Re: Stuck CASPT2 Calculation


[ Molcas user's WWWBoard ]

Posted by Steven Vancoillie on July 17, 2013 at 17:08:11:

In Reply to: Stuck CASPT2 Calculation posted by Oscar on June 19, 2013 at 10:55:01:

: Hello all,

: I tried to run CD-CASPT2(12-in-18) calculations with 94M CSFs several times, but the CASPT2 calculation got stuck >1 week (see below). Not work, even using as much MEM as possible. I appreciate for any suggestion.

: End of the output:
: Estimated memory requirements:
: POLY3 : 17149155588
: MKRHS : 54435453
: SIGMA : 82242839
: DIADNS: 107784432
: PRPCTL: 190523505
: Available workspace: 17234021472

: ********************************************************************************
: Single-state initialization phase begins for state 1
: --------------------------------------------------------------------------------

Why do you think it is stuck? I think it's still busy :)
With ~ 100M CSFs, could take a 7-14 days,
it's probably still in the density matrix routines.

To speed things up, you can compile Molcas with a
threaded lapack library, e.g. the Intel MKL.

Also, use ">> export MOLCAS_PRINT=4" in your input,
this will give you a better progress view. But only
put this just before " &CASPT2", otherwise programs
like e.g. seward will dump each byte they can.

grtz
Steven



Follow Ups:



Post a Followup

Name:
E-Mail:

Subject:

if B is 1s22s22p1, what is Li?

Passfield:

Comments:


[ Follow Ups ] [ Post Followup ] [ Molcas user's WWWBoard ]