[llvm] [AMDGPU] - Generate s_bitreplicate_b64_b32 (PR #69209)

[Jay Foad via llvm-commits]

jayfoad wrote:

> The _intended_ reasoning for why this generic transform is forbidden when `bitreplicate` is `convergent` is as follows.

Thanks for explaining. It would be great if we had an Alive that understood `convergent` that could prove things like this.

https://github.com/llvm/llvm-project/pull/69209