Efficient matrix-free implementation of discontinuous Galerkin methods for compressible flow problems

Abstract

We discuss the matrix-free implementation of Discontinuous Galerkin methods for compressible flow problems, i.e. the compressible Navier-Stokes equations. For the spatial discretization the CDG2 method and for temporal discretization an explicit Runge-Kutta method is used. For the presented matrix-free approach we discuss asynchronous communication, shared memory parallelization, and automated code generation to increase the  oating point performance of the code.