RIKEN International HPC Spring School 2024 - Toward Society 5.0 - (March 13-15, 2024, Kobe/Online)

cf.) R-CCS HPC School List (PDF file: 302KB)

RIKEN Center for Computational Science (R-CCS) will hold a hybrid school in March 2024 titled "RIKEN International HPC Spring School 2024 ~ Toward Society 5.0~ Parallel computation of ill-conditioned sparse matrix in finite element analysis".

The program consists of 3 days in hybrid with on-site and online. The on-site attending is only for those who can commute to R-CCS without accommodation.

The lecture is focusing on parallel linear solver for finite element analysis. Finite element method is a strong tool for various type of industrial problems by utilizing consolidated knowledge of mathematics and engineering. To obtain finite element solution, at the final stage, linear system with large sparse matrix needs to be solved and the condition number of such matrix becomes very large due to spatial jump of physical parameters and to complex coupling of multi-physics.

Participants will gain knowledge to solve linear and nonlinear systems with large sparse matrices on parallel computers. Foundation of linear solver, such as Krylov subspace methods with preconditioner based on domain decomposition, will be explained.

Hands-on session in each day will treat parallel programing using MPI, OpenMP, numerical libraries, METIS and PETSc.

At the end of the lecture, a domain specific language for finite element simulation, FreeFEM is briefly introduced with an example of nonlinear elasticity problem with large deformation, which will be solved by the Newton-Krylov-Schwarz method.

Day 1 : sparse matrix and Krylov subspace method, GMRES and CG
Day 2 : distribution of sparse matrix with overlapping domain and additive Schwarz preconditioner
Day 3 : introduction of domain specific language FreeFEM to generate nonlinear equation with sparse matrix for hyper elasticity problem with Green-St Venant strain tensor

This program will be provided in English.