diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
index 48375a1a323f..e7371d28c745 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
@@ -1432,6 +1432,13 @@ Instruction *InstCombiner::foldICmpWithZero(ICmpInst &Cmp) {
 /// possible that code has been made unnecessary - do we canonicalize IR to
 /// overflow/saturating intrinsics or not?).
 Instruction *InstCombiner::foldICmpWithConstant(ICmpInst &Cmp) {
+  Value *Op0 = Cmp.getOperand(0), *Op1 = Cmp.getOperand(1);
+  ConstantInt *CI;
+
+  // Make sure that the RHS operand is a constant.
+  if (!match(Op1, m_ConstantInt(CI)))
+    return nullptr;
+
   // Match the following pattern, which is a common idiom when writing
   // overflow-safe integer arithmetic functions. The source performs an addition
   // in wider type and explicitly checks for overflow using comparisons against
@@ -1444,10 +1451,9 @@ Instruction *InstCombiner::foldICmpWithConstant(ICmpInst &Cmp) {
   // sum = a + b
   // if (sum+128 >u 255)  ...  -> llvm.sadd.with.overflow.i8
   CmpInst::Predicate Pred = Cmp.getPredicate();
-  Value *Op0 = Cmp.getOperand(0), *Op1 = Cmp.getOperand(1);
   Value *A, *B;
-  ConstantInt *CI, *CI2; // I = icmp ugt (add (add A, B), CI2), CI
-  if (Pred == ICmpInst::ICMP_UGT && match(Op1, m_ConstantInt(CI)) &&
+  ConstantInt *CI2; // I = icmp ugt (add (add A, B), CI2), CI
+  if (Pred == ICmpInst::ICMP_UGT &&
       match(Op0, m_Add(m_Add(m_Value(A), m_Value(B)), m_ConstantInt(CI2))))
     if (Instruction *Res = processUGT_ADDCST_ADD(Cmp, A, B, CI2, CI, *this))
       return Res;