[Mlir-commits] [mlir] [mlir][ROCDL] Add async variants of pre-gfx12 LDS load intrinsics (PR #181072)

[Krzysztof Drewniak]

================
@@ -1121,6 +1164,27 @@ def ROCDL_GlobalLoadLDSOp :
   }];
 }
 
+def ROCDL_GlobalLoadAsyncLDSOp :
+  ROCDL_IntrOp<"global.load.async.lds", [], [], [], 0, 0, 1, 0, [2, 3, 4], ["size", "offset", "aux"]> {
+  dag args = (ins Arg<ROCDLGlobalBuffer, "", [MemRead]>:$globalPtr,
+                 Arg<ROCDLBufferLDS, "", [MemWrite]>:$ldsPtr,
+                 I32Attr:$size,
+                 I32Attr:$offset,
+                 I32Attr:$aux);
+  let arguments = !con(args, baseArgs);
+  let assemblyFormat = [{
+    $globalPtr `,`  $ldsPtr `,` $size `,` $offset `,` $aux
+    attr-dict `:` qualified(type($globalPtr)) `,` qualified(type($ldsPtr))
+  }];
+  let extraClassDefinition = [{
+    ::llvm::SmallVector<::mlir::Value> $cppClass::getAccessedOperands() {
+      return {getGlobalPtr(), getLdsPtr()};
+    }
+  }];
+
----------------
krzysz00 wrote:

What do you think would make sense for a description? I'd rather not repeat `load.async.to.lds`'s docs here

https://github.com/llvm/llvm-project/pull/181072