Better, Faster, and More Scalable: The March to Exascale

According to the Exascale Computing Project (ECP), exascale supercomputers will process a quintillion (1018) calculations each second, more realistically simulating the processes involved in precision, compute-intense usages such as medicine, manufacturing, and climate sciences. And with ECP’s goal to launch such a computationally powerful ecosystem in the U.S. by 2021, the march to Exascale is more like a sprint.

Essential to this type of implementation is the development, tuning, and scaling of MPI* applications—providing more nodes with more cores and more threads, all interconnected by high-speed fabric. (Note, the MPICH* source base from Argonne National Labs—the basis for Intel® MPI Library—has been updated.)

In this webinar, overcoming MPI inefficiencies at scale is addressed, including:

  • How the Intel MPI Library optimizes the MPICH source
  • How Intel® Application Performance Snapshot—part of Intel® VTune™ Amplifier—can help you quickly understand how your distributed and shared memory applications are performing, and where you can focus your optimization efforts
  • A technical preview of Intel MPI Library 2019 that demonstrates architecture improvements and enhanced support for hybrid programming models
Dmitry Durnov, Software Development Engineer, Intel Corporation

Dmitry is a senior software engineer in the Intel® MPI team at Intel Corporation. He is one of lead developers and his current main focus is a full stack Intel® MPI product optimization for new Intel platforms (Intel® Xeon® Scalable processors, Intel® Xeon Phi™ and Intel® Omni-Path Architecture).

For more complete information about compiler optimizations, see our Optimization Notice.