Researchers from Berkeley Lab’s Applied Mathematics and Computational Research Division (AMCR) and Accelerator Technology & Applied Physics Division (ATAP) have been honored with the 2024 Best Paper Award from Computing in Science & Engineering, presented by the IEEE Computer Society Publications Board. The award recognizes their paper, Then and Now: Improving Software Portability, Productivity, and 100× Performance, which reviews and analyzes advances in high‑performance computing software achieved through the U.S. Exascale Computing Project (ECP). The ECP, which concluded in 2023, prepared scientific applications to run efficiently on the world’s first exascale supercomputers — systems capable of performing more than a quintillion calculations per second.

In their paper, the authors compare the state of selected software projects — including flagship mathematical libraries and applications such as the Extreme‑Scale Scientific Software Development Kit (xSDK), SLATE, SuperLU, Ginkgo, and WarpX — before and after the ECP. They highlight how collaborative development, modern programming practices, and a focus on portability have transformed the software ecosystem. The paper also distills lessons learned and offers recommendations to ensure research software remains sustainable, interoperable, and ready for future computing architectures.

The team emphasizes portability as a core design principle, advocating for the adoption of modern programming languages, such as modern C++, to enhance the sustainability, interoperability, and performance of research software. They also note how cross‑institutional collaboration can drive algorithmic advances that go beyond incremental performance gains.

The team emphasizes portability as a core design principle, advocating for the adoption of modern programming languages, such as modern C++, to enhance the sustainability, interoperability, and performance of research software. They also note how cross‑institutional collaboration can drive algorithmic advances that go beyond incremental performance gains.

The review committee praised the paper for its originality, rigor, and significant impact on computational science.

“This award highlights the essential role that innovative software solutions play in driving scientific research forward. As we navigate the evolving landscape of exascale computing and beyond, our work establishes a foundation for future advancements and collaborations that will enhance our understanding and capabilities in this field,” said Xiaoye (Sherry) Li, who leads AMCR’s Scalable Solvers Group and is a co-author of the paper.

Co-author Axel Huebl (ATAP Division) adds: “It was an absolute joy to develop in a fantastic multi-lab team of teams towards the first exascale machines. We and the community are using our resulting application(s) daily in our work for designing future particle accelerators and for fusion energy science. Modularization and efficient code sharing enabled rapid development, performance portability to the world’s largest supercomputers and R&D of new numerical methods. ECP was a bedrock for truly reliable, collaborative, open science and resulting HPC simulations provide excellent training data for upcoming AI/ML efforts.”

Hartwig Anzt (University of Tennessee, Knoxville; Technical University of Munich) complemented the three equally contributing authors.