A randomized Lu-based solver using GPU and Intel Xeon Phi accelerators

Abstract

We present a fast hybrid solver for dense linear systems based on LU factorization. To achieve good performance, we avoid pivoting by using random butterfly transformations for which we developed efficient implementations on heterogeneous architectures. We used both Graphics Processing Units and Intel Xeon Phi as accelerators. The performance results show that the pre-processing due to randomization is negligible and that the solver outperforms the corresponding routines based on partial pivoting.