[Fix] makeblascontractable should return blascontractable objects.
#210
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
|
Wow this is bad indeed, the bug I mean. Strange that this has survived for so long. A test that |
Jutho
reviewed
Jun 10, 2025
| Anew = SV(A_, size(A_), strides(A_), 0, A.op) | ||
| conjA = A.op === conj | ||
| A_ = tensoralloc_add(TC, A, pA, conjA, Val(true), allocator) | ||
| Anew = SV(A_) |
There was a problem hiding this comment.
I don't think conjA is needed in the tensoralloc_add; that information is already in A (not that it matters for AbstractArray types). I think the main bug was that A.op was inserted back into the Anew = SV(...) call.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
I was doing some benchmarking with the
StridedBLASbackend, and found some interesting results in the profiler: somehow I was hitting theStrided.__mul!generic implementation instead of thegemmone from BLAS, and it turns out that for conjugated inputs, themakeblascontractablefunction doesn't actually return a blascontractable object. The result is a slow fallback implementation instead of the actual BLAS call.The current fix seems easy enough, although it's not the only option: in principle we could retain the
A.opwhile permuting such that everything fuses and the second stride is 1, but I'm not sure if this would then make a lot of difference performance-wise?I'm not sure if this is something we should add explicit tests for either. (It's not too hard to check that the output of
makeblascontractablesatisfiesisblascontractable, which is also how I found this "bug").