[llvm-branch-commits] [llvm] AMDGPU: Handle subregister uses in SIFoldOperands constant folding (PR #127485)

[Matt Arsenault via llvm-branch-commits]

arsenm wrote:

> Later patch exposes miscompiles with this one

Not actually this patch's fault, yet more pre-existing bugs 

https://github.com/llvm/llvm-project/pull/127485