[DomTree] Remove unnecessary domtree level check in SemiNCA (NFC)

[nikic]

runSemiNCA() currently checks that the ReverseChildren are below MinLevel in the DT, which is used when performing incremental updates.

However, ReverseChildren is populated during runDFS with only the predecessors that are part of that DFS walk, which will itself be level limited in the relevant cases. As such, I don't believe that this should be checked during runSemiNCA().

This code probably dates back to a time when predecessors were not cached during runDFS and as such not limited to the visited subtree only.

[llvmbot]

@llvm/pr-subscribers-llvm-support

Author: Nikita Popov (nikic)

Changes

runSemiNCA() currently checks that the ReverseChildren are below MinLevel in the DT, which is used when performing incremental updates.

However, ReverseChildren is populated during runDFS with only the predecessors that are part of that DFS walk, which will itself be level limited in the relevant cases. As such, I don't believe that this should be checked during runSemiNCA().

This code probably dates back to a time when predecessors were not cached during runDFS and as such not limited to the visited subtree only.

---

Full diff: https://github.com/llvm/llvm-project/pull/73107.diff

1 Files Affected:

• (modified) llvm/include/llvm/Support/GenericDomTreeConstruction.h (+3-8)

diff --git a/llvm/include/llvm/Support/GenericDomTreeConstruction.h b/llvm/include/llvm/Support/GenericDomTreeConstruction.h
index 6564f98ab0234b0..ce893dbda27acd3 100644
--- a/llvm/include/llvm/Support/GenericDomTreeConstruction.h
+++ b/llvm/include/llvm/Support/GenericDomTreeConstruction.h
@@ -268,7 +268,7 @@ struct SemiNCAInfo {
   }
 
   // This function requires DFS to be run before calling it.
-  void runSemiNCA(DomTreeT &DT, const unsigned MinLevel = 0) {
+  void runSemiNCA(DomTreeT &DT) {
     const unsigned NextDFSNum(NumToNode.size());
     // Initialize IDoms to spanning tree parents.
     for (unsigned i = 1; i < NextDFSNum; ++i) {
@@ -289,11 +289,6 @@ struct SemiNCAInfo {
         assert(NodeToInfo.contains(N) &&
                "ReverseChildren should not contain unreachable predecessors");
 
-        const TreeNodePtr TN = DT.getNode(N);
-        // Skip predecessors whose level is above the subtree we are processing.
-        if (TN && TN->getLevel() < MinLevel)
-          continue;
-
         unsigned SemiU = NodeToInfo[eval(N, i + 1, EvalStack)].Semi;
         if (SemiU < WInfo.Semi) WInfo.Semi = SemiU;
       }
@@ -998,7 +993,7 @@ struct SemiNCAInfo {
     SemiNCAInfo SNCA(BUI);
     SNCA.runDFS(ToIDom, 0, DescendBelow, 0);
     LLVM_DEBUG(dbgs() << "\tRunning Semi-NCA\n");
-    SNCA.runSemiNCA(DT, Level);
+    SNCA.runSemiNCA(DT);
     SNCA.reattachExistingSubtree(DT, PrevIDomSubTree);
   }
 
@@ -1122,7 +1117,7 @@ struct SemiNCAInfo {
                       << BlockNamePrinter(PrevIDom) << "\nRunning Semi-NCA\n");
 
     // Rebuild the remaining part of affected subtree.
-    SNCA.runSemiNCA(DT, MinLevel);
+    SNCA.runSemiNCA(DT);
     SNCA.reattachExistingSubtree(DT, PrevIDom);
   }
 

[kuhar]

Do we have some data to confirm this branch is dynamically dead? IIUC, this patch relies on tis being the case.

[nikic]

I checked that if I replace this with an assertion, it does not fire in our test suite.

I opted not to keep the assertion, because this completely removes the dependency of runSemiNCA() on the dominator tree, which is imho cleaner.

[kuhar]

I think @alinas might have access to some fancy fleet-wide profiles -- could you confirm if this appears dead?

[kuhar]

LGTM, but I would be more confident if we could wait for a confirmation that this if body is dead in some fleed-wide profile.