Two faculty members from the Georgia Institute of Technology College of Computing are contributing to two exascale projects receiving funding from the U.S. Department of Energy’s Exascale Computing Project (ECP).
The three-year, multi- institution projects will receive just over $6.5 million in total funding to be distributed among each organization involved. Supported by School of Computer Science Associate Professor Ada Gavrilovska and Research Scientist Greg Eisenhauer, the projects deal with specific components of software stack development for exascale systems. Collectively, these software projects are the first steps to developing parallel applications that can target distinct exascale architectures.
Gavrilovska supports “A Simplified Complex Memory API and Operating System/Runtime Interface for ECP.” The project addresses the arbitrary process of programming for multiple memory technologies, such as MCDRAM, PCI-E NVM, SATA NVM, 3D stacked memory, PCM, Memristor, and 3D XPoint. The research team plans to devise a unified two-tier, unit-level Application Program Interface (API) for developers. This new memory API will offer scalable control over each memory type, providing an urgently needed tool for the supercomputing community.
The second project, “The ADIOS Framework for Scientific Date on Exascale Systems,” tackles another pressing challenge. Exascale systems are expected to increase computational speed beyond what current input/output (I/O) bandwidth can handle. This creates a time-sensitive need to optimize reading and writing speeds for exascale systems. Eisenhauer’s team aims to satisfy this need by transforming the U.S. Department of Energy’s ADIOS program to maximize I/O bandwidth and efficiently use exascale hardware.
“A significant part of the promise of exascale computing systems lies in the delivery of large-scale simulations and other computations that uses the performance potential of these next-generation machines,” Eisenhauer said. “That can’t happen without the software and system-level infrastructure that is being developed in this program.”
These two software development projects highlight a significant push by the College of Computing to be a vital contributor to the nation’s exascale ambitions. This push includes recent contributions from the School of Computational Science and Engineering (CSE) in related supercomputing projects, such as CSE Chair David Bader’s involvement with the National Strategic Computing Initiative and CSE Professor Edmond Chow’s $2.4 million DOE grant.
In all, 35 software development proposals have received a total of $34 million in grants. Each project confronts a critical research topic in the transition to exascale computing. The awards presented to these projects are a part of the U.S. government’s initiative to build an exascale supercomputer within the decade.