[llvm] [AMDGPU] Fix amdgcn.mbcnt known bits conflicting with the range attribute (PR #227725)

[Teja Alaghari via llvm-commits]

================
@@ -1744,6 +1744,10 @@ GCNTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
       return IC.replaceInstUsesWith(II, II.getArgOperand(1));
     [[fallthrough]];
   case Intrinsic::amdgcn_mbcnt_lo: {
+    // No lanes contribute when the mask is zero.
----------------
TejaX-Alaghari wrote:

Done

https://github.com/llvm/llvm-project/pull/227725