Skip site navigation (1)Skip section navigation (2)
Date:      21 Feb 2002 11:42:27 -0600
From:      Bob Van Valzah <Bob@Talarian.Com>
To:        John Baldwin <jhb@FreeBSD.org>
Cc:        smp@FreeBSD.org
Subject:   RE: Volunteering: Old fart with old CPUs
Message-ID:  <1014313347.1397.32.camel@NewStorm.WhiteBarn.Com>
In-Reply-To: <XFMail.020220172739.jhb@FreeBSD.org>
References:  <XFMail.020220172739.jhb@FreeBSD.org>

next in thread | previous in thread | raw e-mail | index | archive | help
John, Thanks for your reply. I was beginning to think that nobody read
the SMP alias :-)

The box I have available for SMP testing is running -STABLE now, but it
has enough space that I can also install -CURRENT (and Linux or
whatever).

I'm building a -STABLE kernel with config -p now and will take it for a
spin tonight so that I'm up-to-speed on what to expect from a profiled
kernel. Then I'll try the same on -CURRENT and look for differences.
This should help to "quantify" any breakage. Is this the place to report
my findings?

Could you please suggest something to generate a good system call load
for a profiled kernel? Maybe a -j4 buildworld of 4.5-RELEASE? Maybe an
Apache build? Lemmie know.

	Thanks,

	Bob



On Wed, 2002-02-20 at 16:27, John Baldwin wrote:
> 
> On 15-Feb-02 Bob Van Valzah wrote:
> > I just got back from BSDCon where I heard several folks working on SMPng 
> > asking for volunteers.
> > 
> > My "day job" prevents me doing things like writing code or doing serious 
> > debugging, but I think I might still be able to help the SMPng cause. (I 
> > actually did one of the early MP Unix implementations on Gould hardware 
> > back in 1983, but I haven't kept current.) So I guess you could say that 
> > I understand the problem space and I'm excited about what you're trying 
> > to do for 5.0, but I'm rusty.
> > 
> > I haven't written a line of kernel code in over 10 years, so I'm 
> > thinking I might be most helpful on testing, benchmarking, or profiling 
> > work. I'm able to use ddb to help you figure out what's going on, but I 
> > probably can't help you figure out how to fix it. I'm not likely to have 
> > big chunks of time to work on SMPng, but I can probably muster a steady 
> > stream of small blocks of time.
> > 
> > I have an old dual PII-233 box sitting around and I'd be glad to put it 
> > to work if that'd help in some way. (My experience in the past has been 
> > that older, slower machines sometimes shake out latent bugs that you 
> > don't hit on modern screamers.)
> > 
> > I could probably do something like regular builds from -CURRENT and run 
> > various benchmarks or torture tests. I'd be happy to run regular 
> > benchmarks against 3.x, 4.x, other BSDs, Linux or whatever.
> 
> Testing is a big help, yes. :)
> 
> > Also, I think I could do a decent job of SMP "data mining" if that'd 
> > help. For example, I might be able to cobble together some tools to 
> > analyze lock counters or other statistics.
> > 
> > I see "profiling is broken" on the known issues list. I might be able to 
> > tackle that, but don't want to bite off more that I can chew--esp. not 
> > for my first attempt. How badly broken is it?
> 
> Actually, I'm not exactly sure how broken it really is.  If you could figure
> out how kernel profiling works and then figure out if it is actually broken or
> not, that would be great. :)
> 
> > So I'm leaving myself open for suggestions here. Please let me know 
> > what'd be most helpful. I'll try to pitch in where I can.
> > 
> >     Thanks,
> > 
> >     Bob
> 
> -- 
> 
> John Baldwin <jhb@FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
> "Power Users Use the Power to Serve!"  -  http://www.FreeBSD.org/



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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1014313347.1397.32.camel>