I have made some further adjustments to the assembly benchmark in bench/fem/assembly/cpp
The results are here: http://www.fenics.org/hg/dolfin?f=82bebeaa545c;file=bench/fem/assembly/cpp/bench.log Before trying to conclude anything from the results, please run the benchmark and report your results. There may still be bugs. Please take a look at the benchmark code as well, it's fairly short. It would be good if we could settle on a fixed benchmark that we all trust and then continue to discuss what to do about improving the numbers. The benchmark code is here: http://www.fenics.org/hg/dolfin?f=1378b25bf84b;file=bench/fem/assembly/cpp/main.cpp Just to trigger some discussion, Epetra seems to be the worst backend in the assembly event, but the clear winner in the reassembly event. -- Anders
signature.asc
Description: Digital signature
_______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
