This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: [v3] __pool_alloc locking bits
>Benjamin, is this is a feature or could it be changed in a future release?
I'm not quite sure what you mean by "this." In the quoted text, I was
trying to explain why we can't use multiple locks, which, I thought, you
were vehemently against anyway.
>I am relying on allocator<Foo1> and allocator<Foo2> to be *different*
>memory pools (assuming they are memory pools at all). Is this something
>that might change in the future?
Well, this is the case for the other allocators, ie __mt_alloc. I'm
strongly in favor of keeping __mt_alloc like this, so I think you are
ok.
For __pool_alloc, Paolo and others made the case that single-pool was
the historical behavior, and we should try to match old behaviors for
people that rely upon them. I think this is a good point.
The other point made was performance-orientated, but I'm not completely
convinced on that argument, at least as it applies to __mt_alloc, where
we can pick the behavior we want.
best,
benjamin