| Date | Fri, 26 Oct 2012 13:11:09 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH 01/16] math128: Introduce various 128bit primitives |
| |
* Peter Zijlstra <peterz@infradead.org> wrote:
> I had hoped the u128 stuff might be elsewise useful, but if we > don't want to go there, that's fine.
I think it needs a clearer usecase - and even then the 32-bit behavior still looks rather horrible ...
So if we can escape all that with reasonable restrictions then that's far better than taking on this kind of overhead for 32-bit systems. 32-bit still matters, we do the ktime_t complications for 32-bit systems and that's for a far smaller effect.
[ Would be nice to also stick in a WARN_ONCE() in the key place(s) just in case, to make sure the overflow cannot happen silently in the future. ]
Thanks,
Ingo
|