Ceilidh ... Re^7: Parallel geometry optimization with XMC-QDPT2

Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

We are NATO-free zone

Re^7: Parallel geometry optimization with XMC-QDPT2

On Wed Dec 28 '16 10:55pm, Dawid wrote
--------------------------------------
>Thank you again. I guess that this time the calculations have finally
>accelerated to a decent rate. Anyway, one more question. When I
>grep the output file for DISPL, I get this:
>
>
> COMPUTING ENERGY WITH NO DISPLACEMENTS APPLIED
> COMPUTING ENERGY FOR DISPLACEMENT # 28
> COMPUTING ENERGY FOR DISPLACEMENT # 51
> COMPUTING ENERGY FOR DISPLACEMENT # 80
> COMPUTING ENERGY FOR DISPLACEMENT # 98
> COMPUTING ENERGY FOR DISPLACEMENT # 124
> COMPUTING ENERGY WITH NO DISPLACEMENTS APPLIED
> COMPUTING ENERGY FOR DISPLACEMENT # 43
>
>
>Should I expect that? Why don't I see computations for all other
>displacements? Obviously, they must be calculated, right?

>Dawid

Dear Dawid,

There is nothing wrong with your calculations.
Displacements are distributed between all xp groups, but only grand
master process (this process always belongs to the first xp group)
is allowed to write to the output file. Hence output file contains
a log of work performed by the first xp group only.

By the way, do you know that MCQDPT and XMCQDPT code is capable
to run using multithreading? For optimal performance, each xp group
must be run in smppar mode. Do you use these features?

All the best,
Alex Granovsky

Thu Dec 29 '16 1:31pm

This message read 476 times