Date: Wed, 29 Mar 2000 22:00:41 -0800 (PST) From: Matthew Dillon <dillon@apollo.backplane.com> To: Andy Farkas <andyf@speednet.com.au> Cc: freebsd-current@FreeBSD.ORG Subject: Re: SMP buildworld times / performance tests Message-ID: <200003300600.WAA68006@apollo.backplane.com> References: <Pine.BSF.4.10.10003301158360.14426-100000@backup.af.speednet.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
:> time make -j 20 buildworld build FreeBSD-current using 4.0 kernel
:>
:> 4745.607u 1673.646s 1:29:07.45 120.0% 1323+1599k 8237+251565io 1615pf+0w
:>
:> time make -j 20 buildworld build FreeBSD-current using 5.0 kernel
:>
:> 4696.987u 1502.278s 1:10:34.17 146.4% 1359+1641k 10889+4270io 1779pf+0w
:
:Can I ask why is there a huge difference in the number of io (251k vs 4k)?
:What is so different between 4.0 and 5.0 that causes this?
That is very odd. I'm using the same make.conf on both machines
but even if they weren't the same 19 minutes should not make that sort
of difference in I/O statistics.
There are two possibilities: Either 5.0 is doing something very
different in regards to I/O, or I have another patch in 5.0 that
is causing the difference.
I do have one other patch in 5.0 that could be causing the difference.
I added the sequential heuristic code that read() is using to write()
to determine whether to push out the cluster or not. I'm using the
heuristic to 'detect' non-sequential behavior (aka the DBM random I/O
test that was bantied about in an earlier thread which was tripping
over cluster pushouts). I discounted it before because I figured
that compiling would almost always be doing sequential writes anyway
and thus result in the same behavior. Maybe I'm wrong.
I am going to disable the patch in the 5.0 test to see if that accounts
for the reduced I/O. If so then I guess I get to still claim credit,
but it will have been due to the sequential write heuristic instead of
the SMP stuff :-)
-Matt
Matthew Dillon
<dillon@backplane.com>
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?200003300600.WAA68006>