Hi all, I am working on a CFD code which calculates differences, averages, interpolations, etc. by computing matrix-vector and matrix-matrix products. The basis here is formed by (very) sparse matrices and extension to more dimensions is done with kronecker products. This works fine one a single processor; however, before implementing things in parallel I have two questions:
- is there a way to compute kronecker products efficiently in parallel with Petsc? - if problems get big, let's say order of 10-100 million unknowns, is it still efficient to store the entire (sparse) matrices? is this commonly done, or is it better to shift to matrix-free methods? btw, the current system I can run on has several hundreds of processors and 24Gb memory per node (8 proc. per node). Thanks, Ben
