> CPU time / wall clock time / wall clock time spent in solve+update > routines (that in %) / wall clock time spent in assembly routines (that > in %). > All timings performed using the deal timing functions on an 8-core linux > machine. Time is in seconds. The solver used is UMFPACK. > Assembly using TBB: > 744 / 133 / 114.38 (86%) / 15.96 (12%) > Serial assembly: > 750 / 225 / 114.75 (51%) / 108 (48%) > Assembly using the "old style" threading: > 854 / 157 / 114.61 (73%) / 39.25 (25%)
Very nice -- I'm actually quite impressed myself by the speedup for assembly :-) > In conclusion, I must say that I am actually quite happy about how > parallel assembly works. Now I just have to find a way to speed up the > solver - are there any parallel (not MPI based) sparse solvers? Speeding > up refinement would also help... There is a threaded version of SuperLU, another sparse direct solver just like UMFPACK. I made some experiments with the serial version of SuperLU some 3 years ago but it was slower than UMFPACK and I lost interest in this. I still have the patches I created back then -- if you were interested in playing around with them, and in particular if you were interested in trying to get them to work with the threaded version of SuperLU, I'd be more than happy to provide them to you! Another alternative would be to use an iterative solver; matrix-vector products are threaded in deal.II, though most preconditioners are not. Finally, as for the numbers of threads: The TBB starts up 8 threads on an 8-processor machine at the start of the program; this way, whenever there is work to do, there is also a thread waiting as thread creation is an expensive procedure you don't want to do too often. Best W. -- ------------------------------------------------------------------------- Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
