[Mlir-commits] [mlir] [MLIR][NVVM] Add Op for TMA Store with reduction (PR #118853)

[Krzysztof Drewniak]

================
@@ -2029,6 +2029,107 @@ def NVVM_CpAsyncBulkTensorPrefetchOp :
   }];
 }
 
+// List of modes supported for TMA Store and Reduction Ops
+def TMAStoreModeTile   : I32EnumAttrCase<"TILE", 0, "tile">;
+def TMAStoreModeIm2Col : I32EnumAttrCase<"IM2COL", 1, "im2col">;
+
+def TMAStoreMode : I32EnumAttr<"TMAStoreMode", "NVVM TMA Store Mode",
+    [TMAStoreModeTile, TMAStoreModeIm2Col]> {
+  let genSpecializedAttr = 0;
+  let cppNamespace = "::mlir::NVVM";
+}
+def TMAStoreModeAttr : EnumAttr<NVVM_Dialect, TMAStoreMode, "tma_store_mode"> {
+  let assemblyFormat = "`<` $value `>`";
+}
+
+// List of Reduction Ops supported with TMA Store
+def TMAReduxKindAdd : I32EnumAttrCase<"ADD", 0, "add">;
+def TMAReduxKindMin : I32EnumAttrCase<"MIN", 1, "min">;
+def TMAReduxKindMax : I32EnumAttrCase<"MAX", 2, "max">;
+def TMAReduxKindInc : I32EnumAttrCase<"INC", 3, "inc">;
+def TMAReduxKindDec : I32EnumAttrCase<"DEC", 4, "dec">;
+def TMAReduxKindAnd : I32EnumAttrCase<"AND", 5, "and">;
+def TMAReduxKindOr  : I32EnumAttrCase<"OR",  6, "or">;
+def TMAReduxKindXor : I32EnumAttrCase<"XOR", 7, "xor">;
+
+def TMAReduxKind : I32EnumAttr<"TMAReduxKind", "NVVM TMA redux kind",
+    [TMAReduxKindAdd, TMAReduxKindMax, TMAReduxKindMin,
+     TMAReduxKindInc, TMAReduxKindDec, TMAReduxKindAnd,
+     TMAReduxKindOr,  TMAReduxKindXor]> {
+  let genSpecializedAttr = 0;
+  let cppNamespace = "::mlir::NVVM";
+}
+def TMAReduxKindAttr : EnumAttr<NVVM_Dialect, TMAReduxKind, "tma_redux_kind"> {
+  let assemblyFormat = "`<` $value `>`";
+}
+
+def NVVM_CpAsyncBulkTensorReduceOp :
----------------
krzysz00 wrote:

Fair, I'm not going to get in the way of existing code and convention here. And perhaps LLVM intrinsics isn't exactly the right word here - the approach I've taken with AMDGPU/ROCDL is that, when there's a "prickly" feature (like, on our side, DPP, where, there're a lot of magic constants, or MFMA with its implicit type conversions), you get a more user-friendly wrapper under `amdgpu.*` and the raw underlying primitives that trivially translate to the export format in `rocdl.*`.

I was advocating for that approach because I figure it's both better developer experience and because it improves debugability.

https://github.com/llvm/llvm-project/pull/118853