//===-- SPUISelDAGToDAG.cpp - CellSPU pattern matching inst selector ------===// // // The LLVM Compiler Infrastructure // // This file is distributed under the University of Illinois Open Source // License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This file defines a pattern matching instruction selector for the Cell SPU, // converting from a legalized dag to a SPU-target dag. // //===----------------------------------------------------------------------===// #include "SPU.h" #include "SPUTargetMachine.h" #include "SPUISelLowering.h" #include "SPUHazardRecognizers.h" #include "SPUFrameInfo.h" #include "SPURegisterNames.h" #include "SPUTargetMachine.h" #include "llvm/CodeGen/MachineConstantPool.h" #include "llvm/CodeGen/MachineInstrBuilder.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/SelectionDAG.h" #include "llvm/CodeGen/SelectionDAGISel.h" #include "llvm/CodeGen/PseudoSourceValue.h" #include "llvm/Target/TargetOptions.h" #include "llvm/ADT/Statistic.h" #include "llvm/Constants.h" #include "llvm/GlobalValue.h" #include "llvm/Intrinsics.h" #include "llvm/LLVMContext.h" #include "llvm/Support/Debug.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/MathExtras.h" #include "llvm/Support/Compiler.h" #include "llvm/Support/raw_ostream.h" using namespace llvm; namespace { //! ConstantSDNode predicate for i32 sign-extended, 10-bit immediates bool isI64IntS10Immediate(ConstantSDNode *CN) { return isS10Constant(CN->getSExtValue()); } //! ConstantSDNode predicate for i32 sign-extended, 10-bit immediates bool isI32IntS10Immediate(ConstantSDNode *CN) { return isS10Constant(CN->getSExtValue()); } //! ConstantSDNode predicate for i32 unsigned 10-bit immediate values bool isI32IntU10Immediate(ConstantSDNode *CN) { return isU10Constant(CN->getSExtValue()); } //! ConstantSDNode predicate for i16 sign-extended, 10-bit immediate values bool isI16IntS10Immediate(ConstantSDNode *CN) { return isS10Constant(CN->getSExtValue()); } //! SDNode predicate for i16 sign-extended, 10-bit immediate values bool isI16IntS10Immediate(SDNode *N) { ConstantSDNode *CN = dyn_cast(N); return (CN != 0 && isI16IntS10Immediate(CN)); } //! ConstantSDNode predicate for i16 unsigned 10-bit immediate values bool isI16IntU10Immediate(ConstantSDNode *CN) { return isU10Constant((short) CN->getZExtValue()); } //! SDNode predicate for i16 sign-extended, 10-bit immediate values bool isI16IntU10Immediate(SDNode *N) { return (N->getOpcode() == ISD::Constant && isI16IntU10Immediate(cast(N))); } //! ConstantSDNode predicate for signed 16-bit values /*! \arg CN The constant SelectionDAG node holding the value \arg Imm The returned 16-bit value, if returning true This predicate tests the value in \a CN to see whether it can be represented as a 16-bit, sign-extended quantity. Returns true if this is the case. */ bool isIntS16Immediate(ConstantSDNode *CN, short &Imm) { MVT vt = CN->getValueType(0); Imm = (short) CN->getZExtValue(); if (vt.getSimpleVT() >= MVT::i1 && vt.getSimpleVT() <= MVT::i16) { return true; } else if (vt == MVT::i32) { int32_t i_val = (int32_t) CN->getZExtValue(); short s_val = (short) i_val; return i_val == s_val; } else { int64_t i_val = (int64_t) CN->getZExtValue(); short s_val = (short) i_val; return i_val == s_val; } return false; } //! SDNode predicate for signed 16-bit values. bool isIntS16Immediate(SDNode *N, short &Imm) { return (N->getOpcode() == ISD::Constant && isIntS16Immediate(cast(N), Imm)); } //! ConstantFPSDNode predicate for representing floats as 16-bit sign ext. static bool isFPS16Immediate(ConstantFPSDNode *FPN, short &Imm) { MVT vt = FPN->getValueType(0); if (vt == MVT::f32) { int val = FloatToBits(FPN->getValueAPF().convertToFloat()); int sval = (int) ((val << 16) >> 16); Imm = (short) val; return val == sval; } return false; } bool isHighLow(const SDValue &Op) { return (Op.getOpcode() == SPUISD::IndirectAddr && ((Op.getOperand(0).getOpcode() == SPUISD::Hi && Op.getOperand(1).getOpcode() == SPUISD::Lo) || (Op.getOperand(0).getOpcode() == SPUISD::Lo && Op.getOperand(1).getOpcode() == SPUISD::Hi))); } //===------------------------------------------------------------------===// //! MVT to "useful stuff" mapping structure: struct valtype_map_s { MVT VT; unsigned ldresult_ins; /// LDRESULT instruction (0 = undefined) bool ldresult_imm; /// LDRESULT instruction requires immediate? unsigned lrinst; /// LR instruction }; const valtype_map_s valtype_map[] = { { MVT::i8, SPU::ORBIr8, true, SPU::LRr8 }, { MVT::i16, SPU::ORHIr16, true, SPU::LRr16 }, { MVT::i32, SPU::ORIr32, true, SPU::LRr32 }, { MVT::i64, SPU::ORr64, false, SPU::LRr64 }, { MVT::f32, SPU::ORf32, false, SPU::LRf32 }, { MVT::f64, SPU::ORf64, false, SPU::LRf64 }, // vector types... (sigh!) { MVT::v16i8, 0, false, SPU::LRv16i8 }, { MVT::v8i16, 0, false, SPU::LRv8i16 }, { MVT::v4i32, 0, false, SPU::LRv4i32 }, { MVT::v2i64, 0, false, SPU::LRv2i64 }, { MVT::v4f32, 0, false, SPU::LRv4f32 }, { MVT::v2f64, 0, false, SPU::LRv2f64 } }; const size_t n_valtype_map = sizeof(valtype_map) / sizeof(valtype_map[0]); const valtype_map_s *getValueTypeMapEntry(MVT VT) { const valtype_map_s *retval = 0; for (size_t i = 0; i < n_valtype_map; ++i) { if (valtype_map[i].VT == VT) { retval = valtype_map + i; break; } } #ifndef NDEBUG if (retval == 0) { std::string msg; raw_string_ostream Msg(msg); Msg << "SPUISelDAGToDAG.cpp: getValueTypeMapEntry returns NULL for " << VT.getMVTString(); llvm_report_error(Msg.str()); } #endif return retval; } //! Generate the carry-generate shuffle mask. SDValue getCarryGenerateShufMask(SelectionDAG &DAG, DebugLoc dl) { SmallVector ShufBytes; // Create the shuffle mask for "rotating" the borrow up one register slot // once the borrow is generated. ShufBytes.push_back(DAG.getConstant(0x04050607, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0x80808080, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0x0c0d0e0f, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0x80808080, MVT::i32)); return DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, &ShufBytes[0], ShufBytes.size()); } //! Generate the borrow-generate shuffle mask SDValue getBorrowGenerateShufMask(SelectionDAG &DAG, DebugLoc dl) { SmallVector ShufBytes; // Create the shuffle mask for "rotating" the borrow up one register slot // once the borrow is generated. ShufBytes.push_back(DAG.getConstant(0x04050607, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0xc0c0c0c0, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0x0c0d0e0f, MVT::i32)); ShufBytes.push_back(DAG.getConstant(0xc0c0c0c0, MVT::i32)); return DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, &ShufBytes[0], ShufBytes.size()); } //===------------------------------------------------------------------===// /// SPUDAGToDAGISel - Cell SPU-specific code to select SPU machine /// instructions for SelectionDAG operations. /// class SPUDAGToDAGISel : public SelectionDAGISel { SPUTargetMachine &TM; SPUTargetLowering &SPUtli; unsigned GlobalBaseReg; public: explicit SPUDAGToDAGISel(SPUTargetMachine &tm) : SelectionDAGISel(tm), TM(tm), SPUtli(*tm.getTargetLowering()) { } virtual bool runOnMachineFunction(MachineFunction &MF) { // Make sure we re-emit a set of the global base reg if necessary GlobalBaseReg = 0; SelectionDAGISel::runOnMachineFunction(MF); return true; } /// getI32Imm - Return a target constant with the specified value, of type /// i32. inline SDValue getI32Imm(uint32_t Imm) { return CurDAG->getTargetConstant(Imm, MVT::i32); } /// getI64Imm - Return a target constant with the specified value, of type /// i64. inline SDValue getI64Imm(uint64_t Imm) { return CurDAG->getTargetConstant(Imm, MVT::i64); } /// getSmallIPtrImm - Return a target constant of pointer type. inline SDValue getSmallIPtrImm(unsigned Imm) { return CurDAG->getTargetConstant(Imm, SPUtli.getPointerTy()); } SDNode *emitBuildVector(SDValue build_vec) { MVT vecVT = build_vec.getValueType(); MVT eltVT = vecVT.getVectorElementType(); SDNode *bvNode = build_vec.getNode(); DebugLoc dl = bvNode->getDebugLoc(); // Check to see if this vector can be represented as a CellSPU immediate // constant by invoking all of the instruction selection predicates: if (((vecVT == MVT::v8i16) && (SPU::get_vec_i16imm(bvNode, *CurDAG, MVT::i16).getNode() != 0)) || ((vecVT == MVT::v4i32) && ((SPU::get_vec_i16imm(bvNode, *CurDAG, MVT::i32).getNode() != 0) || (SPU::get_ILHUvec_imm(bvNode, *CurDAG, MVT::i32).getNode() != 0) || (SPU::get_vec_u18imm(bvNode, *CurDAG, MVT::i32).getNode() != 0) || (SPU::get_v4i32_imm(bvNode, *CurDAG).getNode() != 0))) || ((vecVT == MVT::v2i64) && ((SPU::get_vec_i16imm(bvNode, *CurDAG, MVT::i64).getNode() != 0) || (SPU::get_ILHUvec_imm(bvNode, *CurDAG, MVT::i64).getNode() != 0) || (SPU::get_vec_u18imm(bvNode, *CurDAG, MVT::i64).getNode() != 0)))) return Select(build_vec); // No, need to emit a constant pool spill: std::vector CV; for (size_t i = 0; i < build_vec.getNumOperands(); ++i) { ConstantSDNode *V = dyn_cast (build_vec.getOperand(i)); CV.push_back(const_cast (V->getConstantIntValue())); } Constant *CP = ConstantVector::get(CV); SDValue CPIdx = CurDAG->getConstantPool(CP, SPUtli.getPointerTy()); unsigned Alignment = cast(CPIdx)->getAlignment(); SDValue CGPoolOffset = SPU::LowerConstantPool(CPIdx, *CurDAG, SPUtli.getSPUTargetMachine()); return SelectCode(CurDAG->getLoad(build_vec.getValueType(), dl, CurDAG->getEntryNode(), CGPoolOffset, PseudoSourceValue::getConstantPool(), 0, false, Alignment)); } /// Select - Convert the specified operand from a target-independent to a /// target-specific node if it hasn't already been changed. SDNode *Select(SDValue Op); //! Emit the instruction sequence for i64 shl SDNode *SelectSHLi64(SDValue &Op, MVT OpVT); //! Emit the instruction sequence for i64 srl SDNode *SelectSRLi64(SDValue &Op, MVT OpVT); //! Emit the instruction sequence for i64 sra SDNode *SelectSRAi64(SDValue &Op, MVT OpVT); //! Emit the necessary sequence for loading i64 constants: SDNode *SelectI64Constant(SDValue &Op, MVT OpVT, DebugLoc dl); //! Alternate instruction emit sequence for loading i64 constants SDNode *SelectI64Constant(uint64_t i64const, MVT OpVT, DebugLoc dl); //! Returns true if the address N is an A-form (local store) address bool SelectAFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index); //! D-form address predicate bool SelectDFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index); /// Alternate D-form address using i7 offset predicate bool SelectDForm2Addr(SDValue Op, SDValue N, SDValue &Disp, SDValue &Base); /// D-form address selection workhorse bool DFormAddressPredicate(SDValue Op, SDValue N, SDValue &Disp, SDValue &Base, int minOffset, int maxOffset); //! Address predicate if N can be expressed as an indexed [r+r] operation. bool SelectXFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index); /// SelectInlineAsmMemoryOperand - Implement addressing mode selection for /// inline asm expressions. virtual bool SelectInlineAsmMemoryOperand(const SDValue &Op, char ConstraintCode, std::vector &OutOps) { SDValue Op0, Op1; switch (ConstraintCode) { default: return true; case 'm': // memory if (!SelectDFormAddr(Op, Op, Op0, Op1) && !SelectAFormAddr(Op, Op, Op0, Op1)) SelectXFormAddr(Op, Op, Op0, Op1); break; case 'o': // offsetable if (!SelectDFormAddr(Op, Op, Op0, Op1) && !SelectAFormAddr(Op, Op, Op0, Op1)) { Op0 = Op; Op1 = getSmallIPtrImm(0); } break; case 'v': // not offsetable #if 1 llvm_unreachable("InlineAsmMemoryOperand 'v' constraint not handled."); #else SelectAddrIdxOnly(Op, Op, Op0, Op1); #endif break; } OutOps.push_back(Op0); OutOps.push_back(Op1); return false; } /// InstructionSelect - This callback is invoked by /// SelectionDAGISel when it has created a SelectionDAG for us to codegen. virtual void InstructionSelect(); virtual const char *getPassName() const { return "Cell SPU DAG->DAG Pattern Instruction Selection"; } /// CreateTargetHazardRecognizer - Return the hazard recognizer to use for /// this target when scheduling the DAG. virtual ScheduleHazardRecognizer *CreateTargetHazardRecognizer() { const TargetInstrInfo *II = TM.getInstrInfo(); assert(II && "No InstrInfo?"); return new SPUHazardRecognizer(*II); } // Include the pieces autogenerated from the target description. #include "SPUGenDAGISel.inc" }; } /// InstructionSelect - This callback is invoked by /// SelectionDAGISel when it has created a SelectionDAG for us to codegen. void SPUDAGToDAGISel::InstructionSelect() { DEBUG(BB->dump()); // Select target instructions for the DAG. SelectRoot(*CurDAG); CurDAG->RemoveDeadNodes(); } /*! \arg Op The ISD instruction operand \arg N The address to be tested \arg Base The base address \arg Index The base address index */ bool SPUDAGToDAGISel::SelectAFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index) { // These match the addr256k operand type: MVT OffsVT = MVT::i16; SDValue Zero = CurDAG->getTargetConstant(0, OffsVT); switch (N.getOpcode()) { case ISD::Constant: case ISD::ConstantPool: case ISD::GlobalAddress: llvm_report_error("SPU SelectAFormAddr: Constant/Pool/Global not lowered."); /*NOTREACHED*/ case ISD::TargetConstant: case ISD::TargetGlobalAddress: case ISD::TargetJumpTable: llvm_report_error("SPUSelectAFormAddr: Target Constant/Pool/Global " "not wrapped as A-form address."); /*NOTREACHED*/ case SPUISD::AFormAddr: // Just load from memory if there's only a single use of the location, // otherwise, this will get handled below with D-form offset addresses if (N.hasOneUse()) { SDValue Op0 = N.getOperand(0); switch (Op0.getOpcode()) { case ISD::TargetConstantPool: case ISD::TargetJumpTable: Base = Op0; Index = Zero; return true; case ISD::TargetGlobalAddress: { GlobalAddressSDNode *GSDN = cast(Op0); GlobalValue *GV = GSDN->getGlobal(); if (GV->getAlignment() == 16) { Base = Op0; Index = Zero; return true; } break; } } } break; } return false; } bool SPUDAGToDAGISel::SelectDForm2Addr(SDValue Op, SDValue N, SDValue &Disp, SDValue &Base) { const int minDForm2Offset = -(1 << 7); const int maxDForm2Offset = (1 << 7) - 1; return DFormAddressPredicate(Op, N, Disp, Base, minDForm2Offset, maxDForm2Offset); } /*! \arg Op The ISD instruction (ignored) \arg N The address to be tested \arg Base Base address register/pointer \arg Index Base address index Examine the input address by a base register plus a signed 10-bit displacement, [r+I10] (D-form address). \return true if \a N is a D-form address with \a Base and \a Index set to non-empty SDValue instances. */ bool SPUDAGToDAGISel::SelectDFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index) { return DFormAddressPredicate(Op, N, Base, Index, SPUFrameInfo::minFrameOffset(), SPUFrameInfo::maxFrameOffset()); } bool SPUDAGToDAGISel::DFormAddressPredicate(SDValue Op, SDValue N, SDValue &Base, SDValue &Index, int minOffset, int maxOffset) { unsigned Opc = N.getOpcode(); MVT PtrTy = SPUtli.getPointerTy(); if (Opc == ISD::FrameIndex) { // Stack frame index must be less than 512 (divided by 16): FrameIndexSDNode *FIN = dyn_cast(N); int FI = int(FIN->getIndex()); DEBUG(cerr << "SelectDFormAddr: ISD::FrameIndex = " << FI << "\n"); if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) { Base = CurDAG->getTargetConstant(0, PtrTy); Index = CurDAG->getTargetFrameIndex(FI, PtrTy); return true; } } else if (Opc == ISD::ADD) { // Generated by getelementptr const SDValue Op0 = N.getOperand(0); const SDValue Op1 = N.getOperand(1); if ((Op0.getOpcode() == SPUISD::Hi && Op1.getOpcode() == SPUISD::Lo) || (Op1.getOpcode() == SPUISD::Hi && Op0.getOpcode() == SPUISD::Lo)) { Base = CurDAG->getTargetConstant(0, PtrTy); Index = N; return true; } else if (Op1.getOpcode() == ISD::Constant || Op1.getOpcode() == ISD::TargetConstant) { ConstantSDNode *CN = dyn_cast(Op1); int32_t offset = int32_t(CN->getSExtValue()); if (Op0.getOpcode() == ISD::FrameIndex) { FrameIndexSDNode *FIN = dyn_cast(Op0); int FI = int(FIN->getIndex()); DEBUG(cerr << "SelectDFormAddr: ISD::ADD offset = " << offset << " frame index = " << FI << "\n"); if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) { Base = CurDAG->getTargetConstant(offset, PtrTy); Index = CurDAG->getTargetFrameIndex(FI, PtrTy); return true; } } else if (offset > minOffset && offset < maxOffset) { Base = CurDAG->getTargetConstant(offset, PtrTy); Index = Op0; return true; } } else if (Op0.getOpcode() == ISD::Constant || Op0.getOpcode() == ISD::TargetConstant) { ConstantSDNode *CN = dyn_cast(Op0); int32_t offset = int32_t(CN->getSExtValue()); if (Op1.getOpcode() == ISD::FrameIndex) { FrameIndexSDNode *FIN = dyn_cast(Op1); int FI = int(FIN->getIndex()); DEBUG(cerr << "SelectDFormAddr: ISD::ADD offset = " << offset << " frame index = " << FI << "\n"); if (SPUFrameInfo::FItoStackOffset(FI) < maxOffset) { Base = CurDAG->getTargetConstant(offset, PtrTy); Index = CurDAG->getTargetFrameIndex(FI, PtrTy); return true; } } else if (offset > minOffset && offset < maxOffset) { Base = CurDAG->getTargetConstant(offset, PtrTy); Index = Op1; return true; } } } else if (Opc == SPUISD::IndirectAddr) { // Indirect with constant offset -> D-Form address const SDValue Op0 = N.getOperand(0); const SDValue Op1 = N.getOperand(1); if (Op0.getOpcode() == SPUISD::Hi && Op1.getOpcode() == SPUISD::Lo) { // (SPUindirect (SPUhi , 0), (SPUlo , 0)) Base = CurDAG->getTargetConstant(0, PtrTy); Index = N; return true; } else if (isa(Op0) || isa(Op1)) { int32_t offset = 0; SDValue idxOp; if (isa(Op1)) { ConstantSDNode *CN = cast(Op1); offset = int32_t(CN->getSExtValue()); idxOp = Op0; } else if (isa(Op0)) { ConstantSDNode *CN = cast(Op0); offset = int32_t(CN->getSExtValue()); idxOp = Op1; } if (offset >= minOffset && offset <= maxOffset) { Base = CurDAG->getTargetConstant(offset, PtrTy); Index = idxOp; return true; } } } else if (Opc == SPUISD::AFormAddr) { Base = CurDAG->getTargetConstant(0, N.getValueType()); Index = N; return true; } else if (Opc == SPUISD::LDRESULT) { Base = CurDAG->getTargetConstant(0, N.getValueType()); Index = N; return true; } else if (Opc == ISD::Register || Opc == ISD::CopyFromReg) { unsigned OpOpc = Op.getOpcode(); if (OpOpc == ISD::STORE || OpOpc == ISD::LOAD) { // Direct load/store without getelementptr SDValue Addr, Offs; // Get the register from CopyFromReg if (Opc == ISD::CopyFromReg) Addr = N.getOperand(1); else Addr = N; // Register Offs = ((OpOpc == ISD::STORE) ? Op.getOperand(3) : Op.getOperand(2)); if (Offs.getOpcode() == ISD::Constant || Offs.getOpcode() == ISD::UNDEF) { if (Offs.getOpcode() == ISD::UNDEF) Offs = CurDAG->getTargetConstant(0, Offs.getValueType()); Base = Offs; Index = Addr; return true; } } else { /* If otherwise unadorned, default to D-form address with 0 offset: */ if (Opc == ISD::CopyFromReg) { Index = N.getOperand(1); } else { Index = N; } Base = CurDAG->getTargetConstant(0, Index.getValueType()); return true; } } return false; } /*! \arg Op The ISD instruction operand \arg N The address operand \arg Base The base pointer operand \arg Index The offset/index operand If the address \a N can be expressed as an A-form or D-form address, returns false. Otherwise, creates two operands, Base and Index that will become the (r)(r) X-form address. */ bool SPUDAGToDAGISel::SelectXFormAddr(SDValue Op, SDValue N, SDValue &Base, SDValue &Index) { if (!SelectAFormAddr(Op, N, Base, Index) && !SelectDFormAddr(Op, N, Base, Index)) { // If the address is neither A-form or D-form, punt and use an X-form // address: Base = N.getOperand(1); Index = N.getOperand(0); return true; } return false; } //! Convert the operand from a target-independent to a target-specific node /*! */ SDNode * SPUDAGToDAGISel::Select(SDValue Op) { SDNode *N = Op.getNode(); unsigned Opc = N->getOpcode(); int n_ops = -1; unsigned NewOpc; MVT OpVT = Op.getValueType(); SDValue Ops[8]; DebugLoc dl = N->getDebugLoc(); if (N->isMachineOpcode()) { return NULL; // Already selected. } if (Opc == ISD::FrameIndex) { int FI = cast(N)->getIndex(); SDValue TFI = CurDAG->getTargetFrameIndex(FI, Op.getValueType()); SDValue Imm0 = CurDAG->getTargetConstant(0, Op.getValueType()); if (FI < 128) { NewOpc = SPU::AIr32; Ops[0] = TFI; Ops[1] = Imm0; n_ops = 2; } else { NewOpc = SPU::Ar32; Ops[0] = CurDAG->getRegister(SPU::R1, Op.getValueType()); Ops[1] = SDValue(CurDAG->getTargetNode(SPU::ILAr32, dl, Op.getValueType(), TFI, Imm0), 0); n_ops = 2; } } else if (Opc == ISD::Constant && OpVT == MVT::i64) { // Catch the i64 constants that end up here. Note: The backend doesn't // attempt to legalize the constant (it's useless because DAGCombiner // will insert 64-bit constants and we can't stop it). return SelectI64Constant(Op, OpVT, Op.getDebugLoc()); } else if ((Opc == ISD::ZERO_EXTEND || Opc == ISD::ANY_EXTEND) && OpVT == MVT::i64) { SDValue Op0 = Op.getOperand(0); MVT Op0VT = Op0.getValueType(); MVT Op0VecVT = MVT::getVectorVT(Op0VT, (128 / Op0VT.getSizeInBits())); MVT OpVecVT = MVT::getVectorVT(OpVT, (128 / OpVT.getSizeInBits())); SDValue shufMask; switch (Op0VT.getSimpleVT()) { default: llvm_report_error("CellSPU Select: Unhandled zero/any extend MVT"); /*NOTREACHED*/ case MVT::i32: shufMask = CurDAG->getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x00010203, MVT::i32), CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x08090a0b, MVT::i32)); break; case MVT::i16: shufMask = CurDAG->getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x80800203, MVT::i32), CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x80800a0b, MVT::i32)); break; case MVT::i8: shufMask = CurDAG->getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x80808003, MVT::i32), CurDAG->getConstant(0x80808080, MVT::i32), CurDAG->getConstant(0x8080800b, MVT::i32)); break; } SDNode *shufMaskLoad = emitBuildVector(shufMask); SDNode *PromoteScalar = SelectCode(CurDAG->getNode(SPUISD::PREFSLOT2VEC, dl, Op0VecVT, Op0)); SDValue zextShuffle = CurDAG->getNode(SPUISD::SHUFB, dl, OpVecVT, SDValue(PromoteScalar, 0), SDValue(PromoteScalar, 0), SDValue(shufMaskLoad, 0)); // N.B.: BIT_CONVERT replaces and updates the zextShuffle node, so we // re-use it in the VEC2PREFSLOT selection without needing to explicitly // call SelectCode (it's already done for us.) SelectCode(CurDAG->getNode(ISD::BIT_CONVERT, dl, OpVecVT, zextShuffle)); return SelectCode(CurDAG->getNode(SPUISD::VEC2PREFSLOT, dl, OpVT, zextShuffle)); } else if (Opc == ISD::ADD && (OpVT == MVT::i64 || OpVT == MVT::v2i64)) { SDNode *CGLoad = emitBuildVector(getCarryGenerateShufMask(*CurDAG, dl)); return SelectCode(CurDAG->getNode(SPUISD::ADD64_MARKER, dl, OpVT, Op.getOperand(0), Op.getOperand(1), SDValue(CGLoad, 0))); } else if (Opc == ISD::SUB && (OpVT == MVT::i64 || OpVT == MVT::v2i64)) { SDNode *CGLoad = emitBuildVector(getBorrowGenerateShufMask(*CurDAG, dl)); return SelectCode(CurDAG->getNode(SPUISD::SUB64_MARKER, dl, OpVT, Op.getOperand(0), Op.getOperand(1), SDValue(CGLoad, 0))); } else if (Opc == ISD::MUL && (OpVT == MVT::i64 || OpVT == MVT::v2i64)) { SDNode *CGLoad = emitBuildVector(getCarryGenerateShufMask(*CurDAG, dl)); return SelectCode(CurDAG->getNode(SPUISD::MUL64_MARKER, dl, OpVT, Op.getOperand(0), Op.getOperand(1), SDValue(CGLoad, 0))); } else if (Opc == ISD::TRUNCATE) { SDValue Op0 = Op.getOperand(0); if ((Op0.getOpcode() == ISD::SRA || Op0.getOpcode() == ISD::SRL) && OpVT == MVT::i32 && Op0.getValueType() == MVT::i64) { // Catch (truncate:i32 ([sra|srl]:i64 arg, c), where c >= 32 // // Take advantage of the fact that the upper 32 bits are in the // i32 preferred slot and avoid shuffle gymnastics: ConstantSDNode *CN = dyn_cast(Op0.getOperand(1)); if (CN != 0) { unsigned shift_amt = unsigned(CN->getZExtValue()); if (shift_amt >= 32) { SDNode *hi32 = CurDAG->getTargetNode(SPU::ORr32_r64, dl, OpVT, Op0.getOperand(0)); shift_amt -= 32; if (shift_amt > 0) { // Take care of the additional shift, if present: SDValue shift = CurDAG->getTargetConstant(shift_amt, MVT::i32); unsigned Opc = SPU::ROTMAIr32_i32; if (Op0.getOpcode() == ISD::SRL) Opc = SPU::ROTMr32; hi32 = CurDAG->getTargetNode(Opc, dl, OpVT, SDValue(hi32, 0), shift); } return hi32; } } } } else if (Opc == ISD::SHL) { if (OpVT == MVT::i64) { return SelectSHLi64(Op, OpVT); } } else if (Opc == ISD::SRL) { if (OpVT == MVT::i64) { return SelectSRLi64(Op, OpVT); } } else if (Opc == ISD::SRA) { if (OpVT == MVT::i64) { return SelectSRAi64(Op, OpVT); } } else if (Opc == ISD::FNEG && (OpVT == MVT::f64 || OpVT == MVT::v2f64)) { DebugLoc dl = Op.getDebugLoc(); // Check if the pattern is a special form of DFNMS: // (fneg (fsub (fmul R64FP:$rA, R64FP:$rB), R64FP:$rC)) SDValue Op0 = Op.getOperand(0); if (Op0.getOpcode() == ISD::FSUB) { SDValue Op00 = Op0.getOperand(0); if (Op00.getOpcode() == ISD::FMUL) { unsigned Opc = SPU::DFNMSf64; if (OpVT == MVT::v2f64) Opc = SPU::DFNMSv2f64; return CurDAG->getTargetNode(Opc, dl, OpVT, Op00.getOperand(0), Op00.getOperand(1), Op0.getOperand(1)); } } SDValue negConst = CurDAG->getConstant(0x8000000000000000ULL, MVT::i64); SDNode *signMask = 0; unsigned Opc = SPU::XORfneg64; if (OpVT == MVT::f64) { signMask = SelectI64Constant(negConst, MVT::i64, dl); } else if (OpVT == MVT::v2f64) { Opc = SPU::XORfnegvec; signMask = emitBuildVector(CurDAG->getNode(ISD::BUILD_VECTOR, dl, MVT::v2i64, negConst, negConst)); } return CurDAG->getTargetNode(Opc, dl, OpVT, Op.getOperand(0), SDValue(signMask, 0)); } else if (Opc == ISD::FABS) { if (OpVT == MVT::f64) { SDNode *signMask = SelectI64Constant(0x7fffffffffffffffULL, MVT::i64, dl); return CurDAG->getTargetNode(SPU::ANDfabs64, dl, OpVT, Op.getOperand(0), SDValue(signMask, 0)); } else if (OpVT == MVT::v2f64) { SDValue absConst = CurDAG->getConstant(0x7fffffffffffffffULL, MVT::i64); SDValue absVec = CurDAG->getNode(ISD::BUILD_VECTOR, dl, MVT::v2i64, absConst, absConst); SDNode *signMask = emitBuildVector(absVec); return CurDAG->getTargetNode(SPU::ANDfabsvec, dl, OpVT, Op.getOperand(0), SDValue(signMask, 0)); } } else if (Opc == SPUISD::LDRESULT) { // Custom select instructions for LDRESULT MVT VT = N->getValueType(0); SDValue Arg = N->getOperand(0); SDValue Chain = N->getOperand(1); SDNode *Result; const valtype_map_s *vtm = getValueTypeMapEntry(VT); if (vtm->ldresult_ins == 0) { std::string msg; raw_string_ostream Msg(msg); Msg << "LDRESULT for unsupported type: " << VT.getMVTString(); llvm_report_error(Msg.str()); } Opc = vtm->ldresult_ins; if (vtm->ldresult_imm) { SDValue Zero = CurDAG->getTargetConstant(0, VT); Result = CurDAG->getTargetNode(Opc, dl, VT, MVT::Other, Arg, Zero, Chain); } else { Result = CurDAG->getTargetNode(Opc, dl, VT, MVT::Other, Arg, Arg, Chain); } return Result; } else if (Opc == SPUISD::IndirectAddr) { // Look at the operands: SelectCode() will catch the cases that aren't // specifically handled here. // // SPUInstrInfo catches the following patterns: // (SPUindirect (SPUhi ...), (SPUlo ...)) // (SPUindirect $sp, imm) MVT VT = Op.getValueType(); SDValue Op0 = N->getOperand(0); SDValue Op1 = N->getOperand(1); RegisterSDNode *RN; if ((Op0.getOpcode() != SPUISD::Hi && Op1.getOpcode() != SPUISD::Lo) || (Op0.getOpcode() == ISD::Register && ((RN = dyn_cast(Op0.getNode())) != 0 && RN->getReg() != SPU::R1))) { NewOpc = SPU::Ar32; if (Op1.getOpcode() == ISD::Constant) { ConstantSDNode *CN = cast(Op1); Op1 = CurDAG->getTargetConstant(CN->getSExtValue(), VT); NewOpc = (isI32IntS10Immediate(CN) ? SPU::AIr32 : SPU::Ar32); } Ops[0] = Op0; Ops[1] = Op1; n_ops = 2; } } if (n_ops > 0) { if (N->hasOneUse()) return CurDAG->SelectNodeTo(N, NewOpc, OpVT, Ops, n_ops); else return CurDAG->getTargetNode(NewOpc, dl, OpVT, Ops, n_ops); } else return SelectCode(Op); } /*! * Emit the instruction sequence for i64 left shifts. The basic algorithm * is to fill the bottom two word slots with zeros so that zeros are shifted * in as the entire quadword is shifted left. * * \note This code could also be used to implement v2i64 shl. * * @param Op The shl operand * @param OpVT Op's machine value value type (doesn't need to be passed, but * makes life easier.) * @return The SDNode with the entire instruction sequence */ SDNode * SPUDAGToDAGISel::SelectSHLi64(SDValue &Op, MVT OpVT) { SDValue Op0 = Op.getOperand(0); MVT VecVT = MVT::getVectorVT(OpVT, (128 / OpVT.getSizeInBits())); SDValue ShiftAmt = Op.getOperand(1); MVT ShiftAmtVT = ShiftAmt.getValueType(); SDNode *VecOp0, *SelMask, *ZeroFill, *Shift = 0; SDValue SelMaskVal; DebugLoc dl = Op.getDebugLoc(); VecOp0 = CurDAG->getTargetNode(SPU::ORv2i64_i64, dl, VecVT, Op0); SelMaskVal = CurDAG->getTargetConstant(0xff00ULL, MVT::i16); SelMask = CurDAG->getTargetNode(SPU::FSMBIv2i64, dl, VecVT, SelMaskVal); ZeroFill = CurDAG->getTargetNode(SPU::ILv2i64, dl, VecVT, CurDAG->getTargetConstant(0, OpVT)); VecOp0 = CurDAG->getTargetNode(SPU::SELBv2i64, dl, VecVT, SDValue(ZeroFill, 0), SDValue(VecOp0, 0), SDValue(SelMask, 0)); if (ConstantSDNode *CN = dyn_cast(ShiftAmt)) { unsigned bytes = unsigned(CN->getZExtValue()) >> 3; unsigned bits = unsigned(CN->getZExtValue()) & 7; if (bytes > 0) { Shift = CurDAG->getTargetNode(SPU::SHLQBYIv2i64, dl, VecVT, SDValue(VecOp0, 0), CurDAG->getTargetConstant(bytes, ShiftAmtVT)); } if (bits > 0) { Shift = CurDAG->getTargetNode(SPU::SHLQBIIv2i64, dl, VecVT, SDValue((Shift != 0 ? Shift : VecOp0), 0), CurDAG->getTargetConstant(bits, ShiftAmtVT)); } } else { SDNode *Bytes = CurDAG->getTargetNode(SPU::ROTMIr32, dl, ShiftAmtVT, ShiftAmt, CurDAG->getTargetConstant(3, ShiftAmtVT)); SDNode *Bits = CurDAG->getTargetNode(SPU::ANDIr32, dl, ShiftAmtVT, ShiftAmt, CurDAG->getTargetConstant(7, ShiftAmtVT)); Shift = CurDAG->getTargetNode(SPU::SHLQBYv2i64, dl, VecVT, SDValue(VecOp0, 0), SDValue(Bytes, 0)); Shift = CurDAG->getTargetNode(SPU::SHLQBIv2i64, dl, VecVT, SDValue(Shift, 0), SDValue(Bits, 0)); } return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(Shift, 0)); } /*! * Emit the instruction sequence for i64 logical right shifts. * * @param Op The shl operand * @param OpVT Op's machine value value type (doesn't need to be passed, but * makes life easier.) * @return The SDNode with the entire instruction sequence */ SDNode * SPUDAGToDAGISel::SelectSRLi64(SDValue &Op, MVT OpVT) { SDValue Op0 = Op.getOperand(0); MVT VecVT = MVT::getVectorVT(OpVT, (128 / OpVT.getSizeInBits())); SDValue ShiftAmt = Op.getOperand(1); MVT ShiftAmtVT = ShiftAmt.getValueType(); SDNode *VecOp0, *Shift = 0; DebugLoc dl = Op.getDebugLoc(); VecOp0 = CurDAG->getTargetNode(SPU::ORv2i64_i64, dl, VecVT, Op0); if (ConstantSDNode *CN = dyn_cast(ShiftAmt)) { unsigned bytes = unsigned(CN->getZExtValue()) >> 3; unsigned bits = unsigned(CN->getZExtValue()) & 7; if (bytes > 0) { Shift = CurDAG->getTargetNode(SPU::ROTQMBYIv2i64, dl, VecVT, SDValue(VecOp0, 0), CurDAG->getTargetConstant(bytes, ShiftAmtVT)); } if (bits > 0) { Shift = CurDAG->getTargetNode(SPU::ROTQMBIIv2i64, dl, VecVT, SDValue((Shift != 0 ? Shift : VecOp0), 0), CurDAG->getTargetConstant(bits, ShiftAmtVT)); } } else { SDNode *Bytes = CurDAG->getTargetNode(SPU::ROTMIr32, dl, ShiftAmtVT, ShiftAmt, CurDAG->getTargetConstant(3, ShiftAmtVT)); SDNode *Bits = CurDAG->getTargetNode(SPU::ANDIr32, dl, ShiftAmtVT, ShiftAmt, CurDAG->getTargetConstant(7, ShiftAmtVT)); // Ensure that the shift amounts are negated! Bytes = CurDAG->getTargetNode(SPU::SFIr32, dl, ShiftAmtVT, SDValue(Bytes, 0), CurDAG->getTargetConstant(0, ShiftAmtVT)); Bits = CurDAG->getTargetNode(SPU::SFIr32, dl, ShiftAmtVT, SDValue(Bits, 0), CurDAG->getTargetConstant(0, ShiftAmtVT)); Shift = CurDAG->getTargetNode(SPU::ROTQMBYv2i64, dl, VecVT, SDValue(VecOp0, 0), SDValue(Bytes, 0)); Shift = CurDAG->getTargetNode(SPU::ROTQMBIv2i64, dl, VecVT, SDValue(Shift, 0), SDValue(Bits, 0)); } return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(Shift, 0)); } /*! * Emit the instruction sequence for i64 arithmetic right shifts. * * @param Op The shl operand * @param OpVT Op's machine value value type (doesn't need to be passed, but * makes life easier.) * @return The SDNode with the entire instruction sequence */ SDNode * SPUDAGToDAGISel::SelectSRAi64(SDValue &Op, MVT OpVT) { // Promote Op0 to vector MVT VecVT = MVT::getVectorVT(OpVT, (128 / OpVT.getSizeInBits())); SDValue ShiftAmt = Op.getOperand(1); MVT ShiftAmtVT = ShiftAmt.getValueType(); DebugLoc dl = Op.getDebugLoc(); SDNode *VecOp0 = CurDAG->getTargetNode(SPU::ORv2i64_i64, dl, VecVT, Op.getOperand(0)); SDValue SignRotAmt = CurDAG->getTargetConstant(31, ShiftAmtVT); SDNode *SignRot = CurDAG->getTargetNode(SPU::ROTMAIv2i64_i32, dl, MVT::v2i64, SDValue(VecOp0, 0), SignRotAmt); SDNode *UpperHalfSign = CurDAG->getTargetNode(SPU::ORi32_v4i32, dl, MVT::i32, SDValue(SignRot, 0)); SDNode *UpperHalfSignMask = CurDAG->getTargetNode(SPU::FSM64r32, dl, VecVT, SDValue(UpperHalfSign, 0)); SDNode *UpperLowerMask = CurDAG->getTargetNode(SPU::FSMBIv2i64, dl, VecVT, CurDAG->getTargetConstant(0xff00ULL, MVT::i16)); SDNode *UpperLowerSelect = CurDAG->getTargetNode(SPU::SELBv2i64, dl, VecVT, SDValue(UpperHalfSignMask, 0), SDValue(VecOp0, 0), SDValue(UpperLowerMask, 0)); SDNode *Shift = 0; if (ConstantSDNode *CN = dyn_cast(ShiftAmt)) { unsigned bytes = unsigned(CN->getZExtValue()) >> 3; unsigned bits = unsigned(CN->getZExtValue()) & 7; if (bytes > 0) { bytes = 31 - bytes; Shift = CurDAG->getTargetNode(SPU::ROTQBYIv2i64, dl, VecVT, SDValue(UpperLowerSelect, 0), CurDAG->getTargetConstant(bytes, ShiftAmtVT)); } if (bits > 0) { bits = 8 - bits; Shift = CurDAG->getTargetNode(SPU::ROTQBIIv2i64, dl, VecVT, SDValue((Shift != 0 ? Shift : UpperLowerSelect), 0), CurDAG->getTargetConstant(bits, ShiftAmtVT)); } } else { SDNode *NegShift = CurDAG->getTargetNode(SPU::SFIr32, dl, ShiftAmtVT, ShiftAmt, CurDAG->getTargetConstant(0, ShiftAmtVT)); Shift = CurDAG->getTargetNode(SPU::ROTQBYBIv2i64_r32, dl, VecVT, SDValue(UpperLowerSelect, 0), SDValue(NegShift, 0)); Shift = CurDAG->getTargetNode(SPU::ROTQBIv2i64, dl, VecVT, SDValue(Shift, 0), SDValue(NegShift, 0)); } return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(Shift, 0)); } /*! Do the necessary magic necessary to load a i64 constant */ SDNode *SPUDAGToDAGISel::SelectI64Constant(SDValue& Op, MVT OpVT, DebugLoc dl) { ConstantSDNode *CN = cast(Op.getNode()); return SelectI64Constant(CN->getZExtValue(), OpVT, dl); } SDNode *SPUDAGToDAGISel::SelectI64Constant(uint64_t Value64, MVT OpVT, DebugLoc dl) { MVT OpVecVT = MVT::getVectorVT(OpVT, 2); SDValue i64vec = SPU::LowerV2I64Splat(OpVecVT, *CurDAG, Value64, dl); // Here's where it gets interesting, because we have to parse out the // subtree handed back in i64vec: if (i64vec.getOpcode() == ISD::BIT_CONVERT) { // The degenerate case where the upper and lower bits in the splat are // identical: SDValue Op0 = i64vec.getOperand(0); ReplaceUses(i64vec, Op0); return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(emitBuildVector(Op0), 0)); } else if (i64vec.getOpcode() == SPUISD::SHUFB) { SDValue lhs = i64vec.getOperand(0); SDValue rhs = i64vec.getOperand(1); SDValue shufmask = i64vec.getOperand(2); if (lhs.getOpcode() == ISD::BIT_CONVERT) { ReplaceUses(lhs, lhs.getOperand(0)); lhs = lhs.getOperand(0); } SDNode *lhsNode = (lhs.getNode()->isMachineOpcode() ? lhs.getNode() : emitBuildVector(lhs)); if (rhs.getOpcode() == ISD::BIT_CONVERT) { ReplaceUses(rhs, rhs.getOperand(0)); rhs = rhs.getOperand(0); } SDNode *rhsNode = (rhs.getNode()->isMachineOpcode() ? rhs.getNode() : emitBuildVector(rhs)); if (shufmask.getOpcode() == ISD::BIT_CONVERT) { ReplaceUses(shufmask, shufmask.getOperand(0)); shufmask = shufmask.getOperand(0); } SDNode *shufMaskNode = (shufmask.getNode()->isMachineOpcode() ? shufmask.getNode() : emitBuildVector(shufmask)); SDNode *shufNode = Select(CurDAG->getNode(SPUISD::SHUFB, dl, OpVecVT, SDValue(lhsNode, 0), SDValue(rhsNode, 0), SDValue(shufMaskNode, 0))); return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(shufNode, 0)); } else if (i64vec.getOpcode() == ISD::BUILD_VECTOR) { return CurDAG->getTargetNode(SPU::ORi64_v2i64, dl, OpVT, SDValue(emitBuildVector(i64vec), 0)); } else { llvm_report_error("SPUDAGToDAGISel::SelectI64Constant: Unhandled i64vec" "condition"); } } /// createSPUISelDag - This pass converts a legalized DAG into a /// SPU-specific DAG, ready for instruction scheduling. /// FunctionPass *llvm::createSPUISelDag(SPUTargetMachine &TM) { return new SPUDAGToDAGISel(TM); }