Model fails quickly with segmentation fault

[Faraz H]

Upon more investigation it seems the segmentation fault depends on the 
number of mpiranks and OMP_NUM_THREADS. As I play with those number the 
simulation progresses to different stages. But it still eventually fails .

Does this suggest some openmpi bug? Or more likely a bug with cp2k?

I also found the same problem when running the regression tests. Certain 
test models would fail depending on the values I chose for mpiranks and 
OMP_NUM_THREADS. Is there some guideline how to best choose those? My 
machines have 20 cpus each.

On Tuesday, September 19, 2017 at 10:56:53 AM UTC-4, Faraz H wrote:
>
> I am helping a user debug why his cp2k model fails in 4.1 with 
> segmentation fault. But the same model runs fine in version 2.6. We built 
> cp2k from source on Red Hat 6.5 with Intel x86_64 architecture. 
>
> Is this a model I can send to cp2k developers to see if they can reproduce 
> the issue? Or  any suggestions how to debug the segmentation fault?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20171005/7ca9170b/attachment.htm>