[Pw_forum] projwfc.x issue crashes in large scale cases

[Martin Gmitra]

Dear QE users,

I am experiencing projwfc.x job crash on IBM LoadLeveler system for large
cases, over thousand of cores. Other codes pw.x, bands.x work fine (qe6.1).

Sysadmins have confirmed that my projwfc.x job run out of memory affecting
system processes on some nodes resulting in their crash.

Is there a way around for the projwfc.x to use less memory by playing with
parallelization options?
I have used for SCF: -npool 18 -ntg 4 -ndiag 32, and wf_collect = .true.
Is there a way to get estimation of used RAM?

Thanks in advance,
Martin Gmitra
Uni Regensburg, Germany
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20171121/5f405543/attachment.html>