While fitting, a significant amount of time and processing power is
spent on computing the probability function for every event and PDF
involved in the fitting model. To speed up this process, roofit can
use the computation functions provided in this library. The functions
provided here process whole data arrays (batches) instead of a single
event at a time, as in the legacy evaluate() function in roofit. In
addition, the code is written in a manner that allows for compiler
optimizations, notably auto-vectorization. This library is compiled
multiple times for different vector instruction set architectures and
the optimal code is executed during runtime, as a result of an
automatic hardware detection mechanism that this library contains.