Performance Comparison of HPX versus MPI+X Threading Models for Finite-Element Models of Environmental Flows

Increasing vector widths and many-core architectures introduce significant challenges to achieving efficient compute resource utilization. High performance parallelX (HPX) is an asynchronous run-time specifically designed to address the bottlenecks associated with the massive concurrency on next generation supercomputers. We present a comparison of a traditional MPI+X threading model vs. an HPX implementation of standard and blended isogeometric discontinuous Galerkin wave kernels. We will present detailed performance profiling and discuss code optimization strategies for each parallelization paradigm. Scaling results will be presented on the Intel Knights Landing chips.