Maximizing LS-DYNA MPP Performance Practices

This technical report presents established practices for enhancing performance in LS-DYNA MPP simulations across various hardware platforms. It begins by discussing the selection of the appropriate LS-DYNA binary tailored to specific system characteristics, emphasizing the importance of aligning model size with available cores and processor capabilities. The report details the benefits of using Intel or AMD compiled binaries based on the hardware in use, as well as the advantages of employing Intel MPI or Open MPI over older standards. Additionally, it outlines strategies for achieving efficient scaling through undersubscription and hybrid configurations, which can optimize resource usage and improve performance. The document also addresses the significance of domain decomposition in load balancing and provides methods for detecting and correcting load imbalances to enhance overall simulation efficiency. Finally, it highlights the impact of high clock speeds and optimal process pinning strategies on simulation performance, ensuring that users can maximize the capabilities of their systems effectively.