[llvm] [AMDGPU] Allow merging unordered and monotonic atomic loads in SILoadStoreOptimizer (PR #189932)

[Harrison Hao via llvm-commits]

harrisonGPU wrote:

> What is the motivation for doing this kind of merging now? 

Because this can merge adjacent atomic buffer loads into one wider load, which reduces load instruction count and memory transactions.

https://github.com/llvm/llvm-project/pull/189932