large water system calculation with cp2k

[Phil G.]

Dear Samuel,

Ok, I will change the cp2k.popt to cp2k.psmp. 

Do you mean that I have to delete the line PURIFICATION_METHOD TRS4 or 
should I write anything different? What is the best choice to use LS_SCF 
method to simulate adsorption of water molecules on solid surface?

Kind regards,

Phil 

Am Donnerstag, 15. September 2016 13:06:23 UTC+2 schrieb Samuel Andermatt:
>
> You will need cp2k.psmp instead of cp2k.popt for the threaded version. I 
> would say cpu's per task should be equal to OMP_NUM_TREADS and 
> OMP_NUM_THREADS times tasks-per-node should be equal to the total number of 
> cores of a node.
>
> About the input file, there is no reason to specify trs4 as optimization 
> method and then include a curvy-steps section. In general curvy-steps might 
> be a little slower but more stable. MATRIX_CLUSTER_TYPE and 
> S_PRECONDITIONER, should be atomic, the molecular cluster type only helps 
> sometimes if GPU's are used for the calculation.
>
> Samuel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160915/17556bcc/attachment.htm>