Zambre Wins Outstanding Doctoral Dissertation Award for Supercomputing Communication

Rohit Zambre’s award-winning dissertation analyzes the problem of supercomputing applications’ slow multithreaded communication and eliminates the communication bottleneck.

July 30, 2021 – Helping supercomputers communicate efficiently was the aim of Rohit Zambre’s doctoral dissertation, which has won the Association for Computing Machinery’s Special Interest Group on High Performance Computing (SIGHPC) 2021 Outstanding Doctoral Dissertation Award.

The award is given each year for the best doctoral dissertation completed in high performance computing (HPC) in the previous year. Zambre earned his doctorate in computer engineering from UC Irvine in 2020 and currently works at AMD Research as an HPC architecture researcher in Washington.

"This award validates all the efforts – not just my own, but also those of my adviser and collaborators – that went toward navigating my Ph.D. journey,” said Zambre. “I’m honored to have been recognized by the ACM SIGHPC for my research contributions. This award strongly motivates me to use the research training I have received to continue my pursuit for impactful research on the most relevant problems as I progress in my career.”

Zambre’s dissertation is titled “Exascalable Communication for Modern Supercomputing.” It analyzes the problem of supercomputing applications’ slow multithreaded communication and eliminates the communication bottleneck by bridging the two ends of the HPC stack – message passing interface (MPI) library developers and domain experts – that typically do not directly talk to each other.

Supercomputing applications play a critical role in solving complex problems. They can range from scientific ones, such as modeling the human brain, to pressing day-to-day problems, such as climate change and the COVID-19 pandemic. The computing hardware that these software applications run on has been continuously evolving to achieve the next frontiers in computing throughput. It is not trivial for an application to capitalize on the capabilities of modern hardware since there are many layers of abstraction in the HPC software stack, explained Zambre. 

“My research focused on enabling applications to utilize the capabilities of modern network hardware since communication between the nodes of a supercomputer occupies a significant portion of an application’s runtime at scale,” said Zambre. “Through collaborations with a variety of system experts and domain scientists, we tackled the problems at each layer of the software stack, and ultimately enabled applications – targeted to run on the upcoming exascale supercomputers – to efficiently utilize modern network hardware and achieve significant performance improvements. The technologies from our research have been incorporated into the most widely used communication library (MPICH) in supercomputing.”

"Rohit's work makes foundational contributions to the problem of slow multithreaded communication, and he demonstrates the applicability of the proposed solutions in real-world supercomputing applications at scale,” said Aparna Chandramowlishwaran, Zambre’s adviser and associate professor of electrical engineering and computer science. “I'm absolutely thrilled to see him win this year's ACM SIGHPC Outstanding Doctoral Dissertation Award for his exemplary work. It was a pleasure to work with Rohit, and I'm sure he'll be a role model for future HPC researchers."

The award is open to students studying anywhere in the world who have completed a Ph.D. dissertation with HPC as a central research theme. Nominations are evaluated on technical depth, the significance of the research contribution, the potential impact on theory and practice, and overall quality of work. It includes a $2,000 cash prize, a plaque and recognition at the International Supercomputing Conference in November.

“The efforts of our work mean faster and more efficient runs on modern supercomputers,” said Zambre. “I hope this means that domain scientists can provide us with answers to the complex problems we face sooner rather than later.”

– Tonya Becerra