LLNL Collaboration with the University of Utah API Center of Excellence

zfp adaptation 092122

The brand new API Heart of Excellence on the College of Utah will embody collaboration with the Lawrence Livermore Nationwide Laboratory Heart for Utilized Scientific Computing and speed up ZFP compression to advance Exascale computing. Developed by LLNL, ZFP is a contemporary lossless and error-controlled compression program for floating-point knowledge that has turn out to be a de facto customary within the HPC neighborhood. An illustration of ZFP’s means to vary the compression ratio to any desired setting is proven, from 10:1 on the left to 250:1 on the suitable, the place the splitting of the info set into small blocks is clear.

The College of Utah has introduced the creation of a brand new single API Heart of Excellence centered on creating extremely environment friendly, moveable, scalable knowledge compression applied sciences.

The oneAPI Heart can be led by the College of Utah Heart for Excessive Knowledge Administration Evaluation and Visualization (CEDMAV) and can embody the collaboration of the Lawrence Livermore Nationwide Laboratory for Utilized Scientific Computing (CASC). It would speed up ZFP compression utilizing open standards-based programming of oneAPI on a number of architectures to boost exascale computing.

Members stated the middle’s efforts develop the long-term collaboration of organizations devoted to creating superior knowledge codecs and layouts for environment friendly storage and offering entry to large-scale scientific knowledge for high-performance computing (HPC) architectures.

“The College of Utah CEDMAV, in collaboration with LLNL’s CASC, has pioneered analysis into managing excessive knowledge purposes that embody scientific simulation and experimental services,” stated Manish Prashar, director of the College of Utah’s Institute of Scientific Computing and Imaging. This collaboration has a confirmed monitor file in creating and publishing open supply scientific software program that finds extensive adoption within the communities of curiosity. This API’s Heart of Excellence will additional this collaboration and assist this tutorial analysis discover sensible adoption on a number of architectural methods. “

Developed by LLNL, ZFP is a state-of-the-art floating level knowledge loss managed compression program that has turn out to be a de-facto customary within the HPC neighborhood, with many scientific, engineering, and consumer purposes. ZFP(de) compression is especially amenable to parallel execution of information by decomposing it into small, unbiased knowledge blocks, and parallel backends have been developed for the OpenMP, CUDA, and HIP programming paradigms, in line with LLNL laptop scientist Peter Lindstrom.

“As a pioneer in ZFP growth, I’m enthusiastic about this chance with our longstanding companions on the College of Utah to increase the capabilities of our ZFP compressor to work effectively on next-generation supercomputers, together with Argonne Nationwide Laboratory’s Aurora system, one of many first Exascale methods on this planet,” Lindstrom stated. world.” “The ensuing compression program will permit for large-scale scientific computing purposes, amongst different issues, to successfully improve reminiscence capability and bandwidth whereas considerably lowering connectivity, I/O time, and offline storage.”

With the ZFP growth crew at LLNL, the API Heart of Excellence will develop a transportable, scalable, high-performance SYCL-based ZFP interface that runs on acceleration architectures throughout completely different distributors, together with Intel datacenter GPUs. As one of many software program applied sciences chosen by the DOE’s Exascale Computing Mission (ECP), ZFP has been licensed by large-scale parallel simulations and applied sciences working on a number of the world’s largest supercomputers, which can profit many high-definition scientific purposes. . Moreover, ZFP’s widespread adoption in business and academia will help the event of many large-scale knowledge administration applied sciences, together with HDF5, ADIOS, OpenZGY, OpenVisus, and Zarr.

The event of ZFP’s high-performance SYCL port on multi-vendor acceleration architectures will profit many high-definition supercomputing purposes and higher showcase the facility of an open, standards-based ecosystem.

“The work of the College of Utah and Lawrence Livermore Nationwide Laboratory in creating the SYCL-based high-performance ZFP library helps make large-scale scientific knowledge out there for high-performance computing architectures, enabling exascale purposes to focus on a number of accelerator architectures,” stated Scott Abeland, senior director. From Intel Developer Ecosystem Packages. “This newest Heart of Excellence will showcase how open, standards-based oneAPI growth advantages the developer neighborhood.”

CEDMAV’s analysis method stems from a scientific evaluation of the wants of HPC purposes and the way they result in new and progressive investigation, adopted by sensible validation and dissemination to broader communities. CEDMAV’s earlier collaborations with LLNL embody joint analysis tasks, dual-appointed employees, and pupil and postdoctoral interns.

“It’s a nice honor for CEDMAV to ascertain this API Heart of Excellence in collaboration with LLNL. This can give a fantastic alternative to consolidate and develop our collaboration with the help and collaboration of Intel engineers,” stated Valerio Pascucci, Founding Director of CEDMAV and former CASC Knowledge Evaluation Group Chief at LLNL Valerio Pascucci. “It’s thrilling to see the emergence of the oneAPI programming mannequin that we plan to totally embrace on this challenge. Particularly, SYCL’s cross-platform abstraction will significantly enhance the productiveness of our groups creating efficiency codes that may run effectively on trendy heterogeneous architectures. Numerous {hardware} and software program architectures have gotten ubiquitous in high-performance methods, and oneAPI know-how will significantly enhance the affect of ZFP in a variety of purposes.”


CEDMAV on the College of Utah is internationally acknowledged for its actions that embody theoretical and algorithmic analysis, methods growth, and the deployment of instruments for coping with excessive knowledge. This analysis lies on the intersection of scientific visualization, massive knowledge administration, HPC, and knowledge analytics.

About CASC

The Heart for Utilized Scientific Computing serves as a window to the LLNL College extra broadly within the laptop science, computational physics, utilized arithmetic, and knowledge science analysis communities. With tutorial, industrial, and different authorities laboratories companions, it conducts world-class scientific analysis and growth on points vital to nationwide safety.

About oneAPI

oneAPI is an open, unified, multi-architecture programming mannequin for CPUs and acceleration architectures (GPUs, FPGAs, and so forth.). Based mostly on the requirements, the programming paradigm simplifies software program growth and delivers unparalleled efficiency for accelerated computing with out proprietary locking, whereas enabling current code integration. With oneAPI, builders can select the most effective structure for the particular drawback they’re making an attempt to unravel with out having to rewrite this system for the subsequent structure and platform.

/Normal launch. This materials from the unique group(s)(s) could also be of a time limit nature, edited for readability, fashion and size. The opinions and opinions expressed are these of the creator(s). View it in full right here.