[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 :
+    AMDGPU_Op<"global_load", [SameVariadicOperandSize]>,
+    Arguments<(ins
----------------
krzysz00 wrote:

This could be copying either one element at a time or be a gather of vectors.

Add a `TypeAttr:$copyType` or some such thing so that we know if it's going to gather `f16` or `vector<2xf16>` from the global memref

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