[llvm] [CodeGen] Emit a more efficient magic number multiplication for exact udivs (PR #87161)

[Craig Topper via llvm-commits]

================
@@ -5197,12 +5249,8 @@ bool CombinerHelper::matchUDivByConst(MachineInstr &MI) {
       return false;
   }
 
-  auto CheckEltValue = [&](const Constant *C) {
-    if (auto *CI = dyn_cast_or_null<ConstantInt>(C))
-      return !CI->isZero();
-    return false;
-  };
-  return matchUnaryPredicate(MRI, RHS, CheckEltValue);
+  return matchUnaryPredicate(
+      MRI, RHS, [](const Constant *C) { return C && !C->isZeroValue(); });
----------------
topperc wrote:

Why using isZeroValue()? that's primarily intended for floating point since it allows -0.0. isNullValue is more correct here.

https://github.com/llvm/llvm-project/pull/87161