for enabeling multithreading in ROOT::Fit::Fitter::Fit. I’m currently using in my fitting class TSimuFit the functor approach for a user defined fit function:
where the class implements double TSimuFit::operator()(const double* par). The (binned or unbinned) data is a class member variable.
Is there a possibility to use the ROOT::EExecutionPolicy::kMultiThread also in this case? To be honest, I’m a little overwhelmed by the many Fitter::Fit and Fitter::FitFCN variants in the interface, but I don’t see any FitFCN that allows to set a ROOT::EExecutionPolicy.
Best and thanks,
Klaus
ROOT Version: 6.36.04 Platform: Linux Mint 21 Compiler: Not Provided
I think I meanwhile found my own error in reasoning. Since my user function loops the data and evaluates the fit model with given parameters itself, there is no simple opportunity for multithreading (by splitting the data into chunks) anymore. So I would like to rephrase my question:
I’d like to implement a simultaneous fit class, that evaluates a set of fit functions sharing several parameters on a set of individual data sets, either binned (histogram bin contents with errors) or unbinned (vector of double). The obvious opportunity for multithreading is splitting the data and evaluating the fit functions in parallel.
Is there a elegant way to do so using Fitter::Fit(..) with ROOT::EExecutionPolicy::kMultiThread enabled?