All, after more than two years of work by Timo Heister, Martin Kronbichler (who deserve most of the credit) and myself, Timo merged the branch on which we have implemented the necessary functionality to completely distribute meshes across many processors. Using this, deal.II can now solve problems on billions (!) of cells and with billions of unknowns if you have enough processors available (we tested this on up to 16,000). The new step-40 tutorial program explains how to do this -- it is a variant of step-6 that requires only some 25 extra lines of code. A (submitted) paper that describes this functionality can be found here: http://iamcs.tamu.edu/file_dl.php?type=preprint&preprint_id=237 I must admit that it's been quite a while since I was so excited about new stuff in deal.II!
I want to give thanks to Carsten Burstedde. The functionality above builds on his p4est library to manage fully distributed meshes on thousands, tens of thousands, and more processors. Carsten helped us out whenever we needed functionality in p4est and generously provided us with his experience in implementing algorithms on extremely large clusters. Best W. -- ------------------------------------------------------------------------- Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
