NextGenTrigger Task 1.7

• CHEP talk together with Jolly got accepted
• Idea to increase memory bandwidth in the CLUE clustering algorithm of CMS

Implement NGT SoA Code in O2 standalone benchmark

• Working on this fork of the AliceO2 repo, with a CI pipeline:
  • Running on NGT hardware with 4 different GPUs (Nvidia and AMD)
  • Extended CI-pipline to fail if GPU.out changes
• Implemented SoA in:
  
  • SectorTracker
    • GPUTPCBaseTrackParam
    • GPUTPCTrackParam
    • GPUTPCTracklet
    • GPUTPCTrack
• The SoA library changed a lot in the course of implementing it in O2. Thus I have now adapted:
  • Unit tests
  • Documentation
  • Benchmarks (see plot below)
• Generate CSV file of the standalone benchmark
• Compute e.g. standard deviations of the benchmark and compare to baseline (dev branch)
• Display the CSV file in the GitHub Pipelines
• Next Steps:
  • Make better use of SoA to improve performance
  • Try David's suggestion

 

Issues

1. Even if I provide a .par file for AMD W7900, the outcome still depends on the settings in GPUParameters.csv
2. Can't really make sense of the standalone benchmarks timers: