[Mlir-commits] [mlir] [mlir][vector] Fix 0-d vector transfer mask inference (PR #116526)

[Kunwar Grover]

Groverkss wrote:

> I’d be cautious about concluding that something is broken - at least not yet! 😊

Sorry, not the right choice of words. The PR is obviously correct, and useful. The only thing that seems weird to me is the lit test. Let me write up why it feels like that quickly. I only blocked the pr so we can discuss it more. I'm happy to remove the block meanwhile, and explain my concerns better.

> This PR strictly follows the documented design, which ensures consistency. Other parts of MLIR should adhere to the same principle as well.

Right, I shouldve been clearer, I don't want to block the change, but want to discuss the masking behavior more.

> All these examples involve broadcasting. It’s worth referring to this point from the vector.transfer_read [documentation](https://mlir.llvm.org/docs/Dialects/Vector/#vectortransfer_read-vectortransferreadop) (emphasis mine):

Right, I am following this definition, but I will write up a response explaining myself better.

Meanwhile, I'm unblocking, feel free to land it :)

https://github.com/llvm/llvm-project/pull/116526