Subject: Re: [boost] [lockfree] review
From: Alexander Terekhov (terekhov_at_[hidden])
Date: 2011-08-29 13:08:49
Dave Abrahams wrote:
>
> on Sat Aug 27 2011, Gottlob Frege <gottlobfrege-AT-gmail.com> wrote:
>
> > There is no "synchronizes with" (standard's wording) because
> > synch-with requires thread 2 to *read* the value written into the
> > atomic by thread 1. We have no reads here - no if statement - so
> > reordering doesn't matter.
> > I don't think the standard disallows the reordering of i and j, even
> > if they are atomics.
>
> IIUC from the point of view of programs without undefined behavior, the
> standard disallows reordering in both cases.
The standard prescribes UB in the case of lock-free non-atomic observer
loads (the program is well-defined in the case of relaxed lock-free
atomic stores and loads to observe the reordering of two stores).
In reality, the generated code is the same (irrespective of whether the
code uses relaxed atomic stores/loads versus ordinary stores/loads).
The standard allows reordering in both cases. The only difference is UB
imposed by fiat in one case and no UB in another case.
regards,
alexander.