Topic: HPC Systems and Software

A new mathematical technique improves the computational efficiency of evaluating the solution in large-scale, high-order meshes on advanced HPC systems.

Nicknamed SUM25, Spack’s first user meeting showcased new development and a thriving community.

LLNL's flagship exascale machine maintained its status as the fastest supercomputer on the planet—claiming the No. 1 spot on not just one, but three of the most prestigious HPC rankings.

Learn how to use a modern, open-source HPC software stack! From mid-July to mid-September, join our tutorials on how to install and use several projects on AWS EC2 instances (provided). All experience levels are welcome.

The award-winning HPC package manager has seen steady growth and community engagement since its open-source release in 2014.

LLNL participates in the ISC High Performance Conference (ISC24) on June 10–13.

Wright’s day-long visit—his first to LLNL since being appointed Energy secretary—included behind-the-scenes tours of some of LLNL’s most advanced facilities and machines, from the National Ignition Facility to El Capitan, the world’s most powerful supercomputer.

Wright’s daylong visit included briefings and discussions with LLNL experts in key mission areas including stockpile modernization, fusion energy, AI, cybersecurity and supercomputing.

This open-source library automatically maps HPC applications to supercomputers efficiently and transparently, so users can run on Day One on any system.

A new just-in-time compilation approach leverages LLVM intermediate representation to optimize GPU kernels with portability to any GPU architecture.

Research recognized at the HiPC IEEE conference proposes using an optimized version of OpenMP for vendor-agnostic GPU performance, portability, and scalability.

In a recent study published in the Astrophysical Journal, LLNL researchers developed an innovative approach to map cosmic shear using linear algebra, statistics, and HPC.

Increased resource utilization is one goal of new architectures. At Livermore Computing, these include AI accelerators such as Samba Nova and Cerebras systems and El Capitan's Rabbits.

People to Watch are at the forefront of HPC trends, adapting new technology to the rapidly changing world in order to unlock the answers to the biggest societal challenges of our time and make the impossible, possible.

Over the next three years, CASC researchers and collaborators will integrate LLMs into HPC software to boost performance and sustainability.

The latest issue of LLNL's magazine explains how the world’s most powerful supercomputer helps scientists safeguard the U.S. nuclear stockpile.

LLNL's Bruce Hendrickson joins other HPC luminaries in this op-ed about the future of the field.

Todd Gamblin has a well-deserved reputation in the HPC software community as a passionate engineer who enjoys rolling up his sleeves and diving into technical problems. It’s not a stretch to see how he got hooked on HPC.

Presented last fall at a conference, a new approach to software binary analysis incorporates large-scale training data and hierarchical embeddings.

Hundreds of current and former LLNL employees, government officials, and industry leaders gathered on January 9 to mark a monumental achievement: the dedication of El Capitan, the world’s fastest supercomputer.

The event marked a monumental achievement with the acceptance and deployment of the first exascale computing system built for the NNSA, a historic milestone in national security and scientific computing recognizing the collaborative efforts of government, industry and scientific leaders.

This interview with HPC-AI Vanguard Kathryn Mohror covers her thoughts on teamwork, her projects, the field, and more.

SC24, held recently in Atlanta, was a landmark event, setting new records and demonstrating LLNL's unparalleled contributions to HPC innovation and impact.

Two teams led by LLNL computer scientists won Editor’s Awards from HPCwire, a leading high performance computing industry publication, at the 2024 Supercomputing Conference in Atlanta.

Held for the first time in a hybrid format, the multi-day MFEM workshop drew participants from around the globe.