| //===------- LegalizeVectorTypes.cpp - Legalization of vector types -------===// |
| // |
| // The LLVM Compiler Infrastructure |
| // |
| // This file is distributed under the University of Illinois Open Source |
| // License. See LICENSE.TXT for details. |
| // |
| //===----------------------------------------------------------------------===// |
| // |
| // This file performs vector type splitting and scalarization for LegalizeTypes. |
| // Scalarization is the act of changing a computation in an illegal one-element |
| // vector type to be a computation in its scalar element type. For example, |
| // implementing <1 x f32> arithmetic in a scalar f32 register. This is needed |
| // as a base case when scalarizing vector arithmetic like <4 x f32>, which |
| // eventually decomposes to scalars if the target doesn't support v4f32 or v2f32 |
| // types. |
| // Splitting is the act of changing a computation in an invalid vector type to |
| // be a computation in two vectors of half the size. For example, implementing |
| // <128 x f32> operations in terms of two <64 x f32> operations. |
| // |
| //===----------------------------------------------------------------------===// |
| |
| #include "LegalizeTypes.h" |
| #include "llvm/IR/DataLayout.h" |
| #include "llvm/Support/ErrorHandling.h" |
| #include "llvm/Support/raw_ostream.h" |
| using namespace llvm; |
| |
| #define DEBUG_TYPE "legalize-types" |
| |
| //===----------------------------------------------------------------------===// |
| // Result Vector Scalarization: <1 x ty> -> ty. |
| //===----------------------------------------------------------------------===// |
| |
| void DAGTypeLegalizer::ScalarizeVectorResult(SDNode *N, unsigned ResNo) { |
| DEBUG(dbgs() << "Scalarize node result " << ResNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"); |
| SDValue R = SDValue(); |
| |
| switch (N->getOpcode()) { |
| default: |
| #ifndef NDEBUG |
| dbgs() << "ScalarizeVectorResult #" << ResNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"; |
| #endif |
| report_fatal_error("Do not know how to scalarize the result of this " |
| "operator!\n"); |
| |
| case ISD::MERGE_VALUES: R = ScalarizeVecRes_MERGE_VALUES(N, ResNo);break; |
| case ISD::BITCAST: R = ScalarizeVecRes_BITCAST(N); break; |
| case ISD::BUILD_VECTOR: R = ScalarizeVecRes_BUILD_VECTOR(N); break; |
| case ISD::EXTRACT_SUBVECTOR: R = ScalarizeVecRes_EXTRACT_SUBVECTOR(N); break; |
| case ISD::FP_ROUND: R = ScalarizeVecRes_FP_ROUND(N); break; |
| case ISD::FP_ROUND_INREG: R = ScalarizeVecRes_InregOp(N); break; |
| case ISD::FPOWI: R = ScalarizeVecRes_FPOWI(N); break; |
| case ISD::INSERT_VECTOR_ELT: R = ScalarizeVecRes_INSERT_VECTOR_ELT(N); break; |
| case ISD::LOAD: R = ScalarizeVecRes_LOAD(cast<LoadSDNode>(N));break; |
| case ISD::SCALAR_TO_VECTOR: R = ScalarizeVecRes_SCALAR_TO_VECTOR(N); break; |
| case ISD::SIGN_EXTEND_INREG: R = ScalarizeVecRes_InregOp(N); break; |
| case ISD::VSELECT: R = ScalarizeVecRes_VSELECT(N); break; |
| case ISD::SELECT: R = ScalarizeVecRes_SELECT(N); break; |
| case ISD::SELECT_CC: R = ScalarizeVecRes_SELECT_CC(N); break; |
| case ISD::SETCC: R = ScalarizeVecRes_SETCC(N); break; |
| case ISD::UNDEF: R = ScalarizeVecRes_UNDEF(N); break; |
| case ISD::VECTOR_SHUFFLE: R = ScalarizeVecRes_VECTOR_SHUFFLE(N); break; |
| case ISD::ANY_EXTEND_VECTOR_INREG: |
| case ISD::SIGN_EXTEND_VECTOR_INREG: |
| case ISD::ZERO_EXTEND_VECTOR_INREG: |
| R = ScalarizeVecRes_VecInregOp(N); |
| break; |
| case ISD::ANY_EXTEND: |
| case ISD::BITREVERSE: |
| 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::FEXP: |
| case ISD::FEXP2: |
| case ISD::FFLOOR: |
| case ISD::FLOG: |
| case ISD::FLOG10: |
| case ISD::FLOG2: |
| case ISD::FNEARBYINT: |
| case ISD::FNEG: |
| case ISD::FP_EXTEND: |
| case ISD::FP_TO_SINT: |
| case ISD::FP_TO_UINT: |
| case ISD::FRINT: |
| case ISD::FROUND: |
| case ISD::FSIN: |
| case ISD::FSQRT: |
| case ISD::FTRUNC: |
| case ISD::SIGN_EXTEND: |
| case ISD::SINT_TO_FP: |
| case ISD::TRUNCATE: |
| case ISD::UINT_TO_FP: |
| case ISD::ZERO_EXTEND: |
| case ISD::FCANONICALIZE: |
| R = ScalarizeVecRes_UnaryOp(N); |
| break; |
| |
| case ISD::ADD: |
| case ISD::AND: |
| case ISD::FADD: |
| case ISD::FCOPYSIGN: |
| case ISD::FDIV: |
| case ISD::FMUL: |
| case ISD::FMINNUM: |
| case ISD::FMAXNUM: |
| case ISD::FMINNAN: |
| case ISD::FMAXNAN: |
| case ISD::SMIN: |
| case ISD::SMAX: |
| case ISD::UMIN: |
| case ISD::UMAX: |
| |
| case ISD::FPOW: |
| case ISD::FREM: |
| case ISD::FSUB: |
| case ISD::MUL: |
| case ISD::OR: |
| case ISD::SDIV: |
| case ISD::SREM: |
| case ISD::SUB: |
| case ISD::UDIV: |
| case ISD::UREM: |
| case ISD::XOR: |
| case ISD::SHL: |
| case ISD::SRA: |
| case ISD::SRL: |
| R = ScalarizeVecRes_BinOp(N); |
| break; |
| case ISD::FMA: |
| R = ScalarizeVecRes_TernaryOp(N); |
| break; |
| } |
| |
| // If R is null, the sub-method took care of registering the result. |
| if (R.getNode()) |
| SetScalarizedVector(SDValue(N, ResNo), R); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_BinOp(SDNode *N) { |
| SDValue LHS = GetScalarizedVector(N->getOperand(0)); |
| SDValue RHS = GetScalarizedVector(N->getOperand(1)); |
| return DAG.getNode(N->getOpcode(), SDLoc(N), |
| LHS.getValueType(), LHS, RHS, N->getFlags()); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_TernaryOp(SDNode *N) { |
| SDValue Op0 = GetScalarizedVector(N->getOperand(0)); |
| SDValue Op1 = GetScalarizedVector(N->getOperand(1)); |
| SDValue Op2 = GetScalarizedVector(N->getOperand(2)); |
| return DAG.getNode(N->getOpcode(), SDLoc(N), |
| Op0.getValueType(), Op0, Op1, Op2); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_MERGE_VALUES(SDNode *N, |
| unsigned ResNo) { |
| SDValue Op = DisintegrateMERGE_VALUES(N, ResNo); |
| return GetScalarizedVector(Op); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_BITCAST(SDNode *N) { |
| EVT NewVT = N->getValueType(0).getVectorElementType(); |
| return DAG.getNode(ISD::BITCAST, SDLoc(N), |
| NewVT, N->getOperand(0)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_BUILD_VECTOR(SDNode *N) { |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| SDValue InOp = N->getOperand(0); |
| // The BUILD_VECTOR operands may be of wider element types and |
| // we may need to truncate them back to the requested return type. |
| if (EltVT.isInteger()) |
| return DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, InOp); |
| return InOp; |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_EXTRACT_SUBVECTOR(SDNode *N) { |
| return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N), |
| N->getValueType(0).getVectorElementType(), |
| N->getOperand(0), N->getOperand(1)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_FP_ROUND(SDNode *N) { |
| EVT NewVT = N->getValueType(0).getVectorElementType(); |
| SDValue Op = GetScalarizedVector(N->getOperand(0)); |
| return DAG.getNode(ISD::FP_ROUND, SDLoc(N), |
| NewVT, Op, N->getOperand(1)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_FPOWI(SDNode *N) { |
| SDValue Op = GetScalarizedVector(N->getOperand(0)); |
| return DAG.getNode(ISD::FPOWI, SDLoc(N), |
| Op.getValueType(), Op, N->getOperand(1)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_INSERT_VECTOR_ELT(SDNode *N) { |
| // The value to insert may have a wider type than the vector element type, |
| // so be sure to truncate it to the element type if necessary. |
| SDValue Op = N->getOperand(1); |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| if (Op.getValueType() != EltVT) |
| // FIXME: Can this happen for floating point types? |
| Op = DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, Op); |
| return Op; |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_LOAD(LoadSDNode *N) { |
| assert(N->isUnindexed() && "Indexed vector load?"); |
| |
| SDValue Result = DAG.getLoad( |
| ISD::UNINDEXED, N->getExtensionType(), |
| N->getValueType(0).getVectorElementType(), SDLoc(N), N->getChain(), |
| N->getBasePtr(), DAG.getUNDEF(N->getBasePtr().getValueType()), |
| N->getPointerInfo(), N->getMemoryVT().getVectorElementType(), |
| N->getOriginalAlignment(), N->getMemOperand()->getFlags(), |
| N->getAAInfo()); |
| |
| // Legalize the chain result - switch anything that used the old chain to |
| // use the new one. |
| ReplaceValueWith(SDValue(N, 1), Result.getValue(1)); |
| return Result; |
| } |
| |
| 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 = 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, DL, TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| } |
| return DAG.getNode(N->getOpcode(), SDLoc(N), DestVT, Op); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_InregOp(SDNode *N) { |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| EVT ExtVT = cast<VTSDNode>(N->getOperand(1))->getVT().getVectorElementType(); |
| SDValue LHS = GetScalarizedVector(N->getOperand(0)); |
| return DAG.getNode(N->getOpcode(), SDLoc(N), EltVT, |
| LHS, DAG.getValueType(ExtVT)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_VecInregOp(SDNode *N) { |
| SDLoc DL(N); |
| SDValue Op = N->getOperand(0); |
| |
| EVT OpVT = Op.getValueType(); |
| EVT OpEltVT = OpVT.getVectorElementType(); |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| |
| if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) { |
| Op = GetScalarizedVector(Op); |
| } else { |
| Op = DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, DL, OpEltVT, Op, |
| DAG.getConstant(0, DL, TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| } |
| |
| switch (N->getOpcode()) { |
| case ISD::ANY_EXTEND_VECTOR_INREG: |
| return DAG.getNode(ISD::ANY_EXTEND, DL, EltVT, Op); |
| case ISD::SIGN_EXTEND_VECTOR_INREG: |
| return DAG.getNode(ISD::SIGN_EXTEND, DL, EltVT, Op); |
| case ISD::ZERO_EXTEND_VECTOR_INREG: |
| return DAG.getNode(ISD::ZERO_EXTEND, DL, EltVT, Op); |
| } |
| |
| llvm_unreachable("Illegal extend_vector_inreg opcode"); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_SCALAR_TO_VECTOR(SDNode *N) { |
| // If the operand is wider than the vector element type then it is implicitly |
| // truncated. Make that explicit here. |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| SDValue InOp = N->getOperand(0); |
| if (InOp.getValueType() != EltVT) |
| return DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, InOp); |
| return InOp; |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_VSELECT(SDNode *N) { |
| SDValue Cond = N->getOperand(0); |
| EVT OpVT = Cond.getValueType(); |
| SDLoc DL(N); |
| // The vselect result and true/value operands needs scalarizing, but it's |
| // not a given that the Cond does. For instance, in AVX512 v1i1 is legal. |
| // See the similar logic in ScalarizeVecRes_VSETCC |
| if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) { |
| Cond = GetScalarizedVector(Cond); |
| } else { |
| EVT VT = OpVT.getVectorElementType(); |
| Cond = DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, DL, VT, Cond, |
| DAG.getConstant(0, DL, TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| } |
| |
| SDValue LHS = GetScalarizedVector(N->getOperand(1)); |
| 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) { |
| case TargetLowering::UndefinedBooleanContent: |
| break; |
| case TargetLowering::ZeroOrOneBooleanContent: |
| assert(VecBool == TargetLowering::UndefinedBooleanContent || |
| VecBool == TargetLowering::ZeroOrNegativeOneBooleanContent); |
| // Vector read from all ones, scalar expects a single 1 so mask. |
| Cond = DAG.getNode(ISD::AND, SDLoc(N), CondVT, |
| Cond, DAG.getConstant(1, SDLoc(N), CondVT)); |
| break; |
| case TargetLowering::ZeroOrNegativeOneBooleanContent: |
| assert(VecBool == TargetLowering::UndefinedBooleanContent || |
| VecBool == TargetLowering::ZeroOrOneBooleanContent); |
| // Vector reads from a one, scalar from all ones so sign extend. |
| Cond = DAG.getNode(ISD::SIGN_EXTEND_INREG, SDLoc(N), CondVT, |
| Cond, DAG.getValueType(MVT::i1)); |
| break; |
| } |
| } |
| |
| return DAG.getSelect(SDLoc(N), |
| LHS.getValueType(), Cond, LHS, |
| GetScalarizedVector(N->getOperand(2))); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_SELECT(SDNode *N) { |
| SDValue LHS = GetScalarizedVector(N->getOperand(1)); |
| return DAG.getSelect(SDLoc(N), |
| LHS.getValueType(), N->getOperand(0), LHS, |
| GetScalarizedVector(N->getOperand(2))); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_SELECT_CC(SDNode *N) { |
| SDValue LHS = GetScalarizedVector(N->getOperand(2)); |
| return DAG.getNode(ISD::SELECT_CC, SDLoc(N), LHS.getValueType(), |
| N->getOperand(0), N->getOperand(1), |
| LHS, GetScalarizedVector(N->getOperand(3)), |
| N->getOperand(4)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_SETCC(SDNode *N) { |
| assert(N->getValueType(0).isVector() == |
| N->getOperand(0).getValueType().isVector() && |
| "Scalar/Vector type mismatch"); |
| |
| if (N->getValueType(0).isVector()) return ScalarizeVecRes_VSETCC(N); |
| |
| SDValue LHS = GetScalarizedVector(N->getOperand(0)); |
| SDValue RHS = GetScalarizedVector(N->getOperand(1)); |
| SDLoc DL(N); |
| |
| // Turn it into a scalar SETCC. |
| return DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS, N->getOperand(2)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_UNDEF(SDNode *N) { |
| return DAG.getUNDEF(N->getValueType(0).getVectorElementType()); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_VECTOR_SHUFFLE(SDNode *N) { |
| // Figure out if the scalar is the LHS or RHS and return it. |
| SDValue Arg = N->getOperand(2).getOperand(0); |
| if (Arg.isUndef()) |
| return DAG.getUNDEF(N->getValueType(0).getVectorElementType()); |
| unsigned Op = !cast<ConstantSDNode>(Arg)->isNullValue(); |
| return GetScalarizedVector(N->getOperand(Op)); |
| } |
| |
| SDValue DAGTypeLegalizer::ScalarizeVecRes_VSETCC(SDNode *N) { |
| assert(N->getValueType(0).isVector() && |
| N->getOperand(0).getValueType().isVector() && |
| "Operand types must be vectors"); |
| SDValue LHS = N->getOperand(0); |
| SDValue RHS = N->getOperand(1); |
| EVT OpVT = LHS.getValueType(); |
| EVT NVT = N->getValueType(0).getVectorElementType(); |
| SDLoc DL(N); |
| |
| // The result needs scalarizing, but it's not a given that the source does. |
| if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) { |
| LHS = GetScalarizedVector(LHS); |
| RHS = GetScalarizedVector(RHS); |
| } else { |
| EVT VT = OpVT.getVectorElementType(); |
| LHS = DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, DL, VT, LHS, |
| DAG.getConstant(0, DL, TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| RHS = DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, DL, VT, RHS, |
| DAG.getConstant(0, DL, TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| } |
| |
| // Turn it into a scalar SETCC. |
| SDValue Res = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS, |
| N->getOperand(2)); |
| // Vectors may have a different boolean contents to scalars. Promote the |
| // value appropriately. |
| ISD::NodeType ExtendCode = |
| TargetLowering::getExtendForContent(TLI.getBooleanContents(OpVT)); |
| return DAG.getNode(ExtendCode, DL, NVT, Res); |
| } |
| |
| |
| //===----------------------------------------------------------------------===// |
| // Operand Vector Scalarization <1 x ty> -> ty. |
| //===----------------------------------------------------------------------===// |
| |
| bool DAGTypeLegalizer::ScalarizeVectorOperand(SDNode *N, unsigned OpNo) { |
| DEBUG(dbgs() << "Scalarize node operand " << OpNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"); |
| SDValue Res = SDValue(); |
| |
| if (!Res.getNode()) { |
| switch (N->getOpcode()) { |
| default: |
| #ifndef NDEBUG |
| dbgs() << "ScalarizeVectorOperand Op #" << OpNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"; |
| #endif |
| llvm_unreachable("Do not know how to scalarize this operator's operand!"); |
| case ISD::BITCAST: |
| Res = ScalarizeVecOp_BITCAST(N); |
| break; |
| case ISD::ANY_EXTEND: |
| 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: |
| Res = ScalarizeVecOp_CONCAT_VECTORS(N); |
| break; |
| case ISD::EXTRACT_VECTOR_ELT: |
| Res = ScalarizeVecOp_EXTRACT_VECTOR_ELT(N); |
| break; |
| case ISD::VSELECT: |
| Res = ScalarizeVecOp_VSELECT(N); |
| break; |
| case ISD::SETCC: |
| Res = ScalarizeVecOp_VSETCC(N); |
| break; |
| case ISD::STORE: |
| Res = ScalarizeVecOp_STORE(cast<StoreSDNode>(N), OpNo); |
| break; |
| case ISD::FP_ROUND: |
| Res = ScalarizeVecOp_FP_ROUND(N, OpNo); |
| break; |
| } |
| } |
| |
| // If the result is null, the sub-method took care of registering results etc. |
| if (!Res.getNode()) return false; |
| |
| // If the result is N, the sub-method updated N in place. Tell the legalizer |
| // core about this. |
| if (Res.getNode() == N) |
| return true; |
| |
| assert(Res.getValueType() == N->getValueType(0) && N->getNumValues() == 1 && |
| "Invalid operand expansion"); |
| |
| ReplaceValueWith(SDValue(N, 0), Res); |
| return false; |
| } |
| |
| /// If the value to convert is a vector that needs to be scalarized, it must be |
| /// <1 x ty>. Convert the element instead. |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_BITCAST(SDNode *N) { |
| SDValue Elt = GetScalarizedVector(N->getOperand(0)); |
| return DAG.getNode(ISD::BITCAST, SDLoc(N), |
| N->getValueType(0), Elt); |
| } |
| |
| /// 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 && |
| "Unexpected vector type!"); |
| SDValue Elt = GetScalarizedVector(N->getOperand(0)); |
| 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.getBuildVector(N->getValueType(0), SDLoc(N), Op); |
| } |
| |
| /// The vectors to concatenate have length one - use a BUILD_VECTOR instead. |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_CONCAT_VECTORS(SDNode *N) { |
| SmallVector<SDValue, 8> Ops(N->getNumOperands()); |
| for (unsigned i = 0, e = N->getNumOperands(); i < e; ++i) |
| Ops[i] = GetScalarizedVector(N->getOperand(i)); |
| return DAG.getBuildVector(N->getValueType(0), SDLoc(N), Ops); |
| } |
| |
| /// If the input is a vector that needs to be scalarized, it must be <1 x ty>, |
| /// so just return the element, ignoring the index. |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_EXTRACT_VECTOR_ELT(SDNode *N) { |
| EVT VT = N->getValueType(0); |
| SDValue Res = GetScalarizedVector(N->getOperand(0)); |
| if (Res.getValueType() != VT) |
| Res = VT.isFloatingPoint() |
| ? DAG.getNode(ISD::FP_EXTEND, SDLoc(N), VT, Res) |
| : DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), VT, Res); |
| return Res; |
| } |
| |
| /// If the input condition is a vector that needs to be scalarized, it must be |
| /// <1 x i1>, so just convert to a normal ISD::SELECT |
| /// (still with vector output type since that was acceptable if we got here). |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_VSELECT(SDNode *N) { |
| SDValue ScalarCond = GetScalarizedVector(N->getOperand(0)); |
| EVT VT = N->getValueType(0); |
| |
| return DAG.getNode(ISD::SELECT, SDLoc(N), VT, ScalarCond, N->getOperand(1), |
| N->getOperand(2)); |
| } |
| |
| /// If the operand is a vector that needs to be scalarized then the |
| /// result must be v1i1, so just convert to a scalar SETCC and wrap |
| /// with a scalar_to_vector since the res type is legal if we got here |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_VSETCC(SDNode *N) { |
| assert(N->getValueType(0).isVector() && |
| N->getOperand(0).getValueType().isVector() && |
| "Operand types must be vectors"); |
| assert(N->getValueType(0) == MVT::v1i1 && "Expected v1i1 type"); |
| |
| EVT VT = N->getValueType(0); |
| SDValue LHS = GetScalarizedVector(N->getOperand(0)); |
| SDValue RHS = GetScalarizedVector(N->getOperand(1)); |
| |
| EVT OpVT = N->getOperand(0).getValueType(); |
| EVT NVT = VT.getVectorElementType(); |
| SDLoc DL(N); |
| // Turn it into a scalar SETCC. |
| SDValue Res = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS, |
| N->getOperand(2)); |
| |
| // Vectors may have a different boolean contents to scalars. Promote the |
| // value appropriately. |
| ISD::NodeType ExtendCode = |
| TargetLowering::getExtendForContent(TLI.getBooleanContents(OpVT)); |
| |
| Res = DAG.getNode(ExtendCode, DL, NVT, Res); |
| |
| return DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VT, Res); |
| } |
| |
| /// If the value to store is a vector that needs to be scalarized, it must be |
| /// <1 x ty>. Just store the element. |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_STORE(StoreSDNode *N, unsigned OpNo){ |
| assert(N->isUnindexed() && "Indexed store of one-element vector?"); |
| assert(OpNo == 1 && "Do not know how to scalarize this operand!"); |
| SDLoc dl(N); |
| |
| if (N->isTruncatingStore()) |
| return DAG.getTruncStore( |
| N->getChain(), dl, GetScalarizedVector(N->getOperand(1)), |
| N->getBasePtr(), N->getPointerInfo(), |
| N->getMemoryVT().getVectorElementType(), N->getAlignment(), |
| N->getMemOperand()->getFlags(), N->getAAInfo()); |
| |
| return DAG.getStore(N->getChain(), dl, GetScalarizedVector(N->getOperand(1)), |
| N->getBasePtr(), N->getPointerInfo(), |
| N->getOriginalAlignment(), N->getMemOperand()->getFlags(), |
| N->getAAInfo()); |
| } |
| |
| /// If the value to round is a vector that needs to be scalarized, it must be |
| /// <1 x ty>. Convert the element instead. |
| SDValue DAGTypeLegalizer::ScalarizeVecOp_FP_ROUND(SDNode *N, unsigned OpNo) { |
| SDValue Elt = GetScalarizedVector(N->getOperand(0)); |
| SDValue Res = DAG.getNode(ISD::FP_ROUND, SDLoc(N), |
| N->getValueType(0).getVectorElementType(), Elt, |
| N->getOperand(1)); |
| return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res); |
| } |
| |
| //===----------------------------------------------------------------------===// |
| // Result Vector Splitting |
| //===----------------------------------------------------------------------===// |
| |
| /// This method is called when the specified result of the specified node is |
| /// found to need vector splitting. At this point, the node may also have |
| /// invalid operands or may have other results that need legalization, we just |
| /// know that (at least) one result needs vector splitting. |
| void DAGTypeLegalizer::SplitVectorResult(SDNode *N, unsigned ResNo) { |
| DEBUG(dbgs() << "Split node result: "; |
| N->dump(&DAG); |
| dbgs() << "\n"); |
| SDValue Lo, Hi; |
| |
| // See if the target wants to custom expand this node. |
| if (CustomLowerNode(N, N->getValueType(ResNo), true)) |
| return; |
| |
| switch (N->getOpcode()) { |
| default: |
| #ifndef NDEBUG |
| dbgs() << "SplitVectorResult #" << ResNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"; |
| #endif |
| report_fatal_error("Do not know how to split the result of this " |
| "operator!\n"); |
| |
| case ISD::MERGE_VALUES: SplitRes_MERGE_VALUES(N, ResNo, Lo, Hi); break; |
| case ISD::VSELECT: |
| case ISD::SELECT: SplitRes_SELECT(N, Lo, Hi); break; |
| case ISD::SELECT_CC: SplitRes_SELECT_CC(N, Lo, Hi); break; |
| case ISD::UNDEF: SplitRes_UNDEF(N, Lo, Hi); break; |
| case ISD::BITCAST: SplitVecRes_BITCAST(N, Lo, Hi); break; |
| 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::FCOPYSIGN: SplitVecRes_FCOPYSIGN(N, Lo, Hi); break; |
| case ISD::INSERT_VECTOR_ELT: SplitVecRes_INSERT_VECTOR_ELT(N, Lo, Hi); break; |
| case ISD::SCALAR_TO_VECTOR: SplitVecRes_SCALAR_TO_VECTOR(N, Lo, Hi); break; |
| case ISD::SIGN_EXTEND_INREG: SplitVecRes_InregOp(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::MGATHER: |
| SplitVecRes_MGATHER(cast<MaskedGatherSDNode>(N), Lo, Hi); |
| break; |
| case ISD::SETCC: |
| SplitVecRes_SETCC(N, Lo, Hi); |
| break; |
| case ISD::VECTOR_SHUFFLE: |
| SplitVecRes_VECTOR_SHUFFLE(cast<ShuffleVectorSDNode>(N), Lo, Hi); |
| break; |
| |
| case ISD::ANY_EXTEND_VECTOR_INREG: |
| case ISD::SIGN_EXTEND_VECTOR_INREG: |
| case ISD::ZERO_EXTEND_VECTOR_INREG: |
| SplitVecRes_ExtVecInRegOp(N, Lo, Hi); |
| break; |
| |
| case ISD::BITREVERSE: |
| case ISD::BSWAP: |
| case ISD::CTLZ: |
| case ISD::CTTZ: |
| case ISD::CTLZ_ZERO_UNDEF: |
| case ISD::CTTZ_ZERO_UNDEF: |
| case ISD::CTPOP: |
| case ISD::FABS: |
| case ISD::FCEIL: |
| case ISD::FCOS: |
| case ISD::FEXP: |
| case ISD::FEXP2: |
| case ISD::FFLOOR: |
| case ISD::FLOG: |
| case ISD::FLOG10: |
| case ISD::FLOG2: |
| case ISD::FNEARBYINT: |
| case ISD::FNEG: |
| case ISD::FP_EXTEND: |
| case ISD::FP_ROUND: |
| case ISD::FP_TO_SINT: |
| case ISD::FP_TO_UINT: |
| case ISD::FRINT: |
| case ISD::FROUND: |
| case ISD::FSIN: |
| case ISD::FSQRT: |
| case ISD::FTRUNC: |
| case ISD::SINT_TO_FP: |
| case ISD::TRUNCATE: |
| case ISD::UINT_TO_FP: |
| case ISD::FCANONICALIZE: |
| SplitVecRes_UnaryOp(N, Lo, Hi); |
| break; |
| |
| case ISD::ANY_EXTEND: |
| case ISD::SIGN_EXTEND: |
| case ISD::ZERO_EXTEND: |
| SplitVecRes_ExtendOp(N, Lo, Hi); |
| break; |
| |
| case ISD::ADD: |
| case ISD::SUB: |
| case ISD::MUL: |
| case ISD::MULHS: |
| case ISD::MULHU: |
| case ISD::FADD: |
| case ISD::FSUB: |
| case ISD::FMUL: |
| case ISD::FMINNUM: |
| case ISD::FMAXNUM: |
| case ISD::FMINNAN: |
| case ISD::FMAXNAN: |
| case ISD::SDIV: |
| case ISD::UDIV: |
| case ISD::FDIV: |
| case ISD::FPOW: |
| case ISD::AND: |
| case ISD::OR: |
| case ISD::XOR: |
| case ISD::SHL: |
| case ISD::SRA: |
| case ISD::SRL: |
| case ISD::UREM: |
| case ISD::SREM: |
| case ISD::FREM: |
| case ISD::SMIN: |
| case ISD::SMAX: |
| case ISD::UMIN: |
| case ISD::UMAX: |
| SplitVecRes_BinOp(N, Lo, Hi); |
| break; |
| case ISD::FMA: |
| SplitVecRes_TernaryOp(N, Lo, Hi); |
| break; |
| } |
| |
| // If Lo/Hi is null, the sub-method took care of registering results etc. |
| if (Lo.getNode()) |
| SetSplitVector(SDValue(N, ResNo), Lo, Hi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_BinOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue LHSLo, LHSHi; |
| GetSplitVector(N->getOperand(0), LHSLo, LHSHi); |
| SDValue RHSLo, RHSHi; |
| GetSplitVector(N->getOperand(1), RHSLo, RHSHi); |
| SDLoc dl(N); |
| |
| const SDNodeFlags Flags = N->getFlags(); |
| unsigned Opcode = N->getOpcode(); |
| Lo = DAG.getNode(Opcode, dl, LHSLo.getValueType(), LHSLo, RHSLo, Flags); |
| Hi = DAG.getNode(Opcode, dl, LHSHi.getValueType(), LHSHi, RHSHi, Flags); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_TernaryOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue Op0Lo, Op0Hi; |
| GetSplitVector(N->getOperand(0), Op0Lo, Op0Hi); |
| SDValue Op1Lo, Op1Hi; |
| GetSplitVector(N->getOperand(1), Op1Lo, Op1Hi); |
| SDValue Op2Lo, Op2Hi; |
| GetSplitVector(N->getOperand(2), Op2Lo, Op2Hi); |
| SDLoc dl(N); |
| |
| Lo = DAG.getNode(N->getOpcode(), dl, Op0Lo.getValueType(), |
| Op0Lo, Op1Lo, Op2Lo); |
| Hi = DAG.getNode(N->getOpcode(), dl, Op0Hi.getValueType(), |
| Op0Hi, Op1Hi, Op2Hi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_BITCAST(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| // We know the result is a vector. The input may be either a vector or a |
| // scalar value. |
| EVT LoVT, HiVT; |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| SDLoc dl(N); |
| |
| SDValue InOp = N->getOperand(0); |
| EVT InVT = InOp.getValueType(); |
| |
| // Handle some special cases efficiently. |
| switch (getTypeAction(InVT)) { |
| case TargetLowering::TypeLegal: |
| case TargetLowering::TypePromoteInteger: |
| case TargetLowering::TypePromoteFloat: |
| case TargetLowering::TypeSoftenFloat: |
| case TargetLowering::TypeScalarizeVector: |
| case TargetLowering::TypeWidenVector: |
| break; |
| case TargetLowering::TypeExpandInteger: |
| case TargetLowering::TypeExpandFloat: |
| // A scalar to vector conversion, where the scalar needs expansion. |
| // If the vector is being split in two then we can just convert the |
| // expanded pieces. |
| if (LoVT == HiVT) { |
| GetExpandedOp(InOp, Lo, Hi); |
| if (DAG.getDataLayout().isBigEndian()) |
| std::swap(Lo, Hi); |
| Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo); |
| Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi); |
| return; |
| } |
| break; |
| case TargetLowering::TypeSplitVector: |
| // If the input is a vector that needs to be split, convert each split |
| // piece of the input now. |
| GetSplitVector(InOp, Lo, Hi); |
| Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo); |
| Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi); |
| return; |
| } |
| |
| // In the general case, convert the input to an integer and split it by hand. |
| EVT LoIntVT = EVT::getIntegerVT(*DAG.getContext(), LoVT.getSizeInBits()); |
| EVT HiIntVT = EVT::getIntegerVT(*DAG.getContext(), HiVT.getSizeInBits()); |
| if (DAG.getDataLayout().isBigEndian()) |
| std::swap(LoIntVT, HiIntVT); |
| |
| SplitInteger(BitConvertToInteger(InOp), LoIntVT, HiIntVT, Lo, Hi); |
| |
| if (DAG.getDataLayout().isBigEndian()) |
| std::swap(Lo, Hi); |
| Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo); |
| Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_BUILD_VECTOR(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| EVT LoVT, HiVT; |
| SDLoc dl(N); |
| 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.getBuildVector(LoVT, dl, LoOps); |
| |
| SmallVector<SDValue, 8> HiOps(N->op_begin()+LoNumElts, N->op_end()); |
| Hi = DAG.getBuildVector(HiVT, dl, HiOps); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_CONCAT_VECTORS(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| assert(!(N->getNumOperands() & 1) && "Unsupported CONCAT_VECTORS"); |
| SDLoc dl(N); |
| unsigned NumSubvectors = N->getNumOperands() / 2; |
| if (NumSubvectors == 1) { |
| Lo = N->getOperand(0); |
| Hi = N->getOperand(1); |
| return; |
| } |
| |
| EVT LoVT, HiVT; |
| 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); |
| |
| SmallVector<SDValue, 8> HiOps(N->op_begin()+NumSubvectors, N->op_end()); |
| Hi = DAG.getNode(ISD::CONCAT_VECTORS, dl, HiVT, HiOps); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_EXTRACT_SUBVECTOR(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue Vec = N->getOperand(0); |
| SDValue Idx = N->getOperand(1); |
| SDLoc dl(N); |
| |
| EVT LoVT, HiVT; |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| |
| Lo = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, LoVT, Vec, Idx); |
| uint64_t IdxVal = cast<ConstantSDNode>(Idx)->getZExtValue(); |
| Hi = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, HiVT, Vec, |
| DAG.getConstant(IdxVal + LoVT.getVectorNumElements(), dl, |
| TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| } |
| |
| 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); |
| |
| EVT VecVT = Vec.getValueType(); |
| unsigned VecElems = VecVT.getVectorNumElements(); |
| unsigned SubElems = SubVec.getValueType().getVectorNumElements(); |
| |
| // If we know the index is 0, and we know the subvector doesn't cross the |
| // boundary between the halves, we can avoid spilling the vector, and insert |
| // into the lower half of the split vector directly. |
| // TODO: The IdxVal == 0 constraint is artificial, we could do this whenever |
| // the index is constant and there is no boundary crossing. But those cases |
| // don't seem to get hit in practice. |
| if (ConstantSDNode *ConstIdx = dyn_cast<ConstantSDNode>(Idx)) { |
| unsigned IdxVal = ConstIdx->getZExtValue(); |
| if ((IdxVal == 0) && (IdxVal + SubElems <= VecElems / 2)) { |
| EVT LoVT, HiVT; |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| Lo = DAG.getNode(ISD::INSERT_SUBVECTOR, dl, LoVT, Lo, SubVec, Idx); |
| return; |
| } |
| } |
| |
| // Spill the vector to the stack. |
| SDValue StackPtr = DAG.CreateStackTemporary(VecVT); |
| SDValue Store = |
| DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, MachinePointerInfo()); |
| |
| // Store the new subvector into the specified index. |
| SDValue SubVecPtr = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Idx); |
| Type *VecType = VecVT.getTypeForEVT(*DAG.getContext()); |
| unsigned Alignment = DAG.getDataLayout().getPrefTypeAlignment(VecType); |
| Store = DAG.getStore(Store, dl, SubVec, SubVecPtr, MachinePointerInfo()); |
| |
| // Load the Lo part from the stack slot. |
| Lo = |
| DAG.getLoad(Lo.getValueType(), dl, Store, StackPtr, MachinePointerInfo()); |
| |
| // Increment the pointer to the other part. |
| unsigned IncrementSize = Lo.getValueSizeInBits() / 8; |
| StackPtr = |
| DAG.getNode(ISD::ADD, dl, StackPtr.getValueType(), StackPtr, |
| DAG.getConstant(IncrementSize, dl, StackPtr.getValueType())); |
| |
| // Load the Hi part from the stack slot. |
| Hi = DAG.getLoad(Hi.getValueType(), dl, Store, StackPtr, MachinePointerInfo(), |
| MinAlign(Alignment, IncrementSize)); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_FPOWI(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDLoc dl(N); |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| Lo = DAG.getNode(ISD::FPOWI, dl, Lo.getValueType(), Lo, N->getOperand(1)); |
| Hi = DAG.getNode(ISD::FPOWI, dl, Hi.getValueType(), Hi, N->getOperand(1)); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_FCOPYSIGN(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue LHSLo, LHSHi; |
| GetSplitVector(N->getOperand(0), LHSLo, LHSHi); |
| SDLoc DL(N); |
| |
| SDValue RHSLo, RHSHi; |
| SDValue RHS = N->getOperand(1); |
| EVT RHSVT = RHS.getValueType(); |
| if (getTypeAction(RHSVT) == TargetLowering::TypeSplitVector) |
| GetSplitVector(RHS, RHSLo, RHSHi); |
| else |
| std::tie(RHSLo, RHSHi) = DAG.SplitVector(RHS, SDLoc(RHS)); |
| |
| |
| Lo = DAG.getNode(ISD::FCOPYSIGN, DL, LHSLo.getValueType(), LHSLo, RHSLo); |
| Hi = DAG.getNode(ISD::FCOPYSIGN, DL, LHSHi.getValueType(), LHSHi, RHSHi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_InregOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue LHSLo, LHSHi; |
| GetSplitVector(N->getOperand(0), LHSLo, LHSHi); |
| SDLoc dl(N); |
| |
| EVT LoVT, HiVT; |
| std::tie(LoVT, HiVT) = |
| DAG.GetSplitDestVTs(cast<VTSDNode>(N->getOperand(1))->getVT()); |
| |
| Lo = DAG.getNode(N->getOpcode(), dl, LHSLo.getValueType(), LHSLo, |
| DAG.getValueType(LoVT)); |
| Hi = DAG.getNode(N->getOpcode(), dl, LHSHi.getValueType(), LHSHi, |
| DAG.getValueType(HiVT)); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_ExtVecInRegOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| unsigned Opcode = N->getOpcode(); |
| SDValue N0 = N->getOperand(0); |
| |
| SDLoc dl(N); |
| SDValue InLo, InHi; |
| |
| if (getTypeAction(N0.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(N0, InLo, InHi); |
| else |
| std::tie(InLo, InHi) = DAG.SplitVectorOperand(N, 0); |
| |
| EVT InLoVT = InLo.getValueType(); |
| unsigned InNumElements = InLoVT.getVectorNumElements(); |
| |
| EVT OutLoVT, OutHiVT; |
| std::tie(OutLoVT, OutHiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| unsigned OutNumElements = OutLoVT.getVectorNumElements(); |
| assert((2 * OutNumElements) <= InNumElements && |
| "Illegal extend vector in reg split"); |
| |
| // *_EXTEND_VECTOR_INREG instructions extend the lowest elements of the |
| // input vector (i.e. we only use InLo): |
| // OutLo will extend the first OutNumElements from InLo. |
| // OutHi will extend the next OutNumElements from InLo. |
| |
| // Shuffle the elements from InLo for OutHi into the bottom elements to |
| // create a 'fake' InHi. |
| SmallVector<int, 8> SplitHi(InNumElements, -1); |
| for (unsigned i = 0; i != OutNumElements; ++i) |
| SplitHi[i] = i + OutNumElements; |
| InHi = DAG.getVectorShuffle(InLoVT, dl, InLo, DAG.getUNDEF(InLoVT), SplitHi); |
| |
| Lo = DAG.getNode(Opcode, dl, OutLoVT, InLo); |
| Hi = DAG.getNode(Opcode, dl, OutHiVT, InHi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDValue Vec = N->getOperand(0); |
| SDValue Elt = N->getOperand(1); |
| SDValue Idx = N->getOperand(2); |
| SDLoc dl(N); |
| GetSplitVector(Vec, Lo, Hi); |
| |
| if (ConstantSDNode *CIdx = dyn_cast<ConstantSDNode>(Idx)) { |
| unsigned IdxVal = CIdx->getZExtValue(); |
| unsigned LoNumElts = Lo.getValueType().getVectorNumElements(); |
| if (IdxVal < LoNumElts) |
| Lo = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, |
| Lo.getValueType(), Lo, Elt, Idx); |
| else |
| Hi = |
| DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, Hi.getValueType(), Hi, Elt, |
| DAG.getConstant(IdxVal - LoNumElts, dl, |
| TLI.getVectorIdxTy(DAG.getDataLayout()))); |
| 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(); |
| SDValue StackPtr = DAG.CreateStackTemporary(VecVT); |
| SDValue Store = |
| DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, MachinePointerInfo()); |
| |
| // Store the new element. This may be larger than the vector element type, |
| // so use a truncating store. |
| SDValue EltPtr = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Idx); |
| Type *VecType = VecVT.getTypeForEVT(*DAG.getContext()); |
| unsigned Alignment = DAG.getDataLayout().getPrefTypeAlignment(VecType); |
| Store = |
| DAG.getTruncStore(Store, dl, Elt, EltPtr, MachinePointerInfo(), EltVT); |
| |
| // Load the Lo part from the stack slot. |
| Lo = |
| DAG.getLoad(Lo.getValueType(), dl, Store, StackPtr, MachinePointerInfo()); |
| |
| // Increment the pointer to the other part. |
| unsigned IncrementSize = Lo.getValueSizeInBits() / 8; |
| StackPtr = DAG.getNode(ISD::ADD, dl, StackPtr.getValueType(), StackPtr, |
| DAG.getConstant(IncrementSize, dl, |
| StackPtr.getValueType())); |
| |
| // Load the Hi part from the stack slot. |
| Hi = DAG.getLoad(Hi.getValueType(), dl, Store, StackPtr, MachinePointerInfo(), |
| MinAlign(Alignment, IncrementSize)); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_SCALAR_TO_VECTOR(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| EVT LoVT, HiVT; |
| SDLoc dl(N); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| Lo = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, LoVT, N->getOperand(0)); |
| Hi = DAG.getUNDEF(HiVT); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_LOAD(LoadSDNode *LD, SDValue &Lo, |
| SDValue &Hi) { |
| assert(ISD::isUNINDEXEDLoad(LD) && "Indexed load during type legalization!"); |
| EVT LoVT, HiVT; |
| SDLoc dl(LD); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(LD->getValueType(0)); |
| |
| ISD::LoadExtType ExtType = LD->getExtensionType(); |
| SDValue Ch = LD->getChain(); |
| SDValue Ptr = LD->getBasePtr(); |
| SDValue Offset = DAG.getUNDEF(Ptr.getValueType()); |
| EVT MemoryVT = LD->getMemoryVT(); |
| unsigned Alignment = LD->getOriginalAlignment(); |
| MachineMemOperand::Flags MMOFlags = LD->getMemOperand()->getFlags(); |
| 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, Alignment, MMOFlags, AAInfo); |
| |
| unsigned IncrementSize = LoMemVT.getSizeInBits()/8; |
| Ptr = DAG.getNode(ISD::ADD, dl, Ptr.getValueType(), Ptr, |
| DAG.getConstant(IncrementSize, dl, Ptr.getValueType())); |
| Hi = DAG.getLoad(ISD::UNINDEXED, ExtType, HiVT, dl, Ch, Ptr, Offset, |
| LD->getPointerInfo().getWithOffset(IncrementSize), HiMemVT, |
| Alignment, MMOFlags, AAInfo); |
| |
| // 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)); |
| |
| // Legalize the chain result - switch anything that used the old chain to |
| // use the new 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(); |
| SDValue Src0 = MLD->getSrc0(); |
| 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; |
| |
| // Split Mask operand |
| SDValue MaskLo, MaskHi; |
| if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Mask, MaskLo, MaskHi); |
| else |
| std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl); |
| |
| EVT MemoryVT = MLD->getMemoryVT(); |
| EVT LoMemVT, HiMemVT; |
| std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT); |
| |
| SDValue Src0Lo, Src0Hi; |
| if (getTypeAction(Src0.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Src0, Src0Lo, Src0Hi); |
| else |
| 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, MLD->isExpandingLoad()); |
| |
| Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, dl, LoMemVT, DAG, |
| MLD->isExpandingLoad()); |
| |
| 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, MLD->isExpandingLoad()); |
| |
| |
| // 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)); |
| |
| // Legalize the chain result - switch anything that used the old chain to |
| // use the new one. |
| ReplaceValueWith(SDValue(MLD, 1), Ch); |
| |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_MGATHER(MaskedGatherSDNode *MGT, |
| SDValue &Lo, SDValue &Hi) { |
| EVT LoVT, HiVT; |
| SDLoc dl(MGT); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(MGT->getValueType(0)); |
| |
| SDValue Ch = MGT->getChain(); |
| SDValue Ptr = MGT->getBasePtr(); |
| SDValue Mask = MGT->getMask(); |
| SDValue Src0 = MGT->getValue(); |
| SDValue Index = MGT->getIndex(); |
| unsigned Alignment = MGT->getOriginalAlignment(); |
| |
| // Split Mask operand |
| SDValue MaskLo, MaskHi; |
| if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Mask, MaskLo, MaskHi); |
| else |
| std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl); |
| |
| EVT MemoryVT = MGT->getMemoryVT(); |
| EVT LoMemVT, HiMemVT; |
| // Split MemoryVT |
| std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT); |
| |
| SDValue Src0Lo, Src0Hi; |
| if (getTypeAction(Src0.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Src0, Src0Lo, Src0Hi); |
| else |
| std::tie(Src0Lo, Src0Hi) = DAG.SplitVector(Src0, dl); |
| |
| SDValue IndexHi, IndexLo; |
| if (getTypeAction(Index.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Index, IndexLo, IndexHi); |
| else |
| std::tie(IndexLo, IndexHi) = DAG.SplitVector(Index, dl); |
| |
| MachineMemOperand *MMO = DAG.getMachineFunction(). |
| getMachineMemOperand(MGT->getPointerInfo(), |
| MachineMemOperand::MOLoad, LoMemVT.getStoreSize(), |
| Alignment, MGT->getAAInfo(), MGT->getRanges()); |
| |
| SDValue OpsLo[] = {Ch, Src0Lo, MaskLo, Ptr, IndexLo}; |
| Lo = DAG.getMaskedGather(DAG.getVTList(LoVT, MVT::Other), LoVT, dl, OpsLo, |
| MMO); |
| |
| SDValue OpsHi[] = {Ch, Src0Hi, MaskHi, Ptr, IndexHi}; |
| Hi = DAG.getMaskedGather(DAG.getVTList(HiVT, MVT::Other), HiVT, dl, OpsHi, |
| MMO); |
| |
| // 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)); |
| |
| // Legalize the chain result - switch anything that used the old chain to |
| // use the new one. |
| ReplaceValueWith(SDValue(MGT, 1), Ch); |
| } |
| |
| |
| void DAGTypeLegalizer::SplitVecRes_SETCC(SDNode *N, SDValue &Lo, SDValue &Hi) { |
| assert(N->getValueType(0).isVector() && |
| N->getOperand(0).getValueType().isVector() && |
| "Operand types must be vectors"); |
| |
| EVT LoVT, HiVT; |
| SDLoc DL(N); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| |
| // Split the input. |
| SDValue LL, LH, RL, RH; |
| std::tie(LL, LH) = DAG.SplitVectorOperand(N, 0); |
| std::tie(RL, RH) = DAG.SplitVectorOperand(N, 1); |
| |
| Lo = DAG.getNode(N->getOpcode(), DL, LoVT, LL, RL, N->getOperand(2)); |
| Hi = DAG.getNode(N->getOpcode(), DL, HiVT, LH, RH, N->getOperand(2)); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_UnaryOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| // Get the dest types - they may not match the input types, e.g. int_to_fp. |
| EVT LoVT, HiVT; |
| SDLoc dl(N); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0)); |
| |
| // If the input also splits, handle it directly for a compile time speedup. |
| // Otherwise split it by hand. |
| EVT InVT = N->getOperand(0).getValueType(); |
| if (getTypeAction(InVT) == TargetLowering::TypeSplitVector) |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| else |
| std::tie(Lo, Hi) = DAG.SplitVectorOperand(N, 0); |
| |
| if (N->getOpcode() == ISD::FP_ROUND) { |
| Lo = DAG.getNode(N->getOpcode(), dl, LoVT, Lo, N->getOperand(1)); |
| Hi = DAG.getNode(N->getOpcode(), dl, HiVT, Hi, N->getOperand(1)); |
| } else { |
| Lo = DAG.getNode(N->getOpcode(), dl, LoVT, Lo); |
| Hi = DAG.getNode(N->getOpcode(), dl, HiVT, Hi); |
| } |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_ExtendOp(SDNode *N, SDValue &Lo, |
| SDValue &Hi) { |
| SDLoc dl(N); |
| EVT SrcVT = N->getOperand(0).getValueType(); |
| EVT DestVT = N->getValueType(0); |
| EVT LoVT, HiVT; |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(DestVT); |
| |
| // We can do better than a generic split operation if the extend is doing |
| // more than just doubling the width of the elements and the following are |
| // true: |
| // - The number of vector elements is even, |
| // - the source type is legal, |
| // - the type of a split source is illegal, |
| // - the type of an extended (by doubling element size) source is legal, and |
| // - the type of that extended source when split is legal. |
| // |
| // This won't necessarily completely legalize the operation, but it will |
| // more effectively move in the right direction and prevent falling down |
| // to scalarization in many cases due to the input vector being split too |
| // far. |
| unsigned NumElements = SrcVT.getVectorNumElements(); |
| if ((NumElements & 1) == 0 && |
| SrcVT.getSizeInBits() * 2 < DestVT.getSizeInBits()) { |
| LLVMContext &Ctx = *DAG.getContext(); |
| EVT NewSrcVT = SrcVT.widenIntegerVectorElementType(Ctx); |
| EVT SplitSrcVT = SrcVT.getHalfNumVectorElementsVT(Ctx); |
| |
| EVT SplitLoVT, SplitHiVT; |
| std::tie(SplitLoVT, SplitHiVT) = DAG.GetSplitDestVTs(NewSrcVT); |
| if (TLI.isTypeLegal(SrcVT) && !TLI.isTypeLegal(SplitSrcVT) && |
| TLI.isTypeLegal(NewSrcVT) && TLI.isTypeLegal(SplitLoVT)) { |
| DEBUG(dbgs() << "Split vector extend via incremental extend:"; |
| N->dump(&DAG); dbgs() << "\n"); |
| // Extend the source vector by one step. |
| SDValue NewSrc = |
| DAG.getNode(N->getOpcode(), dl, NewSrcVT, N->getOperand(0)); |
| // Get the low and high halves of the new, extended one step, vector. |
| std::tie(Lo, Hi) = DAG.SplitVector(NewSrc, dl); |
| // Extend those vector halves the rest of the way. |
| Lo = DAG.getNode(N->getOpcode(), dl, LoVT, Lo); |
| Hi = DAG.getNode(N->getOpcode(), dl, HiVT, Hi); |
| return; |
| } |
| } |
| // Fall back to the generic unary operator splitting otherwise. |
| SplitVecRes_UnaryOp(N, Lo, Hi); |
| } |
| |
| void DAGTypeLegalizer::SplitVecRes_VECTOR_SHUFFLE(ShuffleVectorSDNode *N, |
| SDValue &Lo, SDValue &Hi) { |
| // The low and high parts of the original input give four input vectors. |
| SDValue Inputs[4]; |
| SDLoc dl(N); |
| GetSplitVector(N->getOperand(0), Inputs[0], Inputs[1]); |
| GetSplitVector(N->getOperand(1), Inputs[2], Inputs[3]); |
| EVT NewVT = Inputs[0].getValueType(); |
| unsigned NewElts = NewVT.getVectorNumElements(); |
| |
| // If Lo or Hi uses elements from at most two of the four input vectors, then |
| // express it as a vector shuffle of those two inputs. Otherwise extract the |
| // input elements by hand and construct the Lo/Hi output using a BUILD_VECTOR. |
| SmallVector<int, 16> Ops; |
| for (unsigned High = 0; High < 2; ++High) { |
| SDValue &Output = High ? Hi : Lo; |
| |
| // Build a shuffle mask for the output, discovering on the fly which |
| // input vectors to use as shuffle operands (recorded in InputUsed). |
| // If building a suitable shuffle vector proves too hard, then bail |
| // out with useBuildVector set. |
| unsigned InputUsed[2] = { -1U, -1U }; // Not yet discovered. |
| unsigned FirstMaskIdx = High * NewElts; |
| bool useBuildVector = false; |
| for (unsigned MaskOffset = 0; MaskOffset < NewElts; ++MaskOffset) { |
| // The mask element. This indexes into the input. |
| int Idx = N->getMaskElt(FirstMaskIdx + MaskOffset); |
| |
| // The input vector this mask element indexes into. |
| unsigned Input = (unsigned)Idx / NewElts; |
| |
| if (Input >= array_lengthof(Inputs)) { |
| // The mask element does not index into any input vector. |
| Ops.push_back(-1); |
| continue; |
| } |
| |
| // Turn the index into an offset from the start of the input vector. |
| Idx -= Input * NewElts; |
| |
| // Find or create a shuffle vector operand to hold this input. |
| unsigned OpNo; |
| for (OpNo = 0; OpNo < array_lengthof(InputUsed); ++OpNo) { |
| if (InputUsed[OpNo] == Input) { |
| // This input vector is already an operand. |
| break; |
| } else if (InputUsed[OpNo] == -1U) { |
| // Create a new operand for this input vector. |
| InputUsed[OpNo] = Input; |
| break; |
| } |
| } |
| |
| if (OpNo >= array_lengthof(InputUsed)) { |
| // More than two input vectors used! Give up on trying to create a |
| // shuffle vector. Insert all elements into a BUILD_VECTOR instead. |
| useBuildVector = true; |
| break; |
| } |
| |
| // Add the mask index for the new shuffle vector. |
| Ops.push_back(Idx + OpNo * NewElts); |
| } |
| |
| if (useBuildVector) { |
| EVT EltVT = NewVT.getVectorElementType(); |
| SmallVector<SDValue, 16> SVOps; |
| |
| // Extract the input elements by hand. |
| for (unsigned MaskOffset = 0; MaskOffset < NewElts; ++MaskOffset) { |
| // The mask element. This indexes into the input. |
| int Idx = N->getMaskElt(FirstMaskIdx + MaskOffset); |
| |
| // The input vector this mask element indexes into. |
| unsigned Input = (unsigned)Idx / NewElts; |
| |
| if (Input >= array_lengthof(Inputs)) { |
| // The mask element is "undef" or indexes off the end of the input. |
| SVOps.push_back(DAG.getUNDEF(EltVT)); |
| continue; |
| } |
| |
| // Turn the index into an offset from the start of the input vector. |
| Idx -= Input * NewElts; |
| |
| // Extract the vector element by hand. |
| SVOps.push_back(DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, dl, EltVT, Inputs[Input], |
| DAG.getConstant(Idx, dl, TLI.getVectorIdxTy(DAG.getDataLayout())))); |
| } |
| |
| // Construct the Lo/Hi output using a BUILD_VECTOR. |
| Output = DAG.getBuildVector(NewVT, dl, SVOps); |
| } else if (InputUsed[0] == -1U) { |
| // No input vectors were used! The result is undefined. |
| Output = DAG.getUNDEF(NewVT); |
| } else { |
| SDValue Op0 = Inputs[InputUsed[0]]; |
| // If only one input was used, use an undefined vector for the other. |
| SDValue Op1 = InputUsed[1] == -1U ? |
| DAG.getUNDEF(NewVT) : Inputs[InputUsed[1]]; |
| // At least one input vector was used. Create a new shuffle vector. |
| Output = DAG.getVectorShuffle(NewVT, dl, Op0, Op1, Ops); |
| } |
| |
| Ops.clear(); |
| } |
| } |
| |
| |
| //===----------------------------------------------------------------------===// |
| // Operand Vector Splitting |
| //===----------------------------------------------------------------------===// |
| |
| /// This method is called when the specified operand of the specified node is |
| /// found to need vector splitting. At this point, all of the result types of |
| /// the node are known to be legal, but other operands of the node may need |
| /// legalization as well as the specified one. |
| bool DAGTypeLegalizer::SplitVectorOperand(SDNode *N, unsigned OpNo) { |
| DEBUG(dbgs() << "Split node operand: "; |
| N->dump(&DAG); |
| dbgs() << "\n"); |
| SDValue Res = SDValue(); |
| |
| // See if the target wants to custom split this node. |
| if (CustomLowerNode(N, N->getOperand(OpNo).getValueType(), false)) |
| return false; |
| |
| if (!Res.getNode()) { |
| switch (N->getOpcode()) { |
| default: |
| #ifndef NDEBUG |
| dbgs() << "SplitVectorOperand Op #" << OpNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"; |
| #endif |
| report_fatal_error("Do not know how to split this operator's " |
| "operand!\n"); |
| |
| case ISD::SETCC: Res = SplitVecOp_VSETCC(N); break; |
| case ISD::BITCAST: Res = SplitVecOp_BITCAST(N); break; |
| case ISD::EXTRACT_SUBVECTOR: Res = SplitVecOp_EXTRACT_SUBVECTOR(N); break; |
| case ISD::EXTRACT_VECTOR_ELT:Res = SplitVecOp_EXTRACT_VECTOR_ELT(N); break; |
| case ISD::CONCAT_VECTORS: Res = SplitVecOp_CONCAT_VECTORS(N); break; |
| case ISD::TRUNCATE: |
| Res = SplitVecOp_TruncateHelper(N); |
| break; |
| case ISD::FP_ROUND: Res = SplitVecOp_FP_ROUND(N); break; |
| case ISD::FCOPYSIGN: Res = SplitVecOp_FCOPYSIGN(N); break; |
| case ISD::STORE: |
| Res = SplitVecOp_STORE(cast<StoreSDNode>(N), OpNo); |
| break; |
| case ISD::MSTORE: |
| Res = SplitVecOp_MSTORE(cast<MaskedStoreSDNode>(N), OpNo); |
| break; |
| case ISD::MSCATTER: |
| Res = SplitVecOp_MSCATTER(cast<MaskedScatterSDNode>(N), OpNo); |
| break; |
| case ISD::MGATHER: |
| Res = SplitVecOp_MGATHER(cast<MaskedGatherSDNode>(N), OpNo); |
| break; |
| case ISD::VSELECT: |
| Res = SplitVecOp_VSELECT(N, OpNo); |
| break; |
| case ISD::FP_TO_SINT: |
| case ISD::FP_TO_UINT: |
| if (N->getValueType(0).bitsLT(N->getOperand(0)->getValueType(0))) |
| Res = SplitVecOp_TruncateHelper(N); |
| else |
| Res = SplitVecOp_UnaryOp(N); |
| break; |
| case ISD::SINT_TO_FP: |
| case ISD::UINT_TO_FP: |
| if (N->getValueType(0).bitsLT(N->getOperand(0)->getValueType(0))) |
| Res = SplitVecOp_TruncateHelper(N); |
| else |
| Res = SplitVecOp_UnaryOp(N); |
| break; |
| case ISD::CTTZ: |
| case ISD::CTLZ: |
| case ISD::CTPOP: |
| case ISD::FP_EXTEND: |
| case ISD::SIGN_EXTEND: |
| case ISD::ZERO_EXTEND: |
| case ISD::ANY_EXTEND: |
| case ISD::FTRUNC: |
| case ISD::FCANONICALIZE: |
| Res = SplitVecOp_UnaryOp(N); |
| break; |
| |
| case ISD::ANY_EXTEND_VECTOR_INREG: |
| case ISD::SIGN_EXTEND_VECTOR_INREG: |
| case ISD::ZERO_EXTEND_VECTOR_INREG: |
| Res = SplitVecOp_ExtVecInRegOp(N); |
| break; |
| |
| case ISD::VECREDUCE_FADD: |
| case ISD::VECREDUCE_FMUL: |
| case ISD::VECREDUCE_ADD: |
| case ISD::VECREDUCE_MUL: |
| case ISD::VECREDUCE_AND: |
| case ISD::VECREDUCE_OR: |
| case ISD::VECREDUCE_XOR: |
| case ISD::VECREDUCE_SMAX: |
| case ISD::VECREDUCE_SMIN: |
| case ISD::VECREDUCE_UMAX: |
| case ISD::VECREDUCE_UMIN: |
| case ISD::VECREDUCE_FMAX: |
| case ISD::VECREDUCE_FMIN: |
| Res = SplitVecOp_VECREDUCE(N, OpNo); |
| break; |
| } |
| } |
| |
| // If the result is null, the sub-method took care of registering results etc. |
| if (!Res.getNode()) return false; |
| |
| // If the result is N, the sub-method updated N in place. Tell the legalizer |
| // core about this. |
| if (Res.getNode() == N) |
| return true; |
| |
| assert(Res.getValueType() == N->getValueType(0) && N->getNumValues() == 1 && |
| "Invalid operand expansion"); |
| |
| ReplaceValueWith(SDValue(N, 0), Res); |
| return false; |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_VSELECT(SDNode *N, unsigned OpNo) { |
| // The only possibility for an illegal operand is the mask, since result type |
| // legalization would have handled this node already otherwise. |
| assert(OpNo == 0 && "Illegal operand must be mask"); |
| |
| SDValue Mask = N->getOperand(0); |
| SDValue Src0 = N->getOperand(1); |
| SDValue Src1 = N->getOperand(2); |
| EVT Src0VT = Src0.getValueType(); |
| SDLoc DL(N); |
| assert(Mask.getValueType().isVector() && "VSELECT without a vector mask?"); |
| |
| SDValue Lo, Hi; |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| assert(Lo.getValueType() == Hi.getValueType() && |
| "Lo and Hi have differing types"); |
| |
| EVT LoOpVT, HiOpVT; |
| std::tie(LoOpVT, HiOpVT) = DAG.GetSplitDestVTs(Src0VT); |
| assert(LoOpVT == HiOpVT && "Asymmetric vector split?"); |
| |
| SDValue LoOp0, HiOp0, LoOp1, HiOp1, LoMask, HiMask; |
| std::tie(LoOp0, HiOp0) = DAG.SplitVector(Src0, DL); |
| std::tie(LoOp1, HiOp1) = DAG.SplitVector(Src1, DL); |
| std::tie(LoMask, HiMask) = DAG.SplitVector(Mask, DL); |
| |
| SDValue LoSelect = |
| DAG.getNode(ISD::VSELECT, DL, LoOpVT, LoMask, LoOp0, LoOp1); |
| SDValue HiSelect = |
| DAG.getNode(ISD::VSELECT, DL, HiOpVT, HiMask, HiOp0, HiOp1); |
| |
| return DAG.getNode(ISD::CONCAT_VECTORS, DL, Src0VT, LoSelect, HiSelect); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_VECREDUCE(SDNode *N, unsigned OpNo) { |
| EVT ResVT = N->getValueType(0); |
| SDValue Lo, Hi; |
| SDLoc dl(N); |
| |
| SDValue VecOp = N->getOperand(OpNo); |
| EVT VecVT = VecOp.getValueType(); |
| assert(VecVT.isVector() && "Can only split reduce vector operand"); |
| GetSplitVector(VecOp, Lo, Hi); |
| EVT LoOpVT, HiOpVT; |
| std::tie(LoOpVT, HiOpVT) = DAG.GetSplitDestVTs(VecVT); |
| |
| bool NoNaN = N->getFlags().hasNoNaNs(); |
| unsigned CombineOpc = 0; |
| switch (N->getOpcode()) { |
| case ISD::VECREDUCE_FADD: CombineOpc = ISD::FADD; break; |
| case ISD::VECREDUCE_FMUL: CombineOpc = ISD::FMUL; break; |
| case ISD::VECREDUCE_ADD: CombineOpc = ISD::ADD; break; |
| case ISD::VECREDUCE_MUL: CombineOpc = ISD::MUL; break; |
| case ISD::VECREDUCE_AND: CombineOpc = ISD::AND; break; |
| case ISD::VECREDUCE_OR: CombineOpc = ISD::OR; break; |
| case ISD::VECREDUCE_XOR: CombineOpc = ISD::XOR; break; |
| case ISD::VECREDUCE_SMAX: CombineOpc = ISD::SMAX; break; |
| case ISD::VECREDUCE_SMIN: CombineOpc = ISD::SMIN; break; |
| case ISD::VECREDUCE_UMAX: CombineOpc = ISD::UMAX; break; |
| case ISD::VECREDUCE_UMIN: CombineOpc = ISD::UMIN; break; |
| case ISD::VECREDUCE_FMAX: |
| CombineOpc = NoNaN ? ISD::FMAXNUM : ISD::FMAXNAN; |
| break; |
| case ISD::VECREDUCE_FMIN: |
| CombineOpc = NoNaN ? ISD::FMINNUM : ISD::FMINNAN; |
| break; |
| default: |
| llvm_unreachable("Unexpected reduce ISD node"); |
| } |
| |
| // Use the appropriate scalar instruction on the split subvectors before |
| // reducing the now partially reduced smaller vector. |
| SDValue Partial = DAG.getNode(CombineOpc, dl, LoOpVT, Lo, Hi); |
| return DAG.getNode(N->getOpcode(), dl, ResVT, Partial); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_UnaryOp(SDNode *N) { |
| // The result has a legal vector type, but the input needs splitting. |
| EVT ResVT = N->getValueType(0); |
| SDValue Lo, Hi; |
| SDLoc dl(N); |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| EVT InVT = Lo.getValueType(); |
| |
| EVT OutVT = EVT::getVectorVT(*DAG.getContext(), ResVT.getVectorElementType(), |
| InVT.getVectorNumElements()); |
| |
| Lo = DAG.getNode(N->getOpcode(), dl, OutVT, Lo); |
| Hi = DAG.getNode(N->getOpcode(), dl, OutVT, Hi); |
| |
| return DAG.getNode(ISD::CONCAT_VECTORS, dl, ResVT, Lo, Hi); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_BITCAST(SDNode *N) { |
| // For example, i64 = BITCAST v4i16 on alpha. Typically the vector will |
| // end up being split all the way down to individual components. Convert the |
| // split pieces into integers and reassemble. |
| SDValue Lo, Hi; |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| Lo = BitConvertToInteger(Lo); |
| Hi = BitConvertToInteger(Hi); |
| |
| if (DAG.getDataLayout().isBigEndian()) |
| std::swap(Lo, Hi); |
| |
| return DAG.getNode(ISD::BITCAST, SDLoc(N), N->getValueType(0), |
| JoinIntegers(Lo, Hi)); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_EXTRACT_SUBVECTOR(SDNode *N) { |
| // We know that the extracted result type is legal. |
| EVT SubVT = N->getValueType(0); |
| SDValue Idx = N->getOperand(1); |
| SDLoc dl(N); |
| SDValue Lo, Hi; |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| |
| uint64_t LoElts = Lo.getValueType().getVectorNumElements(); |
| uint64_t IdxVal = cast<ConstantSDNode>(Idx)->getZExtValue(); |
| |
| if (IdxVal < LoElts) { |
| assert(IdxVal + SubVT.getVectorNumElements() <= LoElts && |
| "Extracted subvector crosses vector split!"); |
| return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, SubVT, Lo, Idx); |
| } else { |
| return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, SubVT, Hi, |
| DAG.getConstant(IdxVal - LoElts, dl, |
| Idx.getValueType())); |
| } |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_EXTRACT_VECTOR_ELT(SDNode *N) { |
| SDValue Vec = N->getOperand(0); |
| SDValue Idx = N->getOperand(1); |
| EVT VecVT = Vec.getValueType(); |
| |
| if (isa<ConstantSDNode>(Idx)) { |
| uint64_t IdxVal = cast<ConstantSDNode>(Idx)->getZExtValue(); |
| assert(IdxVal < VecVT.getVectorNumElements() && "Invalid vector index!"); |
| |
| SDValue Lo, Hi; |
| GetSplitVector(Vec, Lo, Hi); |
| |
| uint64_t LoElts = Lo.getValueType().getVectorNumElements(); |
| |
| if (IdxVal < LoElts) |
| return SDValue(DAG.UpdateNodeOperands(N, Lo, Idx), 0); |
| return SDValue(DAG.UpdateNodeOperands(N, Hi, |
| DAG.getConstant(IdxVal - LoElts, SDLoc(N), |
| Idx.getValueType())), 0); |
| } |
| |
| // See if the target wants to custom expand this node. |
| if (CustomLowerNode(N, N->getValueType(0), true)) |
| return SDValue(); |
| |
| // Make the vector elements byte-addressable if they aren't already. |
| SDLoc dl(N); |
| EVT EltVT = VecVT.getVectorElementType(); |
| if (EltVT.getSizeInBits() < 8) { |
| SmallVector<SDValue, 4> ElementOps; |
| for (unsigned i = 0; i < VecVT.getVectorNumElements(); ++i) { |
| ElementOps.push_back(DAG.getAnyExtOrTrunc( |
| DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, EltVT, Vec, |
| DAG.getConstant(i, dl, MVT::i8)), |
| dl, MVT::i8)); |
| } |
| |
| EltVT = MVT::i8; |
| VecVT = EVT::getVectorVT(*DAG.getContext(), EltVT, |
| VecVT.getVectorNumElements()); |
| Vec = DAG.getBuildVector(VecVT, dl, ElementOps); |
| } |
| |
| // Store the vector to the stack. |
| SDValue StackPtr = DAG.CreateStackTemporary(VecVT); |
| SDValue Store = |
| DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, MachinePointerInfo()); |
| |
| // Load back the required element. |
| StackPtr = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Idx); |
| return DAG.getExtLoad(ISD::EXTLOAD, dl, N->getValueType(0), Store, StackPtr, |
| MachinePointerInfo(), EltVT); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_ExtVecInRegOp(SDNode *N) { |
| SDValue Lo, Hi; |
| |
| // *_EXTEND_VECTOR_INREG only reference the lower half of the input, so |
| // splitting the result has the same effect as splitting the input operand. |
| SplitVecRes_ExtVecInRegOp(N, Lo, Hi); |
| |
| return DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(N), N->getValueType(0), Lo, Hi); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_MGATHER(MaskedGatherSDNode *MGT, |
| unsigned OpNo) { |
| EVT LoVT, HiVT; |
| SDLoc dl(MGT); |
| std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(MGT->getValueType(0)); |
| |
| SDValue Ch = MGT->getChain(); |
| SDValue Ptr = MGT->getBasePtr(); |
| SDValue Index = MGT->getIndex(); |
| SDValue Mask = MGT->getMask(); |
| SDValue Src0 = MGT->getValue(); |
| unsigned Alignment = MGT->getOriginalAlignment(); |
| |
| SDValue MaskLo, MaskHi; |
| if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector) |
| // Split Mask operand |
| GetSplitVector(Mask, MaskLo, MaskHi); |
| else |
| std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl); |
| |
| EVT MemoryVT = MGT->getMemoryVT(); |
| EVT LoMemVT, HiMemVT; |
| std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT); |
| |
| SDValue Src0Lo, Src0Hi; |
| if (getTypeAction(Src0.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Src0, Src0Lo, Src0Hi); |
| else |
| std::tie(Src0Lo, Src0Hi) = DAG.SplitVector(Src0, dl); |
| |
| SDValue IndexHi, IndexLo; |
| if (getTypeAction(Index.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Index, IndexLo, IndexHi); |
| else |
| std::tie(IndexLo, IndexHi) = DAG.SplitVector(Index, dl); |
| |
| MachineMemOperand *MMO = DAG.getMachineFunction(). |
| getMachineMemOperand(MGT->getPointerInfo(), |
| MachineMemOperand::MOLoad, LoMemVT.getStoreSize(), |
| Alignment, MGT->getAAInfo(), MGT->getRanges()); |
| |
| SDValue OpsLo[] = {Ch, Src0Lo, MaskLo, Ptr, IndexLo}; |
| SDValue Lo = DAG.getMaskedGather(DAG.getVTList(LoVT, MVT::Other), LoVT, dl, |
| OpsLo, MMO); |
| |
| MMO = DAG.getMachineFunction(). |
| getMachineMemOperand(MGT->getPointerInfo(), |
| MachineMemOperand::MOLoad, HiMemVT.getStoreSize(), |
| Alignment, MGT->getAAInfo(), |
| MGT->getRanges()); |
| |
| SDValue OpsHi[] = {Ch, Src0Hi, MaskHi, Ptr, IndexHi}; |
| SDValue Hi = DAG.getMaskedGather(DAG.getVTList(HiVT, MVT::Other), HiVT, dl, |
| OpsHi, MMO); |
| |
| // 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)); |
| |
| // Legalize the chain result - switch anything that used the old chain to |
| // use the new one. |
| ReplaceValueWith(SDValue(MGT, 1), Ch); |
| |
| SDValue Res = DAG.getNode(ISD::CONCAT_VECTORS, dl, MGT->getValueType(0), Lo, |
| Hi); |
| ReplaceValueWith(SDValue(MGT, 0), Res); |
| return SDValue(); |
| } |
| |
| 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; |
| if (getTypeAction(Data.getValueType()) == TargetLowering::TypeSplitVector) |
| // Split Data operand |
| GetSplitVector(Data, DataLo, DataHi); |
| else |
| std::tie(DataLo, DataHi) = DAG.SplitVector(Data, DL); |
| |
| SDValue MaskLo, MaskHi; |
| if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector) |
| // Split Mask operand |
| GetSplitVector(Mask, MaskLo, MaskHi); |
| else |
| std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, DL); |
| |
| MaskLo = PromoteTargetBoolean(MaskLo, DataLo.getValueType()); |
| MaskHi = PromoteTargetBoolean(MaskHi, DataHi.getValueType()); |
| |
| // 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(), |
| N->isCompressingStore()); |
| |
| Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, DL, LoMemVT, DAG, |
| N->isCompressingStore()); |
| 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(), N->isCompressingStore()); |
| |
| // 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_MSCATTER(MaskedScatterSDNode *N, |
| unsigned OpNo) { |
| SDValue Ch = N->getChain(); |
| SDValue Ptr = N->getBasePtr(); |
| SDValue Mask = N->getMask(); |
| SDValue Index = N->getIndex(); |
| SDValue Data = N->getValue(); |
| EVT MemoryVT = N->getMemoryVT(); |
| unsigned Alignment = N->getOriginalAlignment(); |
| SDLoc DL(N); |
| |
| // Split all operands |
| EVT LoMemVT, HiMemVT; |
| std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT); |
| |
| SDValue DataLo, DataHi; |
| if (getTypeAction(Data.getValueType()) == TargetLowering::TypeSplitVector) |
| // Split Data operand |
| GetSplitVector(Data, DataLo, DataHi); |
| else |
| std::tie(DataLo, DataHi) = DAG.SplitVector(Data, DL); |
| |
| SDValue MaskLo, MaskHi; |
| if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector) |
| // Split Mask operand |
| GetSplitVector(Mask, MaskLo, MaskHi); |
| else |
| std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, DL); |
| |
| SDValue IndexHi, IndexLo; |
| if (getTypeAction(Index.getValueType()) == TargetLowering::TypeSplitVector) |
| GetSplitVector(Index, IndexLo, IndexHi); |
| else |
| std::tie(IndexLo, IndexHi) = DAG.SplitVector(Index, DL); |
| |
| SDValue Lo; |
| MachineMemOperand *MMO = DAG.getMachineFunction(). |
| getMachineMemOperand(N->getPointerInfo(), |
| MachineMemOperand::MOStore, LoMemVT.getStoreSize(), |
| Alignment, N->getAAInfo(), N->getRanges()); |
| |
| SDValue OpsLo[] = {Ch, DataLo, MaskLo, Ptr, IndexLo}; |
| Lo = DAG.getMaskedScatter(DAG.getVTList(MVT::Other), DataLo.getValueType(), |
| DL, OpsLo, MMO); |
| |
| MMO = DAG.getMachineFunction(). |
| getMachineMemOperand(N->getPointerInfo(), |
| MachineMemOperand::MOStore, HiMemVT.getStoreSize(), |
| Alignment, N->getAAInfo(), N->getRanges()); |
| |
| // The order of the Scatter operation after split is well defined. The "Hi" |
| // part comes after the "Lo". So these two operations should be chained one |
| // after another. |
| SDValue OpsHi[] = {Lo, DataHi, MaskHi, Ptr, IndexHi}; |
| return DAG.getMaskedScatter(DAG.getVTList(MVT::Other), DataHi.getValueType(), |
| DL, OpsHi, MMO); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_STORE(StoreSDNode *N, unsigned OpNo) { |
| assert(N->isUnindexed() && "Indexed store of vector?"); |
| assert(OpNo == 1 && "Can only split the stored value"); |
| SDLoc DL(N); |
| |
| bool isTruncating = N->isTruncatingStore(); |
| SDValue Ch = N->getChain(); |
| SDValue Ptr = N->getBasePtr(); |
| EVT MemoryVT = N->getMemoryVT(); |
| unsigned Alignment = N->getOriginalAlignment(); |
| MachineMemOperand::Flags MMOFlags = N->getMemOperand()->getFlags(); |
| AAMDNodes AAInfo = N->getAAInfo(); |
| SDValue Lo, Hi; |
| GetSplitVector(N->getOperand(1), Lo, Hi); |
| |
| EVT LoMemVT, HiMemVT; |
| std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT); |
| |
| unsigned IncrementSize = LoMemVT.getSizeInBits()/8; |
| |
| if (isTruncating) |
| Lo = DAG.getTruncStore(Ch, DL, Lo, Ptr, N->getPointerInfo(), LoMemVT, |
| Alignment, MMOFlags, AAInfo); |
| else |
| Lo = DAG.getStore(Ch, DL, Lo, Ptr, N->getPointerInfo(), Alignment, MMOFlags, |
| AAInfo); |
| |
| // Increment the pointer to the other half. |
| Ptr = DAG.getNode(ISD::ADD, DL, Ptr.getValueType(), Ptr, |
| DAG.getConstant(IncrementSize, DL, Ptr.getValueType())); |
| |
| if (isTruncating) |
| Hi = DAG.getTruncStore(Ch, DL, Hi, Ptr, |
| N->getPointerInfo().getWithOffset(IncrementSize), |
| HiMemVT, Alignment, MMOFlags, AAInfo); |
| else |
| Hi = DAG.getStore(Ch, DL, Hi, Ptr, |
| N->getPointerInfo().getWithOffset(IncrementSize), |
| Alignment, MMOFlags, AAInfo); |
| |
| return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_CONCAT_VECTORS(SDNode *N) { |
| SDLoc DL(N); |
| |
| // The input operands all must have the same type, and we know the result |
| // type is valid. Convert this to a buildvector which extracts all the |
| // input elements. |
| // TODO: If the input elements are power-two vectors, we could convert this to |
| // a new CONCAT_VECTORS node with elements that are half-wide. |
| SmallVector<SDValue, 32> Elts; |
| EVT EltVT = N->getValueType(0).getVectorElementType(); |
| for (const SDValue &Op : N->op_values()) { |
| for (unsigned i = 0, e = Op.getValueType().getVectorNumElements(); |
| i != e; ++i) { |
| Elts.push_back(DAG.getNode( |
| ISD::EXTRACT_VECTOR_ELT, DL, EltVT, Op, |
| DAG.getConstant(i, DL, TLI.getVectorIdxTy(DAG.getDataLayout())))); |
| } |
| } |
| |
| return DAG.getBuildVector(N->getValueType(0), DL, Elts); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_TruncateHelper(SDNode *N) { |
| // The result type is legal, but the input type is illegal. If splitting |
| // ends up with the result type of each half still being legal, just |
| // do that. If, however, that would result in an illegal result type, |
| // we can try to get more clever with power-two vectors. Specifically, |
| // split the input type, but also widen the result element size, then |
| // concatenate the halves and truncate again. For example, consider a target |
| // where v8i8 is legal and v8i32 is not (ARM, which doesn't have 256-bit |
| // vectors). To perform a "%res = v8i8 trunc v8i32 %in" we do: |
| // %inlo = v4i32 extract_subvector %in, 0 |
| // %inhi = v4i32 extract_subvector %in, 4 |
| // %lo16 = v4i16 trunc v4i32 %inlo |
| // %hi16 = v4i16 trunc v4i32 %inhi |
| // %in16 = v8i16 concat_vectors v4i16 %lo16, v4i16 %hi16 |
| // %res = v8i8 trunc v8i16 %in16 |
| // |
| // Without this transform, the original truncate would end up being |
| // scalarized, which is pretty much always a last resort. |
| SDValue InVec = N->getOperand(0); |
| EVT InVT = InVec->getValueType(0); |
| EVT OutVT = N->getValueType(0); |
| unsigned NumElements = OutVT.getVectorNumElements(); |
| bool IsFloat = OutVT.isFloatingPoint(); |
| |
| // Widening should have already made sure this is a power-two vector |
| // if we're trying to split it at all. assert() that's true, just in case. |
| assert(!(NumElements & 1) && "Splitting vector, but not in half!"); |
| |
| unsigned InElementSize = InVT.getScalarSizeInBits(); |
| unsigned OutElementSize = OutVT.getScalarSizeInBits(); |
| |
| // If the input elements are only 1/2 the width of the result elements, |
| // just use the normal splitting. Our trick only work if there's room |
| // to split more than once. |
| if (InElementSize <= OutElementSize * 2) |
| return SplitVecOp_UnaryOp(N); |
| SDLoc DL(N); |
| |
| // Extract the halves of the input via extract_subvector. |
| SDValue InLoVec, InHiVec; |
| std::tie(InLoVec, InHiVec) = DAG.SplitVector(InVec, DL); |
| // Truncate them to 1/2 the element size. |
| EVT HalfElementVT = IsFloat ? |
| EVT::getFloatingPointVT(InElementSize/2) : |
| EVT::getIntegerVT(*DAG.getContext(), InElementSize/2); |
| EVT HalfVT = EVT::getVectorVT(*DAG.getContext(), HalfElementVT, |
| NumElements/2); |
| SDValue HalfLo = DAG.getNode(N->getOpcode(), DL, HalfVT, InLoVec); |
| SDValue HalfHi = DAG.getNode(N->getOpcode(), DL, HalfVT, InHiVec); |
| // Concatenate them to get the full intermediate truncation result. |
| EVT InterVT = EVT::getVectorVT(*DAG.getContext(), HalfElementVT, NumElements); |
| SDValue InterVec = DAG.getNode(ISD::CONCAT_VECTORS, DL, InterVT, HalfLo, |
| HalfHi); |
| // Now finish up by truncating all the way down to the original result |
| // type. This should normally be something that ends up being legal directly, |
| // but in theory if a target has very wide vectors and an annoyingly |
| // restricted set of legal types, this split can chain to build things up. |
| return IsFloat |
| ? DAG.getNode(ISD::FP_ROUND, DL, OutVT, InterVec, |
| DAG.getTargetConstant( |
| 0, DL, TLI.getPointerTy(DAG.getDataLayout()))) |
| : DAG.getNode(ISD::TRUNCATE, DL, OutVT, InterVec); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_VSETCC(SDNode *N) { |
| assert(N->getValueType(0).isVector() && |
| N->getOperand(0).getValueType().isVector() && |
| "Operand types must be vectors"); |
| // The result has a legal vector type, but the input needs splitting. |
| SDValue Lo0, Hi0, Lo1, Hi1, LoRes, HiRes; |
| SDLoc DL(N); |
| GetSplitVector(N->getOperand(0), Lo0, Hi0); |
| GetSplitVector(N->getOperand(1), Lo1, Hi1); |
| unsigned PartElements = Lo0.getValueType().getVectorNumElements(); |
| EVT PartResVT = EVT::getVectorVT(*DAG.getContext(), MVT::i1, PartElements); |
| EVT WideResVT = EVT::getVectorVT(*DAG.getContext(), MVT::i1, 2*PartElements); |
| |
| LoRes = DAG.getNode(ISD::SETCC, DL, PartResVT, Lo0, Lo1, N->getOperand(2)); |
| HiRes = DAG.getNode(ISD::SETCC, DL, PartResVT, Hi0, Hi1, N->getOperand(2)); |
| SDValue Con = DAG.getNode(ISD::CONCAT_VECTORS, DL, WideResVT, LoRes, HiRes); |
| return PromoteTargetBoolean(Con, N->getValueType(0)); |
| } |
| |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_FP_ROUND(SDNode *N) { |
| // The result has a legal vector type, but the input needs splitting. |
| EVT ResVT = N->getValueType(0); |
| SDValue Lo, Hi; |
| SDLoc DL(N); |
| GetSplitVector(N->getOperand(0), Lo, Hi); |
| EVT InVT = Lo.getValueType(); |
| |
| EVT OutVT = EVT::getVectorVT(*DAG.getContext(), ResVT.getVectorElementType(), |
| InVT.getVectorNumElements()); |
| |
| Lo = DAG.getNode(ISD::FP_ROUND, DL, OutVT, Lo, N->getOperand(1)); |
| Hi = DAG.getNode(ISD::FP_ROUND, DL, OutVT, Hi, N->getOperand(1)); |
| |
| return DAG.getNode(ISD::CONCAT_VECTORS, DL, ResVT, Lo, Hi); |
| } |
| |
| SDValue DAGTypeLegalizer::SplitVecOp_FCOPYSIGN(SDNode *N) { |
| // The result (and the first input) has a legal vector type, but the second |
| // input needs splitting. |
| return DAG.UnrollVectorOp(N, N->getValueType(0).getVectorNumElements()); |
| } |
| |
| |
| //===----------------------------------------------------------------------===// |
| // Result Vector Widening |
| //===----------------------------------------------------------------------===// |
| |
| void DAGTypeLegalizer::WidenVectorResult(SDNode *N, unsigned ResNo) { |
| DEBUG(dbgs() << "Widen node result " << ResNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"); |
| |
| // See if the target wants to custom widen this node. |
| if (CustomWidenLowerNode(N, N->getValueType(ResNo))) |
| return; |
| |
| SDValue Res = SDValue(); |
| switch (N->getOpcode()) { |
| default: |
| #ifndef NDEBUG |
| dbgs() << "WidenVectorResult #" << ResNo << ": "; |
| N->dump(&DAG); |
| dbgs() << "\n"; |
| #endif |
| llvm_unreachable("Do not know how to widen the result of this operator!"); |
| |
| case ISD::MERGE_VALUES: Res = WidenVecRes_MERGE_VALUES(N, ResNo); break; |
| case ISD::BITCAST: Res = WidenVecRes_BITCAST(N); break; |
| case ISD::BUILD_VECTOR: Res = WidenVecRes_BUILD_VECTOR(N); break; |
| case ISD::CONCAT_VECTORS: Res = WidenVecRes_CONCAT_VECTORS(N); break; |
| case ISD::EXTRACT_SUBVECTOR: Res = WidenVecRes_EXTRACT_SUBVECTOR(N); break; |
| case ISD::FP_ROUND_INREG: Res = WidenVecRes_InregOp(N); break; |
| case ISD::INSERT_VECTOR_ELT: Res = WidenVecRes_INSERT_VECTOR_ELT(N); break; |
| case ISD::LOAD: Res = WidenVecRes_LOAD(N); break; |
| case ISD::SCALAR_TO_VECTOR: Res = WidenVecRes_SCALAR_TO_VECTOR(N); break; |
| case ISD::SIGN_EXTEND_INREG: Res = WidenVecRes_InregOp(N); break; |
| case ISD::VSELECT: |
| case ISD::SELECT: Res = WidenVecRes_SELECT(N); break; |
| case ISD::SELECT_CC: Res = WidenVecRes_SELECT_CC(N); break; |
| case ISD::SETCC: Res = WidenVecRes_SETCC(N); break; |
| case ISD::UNDEF: Res = WidenVecRes_UNDEF(N); break; |
| case ISD::VECTOR_SHUFFLE: |
| Res = WidenVecRes_VECTOR_SHUFFLE(cast<ShuffleVectorSDNode>(N)); |
| break; |
| case ISD::MLOAD: |
| Res = WidenVecRes_MLOAD(cast<MaskedLoadSDNode>(N)); |
| break; |
| case ISD::MGATHER: |
| Res = WidenVecRes_MGATHER(cast<MaskedGatherSDNode>(N)); |
| break; |
| |
| case ISD::ADD: |
| case ISD::AND: |
| case ISD::MUL: |
| case ISD::MULHS: |
| case ISD::MULHU: |
| case ISD::OR: |
| case ISD::SUB: |
| case ISD::XOR: |
| case ISD::FMINNUM: |
| case ISD::FMAXNUM: |
| case ISD::FMINNAN: |
| case ISD::FMAXNAN: |
| case ISD::SMIN: |
| case ISD::SMAX: |
| case ISD::UMIN: |
| case ISD::UMAX: |
| Res = WidenVecRes_Binary(N); |
| break; |
| |
| case ISD::FADD: |
| case ISD::FMUL: |
| case ISD::FPOW: |
| case ISD::FSUB: |
| case ISD::FDIV: |
| case ISD::FREM: |
| case ISD::SDIV: |
| case ISD::UDIV: |
| case ISD::SREM: |
| case ISD::UREM: |
| Res = WidenVecRes_BinaryCanTrap(N); |
| break; |
| |
| case ISD::FCOPYSIGN: |
| Res = WidenVecRes_FCOPYSIGN(N); |
| break; |
| |
| case ISD::FPOWI: |
| Res = WidenVecRes_POWI(N); |
| break; |
| |
| case ISD::SHL: |
| case ISD::SRA: |
| case ISD::SRL: |
| Res = WidenVecRes_Shift(N); |
| break; |
| |
| case ISD::ANY_EXTEND_VECTOR_INREG: |
| case ISD::SIGN_EXTEND_VECTOR_INREG: |
| case ISD::ZERO_EXTEND_VECTOR_INREG: |
| Res = WidenVecRes_EXTEND_VECTOR_INREG(N); |
| break; |
| |
| case ISD::ANY_EXTEND: |
| case ISD::FP_EXTEND: |
| case ISD::FP_ROUND: |
| case ISD::FP_TO_SINT: |
| case ISD::FP_TO_UINT: |
| case ISD::SIGN_EXTEND: |
| case ISD::SINT_TO_FP: |
| case ISD::TRUNCATE: |
| case ISD::UINT_TO_FP: |
| case ISD::ZERO_EXTEND: |
| Res = WidenVecRes_Convert(N); |
| break; |
| |
| case ISD::BITREVERSE: |
|