Reproducible Triangular Solvers for High-Performance Computing

Roman Iakymchuk, David Defour, Sylvain Collange, Stef Graillat
2015 2015 12th International Conference on Information Technology - New Generations  
On modern parallel architectures, floating-point computations may become non-deterministic and, therefore, nonreproducible mainly due to non-associativity of floating-point operations. We propose an algorithm to solve dense triangular systems by leveraging the standard parallel triangular solver and our, recently introduced, multi-level exact summation approach. Finally, we present implementations of the proposed fast reproducible triangular solver and results on recent NVIDIA GPUs.
doi:10.1109/itng.2015.63 fatcat:mhvv2rksy5gobiplofz3xs7mf4