One of the more compelling HPC community-centered questions that tends to arise is how to snag and maintain interest in supercomputing among the younger ranks of researchers. With opportunities at large companies with massive datacenter and coding needs, it can be difficult for research centers to cling to young talent.
To demonstrate that there is plenty of interest from the younger set, this summer at the International Supercomputing Conference in Leipzig, Germany, a special segment will be devoted to young researchers in high performance computing. Conference organizers are bolstering interest in what’s happening with emerging researchers (in addition to the activity around the cluster competition and tutorial and other resources for young researchers) by highlighting the work of three select young researchers from around the world.
Among those hand-picked to present the next wave of HPC research is Dr. Ian Karlin, a computer scientist in the Development Environment Group at Lawrence Livermore National Laboratory has spent the early part of his career modeling memory systems for linear algebra computation, but now, at LLNL, he remains focused on strategies for future computing challenges on large integrated multi-physics codes.
Karlin’s talk will focus on a topic fresh in the minds of many, especially with the upcoming new release of the next Top 500 list—the question of whether measuring FLOPS or bandwidth is more valuable—or if a new benchmark paradigm is needed altogether. As Karlin describes, the performance bottleneck for HPC applications continues to change with new system innovations. “For implicit codes, FLOPs were the first performance bottleneck, while more recently memory bandwidth has limited performance. However, many codes use other methods to solve PDEs and for energy transport. These methods, such as Monte Carlo and explicit hydrodynamics have performance bottlenecks that are different than linear algebra based applications.”
Karlin will present data showing the many machine design vectors that must be balanced for general purpose exascale HPC systems. He also plans to demonstrate that latency to memory needs to be more strongly considered in system design and that the balance of compute capabilities in today’s HPC microprocessors “make it impossible for many modern applications to be limited by floating-point capability even if the bandwidth constraints on modern systems were solved.”
On the theme of keeping up with ever-larger system capabilities, another young researcher, Hiroyuki Takizawa, from Tohoku University, will offer an overview of the evolutionary adaption of HPC applications to large system changes—an important topic as applications are prepared to meet the exascale era.
Takizawa will reiterate that since applications cannot constantly change to meet new architectures or system capabilities, there is a need for a systematic way to help software evolve more dynamically. To do this, he proposes a separation between system-specific application optimization and what is more broadly useful for applications to more naturally evolve with system enhancements—versus constantly playing a game of “catch up” to the latest architectures.
Adding to this lineup of young HPC researchers is Dr. Jens Zumbrägel, Marie Curie Postdoctoral Fellow at Technische Universität in Dresden, who will put some of this work in practical context in a cryptographic research frame. He plants to present his work on one of the most famous numerical theory issues for cryptographic systems, the discrete logarithm problem. He plans to highlight the work that’s been done in this area over the last 20 years, noting that there has been very little innovation since it first appeared on the security and cryptographic front. Zumbrägel will describe how some new methods to tackling the problem have pushed significant cryptanalysis record computations forward, showcasing his ideas in real-world context.
The session that puts these three on stage takes place on Monday–we’ll make sure to follow up with all three during our live coverage of the event to see what questions the community found most compelling about their work.