Skip to content

Conversation

@sv2518
Copy link
Contributor

@sv2518 sv2518 commented Dec 2, 2021

This PR makes it possible to run a preconditioner based on Slate globally matrix-free but with jacobi preconditioner. In that case we ask to assemble the diagonal of an expression. Ideally this would have been part of the PR #2259 which we merged a month ago, since it's when I dropped the error message that "Diagonal + slate not supported" and introduced DiagonalTensors in Slate. The existing diagonal infrastructure only had to be adapted to give the diagonal tensor back as a vector. Before this PR we were not able to run a solve in a Slate based preconditioner with Jacobi and 'mat_type': 'matfree' in its parameters

  'hybridization': {'ksp_type': 'cg',
                    'pc_type': 'jacobi',
                    'mat_type': 'matfree'}}

@sv2518 sv2518 force-pushed the sv/hybridizationpc-with-jacobi-and-global-matfree branch 2 times, most recently from eaafc30 to be838e9 Compare December 6, 2021 10:24
@sv2518 sv2518 marked this pull request as ready for review December 6, 2021 10:24
sv2518 added 2 commits March 30, 2022 12:33
make it possible to run a preconditioner which is based on Slate globally matrix-free but with jacobi. The infrastructure only had to be adapted to give the diagonal tensor in vector shape back.
@sv2518 sv2518 force-pushed the sv/hybridizationpc-with-jacobi-and-global-matfree branch from be838e9 to f4fab16 Compare March 30, 2022 13:34
@dham dham closed this Jun 30, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants