[llvm] [Instcombine] Fold away shift in or reduction chain. (PR #137875)

[David Green via llvm-commits]

================
@@ -7742,6 +7765,11 @@ Instruction *InstCombinerImpl::visitICmpInst(ICmpInst &I) {
     }
   }
 
+  // icmp eq/ne or(shl(a), b), 0 -> icmp eq/ne or(a, b)
+  if (I.isEquality() && match(Op1, m_Zero()) && isa<Instruction>(Op0))
----------------
davemgreen wrote:

Thanks - I wasn't sure which of foldICmpEquality, foldICmpBinOp, foldICmpWithZero, foldICmpWithConstant or.. foldICmpInstWithConstant (?) it would live best in. I've moved it to the end of foldICmpWithZero (and had to move the function earlier in the file).

https://github.com/llvm/llvm-project/pull/137875