This paper presents a library dedicated to artificial neural networks development. The code, in classic C, obtained while using the functions of our library may be compiled and run on both types of computer architectures, namely sequential and general purpose shared memory MIMD parallel. Moreover, our library allows, with its small set of functions, artificial neural networks developers to design and implement their models while respecting the parallel characteristics of neuronal formalism. More generally, our library is a bridge between two parallel granularities, the fine grain of neuronal models and the medium grain of the MIMD computers.