Branch of Float16_t datatype

ROOT
1

Hi,
I am doing some experimentations to reduce the size of the root files we have.
I think converting branches from Float_t to Float16_t is a very good idea since we do not need all these digits especially when the values are in MeV.
In my experiments I used root 6.16/00 and actually I did not see any size reduction at all

Is this feature available in this root version?

This is the code I am using

TFile f("test.root", "RECREATE");
auto t = new TTree("t", "t");
TRandom3 r(1);
Float16_t  f1 = 0.;
t->Branch("f1", &f1, "f1/f");

for(int i=0;i<10000;i++){
    f1=r.Uniform(0,1);
    t->Fill();
    
}

f.cd();
t->Write();
2

This is supported only in the soon to be released v6.18/00 (and/or the master branch).

Cheers,
Philippe.

3

Thanks a lot,
I am eager to use this feature :sunglasses:
few questions:

  1. will vector<Float16_t> be supported?
  2. will we have a method to determine the number of digits to keep?
    for example 3.14159265359 can be reduced 3.1415.
4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.