I get a bunch of numbers per event and would like to store them.
I have implemented a tree to do it. But what I noticed was, hadd takes incredibly large amount of time, of the order of 10 hours for what used to take just under a minute, when there’s no tree. the populated data is also not much (<50MB with tree) per file over 100s of files.
With tree : < 50MB per file, hadd : 10+ hours
Without tree < 1 MB per file , hadd : < 1 minute
Is there an alternative data structure that I can use or a way to reduce the time it takes to hadd the files ?
hadd is usually very fast even with TTrees … What command line do you use to invoke hadd? Can you share a few of your files for us to try to reproduce this problem? Is the problem linear? (i.e. i you have 100 files does it take 10 times more time than to hadd 10 files?)