[QE-users] Different SCF energy when # of MPI is changed

[Takahiro Chiba]

Hi all,

I found a discrepancy in total energy and total_magnetization even though I
used the same input, the same pw.x, the same Intel libraries, and the same
host. The only difference is the number of MPI processes.
[image: mail2.png]

Is it expected behavior? If not, can anyone give me advice? Is there any
mistake in my input?


==Input and output files==
https://www.scl.kyoto-u.ac.jp/~tchiba/SCFdiff.tar.gz
==

==QE compilation==
QE 7.1 is prepared by "./configure --enable-openmp --enable-signals
--with-scalapack=intel && make && make install". No issue was found with
the test-suite. Version of Intel Compiler, MKL, and MPI was 2018.5.274.
==QE execution==
mpirun -genv I_MPI_DEBUG 5 \
-env I_MPI_FABRICS shm:tcp -bootstrap sge -env OMP_NUM_THREADS 1 \
pw.x -ni 1 -nk 1 -ndiag 1 -i ${INP} > ${OUT}
==Hardware==
2x Xeon Gold 6330, 256GB DDR4
(2x 28 Cores Ice Lake)
==

==I tried below, but no luck==
1. Different hosts with the same CPU
2. QE 7.0
3. Other pseudo (*.pbe-spn-kjpaw_psl.1.0.0.UPF)
4. More ecutwfc (120, 160 Ry)
==


Sincerely,
Takahiro

---Sender---
Takahiro Chiba
Grad. school of chem. sci. and eng., Hokkaido Univ.
Group:   https://www.cat.hokudai.ac.jp/hasegawa/HOME.html
Mail:      takahiro_chiba at eis.hokudai.ac.jp
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221006/560a59a7/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mail2.png
Type: image/png
Size: 4094 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221006/560a59a7/attachment.png>