Abstract
MPI has been the de-facto programming model for scientific parallel applications. However, it is hard to extract the maximum performance for irregular data-driven applications using MPI. The Partitioned Global Address Space (PGAS) programming models present an alternative approach to improve programmability. The lower overhead in one-sided communication and the global view of data in PGAS models have the potential to increase the performance at scale. In this study, we take up ‘Concurrent Search’ kernel of Graph500 — a highly data driven irregular benchmark — and redesign it using both MPI and OpenSHMEM constructs. We also implement load balancing in Graph500. Our performance evaluations using MVAPICH2-X (Unified MPI+PGAS Communication Runtime over InfiniBand) indicate a 59% reduction in execution time for the hybrid design, compared to the best performing MPI based design at 8,192 cores.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
