From mboxrd@z Thu Jan  1 00:00:00 1970
From: Martin Olsen <triplefault@gmail.com>
Date: Wed, 10 Jan 2007 17:53:40 +0000
Subject: Re: [KJ] powers of 2, and the boundary case of zero
Message-Id: <200701101853.41058.triplefault@gmail.com>
List-Id: <kernel-janitors.vger.kernel.org>
References: <Pine.LNX.4.64.0701090539520.7476@localhost.localdomain>
In-Reply-To: <Pine.LNX.4.64.0701090539520.7476@localhost.localdomain>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: kernel-janitors@vger.kernel.org

On Wednesday 10 January 2007 16:40, you wrote:
> On Wed, 10 Jan 2007, Martin Olsen wrote:
> > Using "unsigned long" certainly seems more sane than int.
> >
> > I've split the code up in is_power_of_2{,_or_zero} functions, changed
> > "int" to "unsigned long" and added the circular inclusion protection.
> >
> > ---
> >
> > diff -uprN -X linux-2.6.19-vanilla/Documentation/dontdiff
> > linux-2.6.19-vanilla/include/linux/power_of_2.h
> > linux-2.6.19-changed/include/linux/power_of_2.h
> > --- linux-2.6.19-vanilla/include/linux/power_of_2.h	1970-01-01
> > 01:00:00.000000000 +0100
> > +++ linux-2.6.19-changed/include/linux/power_of_2.h	2007-01-10
> > 16:05:28.000000000 +0100
> > @@ -0,0 +1,30 @@
> > +#ifndef _LINUX_POWER_OF_2_H
> > +#define _LINUX_POWER_OF_2_H
> > +
> > +static inline bool is_power_of_2_or_zero(unsigned long i)
> > +{
> > +	return (i & (i - 1)) = 0;
> > +}
> > +
> > +static inline bool is_power_of_2(unsigned long i)
> > +{
> > +	return i && is_power_of_2_or_zero(i);
> > +}
> > +
> > +static inline unsigned long round_up_to_power_of_2(unsigned long i)
> > +{
> > +	while ((i & (i - 1)) != 0)
> > +		i += i & ~(i - 1);
> > +
> > +	return i;
> > +}
> > +
> > +static inline unsigned long round_down_to_power_of_2(unsigned long i)
> > +{
> > +	while (i & (i - 1))
> > +		i &= (i - 1);
> > +
> > +	return i;
> > +}
> > +
> > +#endif /* !_LINUX_POWER_OF_2_H */
>
>              ^^^^^^^^^^^^^^^^^^^^  i think that might be overkill.  at
> most, the existing header files rarely add the "!".
>
> looks reasonably sane, i don't see any immediate problems.  i'm still
> interested if there's a single (bitwise?) operation to do the rounding
> up or down without involving a loop, which would be cool, but i don't
> know if that's possible.  anyone?

A quick search on Google Code found this snippet:

/* rounding macros, assumes ALIGN is a power of two */
#define ROUND_UP(N,ALIGN)       (((N) + ((ALIGN)-1)) & ~((ALIGN)-1))
#define ROUND_DOWN(N,ALIGN)     ((N) & ~((ALIGN)-1))

I am no mathematician, but wouldn't the above code work if ALIGN was set to 2?

> at this point, i would submit this as an official patch to the KJ
> list.  um ... you *have* tested this code to see what it does in
> various situations, right?  :-)

Heh.. No, I forgot that. A quick testcase compiled against my header file 
correctly denies any knowledge of "bool" (not defined in C). Should I include 
linux/types.h or just use "_Bool" (is _Bool part of C? I cant find any 
definitions of it in my include directories)?

But other than that the code looks okay.
_______________________________________________
Kernel-janitors mailing list
Kernel-janitors@lists.osdl.org
https://lists.osdl.org/mailman/listinfo/kernel-janitors