[Mlir-commits] [mlir] [MLIR][AMDGPU] Add a wrapper for global LDS load intrinsics in AMDGPU (PR #133498)

[Krzysztof Drewniak]

================
@@ -765,4 +771,37 @@ def AMDGPU_WMMAOp :
   let hasVerifier = 1;
 }
 
+def GlobalLoadMemRefType : MemRefOf<[GlobalLoadTypes]>;
+
+def AMDGPU_GlobalLoadLDSOp :
----------------
krzysz00 wrote:

1. This doesn't need "global" in the name - it'll work for buffer intrinsic as well
2. This operation has some rather strong and nonobvious gather semantics - could it be `amdgpu.gather_to_lds` instead? Or `amdgpu.direct_lds_gather`? ... 

https://github.com/llvm/llvm-project/pull/133498