Hello,
How to estimate the memory usage of epw with etf_mem=0? How many times it should be bigger than the memory usage of epw with etf_mem=1? Is the memory usage should be the summation of prefix.epmatwp*?
I changed one of my calculation from eft_mem=1 to etf_mem=0, and the output ends with "=BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES" and "EXIT CODE: 9". I doubt that process ran out of memory.
Thanks.
Best regards,
Yuhao Gu
How to estimate the memory usage of epw with etf_mem=0
Moderator: stiwari
-
fabiocosta4
Re: How to estimate the memory usage of epw with etf_mem=0
Dear EPW community
Guyuhao's question was acctually very similar to one I have, and I would like to add to it. In my case, It seems I'm having some memory issues, as the code crashes with an mpi abort message, and no error code from EPW is showed in the output file. I'm trying to interpolate matrices calculated with 12 12 12 kpoints and 6 6 6 qpoins, to a grid of 10q-20k, 20q-40k, and so on. I tryed many combinations of etf_mem (=1 and =2), max_memlt, but only the 10q-20k runs without crashing. To add to Guyhao's question, I also wanted to know how to estimate the memory requirement, so I can determine if there is anything wrong with my calculations, or if it's the case where the cluster where I'm running the calculations dont have enough memory to complete the task.
Thanks for the assistance
Fbio Costa
Guyuhao's question was acctually very similar to one I have, and I would like to add to it. In my case, It seems I'm having some memory issues, as the code crashes with an mpi abort message, and no error code from EPW is showed in the output file. I'm trying to interpolate matrices calculated with 12 12 12 kpoints and 6 6 6 qpoins, to a grid of 10q-20k, 20q-40k, and so on. I tryed many combinations of etf_mem (=1 and =2), max_memlt, but only the 10q-20k runs without crashing. To add to Guyhao's question, I also wanted to know how to estimate the memory requirement, so I can determine if there is anything wrong with my calculations, or if it's the case where the cluster where I'm running the calculations dont have enough memory to complete the task.
Thanks for the assistance
Fbio Costa