Heterogeneous Computer System for Code Development, Large-Scale Simulations, and Data Post-Processing

Abstract

Accurate first-principles based computational fluid dynamics (CFD) simulations of flows of engineering interest require grid resolutions that challenge existing Central Processing Unit (CPU) based computer systems. Recently a shift towards heterogeneous systems occurred at the DoD High Performance Computing (HPC) centers. The combination of CPUs, Graphics Processing Units (GPUs) and other purpose-designed hardware (such as the Intel Phi chip) into heterogeneous systems resulted in a leap in compute power over older, purely CPU-based systems. Novel simulation codes and algorithms, ideally based on an open standard such as OpenCL, have to be devised to unleash the compute power inherent in these new hardware architectures for cutting edge CFD research. The computing hardware at New Mexico State University (NMSU) has not kept up with these latest developments at the DoD HPC centers and does not allow for the local development and testing of efficient OpenCL codes. Despite the minimal computing power available, NMSU has considerable expertise in developing CFD codes that utilize the Message Passing Interface (MPI) for parallel execution on distributed-memory CPU-based HPC systems but is behind in the development of codes and algorithms for heterogeneous systems. The proposed system will significantly enhance and accelerate our DoD funded research by allowing us to develop codes and algorithms that exploit the massive parallelism of the latest heterogeneous systems at the DoD HPC centers and by providing (for the first time ever) a local computer system that is large enough for production runs in support of our DoD-funded research. The proposed system will also constitute an ideal education and training environment for our graduate students in all aspects of modern supercomputing.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 14, 2019
Source ID
W911NF1810454

Entities

People

  • Andreas Gross

Organizations

  • Army Contracting Command
  • New Mexico State University
  • United States Army

Tags

Fields of Study

  • Engineering

Readers

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