TIME: 3:30 PM

LOCATION: GMCS 314

SPEAKER: Damian Rouson, Lawrence Berkeley National Laboratory

ABSTRACT: The vast majority of parallel programs running on high-performance computing (HPC) platforms leverage parallel programming models often referred to as “MPI+X”, where MPI is the Message-Passing Interface and X typically stands for OpenMP, OpenACC, or CUDA. Over the course of the Fortran 2008, 2018, and 2023 standards, the world’s first widely used programming language became a parallel language, eliminating most needs for directly referencing MPI+X in source code. This talk will present some of Fortran’s native parallel features for executing on central processing units (CPUs) and graphics processing units (GPUs). The talk will describe uses of these features in open-source software developed by Berkeley Lab’s Computer Languages and Systems Software (CLaSS) Group and our collaborators: the Julienne correctness-checking framework [1], the Fiats deep learning library [2], and the Matcha T-cell motility simulator [3]. The talk will also briefly touch on the use of Julienne in AI-assisted vibe coding and plans for implementing the mimetic differences of Corbino and Castillo [4] using the Abstract Calculus pattern of Rouson, Xia, and Xu [5].

[1] https://go.lbl.gov/julienne
[2] https://go.lbl.gov/fiats
[3] https://go.lbl.gov/matcha
[4] Corbino, J., & Castillo, J. E. (2020) “High-order mimetic finite-difference operators satisfying the extended Gauss divergence theorem,” Journal of Computational and Applied Mathematics, 364, 112326.
[5] Rouson, D., Xia, J., & Xu, X. (2011) Scientific software design: the object-oriented way. Cambridge University Press.

HOST: Christopher Paolini