R-CCS Cafe 特別版(2)（2024年9月18日）

Title:
Interoperability between libraries for high-performance distributed-memory computing
Abstract:
Having state-of-the-art libraries interact seamlessly is of paramount importance for scientists who need to harness the power of supercomputers without knowing all the details of the underlying architectures. In this presentation, I will focus on libraries for performing finite-element analyses. These involve assembly procedures with a high-level of abstraction and an algebraic backend which is tightly coupled with the discretization kernel so that linear and nonlinear solvers can benefit from additional physics-based inputs. Examples will be based on the following libraries and languages: FreeFEM, HPDDM, PETSc, SLEPc, in C, C++, Python, Fortran, or a custom domain-specific language.