[llvm-branch-commits] [llvm] [AMDGPU] Support one immediate folding for global load (PR #178608)

[via llvm-branch-commits]

ruiling wrote:

@jmmartinez has suggested to simplify the test a little bit. But after some try, it generates more complex pattern to match. This maybe not the right solution for the problem. I will check whether we can turn such address calculation into more simple form before instruction selection. So we only need to care about less patterns during selection.

https://github.com/llvm/llvm-project/pull/178608