If you are using the parallel version of the Ansys Polyflow solver, you will need to divide the computational domain into sub-parts. This decomposition into sub-parts can also be used to reduce the CPU cost of a calculation using the standard serial solver. Two decomposition methods are available: Metis (developed at the University of Minnesota) and Chaco (developed and licensed by Sandia National Laboratories). Metis is provided with the standard release of Ansys Polyflow, and Chaco is available as an additional option. They both yield good results, but Chaco may provide better results in some cases.

The number of sub-parts must be a power of 2. Note that the number of sub-parts can be different from the number of processors used by the parallel solver. The number of sub-parts must be higher than the number of active processors, which is in turn arbitrary and not constrained to be a power of 2. It is not uncommon to solve a large 3D problem with 512 sub-parts on, say, 4 or 8 processors or even on a single-processor machine.