Skip to content

Rich future clustering improvements + small RichTools cleanup.

Christopher Rob Jones requested to merge RichFutureClusteringCleanup into master

Updates the future clustering algorithm to make it much more CPU efficient :

  • Sort the data by PD (pointer) before the clustering, instead of by cluster object afterwards.

  • Remove the need to copy the hit containers due to the 'cleaning' and the associated private tool instances. In practise the only cleaning we use is an 'all or nothing' maximum PD occupancy cut and this can be implemented much more efficiently (and sensibly, its a little overkill to have a tool to just apply a single < cut...) directly in the algorithm rather than via a tool.

Overall, gains just over a factor 5 in speed...

Merge request reports