MPAS-Ocean NESAP Status Report
NESAP performance improvements on MPAS-Ocean have resulted in a 5% to 7% speed-up on each of the examined systems including Cori-KNL, Cori-Haswell, and Edison. These tests were configured to emulate a production workload by using 128 nodes and a high-resolution ocean domain. Overall, the gap between standard and many-core architecture performance has been narrowed, but Cori-KNL remains considerably under-performing relative to Edison. NESAP code alterations affected 600 lines of code, and most of these improvements will benefit other MPAS codes (sea ice, land ice) that are also components within ACME. Modifications are fully tested within MPAS. Testing in ACME across many platforms is underway, and must be completed before the code is merged. In addition, a ten-year production ACME global simulation was conducted on Cori-KNL in late 2016 with the pre-NESAP code in order to test readiness and configurations for scientific studies. Next steps include assessing performance across a range of nodes, threads per node, and ocean resolutions on Cori-KNL.
- Research Organization:
- Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
- Sponsoring Organization:
- USDOE Office of Science (SC). Advanced Scientific Computing Research (ASCR) (SC-21)
- DOE Contract Number:
- AC52-06NA25396
- OSTI ID:
- 1352437
- Report Number(s):
- LA-UR-17-23193
- Country of Publication:
- United States
- Language:
- English
Similar Records
Roofline Analysis in the Intel® Advisor to Deliver Optimized Performance for applications on Intel® Xeon Phi™ Processor
Preparing NERSC users for Cori, a Cray XC40 system with Intel many integrated cores