[WIP][AMDAIEPackAndTranpose] Undo folding of rank-preserving packs #787
Conversation
This file contains 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
newling
requested review from
MaheshRavishankar,
nirvedhmeshram,
yzhang93,
Abhishek-Varma and
jtuyls
as code owners
newling
changed the title
newling
changed the title
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Outdated suggestions cannot be applied.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.
The function
linalg::packdoesn't create rank-preserving pack ops, see: https://github.com/llvm/llvm-project/blob/644899addd8fd789c93e9a0f0727d37eb1b29c55/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp#L542The upstream design cannot easily be undone, because it is thoroughly tested for. See: https://github.com/llvm/llvm-project/blob/644899addd8fd789c93e9a0f0727d37eb1b29c55/mlir/test/Dialect/Linalg/transform-op-pack.mlir
In our use case it is sometimes useful to have identity pack ops, specifically if we don't want to pack any dimensions of an operand, but still want to set up a new tensor for bufferization.
The use case I have for this is channel-first convolution. This PR is not strictly necessary because I have an alternative (packing input and output channel dimensions with size 1) but IMO this is maybe neater.
If there's general approval I'll add tests.