[llvm] [AArch64] Eliminate Common Subexpression of CSEL by Reassociation (PR #121350)

[David Green via llvm-commits]

================
@@ -24838,6 +24838,122 @@ static SDValue foldCSELOfCSEL(SDNode *Op, SelectionDAG &DAG) {
   return DAG.getNode(AArch64ISD::CSEL, DL, VT, L, R, CCValue, Cond);
 }
 
+// Reassociate the true/false expressions of a CSEL instruction to obtain a
+// common subexpression with the comparison instruction. For example, change
+// (CSEL (ADD (ADD x y) -c) f LO (SUBS x c)) to
+// (CSEL (ADD (SUBS x c) y) f LO (SUBS x c)) such that (SUBS x c) is a common
+// subexpression.
+static SDValue reassociateCSELOperandsForCSE(SDNode *N, SelectionDAG &DAG) {
+  SDValue SubsNode = N->getOperand(3);
+  if (SubsNode.getOpcode() != AArch64ISD::SUBS || !SubsNode.hasOneUse())
+    return SDValue();
+  auto *CmpOpConst = dyn_cast<ConstantSDNode>(SubsNode.getOperand(1));
+  if (!CmpOpConst)
+    return SDValue();
+
+  auto CC = static_cast<AArch64CC::CondCode>(N->getConstantOperandVal(2));
+  bool IsEquality = CC == AArch64CC::EQ || CC == AArch64CC::NE;
+  if (IsEquality && !CmpOpConst->isZero())
+    return SDValue();
+
+  SDValue CmpOpOther = SubsNode.getOperand(0);
+  EVT VT = N->getValueType(0);
+
+  auto Reassociate = [&](SDValue Op, APInt ExpectedConst, SDValue NewCmp) {
+    if (Op.getOpcode() != ISD::ADD)
+      return SDValue();
+    auto *AddOpConst = dyn_cast<ConstantSDNode>(Op.getOperand(1));
+    if (!AddOpConst)
+      return SDValue();
+    if (AddOpConst->getAPIntValue() != ExpectedConst)
+      return SDValue();
+    if (Op.getOperand(0).getOpcode() != ISD::ADD ||
+        !Op.getOperand(0).hasOneUse())
+      return SDValue();
+    SDValue X = Op.getOperand(0).getOperand(0);
+    SDValue Y = Op.getOperand(0).getOperand(1);
+    if (X != CmpOpOther)
+      std::swap(X, Y);
+    if (X != CmpOpOther)
+      return SDValue();
+    SDNodeFlags Flags;
+    if (Op.getOperand(0).getNode()->getFlags().hasNoUnsignedWrap())
+      Flags.setNoUnsignedWrap(true);
+    return DAG.getNode(ISD::ADD, SDLoc(Op), VT, NewCmp.getValue(0), Y, Flags);
+  };
+
+  auto Fold = [&](APInt NewCmpConst, AArch64CC::CondCode NewCC) {
+    SDValue NewCmp = DAG.getNode(AArch64ISD::SUBS, SDLoc(SubsNode),
----------------
davemgreen wrote:

It is best to only create nodes if we know they are needed, if possible.

https://github.com/llvm/llvm-project/pull/121350