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

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Wed, 2012-05-23 at 18:19 +0200, Borislav Petkov wrote:
> On Wed, May 23, 2012 at 09:11:51AM -0700, H. Peter Anvin wrote:
> > On 05/23/2012 09:10 AM, Peter Zijlstra wrote:
> > > On Wed, 2012-05-23 at 09:07 -0700, tip-bot for Borislav Petkov wrote:
> > >> +#define BIT_64(n)                      (U64_C(1) << (n))
> > > 
> > > Because writing 1ULL << n is too much work? 
> > > 
> > 
> > Because writing 1ULL << n is broken in anything that needs to be used in
> > assembly, for example.
> 
> Actually we need a BIT() macro that works both
> on 32- and 64-bit. But that won't be that easy:
> http://lkml.indiana.edu/hypermail/linux/kernel/1010.1/02335.html
> 
> And it should return UL for shift values < 32 and ULL otherwise.

If I remember my type rules correctly you'll get something like that
with:

#define BIT(n)	({ typeof(n) __n = (n); (__n < 32) ? (1UL << __n) : (1ULL << __n); })

That said, having it do this might be unexpected, also hpa mentioned
something about assembly magics which will obviously not work with the
above either.

Anyway, ignore me, I just thought the BIT_64() thing looked funny, but
apparently there's good reasons for it.