[Mlir-commits] [mlir] [mlir][Vector] add vector.insert canonicalization pattern to convert a chain of insertions to vector.from_elements (PR #142944)

[Yang Bai]

yangtetris wrote:

> Thanks for the updates. I am wondering whether this is robust enough to handle overlapping vector.insert operations?

Please refer to another [thread](https://github.com/llvm/llvm-project/pull/142944#discussion_r2250719079) on this page.
> Also, I am not sure whether this would be beneficial? Also, are there any tests for this case?

Sorry I didn't update the PR description in time. With @dcaballe 's advice, we now use `vector.to_elements`, which looks much better. The case `fully_insert_to_vector_overlap` covers this scenario.
@banach-space  Please take another look at the new PR description. Thanks!

https://github.com/llvm/llvm-project/pull/142944