| 12:14:50 PM PDT - Tue, Mar 14th 2017 |
|
The memory specification is per MPI process. So specifying
memory 1000 mb total
for a calculation running 16 MPI processes/node and 4 nodes would give the code 64,000 MB of memory to work with (16,000 MB/node).
The total memory requirements for TDDFT are going to be on the order of the number of occupied orbitals times the number of virtual orbitals since this is the length of a trial vector. You could sit and figure out approximately how much memory you need based on that length and the expected number of trial vectors (maybe 10-20 times the number of roots), or you could just give the calculation all the memory the system has and let it go. The TDDFT stores most quantities in global arrays, so you can always bias your memory allocation scheme towards global if need be (http://nwchemgit.github.io/index.php/Release66:Top-level#MEMORY).
|