Hi all, I want to check the influence of neighbour list cutoff radii on total (CFD-DEM) memory consumption.
If I grep "Memory usage per processor" from the log file I get the memory used by liggghts, but this value is only a fraction than the value that htop gives me. Any suggestion?
I saw also running ErgunTestMPI that at the end gives a memory histogram per proccessor, this is done by selecting standardClock from couplingProperties file? Because I do that but I don't get the same type of message on my simulation case.
Regards,
Mauyro
j-kerbl | Wed, 05/10/2017 - 16:45
Hi Mauyro,
Hi Mauyro,
memory measurement itself is actually a non-trivial problem. There are different versions available, depending on which type of memory allocated is counted
In CFDEMcoupling the RSS memory (https://en.wikipedia.org/wiki/Resident_set_size) is measured at THE END by enabling the standardClock. The LIGGGHTS memory tracking works only for the liggghts part and works differently.
In htop you can track e.g. M_RESIDENT (RES).
Cheers,
Josef
mbaldini | Wed, 05/10/2017 - 22:14
Hi Josef, thanks for your
Hi Josef, thanks for your answer.
Regards,
Mauro