[DAG] visitSHL - move SimplifyDemandedBits after all standard folds to give them a chance to match

Pulled out of D155472

Author: RKSimon

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -9857,9 +9857,6 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
       return DAG.getNode(ISD::SHL, SDLoc(N), VT, N0, NewOp1);
   }
 
-  if (SimplifyDemandedBits(SDValue(N, 0)))
-    return SDValue(N, 0);
-
   // fold (shl (shl x, c1), c2) -> 0 or (shl x, (add c1, c2))
   if (N0.getOpcode() == ISD::SHL) {
     auto MatchOutOfRange = [OpSizeInBits](ConstantSDNode *LHS,
@@ -10077,6 +10074,9 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
     if (SDValue NewSHL = visitShiftByConstant(N))
       return NewSHL;
 
+  if (SimplifyDemandedBits(SDValue(N, 0)))
+    return SDValue(N, 0);
+
   // Fold (shl (vscale * C0), C1) to (vscale * (C0 << C1)).
   if (N0.getOpcode() == ISD::VSCALE && N1C) {
     const APInt &C0 = N0.getConstantOperandAPInt(0);

llvm/test/CodeGen/AArch64/ushl_sat.ll

@@ -128,9 +128,9 @@ define i16 @combine_shlsat_to_shl(i16 %x) nounwind {
 define i16 @combine_shlsat_to_shl_no_fold(i16 %x) nounwind {
 ; CHECK-LABEL: combine_shlsat_to_shl_no_fold:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    and w8, w0, #0xfffc
-; CHECK-NEXT:    lsl w9, w8, #17
-; CHECK-NEXT:    lsl w8, w8, #14
+; CHECK-NEXT:    lsl w8, w0, #14
+; CHECK-NEXT:    and w8, w8, #0x3fff0000
+; CHECK-NEXT:    lsl w9, w8, #3
 ; CHECK-NEXT:    cmp w8, w9, lsr #3
 ; CHECK-NEXT:    csinv w8, w9, wzr, eq
 ; CHECK-NEXT:    lsr w0, w8, #16