September 22, 2022 – The University of Utah announced the establishment of a new oneAPI Center of Excellence focused on developing portable, scalable, and high-performance data compression techniques.
The oneAPI Center is led from the University of Utah’s Center for Extreme Data Management Analysis and Visualization (CEDMAV) and conducted in collaboration with the Lawrence Livermore National Laboratory’s Center for Applied Scientific Computing (CASC). It accelerates ZFP compression software using oneAPI’s open, standards-based programming on multiple architectures to power exascale computing.
Participants said the center’s efforts expand the long-standing collaboration of the organizations dedicated to developing advanced data formats and layouts for efficient storage and providing access to big scientific data for High Performance Computing (HPC) architectures.
“The University of Utah’s CEDMAV, in collaboration with LLNL’s CASC, pioneered research into managing extreme data applications involving scientific simulations and experimental facilities,” said Manish Prashar, director of the Scientific Computing and Imaging Institute at the University of Utah. “This collaboration has a long track record of developing and delivering open source scientific software that is widely accepted by communities of interest. This oneAPI Competence Center will strengthen this collaboration and help this academic research find practical application on multi-architecture systems.”
Developed by LLNL, ZFP is a state-of-the-art software for lossless and error-controlled lossy compression of floating-point data, which is becoming a de facto standard in the HPC community with numerous scientific and technical applications and users. ZFP (de)compression lends itself particularly well to data-parallel execution because it is broken up into small, independent data chunks, and parallel backends have been developed for OpenMP, CUDA, and HIP programming models, according to LLNL computer scientist Peter Lindstrom.
“As director of NDT development, I am excited to have this opportunity with our longtime collaborators at the University of Utah to extend the capabilities of our NDT compressor to run efficiently on next-generation supercomputers, including the Argonne’s Aurora system National Laboratory, one of the world’s first exascale systems,” Lindstrom said. “The resulting compression software will enable large scientific computing applications, among others, to effectively increase storage capacity and bandwidth while significantly reducing communication and I/O time and offline storage.”
With LLNL’s ZFP development team, the oneAPI Competence Center will develop a SYCL-based, portable, scalable and high-performance ZFP backend running on accelerator architectures from various vendors, including Intel data center GPUs. As one of the software technologies selected by the Department of Energy’s (DOE) Exascale Computing Project (ECP), NDT is being adopted by massively parallel simulations and technologies running on some of the world’s largest supercomputers, benefiting multiple high-visibility scientific applications becomes. In addition, the widespread adoption of ZFP in industry and academia will help drive many large-scale data management technologies, including HDF5, ADIOS, OpenZGY, OpenVisus, and Zarr.
The development of a high-performance ZFP SYCL port on accelerator architectures supporting multiple vendors will benefit multiple high-visibility supercomputing applications and better demonstrate the power of an open, standards-based software ecosystem.
“The work of the University of Utah and Lawrence Livermore National Laboratory in developing a high-performance SYCL-based ZFP library supports the availability of rich scientific data for high-performance computing architectures and enables exascale applications targeting multiple accelerator architectures,” said Scott Apeland, senior director of Intel Developer Ecosystem programs. “This latest Center of Excellence will demonstrate how the developer community benefits from open, standards-based oneAPI development.”
CEDMAV’s research approach is based on a systematic assessment of HPC application requirements and how they lead to new investigations and innovations, followed by practical validation and deployment in wider communities. CEDMAV’s previous collaborations with LLNL include joint research projects, dual-occupational staff, student interns and postdocs.
“It is an honor for CEDMAV to set up this oneAPI center of excellence in cooperation with LLNL. This presents a great opportunity to consolidate and expand our collaboration with the support and collaboration of Intel engineers,” said Valerio Pascucci, founding director of CEDMAV and former group leader for CASC data analysis at LLNL. “It is exciting to see the emergence of the oneAPI programming model, which we aim to fully incorporate in this project. In particular, the cross-platform abstraction of SYCL will enormously increase the productivity of our teams when creating high-performance code that runs efficiently on modern, heterogeneous architectures. Diverse hardware-software architectures are becoming ubiquitous in high-performance systems, and oneAPI technology will greatly increase the impact of ZFP in a wide range of applications.”
The University of Utah’s CEDMAV is internationally recognized for its activities, which include theoretical and algorithmic research, systems development, and the use of tools to deal with extreme data. This research lies at the intersection of scientific visualization, big data management, HPC and data analysis.
The Center for Applied Scientific Computing serves as the LLNL’s window to the broader research communities in computer science, computational physics, applied mathematics and data science. With academic, industrial, and other government laboratory partners, it conducts world-class scientific research and development on issues critical to national security.
oneAPI is an open, unified, cross-architecture programming model for CPUs and accelerator architectures (GPUs, FPGAs and others). The standards-based programming model simplifies software development and delivers uncompromised performance for accelerated computing without proprietary lock-in, while allowing integration with existing code. With oneAPI, developers can choose the best architecture for the specific problem they are trying to solve without having to rewrite software for the next architecture and platform.
Source: Jeremy Thomas, LLNL