[PATCH] D152298: [RISCV] Use vslidedown.vi vN, vN, 0 instead of vslideup.vi vN, vM, 0 for subvector insertion

[Luke Lau via Phabricator via llvm-commits]

luke added a comment.

In D152298#4407683 <https://reviews.llvm.org/D152298#4407683>, @craig.topper wrote:

> Should we be using vmv.v.v instead?

I noticed that these slides of zero are also equivalent to vmerge with an all ones mask and a tail undisturbed policy, so I tried combining to that to see what changed. The vmerges then got converted to `vadd.v.i vd, vs, 0` in `performCombineVMergeAndVOps`. Should all these forms be canonicalised to `vmv.v.v`?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152298/new/

https://reviews.llvm.org/D152298