Support: Support: US Air Force Office of Scientific Research (AFOSR) under Defense University Research Instrumentation Program (DURIP)

Period of Performance: February 1, 2023 — January 31, 2025

Budget: $228,000

Summary: We seek to fill in the gap in computing capacity between the personal desktops available to researchers in the PI’s group at the Center for Research in Scientific Computation (CRSC) and Department of Mathematics, North Carolina State University (NCSU), and extremely large-scale supercomputers such as those installed at the Air Force Research Laboratory (AFRL). The former are ready and convenient to use but can offer only a limited numerical performance. The latter pro-vide a virtually unlimited simulation capacity but are designed as multi-node distributed memory clusters and thus require a substantial array of additional software tools, as well as human skills, for the implementation of numerical algorithms. The proposed system will have a shared memory architecture and offer decent performance on the large scale (albeit not extreme). It will also have special hardware and software installed for the efficient execution of deep learning codes. Its performance will be sufficient for the development, verification, and validation of numerical methods, as well as addressing various radar applications of interest to the Air Force. At the same time, achieving top performance on this system will not require mastering the cluster architectures with distributed memory, which often appears a full-fledged development task in its own right, and will therefore be better suited for the University-based research in applied mathematics, as well as relevant objectives in Graduate research and education.