oneAPI, oneCCL and OFI: Path to Heterogeneous Architecture Programming with Scalable Collective Communications
Sayantan Sur, Intel, Corp.
Sayantan Sur is a Principal Engineer at Intel Corporation. He is an expert in MPI, Fabrics and High-Performance computing. He is currently focusing on the oneAPI stack for Intel’s Xe GPUs. Sayantan is well known in the Fabrics community for his work in scaling MPI implementations over InfiniBand. His work enabled MPI to scale from a few InfiniBand connected nodes to thousands. He has published more than 30 papers in major conferences related to these research areas and delivered several tutorials at conferences such as Supercomputing and Hot Interconnects. Sayantan received his Ph.D. degree from The Ohio State University in 2007.
Hetergeneous architectures are becoming more prevalent as a mainstream compute resource. The challenges of having disparate programming models, tools and workflows of development is difficult especially as there are no common standards or broad community efforts to get to a solution. oneAPI is an industry initiative that Intel is driving along with the community to develop standards-based programming models along with libraries that focus on delivering cross architecture support. Within oneAPI, the oneCCL Collective Communication Library is designed to offer easy integration of high-performance collective communication patterns into popular Deep Learning Frameworks. In this talk, we will describe the how the combination of oneAPI, oneCCL and OFI provide a pathway towards portable heterogenous architecture programming with performance.
Watch video oneAPI, oneCCL and OFI: Path to Heterogeneous Architecture Programming with Scalable Collective... online without registration, duration hours minute second in high quality. This video was added by user insideHPC Report 22 June 2020, don't forget to share it with your friends and acquaintances, it has been viewed on our site 45 once and liked it people.