[molpro-user] Parallel codes

H. -J. Werner werner at theochem.uni-stuttgart.de
Wed Jul 5 13:25:12 BST 2006

In Molpro both mrci and ccsd(t) are fully parallelized. caspt2
is not yet parallelized. In multi the determinant ci is parallel 
(which takes most of the time in casscf calculations with large 
active spaces). Also parallel are the integrals, hf, gradients, 
analytical second derivatives (McKinley), most of the
direct codes, df-hf, and df-lmp2. df-lccsd is currently only
partly parallelized, but the local (t) correction is already
fully parallel. Naturally, the performance depends
on the communication speed, best performance is obtained
on shared memory machines.
Best regards
Joachim Werner

On Mi, 05 Jul 2006, Pawel Wielgus wrote:

>Dear Molpro Users,
>just a small off topic... for several days I was fighting with parallel
>compilation of Molcas on IBM machine, and when I did it I realized that
>... MRCI code is not parallelized, neither CASPT2, and RASSCF only
>partially. Efficiently parallelized are Seward, Scf, McKinley, Alaska,
>So how does this look like with Molpro ? Especially with MULTI, MRCI and
>CCSD ? Are they really parallel codes ?
>Have a nice time
>Pawel Wielgus

Prof. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart, Germany
Tel.: (0049) 711 / 685 64400
Fax.: (0049) 711 / 685 64442
e-mail: werner at theochem.uni-stuttgart.de

More information about the Molpro-user mailing list