In this bimonthly feature, HPCwire highlights newly published research in the high-performance computing community and related domains. From parallel programming to exascale to quantum computing, the details are here.
Many new HPC systems utilize many nodes with several processors per node. In this paper, researchers discuss how they designed a simulator to help estimate power use in these clusters and assist designers in experimenting with execution of different applications. They propose a method for managing the temporary shutdown of CPUs when they are idle and claim that their method achieves a nearly 20 percent power reduction.
Authors: Maaz Ahmed, Mohsin Khan, Waseem Ahmed, Rashid Mehmood, Abdullah Algarni, Aiiad Albeshri and Iyad Katib.
In the race for quantum computing, cooperation can be just as important as competition. In this paper – written by a team from Clemson University – the authors examine successful contributors to the quantum computing field and community with the goal of understanding the roots of their success. They outline the key characteristics of successful contributors and suggest objectives for opening the field to more contributors.
Authors: Ruslan Shaydulin, Caleb Thomas and Paige Rodeghero.
This article, written by a team from France and Poland, surveys web interfaces for HPC simulation software. They present a brief history beginning in the 1990s, outline a series of web-based HPC portal use cases, and identify and discuss the key features that characterize those portals. They discuss usability requirements and outline a path forward for future work.
Authors: Patrice Calegari, Marc Levrier and Paweł Balczyński.
This paper – written by a team of research from NREL – discusses the ExaWind project, which aims to “enable scientific discovery through predictive simulations of wind farms comprised of many megawatt-scale turbines situated in complex terrain.” Simulations of a single turbine are petascale-class – meaning that the ExaWind project will require exascale computing. In this paper, the authors discuss a “production run” of a megawatt-scale turbine simulation.
Authors: Michael J. Lawson, Jeremy Melvin, Shreyas Ananthan, Kenny M. Gruchalla, Jonathan S. Rood and Michael A. Sprague.
Isambard, the first Cray XC50 “Scout” system, will be delivered in the summer of 2018 and contain over 10,000 Arm cores. In this paper, researchers from the University of Bristol present node-level results from eight early-access nodes. They discuss the benchmark results, comparing them to mainstream CPUs, with a focus on applications of interest to the UK national HPC service.
Authors: Simon McIntosh-Smith, James Price, Tom Deakin and Andrei Poenaru.
With error rates expected to increase dramatically in the exascale era, this researcher – a graduate student from the University of Illinois at Urbana-Champaign – suggests that application-level checkpointing may no longer be enough. The author uses fault injection to emulate failures, providing insight into application- and system-level resiliences. They find deficiencies in fault tolerance and a previously-unobserved failure phenomenon.
Author: Sharon S. Tang
Modeling gravitational lenses – areas with gravitational fields strong enough to bend light – can help astronomers understand cosmology and the distribution of dark matter. In this paper – written by a team from Switzerland and France – the authors explore the possible impacts of HPC techniques on a performance-intensive process in a piece of lens modeling software. The authors experiment with different precision levels and the choice of CPU versus GPU. The authors propose a “mixed precision” algorithm combined with consumer GPUs to achieve the best results.
Authors: Markus Rexroth, Christoph Schäfer, Gilles Fourestey and Jean-Paul Kneib.
Do you know about research that should be included in next month’s list? If so, send us an email at [email protected]. We look forward to hearing from you.