Re: [molpro-user] Problem with parallel mrci

[Andy May]

John,

Can you clarify what you mean by 'stopped'. Did the program actually 
terminate, and if so was there an error or any kind? If it was just a 
lack of activity or a hang I'd be suspicious of low memory/swapping, but 
this seems unlikely given the amount of memory available. Was anything 
else running on the nodes at the time?

If you send me a copy of the CONFIG file, input and output files I'll 
attempt to run with the same compilers etc. on our system.

Best wishes,

Andy

On 29/08/11 14:58, John Travers wrote:
> Dear Molpro users,
>
> I had a strange problem running mrci jobs in parallel. For a test
> calculation by requesting totally 12 CPUs, if I assigned 3 CPUs per
> node, the calculation finished smoothly. But if I requested 4 CPUs per
> node then the calculation stopped at a point when the program started to
> calculate the cluster corrections for the MRCISD energy and never complete.
>
> The Molpro version is 2010.1 patched at the level of 23. I compiled with
> openmpi-1.5.3&intel-12.0. After the compilation, I run the test jobs and
> no error was found. The configuration of the computer node is 8 CPU with
> 32 Gb mem. My test mrci job uses less 2gb mem per CPU.
>
> Any insight?
>
> Thanks
>
> John
>
>
>
>
>
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user