[llvm] [AArch64] Convert concat(uhadd(a,b), uhadd(c,d)) to uhadd(concat(a,c), concat(b,d)) (PR #80674)

[David Sherwood via llvm-commits]

================
@@ -18218,50 +18218,23 @@ static SDValue performConcatVectorsCombine(SDNode *N,
   if (DCI.isBeforeLegalizeOps())
     return SDValue();
 
-  // Optimise concat_vectors of two [us]avgceils or [us]avgfloors that use
-  // extracted subvectors from the same original vectors. Combine these into a
-  // single avg that operates on the two original vectors.
-  // avgceil is the target independant name for rhadd, avgfloor is a hadd.
-  // Example:
-  //  (concat_vectors (v8i8 (avgceils (extract_subvector (v16i8 OpA, <0>),
-  //                                   extract_subvector (v16i8 OpB, <0>))),
-  //                  (v8i8 (avgceils (extract_subvector (v16i8 OpA, <8>),
-  //                                   extract_subvector (v16i8 OpB, <8>)))))
-  // ->
-  //  (v16i8(avgceils(v16i8 OpA, v16i8 OpB)))
-  if (N->getNumOperands() == 2 && N0Opc == N1Opc &&
+  // Optimise concat_vectors of two [us]avgceils or [us]avgfloors with a 128-bit
+  // destination size, combine into an avg of two contacts of the source
+  // vectors. eg: concat(uhadd(a,b), uhadd(c, d)) -> uhadd(concat(a, c),
+  // concat(b, d))
+  if (N->getNumOperands() == 2 && N0Opc == N1Opc && VT.is128BitVector() &&
       (N0Opc == ISD::AVGCEILU || N0Opc == ISD::AVGCEILS ||
-       N0Opc == ISD::AVGFLOORU || N0Opc == ISD::AVGFLOORS)) {
+       N0Opc == ISD::AVGFLOORU || N0Opc == ISD::AVGFLOORS) &&
+      N0->hasOneUse() && N1->hasOneUse()) {
     SDValue N00 = N0->getOperand(0);
     SDValue N01 = N0->getOperand(1);
     SDValue N10 = N1->getOperand(0);
     SDValue N11 = N1->getOperand(1);
 
-    EVT N00VT = N00.getValueType();
-    EVT N10VT = N10.getValueType();
-
-    if (N00->getOpcode() == ISD::EXTRACT_SUBVECTOR &&
-        N01->getOpcode() == ISD::EXTRACT_SUBVECTOR &&
-        N10->getOpcode() == ISD::EXTRACT_SUBVECTOR &&
-        N11->getOpcode() == ISD::EXTRACT_SUBVECTOR && N00VT == N10VT) {
-      SDValue N00Source = N00->getOperand(0);
-      SDValue N01Source = N01->getOperand(0);
-      SDValue N10Source = N10->getOperand(0);
-      SDValue N11Source = N11->getOperand(0);
-
-      if (N00Source == N10Source && N01Source == N11Source &&
-          N00Source.getValueType() == VT && N01Source.getValueType() == VT) {
-        assert(N0.getValueType() == N1.getValueType());
-
-        uint64_t N00Index = N00.getConstantOperandVal(1);
-        uint64_t N01Index = N01.getConstantOperandVal(1);
-        uint64_t N10Index = N10.getConstantOperandVal(1);
-        uint64_t N11Index = N11.getConstantOperandVal(1);
-
-        if (N00Index == N01Index && N10Index == N11Index && N00Index == 0 &&
-            N10Index == N00VT.getVectorNumElements())
-          return DAG.getNode(N0Opc, dl, VT, N00Source, N01Source);
-      }
+    if (!N00.isUndef() && !N01.isUndef() && !N10.isUndef() && !N11.isUndef()) {
----------------
david-arm wrote:

Is it worth having a negative test for the undef case to show that we don't optimise this case?

https://github.com/llvm/llvm-project/pull/80674