Project

Domain Decomposition

Status: Planned · Study local solves, overlap, and coarse spaces as preconditioners for discretized PDEs.

Question

How do overlap and a coarse correction affect convergence, setup cost, and parallel communication?

Planned sequence

  1. Establish a block-Jacobi baseline for a system from the FEM project.
  2. Define restriction, extension, overlap, and local subdomain operators.
  3. Implement additive and restricted additive Schwarz (ASM/RAS).
  4. Add a two-level correction and state the coarse-space construction.
  5. Evaluate convergence and communication with the parallel sparse-computing project.

Evidence to record

Report the partition, overlap width, number of subdomains, local solver, coarse-space dimension, outer Krylov method, residual criterion, and timings. State symmetry and definiteness assumptions when selecting the outer solver.

Current evidence

This is a planned project; no ASM/RAS implementation, coarse-space experiment, or scaling claim is published here yet.

Study prerequisites

Basic preconditioners → Domain Decomposition Bridge

Linked notes

No project notes have been published yet. The steps above describe planned work.

Linked paper notes

No paper notes linked yet.

All projects · Research directions