yanliusp / antinu Goto Github PK

what does backfill mean?

Not an issue per say, but here is how:

https://root-forum.cern.ch/t/looping-over-branches/18709

I'm a bit confused with what's the best way to apply the cuts.

The most straight-forward way is just apply the global (by global I meant both the prompt and delayed events) cuts in skim.cc. These cuts includes trigger, datacleaning, fitValid, and possibly u*r, FV, etc..

However, this is computational very expensive, because every time I adjust FV, or u*r, I would need to loop through all data, which 99% are useless events.

Another possible idea is to do the necessary skim first (trigger, datacleaning and fitValid), and treat the skim files as the base files that I will operate on.

This means I'd have to skim (here skim means applying FV, u*r, .. global cuts)again the skimmed files.

I think I will go with the second method (otherwise it would take too much time). but I need to think more clear the sequence of the cuts applied. Maybe I need to take into consideration of all cuts I might need to apply.

Anyway, I need to think more about adding the branches, it looks like adding the branches too early is not a clevel thing to do.

triggerWord, dataCleaning and fitValid would remove:

(16195593-49965)/16195593 = 99.691%

of the events....

fitValid of all has the single biggest effect.

Currently with the basic cuts (trigger, datacleaning and fitValid) it's taking too much time to preprocess. (~40 min per 1,000,000 skim_s%.root file)

Tow possible ways to reduce the time:

1. apply beta14, itr, and preliminary posr and udotr cuts.
2. not copy all the branches -> instead only copy branches of interests.

This issue is not urgent.