FitQualityOnSurface make it trivial
We seem to have
-
FitQuality
(refers to Track or Segment) that needs to beexpandable
and so have virtual methods (and therefore vptr) -
FitQualityOnSurface
. We have much more of them around . And for them we seem to need just thebasic
part -
This tries to achieve this. The only gain here for this MR would be the
vptr
so FitQualityOnSurface will have 2/3 the size wrt before. -
In principle as a second step one could even attempt to change here [https://gitlab.cern.ch/atlas/athena/-/blob/master/Tracking/TrkEvent/TrkTrack/TrkTrack/TrackStateOnSurface.h] to using "value semantics" which prb would get rid of quite a few small allocations all over the place , per event (and prb improve
locality
) -
There are other things one could also fix (like having a bool operator checking both doubles are not 0 or so ...)
clean version of [!57973 (closed)]