Page 1 of 1

magnetic job hang with vasp/6.1 pgi

Posted: Thu Sep 03, 2020 9:36 am
by mariella_ippolito1
Dear all,
I have a problem with magnetic calculations that hang using vasp/6.1.1 compiled with pgi: the calculations fail on both power9 architecture and x86 architecture. They fail also running on CPUs only.
The same jobs run properly using vasp compiled with intel or gnu/8.4.0, even using the gpu version on Power9 (gpu_ncl).
The problem seams therefore related to pgi compiler.
I attach the tar.gz with the input and output files.
I have already reported the problem to Nvidia.

Thanks,
Mariella Ippolito


--
PhD Mariella Ippolito
CINECA SuperComputing Applications and Innovation Department - SCAI
via dei Tizii 6, 00185 Roma - ITALY
Phone: +39 0644486719
Skype: mariellaippolito

Re: magnetic job hang with vasp/6.1 pgi

Posted: Fri Sep 18, 2020 8:28 am
by merzuk.kaltak
This issue has been reported to PGI. This thread will be updated as soon as we have a solution to this problem.
Thank you for report this issue.

Re: magnetic job hang with vasp/6.1 pgi

Posted: Tue Sep 22, 2020 11:02 am
by mariella_ippolito1
Dear Vasp support,
these calculations that hang with vasp/6.1.1 compiled with pgi19 run if compiling vasp using hpc-sdk.
However some magnetic calculations that hang with vasp compiled with pgi19, even if they run using vasp compiled with hpc-sdk they produce wrong results (non converging energies).
I think it is better to open a new issue for this problem.

Regards,
Mariella