Skip to content
GitLab
Projects Groups Topics Snippets
  • /
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Register
  • Sign in
  • dumux dumux
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributor statistics
    • Graph
    • Compare revisions
  • Issues 100
    • Issues 100
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 92
    • Merge requests 92
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Environments
    • Releases
  • Packages and registries
    • Packages and registries
    • Container Registry
  • Monitor
    • Monitor
    • Incidents
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Wiki
    • Wiki
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • dumux-repositoriesdumux-repositories
  • dumuxdumux
  • Merge requests
  • !2113

WIP: Feature/new istl linear solvers

  • Review changes

  • Download
  • Patches
  • Plain diff
Closed Timo Koch requested to merge feature/new-istl-linear-solvers into master May 03, 2020
  • Overview 152
  • Commits 42
  • Pipelines 10
  • Changes 25
  • Revisit template arguments
  • Introduce linear algebra backend or similar to make usage simpler
  • Adapt all linear solvers to provide norm
  • In Newton call linearSolver->norm(r) if available otherwise call assembler->residualNorm(); (depends on !2311 (merged) to be merged)
  • Deprecate conversion of multitype matrices in Newton (do in solver instead if necessary)

Depends on !2310 (merged) to be merged.

The matrix and vector type now have to be known to construct the solver. This was previously delayed until the solve call but made the structure kind of intransparent because it wasn't really clear which vector type has to come in but only specific types work anyway.

Hard coding the solver hopefully reduces compile times wrt the factory. Also the implementation should be

  • more compact than the old backends
  • have more runtime option due to parameter tree-based params
  • work in parallel as well

If this works out, we would deprecated the old solver backends.

Edited Jan 02, 2021 by Timo Koch
Assignee
Assign to
Reviewers
Request review from
Time tracking
Source branch: feature/new-istl-linear-solvers