Matti Kortelainen
Matti Kortelainen
In case IgProf+cmsRun combination crashes, is any information on the job timings saved that can be used for comparison?
The crashes under profilers are quite likely caused by the memory corruption inside Tensorflow (when ran through IgProf or VTune) that has been investigated in https://github.com/cms-sw/cmssw/issues/42444.
Is the most burning problem that there is no timing information (entire job, per module) or that the real IgProf/VTune profile (with function-level information) is missing (because of crash)?
> due to initialization times reading inputs we are having too many false positives Do you mean something along the the baseline being run first, and that leading the input...
In the table, I'd suggest to also add units to both time and memory, and in the cells present first the baseline and then the PR value (but keep the...
Comparison differences are in 29634.911, which is a "known" instability (although we don't seem to have a dedicated issue for that?)
+core
Any news?
assign operations, core
On the automation side, would it be more important to try to cover all such PRs at the expense of some PRs being misclassified (i.e. some amount of false positives...