[llvm] [X86] Combine `uitofp <v x i32> to <v x half>` (PR #121809)

[via llvm-commits]

abhishek-kaushik22 wrote:

> I'd prefer this was handled in VectorLegalizer::ExpandUINT_TO_FLOAT since the issue isn't x86-specific

Can you please provide an example for other targets that this is failing on?

https://github.com/llvm/llvm-project/pull/121809