[llvm] [InstCombine] Missed optimization: Fold (sext(a) & c1) == c2 to (a & c3) == trunc(c2) (PR #112646)

[Nikita Popov via llvm-commits]

================
@@ -7716,6 +7716,32 @@ Instruction *InstCombinerImpl::visitICmpInst(ICmpInst &I) {
   if (Instruction *Res = foldReductionIdiom(I, Builder, DL))
     return Res;
 
+  {
+    Value *A;
+    const APInt *C1, *C2;
+    ICmpInst::Predicate Pred = I.getPredicate();
+    if (ICmpInst::isEquality(Pred)) {
+      // sext(a) & c1 == c2 --> a & c3 == trunc(c2)
+      // sext(a) & c1 != c2 --> a & c3 != trunc(c2)
+      if (match(Op0, m_And(m_SExtLike(m_Value(A)), m_APInt(C1))) &&
----------------
nikic wrote:

If m_SExtLike doing anything here? I'd expect that if this is a zext nneg this will already get folded via other code. (If it *is* doing anything, there should be a test that demonstrates that.)

https://github.com/llvm/llvm-project/pull/112646