On Thu, 23 Apr 2009, Robert P. J. Day wrote:
>
> once upon a time, the kernel source code was replete with
> conditional constructs of the form:
>
> if ((n & (n-1)) == 0)
>
> as a way of testing whether something was a power of two. mercifully,
> include/linux/log2.h was created which introduced, among other things:
>
> static inline __attribute__((const))
> bool is_power_of_2(unsigned long n)
> {
> return (n != 0 && ((n & (n - 1)) == 0));
> }
>
> so we could simply ask whether "is_power_of_2(n)", which is convenient
> when we're testing things like, oh, blocksize.
>
> similarly, there are bunches of places which need to test whether an
> integer value has only a single set bit (for instance, to make sure
> only one flag bit out of a number of mutually exclusive bits are set).
>
> mathematically, that would be the same test, of course, but
> semantically, it would be ugly and inappropriate. is there,
> somewhere, a corresponding macro/function that asks:
>
> single_bit_set(n)
>
> if not, that would be handy, could be plopped into
> include/linux/bitops.h and could be defined exactly the same way, and
> would allow piles of code to be simplified.
>
> thoughts?
I found 8 occurrences with the following Coccinelle semantic match:
@@ expression n; @@
* (n & (n-1)) == 0
The most unpleasant was:
(!(((fp)->ipend & ~0x30) & (((fp)->ipend & ~0x30) - 1)))
in the file arch/blackfin/kernel/traps.c.
At least some of them seemed to have comments nearby that suggested that
searching for a single 1-bit was indeed the goal. I haven't looked at all
of them, though.
julia
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
