[llvm] AMDGPU: Fold frame indexes into disjoint s_or_b32 (PR #102345)

[Jan Patrick Lehr via llvm-commits]

jplehr wrote:

This patch may be the triggering issue for the bot breakage in https://lab.llvm.org/buildbot/#/builders/10/builds/25403

https://github.com/llvm/llvm-project/pull/102345