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;
case TargetLowering::TypeSoftenFloat:
// Promote the integer operand by hand.
return DAG.getNode(ISD::ANY_EXTEND, dl, NOutVT, GetSoftenedFloat(InOp));
+ case TargetLowering::TypePromoteFloat: {
+ // Convert the promoted float by hand.
+ if (NOutVT.bitsEq(NInVT)) {
+ SDValue PromotedOp = GetPromotedFloat(InOp);
+ SDValue Trunc = DAG.getNode(ISD::FP_TO_FP16, dl, NOutVT, PromotedOp);
+ return DAG.getNode(ISD::AssertZext, dl, NOutVT, Trunc,
+ DAG.getValueType(OutVT));
+ }
+ break;
+ }
case TargetLowering::TypeExpandInteger:
case TargetLowering::TypeExpandFloat:
break;
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.
SDValue DAGTypeLegalizer::PromoteIntOp_MSTORE(MaskedStoreSDNode *N, unsigned OpNo){
- assert(OpNo == 2 && "Only know how to promote the mask!");
- EVT DataVT = N->getOperand(3).getValueType();
- 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 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){
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);
}
/// and the shift amount is a constant 'Amt'. Expand the operation.
void DAGTypeLegalizer::ExpandShiftByConstant(SDNode *N, unsigned Amt,
SDValue &Lo, SDValue &Hi) {
- assert(Amt && "Expected zero shifts to be already optimized away.");
SDLoc DL(N);
// Expand the incoming operand to be shifted, so that we have its parts
SDValue InL, InH;
GetExpandedInteger(N->getOperand(0), InL, InH);
+ // Though Amt shouldn't usually be 0, it's possible. E.g. when legalization
+ // splitted a vector shift, like this: <op1, op2> SHL <0, 2>.
+ if (!Amt) {
+ Lo = InL;
+ Hi = InH;
+ return;
+ }
+
EVT NVT = InL.getValueType();
unsigned VTBits = N->getValueType(0).getSizeInBits();
unsigned NVTBits = NVT.getSizeInBits();
SDValue AmtLack = DAG.getNode(ISD::SUB, dl, ShTy, NVBitsNode, Amt);
SDValue isShort = DAG.getSetCC(dl, getSetCCResultType(ShTy),
Amt, NVBitsNode, ISD::SETULT);
+ SDValue isZero = DAG.getSetCC(dl, getSetCCResultType(ShTy),
+ Amt, DAG.getConstant(0, ShTy),
+ ISD::SETEQ);
SDValue LoS, HiS, LoL, HiL;
switch (N->getOpcode()) {
LoS = DAG.getNode(ISD::SHL, dl, NVT, InL, Amt);
HiS = DAG.getNode(ISD::OR, dl, NVT,
DAG.getNode(ISD::SHL, dl, NVT, InH, Amt),
- // FIXME: If Amt is zero, the following shift generates an undefined result
- // on some architectures.
DAG.getNode(ISD::SRL, dl, NVT, InL, AmtLack));
// Long: ShAmt >= NVTBits
HiL = DAG.getNode(ISD::SHL, dl, NVT, InL, AmtExcess); // Hi from Lo part.
Lo = DAG.getSelect(dl, NVT, isShort, LoS, LoL);
- Hi = DAG.getSelect(dl, NVT, isShort, HiS, HiL);
+ Hi = DAG.getSelect(dl, NVT, isZero, InH,
+ DAG.getSelect(dl, NVT, isShort, HiS, HiL));
return true;
case ISD::SRL:
// Short: ShAmt < NVTBits
HiL = DAG.getConstant(0, NVT); // Hi part is zero.
LoL = DAG.getNode(ISD::SRL, dl, NVT, InH, AmtExcess); // Lo from Hi part.
- Lo = DAG.getSelect(dl, NVT, isShort, LoS, LoL);
+ Lo = DAG.getSelect(dl, NVT, isZero, InL,
+ DAG.getSelect(dl, NVT, isShort, LoS, LoL));
Hi = DAG.getSelect(dl, NVT, isShort, HiS, HiL);
return true;
case ISD::SRA:
HiS = DAG.getNode(ISD::SRA, dl, NVT, InH, Amt);
LoS = DAG.getNode(ISD::OR, dl, NVT,
DAG.getNode(ISD::SRL, dl, NVT, InL, Amt),
- // FIXME: If Amt is zero, the following shift generates an undefined result
- // on some architectures.
DAG.getNode(ISD::SHL, dl, NVT, InH, AmtLack));
// Long: ShAmt >= NVTBits
DAG.getConstant(NVTBits-1, ShTy));
LoL = DAG.getNode(ISD::SRA, dl, NVT, InH, AmtExcess); // Lo from Hi part.
- Lo = DAG.getSelect(dl, NVT, isShort, LoS, LoL);
+ Lo = DAG.getSelect(dl, NVT, isZero, InL,
+ DAG.getSelect(dl, NVT, isShort, LoS, LoL));
Hi = DAG.getSelect(dl, NVT, isShort, HiS, HiL);
return true;
}
SDValue &Hi) {
SDLoc dl(N);
EVT VT = N->getValueType(0);
+
SDValue Op = N->getOperand(0);
+ if (getTypeAction(Op.getValueType()) == TargetLowering::TypePromoteFloat)
+ Op = GetPromotedFloat(Op);
+
RTLIB::Libcall LC = RTLIB::getFPTOSINT(Op.getValueType(), VT);
assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unexpected fp-to-sint conversion!");
SplitInteger(TLI.makeLibCall(DAG, LC, VT, &Op, 1, true/*irrelevant*/,
SDValue &Hi) {
SDLoc dl(N);
EVT VT = N->getValueType(0);
+
SDValue Op = N->getOperand(0);
+ if (getTypeAction(Op.getValueType()) == TargetLowering::TypePromoteFloat)
+ Op = GetPromotedFloat(Op);
+
RTLIB::Libcall LC = RTLIB::getFPTOUINT(Op.getValueType(), VT);
assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unexpected fp-to-uint conversion!");
SplitInteger(TLI.makeLibCall(DAG, LC, VT, &Op, 1, false/*irrelevant*/,
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