The development and maintenance of high-performance scientific computing software is a complex task. This complexity results from the fact that software and hardware are tightly coupled. Furthermore current parallel programming approaches lack of accessibility and lead to a mixing of concerns within the source code. In this thesis we define an approach for the development of high-performance scientific computing software which relies on model-driven engineering. In order to reduce both duration and cost of migration phases toward new hardware architectures and also to focus on tasks with higher added value this approach called MDE4HPC defines a domain-specific modeling language. This language enables applied mathematicians to describe their numerical model in a both user-friendly and hardware independent way. The different concerns are separated thanks to the use of several models as well as several modeling viewpoints on these models. Depending on the targeted execution platforms, these abstract models are translated into executable implementations with model transformations that can be shared among several software developments. To evaluate the effectiveness of this approach we developed a tool called ArchiMDE. Using this tool we developed different numerical simulation software to validate the design choices made regarding the modeling language.
