[RDanks]

Hello Foamers,

I'm hoping someone here can help answer this question regarding how the decomposition process works in decomposePar.

Basically, I want to know if I run decomposePar on the same geometry with the same settings, will the mesh always be broken up the exact same way. Or will points on "the the borderline" between two processors not be consistantly mapped to the same processor.

I've done some basic testing with hierarchical decomp and comparing the processor's points file and they look to be the same. But I'm wondering if this is true in the general case?

If you're wondering why I'm asking this, I'm running an exterior simulation from multiple wind directions where each direction is a separate case and I want to know if I can just copy field files between the different cases without reconstructing. I realize that mapFields would do this for me, but in my previous experience mapFields with parallel sources and destinations doesn't work quite right.

Thanks!

[olesen]

Originally Posted by RDanks

Hello Foamers,

I'm hoping someone here can help answer this question regarding how the decomposition process works in decomposePar.

Basically, I want to know if I run decomposePar on the same geometry with the same settings, will the mesh always be broken up the exact same way. Or will points on "the the borderline" between two processors not be consistantly mapped to the same processor.

I've done some basic testing with hierarchical decomp and comparing the processor's points file and they look to be the same. But I'm wondering if this is true in the general case?

If you're wondering why I'm asking this, I'm running an exterior simulation from multiple wind directions where each direction is a separate case and I want to know if I can just copy field files between the different cases without reconstructing. I realize that mapFields would do this for me, but in my previous experience mapFields with parallel sources and destinations doesn't work quite right.

Thanks!

You can avoid this worries (and save some time too) if you just use symlinks or rsync the decomposed geometry to your sub-cases. You can then use

 
decomposePar -fields

to decompose your initial conditions without affecting the geometry.
Even if this is not the whole solution, it might be a good approach for you. This link might be of interest too: http://olesenm.github.com/2009/11/17...ludeIfPresent/

[RDanks]

Thanks for the tip! It had never occurred to me to use symlinks before. That blog post was also very enlightening.