LLNL teams win two HPCwire awards at SC24

Two teams led by Lawrence Livermore National Laboratory (LLNL) computer scientists won Editor’s Awards from HPCwire, a leading high-performance computing industry publication, at the 2024 Supercomputing Conference (SC24) in Atlanta.

Researchers at LLNL, Los Alamos National Laboratory, Sandia National Laboratories and Cerebras Systems took home the publication’s award for Best Use of HPC in Physical Science for significantly speeding up molecular dynamics (MD) simulations using the second-generation Cerebras Wafer-Scale Engine (WSE-2), the world’s largest computer chip.

Running MD simulations on the WSE-2 — a cutting-edge processor boasting 850,000 cores — the team achieved a 457-fold improvement over exascale systems such as Oak Ridge National Laboratory’s Frontier. By dedicating a processor core to each simulated atom, the WSE-2 achieved a simulation rate of more than 699,000 timesteps per second for problems involving 800,000 tantalum atoms, setting a new standard for general-purpose processing cores and for what is possible in the realm of molecular dynamics.

The work was a finalist for the 2024 ACM Gordon Bell Prize. It was conducted under the NNSA’s Advanced Simulation and Computing (ASC) program’s Advanced Memory Technology (AMT) project, which aims to sustain technology research and development momentum in the post-exascale world through industry engagement. MD simulations are of particular interest to the NNSA Tri-Labs, where they are essential for exploring material behavior under extreme conditions.

“We are happy and thankful to be recognized by the editors at HPCwire for this award,” said LLNL computer scientist and team member Edgar Leon. “This is a testament to the hard work the NNSA Tri-Lab team and Cerebras have done to advance the field of molecular dynamics. It is exciting to think what this multi-disciplinary team will continue to achieve in this and other scientific domains — the future is bright.”

For the second year in a row, LLNL computer scientist Todd Gamblin and his team won the publication’s Editor’s Choice award for Best HPC Programming Tool or Technology for Spack, the highly popular package-management tool that has become a standard bearer in the HPC community.

A versatile, configurable and multi-platform package manager, Spack was created by Gamblin in 2013, and the Spack developer and user communities have grown rapidly since then. Spack has played a crucial role in running exascale machines at the Oak Ridge, Argonne and Lawrence Livermore national laboratories. Spack was the official package management tool for the Exascale Computing Project, and it has been deployed on HPC machines at many of the world’s top computing centers.

Spack has about 6,000 monthly users and more than 1,400 contributors on Github. Spack is no longer solely led by LLNL; it recently joined the High Performance Software Foundation (HPSF) within the Linux Foundation, providing Spack with a formal governance structure, broader visibility and more pathways to continue to sustain and grow its community, Gamblin said.  

“We’re really proud to have so many contributors here to accept this award,” Gamblin said. “This a great recognition for the Spack community. It feels great to be nominated by the editors who know this community well. Many contributors got into Spack because their HPC centers were using it. Usage has also grown rapidly simply by word-of-mouth. We try to address a very wide range of use cases — from end-users to developers to administrators to user support teams. I didn’t think in 2013 that we would still be refining Spack for so many different audiences. It's become a truly massive community.”

Gamblin added that Spack will hold its first in-person user meeting in Chicago in May 2025, sponsored by HPSF, with a goal of further fostering community growth.

The annual HPCwire Readers’ and Editors’ Choice Awards are determined through a nomination and voting process with the global HPCwire community, as well as selections from the HPCwire editors. The awards are an annual feature of the publication and constitute prestigious recognition from the HPC community. They are revealed each year to kick off the annual supercomputing conference, which showcases high-performance computing, networking, storage and data analysis.