Conversation
Member
Author
|
Ah but needs the TLC for 1.10 and 1.11, a task for tomorrow... |
Member
Author
|
OK, this actually looks ... somewhat ok? Windows fail seems unrelated, CUDA 1.10 is happening elsewhere also and is due to Pkg issues. |
The redundant `=true` definitions are dropped (base already short-circuits the matching direction), and the load-bearing opposite-triangle queries move from JLArrays to core, keyed on the GPU array parent so every backend gets them. The matmul kernels now use istriu/istril uniformly. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
`mul!(C, A, B, α, 0)` with α≠1 reaches the kernel, where `β * C[i,j]` turned an uninitialized NaN/Inf into NaN. Guard the β term and add a regression test that initializes C with NaN. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
upperB was never read (B's structure comes from its getindex), and the 1.10 kernel's wA/wB wraps were computed but never passed to the kernel. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The istriu/istril overrides were unsound: UpperTriangular(Diagonal(...)) is also lower triangular, so a type-only `istril = false` lies. The kernels only need the wrapper's structural guarantee, which a plain `isa` check gives directly and without scalar indexing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The three version-gated tri×tri methods shared a near-identical body and kernel. Extract _triangular_matmatmul!(C, A, B, α, β) and reduce each to a thin forwarder (the pre-1.12 ones unpack α/β from the MulAddMul). Net -47 lines; verified on Julia 1.10, 1.11 and 1.12. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Existing cases only covered α=1 (scaling is a no-op) or β=0 (the β·C term drops), so the general α·(A·B) + β·C path was untested. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Member
|
OpenCL.jl crash is an LLVM bug, fixed in llvm/llvm-project#201417, back-ported as part of JuliaPackaging/Yggdrasil#13884, |
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
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.
Should hopefully fix #726