Skip to content

Nesting and domain-decomposition machinery (optional capstone) #53

Description

@awickert

Context

With inhomogeneous BCs as transmission data (geometric couplet = Dirichlet, mechanical couplet = Neumann) and the FFT preconditioner (#49), gFlex can support nested/downscaled solves and, optionally, parallel domain decomposition. The foundation term's exponential localization (decay length α) makes Schwarz-type iteration converge fast, and a one-way coarse→fine downscale is often accurate enough without iteration. Serial nesting/coupling is the immediate payoff; parallelism is one further use — not the point of this item.

Scope (gated — build only if nesting/scaling demand materializes)

  • One-way downscale: solve coarse, impose coarse (w, slope) on a fine subdomain boundary via the Issue-4 boundary object (Interface, validation, and back-compat presets for boundary conditions #52).
  • Optional two-way / overlapping-Schwarz iteration reusing the same boundary object as transmission data, with the FFT preconditioner applied per subdomain.
  • Keep serial nesting/coupling working without requiring parallel infrastructure.

Acceptance criteria

  • One-way nested solve reproduces the monolithic solution within tolerance away from the refinement boundary.
  • (If implemented) iterative DD converges to the monolithic solution.

Dependencies

Notes

Optional stretch goal. Nothing in #51 or #52 waits on this issue; it shapes them only via the "expose boundary data as a first-class object" requirement already in #52.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions