Nowadays, parallel computers have become ubiquitous and current processors contain several execution cores, anywhere from a couple to hundreds. This multi-core tendency is due to constraints preventing the increase of clock frequencies, such as heat generation and power consumption. A variety of low-level tools exist to program these chips efficiently, but they are considered hard to program, to maintain, and to debug, because they may exhibit non-deterministic behaviors. This project focuses on adding an abstraction level in order to have as much performance as possible while not dealing with low-level mechanisms. The approach is based on data flow programming, which allows the programmer to specify only the operations to perform and their dependencies, without actually scheduling them. This project combines this paradigm with the Polyhedral Model, which allows automatic parallelization and optimization of loop nests, in order to make the programming easier by delegating work to the compilers and static analyzers.