[Mlir-commits] [mlir] [MLIR][NVVM] Update TMA Store Op (PR #155435)

[Guray Ozen]

================
@@ -2353,6 +2353,20 @@ def TMALoadModeAttr : EnumAttr<NVVM_Dialect, TMALoadMode, "tma_load_mode"> {
   let assemblyFormat = "`<` $value `>`";
 }
 
+// List of modes supported for TMA Store and Reduction Ops
+def TMAStoreModeTile   : I32EnumAttrCase<"TILE", 0, "tile">;
+def TMAStoreModeIm2Col : I32EnumAttrCase<"IM2COL", 1, "im2col">;
+def TMAStoreModeTileScatter4 : I32EnumAttrCase<"TILE_SCATTER4", 2, "tile_scatter4">;
----------------
grypp wrote:

I don't have a strong opinion. It reduces some tablegen code, but increases verifier code. 

https://github.com/llvm/llvm-project/pull/155435