[CP2K:4952] Re: too many steps in some test jobs

[Jörg Saßmannshausen]

Hi Matt,

thanks for the feedback. I was just about to report back to the list.
I have compiled ATLAS 3.10.1 and the latest SCALAPACK and that is working now. 
Thus, there seems to be something funny going on in the MKLs I was using. I 
will try and confirm that once I got a bit more time on my test cluster, rather 
then the production cluster. 
So now I need to build the libsmm against that ATLAS version and run the 
testjobs again.

All the best from a damp London (will that ever change??)

Jörg

On Tuesday 18 Feb 2014 14:33:55 Matt W wrote:
> > Hi Jorg,
> 
> nothing much to add except a couple observations as nothing else has been
> suggested:
> 
> I think libxc should only be used  in regtest-libxc, so if you're failing
> elsewhere probably not the culprit
> gfortran 4.6.? is pretty well tested with CP2K, as you heard from Joost a
> week or two ago, so probably not the culprit
> 
> So, I'd guess either MKL, or maybe (hopefully not) MPI libraries. Is this
> Legion?
> 
> Matt

-- 
*************************************************************
Jörg Saßmannshausen
University College London
Department of Chemistry
Gordon Street
London
WC1H 0AJ 

email: j.sassma... at ucl.ac.uk
web: http://sassy.formativ.net

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html
-------------- next part --------------
Q29udGVudC1UeXBlOiBhcHBsaWNhdGlvbi9wZ3Atc2lnbmF0dXJlOyBuYW1lPSJzaWduYXR1cmUu
YXNjIg0KQ29udGVudC1EZXNjcmlwdGlvbjogVGhpcyBpcyBhIGRpZ2l0YWxseSBzaWduZWQgbWVz
c2FnZSBwYXJ0Lg0KDQotLS0tLUJFR0lOIFBHUCBTSUdOQVRVUkUtLS0tLQ0KVmVyc2lvbjogR251
UEcgdjEuNC4xMiAoR05VL0xpbnV4KQ0KDQppRVlFQUJFQ0FBWUZBbE1EY1lFQUNna1FFSTFhS1Br
OHZBdDV4Z0NnZ3g5WUJNclJlNG5XZVFFdzRlVkpmOTI5DQpldnNBb0wxa2xaL0xQcWN0ZjUyanZ3
eFNkQXYvT1BYUg0KPUFEclUNCi0tLS0tRU5EIFBHUCBTSUdOQVRVUkUtLS0tLQ0K