Add codegen support for vector select (in the IR this means a select
with a vector condition); such selects become VSELECT codegen nodes.
This patch also removes VSETCC codegen nodes, unifying them with SETCC
nodes (codegen was actually often using SETCC for vector SETCC already).
This ensures that various DAG combiner optimizations kick in for vector
comparisons. Passes dragonegg bootstrap with no testsuite regressions
(nightly testsuite as well as "make check-all"). Patch mostly by
Nadav Rotem.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@139159 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index ec7bfbe..5f5a880 100644
--- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -927,13 +927,25 @@
assert(VT.isInteger() && "Cannot create FP integer constant!");
EVT EltVT = VT.getScalarType();
- assert(Val.getBitWidth() == EltVT.getSizeInBits() &&
- "APInt size does not match type size!");
+ const ConstantInt *Elt = &Val;
+ // In some cases the vector type is legal but the element type is illegal and
+ // needs to be promoted, for example v8i8 on ARM. In this case, promote the
+ // inserted value (the type does not need to match the vector element type).
+ // Any extra bits introduced will be truncated away.
+ if (VT.isVector() && TLI.getTypeAction(*getContext(), EltVT) ==
+ TargetLowering::TypePromoteInteger) {
+ EltVT = TLI.getTypeToTransformTo(*getContext(), EltVT);
+ APInt NewVal = Elt->getValue().zext(EltVT.getSizeInBits());
+ Elt = ConstantInt::get(*getContext(), NewVal);
+ }
+
+ assert(Elt->getBitWidth() == EltVT.getSizeInBits() &&
+ "APInt size does not match type size!");
unsigned Opc = isT ? ISD::TargetConstant : ISD::Constant;
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, getVTList(EltVT), 0, 0);
- ID.AddPointer(&Val);
+ ID.AddPointer(Elt);
void *IP = 0;
SDNode *N = NULL;
if ((N = CSEMap.FindNodeOrInsertPos(ID, IP)))
@@ -941,7 +953,7 @@
return SDValue(N, 0);
if (!N) {
- N = new (NodeAllocator) ConstantSDNode(isT, &Val, EltVT);
+ N = new (NodeAllocator) ConstantSDNode(isT, Elt, EltVT);
CSEMap.InsertNode(N, IP);
AllNodes.push_back(N);
}
@@ -1720,8 +1732,8 @@
// The boolean result conforms to getBooleanContents. Fall through.
case ISD::SETCC:
// If we know the result of a setcc has the top bits zero, use this info.
- if (TLI.getBooleanContents() == TargetLowering::ZeroOrOneBooleanContent &&
- BitWidth > 1)
+ if (TLI.getBooleanContents(Op.getValueType().isVector()) ==
+ TargetLowering::ZeroOrOneBooleanContent && BitWidth > 1)
KnownZero |= APInt::getHighBitsSet(BitWidth, BitWidth - 1);
return;
case ISD::SHL:
@@ -2155,7 +2167,7 @@
// The boolean result conforms to getBooleanContents. Fall through.
case ISD::SETCC:
// If setcc returns 0/-1, all bits are sign bits.
- if (TLI.getBooleanContents() ==
+ if (TLI.getBooleanContents(Op.getValueType().isVector()) ==
TargetLowering::ZeroOrNegativeOneBooleanContent)
return VTBits;
break;
@@ -5965,8 +5977,8 @@
case ISD::FPOWI: return "fpowi";
case ISD::SETCC: return "setcc";
- case ISD::VSETCC: return "vsetcc";
case ISD::SELECT: return "select";
+ case ISD::VSELECT: return "vselect";
case ISD::SELECT_CC: return "select_cc";
case ISD::INSERT_VECTOR_ELT: return "insert_vector_elt";
case ISD::EXTRACT_VECTOR_ELT: return "extract_vector_elt";