DK Panda and Karen Tomko
Ohio Supercomputer Center
The fields of computer architecture, interconnection networks and system design are undergoing rapid change enabling very large supercomputers to be built. System advances have come in the form of increased parallelism from many-core accelerators and improved communication interfaces. The MPI library is also changing with several new capabilities introduced to enable highly scalable parallel programs on such systems. This three-day webinar series introduces several of these capabilities.
Thursday, May 8 - MPI for many-core processors
• MVAPICH2-GPU: CUDA-aware MPI for nVidia GPU
• MVAPICH2-MIC design for clusters with Infiniband and Intel Xeon Phi