[Mlir-commits] [mlir] [mlir][vector] Rename vector type TD definitions (nfc) (PR #117150)

[Andrzej Warzyński]

banach-space wrote:

> But another suggestion based on problems I've had with these tablegen defs: AnyVector not allowing 0-d rank vectors is a bit confusing. AnyVector is really AnyVectorOfAnyRank + additional constraints. It took some debug time in the past to figure out that the naming is weird. The naming should maybe be AnyVectorOfNonZeroRank and AnyVector, or even AnyVectorOfNonZeroRank and AnyVectorOfAnyRank, to better reflect that the current AnyVector is more constrainted definition. What do you think? (My suggestion is non blocking, feel free to land)

I agree that `AnyVector` is quite misleading and deserves a re-name as well. I assume that those names were added as a temporary workaround and we just never got round to unifying them (or "fixing" 0-D support).

I am happy to change that in a separate PR. First, I wanted to send this one that fixes inconsistencies to make sure that there are no objections 😅 

Thanks for reviewing 🙏🏻 

https://github.com/llvm/llvm-project/pull/117150