[llvm] [X86] optimize ssse3 horizontal saturating add/sub (PR #169591)

[Phoebe Wang via llvm-commits]

================
@@ -54084,6 +54090,27 @@ static SDValue combineToHorizontalAddSub(SDNode *N, SelectionDAG &DAG,
       }
     }
     break;
+  case ISD::SADDSAT:
+  case ISD::SSUBSAT:
+    if (Subtarget.hasSSSE3() && VT == MVT::v8i16) {
----------------
phoebewang wrote:

Yes, AVX/AVX2 patterns are both required.

https://github.com/llvm/llvm-project/pull/169591