Subject: Re: Thread benchmarks
To: None <darrenr@NetBSD.org>
From: M. Warner Losh <firstname.lastname@example.org>
Date: 10/02/2007 03:46:14
In message: <47020C5F.email@example.com>
Darren Reed <darrenr@NetBSD.org> writes:
: Kris Kennaway wrote:
: > Andrew Doran wrote:
: >>> In tests that have been run on p4 hardware, the FreeBSD system's graph
: >>> looks more like NetBSD's than the one presented here. FreeBSD's kernel
: >>> has a lot of debugging options that hurt performance on by default.
: >>> Also,
: >>> FreeBSD's malloc defaults to 'AJ' in head, which would result in reduced
: >>> performance.
: >> I can try turning off debugging in the allocator. What else would you
: >> like
: >> me to try? I would like to provide remote access to the two systems but
: >> unfortunatley my Internet link is unreliable and I'm not in a position to
: >> leave them on 24x7. Some details on the test. I grabbed my.cnf from Jeff
: >> Roberson's weblog:
: > You should rebuild malloc with MALLOC_PRODUCTION defined (edit
: > lib/libc/stdlib/malloc.c) as well as making sure that either
: > /etc/malloc.conf is removed or symlinked to 'aj'. This is pretty
: > important.
: When does this get turned on for normal FreeBSD builds?
: Just those that are "releases" (vs current)?
Yes. -HEAD has that turned off so that we maximum sanity testing
during development cycles. After we branch, one of the things done on
the branch before a release is to turn off all the performance
degrading debugging/sanity code. Once off on a branch, it stays off
for the life of the branch.