Emerging High Performance Computing Systems and Next Generation Engineering Analysis Applications
James Ang, Richard Frederick Barrett, Simon David Hammond, Arun F. Rodrigues · 2013
This paper provides a high level overview of the intersection between the broad fields of Infrastructure Engineering and Computer Systems Engineering. The last two decades of technical high performance computing (HPC) have been remarkably stable, with high-end scientific and engineering applications able to leverage the increases in performance of commodity processors in massively parallel supercomputers. But issues began to arise with the advent of the dual core processor in 2004. While many commercial workloads and some technical applications such as materials science can still achieve good performance on multi-core processors and many-core based systems, most finite element engineering analysis applications are sensitive to data locality and data movement and thus have difficulty realizing the performance potential of these systems. This paper describes the HPC co-design methodology we are using to guide the development of advanced concepts for HPC computer architectures and future engineering analysis applications that will execute on them.