NERSCPowering Scientific Discovery Since 1974

NERSC / Cray Center of Excellence

| Tags: Math & Computer Science

COE.png

The figure shows a plot of runtime vs. MPI and OpenMP cores for the NERSC PARATEC benchmark. Red=time in ZGEMM; Blue=time in FFT; Green=total memory required.

Key Challenges: Determining the best ways to use massive on-chip parallelism expected in future systems and refine "best
practices" in parallel programming and I/O techniques to fully exploit the performance potential of current systems, especially the NERSC Hopper XE6.  Applying advanced performance modeling methods to characterize system performance and application requirements so as to use the NERSC workload requirements to drive changes in computing architecture that will result in better HPC system architectures for scientific computing in future generation machines. 

Why it Matters:  Although Moore’s Law continues unabated, power constraints mean that cores will double every 18 months, not clock speed, as has been the case for 20 years.  Memory capacity is not doubling at the same rate so that GB/core will decrease.  Users may be faced with the need to change the way they program parallel machines in order to avoid performance flatlining.   The memory hierarchy is deepening, as well.  

The NERSC/Cray “Programming Models Center of Excellence” combines Berkeley Lab strength in advanced programming models, multicore tuning, and application benchmarking with Cray strength in advanced programming models, optimizing compilers, and benchmarking.

Accomplishments: Several performance studies in collaboration with benchmarking experts at Cray Inc. have been done to understand the performance of hybrid programming models on the franklin and hopper.  The results have been presented at NERSC User Group training sessions and have been adopted for inclusion in NERSC User Services Group documentation. 

Investigators: Nicholas J Wright and John Shalf (NERSC); Hongzhang Shan, Tony Drummond,and Andrew Canning (LBNL); Marcus Wagner, Sarah Anderson,Ryan Olsen and Mike Aamodt (Cray)

More Information: See, for example, Nicholas Wright, et al., "The NERSC- Cray Center of Excellence: Performance Optimization for the Multicore Era," Proceedings of the 2011 Cray User Group Conference, May 23-26, 2011. 

 


About NERSC and Berkeley Lab
The National Energy Research Scientific Computing Center (NERSC) is the primary high-performance computing facility for scientific research sponsored by the U.S. Department of Energy's Office of Science. Located at Lawrence Berkeley National Laboratory, the NERSC Center serves more than 4,000 scientists at national laboratories and universities researching a wide range of problems in combustion, climate modeling, fusion energy, materials science, physics, chemistry, computational biology, and other disciplines. Berkeley Lab is a U.S. Department of Energy national laboratory located in Berkeley, California. It conducts unclassified scientific research and is managed by the University of California for the U.S. DOE Office of Science. »Learn more about computing sciences at Berkeley Lab.