Fast, good, and repeatable: Summations, vectorization, and reproducibility
Brett Neuman, Andy J Dubois, Laura Monroe, Robert Robey · The International Journal of High Performance Computing Applications · 2020
Enhanced-precision global sums are key to reproducibility in exascale applications. We examine two classic summation algorithms and show that vectorized versions are fast, good and reproducible at exascale. Both 256-bit and 512-bit implementations speed up the operation by almost a factor of four over the serial version. They thus demonstrate improved performance on global summations while retaining the numerical reproducibility of these methods.