This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Our proposal for a MATDS
- From: Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>
- To: libstdc++ at gcc dot gnu dot org
- Cc: rittle at latour dot rsch dot comm dot mot dot com
- Date: Fri, 7 Mar 2003 04:07:41 -0600 (CST)
- Subject: Re: Our proposal for a MATDS
- Organization: Networks and Infrastructure Lab (IL02/2240), Motorola Labs
- References: <3E63F7E9.7090908@snon.net>
In article <20030306115523 dot 4b19ceae dot bkoz at redhat dot com> Benjamin writes:
> This looks good. Can somebody give non-linux benchmark results?
I tested an earlier version of this code on FreeBSD. I posted my
results (brief summary: it reconfirmed that the FreeBSD malloc has a
different profile than Linux malloc under MT tests). Since this patch
is *not* changing the default allocator, I'd approve of it without
more retesting. Nathan raised interesting points (in fact, the
FreeBSD malloc implements that type of stuff, which is why it beats
the libstdc++-v3 pool allocators, especially in MT cases) but I don't
think they are a strike against this incremental improvement.
Regards,
Loren