[llvm] [AMDGPU] Apply target flag specifier when lowering MO_ExternalSymbol operands (PR #202389)

[Arseniy Obolenskiy via llvm-commits]

aobolensk wrote:

> Can you add an end to end IR test

(Un)fortunately, I could not reproduce the problem on end-to-end path

https://github.com/llvm/llvm-project/pull/202389