This is the mail archive of the
mailing list for the libstdc++ project.
Re: basic_string atomicity
On Fri, Jan 6, 2012 at 8:30 AM, Paolo Carlini <firstname.lastname@example.org> wrote:
> sorry I'm traveling, thus I cannot follow the full exchange. Just wanted to make sure it's well clear to everybody that all these issues are moot in C++11 and in general I mean to resist to invasive changes in this area. The implementation isn't perfect but served us decently well so many years, is close to its end of life, and, as well known, the basic_string bits are critical in terms of ABI stability. Thus before any further investigation I would encourage the interested parties to at least double check and triple check too ;) that at least there are no ABI issues in all possible situations, possible targets, accidentally differnt inlining choices, etc.
> Atomicity (eg for locale) outside string seems also interesting, some of it will be probably different in the next ABI, we can surely start on those issues for 4.8.
Unfortunately the changes to __sync_fetch_and_add because of the C++11
memory model changes to GCC ALREADY HAVE CAUSED A REGRESSION that
affects libstdc++. G++ now produces worse code, at least for
architectures that can make a distinction among the various memory