This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: RFC on mt_allocator.h
- From: Benjamin Kosnik <bkoz at redhat dot com>
- To: Felix Yen <fwy at alumni dot brown dot edu>
- Cc: stefan at xapa dot se, libstdc++ at gcc dot gnu dot org, ola at xapa dot se
- Date: Tue, 13 Jan 2004 21:31:28 -0600
- Subject: Re: RFC on mt_allocator.h
- Organization: Red Hat / Chicago
- References: <78261414-4637-11D8-A819-0003931E0BB8@alumni.brown.edu>
Wow! There are a lot of interesting things in your reply that I'm not
going to mention at the moment, as I'm trying to do something else.
Thanks though, for your introduction and the information you supplied.
>At the very least, an allocator as sophisticated as the one you're
>discussing should be accompanied by documentation that specifies its
>intended applications and describes its design as well as its
>limitations.
Yep. If you are interested, you could probably help with this documentation.
> I would also like to see a brief description of the
>performance test(s) used to qualify the allocator.
We have made some progress in the performance test area. They are still
evolving, but we do actually have two tests/code paths that we'd like to
make as fast as possible, for as many containers as possible.
You can find them in the recent gcc snapshots, in the
libstdc++-v3/testsuite/performance directory. One is the simple
push_back test (allocator.cc) and the other is the multi-threaded test
case with reclaimed memory (allocator_thread.cc).
The latter demonstrates known performance issues in the current default allocator.
-benjamin