Print cumulative timing information for each module at end of job when Timing service is used
It would be nice to have a summary of the amount of time spent in each module printed at the end of a job, with wall time, user time and system time, ie
#1 Updated by Rob Kutschke over 9 years ago
I would like to add to this request. I would like to have the same info printed at different times and the option to configure any/all of the printout:
1) During each event at the end of each module so that we can find nasty events. ( Like the existing timer service ).
2) Per module once at the end of the job (Brian's request)
3) Summed over all modules, once at the end of the job.
Can the printout be decorated to make it easy to parse to build statistics over many jobs? Alternatively does it make sense to have any of the above written to the event-data output file? Not sure yet if I really want this since there will be jobs without event-data output file and other jobs with many such files.
#10 Updated by Kyle Knoepfel almost 6 years ago
- Status changed from Accepted to Resolved
- % Done changed from 0 to 100
This feature request has been resolved as described in issue #3598. We have elected to not provide separate user- and system-time quantities, which can be determined only by calling a function that we think is rather expensive. Instead, we have added a new service called
TimeTracker, which provides a more descriptive printout that includes the minimum, maximum, mean, median, and root-mean-square of time taken to execute a full event or a specific module.
Implemented with e7078a47887c46d97e977b05f4f985b0cc5b3aec.