case ISD::EXTRACT_VECTOR_ELT:
Res = PromoteIntRes_EXTRACT_VECTOR_ELT(N); break;
case ISD::LOAD: Res = PromoteIntRes_LOAD(cast<LoadSDNode>(N));break;
+ case ISD::MLOAD: Res = PromoteIntRes_MLOAD(cast<MaskedLoadSDNode>(N));break;
case ISD::SELECT: Res = PromoteIntRes_SELECT(N); break;
case ISD::VSELECT: Res = PromoteIntRes_VSELECT(N); break;
case ISD::SELECT_CC: Res = PromoteIntRes_SELECT_CC(N); break;
EVT NVT = Op.getValueType();
Op = DAG.getNode(N->getOpcode(), dl, NVT, Op);
// Subtract off the extra leading bits in the bigger type.
- return DAG.getNode(ISD::SUB, dl, NVT, Op,
- DAG.getConstant(NVT.getSizeInBits() -
- OVT.getSizeInBits(), NVT));
+ return DAG.getNode(
+ ISD::SUB, dl, NVT, Op,
+ DAG.getConstant(NVT.getScalarSizeInBits() - OVT.getScalarSizeInBits(),
+ NVT));
}
SDValue DAGTypeLegalizer::PromoteIntRes_CTPOP(SDNode *N) {
// The count is the same in the promoted type except if the original
// value was zero. This can be handled by setting the bit just off
// the top of the original type.
- APInt TopBit(NVT.getSizeInBits(), 0);
- TopBit.setBit(OVT.getSizeInBits());
+ auto TopBit = APInt::getOneBitSet(NVT.getScalarSizeInBits(),
+ OVT.getScalarSizeInBits());
Op = DAG.getNode(ISD::OR, dl, NVT, Op, DAG.getConstant(TopBit, NVT));
}
return DAG.getNode(N->getOpcode(), dl, NVT, Op);
return Res;
}
+SDValue DAGTypeLegalizer::PromoteIntRes_MLOAD(MaskedLoadSDNode *N) {
+ EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), N->getValueType(0));
+ SDValue ExtSrc0 = GetPromotedInteger(N->getSrc0());
+
+ SDValue Mask = N->getMask();
+ EVT NewMaskVT = getSetCCResultType(NVT);
+ if (NewMaskVT != N->getMask().getValueType())
+ Mask = PromoteTargetBoolean(Mask, NewMaskVT);
+ SDLoc dl(N);
+
+ SDValue Res = DAG.getMaskedLoad(NVT, dl, N->getChain(), N->getBasePtr(),
+ Mask, ExtSrc0, N->getMemoryVT(),
+ N->getMemOperand(), ISD::SEXTLOAD);
+ // 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;
+}
/// Promote the overflow flag of an overflowing arithmetic node.
SDValue DAGTypeLegalizer::PromoteIntRes_Overflow(SDNode *N) {
// Simply change the return type of the boolean result.
case ISD::SINT_TO_FP: Res = PromoteIntOp_SINT_TO_FP(N); break;
case ISD::STORE: Res = PromoteIntOp_STORE(cast<StoreSDNode>(N),
OpNo); break;
+ case ISD::MSTORE: Res = PromoteIntOp_MSTORE(cast<MaskedStoreSDNode>(N),
+ OpNo); break;
+ case ISD::MLOAD: Res = PromoteIntOp_MLOAD(cast<MaskedLoadSDNode>(N),
+ OpNo); break;
case ISD::TRUNCATE: Res = PromoteIntOp_TRUNCATE(N); break;
case ISD::FP16_TO_FP:
case ISD::UINT_TO_FP: Res = PromoteIntOp_UINT_TO_FP(N); break;
switch (CCCode) {
default: llvm_unreachable("Unknown integer comparison!");
case ISD::SETEQ:
- case ISD::SETNE:
+ case ISD::SETNE: {
+ SDValue OpL = GetPromotedInteger(NewLHS);
+ SDValue OpR = GetPromotedInteger(NewRHS);
+
+ // We would prefer to promote the comparison operand with sign extension,
+ // if we find the operand is actually to truncate an AssertSext. With this
+ // optimization, we can avoid inserting real truncate instruction, which
+ // is redudant eventually.
+ if (OpL->getOpcode() == ISD::AssertSext &&
+ cast<VTSDNode>(OpL->getOperand(1))->getVT() == NewLHS.getValueType() &&
+ OpR->getOpcode() == ISD::AssertSext &&
+ cast<VTSDNode>(OpR->getOperand(1))->getVT() == NewRHS.getValueType()) {
+ NewLHS = OpL;
+ NewRHS = OpR;
+ } else {
+ NewLHS = ZExtPromotedInteger(NewLHS);
+ NewRHS = ZExtPromotedInteger(NewRHS);
+ }
+ break;
+ }
case ISD::SETUGE:
case ISD::SETUGT:
case ISD::SETULE:
EVT VecVT = N->getValueType(0);
unsigned NumElts = VecVT.getVectorNumElements();
assert(!((NumElts & 1) && (!TLI.isTypeLegal(VecVT))) &&
- "Legal vector of one illegal element?");
+ "Legal vector of one illegal element?");
// Promote the inserted value. The type does not need to match the
// vector element type. Check that any extra bits introduced will be
N->getMemoryVT(), N->getMemOperand());
}
+SDValue DAGTypeLegalizer::PromoteIntOp_MSTORE(MaskedStoreSDNode *N, unsigned OpNo){
+
+ SDValue DataOp = N->getValue();
+ EVT DataVT = DataOp.getValueType();
+ SDValue Mask = N->getMask();
+ EVT MaskVT = Mask.getValueType();
+ SDLoc dl(N);
+
+ bool TruncateStore = false;
+ if (!TLI.isTypeLegal(DataVT)) {
+ if (getTypeAction(DataVT) == TargetLowering::TypePromoteInteger) {
+ DataOp = GetPromotedInteger(DataOp);
+ if (!TLI.isTypeLegal(MaskVT))
+ Mask = PromoteTargetBoolean(Mask, DataOp.getValueType());
+ TruncateStore = true;
+ }
+ else {
+ assert(getTypeAction(DataVT) == TargetLowering::TypeWidenVector &&
+ "Unexpected data legalization in MSTORE");
+ DataOp = GetWidenedVector(DataOp);
+
+ if (getTypeAction(MaskVT) == TargetLowering::TypeWidenVector)
+ Mask = GetWidenedVector(Mask);
+ else {
+ EVT BoolVT = getSetCCResultType(DataOp.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);
+ }
+ }
+ }
+ else
+ Mask = PromoteTargetBoolean(N->getMask(), DataOp.getValueType());
+ return DAG.getMaskedStore(N->getChain(), dl, DataOp, N->getBasePtr(), Mask,
+ N->getMemoryVT(), N->getMemOperand(),
+ TruncateStore);
+}
+
+SDValue DAGTypeLegalizer::PromoteIntOp_MLOAD(MaskedLoadSDNode *N, unsigned OpNo){
+ assert(OpNo == 2 && "Only know how to promote the mask!");
+ EVT DataVT = N->getValueType(0);
+ SDValue Mask = PromoteTargetBoolean(N->getOperand(OpNo), DataVT);
+ SmallVector<SDValue, 4> NewOps(N->op_begin(), N->op_end());
+ NewOps[OpNo] = Mask;
+ return SDValue(DAG.UpdateNodeOperands(N, NewOps), 0);
+}
+
SDValue DAGTypeLegalizer::PromoteIntOp_TRUNCATE(SDNode *N) {
SDValue Op = GetPromotedInteger(N->getOperand(0));
return DAG.getNode(ISD::TRUNCATE, SDLoc(N), N->getValueType(0), Op);
std::pair <SDValue, SDValue> DAGTypeLegalizer::ExpandAtomic(SDNode *Node) {
unsigned Opc = Node->getOpcode();
MVT VT = cast<AtomicSDNode>(Node)->getMemoryVT().getSimpleVT();
- RTLIB::Libcall LC;
-
- switch (Opc) {
- default:
- llvm_unreachable("Unhandled atomic intrinsic Expand!");
- case ISD::ATOMIC_SWAP:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_LOCK_TEST_AND_SET_1; break;
- case MVT::i16: LC = RTLIB::SYNC_LOCK_TEST_AND_SET_2; break;
- case MVT::i32: LC = RTLIB::SYNC_LOCK_TEST_AND_SET_4; break;
- case MVT::i64: LC = RTLIB::SYNC_LOCK_TEST_AND_SET_8; break;
- case MVT::i128:LC = RTLIB::SYNC_LOCK_TEST_AND_SET_16;break;
- }
- break;
- case ISD::ATOMIC_CMP_SWAP:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_VAL_COMPARE_AND_SWAP_1; break;
- case MVT::i16: LC = RTLIB::SYNC_VAL_COMPARE_AND_SWAP_2; break;
- case MVT::i32: LC = RTLIB::SYNC_VAL_COMPARE_AND_SWAP_4; break;
- case MVT::i64: LC = RTLIB::SYNC_VAL_COMPARE_AND_SWAP_8; break;
- case MVT::i128:LC = RTLIB::SYNC_VAL_COMPARE_AND_SWAP_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_ADD:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_ADD_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_ADD_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_ADD_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_ADD_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_ADD_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_SUB:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_SUB_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_SUB_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_SUB_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_SUB_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_SUB_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_AND:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_AND_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_AND_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_AND_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_AND_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_AND_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_OR:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_OR_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_OR_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_OR_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_OR_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_OR_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_XOR:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_XOR_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_XOR_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_XOR_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_XOR_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_XOR_16;break;
- }
- break;
- case ISD::ATOMIC_LOAD_NAND:
- switch (VT.SimpleTy) {
- default: llvm_unreachable("Unexpected value type for atomic!");
- case MVT::i8: LC = RTLIB::SYNC_FETCH_AND_NAND_1; break;
- case MVT::i16: LC = RTLIB::SYNC_FETCH_AND_NAND_2; break;
- case MVT::i32: LC = RTLIB::SYNC_FETCH_AND_NAND_4; break;
- case MVT::i64: LC = RTLIB::SYNC_FETCH_AND_NAND_8; break;
- case MVT::i128:LC = RTLIB::SYNC_FETCH_AND_NAND_16;break;
- }
- break;
- }
+ RTLIB::Libcall LC = RTLIB::getATOMIC(Opc, VT);
+ assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unexpected atomic op or value type!");
return ExpandChainLibCall(LC, Node, false);
}
bool isVolatile = N->isVolatile();
bool isNonTemporal = N->isNonTemporal();
bool isInvariant = N->isInvariant();
- const MDNode *TBAAInfo = N->getTBAAInfo();
+ AAMDNodes AAInfo = N->getAAInfo();
SDLoc dl(N);
assert(NVT.isByteSized() && "Expanded type not byte sized!");
EVT MemVT = N->getMemoryVT();
Lo = DAG.getExtLoad(ExtType, dl, NVT, Ch, Ptr, N->getPointerInfo(),
- MemVT, isVolatile, isNonTemporal, Alignment, TBAAInfo);
+ MemVT, isVolatile, isNonTemporal, isInvariant,
+ Alignment, AAInfo);
// Remember the chain.
Ch = Lo.getValue(1);
// Little-endian - low bits are at low addresses.
Lo = DAG.getLoad(NVT, dl, Ch, Ptr, N->getPointerInfo(),
isVolatile, isNonTemporal, isInvariant, Alignment,
- TBAAInfo);
+ AAInfo);
unsigned ExcessBits =
N->getMemoryVT().getSizeInBits() - NVT.getSizeInBits();
DAG.getConstant(IncrementSize, Ptr.getValueType()));
Hi = DAG.getExtLoad(ExtType, dl, NVT, Ch, Ptr,
N->getPointerInfo().getWithOffset(IncrementSize), NEVT,
- isVolatile, isNonTemporal,
- MinAlign(Alignment, IncrementSize), TBAAInfo);
+ isVolatile, isNonTemporal, isInvariant,
+ MinAlign(Alignment, IncrementSize), AAInfo);
// Build a factor node to remember that this load is independent of the
// other one.
Hi = DAG.getExtLoad(ExtType, dl, NVT, Ch, Ptr, N->getPointerInfo(),
EVT::getIntegerVT(*DAG.getContext(),
MemVT.getSizeInBits() - ExcessBits),
- isVolatile, isNonTemporal, Alignment, TBAAInfo);
+ isVolatile, isNonTemporal, isInvariant, Alignment,
+ AAInfo);
// Increment the pointer to the other half.
Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr,
Lo = DAG.getExtLoad(ISD::ZEXTLOAD, dl, NVT, Ch, Ptr,
N->getPointerInfo().getWithOffset(IncrementSize),
EVT::getIntegerVT(*DAG.getContext(), ExcessBits),
- isVolatile, isNonTemporal,
- MinAlign(Alignment, IncrementSize), TBAAInfo);
+ isVolatile, isNonTemporal, isInvariant,
+ MinAlign(Alignment, IncrementSize), AAInfo);
// Build a factor node to remember that this load is independent of the
// other one.
unsigned Alignment = N->getAlignment();
bool isVolatile = N->isVolatile();
bool isNonTemporal = N->isNonTemporal();
- const MDNode *TBAAInfo = N->getTBAAInfo();
+ AAMDNodes AAInfo = N->getAAInfo();
SDLoc dl(N);
SDValue Lo, Hi;
GetExpandedInteger(N->getValue(), Lo, Hi);
return DAG.getTruncStore(Ch, dl, Lo, Ptr, N->getPointerInfo(),
N->getMemoryVT(), isVolatile, isNonTemporal,
- Alignment, TBAAInfo);
+ Alignment, AAInfo);
}
if (TLI.isLittleEndian()) {
GetExpandedInteger(N->getValue(), Lo, Hi);
Lo = DAG.getStore(Ch, dl, Lo, Ptr, N->getPointerInfo(),
- isVolatile, isNonTemporal, Alignment, TBAAInfo);
+ isVolatile, isNonTemporal, Alignment, AAInfo);
unsigned ExcessBits =
N->getMemoryVT().getSizeInBits() - NVT.getSizeInBits();
Hi = DAG.getTruncStore(Ch, dl, Hi, Ptr,
N->getPointerInfo().getWithOffset(IncrementSize),
NEVT, isVolatile, isNonTemporal,
- MinAlign(Alignment, IncrementSize), TBAAInfo);
+ MinAlign(Alignment, IncrementSize), AAInfo);
return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo, Hi);
}
// Store both the high bits and maybe some of the low bits.
Hi = DAG.getTruncStore(Ch, dl, Hi, Ptr, N->getPointerInfo(),
- HiVT, isVolatile, isNonTemporal, Alignment, TBAAInfo);
+ HiVT, isVolatile, isNonTemporal, Alignment, AAInfo);
// Increment the pointer to the other half.
Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr,
N->getPointerInfo().getWithOffset(IncrementSize),
EVT::getIntegerVT(*DAG.getContext(), ExcessBits),
isVolatile, isNonTemporal,
- MinAlign(Alignment, IncrementSize), TBAAInfo);
+ MinAlign(Alignment, IncrementSize), AAInfo);
return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo, Hi);
}
FudgePtr,
MachinePointerInfo::getConstantPool(),
MVT::f32,
- false, false, Alignment);
+ false, false, false, Alignment);
return DAG.getNode(ISD::FADD, dl, DstVT, SignedConv, Fudge);
}
EVT VT = N->getValueType(0);
SDLoc dl(N);
- unsigned NumElts = VT.getVectorNumElements();
- SmallVector<int, 8> NewMask;
- for (unsigned i = 0; i != NumElts; ++i) {
- NewMask.push_back(SV->getMaskElt(i));
- }
+ ArrayRef<int> NewMask = SV->getMask().slice(0, VT.getVectorNumElements());
SDValue V0 = GetPromotedInteger(N->getOperand(0));
SDValue V1 = GetPromotedInteger(N->getOperand(1));
EVT OutVT = V0.getValueType();
- return DAG.getVectorShuffle(OutVT, dl, V0, V1, &NewMask[0]);
+ return DAG.getVectorShuffle(OutVT, dl, V0, V1, NewMask);
}
projects / oota-llvm.git / blobdiff