Presenter

Michael Mascagni is professor of Computer Science, Mathematics, and Scientific Computing at Florida State University. He is also a Faculty Appointee at the National Institute for Standards and Technology. Prof. Mascagni expertise is in numerical and scientific computing, especially in aspects of stochastic computing. In particular, he is an expert in Monte Carlo methods and random number generation and their application to scientific problems and their implementation on high performance computing (HPC) architectures. In particular, his work on random number generation, and the Scalable Parallel Random Number Generators (SPRNG) library has included consideration of the reproducibility of random number streams when computations are undertaken in diverse HPC environments. SPRNG was absolutely reproducible on distributed memory parallel machines, but the notion of reproducibility has had to be modified for multicore and accelerator based architectures. This motivated his interest in the more general problem of numerical reproducibility for HPC systems, especially at the Exascale.