Skip to content

Conversation

@mgehre-amd
Copy link
Collaborator

No description provided.

aldesilv and others added 3 commits November 12, 2024 12:54
(nod-ai/SHARK-TestSuite#391)
Repro (using SHARK TestSuite):
1. `python run.py --torchtolinalg -m cl-onnx-iree -t cubic_test`

---------

Co-authored-by: zjgarvey <[email protected]>
…ainderTensorOp` (llvm#3861)

1. adds a lowering for `aten.neg.int` and `aten.remainder.int` to arith.
2. adds a scalarization pattern for `aten.neg` and
`aten.remainder.Tensor` ops.
3. improves folding of `aten.mul.int`
4. adds a scalarization pattern for `aten.to.dtype` which relies on
scalar cast ops and basic C++ casting between `double` and `int64_t`.
5. improves rank-0 case handling for `FoldAtenSplatPattern`
6. removes a bug with `aten.unflatten.int` decomposition incorrectly
generating a constant size int from a dynamic shape.
7. simplifies the dim list for `aten.unflatten.int` ops generated from
the `aten.view` canonicalization in scalarize shapes.

All of these changes were necessary to unblock
<iree-org/iree#18899>.
Base automatically changed from bump_to_17c1985c to bump_to_b6f04fa3 February 4, 2025 08:32
@mgehre-amd mgehre-amd merged commit e33b6e9 into bump_to_b6f04fa3 Feb 4, 2025
3 of 4 checks passed
@mgehre-amd mgehre-amd deleted the bump_to_cd38ecf6 branch February 4, 2025 08:33
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants