Towards best achievable floating point performance for linear algebra computations on COTS Beowulf clusters
S. Fourmanoit, R. Roy, François Bertrand · PolyPublie (École Polytechnique de Montréal) · 2002
The solution of linear systems is required for many modern engineering applications from computational fluid dynamics to biomedical imaging. This paper shows how a low cost Beowulf cluster can be optimized to handle various CPU-intensive linear algebra computations using the scalable linear algebra package (ScaLAPACK). The paper focuses on the techniques and tools that we have developed to enhance the performance of a basic reference implementation. The tuning of system performance includes both partial recoding of some modules and careful engineering of clusters design, doubling speedup as compared to the reference implementation.