An architecture for translating sequential code to parallel

With the increasing progress of high performance computing infrastructure, it has become necessary to design and implement easy to use tools that are able to translate any legacy software application to parallel. A deep observation of the existing tools allowing such translation or conversion reveals that all of them are still far away from their expectations. Indeed, some recent studies stress the urgent need for such tools in order to respond better to users’ needs in term of performance. In this paper, we propose a novel architecture based on web services which is able to translate any legacy software application into a parallel code. The resulting parallel code can be generated for any parallel programming model, also called the parallelization technique, such as MPI, OpenMP, CUDA, OpenCL or hybrid model. Our ultimate objective through this research is twofold. The first one is providing users a flexible architecture which is able to translate any sequential code into a parallel one, whatever the target parallel programming model. The second objective is providing parallel application developers a wide range of useful, easy to use parallel codes built as web services which can be included and used in their applications.