Re: [tip:x86/mce] x86/bitops: Move BIT_64() for a wider use

[Linus Torvalds]

On Wed, May 23, 2012 at 9:19 AM, Borislav Petkov
<borislav.petkov@xxxxxxx> wrote:
>
> Actually we need a BIT() macro that works both
> on 32- and 64-bit. But that won't be that easy:

We could use __builtin_choose_expr(), but that *only* works with constants.

So we could do this:

  static inline unsigned long bit(unsigned int x)
  {
      return 1ul << x;
  }

  static inline u64 bit64(unsigned int x)
  {
      return 1ull << x;
  }

  #define BIT(x) \
    __builtin_choose_expr((x) < 8*sizeof(unsigned long), bit(x), bit64(x))

but then you *have* to use a plain constant for the BIT() macro.
Anything else will error out in a big way. Non-constant users would
have to be modified to use bit() and bit64() instead.

And no, I tested. You apparently can't do

  #define __is_longlongshift(x) \
        (__builtin_constant_p(x) && (x) < 8*(sizeof(long)))

because while that is a compile-time constant expression, it's not
"constant enough" for __builtin_choose_expr().

                   Linus
--
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/