Hamed, 2016-10-28 10:36 GMT-04:00 Hamed Babaei <hamedbabae...@gmail.com>: > If I solve a problem with one million DOFs on a cluster of 64 processors, > every processor will be dealing with around 15000 DOFs. > So you mean even in that case direct solver would be slow? I thought direct > solver would solve the problem only on those DOFs that it owns. But you are not solving 64 independent problems. Solving one problem with one million unknowns divided among 64 processors is more work than solving 64 problems which have a total of one million unknowns. Let say for simplification that the amount of work for the LU decomposition is the square of the number of unknowns (it is actually bigger than that). Solving a system of 100 unknowns will require 10000 operations while solving 10 systems of 10 unknowns will require only 1000 operations.
Best, Bruno -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
