[Mlir-commits] [mlir] [mlir][bufferization] Add tensor-like and memref-like interfaces (PR #134220)

[Andrei Golubev]

andrey-golubev wrote:

> I'll leave it up to @matthias-springer , but I find it weird to attach interfaces in dialect initialization of another, irrelevant dialect. We already have external models for bufferization op interfaces, e.g.,

I think it's just me unaware of what's the correct way of doing this. Thanks for the suggestion!

> . I'd expect we do the same for bufferization type interfaces. AFAICS, `one-shot-bufferize` does not list the bufferization dialect as dependent, so there is no guarantee it is loaded when the pass runs. I'd find it worryingly inconsistent if built-in types wouldn't bufferize when another dialect is not loaded.

I agree. Which is why I thought that dialect initialization is the best place.

For the "external model" my only concern is: can we actually have `void builtin::registerBufferizableOpInterfaceExternalModels()`? Wouldn't it introduce unintentional coupling of builtins and bufferization?

https://github.com/llvm/llvm-project/pull/134220