Re: [PATCH] Add "is_power_of_2" checking to log2.h.

["H. Peter Anvin" <hpa@zytor.com>]

Robert P. J. Day wrote:
> 
> seriously, though, there is the potential of breaking something with
> this change since you can see how there is some inconsistency in how
> it's done *now* just for powerpc which, in some places, defines its
> own versions of this:
> 
> ./arch/ppc/mm/pgtable.c:
> 	#define is_power_of_2(x) ((x) != 0 && (((x) & ((x) - 1)) == 0))
> ./arch/ppc/syslib/ppc85xx_rio.c:
> 	#define is_power_of_2(x) (((x) & ((x) - 1)) == 0)
> ./arch/powerpc/mm/pgtable_32.c:
> 	#define is_power_of_2(x) ((x) != 0 && (((x) & ((x) - 1)) == 0))
> 
> note how the first and third macros *won't* consider zero a power of
> two, while the second one *will*.  hence the need for a single
> standard for all of this, just to play it safe.
> 

I suspect the reason the test for zero was omitted is because the author
didn't want the extra cost (the test for zero needs an extra branch on a
lot of architectures.)

	-hpa