This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: Our proposal for a MATDS
- From: Nathan Myers <ncm-nospam at cantrip dot org>
- To: libstdc++ <libstdc++ at gcc dot gnu dot org>
- Date: Mon, 3 Mar 2003 19:59:54 -0800
- Subject: Re: Our proposal for a MATDS
- References: <3E63F7E9.7090908@snon.net>
On Tue, Mar 04, 2003 at 01:48:41AM +0100, Stefan Olsson wrote:
> As mentioned erlier on this list
> (http://gcc.gnu.org/ml/libstdc++/2003-02/msg00098.html) we (that's me and
> Ola R?nnerup) have been working on a allocator suitable for MT applications
> for quite some time now.
Have you read Paul McKenney's papers on the subject? He did the
allocator for the Sequent Dynix kernel.
http://www.rdrop.com/users/paulmck/
I expect an order of magnitude or better improvement is available,
particularly if you bypass malloc() and go directly to mmap(), and
use modern page-tagging and RCU primitives. Glibc's malloc would
would benefit from the same kind of attention.
Nathan Myers
ncm-nospam at cantrip dot org