[llvm] [AArch64] Support scalable vp.udiv/vp.sdiv with SVE (PR #154327)

[Luke Lau via llvm-commits]

lukel97 wrote:

> In which case I'd rather see the addition of llvm.masked.sdiv because it's more in keeping with the existing masked operations rather than mixing between the two styles.

Unfortunately we still need the EVL operand though for RISC-V, which is why I was hoping to emit the VP intrinsic from the loop vectorizer.

If it's of any consolation, the ExpandVectorPredication pass will "discard" the EVL operand by the time it hits the AArch64 backend, so it can just be ignored during isel.

https://github.com/llvm/llvm-project/pull/154327