[llvm] [AMDGPU] Avoid setting GLC for image atomics when result is unused (PR #150742)

[Harrison Hao via llvm-commits]

harrisonGPU wrote:

> > Thanks, Matt. I will try to implement it! Do you think it makes sense to merge this PR?
> 
> Not without mutating the opcode to the no-return variant

Okay, thanks. :-)

https://github.com/llvm/llvm-project/pull/150742