GPU Acceleration of Large-Scale Full-Frequency GW Calculations

TL;DR

This paper demonstrates a GPU-accelerated implementation of the full-frequency GW method in the WEST code, achieving significant speedups and scalability on high-performance computing systems for large-scale electronic excitation simulations.

Contribution

The paper introduces optimized GPU-based algorithms and parallelization strategies for the GW method, enabling efficient large-scale calculations on thousands of GPUs.

Findings

Substantial speedup over CPU version

Good strong and weak scaling up to 25920 GPUs

Successful large-scale calculations of systems with over 10,000 electrons

Abstract

Many-body perturbation theory is a powerful method to simulate electronic excitations in molecules and materials starting from the output of density functional theory calculations. By implementing the theory efficiently so as to run at scale on the latest leadership high-performance computing systems it is possible to extend the scope of GW calculations. We present a GPU acceleration study of the full-frequency GW method as implemented in the WEST code. Excellent performance is achieved through the use of (i) optimized GPU libraries, e.g., cuFFT and cuBLAS, (ii) a hierarchical parallelization strategy that minimizes CPU-CPU, CPU-GPU, and GPU-GPU data transfer operations, (iii) nonblocking MPI communications that overlap with GPU computations, and (iv) mixed-precision in selected portions of the code. A series of performance benchmarks have been carried out on leadership high-performance computing systems, showing a substantial speedup of the GPU-accelerated version of WEST with respect to its CPU version. Good strong and weak scaling is demonstrated using up to 25920 GPUs. Finally, we showcase the capability of the GPU version of WEST for large-scale, full-frequency GW calculations of realistic systems, e.g., a nanostructure, an interface, and a defect, comprising up to 10368 valence electrons.