Technical scope

The technical directions below shape the software developments and scientific themes of the project.

Key directions

• Algorithmic optimization: improve efficiency and scalability of numerical kernels with attention to memory, data locality and vectorization.

• Solvers and preconditioners: integrate and evaluate PSCToolkit and MUMPS as robust coarse/problem solvers for ill-conditioned cases.

• Accelerators and heterogeneity: adapt critical phases (assembly, preconditioner setup) for GPUs and hybrid architectures, including matrix-free and mixed-precision strategies.

• Multiscale and coupling: tooling for 3D/1D coupling, mixed-dimensional formulations and immersed methods for bio-physical problems.

• Validation and reproducibility: test workflows, datasets for ML, and benchmark suites to compare methods and optimizations.

These themes guide the development of the components listed in the technical repositories section. See project_overview.md for a short summary.