Posted by Valera Veryazov on June 18, 2009 at 12:21:51:
In Reply to: Molcas parallel implentation. Some doubts posted by Javier on June 18, 2009 at 11:57:38:
: I have some doubts about running Molcas in parallel mode:
If you use RI/CD integrals, the construction of Fock matrix in RASSCF is parallelized.
But CI part is not.
So, if you have a very large basis set and reasonable CI - yes, there is a perfect
sense to run molcas in parallel. If you have a small basis set and large active space - running in parallel will not improve the performance.
CASPT2 is not parallel code (yet). However, if you use numerical gradients with CASPT2 -
it can (and should) run in parallel - since all points are computed separately.
: In SEWARD, can the Cholesky decomposition affect the performance of a parallel running?
construction of RI/CD vectors in seward runs in parallel.