[SDAG] Address post commit review feedback from f8c63a7f

The major change is falling through to ComputeKnownBits when we don't have an implementation of ComputeNumSignBits due to conservatism over scalable vectors.  Right now, we're mostly conservative in the same cases, but this allows our results to improve when we change ComputeKnownBits without also needing to improve ComputeNumSignBits at the same time.

diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index b562949..cba592f 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -4058,7 +4058,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
 
   case ISD::BITCAST: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     SDValue N0 = Op.getOperand(0);
     EVT SrcVT = N0.getValueType();
     unsigned SrcBits = SrcVT.getScalarSizeInBits();
@@ -4117,7 +4117,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
     return std::max(Tmp, Tmp2);
   case ISD::SIGN_EXTEND_VECTOR_INREG: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     SDValue Src = Op.getOperand(0);
     EVT SrcVT = Src.getValueType();
     APInt DemandedSrcElts = DemandedElts.zext(SrcVT.getVectorNumElements());
@@ -4336,7 +4336,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
   }
   case ISD::EXTRACT_ELEMENT: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     const int KnownSign = ComputeNumSignBits(Op.getOperand(0), Depth+1);
     const int BitWidth = Op.getValueSizeInBits();
     const int Items = Op.getOperand(0).getValueSizeInBits() / BitWidth;
@@ -4351,7 +4351,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
   }
   case ISD::INSERT_VECTOR_ELT: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     // If we know the element index, split the demand between the
     // source vector and the inserted element, otherwise assume we need
     // the original demanded vector elements and the value.
@@ -4382,8 +4382,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
     return Tmp;
   }
   case ISD::EXTRACT_VECTOR_ELT: {
-    if (VT.isScalableVector())
-      return 1;
+    assert(!VT.isScalableVector());
     SDValue InVec = Op.getOperand(0);
     SDValue EltNo = Op.getOperand(1);
     EVT VecVT = InVec.getValueType();
@@ -4423,7 +4422,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
   }
   case ISD::CONCAT_VECTORS: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     // Determine the minimum number of sign bits across all demanded
     // elts of the input vectors. Early out if the result is already 1.
     Tmp = std::numeric_limits<unsigned>::max();
@@ -4443,7 +4442,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
   }
   case ISD::INSERT_SUBVECTOR: {
     if (VT.isScalableVector())
-      return 1;
+      break;
     // Demand any elements from the subvector and the remainder from the src its
     // inserted into.
     SDValue Src = Op.getOperand(0);
@@ -4551,12 +4550,12 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
       Opcode == ISD::INTRINSIC_VOID) {
     // TODO: This can probably be removed once target code is audited.  This
     // is here purely to reduce patch size and review complexity.
-    if (VT.isScalableVector())
-      return 1;
-    unsigned NumBits =
+    if (!VT.isScalableVector()) {
+      unsigned NumBits =
         TLI->ComputeNumSignBitsForTargetNode(Op, DemandedElts, *this, Depth);
-    if (NumBits > 1)
-      FirstAnswer = std::max(FirstAnswer, NumBits);
+      if (NumBits > 1)
+        FirstAnswer = std::max(FirstAnswer, NumBits);
+    }
   }
 
   // Finally, if we can prove that the top bits of the result are 0's or 1's,