How to run aoforce in parallel?

[Tiuri]

Hey,

I got one question that is hopefully not too stupid.

On our cluster we managed to get TM running in parallel, using "ridft_mpi" instead of "ridft".
Geometry optimizations with "jobex" run just fine, but how do I get aoforce to run in parallel (on 2 cpus)?

"nohup aoforce > force.out &" --> aoforce running fine, but on one cpu
"nohup aoforce -np2 > force.out &" --> still one cpu (jobex runs fine with the -np command, why not aoforce)
"nohup aoforce -n 2 > force.out &" --> same
"nohup aoforce_mpi -np 2 > force.out" --> "nohup: cannot run command `aoforce_mpi': Permission denied" is what i get

I'm probably just stupid, can you help me?
Thanks a lot in advance!

[antti_karttunen]

Hi,

the answer to your question is (unfortunately) simpler than you might have thought : aoforce cannot be run in parallel in the current release version of Turbomole. But maybe the Turbomole crew can shed light on the future developments of aoforce.

On the other hand, you can parallelize NumForce calculations, but then you lose the speed advantage of analytic frequency calculations for HF/DFT...

[Tiuri]

Thanks for your help

Well then, I'll do it on one cpu...

I was thinking about using SNF (http://www.theochem.ethz.ch/software/snf), especially for doing Raman calculations. Hope that is a good idea

Have a nice weekend!

[turbomaster]

Hi,

parallelization of aoforce is underway, but its hard to say at the moment how
long this way is ...