Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 24 Jul 2001 12:32:29 -0700
From:      Kent Stewart <kstewart@urx.com>
To:        mikea <mikea@mikea.ath.cx>
Cc:        freebsd-stable@FreeBSD.ORG
Subject:   Re: question about wc on for NON-critical workstation (no flamebait)
Message-ID:  <3B5DCD4D.9B0BE80F@urx.com>
References:  <20010724180907.A71575@dogma.freebsd-uk.eu.org> <3B5DAECF.E9FE622D@mitre.org> <3B5DB1FA.1B894F9E@urx.com> <20010724184624.B71800@dogma.freebsd-uk.eu.org> <20010724141346.A5878@mikea.ath.cx>

next in thread | previous in thread | raw e-mail | index | archive | help


mikea wrote:
> 
> On Tue, Jul 24, 2001 at 06:46:24PM +0100, j mckitrick wrote:
> > On Tue, Jul 24, 2001 at 10:35:54AM -0700, Kent Stewart wrote:
> > | My buildworld on a dual 866 coppermine system went from 42 minutes to 29
> > | minutes. The other side effect was the -j8 parameter finally did
> > | something. Before that anything from -j2 on, actually made the
> >
> > I thought over -j2 was not recommended, and often broke?
> 
> The recommendation I've seen is two jobs per processor. In Kent's
> case and mine, with dual processors, we could run with -j4 and
> not exceed the recommendation. Im practice, I find that -j8 (four
> jobs per processor, on average) works very handily in almost all
> cases.

Where I got into disagreement was the Handbook. It stated -j4 was good
for single cpus and -j6 to -j10 for multi-processor systems. I never saw
that and started refering to it as a "FreeBSD Urban Legend", i.e.,
something that only worked on someone elses system.

> 
> >From time to time, a job will require a module that hasn't been
> built then by another job running in parallel, but a rerun of
> the make generally succeeds.
> 
> > | buildworld run longer. I think the cpu's were starved for I/O. The
> > | system is built around 3-ATA-100 Maxtor 30GB HD's. The motherboard is a
> > | VP6 and each HD is on its own controller. Using raid-0 also slowed the
> > | compile down.
> >
> > A far cry from my little laptop.  ;-)
> 
> My box isn't quite the horse that Kent's is, but I have noticed
> a significant speed increase with -j8 over -j4, even with two
> 30GB ATA-66 HDs.
> 
> I think I'll turn on write caching and live dangerously for a
> while.

This system is basically a test system. Occasionally I see a message
that -stable is broken. I want a fast build. If I am around, I can see
if your failure is real. If my system builds, it was your error or you
need to recvsup. If my build fails, than we have two systems failing. My
system is pretty simple and a failure on it could be a real problem or
we both need to recvsup. It is built often enough and I have an HTML
version of all of the recent source changes. It is usually pretty simple
to point a finger in the right direction. I just can't pass the pointy
hat :).

BTW, I didn't know they called them sysadmins in 1964. There were some
systems in the late 60's that I was challenged to break. I was given the
procedure to access system functions that could panic a system and told
that wasn't fair.

Kent

-- 
Kent Stewart
Richland, WA

Cool site
http://www.bmwfilms.com

mailto:kbstew99@hotmail.com
http://kstewart.urx.com/kstewart/index.html
http://daily.daemonnews.org/

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3B5DCD4D.9B0BE80F>