[llvm] [AMDGPU] Use immediates for stack accesses in chain funcs (PR #71913)

[via llvm-commits]

rovka wrote:

> Was the generated code wrong before? Or just suboptimal, because it set up an sp register in cases where it didn't really need to?

Just suboptimal.

> Was the generated code wrong before? Or just suboptimal, because it set up an sp register in cases where it didn't really need to?

Just suboptimal.

https://github.com/llvm/llvm-project/pull/71913