# nc2008/grass7 mapset OUTFILE=callgrind.out.1196 g.region n=221230 s=220404 w=638634 e=639530 res=2 CMD="v.surf.bspline in=lidar_test raster=elev.lidar_test_cubic method=bicubic --o" valgrind --tool=callgrind --callgrind-out-file=$OUTFILE --trace-children=yes $CMD # http://kcachegrind.sourceforge.net kcachegrind $OUTFILE 99% of time is spent in gmath lib's G_math_cholesky_sband_decomposition() time $CMD ========= base serial: real 2m58.174s user 2m57.759s sys 0m0.428s -- real 2m58.215s user 2m57.723s sys 0m0.508s ## with OpenMP x6 cpus: real 0m50.322s 3.5x speedup in clock time (42% efficiency loss) user 4m44.006s 1.6x increase in overall CPU time sys 0m1.048s --nostaic real 0m50.792s user 4m46.426s sys 0m1.248s --static again real 0m50.168s user 4m42.914s sys 0m1.180s --nostatic again real 0m52.104s user 4m54.334s sys 0m1.244s (static wins) ##OMP_NUM_THREADS=4: real 1m30.730s user 5m52.274s sys 0m0.696s ##OMP_NUM_THREADS=2: real 1m48.180s 1.6x speedup in clocktime user 3m32.477s 1.2x increase in overall CPU time sys 0m0.560s ##OMP_NUM_THREADS=1: real 2m51.480s user 2m51.031s sys 0m0.464s