In this video, Jeff Kuehn from LANL, Pavel Shamis from ARM, and Gilad Shainer from Mellanox describe progress on the new UCF consortium, a collaboration between industry, laboratories, and academia to create an open-source production grade communication framework for data centric and high-performance applications.
The first project for the Consortium is UCX, which unites the strengths and capabilities of MXM (Mellanox), PAMI (IBM) and UCCS (Oak Ridge National Laboratory) programming API and provides support for all of the leading communication libraries, including MPI, SHMEM/PGAS and UPC. This unified communication framework encourages co-design between software and hardware, delivers the building blocks essential for the development of a high-performance communication ecosystem that is critical to power tomorrow’s leading HPC systems.
“The path to Exascale, in addition to many other challenges, requires programming models where communications and computations unfold together, collaborating instead of competing for the underlying resources,” said George Bosilca, research director, Innovative Computing Laboratory, University of Tennessee, Knoxville. “In such environments, providing holistic access to the hardware is a major component of any programming model or communication library. With UCX, we now have a vehicle for production quality software, a flexible and efficient low-level research infrastructure for dynamic and portable support for future-ready programming models.”
UCX provides the fine grain flexibility for researchers to customize and adjust the communication software framework for their unique and specific needs,” said Pavel Shamis, Principal Research Engineer at Arm. “Co-design was critical in order to make this type of individual optimization possible. Through the UCX framework researchers have the ability to influence hardware architecture and can learn about new features or capabilities of the hardware, which, in turn, enables the development of Exascale programming models.”