Pre-exascale accelerated application development: The ORNL Summit experience

Journal Article · Fri May 01 00:00:00 EDT 2020 · IBM Journal of Research and Development

DOI:https://doi.org/10.1147/jrd.2020.2965881· OSTI ID:1649509

Luo, L. ^[1]; P. Straatsma, T. ^[2]; Suarez, L. Aguilar ^[3]; Broer, R. ^[3]; Bykov, D. ^[2]; D'Azevedo, E. F. ^[2]; S. Faraji, S. ^[3]; C. Gottiparthi, K. ^[2]; De Graaf, C. ^[4]; A. Harris, J. ^[2]; A. Havenith, R. W. ^[3]; Jensen, H. Aa. ^[5]; Joubert, W. ^[2]; K. Kathir, R. ^[3]; Larkin, J. ^[6]; W. Li, Y. ^[7]; I. Lyakh, D. ^[2]; B. Messer, O. E. ^[2]; R. Norman, M. ^[2]; C. Oefelein, J. ^[8] ; Sankaran, R. ^[2]; F. Tillack, A. ^[2]; L. Barnes, A. ^[2]; Visscher, L. ^[9]; C. Wells, J. ^[2]; Wibowo, M. ^[3]

---

1. IBM Research, Oak Ridge, TN (United States)
2. Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
3. Univ. of Groningen (The Netherlands)
4. Univ. Rovira i Virgili, Tarragona (Spain)
5. Univ. of Southern Denmark, Odense (Denmark)
6. Nvidia Corporation, Oak Ridge, TN (United States)
7. Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
8. Georgia Inst. of Technology, Atlanta, GA (United States)
9. Univ. of Amsterdam (Netherlands)

High-performance computing (HPC) increasingly relies on heterogeneous architectures to achieve higher performance. In the Oak Ridge Leadership Facility (OLCF), Oak Ridge, TN, USA, this trend continues as its latest supercomputer, Summit, entered production in early 2019. The combination of IBM POWER9 CPU and NVIDIA V100 GPU, along with a fast NVLink2 interconnect and other latest technologies, pushes system performance to a new height and breaks the exascale barrier by certain measures. Due to Summit's powerful GPUs and much higher GPU–CPU ratio, offloading to accelerators becomes a requirement for any application, which intends to effectively use the system. To facilitate navigating a complex landscape of competing heterogeneous architectures, a collection of applications from a wide spectrum of scientific domains is selected for early adoption on Summit. In this article, the experience and lessons learned are summarized, in the hope of providing useful guidance to address new programming challenges, such as scalability, performance portability, and software maintainability, for future application development efforts on heterogeneous HPC systems.

• Research Organization: Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States)
• Sponsoring Organization: USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)
• Grant/Contract Number: AC05-00OR22725
• OSTI ID: 1649509
• Journal Information: IBM Journal of Research and Development, Vol. 64, Issue 3/4; ISSN 0018-8646
• Publisher: IEEE
• Country of Publication: United States
• Language: English

Related Subjects

97 MATHEMATICS AND COMPUTING
High performance computing
accelerated computing
GPU