Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 3 Aug 2000 16:46:32 +0930
From:      Greg Lehey <grog@lemis.com>
To:        FreeBSD SMP list <FreeBSD-smp@FreeBSD.org>
Subject:   Functional interrupt-thread based SMPng code is now available
Message-ID:  <20000803164632.B30009@wantadilla.lemis.com>

next in thread | raw e-mail | index | archive | help
I now have interrupt threads running here on a single processor Intel
machine.  Diffs against today's -CURRENT are at
http://people.freebsd.org/~grog/patches4.gz.

This code still needs a lot of work, of course.  In particular,

1.  There's no alpha MD support.
2.  There's no Intel SMP support.
3.  There are various minor things which need fixing.  For example:
    - CPU time usage doesn't work correctly: vmstat and top both show
      100% system CPU usage, no user time, no idle.  ps shows that the
      rtc interrupt (irq8) is using between 80% and 100% of CPU time,
      and the interrupt processes don't accumulate CPU time.  I'll
      look at these in the near future.
    - We had some cases where a looping process would cause the system
      to go catatonic, presumably because the interrupt thread wasn't
      getting scheduled.  It looks as if that is OK now, but it hasn't
      been tested very heavily.

The good news is that it seems stable enough, and performance isn't
nearly as bad as I feared.  I ran buildworlds in a loop for 14 hours
from last night, and I didn't have any stability problems.  I'm
currently running a buildworld with a GENERIC kernel, and I'll then
compare the time with the SMPng kernel.

Greg
--
Finger grog@lemis.com for PGP public key
See complete headers for address and phone numbers


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?20000803164632.B30009>