[Mlir-commits] [mlir] [mlir][scf] refactor loop peeling logic to use arith operations (PR #235260)

[llvmlistbot at llvm.org]

StarOne01 wrote:

@matthias-springer the current signed path has an affine check that skips peeling when `(ub - lb) mod step` is provably 0, but arith can’t do that, so kept the affine check for signed index loops. is that fine?

https://github.com/llvm/llvm-project/pull/235260