This new center will use oneAPI capabilities to advance energy-efficient computing for material sciences and life sciences codes.
The Zuse Institute Berlin (ZIB) has formed an Intel oneAPI Center of Excellence focused on using oneAPI cross-architecture programming for energy-efficient HPC computing by delivering portable implementations on GPUs and FPGAs.
With the Intel oneAPI Center of Excellence, ZIB is extending its long-lasting collaboration with Intel after establishing the Intel® Parallel Computing Center in 2013. To foster portable applications on accelerator architectures across different vendors, oneAPI’s open, unified programming framework was chosen as it supports different levels of abstraction and a standardized language support for CPUs, GPUs and FPGAs. As energy costs are increasingly critical for operating HPC resources, ZIB is particularly interested in exploring oneAPI’s capabilities for energy-efficient computing with two selected workloads from material sciences and life sciences, respectively.
Accelerating Material and Life Sciences Codes
The Algorithms for Innovative Architectures (A4IA) group of the Supercomputing department and the NHR Center at ZIB are working on the migration and optimization of the VASP code, one of the key workloads for material science studies worldwide for faster atomistic simulations of materials, on Intel Xe architecture GPUs. This continues collaborations with experts from the VASP developer team at the University of Vienna and the VASP Software GmbH.
Utilizing ZIB’s expertise with custom computing in life science, the A4IA group is developing portable and fast implementations of SeqAn library functions with oneAPI to enable energy-efficient solutions of genomic analysis workflows on FPGAs and CPUs. The SeqAn library, which is widely-used around the word, is developed by the Reinert group at the Free University Berlin, our partner in a related project.
The research and development on efficient designs and implementations in real-world workloads with oneAPI are accompanied by teaching and dissemination activities. We continue to organize annual oneAPI workshops jointly with Intel. Further, we will integrate SYCL and oneAPI in our planned course offerings for students and scientists.
"As the need for increased computational power can be achieved by accelerators, programming these devices with a widely-supported standard is crucial to prevent being bound to a particular vendor,” says Steffen Christgau, research associate, ZIB Algorithms for Innovative Architectures group. “With oneAPI's implementation of SYCL, we developed successful cross-architecture deployment capabilities for demonstrator applications. Within the Centre’s activity, we will leverage these proficiencies to port key applications of our data center/users to oneAPI. To further increase knowledge transfer and prepare the future generation of scientists for the unavoidable move to heterogeneous computing, we will include SYCL and oneAPI in our teaching portfolio."
Thomas Steinke, head of the Supercomputing department elaborates, “We are excited to continue our successful collaboration with Intel, which started nearly 10 years ago. This new format of joint work is important for us to explore in a much detailed way future paths toward portable, performant and energy-efficient solutions based on standardized parallel programming models for accelerators like GPUs, FPGAs and beyond.”
“Intel’s ongoing collaboration with the Zuse Institute Berlin aided scientific computing developers to not only understand and implement best methods for parallel computing, but helped apply them to applications in Germany and globally,” said Joe Curley, vice president and general manager of Intel Software Products and Ecosystem group. “Our latest Center of Excellence will now focus on solving the problems of cross-architecture programming with the intent to benefit developers in productivity, code longevity, and real freedom to choose the best hardware for their applications.”
About ZIB
The Zuse Institute Berlin (ZIB) is an interdisciplinary research institute for applied mathematics and data- intensive high-performance computing. Its research focuses on modeling, simulation, and optimization with scientific cooperation partners from academia and industry. Since January 2021, the ZIB has been extending its scientific services by offering HPC consulting for scientists in Germany and international projects as part of the nationwide HPC initiative “Nationales Hochleistungsrechnen” (NHR). We operate compute and storage resources at a top level, currently our 8 PFLOP/s HPC system “Lise” with Intel Cascade Lake AP CPU nodes which will be extended by GPU compute resources including the Intel Xe GPU architecture.
About Intel oneAPI
oneAPI is an open, unified and cross-architecture programming model for CPUs and accelerator architectures (GPUs, FPGAs, and others). Based on standards, the programming model simplifies software development and delivers uncompromised performance for accelerated compute without proprietary lock-in, while enabling the integration of existing code. With oneAPI, developers can choose the best architecture for the specific problem they are trying to solve without rewriting software for the next architecture and platform.
Learn more about Intel oneAPI Centers of Excellence.
