Cython: fast as GPUs without GPUs

Nowadays it is very common to heavily rely on intense computational power. For example, neural networks and deep learning are intensively employing graphical cards (GPGPU or GPUs) to extract weights in a reasonable computational time (e.g. seconds or minutes). This abundant usage of GPUs requires money, an API which is able to translate the source code in CUDA, for running and also a user-friendly written Python API.