Skip to content

Only run Pixel, SCT and TRT conversion to xAOD for TIDE preselected events.

Goetz Gaycken requested to merge goetz/athena:23.0_IDTIDE_fix_exessive_cpu_consumption into 23.0

The conversion of Pixel, SCT and TRT raw data to xAOD is very time consuming. Since the content is only written to the IDTIDE stream these algorithms only have to run for IDTIDE selected events.

This is realised by extra sequences. There is now one sequence operated sequentially until an algorithm (i.e. the IDTIDEKernelPresel or IDTIDEKernel) set the filter status to reject the event. Then there is a sequence (IDTIDEPreselAlgSequence) for algorithms like the conversion algorithms which are executed after IDTIDEKernelPresel in parallel. Finally there is a sequence (IDTIDEPostProcSequence) for the augmentation and thinning kernels which are run after the final selection of the IDTIDEKernel (also in parallel):

  • IDTIDEPreselSequence (sequential, stop depending on algorithm filter status)
    • DerivationKernel/IDTIDEKernelPresel
    • IDTIDEPreselAlgSequence (parallel)
      • TRT_PrepDataToxAOD/xAOD_TRT_PrepDataToxAOD
      • SCT_PrepDataToxAOD/xAOD_SCT_PrepDataToxAOD
      • PixelChargeToTConversion/InDetPixelChargeToTConversion
      • PixelPrepDataToxAOD/xAOD_Pixel_PrepDataToxAOD
    • DerivationKernel/IDTIDEKernel
    • IDTIDEPostProcSequence (parallel)
      • DerivationKernel/DFTSOSKernel
      • DerivationKernel/IDTIDEThinningKernel

Merge request reports