12:34:32 AM PST - Tue, Nov 5th 2013 |
|
Thank you, Huub.
I will try it - 500 mw for global and 100 mw spread between heap and stack.
Also, maybe I didn't clearly described my system - it isn't 14-core processor (unfortunately), but only one 4-core processor, so I allow NWChem to use 6 Gb per process.
And another one question, can this problem be connected with MPI? I propose it due to some ARMCI offset warnings which appear only for 3 processes from 4, such as "WARNING: armci_set_mem_offset: offset changed". Or they are also connected with global memory problems?
Upd. It really works, iterative process began. Many thanks!
|
|