Computing Conferences and Events
ALCF researchers regularly contribute to some of the world’s leading computing conferences and events to share their latest advances in areas ranging from computational science and AI to HPC software and exascale technologies. In 2022, Argonne staff participated in a wide range of events including SC22, ISC High Performance, Grace Hopper Celebration, SIAM Conference on Computational Science and Engineering, Richard Tapia Celebration of Diversity in Computing Conference, IEEE International Parallel & Distributed Processing Symposium, International Conference on Parallel Processing, International Symposium on Cluster, Cloud and Grid Computing, International Workshop on OpenCL and SYCL, Platform for Advanced Scientific Computing Conference, HPC User Forum, Energy High Performance Computing Conference, Lustre User Group Conference, Intel eXtreme Performance Users Group Conference, Conference on Machine Learning and Systems, and more.
Exascale Computing Project
DOE’s Exascale Computing Project (ECP) is a multi-lab initiative aimed at accelerating the delivery of a capable exascale computing ecosystem. Launched in 2016, the ECP’s mission is to pave the way for the deployment of the nation’s first exascale systems by building an ecosystem that encompasses applications, system software, hardware technologies, architectures, and workforce development. Researchers from the ALCF and across Argonne—one of the six ECP core labs—are helping the project deliver on its ambitious goals. The laboratory has a strong presence on the ECP leadership team and has several researchers engaged in ECP projects and working groups focused on application development, software development, and hardware technology. In the workforce development space, the ECP continues to fund the annual Argonne Training Program on Extreme-Scale Computing (ATPESC), which is organized and managed by ALCF staff.
HPC Standards and Community Groups
ALCF staff members remain actively involved in a number of HPC standards and community groups that help drive improvements in the usability and efficiency of scientific computing tools, technologies, and applications. Staff activities include contributions to the C++ Standards Committee, Cray User Group, HPC User Forum, Intel eXtreme Performance Users Group, Khronos OpenCL and SYCL Working Groups, MLPerf (HPC, Science, and Storage Working Groups), MPI Forum, NITRD Middleware and Grid Infrastructure Team, oneAPI Steering Committee, oneAPI Special Interest Groups (Hardware, Language, and Math), OpenMP Architecture Review Board, OpenMP Language Committee, Open Scalable File Systems (OpenSFS) Board, and SPEC High-Performance Group.
The ALCF continued its collaboration with NERSC and OLCF to operate and maintain a website dedicated to enabling performance portability across the DOE Office of Science HPC facilities. The website serves a documentation hub and guide for applications teams targeting systems at multiple computing facilities. Staff from the DOE computing facilities also collaborate on various projects and training events to maximize the portability of scientific applications on diverse supercomputer architectures.
The ALCF works closely with many companies in the HPC and AI industries to develop and deploy cutting-edge hardware and software for the research community. This includes collaborating with Intel and HPE to deliver the Aurora exascale system, working with HPE to deploy the Polaris testbed supercomputer, and partnering with NVIDIA on system enhancements and training related to ThetaGPU. Such partnerships are critical to ensuring the facility’s supercomputing resources meet the requirements of the scientific computing community. In addition, the ALCF is working with several AI start-up companies, including Cerebras, Graphcore, SambaNova, Groq, and Habana, to deploy a diverse set of AI accelerators as part of the ALCF AI Testbed. The testbed, which opened up to the broader research community in 2022, is playing a key role in determining how AI accelerators can be applied to scientific research, while also allowing vendors to prepare their software and hardware for scientific AI workloads.