HYBRID FPGA-GPU-CPU HPC PLATFORM FOR HIGH-FIDELITY DIRECT NUMERICAL SIMULATIONS OF LAMINAR-TURBULENT TRANSITION

Abstract

Funding is requested for a hybrid FPGA-GPU-CPU-system consisting of 16 FPGAs (Xilinx U200 Alveo), 16 GPUs (NVidia Tesla V100 32GB PCIe), and 8 CPUs (Intel Xeon Silver 4216 Sixteen-Core). Because GPU and FPGA architectures are very different compared to conventional CPUs, our DNS codes that have been developed for CPU systems cannot simply be ported to a hybrid FPGA-GPU-CPU platform.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 25, 2023
Source ID
FA95502110026

Entities

People

  • Hermann Fasel

Organizations

  • Air Force Office of Scientific Research
  • United States Air Force
  • University of Arizona

Tags

Fields of Study

  • Physics

Readers

  • Computational Fluid Dynamics (CFD)
  • Parallel and Distributed Computing.