Parallelization of maximum likelihood fits with OpenMP and CUDA