All the mail mirrored from lore.kernel.org
 help / color / mirror / Atom feed
* Re: Good MM benchmarks utilities?
@ 1999-01-04 22:35 Albert D. Cahalan
  0 siblings, 0 replies; only message in thread
From: Albert D. Cahalan @ 1999-01-04 22:35 UTC (permalink / raw
  To: linux-kernel


Steve Bergman writes:

> I've been doing some benchmarking of the 2.2pre kernels and various
> patches in development.  I am looking for memory management benchmark
> utilities.  Something like Bonnie for vm subsystem benchmarking.
> I'm concerned about the comprehensiveness of my own simple tests.
> What I'm finding is that while 2.2 is outperforming 2.0.36 in some
> situations, it is falling far behind in others.  Particularly the very
> low memory case (6M with X running) 2.0.36 performs far better (Even
> though the overall 'available memory' as reported by top is actually
> slightly *greater* for 2.2.  Andrea Arcangeli's recent patches have
> improved things a great deal but I still haven't found anything to
> beat 2.0.36 in very low memory.  It would help to have a suite that
> gives a more comprehensive view of mm performance.
> 
> Also, no matter how good the suite is, it will only tell me about the
> performance of my particular hardware layout.  The more people who
> test and submit results, the better 2.2/2.3 mm will be.  

If gcc could be IO-bound, you'd have a great benchmark.
You could get a 450 MHZ Xeon or 666 MHz Alpha with old IDE drives,
or slow down your IO in software to adjust for a normal CPU.

There might be a place in the block device code where you could add
a nice delay. You could get the NFS swap patches and use a PPP link
over a 14.4 modem. I suspect that would make gcc highly IO-bound.

If you had the kernel log user IO operations and page faults, you
could have a benchmark play back gcc IO behavior without all the
normal CPU usage.



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~1999-01-05  9:38 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
1999-01-04 22:35 Good MM benchmarks utilities? Albert D. Cahalan

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.