Skip to content

GitLab

FakeBkgTools: speed improvement for saveProgress() function of likelihood matrix method tool

Merged Julien Maurer requested to merge jmaurer/athena:b21_2-FakeBkgTools-Jan23-speed into 21.2

Three proposed improvements to the LhoodMM_tools class of the FakeBkgTools package, without impact on physics results.

  1. @slawlor reported that the saveProgress() function of the LhoodMM_tools tool could be pretty slow (hours) for large number of events (>100 k). This seems to be due mostly to a few places where heavy objects are copied inside a loop, while working with references would avoid the overhead. After changing to the latter, @slawlor indicated that the running time is largely reduced.

  2. a couple small memory leaks were removed in saveProgress() by using smart pointers.

  3. the mergeSubJobs() function seemed to be ignoring the ProgressFileDirectory property, making it impossible to process a ROOT file where the results are not stored at top level. This is now fixed.

Bringing this to the attention of @evarnes, the tool expert...