[QE-users] about the ph.x serial (MPI) execution
Gang Yang
gang.yang at ec-lyon.fr
Mon Dec 13 13:20:46 CET 2021
Dear QE team and users:
I am doing phonon dynamics calculation using QE V6.8 in my personal laptop. I installed openmpi 4.1.1 and QE V6.8(QE based on intel compiler using intel mkl library) .
after PWscf calculation, when i try to carry on " mpirun np 6 ph.x<AAA.ph.in>xxx.ph.out" . i found the ph.x calculation is very slowly for 10 atoms. it needs few days in my personal laptop (6 core i7, 128G RAM)
i doubt that it was running np copies of the same process.
meanwhile. i found that if i simply use " pw.x -i AAA.in> AAA.out " instead of " mpirun np 6 pw.x<AAA.in> AAA.out. the calculation speed is very fast in my personal laptop.
but if I using ph.x -i or just ph.x<AAA.ph.in>AAA.ph.out instead " mpirun np 6 ph.x<AAA.ph.in>xxx.ph.out"
some error will happen after a succeful Pwscf calculation. ( the prefix and outdir is same for the pw.x and ph.x input file)
So, my question is that does ph.x support serical (MPI) running? how can I speed up ph.x computing?
Thank you very much.
Yang Gang
Institut des Nanotechnologies de Lyon (INL) - UMR5270
Ecole Centrale de Lyon
36, Avenue Guy de Collongue
69 134 Ecully Cedex
Tél.: +33 (0)7 64 83 01 70
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20211213/b01e7d08/attachment.html>
More information about the users
mailing list