UT’s Innovative Computing Laboratory (ICL) will participate in six of the 35 software development awards and one of four co-design center awards for the US Department of Energy’s Exascale Computing Project (ECP).
The project is aimed at developing systems at least 50 times faster than the nation’s most powerful current supercomputers.
Doug Kothe, ECP director of application development, explained that co-design entails an interchange of the best that hardware technologies, software technologies, and applications have to offer each other.
Kothe said such work will be a catalyst for delivery of exascale-enabling science and engineering solutions for the United States. Exascale systems are capable of one billion billion computations—1,000,000,000,000,000,000—per second.
“By targeting common patterns of computation and communication known as application motifs, we are confident that these ECP co-design centers will knock down key performance barriers,” said Kothe. “This will pave the way for applications to exploit all that capable exascale has to offer.”
For its role, the ICL—which is housed in UT’s Tickle College of Engineering—will receive about $3.3 million in funding the first year and more than $3.4 million in each of the next two years for an overall total of approximately $10.2 million. First-year funding for all the ECP software development awards totals $34 million.
“ECP is a very important project for our country, and we’re grateful for the opportunity to be able to contribute to it,” said UT Distinguished Professor and ICL Director Jack Dongarra. “This vote of confidence that DOE has given us is a product of many years of hard work by the members of the Innovative Computing Laboratory to deliver solid open-source software to the research community, industry, and society in general.”
Three of the software development awards will fund projects conducted solely by ICL, while the other three will involve collaborations with DOE laboratories and other academic institutions.
The projects that the ICL will conduct independently are titled Software for Linear Algebra Targeting at Exascale (SLATE), Distributed Tasking for Exascale, and the Exascale Performance Application Programming Interface (EXA-PAPI).
During the course of nearly 10 years, Dongarra and ICL research team members applied algorithmic and technological innovations to pioneer, implement, and disseminate dense linear algebra software. The SLATE project will converge and consolidate that software into a dense linear algebra library that will integrate seamlessly into the ECP ecosystem.
The Distributed Tasking for Exascale project will transition a generic framework developed by the ICL, called the Parallel Runtime Scheduling and Execution Controller, to exascale supercomputers, while extending it to further facilitate programmability and increase scientific productivity.
EXA-PAPI will perform two fundamental roles within ECP. One of those will be to assist in finding the right balance between performance and power efficiency at the appropriate software and hardware levels. The other major role will be to enable what is referred to as cross-layer and integrated modeling and analysis of the entire ECP hardware–software ecosystem.
Sandia National Laboratories (SNL) and the ICL will collaborate on an ECP project named Production-ready, Exascale-enabled Krylov Solvers for Exascale Computing. It will explore the redesign of pieces of mathematical software, known as solvers, to address new challenges in supercomputing.
Another of the joint projects—the Extreme-scale Scientific Software Kit for the ECP, led by multiple DOE laboratories—will work to enable the seamless combined use of diverse independently developed software packages as needed by ECP applications. In addition to the ICL, partners in this project are Argonne National Laboratory; Lawrence Berkeley National Laboratory; Lawrence Livermore National Laboratory (LLNL); SNL; and the University of California, Berkeley.
A standard system known as MPI (message passing interface), used in a variety of parallel computing architectures and available in many open-source or public-domain implementations, needs to be made ready for exascale. To address the associated challenges, Oak Ridge National Laboratory, Los Alamos National Laboratory, LLNL, SNL, and the ICL will collaborate on a project named Open MPI for Exascale.
The co-design center of which ICL will be a part is the Center for Efficient Exascale Discretizations (CEED), led by LLNL. CEED’s focus will be on developing the next-generation discretization software and algorithms that will enable a wide range of DOE and National Nuclear Security Administration (NNSA) applications such as global climate modeling, turbulent combustion in internal combustion engines, nuclear reactor modeling, additive manufacturing, subsurface flow, and national security applications to run efficiently on future hardware. Working with LLNL and the ICL on CEED will be Argonne; the University of Illinois at Urbana-Champaign; Virginia Tech; Colorado University, Boulder; and the Rensselaer Polytechnic Institute.
The overall ECP pairs in collaboration two DOE organizations—the Office of Science and the NNSA. As part of President Barack Obama’s National Strategic Computing Initiative, ECP was established to develop a capable exascale ecosystem, encompassing applications, system software, hardware technologies and architectures, and workforce development, to meet the scientific and national security mission needs of DOE in the early 2020s. ECP is expected to last 10 years and receive $1 billion in funding.