A Parallelization Issue
I wrote a synthetic turbulence model that is based on a filter equation, so:
u(x,y,z) = Int over Vol (A(x',y',z')*K(x',y',z')*u'(x',y',z') dV)
Int is integral, and Vol is a portion of the volume around the cell (x,y,z).
The code is ok for scalar computation, now I would like to parallelize it.
The problem is that, as you can see, the filter equation require informations from several neighbouring cells. I thought that one possible approach is to save the needed cells that are across a processor boundary at the beginning, and then pass with Pstream them on run time. Could it be a sufficient solution? And, can I access, for each single processor, a "global" mesh index?
|All times are GMT -4. The time now is 18:08.|