Optimizing weight evaluations for convolution fits in RooFit using vectorization

Convolutions play an important role in high energy physics, since they are used when fitting probability density functions (PDFs) to data. When fitting convolutions of PDFs to data with RooFit, the PDFs are convoluted with a discrete fast Fourier transform and the result is stored in a histogram that is interpolated at its bin centres. This report concerns optimizations of weight evaluations in RooFit in the case of one dimensional histograms with up to second order interpolations. By implementing vectorized versions of the evaluation functions, this project resulted in approximately 2-5 times faster weight evaluations, depending on interpolation order and histogram properties. A suggested continuation of the project is to implement vectorization for the convolution operations.