Date: Tue, 28 Jan 2003 19:46:22 +0100 From: "D. Rock" <D.Rock@t-online.de> To: Dan Nelson <dnelson@allantgroup.com> Cc: Rahul Siddharthan <rsidd@online.fr>, Kenneth Culver <culverk@yumyumyum.org>, Atte Peltomaki <atte.peltomaki@iki.fi>, freebsd-current@FreeBSD.ORG Subject: Re: Performance problems with 5.0-RELEASE Message-ID: <3E36CFFE.9080401@t-online.de> In-Reply-To: <20030123185657.GB54656@dan.emsphone.com> References: <20030123123522.C9068-100000@alpha.yumyumyum.org> <20030123181628.GA2185@papagena.rockefeller.edu> <20030123185657.GB54656@dan.emsphone.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Dan Nelson schrieb: > In the last episode (Jan 23), Rahul Siddharthan said: > >>Kenneth Culver wrote: >> >>>Did you by any chance build your own kernel? If so did you leave >>>things like this in: >>> >>>options INVARIANTS #Enable calls of extra sanity >>>options INVARIANT_SUPPORT #Extra sanity checks of internal >>>options WITNESS #Enable checks to detect deadlocks >>>options WITNESS_SKIPSPIN #Don't run witness on spinlocks >> >>I'd like to add that even with these removed, I was experiencing >>terrible performance in building ports, etc (anything involving heavy >>filesystem activity or memory usage). Setting up an /etc/malloc.conf >>fixed this (this is also briefly noted in UPDATING). Specifically I >>use (don't know whether it's optimal, but it works): >> >># ls -l /etc/malloc.conf >>lrwxr-xr-x 1 root wheel 4 Jan 23 11:52 /etc/malloc.conf -> HR< > > > H and < should only make a difference if you are low on memory. R is on > by default in 5.0 anyway, due to A and J being on by default. Setting > malloc.conf to "aj" makes it work like it does in 4.*. > Ahh, thanks for the info. Just a notice. With malloc.conf pointing to "aj" I got a speedup of over 85% for a specific test program for Perl 5.8.0: To reproduce: cd /usr/ports/lang/perl make cd work/perl-5.8.0 time ./perl t/op/pat.t Results: no /etc/malloc.conf 125 seconds user time /etc/malloc.conf -> aj 18 seconds user time on an AMD K6-2 300. I'd say this is a significant difference... Daniel To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3E36CFFE.9080401>