[llvm] [AArch64][SVE] Fold predicate SETCC negations (PR #206931)

[Matthew Devereau via llvm-commits]

MDevereau wrote:

> > I wanted to do it in AArch64 InstCombine but the generic XOR with an Intrinsic pushed me into DAG.
> 
> I know what you mean. What if you take a similar approach as #204796? I don't normally like walking the IR forward but it's only one level up and well contained so is probably worth it to make the IR more optimal.

Sounds good to me, I'll re-submit it.

https://github.com/llvm/llvm-project/pull/206931