Accelerating the FlowSimulator: Low-level Analysis and Optimization of CODA Kernels

This paper investigates the performance of the solver CFD by Onera, DLR and Airbus (CODA)}. Performance of an HPC code mainly consists of two factors: scalability and node-level performance. Previous studies showed good scalability for CODA, but low node-level performance, with relevant code parts being neither compute nor memory bound. In this study we analyze the code on assembly instruction level to now get a deeper look into what the actual bottlenecks are. In conjunction static and dynamic analysis tools provide very detailed data about the program execution. Using the dynamic analysis tools LIKWID, Linux perf-tools and MAQAO and the static analysis tools OSACA and MAQAO, we investigate a selection of the most relevant kernels in CODA. We then develop optimization strategies, implement them and evaluate their impact.