geoclaw
geoclaw copied to clipboard
clawpack
Provide better timing info
Currently geoclaw provides output like what's below, which is confusing. This is also different from what is provided by amrclaw. Clean this up since most users don't care about Bufnst or gfixup -- were these left over from debugging?
It's not clear what Total time vs. (clock) vs. (wall) vs. (all cores) means below.
In the example below it reports total time 2728 seconds, about 45 minutes, which was the wall time. It would be useful to also report the total CPU time to see how good the OpenMP efficiency is.
Total time to solution = 2728.37300000 s, using 7 threads
Total advanc time on level 1 = 1.44800000 s
Total advanc time on level 2 = 8.35200000 s
Total advanc time on level 3 = 350.64700000 s
Total advanc time on level 4 = 2073.11100000 s
Total updating time 18.57400000 s
Total valout time 12.49000000 s
Total regridding time (clock) 261.31800000 s
Total Grdfit time (clock) 115.17100000 s
Total Flglvl time 109.02200000 s
Total Flagger time (clock) 22.80100000 s
Total Bufnst time (clock) 45.66100000 s
Total gfixup time (clock) 19.96700000 s
Total filval (wall) time 19.72300000 s
Total filval (all cores) time 134.37400000 s
Total setaux (all cores) time 29.65100000 s
integration time, still not counting saveqc
Total Bound (all levels) wall time 49.18100000 s
Total stepgrid (all levels) wall time 2352.02700000 s
would you like to see only the left most times? or should all of it be hidden? Or just better explanations?
I don’t think it’s really possible to provide total cpu time, unless someone knows otherwise?
Marsha
On May 16, 2015, at 6:36 PM, Randall J. LeVeque [email protected] wrote:
Currently geoclaw provides output like what's below, which is confusing. This is also different from what is provided by amrclaw. Clean this up since most users don't care about Bufnst or gfixup -- were these left over from debugging?
It's not clear what Total time vs. (clock) vs. (wall) vs. (all cores) means below.
In the example below it reports total time 2728 seconds, about 45 minutes, which was the wall time. It would be useful to also report the total CPU time to see how good the OpenMP efficiency is.
Total time to solution = 2728.37300000 s, using 7 threads Total advanc time on level 1 = 1.44800000 s Total advanc time on level 2 = 8.35200000 s Total advanc time on level 3 = 350.64700000 s Total advanc time on level 4 = 2073.11100000 s Total updating time 18.57400000 s Total valout time 12.49000000 s
Total regridding time (clock) 261.31800000 s Total Grdfit time (clock) 115.17100000 s Total Flglvl time 109.02200000 s Total Flagger time (clock) 22.80100000 s Total Bufnst time (clock) 45.66100000 s Total gfixup time (clock) 19.96700000 s Total filval (wall) time 19.72300000 s Total filval (all cores) time 134.37400000 s
Total setaux (all cores) time 29.65100000 s
integration time, still not counting saveqc Total Bound (all levels) wall time 49.18100000 s Total stepgrid (all levels) wall time 2352.02700000 s — Reply to this email directly or view it on GitHub.
