[Mlir-commits] [mlir] [mlir][OperationEquivalence] Add an extra callback to hook operation equivalence check (PR #73455)

[Ivan Butygin]

================
@@ -776,24 +775,39 @@ OperationEquivalence::isRegionEquivalentTo(Region *lhs, Region *rhs,
       flags);
 }
 
+/*static*/ llvm::hash_code OperationEquivalence::simpleOpHash(Operation *op) {
+  return llvm::hash_combine(op->getName(), op->getResultTypes(),
+                            op->hashProperties(),
+                            op->getDiscardableAttrDictionary());
+}
+
+/*static*/ LogicalResult
+OperationEquivalence::simpleOpEquivalence(Operation *lhs, Operation *rhs) {
+  return LogicalResult::success(
+      lhs->getName() == rhs->getName() &&
+      lhs->getDiscardableAttrDictionary() ==
+          rhs->getDiscardableAttrDictionary() &&
+      lhs->getNumRegions() == rhs->getNumRegions() &&
+      lhs->getNumSuccessors() == rhs->getNumSuccessors() &&
+      lhs->getNumOperands() == rhs->getNumOperands() &&
+      lhs->getNumResults() == rhs->getNumResults() &&
+      lhs->getName().compareOpProperties(lhs->getPropertiesStorage(),
+                                         rhs->getPropertiesStorage()));
+}
----------------
Hardcode84 wrote:

I agree with this argument, but I'm also not sure about the best granularity.
Flags give too little control, and introducing individual callbacks for each of those comparisons will be overkill.
Any suggestions?

(and we definitely need to switch to abstract class if we are going to introduce even more callbacks)

https://github.com/llvm/llvm-project/pull/73455