Performance and Productivity Advantages of Partitioned Global Address Space Programming
May 24, 2007
As high end computing moves towards a petascale with many cores per chip, there are concerns that two-sided message passing will not be the best programming model. Unified Parallel C (UPC) is an example of a Partitioned Global Address Space (PGAS) language, originally developed for ease of programming. I this talk I will give a short overview of the programming features of UPC and describe some of the performance advantages of UPC due its use of one-sided communication. One-sided communication allows for more effective use of network bandwidth and better overlap of communication with computations. The results show that, on machines with RDMA support, UPC outperforms optimized MPI code on both microbenchmarks and some applications.
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. For more information about computing sciences at Berkeley Lab, please visit www.lbl.gov/cs.
