"Fulton CR (Charles)" wrote:
>
> oops hit the send button too quick.
>
> It would be nice to have a "design set" to test out different configurations
> with a unified set of "stress tests". e.g. everyone that can, running the
> tests with and w/o apic, nosmp, diff bios.., etc, etc. And then also have a
> formalized way of reporting back your test data, using xml or something ;)
That sounds good. There is a stress test list at
http://www.bitwizard.nl/cgi-bin/abit but I don't find it very easy to
use. IMHO there should be some kind of multidimensional significance
thing...
My stress test currently focuses on hard drive throughput combined with
- with "something". I can't really find out what that "something" is...
I did all the "this really helps" things (flashed the bios with the
latest QQ version, cleared the CMOS) and then built 2 kernels: an SMP
and a non SMP.
My setup has the HPT366 active, and a 100Mbit card. However, I don't
think this really counts - the setups seem quite different, you even
have people with SCSI cards and hangs.
The "stress test" used to be:
mount /dev/hde1 /mnt/hd0
mount /dev/hde3 /mnt/hd1
cp -av /mnt/hd0/* /mnt/hd1
... then try "some other things".
Now I figured out that:
cat /dev/hde > /dev/null & ping -f other-machine
works just as well. You can do this in single user mode if you like.
This will hang my machine within 30 seconds.
I don't know why. Tried a kernel with sysrq builtin, but I cannot find
out where it goes wrong.
The "other" tests are about the SMP thing. If I start an SMP kernel with
the "nosmp" option, it just hangs without any effort, and the strangest
thing is that this seems IRQ related (so: first mouse and keyboard
really hang, but the network is reacheable; in fact, I could log in and
kill X! Then again, trying kbdrate or disconnect/connect keyboard made
no difference.
Maybe this helps - I'd be glad to hear your results.
Valentijn
--
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
