[molpro-user] memory settings in parallel runs

Manuel Dömer manuel.doemer at epfl.ch
Mon Nov 16 13:53:54 GMT 2009


Dear Mikael,

many thanks for your answer!

So is there indeed no possibility for a check-run on the memory  
requirements?
Has anybody else experience with finding the optimal settings for the  
global array memory?

Thanks again and have a nice week!

Manuel

On Nov 12, 2009, at 12:39 PM, Mikael Johansson wrote:

>
> Hello Manuel!
>
> A short answer based on empirical studies here, someone with more  
> insight
> of the inner workings of the program can provide more.
>
> I've never found the need to play around with the global memory  
> option.
> Instead, adjusting the memory per task has always been enough. I  
> normally
> do it with the "memory" keyword in the input file, for example:
> memory,1800,m
>
> This seems to be the most critical for getting jobs done, at least  
> for the
> CCSD(T) calcs which I've mainly used Molpro for. And yes, having fewer
> task with more memory available per node is what you usually will  
> need to
> do, the memory requirements per task do not decrease that much with a
> larger number of total tasks (again, at least for the calcs I've been
> running).
>
> Hope this helps, have a nice day,
>     Mikael J.
>     http://www.iki.fi/~mpjohans/




More information about the Molpro-user mailing list