2020/03/06 by David Goz, Georgios Ieronymakis, Goz, David +17
Computer Science · #Advanced Data Storage Technologies #Distributed and Parallel Computing Systems #FOS: Computer and information sciences #FOS: Physical sciences #Instrumentation and Methods for Astrophysics (astro-ph.IM) #Parallel Computing and Optimization Techniques #Performance (cs.PF)
paper · pdf · doi:10.48550/arxiv.2003.03283
openalex publication_date 2020/03/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
New challenges in Astronomy and Astrophysics (AA) are urging the need for a\nlarge number of exceptionally computationally intensive simulations. "Exascale"\n(and beyond) computational facilities are mandatory to address the size of\ntheoretical problems and data coming from the new generation of observational\nfacilities in AA. Currently, the High Performance Computing (HPC) sector is\nundergoing a profound phase of innovation, in which the primary challenge to\nthe achievement of the "Exascale" is the power-consumption. The goal of this\nwork is to give some insights about performance and energy footprint of\ncontemporary architectures for a real astrophysical application in an HPC\ncontext. We use a state-of-the-art N-body application that we re-engineered and\noptimized to exploit the heterogeneous underlying hardware fully. We\nquantitatively evaluate the impact of computation on energy consumption when\nrunning on four different platforms. Two of them represent the current HPC\nsystems (Intel-based and equipped with NVIDIA GPUs), one is a micro-cluster\nbased on ARM-MPSoC, and one is a "prototype towards Exascale" equipped with\nARM-MPSoCs tightly coupled with FPGAs. We investigate the behavior of the\ndifferent devices where the high-end GPUs excel in terms of time-to-solution\nwhile MPSoC-FPGA systems outperform GPUs in power consumption. Our experience\nreveals that considering FPGAs for computationally intensive application seems\nvery promising, as their performance is improving to meet the requirements of\nscientific applications. This work can be a reference for future platforms\ndevelopment for astrophysics applications where computationally intensive\ncalculations are required.\n