Date: Thu, 05 Apr 2001 19:27:11 +0300 From: Maxim Sobolev <sobomax@FreeBSD.org> To: Poul-Henning Kamp <phk@critter.freebsd.dk>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: ports/devel Makefile ports/devel/dlmalloc Makefile distinfo pkg-comment pkg-descr pkg-message pkg-plist ports/devel/dlmalloc/files Makefile README.FreeBSD Message-ID: <3ACC9CDF.9462C2CA@FreeBSD.org> References: <22779.986485542@critter> <3ACC996A.98218FBE@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Maxim Sobolev wrote: > Poul-Henning Kamp wrote: > > > In message <3ACC91AB.2A183123@FreeBSD.org>, Maxim Sobolev writes: > > > > >> > Moreover, due to not very clear for me reasons Python benchmark (pybench) > > >> > is about 60% (!!!) faster with this allocator comparing to the libc one. > > >> > Obviously we should investigate this further and if there is no error > > >> > then tune either Python or our own malloc. > > >> > > >> Did you measure this with the right malloc flags for the libc version ? > > >> > > >> Remember that -current runs with malloc flags "AJ" as default, that is not > > >> fair if it comes to performance comparisons (but it does wonders for finding > > >> bugs :-) > > > > > >I don't use default value: > > > > > >$ ls -l /etc/malloc.conf > > >lrwxr-xr-x 1 root wheel 7 Jul 22 2000 /etc/malloc.conf@ -> HRajuxz > > > > I think you should use 'r' for performance comparisons... > > Bah, you are as usually right, however even with this settings python+dlmalloc in > -current is about 10-12% faster (only 5% in -stable). And the size argument still holds - memory footprint of my X server with dlmalloc() is about 1.5MB less than with default one (12.5MB vs. 14MB). -Maxim To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3ACC9CDF.9462C2CA>