[Mlir-commits] [mlir] [mlir][scf] Refactor and improve ParallelLoopFusion (PR #179284)

[llvmlistbot at llvm.org]

================
@@ -43,6 +43,12 @@ def SCFForLoopSpecialization : Pass<"scf-for-loop-specialization"> {
 def SCFParallelLoopFusion : Pass<"scf-parallel-loop-fusion"> {
   let summary = "Fuse adjacent parallel loops";
   let constructor = "mlir::createParallelLoopFusionPass()";
+  let options = [
+    Option<"strict", "strict", "bool",
----------------
fabrizio-indirli wrote:

On second thought, I removed the option. I think it could have given the wrong impression that not doing that check was fine, and that one would check that only if being ultra-strict...while in reality it's something that we should always be doing, but for now we skip it because of implementation details. 
I also didn't want to add a test with some op that currently doesn't implement the _MemoryEffectInterface_, only for it to start failing the moment that someone (rightfully) implements the interface for that op in the future.
So a TODO is a better choice IMHO.

https://github.com/llvm/llvm-project/pull/179284