QMCPACK: Journey to Exascale on Aurora

Ye Luo, Argonne National Laboratory
Developer Session Graphic featuring title, date, and image of aurora.

Join us on April 24, 2024, at 11:00 AM for a webinar that details the QMCPACK team's journey to exascale on Aurora. 

In this talk, Argonne's Ye Luo will share the team's experience porting QMCPACK to Aurora and demonstrate the impressive performance achieved on Intel Max GPUs thus far.

As exascale supercomputers are being deployed in U.S., QMCPACK (https://qmcpack.org) developers have migrated the code base to a performance portable implementation for maximal production. With a fresh design of code architecture, historically divergent code paths for CPUs and GPUs have been unified and a core set of features are available on all the computing platforms including CPUs and GPUs today. The new design was first implemented for NVIDIA GPUs and then extended to AMD GPUs with minimal change due to the HIP portable layer resemblance with CUDA. As Aurora offers SYCL programming model, we extended customized kernels to SYCL and integrated optimized linear algebra routines from MKL. OpenMP offload kernels worked out-of-the-box with Intel compilers. 

Ye Luo Ph.D is a computational scientist specialized in condensed mater physics and high-performance computing. He focuses on developing novel methods and algorithms enabling challenging atomistic and quantum-mechanics based simulations of materials on supercomputers. He is all-time number one committer to QMCPACK, a highly optimized simulation package for Quantum Monte Carlo methods. He also works on the Quantum ESPRESSO density functional theory simulation package. Dr. Luo earned his Ph.D. degree in Theory and Numerical Simulation of Condensed Matter from Scuola Internazionale Superiore di Studi Avanzati, Italy.