERO(x)        (IS_POWER_OF_2(x) || x = 0)
> > > >
> > > > Hmm mathematically you are right, so I guess that this naming would
> > > > be better.
> > >
> > > just for some final entertainment value, i googled on the notion of
> > > testing for power-of-2ness in C, and found the following:
> > >
> > >   http://www.parashift.com/c++-faq-lite/intrinsic-types.html#faq-26.12
> > >
> > > so that seems to finalize the rule that zero is not considered a power
> > > of two.  now if someone wants to propose an initial power-of-2.h
> > > header file with the appropriate routines, we can argue about *that*.
> > > :-)
> >
> > Looks like using an inline function would be safer and probably
> > more in the current vogue.
> 
> agreed.  and, as i wrote earlier, i'd like to see this stuff in its
> own header file in include/linux, and not dumped into kernel.h to make
> that file even more chaotic than it already is.
> 
> so ... who's up for it?

You? Step up! :-) I'd gladly sign-off it

- arnaldo
_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/kernel-janitors