Palestra do Prof. Juan Alonso, Stanford University

Abstract: The advent of ever-more-powerful computing architectures that are faster and more heterogeneous in nature (accelerators, many-core, GPUs) and that require explicit memory management by the developer present significant opportunities to improve the level of resolution of fluid flows of interest.  In addition, such awesome computing power allows us to explore complex design spaces and pursue both aerodynamic shape optimization and even multi-disciplinary problems.  But extracting performance from new computing architectures is not trivial and requires sustained efforts from an interdisciplinary team of researchers that are knowledgeable in flow physics, computing science / HPC, and applied mathematics.  This talk discusses our efforts at Stanford University and in partnership with many institutions around the world to develop a community, open-source code and framework named SU2 (su2.stanford.edu) with two flow solvers that incorporate the latest techniques in HPC, numerical methods, and flow models.  The talk discusses the challenges of developing such a system, the improvements in performance that can be obtained from using modern computing architectures, and highlights some recent examples of supersonic aircraft shape optimization and the new Discontinuous-Galerkin higher order solver that can be used for large-eddy simulations.