[molpro-user] Molpro hangs with RASSCF with large active space

[Gerald Knizia]

On Wednesday 02 March 2011 22:14, Leon Freitag wrote:
> On Tuesday 01 March 2011 09:48:22 Peter Knowles wrote:
>> It wouldn't be a surprise to me if this job happened to need a very long
>> time to complete. The implementation of 'restrict' in Molpro's MCSCF is
>> functional but not necessarily optimal, and so hopeful comparison against
>> other software options might not be the best way to estimate feasibility.
>
> thank you for your reply. Normally MULTI prints iteration controls and
> attempts to read the starting orbitals after the CSFs # -- so I was just
> surprised that molpro did not even attempt to do this after such a long
> time -- but if just gets stuck in some inefficient subroutine before
> attempting to read in the initial orbital coefficients, there's not much
> you can do about it, right?

Well, the routine generating the configuration spaces could be fixed, so there 
is something that could be done about it. We have changed parts of it for a 
new MRCI program, and intend to extend and integrate these changes with the 
base version for MCSCF and the other multi-reference programs. But in this 
concrete context this is not really trivial due to program interdependencies 
and special cases, and therefore this currently still has a status as "todo" 
for a future Molpro release, I'm afraid.

The main problem with these kinds of things is that they only show up in 
calculations which take very long even if everything is optimal. This makes 
such problems *very* painful to debug.
-- 
Gerald Knizia