EPW Forum

Dear developers:

I would like to report two very minor issues as follows:

(1) When iverbosity=2 and nbndfs>10, the cube files with the band index greater than 10 are overwritten due to the formatted output with "i1".
You can find these cases in the files of eliashberg_setup.f90 and eliashberg_write.f90.

(2) Currently, the subroutine "set_ndnmbr" doesn't consider the case using the number of cores greater than 1000. Is there a specific reason to stick to the case using the number of cores small than 1000? I can easily modify the code, but I am just wondering whether there is a part in the code which limits the number of cores.

Sincerely,

Hyungjun Lee

Dear Hyungjun Lee,

Thank you for reporting those two small bugs.

1) Indeed. I modified this and commit it (see Github trunk).

2) This was for historical reasons. I raised it to 99,999 cores (also committed).
Note however that I would not recommend to use much more than 1000 cores at the moment. See benchmark: http://epw.org.uk/Main/Benchmarks

Best,
Samuel

Dear Samuel Ponc:

Thank you for your reply and for giving me useful advice for scalability in EPW.

Sincerely,

Hyungjun Lee