Dr Jorge Luis Galvez Vallejo1
1National Computational Infrastructure, Canberra, Australia
Biography:
I am an High Performance Computing specialist at the National Computational Infrastructure in Canberra, Australia. I am formally trained in quantum computational chemistry but branched out into high performance computing during my PhD and Postdoctoral Fellow appointments. During my PhD I was part of the Exascale Computing Project under the banner of GAMESS under professor Mark Gordon. I played a crucial role in the porting of key routines to GPUs in the LibCChem/GMSHPC program using C++/CUDA/HIP. Upon completing my PhD I came to Canberra to the ANU to work under A. P. Giuseppe Barca, who lead the C++/CUDA efforts of the GAMESS-ECP program. During my time at the ANU I became the lead developer and key maintainer of the EXESS program, a from scratch written electronic structure software that specializes in exploiting massively parallel architectures with GPU accelerators to its full extent. The highlight of my tenure was being a part of the team that was awarded the 2024 Gordon Bell prize by the ACM for our ab initio molecular dynamics simulation that was able to use the entirety of Frontier effectively running at more than 1 Exaflop. I moved to the NCI in May 2025 to seek new challenges in the areas of weather, climate, and geophysics.
Abstract:
The General Atomic and Molecular Electronic Structure System (GAMESS) is one of the most widely used quantum chemistry programs in the world and has been in continuous development since the late 70s. GAMESS has been a pioneer in many areas of computational chemistry and software technology.
GAMESS has always been a free to use, source available program; however, it has lacked key software engineering technologies that have limited it from being FAIR software. In this work, I present a series of innovations introduced to GAMESS to abide by the FAIR principles. First, the build system of GAMESS has been entirely rewritten to use CMake with modularity in mind, being able to select external packages, libraries to use with ease.
Using the new CMake infrastructure, I have made GAMESS more accessible by introducing a conda environment for dependency management and a spack recipe for deployment in HPC centres.
By reworking the build system, I provide an accessible and reusable GAMESS library which other programs can link to and use. This is made easier with a CMake package included in with the program. The new infrastructure makes it seamless to link external packages to GAMESS.
Additionally, a unit-testing system has been introduced to GAMESS for further validation of the program.
By introducing these key improvements to GAMESS, software modernization incentives can begin, such as migrating from Fortran77 to modern Fortran constructs to ensure portability in edge computing systems. Finally, a full rework of the documentation is presented.