timing output bug?
Closed this issue · 3 comments
donnaaboise commented
Originally reported by: Donna Calhoun (Bitbucket: donnaaboise, GitHub: donnaaboise)
Something wrong with the timing output - and this was a serial run ..
[libsc] Maximum attained at rank 0: 11.7025
[libsc] Statistics for GHOSTPATCH_COMM
[libsc] Global number of values: 1
[libsc] Mean value (std. dev.): [libsc Minimum attained at rank 0: 1
[libsc Maximum attained at rank 0: 1
[libscStatistics for GRIDS_LOCAL_BOUNDARY
[libsc Global number of values: 1
[libsc Mean value (std. dev.): 0 (0)
[libsc Minimum attained at rank 0: 0
[libsc Maximum attained at rank 0: 0
[libscStatistics for GRIDS_INTERIOR
[libsc Global number of values: 1
[libsc Mean value (std. dev.): 1 (0 = 0%)
[libsc Minimum attained at rank 0: 1
[libsc Maximum attained at rank 0: 1
[libscStatistics for GRIDS_LOCAL_BOUNDARY_RATIO
[libsc Global number of values: 1
[libsc Mean value (std. dev.): 1 (0 = 0%)
[libsc Minimum attained at rank 0: 1
[libsc Maximum attained at rank 0: 1
cburstedde commented
Weird. Is this reproducible?
donnaaboise commented
It happens fairly rarely, and mostly on supercomputers. I'll try to keep track of when it happens. Of course, it would be less of an issue if I weren't try to parse the output file, but instead just used the summary statistics available at the end of the output.
cburstedde commented
It happens fairly rarely, and mostly on supercomputers. I'll try to keep track of when it happens. Of course, it would be less of an issue if I weren't try to parse the output file, but instead just used the summary statistics available at the end of the output.
It is very much advised to process the summary line. The multi-line output is
really just for humans and not as reliable in parallel.