Performances comparison: go-hep v.s. ROOT

This repository hosts a couple of small benchmark based on a TTree reading in order to compare performances of go-hep and ROOT.

Toy data

In these test, fake data are generated, with 30 branches of either floats or slices (slice size varies between 1 and n=30). The test consists in loading all the branches and computing three sums (involving all values of all branches). One can regenerate the fake data with:

$> go run ./gen-data.go

The following table shows the time (milliseconds per 10k events) for different codes, using the exact same input file, both for ROOT (left) and go-hep (middle), as well as the ratio between go-hep and the ROOT branch approach (right):

Root	go-hep	go-hep / branch
Method Floats Slices branch 2.4 23 reader 1.8 11	Commit Floats Slices b5f3896 1.2 12 839f662 3.8 45 feb48b1 4.7 59	go-hep commit Floats Slices b5f3896 0.5 0.5 839f662 1.6 2.0 feb48b1 1.9 2.7

Realistic data

In this test, 42 float branches are loaded for 720k events (contact me if you want to access the input file).

GOHEP

• b5f3896: 1.6 ms/kEvt (1.1 s for 720 kEvts)
• 839f662: 4.5 ms/kEvt (3.3 s for 720 kEvts)
• feb48b1: 5.2 ms/kEvt (3.7 s for 720 kEvts)

ROOT

• 2.6 ms/kEvt (1.9 s for 720 kEvts)