Algorithmic skeleton library for scientific simulations: SkelGIS

Scientific simulations give rise to complex codes where data size and computation time become very important issues and sometimes a scientific barrier. Therefore, aiding parallelization has become a main topic in computer science. In this paper is introduced a new type of solution to give a totally transparent access to parallel programming for noncomputer scientists. This solution, named SkelGIS, takes place at the crossroads of domain specific languages, domain specific libraries and skeleton libraries, getting advantages of each domain. SkelGIS aims at providing a general solution for simulations in heterogeneous scientific domains. SkelGIS does not require a new language learning, nor to manipulate specific parallel tools, and it ensures a sequential programming style to the user with a totally hidden parallelization. Concepts and technical choices of the library are described in this paper. Furthermore, two very different benchmarks are studied to bring out performances of the library on light and complex simulations.