fix default solver selection for CuSparseMatrix #834
Merged
+93
−9
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
hexaeder
changed the title
Comment on lines
+348
to
+350
| nothing | ||
| elseif LinearSolve.is_cusparse_csr(A) && !LinearSolve.cudss_loaded(A) | ||
| nothing |
Member
There was a problem hiding this comment.
Would it instead make sense to just fix cholesky instance for GPUs?
Contributor
Author
There was a problem hiding this comment.
We could have a extension (either in ArrayInterface or CUDSS) to define
ArrayInterface.cholesky_instance(::CuSparseCSR)the Problem is that if CUDSS is not loaded, the extension would not trigger, the specialization would not load and ArrayInterface would still fallback to LinearAlgebra.cholesky for A::Any which errors out.
Or do you mean something else?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR attempts to fix #827 . It requires a revert of #828 (PRed in #833)
This fixes mainly cache creaction problems. Now, based on whether CUDSS is loaded and whether a CSC or CSR matrix is provided it skips the cholesky and/or lu cache build.
Also changes the default solver logic:
CuSparseMatrixCSR-> Krylov if CUDSS not loaded (warns once), LU otherwiseCuSparseMatrixCSR-> Krylov (warns once about this)It also adds some tests for the default solver choice before and after
using CUDSSand fixes some missing import in the cuda tests.Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC. (new code does, surruding code does not)
Additional context
Add any other context about the problem here.