[Mlir-commits] [mlir] [mlir][amdgpu] Add make_dma_base operation (PR #169086)

[Krzysztof Drewniak]

================
@@ -79,6 +80,36 @@ def AMDGPU_AddressSpaceAttr : EnumAttr<AMDGPU_Dialect, AMDGPU_AddressSpace,
   let assemblyFormat = "`<` $value `>`";
 }
 
+class AMDGPU_Type<string name, string typeMnemonic, list<Trait> traits = []>
+    : TypeDef<AMDGPU_Dialect, name, traits> {
+  let mnemonic = typeMnemonic;
+}
+
+//===----------------------------------------------------------------------===//
+// AMDGPU Type definitions
+//===----------------------------------------------------------------------===//
+
+def AMDGPU_TDMBaseType : AMDGPU_Type<"TDMBase", "tdm_base"> {
+  // TODO:
+  // * Add verifiers such that one of the memrefs is from LDS and the other global.
+  // * Add verifiers to make sure that the type is in the correct direction.
+  // * Add verifiers to make sure that the number of indices do not exceed the number of dimensions.
+
+  let summary = "Pair of base addresses that move data between LDS and global storage.";
+  let description = [{
+    This type is opaque and it is used to represent a struct of two addresses.
+    One address is in LDS while the other is in global memory.
+  }];
+  let parameters = (ins "Type":$elementType);
+  let builders = [
+    TypeBuilderWithInferredContext<(ins "Type":$elementType), [{
+      return $_get(elementType.getContext(), elementType);
+    }]>
+  ];
+  let assemblyFormat = "`<` $elementType `>`";
----------------
krzysz00 wrote:

I'd disagree 

It's entirely possible to use the same descriptior to load a tyle and then store it back

That's not an error, that's an intended usecase

https://github.com/llvm/llvm-project/pull/169086