[llvm] [AArch64] Don't try to vectorize fixed point to fp narrowing conversion (PR #130665)

[Eric Christopher via llvm-commits]

echristo wrote:

> There will be a lot of test cases that need updating. We should attempt to limit the damage this does, and I don't believe the default expansion is very efficient for unsigned operations.

I did notice that arguably turning off vectorization of this operation might be the most efficient way. It does leave the loop rather than unrolling it, but the code is quite a bit tighter and the scalar conversion is precise.

https://github.com/llvm/llvm-project/pull/130665