[llvm] [ARM] Speedups for CombineBaseUpdate. (PR #129725)

Peter Smith via llvm-commits llvm-commits at lists.llvm.org
Wed Mar 5 03:31:53 PST 2025


================
@@ -16360,8 +16367,8 @@ static SDValue PerformMVEVLDCombine(SDNode *N,
     Visited.insert(Addr.getNode());
     Worklist.push_back(N);
     Worklist.push_back(User);
-    if (SDNode::hasPredecessorHelper(N, Visited, Worklist) ||
-        SDNode::hasPredecessorHelper(User, Visited, Worklist))
+    if (SDNode::hasPredecessorHelper(N, Visited, Worklist, 1024) ||
----------------
smithp35 wrote:

We've got 1024 used in two different functions. Would it be worth making it a global constant with a comment explaining it.

The x86 and SystemZ backend seem to use Max uncommented, although it is local to the function.

https://github.com/llvm/llvm-project/pull/129725


More information about the llvm-commits mailing list