Thanks, Dear Wolfgang for your reply, I always have the problem with dealii installation as well:
[image: Screenshot from 2022-02-10 11-09-05.png] I use candi for the installation as I have not found a straightforward installation process on dealii documentation. It might be possible for my case the problem comes from my installation. Therefore, how do you usually install dealli on Ubuntu along with its dependencies (like P4est, Trilinos, METIS, ...)? I am struggling with this issue for almost 3 weeks without any progress. On Thursday, February 10, 2022 at 8:35:10 AM UTC+11 Wolfgang Bangerth wrote: > On 2/9/22 01:49, Mohammad Fazli wrote: > > I am using dealii and I faced a big problem with parallel scaling on > > ubuntu 18.04. > > For instance, I have compiled step-17 with different core numbers on my > > machine: > > Memory: 31.2 GB, > > Processor: Intel® Core™ i7-8700 CPU @ 3.20GHz × 12, > > which shows me: > > > > *time mpirun -np 1 ./step-17* > > *real 0m17.701s* > > *time mpirun -np 2 ./step-17* > > *real 0m16.507s* > > *time mpirun -np 4 ./step-17* > > *real 0m33.934s* > > * > > * > > I have tried to install dealii several times but all the time the same > > issue happens. Someone told me the problem can be with p4est. But I > > don't know what I should do to resolve this issue. > > Mohammad: step-17 does not actually use p4est, so that is not the cause. > Your results match what I see on my machine reasonably closely -- this > is what I get here: > > 1 processors: 0m25.833s > 2 processors: 0m20.885s > 4 processors: 0m14.838s > 8 processors: 0m15.463s > 16 processors: 0m15.858s > 32 processors: 0m17.176s > > In other words, there is not very much speedup. There can be many > reasons for that, among which are: > > * You have only a single CPU. Using multiple MPI jobs will eventually > run into memory bandwidth limitations between the CPU and main memory. > > * The problem is quite small (on the finest level, step-17 has only > about 40,000 unknowns). Breaking it into even smaller pieces does not > leave very much work for each MPI process, but requires substantial > communication. > > * The program is intended to demonstrate some basic functionality, but > it is not tuned for efficiency, and certainly not for such small problems. > > So there may not actually be anything wrong with what you do. If you > wanted to have a more realistic test, make sure that you increase the > initial number of refinement steps so that you have several 100,000 > unknowns, and then try again with that. > > Best > W. > > -- > ------------------------------------------------------------------------ > Wolfgang Bangerth email: [email protected] > www: http://www.math.colostate.edu/~bangerth/ > -- 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]. To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/13323a75-39e3-4411-bdee-b99c267223ecn%40googlegroups.com.
