[Mlir-commits] [mlir] [MLIR][LLVM] Support Recursive DITypes (PR #80251)

[Billy Zhu]

================
@@ -289,7 +316,19 @@ DINodeAttr DebugImporter::translate(llvm::DINode *node) {
     return nullptr;
   };
   if (DINodeAttr attr = translateNode(node)) {
-    nodeToAttr.insert({node, attr});
+    // If this node was marked as recursive, set its recId.
+    if (auto recType = dyn_cast<DIRecursiveTypeAttrInterface>(attr)) {
+      if (DistinctAttr recId = translationStack.lookup(node)) {
+        attr = cast<DINodeAttr>(recType.withRecId(recId));
+        // Remove the unbound recursive ID from the set of unbound self
+        // references in the translation stack.
+        unboundRecursiveSelfRefs.back().erase(recId);
+      }
+    }
+
+    // Only cache fully self-contained nodes.
+    if (unboundRecursiveSelfRefs.back().empty())
----------------
zyx-billy wrote:

oh very interesting... Is it something like this
```
// LLVM
+---> B ----+
|           v
A <-------- D
|           ^
+---> C ----+

// MLIR
A -> B -> D -> A'
+--> C -> D -> A'
```
So now in MLIR you have two copies of the subgraph `D -> A'`, but they're different because we don't cache it?

I like your idea of a temporary cache that keys on the set of unbounded references. This way repeated subgraphs can be merged too. I can get started on something for this. Please let me know if this is not what you were looking for though 😂.

https://github.com/llvm/llvm-project/pull/80251