[Mlir-commits] [mlir] [acc] Consistency between acc.loop and acc compute ops (PR #114549)

[Valentin Clement バレンタイン クレメン]

================
@@ -1145,6 +1145,9 @@ def OpenACC_ParallelOp : OpenACC_Op<"parallel",
     /// The i-th data operand passed.
     Value getDataOperand(unsigned i);
 
+    /// Used to retrieve the block inside the op's region.
+    Block &getBody() { return getRegion().front(); }
----------------
clementval wrote:

I think in the long term it would be nice to define an interface `OpenACCComputeRegionOp` or smth like that so we make sure all ops implement what we expect. This can be done later. 

https://github.com/llvm/llvm-project/pull/114549