[Mlir-commits] [mlir] [mlir][bufferization] Support custom types (1/N) (PR #142986)

[Han-Chung Wang]

https://github.com/hanhanW commented:

Many thanks for the upstream contribution! I've seen some changes like this that'd break downstream projects. It is very reasonable because you're improving the bufferization.

I'd appreciate it if you can list down some naming changes in the PR description before you land the PR. It'd save us some time to look at all the details. E.g., the operand name of `ToTensorOp` becomes `buffer`; the return type of `bufferization::getBufferType` becomes `FailureOr<BufferLikeType>`, etc;

(It is definitely not a requirement, and it is okay if you don't do it. I appreciate your contribution. Thanks again! 🙏)

https://github.com/llvm/llvm-project/pull/142986