From: robs <aq...@ya...> - 2012-03-11 13:49:20
|
I've done some more testing and it looks like what we've currently got is in fact pretty good: This is before the interleave optimization: single-threaded real 37.51 user 37.06 sys 0.32 omp multi-threaded real 27.60 user 52.13 sys 0.51 fully multi-threaded (simulated using two processes) real 18.41 user 33.01 sys 1.22 but after: single-threaded real 31.58 user 30.64 sys 0.40 omp multi-threaded real 20.37 user 38.23 sys 0.39 fully multi-threaded (simulated using two processes) real 18.93 user 32.89 sys 1.47 So OMP seems to be doing a good job (on this linux32 dual-core system at least) Script attached if anyone else wants to try. (run with parameter "gen" the 1st time) Cheers, Rob |