case ISD::ANY_EXTEND:
case ISD::BSWAP:
case ISD::CTLZ:
+ case ISD::CTLZ_ZERO_UNDEF:
case ISD::CTPOP:
case ISD::CTTZ:
+ case ISD::CTTZ_ZERO_UNDEF:
case ISD::FABS:
case ISD::FCEIL:
case ISD::FCOS:
case ISD::FCOPYSIGN:
case ISD::FDIV:
case ISD::FMUL:
+ case ISD::FMINNUM:
+ case ISD::FMAXNUM:
+
case ISD::FPOW:
case ISD::FREM:
case ISD::FSUB:
N->getMemoryVT().getVectorElementType(),
N->isVolatile(), N->isNonTemporal(),
N->isInvariant(), N->getOriginalAlignment(),
- N->getTBAAInfo());
+ N->getAAInfo());
// Legalized the chain result - switch anything that used the old chain to
// use the new one.
SDValue DAGTypeLegalizer::ScalarizeVecRes_UnaryOp(SDNode *N) {
// Get the dest type - it doesn't always match the input type, e.g. int_to_fp.
EVT DestVT = N->getValueType(0).getVectorElementType();
- SDValue Op = GetScalarizedVector(N->getOperand(0));
+ SDValue Op = N->getOperand(0);
+ EVT OpVT = Op.getValueType();
+ SDLoc DL(N);
+ // The result needs scalarizing, but it's not a given that the source does.
+ // This is a workaround for targets where it's impossible to scalarize the
+ // result of a conversion, because the source type is legal.
+ // For instance, this happens on AArch64: v1i1 is illegal but v1i{8,16,32}
+ // are widened to v8i8, v4i16, and v2i32, which is legal, because v1i64 is
+ // legal and was not scalarized.
+ // See the similar logic in ScalarizeVecRes_VSETCC
+ if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
+ Op = GetScalarizedVector(Op);
+ } else {
+ EVT VT = OpVT.getVectorElementType();
+ Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Op,
+ DAG.getConstant(0, TLI.getVectorIdxTy()));
+ }
return DAG.getNode(N->getOpcode(), SDLoc(N), DestVT, Op);
}
SDValue DAGTypeLegalizer::ScalarizeVecRes_VSELECT(SDNode *N) {
SDValue Cond = GetScalarizedVector(N->getOperand(0));
SDValue LHS = GetScalarizedVector(N->getOperand(1));
- TargetLowering::BooleanContent ScalarBool = TLI.getBooleanContents(false);
- TargetLowering::BooleanContent VecBool = TLI.getBooleanContents(true);
+ TargetLowering::BooleanContent ScalarBool =
+ TLI.getBooleanContents(false, false);
+ TargetLowering::BooleanContent VecBool = TLI.getBooleanContents(true, false);
+
+ // If integer and float booleans have different contents then we can't
+ // reliably optimize in all cases. There is a full explanation for this in
+ // DAGCombiner::visitSELECT() where the same issue affects folding
+ // (select C, 0, 1) to (xor C, 1).
+ if (TLI.getBooleanContents(false, false) !=
+ TLI.getBooleanContents(false, true)) {
+ // At least try the common case where the boolean is generated by a
+ // comparison.
+ if (Cond->getOpcode() == ISD::SETCC) {
+ EVT OpVT = Cond->getOperand(0)->getValueType(0);
+ ScalarBool = TLI.getBooleanContents(OpVT.getScalarType());
+ VecBool = TLI.getBooleanContents(OpVT);
+ } else
+ ScalarBool = TargetLowering::UndefinedBooleanContent;
+ }
+
if (ScalarBool != VecBool) {
EVT CondVT = Cond.getValueType();
switch (ScalarBool) {
// Vectors may have a different boolean contents to scalars. Promote the
// value appropriately.
ISD::NodeType ExtendCode =
- TargetLowering::getExtendForContent(TLI.getBooleanContents(true));
+ TargetLowering::getExtendForContent(TLI.getBooleanContents(OpVT));
return DAG.getNode(ExtendCode, DL, NVT, Res);
}
case ISD::ZERO_EXTEND:
case ISD::SIGN_EXTEND:
case ISD::TRUNCATE:
+ case ISD::FP_TO_SINT:
+ case ISD::FP_TO_UINT:
+ case ISD::SINT_TO_FP:
+ case ISD::UINT_TO_FP:
Res = ScalarizeVecOp_UnaryOp(N);
break;
case ISD::CONCAT_VECTORS:
N->getValueType(0), Elt);
}
-/// ScalarizeVecOp_EXTEND - If the value to extend is a vector that needs
-/// to be scalarized, it must be <1 x ty>. Extend the element instead.
+/// ScalarizeVecOp_UnaryOp - If the input is a vector that needs to be
+/// scalarized, it must be <1 x ty>. Do the operation on the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_UnaryOp(SDNode *N) {
assert(N->getValueType(0).getVectorNumElements() == 1 &&
- "Unexected vector type!");
+ "Unexpected vector type!");
SDValue Elt = GetScalarizedVector(N->getOperand(0));
- SmallVector<SDValue, 1> Ops(1);
- Ops[0] = DAG.getNode(N->getOpcode(), SDLoc(N),
- N->getValueType(0).getScalarType(), Elt);
+ SDValue Op = DAG.getNode(N->getOpcode(), SDLoc(N),
+ N->getValueType(0).getScalarType(), Elt);
// Revectorize the result so the types line up with what the uses of this
// expression expect.
- return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N), N->getValueType(0),
- &Ops[0], 1);
+ return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N), N->getValueType(0), Op);
}
/// ScalarizeVecOp_CONCAT_VECTORS - The vectors to concatenate have length one -
SmallVector<SDValue, 8> Ops(N->getNumOperands());
for (unsigned i = 0, e = N->getNumOperands(); i < e; ++i)
Ops[i] = GetScalarizedVector(N->getOperand(i));
- return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N), N->getValueType(0),
- &Ops[0], Ops.size());
+ return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N), N->getValueType(0), Ops);
}
/// ScalarizeVecOp_EXTRACT_VECTOR_ELT - If the input is a vector that needs to
N->getBasePtr(), N->getPointerInfo(),
N->getMemoryVT().getVectorElementType(),
N->isVolatile(), N->isNonTemporal(),
- N->getAlignment(), N->getTBAAInfo());
+ N->getAlignment(), N->getAAInfo());
return DAG.getStore(N->getChain(), dl, GetScalarizedVector(N->getOperand(1)),
N->getBasePtr(), N->getPointerInfo(),
N->isVolatile(), N->isNonTemporal(),
- N->getOriginalAlignment(), N->getTBAAInfo());
+ N->getOriginalAlignment(), N->getAAInfo());
}
/// ScalarizeVecOp_FP_ROUND - If the value to round is a vector that needs
case ISD::BUILD_VECTOR: SplitVecRes_BUILD_VECTOR(N, Lo, Hi); break;
case ISD::CONCAT_VECTORS: SplitVecRes_CONCAT_VECTORS(N, Lo, Hi); break;
case ISD::EXTRACT_SUBVECTOR: SplitVecRes_EXTRACT_SUBVECTOR(N, Lo, Hi); break;
+ case ISD::INSERT_SUBVECTOR: SplitVecRes_INSERT_SUBVECTOR(N, Lo, Hi); break;
case ISD::FP_ROUND_INREG: SplitVecRes_InregOp(N, Lo, Hi); break;
case ISD::FPOWI: SplitVecRes_FPOWI(N, Lo, Hi); break;
case ISD::INSERT_VECTOR_ELT: SplitVecRes_INSERT_VECTOR_ELT(N, Lo, Hi); break;
case ISD::LOAD:
SplitVecRes_LOAD(cast<LoadSDNode>(N), Lo, Hi);
break;
+ case ISD::MLOAD:
+ SplitVecRes_MLOAD(cast<MaskedLoadSDNode>(N), Lo, Hi);
+ break;
case ISD::SETCC:
SplitVecRes_SETCC(N, Lo, Hi);
break;
case ISD::FCOPYSIGN:
case ISD::FSUB:
case ISD::FMUL:
+ case ISD::FMINNUM:
+ case ISD::FMAXNUM:
case ISD::SDIV:
case ISD::UDIV:
case ISD::FDIV:
std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
unsigned LoNumElts = LoVT.getVectorNumElements();
SmallVector<SDValue, 8> LoOps(N->op_begin(), N->op_begin()+LoNumElts);
- Lo = DAG.getNode(ISD::BUILD_VECTOR, dl, LoVT, &LoOps[0], LoOps.size());
+ Lo = DAG.getNode(ISD::BUILD_VECTOR, dl, LoVT, LoOps);
SmallVector<SDValue, 8> HiOps(N->op_begin()+LoNumElts, N->op_end());
- Hi = DAG.getNode(ISD::BUILD_VECTOR, dl, HiVT, &HiOps[0], HiOps.size());
+ Hi = DAG.getNode(ISD::BUILD_VECTOR, dl, HiVT, HiOps);
}
void DAGTypeLegalizer::SplitVecRes_CONCAT_VECTORS(SDNode *N, SDValue &Lo,
std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
SmallVector<SDValue, 8> LoOps(N->op_begin(), N->op_begin()+NumSubvectors);
- Lo = DAG.getNode(ISD::CONCAT_VECTORS, dl, LoVT, &LoOps[0], LoOps.size());
+ Lo = DAG.getNode(ISD::CONCAT_VECTORS, dl, LoVT, LoOps);
SmallVector<SDValue, 8> HiOps(N->op_begin()+NumSubvectors, N->op_end());
- Hi = DAG.getNode(ISD::CONCAT_VECTORS, dl, HiVT, &HiOps[0], HiOps.size());
+ Hi = DAG.getNode(ISD::CONCAT_VECTORS, dl, HiVT, HiOps);
}
void DAGTypeLegalizer::SplitVecRes_EXTRACT_SUBVECTOR(SDNode *N, SDValue &Lo,
TLI.getVectorIdxTy()));
}
+void DAGTypeLegalizer::SplitVecRes_INSERT_SUBVECTOR(SDNode *N, SDValue &Lo,
+ SDValue &Hi) {
+ SDValue Vec = N->getOperand(0);
+ SDValue SubVec = N->getOperand(1);
+ SDValue Idx = N->getOperand(2);
+ SDLoc dl(N);
+ GetSplitVector(Vec, Lo, Hi);
+
+ // Spill the vector to the stack.
+ EVT VecVT = Vec.getValueType();
+ EVT SubVecVT = VecVT.getVectorElementType();
+ SDValue StackPtr = DAG.CreateStackTemporary(VecVT);
+ SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr,
+ MachinePointerInfo(), false, false, 0);
+
+ // Store the new subvector into the specified index.
+ SDValue SubVecPtr = GetVectorElementPointer(StackPtr, SubVecVT, Idx);
+ Type *VecType = VecVT.getTypeForEVT(*DAG.getContext());
+ unsigned Alignment = TLI.getDataLayout()->getPrefTypeAlignment(VecType);
+ Store = DAG.getStore(Store, dl, SubVec, SubVecPtr, MachinePointerInfo(),
+ false, false, 0);
+
+ // Load the Lo part from the stack slot.
+ Lo = DAG.getLoad(Lo.getValueType(), dl, Store, StackPtr, MachinePointerInfo(),
+ false, false, false, 0);
+
+ // Increment the pointer to the other part.
+ unsigned IncrementSize = Lo.getValueType().getSizeInBits() / 8;
+ StackPtr =
+ DAG.getNode(ISD::ADD, dl, StackPtr.getValueType(), StackPtr,
+ DAG.getConstant(IncrementSize, StackPtr.getValueType()));
+
+ // Load the Hi part from the stack slot.
+ Hi = DAG.getLoad(Hi.getValueType(), dl, Store, StackPtr, MachinePointerInfo(),
+ false, false, false, MinAlign(Alignment, IncrementSize));
+}
+
void DAGTypeLegalizer::SplitVecRes_FPOWI(SDNode *N, SDValue &Lo,
SDValue &Hi) {
SDLoc dl(N);
return;
}
+ // See if the target wants to custom expand this node.
+ if (CustomLowerNode(N, N->getValueType(0), true))
+ return;
+
// Spill the vector to the stack.
EVT VecVT = Vec.getValueType();
EVT EltVT = VecVT.getVectorElementType();
bool isVolatile = LD->isVolatile();
bool isNonTemporal = LD->isNonTemporal();
bool isInvariant = LD->isInvariant();
- const MDNode *TBAAInfo = LD->getTBAAInfo();
+ AAMDNodes AAInfo = LD->getAAInfo();
EVT LoMemVT, HiMemVT;
std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
Lo = DAG.getLoad(ISD::UNINDEXED, ExtType, LoVT, dl, Ch, Ptr, Offset,
LD->getPointerInfo(), LoMemVT, isVolatile, isNonTemporal,
- isInvariant, Alignment, TBAAInfo);
+ isInvariant, Alignment, AAInfo);
unsigned IncrementSize = LoMemVT.getSizeInBits()/8;
Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr,
Hi = DAG.getLoad(ISD::UNINDEXED, ExtType, HiVT, dl, Ch, Ptr, Offset,
LD->getPointerInfo().getWithOffset(IncrementSize),
HiMemVT, isVolatile, isNonTemporal, isInvariant, Alignment,
- TBAAInfo);
+ AAInfo);
// Build a factor node to remember that this load is independent of the
// other one.
ReplaceValueWith(SDValue(LD, 1), Ch);
}
+void DAGTypeLegalizer::SplitVecRes_MLOAD(MaskedLoadSDNode *MLD,
+ SDValue &Lo, SDValue &Hi) {
+ EVT LoVT, HiVT;
+ SDLoc dl(MLD);
+ std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(MLD->getValueType(0));
+
+ SDValue Ch = MLD->getChain();
+ SDValue Ptr = MLD->getBasePtr();
+ SDValue Mask = MLD->getMask();
+ unsigned Alignment = MLD->getOriginalAlignment();
+ ISD::LoadExtType ExtType = MLD->getExtensionType();
+
+ // if Alignment is equal to the vector size,
+ // take the half of it for the second part
+ unsigned SecondHalfAlignment =
+ (Alignment == MLD->getValueType(0).getSizeInBits()/8) ?
+ Alignment/2 : Alignment;
+
+ SDValue MaskLo, MaskHi;
+ std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl);
+
+ EVT MemoryVT = MLD->getMemoryVT();
+ EVT LoMemVT, HiMemVT;
+ std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
+
+ SDValue Src0 = MLD->getSrc0();
+ SDValue Src0Lo, Src0Hi;
+ std::tie(Src0Lo, Src0Hi) = DAG.SplitVector(Src0, dl);
+
+ MachineMemOperand *MMO = DAG.getMachineFunction().
+ getMachineMemOperand(MLD->getPointerInfo(),
+ MachineMemOperand::MOLoad, LoMemVT.getStoreSize(),
+ Alignment, MLD->getAAInfo(), MLD->getRanges());
+
+ Lo = DAG.getMaskedLoad(LoVT, dl, Ch, Ptr, MaskLo, Src0Lo, LoMemVT, MMO,
+ ExtType);
+
+ unsigned IncrementSize = LoMemVT.getSizeInBits()/8;
+ Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr,
+ DAG.getConstant(IncrementSize, Ptr.getValueType()));
+
+ MMO = DAG.getMachineFunction().
+ getMachineMemOperand(MLD->getPointerInfo(),
+ MachineMemOperand::MOLoad, HiMemVT.getStoreSize(),
+ SecondHalfAlignment, MLD->getAAInfo(), MLD->getRanges());
+
+ Hi = DAG.getMaskedLoad(HiVT, dl, Ch, Ptr, MaskHi, Src0Hi, HiMemVT, MMO,
+ ExtType);
+
+
+ // Build a factor node to remember that this load is independent of the
+ // other one.
+ Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
+ Hi.getValue(1));
+
+ // Legalized the chain result - switch anything that used the old chain to
+ // use the new one.
+ ReplaceValueWith(SDValue(MLD, 1), Ch);
+
+}
+
void DAGTypeLegalizer::SplitVecRes_SETCC(SDNode *N, SDValue &Lo, SDValue &Hi) {
assert(N->getValueType(0).isVector() &&
N->getOperand(0).getValueType().isVector() &&
}
// Construct the Lo/Hi output using a BUILD_VECTOR.
- Output = DAG.getNode(ISD::BUILD_VECTOR,dl,NewVT, &SVOps[0], SVOps.size());
+ Output = DAG.getNode(ISD::BUILD_VECTOR, dl, NewVT, SVOps);
} else if (InputUsed[0] == -1U) {
// No input vectors were used! The result is undefined.
Output = DAG.getUNDEF(NewVT);
case ISD::STORE:
Res = SplitVecOp_STORE(cast<StoreSDNode>(N), OpNo);
break;
+ case ISD::MSTORE:
+ Res = SplitVecOp_MSTORE(cast<MaskedStoreSDNode>(N), OpNo);
+ break;
case ISD::VSELECT:
Res = SplitVecOp_VSELECT(N, OpNo);
break;
Idx.getValueType())), 0);
}
+ // See if the target wants to custom expand this node.
+ if (CustomLowerNode(N, N->getValueType(0), true))
+ return SDValue();
+
// Store the vector to the stack.
EVT EltVT = VecVT.getVectorElementType();
SDLoc dl(N);
// Load back the required element.
StackPtr = GetVectorElementPointer(StackPtr, EltVT, Idx);
return DAG.getExtLoad(ISD::EXTLOAD, dl, N->getValueType(0), Store, StackPtr,
- MachinePointerInfo(), EltVT, false, false, 0);
+ MachinePointerInfo(), EltVT, false, false, false, 0);
+}
+
+SDValue DAGTypeLegalizer::SplitVecOp_MSTORE(MaskedStoreSDNode *N,
+ unsigned OpNo) {
+ SDValue Ch = N->getChain();
+ SDValue Ptr = N->getBasePtr();
+ SDValue Mask = N->getMask();
+ SDValue Data = N->getValue();
+ EVT MemoryVT = N->getMemoryVT();
+ unsigned Alignment = N->getOriginalAlignment();
+ SDLoc DL(N);
+
+ EVT LoMemVT, HiMemVT;
+ std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
+
+ SDValue DataLo, DataHi;
+ GetSplitVector(Data, DataLo, DataHi);
+ SDValue MaskLo, MaskHi;
+ GetSplitVector(Mask, MaskLo, MaskHi);
+
+ // if Alignment is equal to the vector size,
+ // take the half of it for the second part
+ unsigned SecondHalfAlignment =
+ (Alignment == Data->getValueType(0).getSizeInBits()/8) ?
+ Alignment/2 : Alignment;
+
+ SDValue Lo, Hi;
+ MachineMemOperand *MMO = DAG.getMachineFunction().
+ getMachineMemOperand(N->getPointerInfo(),
+ MachineMemOperand::MOStore, LoMemVT.getStoreSize(),
+ Alignment, N->getAAInfo(), N->getRanges());
+
+ Lo = DAG.getMaskedStore(Ch, DL, DataLo, Ptr, MaskLo, LoMemVT, MMO,
+ N->isTruncatingStore());
+
+ unsigned IncrementSize = LoMemVT.getSizeInBits()/8;
+ Ptr = DAG.getNode(ISD::ADD, DL, Ptr.getValueType(), Ptr,
+ DAG.getConstant(IncrementSize, Ptr.getValueType()));
+
+ MMO = DAG.getMachineFunction().
+ getMachineMemOperand(N->getPointerInfo(),
+ MachineMemOperand::MOStore, HiMemVT.getStoreSize(),
+ SecondHalfAlignment, N->getAAInfo(), N->getRanges());
+
+ Hi = DAG.getMaskedStore(Ch, DL, DataHi, Ptr, MaskHi, HiMemVT, MMO,
+ N->isTruncatingStore());
+
+
+ // Build a factor node to remember that this store is independent of the
+ // other one.
+ return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
+
}
SDValue DAGTypeLegalizer::SplitVecOp_STORE(StoreSDNode *N, unsigned OpNo) {
unsigned Alignment = N->getOriginalAlignment();
bool isVol = N->isVolatile();
bool isNT = N->isNonTemporal();
- const MDNode *TBAAInfo = N->getTBAAInfo();
+ AAMDNodes AAInfo = N->getAAInfo();
SDValue Lo, Hi;
GetSplitVector(N->getOperand(1), Lo, Hi);
if (isTruncating)
Lo = DAG.getTruncStore(Ch, DL, Lo, Ptr, N->getPointerInfo(),
- LoMemVT, isVol, isNT, Alignment, TBAAInfo);
+ LoMemVT, isVol, isNT, Alignment, AAInfo);
else
Lo = DAG.getStore(Ch, DL, Lo, Ptr, N->getPointerInfo(),
- isVol, isNT, Alignment, TBAAInfo);
+ isVol, isNT, Alignment, AAInfo);
// Increment the pointer to the other half.
Ptr = DAG.getNode(ISD::ADD, DL, Ptr.getValueType(), Ptr,
if (isTruncating)
Hi = DAG.getTruncStore(Ch, DL, Hi, Ptr,
N->getPointerInfo().getWithOffset(IncrementSize),
- HiMemVT, isVol, isNT, Alignment, TBAAInfo);
+ HiMemVT, isVol, isNT, Alignment, AAInfo);
else
Hi = DAG.getStore(Ch, DL, Hi, Ptr,
N->getPointerInfo().getWithOffset(IncrementSize),
- isVol, isNT, Alignment, TBAAInfo);
+ isVol, isNT, Alignment, AAInfo);
return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
}
}
}
- return DAG.getNode(ISD::BUILD_VECTOR, DL, N->getValueType(0),
- &Elts[0], Elts.size());
+ return DAG.getNode(ISD::BUILD_VECTOR, DL, N->getValueType(0), Elts);
}
SDValue DAGTypeLegalizer::SplitVecOp_TRUNCATE(SDNode *N) {
case ISD::VECTOR_SHUFFLE:
Res = WidenVecRes_VECTOR_SHUFFLE(cast<ShuffleVectorSDNode>(N));
break;
+ case ISD::MLOAD:
+ Res = WidenVecRes_MLOAD(cast<MaskedLoadSDNode>(N));
+ break;
case ISD::ADD:
case ISD::AND:
- case ISD::BSWAP:
case ISD::MUL:
case ISD::MULHS:
case ISD::MULHU:
case ISD::OR:
case ISD::SUB:
case ISD::XOR:
+ case ISD::FMINNUM:
+ case ISD::FMAXNUM:
Res = WidenVecRes_Binary(N);
break;
Res = WidenVecRes_Convert(N);
break;
+ case ISD::BSWAP:
case ISD::CTLZ:
case ISD::CTPOP:
case ISD::CTTZ:
while (SubConcatEnd < OpsToConcat)
SubConcatOps[SubConcatEnd++] = undefVec;
ConcatOps[SubConcatIdx] = DAG.getNode(ISD::CONCAT_VECTORS, dl,
- NextVT, &SubConcatOps[0],
- OpsToConcat);
+ NextVT, SubConcatOps);
ConcatEnd = SubConcatIdx + 1;
}
}
for (unsigned j = ConcatEnd; j < NumOps; ++j)
ConcatOps[j] = UndefVal;
}
- return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, &ConcatOps[0], NumOps);
+ return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT,
+ makeArrayRef(ConcatOps.data(), NumOps));
}
SDValue DAGTypeLegalizer::WidenVecRes_Convert(SDNode *N) {
SDValue UndefVal = DAG.getUNDEF(InVT);
for (unsigned i = 1; i != NumConcat; ++i)
Ops[i] = UndefVal;
- SDValue InVec = DAG.getNode(ISD::CONCAT_VECTORS, DL, InWidenVT,
- &Ops[0], NumConcat);
+ SDValue InVec = DAG.getNode(ISD::CONCAT_VECTORS, DL, InWidenVT, Ops);
if (N->getNumOperands() == 1)
return DAG.getNode(Opcode, DL, WidenVT, InVec);
return DAG.getNode(Opcode, DL, WidenVT, InVec, N->getOperand(1));
for (; i < WidenNumElts; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, DL, WidenVT, &Ops[0], WidenNumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, DL, WidenVT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecRes_POWI(SDNode *N) {
SDValue NewVec;
if (InVT.isVector())
- NewVec = DAG.getNode(ISD::CONCAT_VECTORS, dl,
- NewInVT, &Ops[0], NewNumElts);
+ NewVec = DAG.getNode(ISD::CONCAT_VECTORS, dl, NewInVT, Ops);
else
- NewVec = DAG.getNode(ISD::BUILD_VECTOR, dl,
- NewInVT, &Ops[0], NewNumElts);
+ NewVec = DAG.getNode(ISD::BUILD_VECTOR, dl, NewInVT, Ops);
return DAG.getNode(ISD::BITCAST, dl, WidenVT, NewVec);
}
}
assert(WidenNumElts >= NumElts && "Shrinking vector instead of widening!");
NewOps.append(WidenNumElts - NumElts, DAG.getUNDEF(EltVT));
- return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, &NewOps[0], NewOps.size());
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, NewOps);
}
SDValue DAGTypeLegalizer::WidenVecRes_CONCAT_VECTORS(SDNode *N) {
Ops[i] = N->getOperand(i);
for (unsigned i = NumOperands; i != NumConcat; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, &Ops[0], NumConcat);
+ return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, Ops);
}
} else {
InputWidened = true;
SDValue UndefVal = DAG.getUNDEF(EltVT);
for (; Idx < WidenNumElts; ++Idx)
Ops[Idx] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, &Ops[0], WidenNumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecRes_CONVERT_RNDSAT(SDNode *N) {
for (unsigned i = 1; i != NumConcat; ++i)
Ops[i] = UndefVal;
- InOp = DAG.getNode(ISD::CONCAT_VECTORS, dl, InWidenVT, &Ops[0],NumConcat);
+ InOp = DAG.getNode(ISD::CONCAT_VECTORS, dl, InWidenVT, Ops);
return DAG.getConvertRndSat(WidenVT, dl, InOp, DTyOp, STyOp, RndOp,
SatOp, CvtCode);
}
for (; i < WidenNumElts; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, &Ops[0], WidenNumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecRes_EXTRACT_SUBVECTOR(SDNode *N) {
SDValue UndefVal = DAG.getUNDEF(EltVT);
for (; i < WidenNumElts; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, &Ops[0], WidenNumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecRes_INSERT_VECTOR_ELT(SDNode *N) {
if (LdChain.size() == 1)
NewChain = LdChain[0];
else
- NewChain = DAG.getNode(ISD::TokenFactor, SDLoc(LD), MVT::Other,
- &LdChain[0], LdChain.size());
+ NewChain = DAG.getNode(ISD::TokenFactor, SDLoc(LD), MVT::Other, LdChain);
// Modified the chain - switch anything that used the old chain to use
// the new one.
return Result;
}
+SDValue DAGTypeLegalizer::WidenVecRes_MLOAD(MaskedLoadSDNode *N) {
+
+ EVT WidenVT = TLI.getTypeToTransformTo(*DAG.getContext(),N->getValueType(0));
+ SDValue Mask = N->getMask();
+ EVT MaskVT = Mask.getValueType();
+ SDValue Src0 = GetWidenedVector(N->getSrc0());
+ ISD::LoadExtType ExtType = N->getExtensionType();
+ SDLoc dl(N);
+
+ if (getTypeAction(MaskVT) == TargetLowering::TypeWidenVector)
+ Mask = GetWidenedVector(Mask);
+ else {
+ EVT BoolVT = getSetCCResultType(WidenVT);
+
+ // We can't use ModifyToType() because we should fill the mask with
+ // zeroes
+ unsigned WidenNumElts = BoolVT.getVectorNumElements();
+ unsigned MaskNumElts = MaskVT.getVectorNumElements();
+
+ unsigned NumConcat = WidenNumElts / MaskNumElts;
+ SmallVector<SDValue, 16> Ops(NumConcat);
+ SDValue ZeroVal = DAG.getConstant(0, MaskVT);
+ Ops[0] = Mask;
+ for (unsigned i = 1; i != NumConcat; ++i)
+ Ops[i] = ZeroVal;
+
+ Mask = DAG.getNode(ISD::CONCAT_VECTORS, dl, BoolVT, Ops);
+ }
+
+ SDValue Res = DAG.getMaskedLoad(WidenVT, dl, N->getChain(), N->getBasePtr(),
+ Mask, Src0, N->getMemoryVT(),
+ N->getMemOperand(), ExtType);
+ // Legalized the chain result - switch anything that used the old chain to
+ // use the new one.
+ ReplaceValueWith(SDValue(N, 1), Res.getValue(1));
+ return Res;
+}
+
SDValue DAGTypeLegalizer::WidenVecRes_SCALAR_TO_VECTOR(SDNode *N) {
EVT WidenVT = TLI.getTypeToTransformTo(*DAG.getContext(), N->getValueType(0));
return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N),
case ISD::EXTRACT_SUBVECTOR: Res = WidenVecOp_EXTRACT_SUBVECTOR(N); break;
case ISD::EXTRACT_VECTOR_ELT: Res = WidenVecOp_EXTRACT_VECTOR_ELT(N); break;
case ISD::STORE: Res = WidenVecOp_STORE(N); break;
+ case ISD::MSTORE: Res = WidenVecOp_MSTORE(N, OpNo); break;
case ISD::SETCC: Res = WidenVecOp_SETCC(N); break;
+ case ISD::ANY_EXTEND:
+ case ISD::SIGN_EXTEND:
+ case ISD::ZERO_EXTEND:
+ Res = WidenVecOp_EXTEND(N);
+ break;
+
case ISD::FP_EXTEND:
case ISD::FP_TO_SINT:
case ISD::FP_TO_UINT:
case ISD::SINT_TO_FP:
case ISD::UINT_TO_FP:
case ISD::TRUNCATE:
- case ISD::SIGN_EXTEND:
- case ISD::ZERO_EXTEND:
- case ISD::ANY_EXTEND:
Res = WidenVecOp_Convert(N);
break;
}
return false;
}
+SDValue DAGTypeLegalizer::WidenVecOp_EXTEND(SDNode *N) {
+ SDLoc DL(N);
+ EVT VT = N->getValueType(0);
+
+ SDValue InOp = N->getOperand(0);
+ // If some legalization strategy other than widening is used on the operand,
+ // we can't safely assume that just extending the low lanes is the correct
+ // transformation.
+ if (getTypeAction(InOp.getValueType()) != TargetLowering::TypeWidenVector)
+ return WidenVecOp_Convert(N);
+ InOp = GetWidenedVector(InOp);
+ assert(VT.getVectorNumElements() <
+ InOp.getValueType().getVectorNumElements() &&
+ "Input wasn't widened!");
+
+ // We may need to further widen the operand until it has the same total
+ // vector size as the result.
+ EVT InVT = InOp.getValueType();
+ if (InVT.getSizeInBits() != VT.getSizeInBits()) {
+ EVT InEltVT = InVT.getVectorElementType();
+ for (int i = MVT::FIRST_VECTOR_VALUETYPE, e = MVT::LAST_VECTOR_VALUETYPE; i < e; ++i) {
+ EVT FixedVT = (MVT::SimpleValueType)i;
+ EVT FixedEltVT = FixedVT.getVectorElementType();
+ if (TLI.isTypeLegal(FixedVT) &&
+ FixedVT.getSizeInBits() == VT.getSizeInBits() &&
+ FixedEltVT == InEltVT) {
+ assert(FixedVT.getVectorNumElements() >= VT.getVectorNumElements() &&
+ "Not enough elements in the fixed type for the operand!");
+ assert(FixedVT.getVectorNumElements() != InVT.getVectorNumElements() &&
+ "We can't have the same type as we started with!");
+ if (FixedVT.getVectorNumElements() > InVT.getVectorNumElements())
+ InOp = DAG.getNode(ISD::INSERT_SUBVECTOR, DL, FixedVT,
+ DAG.getUNDEF(FixedVT), InOp,
+ DAG.getConstant(0, TLI.getVectorIdxTy()));
+ else
+ InOp = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, FixedVT, InOp,
+ DAG.getConstant(0, TLI.getVectorIdxTy()));
+ break;
+ }
+ }
+ InVT = InOp.getValueType();
+ if (InVT.getSizeInBits() != VT.getSizeInBits())
+ // We couldn't find a legal vector type that was a widening of the input
+ // and could be extended in-register to the result type, so we have to
+ // scalarize.
+ return WidenVecOp_Convert(N);
+ }
+
+ // Use special DAG nodes to represent the operation of extending the
+ // low lanes.
+ switch (N->getOpcode()) {
+ default:
+ llvm_unreachable("Extend legalization on on extend operation!");
+ case ISD::ANY_EXTEND:
+ return DAG.getAnyExtendVectorInReg(InOp, DL, VT);
+ case ISD::SIGN_EXTEND:
+ return DAG.getSignExtendVectorInReg(InOp, DL, VT);
+ case ISD::ZERO_EXTEND:
+ return DAG.getZeroExtendVectorInReg(InOp, DL, VT);
+ }
+}
+
SDValue DAGTypeLegalizer::WidenVecOp_Convert(SDNode *N) {
// Since the result is legal and the input is illegal, it is unlikely
// that we can fix the input to a legal type so unroll the convert
DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, InEltVT, InOp,
DAG.getConstant(i, TLI.getVectorIdxTy())));
- return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, &Ops[0], NumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecOp_BITCAST(SDNode *N) {
Ops[Idx++] = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, EltVT, InOp,
DAG.getConstant(j, TLI.getVectorIdxTy()));
}
- return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, &Ops[0], NumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, Ops);
}
SDValue DAGTypeLegalizer::WidenVecOp_EXTRACT_SUBVECTOR(SDNode *N) {
if (StChain.size() == 1)
return StChain[0];
else
- return DAG.getNode(ISD::TokenFactor, SDLoc(ST),
- MVT::Other,&StChain[0],StChain.size());
+ return DAG.getNode(ISD::TokenFactor, SDLoc(ST), MVT::Other, StChain);
+}
+
+SDValue DAGTypeLegalizer::WidenVecOp_MSTORE(SDNode *N, unsigned OpNo) {
+ MaskedStoreSDNode *MST = cast<MaskedStoreSDNode>(N);
+ SDValue Mask = MST->getMask();
+ EVT MaskVT = Mask.getValueType();
+ SDValue StVal = MST->getValue();
+ // Widen the value
+ SDValue WideVal = GetWidenedVector(StVal);
+ SDLoc dl(N);
+
+ if (OpNo == 2 || getTypeAction(MaskVT) == TargetLowering::TypeWidenVector)
+ Mask = GetWidenedVector(Mask);
+ else {
+ // The mask should be widened as well
+ EVT BoolVT = getSetCCResultType(WideVal.getValueType());
+ // We can't use ModifyToType() because we should fill the mask with
+ // zeroes
+ unsigned WidenNumElts = BoolVT.getVectorNumElements();
+ unsigned MaskNumElts = MaskVT.getVectorNumElements();
+
+ unsigned NumConcat = WidenNumElts / MaskNumElts;
+ SmallVector<SDValue, 16> Ops(NumConcat);
+ SDValue ZeroVal = DAG.getConstant(0, MaskVT);
+ Ops[0] = Mask;
+ for (unsigned i = 1; i != NumConcat; ++i)
+ Ops[i] = ZeroVal;
+
+ Mask = DAG.getNode(ISD::CONCAT_VECTORS, dl, BoolVT, Ops);
+ }
+ assert(Mask.getValueType().getVectorNumElements() ==
+ WideVal.getValueType().getVectorNumElements() &&
+ "Mask and data vectors should have the same number of elements");
+ return DAG.getMaskedStore(MST->getChain(), dl, WideVal, MST->getBasePtr(),
+ Mask, MST->getMemoryVT(), MST->getMemOperand(),
+ false);
}
SDValue DAGTypeLegalizer::WidenVecOp_SETCC(SDNode *N) {
bool isVolatile = LD->isVolatile();
bool isNonTemporal = LD->isNonTemporal();
bool isInvariant = LD->isInvariant();
- const MDNode *TBAAInfo = LD->getTBAAInfo();
+ AAMDNodes AAInfo = LD->getAAInfo();
int LdWidth = LdVT.getSizeInBits();
int WidthDiff = WidenWidth - LdWidth; // Difference
int NewVTWidth = NewVT.getSizeInBits();
SDValue LdOp = DAG.getLoad(NewVT, dl, Chain, BasePtr, LD->getPointerInfo(),
isVolatile, isNonTemporal, isInvariant, Align,
- TBAAInfo);
+ AAInfo);
LdChain.push_back(LdOp.getValue(1));
// Check if we can load the element with one instruction
ConcatOps[0] = LdOp;
for (unsigned i = 1; i != NumConcat; ++i)
ConcatOps[i] = UndefVal;
- return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, &ConcatOps[0],
- NumConcat);
+ return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, ConcatOps);
}
// Load vector by using multiple loads from largest vector to scalar
L = DAG.getLoad(NewVT, dl, Chain, BasePtr,
LD->getPointerInfo().getWithOffset(Offset), isVolatile,
isNonTemporal, isInvariant, MinAlign(Align, Increment),
- TBAAInfo);
+ AAInfo);
LdChain.push_back(L.getValue(1));
if (L->getValueType(0).isVector()) {
SmallVector<SDValue, 16> Loads;
Loads.push_back(DAG.getUNDEF(L->getValueType(0)));
size += L->getValueSizeInBits(0);
}
- L = DAG.getNode(ISD::CONCAT_VECTORS, dl, LdOp->getValueType(0),
- &Loads[0], Loads.size());
+ L = DAG.getNode(ISD::CONCAT_VECTORS, dl, LdOp->getValueType(0), Loads);
}
} else {
L = DAG.getLoad(NewVT, dl, Chain, BasePtr,
LD->getPointerInfo().getWithOffset(Offset), isVolatile,
isNonTemporal, isInvariant, MinAlign(Align, Increment),
- TBAAInfo);
+ AAInfo);
LdChain.push_back(L.getValue(1));
}
if (NewLdTy != LdTy) {
// Create a larger vector
ConcatOps[End-1] = DAG.getNode(ISD::CONCAT_VECTORS, dl, NewLdTy,
- &ConcatOps[Idx], End - Idx);
+ makeArrayRef(&ConcatOps[Idx], End - Idx));
Idx = End - 1;
LdTy = NewLdTy;
}
if (WidenWidth == LdTy.getSizeInBits()*(End - Idx))
return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT,
- &ConcatOps[Idx], End - Idx);
+ makeArrayRef(&ConcatOps[Idx], End - Idx));
// We need to fill the rest with undefs to build the vector
unsigned NumOps = WidenWidth / LdTy.getSizeInBits();
for (; i != NumOps; ++i)
WidenOps[i] = UndefVal;
}
- return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, &WidenOps[0],NumOps);
+ return DAG.getNode(ISD::CONCAT_VECTORS, dl, WidenVT, WidenOps);
}
SDValue
unsigned Align = LD->getAlignment();
bool isVolatile = LD->isVolatile();
bool isNonTemporal = LD->isNonTemporal();
- const MDNode *TBAAInfo = LD->getTBAAInfo();
+ bool isInvariant = LD->isInvariant();
+ AAMDNodes AAInfo = LD->getAAInfo();
EVT EltVT = WidenVT.getVectorElementType();
EVT LdEltVT = LdVT.getVectorElementType();
unsigned Increment = LdEltVT.getSizeInBits() / 8;
Ops[0] = DAG.getExtLoad(ExtType, dl, EltVT, Chain, BasePtr,
LD->getPointerInfo(),
- LdEltVT, isVolatile, isNonTemporal, Align, TBAAInfo);
+ LdEltVT, isVolatile, isNonTemporal, isInvariant,
+ Align, AAInfo);
LdChain.push_back(Ops[0].getValue(1));
unsigned i = 0, Offset = Increment;
for (i=1; i < NumElts; ++i, Offset += Increment) {
BasePtr.getValueType()));
Ops[i] = DAG.getExtLoad(ExtType, dl, EltVT, Chain, NewBasePtr,
LD->getPointerInfo().getWithOffset(Offset), LdEltVT,
- isVolatile, isNonTemporal, Align, TBAAInfo);
+ isVolatile, isNonTemporal, isInvariant, Align,
+ AAInfo);
LdChain.push_back(Ops[i].getValue(1));
}
for (; i != WidenNumElts; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, &Ops[0], Ops.size());
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, WidenVT, Ops);
}
unsigned Align = ST->getAlignment();
bool isVolatile = ST->isVolatile();
bool isNonTemporal = ST->isNonTemporal();
- const MDNode *TBAAInfo = ST->getTBAAInfo();
+ AAMDNodes AAInfo = ST->getAAInfo();
SDValue ValOp = GetWidenedVector(ST->getValue());
SDLoc dl(ST);
StChain.push_back(DAG.getStore(Chain, dl, EOp, BasePtr,
ST->getPointerInfo().getWithOffset(Offset),
isVolatile, isNonTemporal,
- MinAlign(Align, Offset), TBAAInfo));
+ MinAlign(Align, Offset), AAInfo));
StWidth -= NewVTWidth;
Offset += Increment;
Idx += NumVTElts;
StChain.push_back(DAG.getStore(Chain, dl, EOp, BasePtr,
ST->getPointerInfo().getWithOffset(Offset),
isVolatile, isNonTemporal,
- MinAlign(Align, Offset), TBAAInfo));
+ MinAlign(Align, Offset), AAInfo));
StWidth -= NewVTWidth;
Offset += Increment;
BasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(), BasePtr,
unsigned Align = ST->getAlignment();
bool isVolatile = ST->isVolatile();
bool isNonTemporal = ST->isNonTemporal();
- const MDNode *TBAAInfo = ST->getTBAAInfo();
+ AAMDNodes AAInfo = ST->getAAInfo();
SDValue ValOp = GetWidenedVector(ST->getValue());
SDLoc dl(ST);
StChain.push_back(DAG.getTruncStore(Chain, dl, EOp, BasePtr,
ST->getPointerInfo(), StEltVT,
isVolatile, isNonTemporal, Align,
- TBAAInfo));
+ AAInfo));
unsigned Offset = Increment;
for (unsigned i=1; i < NumElts; ++i, Offset += Increment) {
SDValue NewBasePtr = DAG.getNode(ISD::ADD, dl, BasePtr.getValueType(),
StChain.push_back(DAG.getTruncStore(Chain, dl, EOp, NewBasePtr,
ST->getPointerInfo().getWithOffset(Offset),
StEltVT, isVolatile, isNonTemporal,
- MinAlign(Align, Offset), TBAAInfo));
+ MinAlign(Align, Offset), AAInfo));
}
}
for (unsigned i = 1; i != NumConcat; ++i)
Ops[i] = UndefVal;
- return DAG.getNode(ISD::CONCAT_VECTORS, dl, NVT, &Ops[0], NumConcat);
+ return DAG.getNode(ISD::CONCAT_VECTORS, dl, NVT, Ops);
}
if (WidenNumElts < InNumElts && InNumElts % WidenNumElts)
SDValue UndefVal = DAG.getUNDEF(EltVT);
for ( ; Idx < WidenNumElts; ++Idx)
Ops[Idx] = UndefVal;
- return DAG.getNode(ISD::BUILD_VECTOR, dl, NVT, &Ops[0], WidenNumElts);
+ return DAG.getNode(ISD::BUILD_VECTOR, dl, NVT, Ops);
}