[llvm] [InstCombine] Fold Minimum over trailing or leading zeros (PR #173768)

[Aryan Kadole via llvm-commits]

================
@@ -1700,12 +1702,18 @@ foldMinimumOverTrailingOrLeadingZeroCount(Value *I0, Value *I1,
   static_assert(IntrID == Intrinsic::cttz || IntrID == Intrinsic::ctlz,
                 "This helper only supports cttz and ctlz intrinsics");
 
-  Value *CtOp;
-  Value *ZeroUndef;
-  if (!match(I0,
-             m_OneUse(m_Intrinsic<IntrID>(m_Value(CtOp), m_Value(ZeroUndef)))))
+  Value *CtOp1, *CtOp2;
+  Value *ZeroUndef1, *ZeroUndef2;
+  if (!match(I0, m_OneUse(
+                     m_Intrinsic<IntrID>(m_Value(CtOp1), m_Value(ZeroUndef1)))))
     return nullptr;
 
+  if (match(I1,
+            m_OneUse(m_Intrinsic<IntrID>(m_Value(CtOp2), m_Value(ZeroUndef2)))))
+    return Builder.CreateBinaryIntrinsic(
+        IntrID, Builder.CreateOr(CtOp1, CtOp2),
+        ConstantInt::getTrue(ZeroUndef1->getType()));
----------------
ak1932 wrote:

Yes, I forgot to comment that I did not know what to set the flag to. Should it be true if either of the flags are true?

https://github.com/llvm/llvm-project/pull/173768