Automatic translation from CUDA to C++

Performance portability is a major challenge faced today by developers on the heterogeneous high performance computers. This project consisted in the development of a tool for translating CUDA programs in C++. Our approach consists in writing the best possible code using the best possible frameworks and run it on the best possible hardware. This means that when we write our code, we don't think in advance about the legacy hardware on which our code will run, thus we can work with modern frameworks like CUDA. Another fundamental aspect is that we want to discover if CUDA can be an effective data-parallel programming model for more than just GPU architectures