Bridges-2 Webinar Series: High performance GPU accelerated MuST software

The MuST package is computational software designed for ab-initio electronic structure calculations for solids. The Locally Self-consistent Multiple Scattering (LSMS) method implemented in MuST allows one to perform the electronic structure calculation for systems with a large number of atoms per unit cell. For the LSMS method, the major computational challenge is the matrix inversion for the scattering matrix calculation, which could take more than 90% of the computing time. However, the matrix inversion can be significantly accelerated by modern graphical processing units (GPUs). In this presentation, Dr. Liang will discuss the team’s approach to code acceleration by offloading the matrix inverse tasks to the GPUs through a Fortran-C interface to the CUDA code. The team found that a high speedup ratio can be achieved by GPU acceleration. Dr. Liang will discuss specific source code and job script examples to show how they parallelize the computational tasks over multiple GPUs on Bridges-2. Finally they will report their performance results on the calculation of NiAu alloy, a candidate for thermoelectric material. This work was supported by the National Science Foundation through the OAC-21939536 Characteristic Science Applications for the Leadership Class Computing Facility award.