Indirection in matrix multiplication to avoid ambiguities #1210
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
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #1210 +/- ##
==========================================
+ Coverage 91.87% 91.89% +0.01%
==========================================
Files 34 34
Lines 15368 15369 +1
==========================================
+ Hits 14120 14123 +3
+ Misses 1248 1246 -2 ☔ View full report in Codecov by Sentry. |
dkarrasch
approved these changes
Feb 20, 2025
Member
dkarrasch
left a comment
dkarrasch
left a comment
There was a problem hiding this comment.
I think this is a good idea. Packages may keep overloading *, but if we can move away the "internal" to LinearAlgebra.jl dispatch toward mul, then this should actually greatly help the ecosystem. I think we have done something very similar with the inplace multiplication methods. I wonder if basically "all" specialized * methods should/could be moved to mul? I see methods in bidiag.jl and lots of triangular stuff. Could be in another PR, though.
KristofferC
pushed a commit
that referenced
this pull request
Feb 25, 2025
(cherry picked from commit 5cf41c4)
KristofferC
added a commit
that referenced
this pull request
Feb 25, 2025
Backported PRs: - [x] #1194 - [x] #1207 - [x] #1196 <!-- Explicitly declare type constructor imports --> - [x] #1202 <!-- Add fast path in generic matmul --> - [x] #1203 <!-- Restrict Diagonal sqrt branch to positive diag --> - [x] #1210 <!-- Indirection in matrix multiplication to avoid ambiguities -->
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 adds a level of indirection in
*(::AbstractMatrix, ::AbstractMatrix)to avoid method ambiguities with packages. This generally happens when a package defines*(::SpecializedMatrix, ::SomeMatrix)andLinearAlgebradefines*(::SomeMatrix, ::AnotherSpecializedMatrix). By changing the latter to add a method to the internal functionmul, we avoid such method ambiguities.Practically speaking, this is mainly necessary when a method is being added to
*with one of the arguments being quite specialized, e.g.*(::Diagonal, ::UpperTriangular{<:Any, <:StridedMatrix}). In such cases, we may add the method tomulinstead. This should usually not be necessary if one is just defining*(::Diagonal, ::UpperTriangular). However, seeing that packages may add methods like*(::Diagonal{<:Any, <:SpecialMatrix}, ::AbstractMatrix), we may decide to change these to usemulas well in the future. I've not made too many changes in this PR to avoid clutter, and this mainly focusses on addressing regressions arising from the diagonal-triangular multiplications.This PR resolves a couple of method ambiguity errors in the
ArrayLayouts.jltest set that are seen on nightly.