RUNTIME PARALLELIZATION OF THE FINITE-ELEMENT CODE PERMAS

The general-purpose finite element system PERMAS has been pelted to pa rallel computer architectures within the scope of the ESPRIT project E UROPORT-1. The kernel of this 1M lines of code has been restructured i n such a way that parallelism is automatically identified and exploite d at runtime. This paper describes the general structure of the approa ch and the evaluations and models carried out to analyze the potential performance of different algorithms. The study shows the performance bounds of the application, the efficiency of the scheduling algorithms , and the influence of architectural parameters of the target machine. The paper also describes some issues of relevance of the approach for such an industrial code. Among them, issues of compatibility with pre vious versions, reduction of future development, and maintenance costs are addressed.