schedulers: use num_cores_per_mpiproc in direct

[dev-zero]

No description provided.

[codecov]

Codecov Report

Merging #5051 (87faf08) into develop (78ef633) will decrease coverage by 0.01%.
The diff coverage is 37.50%.

@@             Coverage Diff             @@
##           develop    #5051      +/-   ##
===========================================
- Coverage    80.23%   80.23%   -0.00%     
===========================================
  Files          515      515              
  Lines        36757    36762       +5     
===========================================
+ Hits         29490    29493       +3     
- Misses        7267     7269       +2     

Flag	Coverage Δ
django	74.72% <37.50%> (-<0.01%)	⬇️
sqlalchemy	73.62% <37.50%> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
aiida/schedulers/plugins/direct.py	61.18% <37.50%> (-0.03%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 78ef633...87faf08. Read the comment docs.

[giovannipizzi]

Thanks @dev-zero !

Can you please add one test along the lines of, e.g., this one, in the appropriate file?

If @sphuber has no further comment, then for me it can be merged
(I guess it can be useful also on e.g. Quantum Mobile, related to the issues of the number of OpenMP threads, so pinging @chrisjsewell for info)

[sphuber]

Looks fine to me as well, but I must say that I am not too experienced with this stuff. Is there any environment thinkable where exporting OMP_NUM_THREADS can have an effect that one wouldn't want? Or is it that at worst it does nothing? If so, then it is fine for me to add, after a small test is added.

[sphuber]

I have added tests, but I don't have writing rights to @dev-zero 's fork. To get this merged I pushed this to my fork and created a new PR #5126 Attribution is kept of course. Hope you don't mind