@tenderlove

Back, in my research centre, we often used to say that the best way to make things faster is first to change the algorithm itself if we can. That can have a many orders of magnitude difference. Everything else could also help, if you couldn't change the algorithm.