[llvm] [RegAlloc][X86] Rematerialization of instructions with physical register (PR #198853)

[Matt Arsenault via llvm-commits]

https://github.com/arsenm commented:

If you're going to handle this in X86 code, doing in the post-RA pseudo expansion is probably the wrong choice. You'd probably want a pre-assignment optimization pass.

However I do think there is value in rematerializing the dead def case independently, but that doesn't require the target hook 

https://github.com/llvm/llvm-project/pull/198853