[Mlir-commits] [mlir] [MLIR][Vector] Add Lowering for vector.step (PR #113655)

Kunwar Grover llvmlistbot at llvm.org
Tue Oct 29 07:17:44 PDT 2024


Groverkss wrote:

> Cool, why not upstream that? That's a good optimisation and nice justification for this change.

It would eventually be upstreamed. But upstreaming a transformation, which can never actually be used because the op folds itself away, is upstreaming something that cannot get used.

I'm not sure this change by itself needs justification though. In the original review for vector.step, moving this transformation out of canonicalization was already requested but not addressed https://github.com/llvm/llvm-project/pull/96776/files#r1667009554 . 

https://github.com/llvm/llvm-project/pull/113655


More information about the Mlir-commits mailing list