This fall a diverse group of several hundred engineers, scientists, programmers and system administrators will converge on the Intel campus in Hillsboro, Oregon for the Intel Extreme Performance Users Group (IXPUG) conference.
Despite the fact that the attendees represent a wide variety of organizations and job functions, they all have one thing in common: they want to explore how to get optimum performance from Intel’s new many-core systems.
This is the semi-annual meeting of the Intel Extreme Performance Users Group (IXPUG), the independent users group that was formed four years ago. The IXPUG conference is focused on all aspects of employing and adopting many-core processing technologies and techniques for optimal application execution. The conference also covers system hardware beyond the processor (memory, interconnect, etc.), software tools, programming models, new workloads (visualization, data analytics, machine learning, etc.) and more. The conference will provide an interactive experience focused on key topics associated with high-performance computing, artificial intelligence, and cloud computing.
The conference is being co-organized by Sandia National Laboratories and Los Alamos National Laboratory, both IXPUG charter members. Although at the time of this writing the final conference program was still being developed, you can get more insights into the event content at IXPUG.org.
IXPUG held its first meeting in 2014 at the Texas Advanced Computing Center (TACC) in Austin TX, which had an Intel® Xeon Phi™ coprocessor based machine. At the time the community was known as the Intel Xeon Phi Users Group.
“TACC had one of the first large installations of the Intel Xeon Phi coprocessor,“ recalls David Martin, president of the IXPUG steering committee and manager of industry partnerships and outreach at the Argonne Leadership Computing Facility (ALCF), a DOE Office of Science User Facility.
“Like any other technology product, there was a manual that came with the merchandise,“ Martin says. “But it didn’t really tell you how this complex CPU actually works when it’s installed and interacting within a variety of other systems. TACC realized they had to talk not only with Intel, but many other users in the HPC community to figure out this new technology.”
This required more than a one-way communication between Intel and the HPC center and applications scientists, the IXPUG web site notes. It involved building a community dedicated to making the most of the powerful new processor.
Seeing the wide interest in IXPUG, TACC helped put together a steering committee composed of representatives from HPC sites around the world that were building supercomputers powered by Intel processors. The steering committee wrote a charter and the community took off. Many hundreds of people have participated in events small and large-from annual meetings in Europe and North America, to regional and single-topic workshops.
Pushing the Envelope
The steering committee found that the scientists and engineers who were investigating extreme scale computing were focusing on more than the CPU. They were exploring the capabilities and interconnections of the entire system.
“We were initially focused on the Intel Xeon Phi coprocessor,“ says Martin.
“When, Intel next introduced the Intel Xeon Phi processor targeting a much bigger, worldwide community, people began building larger systems and tackling tough scientific challenges and looked to IXPUG to help. Earlier this year, we changed the name of our organization to the Intel Extreme Performance User’s Group to emphasize our broader focus.”
Today user investigations include software tools and programming models, as well as new workloads such as visualization, data analytics, machine learning, and artificial intelligence.
Online Working Groups – Everyone is welcome
One of the ways IXPUG supports its members is to provide forums such as online working groups. These are open meetings – anyone who wishes to join is welcome.
For example, recent working group sessions were held to develop and share tribal knowledge of how to make the most of Intel® Xeon® Scalable processors. And later this month, a group will meet to discuss how to apply deep learning to problems in cosmology. Two computer scientists at NERSC will lead the session titled “Machine Learning at Scale” that, among other things, will explore how to develop a deep learning-based emulator for cosmological observables that can reduce the need for computationally expensive simulations.
Also of interest to the IXPUG community is the Intel® Omni-Path Architecture (Intel® OPA), a next generation fabric that scales from entry-level HPC clusters to clusters with 10,000 nodes or more. Intel OPA contains enhancements specifically designed to help supercomputer architects move toward Exascale. In coordination with the Omni-Path Users Group (OPUG), the technology will be one of the many topics discussed at this year’s IXPUG Annual Fall Conference.
And several years from now, when Intel releases the nation’s first exascale system, IXPUG will provide the venues and organization needed for its members to meet and assist one another. This group of dedicated individuals will help the HPC community make the most of the next revolution in supercomputer technology.