[gmx-users] Segmentation Violation

[Giacomo Fiorin]

On 13 Mar 2003, David van der Spoel wrote:

> Could it be the problem recently noted on the list that you have no
> water on the first processor anymore? Is it the first processor that
> crashes actually?

Well, I think this could not be the point, since 16 procs was the 
same "magic" number of procs after which my jobs crashed too, 
systematically.

One thing I think interesting for all the list is that I recompiled
FFTW (I use PME) with the debug option turned on: I saw one error
message for EACH task.  I do not remember exactly which, I was simply
thinking of an error on the network, which at that time was working
really bad for other reasons.

My system is composed by 2260 protein atoms and 32000 total, hence the
first node should have indeed no waters: but I cannot explain the
several error messages.

At the moment I'm outside, when I'm back I will give you more details
privately. I'm just a beginner in parallel debugging, but if I can be
useful the same please ask: I am really interested in getting it
working.

Thank you (and good luck)
Giacomo