[CP2K-user] [CP2K:11401] Convergence difference with different number of pocessors for an input

[Krack Matthias (PSI)]

Hi

The output files from the water benchmark which you attached show that you have an invalid CP2K installation (wrong results). You need first to fix your CP2K installation and then validate it using the regression tests, before you start your project runs.

Matthias

From: cp... at googlegroups.com <cp... at googlegroups.com> On Behalf Of Mohammad Shakiba
Sent: Sonntag, 10. März 2019 21:24
To: cp2k <cp... at googlegroups.com>
Subject: Re: [CP2K:11401] Convergence difference with different number of pocessors for an input

Thanks for fixing the input problems. I changed the PSOLVER to MT and add PERIODIC NONE in the CELL section. Its convergence was better when started but does not change from 0.01 in further steps.

I tried that with a test input of cp2k in the 'test' folder ('/test/QS/benchmark/H2O-64.inp') with 32, 16 and 4 number of cores with OMP_NUM_THREADS=1. The results are attached. The input with 32 number of cores diverges and stops. With the other two also it does not converge for the time I put them for a short run.

I have 88 CPU cores on one node. Do you think the problem is with compiling? I used the method in https://xconfigure.readthedocs.io/en/latest/cp2k/README/ with

source /opt/intel/compilers_and_libraries_2017.1.132/linux/bin/compilervars.sh intel64

and libxsmm-1.9 since I could not install that with ./install_cp2k_toolchain --with-gcc=install (there was gcc 4.4 installed on the cluster but with ifort 17.0.1). I also used both cp2k 6.1 and cp2k 7.0 with 'psmp' and 'popt' versions but it did not change.
Besides all the problems above, I have one more question: does this happen when we use a high number of processors with cp2k or with a higher the number of processors we always get a better performance?

Thanks.
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
To post to this group, send email to cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190311/ec13e2ea/attachment.htm>