2016-10-05 10:19 GMT-04:00 Chenchen Liu <[email protected]>: > For "#PBS -l nodes=1:ppn=x", > when x=1, and use "mpirun -np 1 ./step-17", it works well as a single > processor. > when x=1, and use "mpirun -np 4 ./step-17", memory error happens. > For "#PBS -l nodes=4:ppn=1", > when I use "mpirun -np 1 ./step-17", it works well as a single processor. > when I use "mpirun -np 4 ./step-17", memory error happens. > > Everytime I use " GridTools::partition_triangulation (n_mpi_processes, > triangulation)", error happens unless n_mpi_processes=1. > By the way, I am still using deal.ii 8.3.0, and I saw some upgrades for > step-17 and 18 in new version of deal.ii. Especially, in new step-18, it > does not use "GridTools::partition_triangulation (n_mpi_processes, > triangulation)", and it uses automatic partitioning. I am not sure whether > it will solve my problem with METIS. Do you know whether automatic > partitioning uses METIS? But still, I think the old step-17 and 18 should > work well with version 8.3... Automatic partitioning does not use METIS and does not work across nodes. It might be a problem with your METIS or deal with METIS install. Since your are loading a module for deal.II, I assume that deal was installed by an administrator, right? If that's the case, you should tell them that step-17 doesn't work. I don't know what else can be wrong.
Sorry 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 [email protected]. For more options, visit https://groups.google.com/d/optout.
