[gmx-users] Poor stupid Athlons...
lindahl at csb.stanford.edu
Tue Oct 7 03:43:01 CEST 2003
(Un)fortunately I came to the same conclusion after weeks of testing,
1. I cannot find any bug in the Gromacs SSE code (which is good news I
guess :-). I've gone through line by line several times, so this can
more or less be ruled out as an explanating I think...
2. Since it works great on Pentium, the next step would be to at least
isolate where the crash happens on Athlons. Even this doesn't work -
when the code is changed (debug info, etc), the crash happens somewhere
else or not at all.
3. With 3DNow everythings works great.
I'll try to get in touch with AMD about this since it has all the signs
of bad hardware, so keep your executables that crash - I'm particularly
interested in reproducible crashes that take the OS and "everything"
down (since that simply shouldn't be possible...)
On Monday, October 6, 2003, at 06:22 PM, Justin MacCallum wrote:
> Hi all,
> we have been having problems with our Athlon machines (dual MP2200+'s)
> crashing roughly once every several weeks. As other people have
> reported, the crash brings down the entire machine, OS and all,
> requiring a hard reset.
> We have tried everything to get them to work, including moving them to
> different building that has a better machine room!
> On Erik's advice, I recompiled gromacs with a hacked version of
> detectcpu.c. I edited the file to skip the check for SSE, forcing the
> use of 3DNow.
> I'm happy to report that after a little more than 2 weeks of solid use,
> we have not had a single crash on 16 machines. Normally I would have
> expected a least 2 or 3 machines to be down.
> I'll report back in a couple more weeks, but so far it looks like this
> fix works (for us at least).
> gmx-users mailing list
> gmx-users at gromacs.org
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
More information about the gromacs.org_gmx-users