//===---- IA64ISelDAGToDAG.cpp - IA64 pattern matching inst selector ------===// // // The LLVM Compiler Infrastructure // // This file was developed by Duraid Madina and is distributed under // the University of Illinois Open Source License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // // This file defines a pattern matching instruction selector for IA64, // converting a legalized dag to an IA64 dag. // //===----------------------------------------------------------------------===// #include "IA64.h" #include "IA64TargetMachine.h" #include "IA64ISelLowering.h" #include "llvm/CodeGen/MachineInstrBuilder.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/SSARegMap.h" #include "llvm/CodeGen/SelectionDAG.h" #include "llvm/CodeGen/SelectionDAGISel.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/Support/Debug.h" #include "llvm/Support/MathExtras.h" #include #include using namespace llvm; namespace { Statistic<> FusedFP ("ia64-codegen", "Number of fused fp operations"); Statistic<> FrameOff("ia64-codegen", "Number of frame idx offsets collapsed"); //===--------------------------------------------------------------------===// /// IA64DAGToDAGISel - IA64 specific code to select IA64 machine /// instructions for SelectionDAG operations. /// class IA64DAGToDAGISel : public SelectionDAGISel { IA64TargetLowering IA64Lowering; unsigned GlobalBaseReg; public: IA64DAGToDAGISel(IA64TargetMachine &TM) : SelectionDAGISel(IA64Lowering), IA64Lowering(*TM.getTargetLowering()) {} virtual bool runOnFunction(Function &Fn) { // Make sure we re-emit a set of the global base reg if necessary GlobalBaseReg = 0; return SelectionDAGISel::runOnFunction(Fn); } /// getI64Imm - Return a target constant with the specified value, of type /// i64. inline SDOperand getI64Imm(uint64_t Imm) { return CurDAG->getTargetConstant(Imm, MVT::i64); } /// getGlobalBaseReg - insert code into the entry mbb to materialize the PIC /// base register. Return the virtual register that holds this value. // SDOperand getGlobalBaseReg(); TODO: hmm // Select - Convert the specified operand from a target-independent to a // target-specific node if it hasn't already been changed. void Select(SDOperand &Result, SDOperand N); SDNode *SelectIntImmediateExpr(SDOperand LHS, SDOperand RHS, unsigned OCHi, unsigned OCLo, bool IsArithmetic = false, bool Negate = false); SDNode *SelectBitfieldInsert(SDNode *N); /// SelectCC - Select a comparison of the specified values with the /// specified condition code, returning the CR# of the expression. SDOperand SelectCC(SDOperand LHS, SDOperand RHS, ISD::CondCode CC); /// SelectAddr - Given the specified address, return the two operands for a /// load/store instruction, and return true if it should be an indexed [r+r] /// operation. bool SelectAddr(SDOperand Addr, SDOperand &Op1, SDOperand &Op2); /// InstructionSelectBasicBlock - This callback is invoked by /// SelectionDAGISel when it has created a SelectionDAG for us to codegen. virtual void InstructionSelectBasicBlock(SelectionDAG &DAG); virtual const char *getPassName() const { return "IA64 (Itanium) DAG->DAG Instruction Selector"; } // Include the pieces autogenerated from the target description. #include "IA64GenDAGISel.inc" private: SDOperand SelectDIV(SDOperand Op); }; } /// InstructionSelectBasicBlock - This callback is invoked by /// SelectionDAGISel when it has created a SelectionDAG for us to codegen. void IA64DAGToDAGISel::InstructionSelectBasicBlock(SelectionDAG &DAG) { DEBUG(BB->dump()); // The selection process is inherently a bottom-up recursive process (users // select their uses before themselves). Given infinite stack space, we // could just start selecting on the root and traverse the whole graph. In // practice however, this causes us to run out of stack space on large basic // blocks. To avoid this problem, select the entry node, then all its uses, // iteratively instead of recursively. std::vector Worklist; Worklist.push_back(DAG.getEntryNode()); // Note that we can do this in the IA64 target (scanning forward across token // chain edges) because no nodes ever get folded across these edges. On a // target like X86 which supports load/modify/store operations, this would // have to be more careful. while (!Worklist.empty()) { SDOperand Node = Worklist.back(); Worklist.pop_back(); if ((Node.Val->getOpcode() >= ISD::BUILTIN_OP_END && Node.Val->getOpcode() < IA64ISD::FIRST_NUMBER) || CodeGenMap.count(Node)) continue; for (SDNode::use_iterator UI = Node.Val->use_begin(), E = Node.Val->use_end(); UI != E; ++UI) { // Scan the values. If this use has a value that is a token chain, add it // to the worklist. SDNode *User = *UI; for (unsigned i = 0, e = User->getNumValues(); i != e; ++i) if (User->getValueType(i) == MVT::Other) { Worklist.push_back(SDOperand(User, i)); break; } } // Finally, legalize this node. SDOperand Dummy; Select(Dummy, Node); } // Select target instructions for the DAG. DAG.setRoot(SelectRoot(DAG.getRoot())); assert(InFlightSet.empty() && "ISel InFlightSet has not been emptied!"); CodeGenMap.clear(); HandleMap.clear(); ReplaceMap.clear(); DAG.RemoveDeadNodes(); // Emit machine code to BB. ScheduleAndEmitDAG(DAG); } SDOperand IA64DAGToDAGISel::SelectDIV(SDOperand Op) { SDNode *N = Op.Val; SDOperand Chain, Tmp1, Tmp2; Select(Chain, N->getOperand(0)); Select(Tmp1, N->getOperand(0)); Select(Tmp2, N->getOperand(1)); bool isFP=false; if(MVT::isFloatingPoint(Tmp1.getValueType())) isFP=true; bool isModulus=false; // is it a division or a modulus? bool isSigned=false; switch(N->getOpcode()) { case ISD::FDIV: case ISD::SDIV: isModulus=false; isSigned=true; break; case ISD::UDIV: isModulus=false; isSigned=false; break; case ISD::FREM: case ISD::SREM: isModulus=true; isSigned=true; break; case ISD::UREM: isModulus=true; isSigned=false; break; } // TODO: check for integer divides by powers of 2 (or other simple patterns?) SDOperand TmpPR, TmpPR2; SDOperand TmpF1, TmpF2, TmpF3, TmpF4, TmpF5, TmpF6, TmpF7, TmpF8; SDOperand TmpF9, TmpF10,TmpF11,TmpF12,TmpF13,TmpF14,TmpF15; SDNode *Result; // we'll need copies of F0 and F1 SDOperand F0 = CurDAG->getRegister(IA64::F0, MVT::f64); SDOperand F1 = CurDAG->getRegister(IA64::F1, MVT::f64); // OK, emit some code: if(!isFP) { // first, load the inputs into FP regs. TmpF1 = SDOperand(CurDAG->getTargetNode(IA64::SETFSIG, MVT::f64, Tmp1), 0); Chain = TmpF1.getValue(1); TmpF2 = SDOperand(CurDAG->getTargetNode(IA64::SETFSIG, MVT::f64, Tmp2), 0); Chain = TmpF2.getValue(1); // next, convert the inputs to FP if(isSigned) { TmpF3 = SDOperand(CurDAG->getTargetNode(IA64::FCVTXF, MVT::f64, TmpF1), 0); Chain = TmpF3.getValue(1); TmpF4 = SDOperand(CurDAG->getTargetNode(IA64::FCVTXF, MVT::f64, TmpF2), 0); Chain = TmpF4.getValue(1); } else { // is unsigned TmpF3 = SDOperand(CurDAG->getTargetNode(IA64::FCVTXUFS1, MVT::f64, TmpF1), 0); Chain = TmpF3.getValue(1); TmpF4 = SDOperand(CurDAG->getTargetNode(IA64::FCVTXUFS1, MVT::f64, TmpF2), 0); Chain = TmpF4.getValue(1); } } else { // this is an FP divide/remainder, so we 'leak' some temp // regs and assign TmpF3=Tmp1, TmpF4=Tmp2 TmpF3=Tmp1; TmpF4=Tmp2; } // we start by computing an approximate reciprocal (good to 9 bits?) // note, this instruction writes _both_ TmpF5 (answer) and TmpPR (predicate) if(isFP) TmpF5 = SDOperand(CurDAG->getTargetNode(IA64::FRCPAS0, MVT::f64, MVT::i1, TmpF3, TmpF4), 0); else TmpF5 = SDOperand(CurDAG->getTargetNode(IA64::FRCPAS1, MVT::f64, MVT::i1, TmpF3, TmpF4), 0); TmpPR = TmpF5.getValue(1); Chain = TmpF5.getValue(2); SDOperand minusB; if(isModulus) { // for remainders, it'll be handy to have // copies of -input_b minusB = SDOperand(CurDAG->getTargetNode(IA64::SUB, MVT::i64, CurDAG->getRegister(IA64::r0, MVT::i64), Tmp2), 0); Chain = minusB.getValue(1); } SDOperand TmpE0, TmpY1, TmpE1, TmpY2; TmpE0 = SDOperand(CurDAG->getTargetNode(IA64::CFNMAS1, MVT::f64, TmpF4, TmpF5, F1, TmpPR), 0); Chain = TmpE0.getValue(1); TmpY1 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpF5, TmpE0, TmpF5, TmpPR), 0); Chain = TmpY1.getValue(1); TmpE1 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpE0, TmpE0, F0, TmpPR), 0); Chain = TmpE1.getValue(1); TmpY2 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpY1, TmpE1, TmpY1, TmpPR), 0); Chain = TmpY2.getValue(1); if(isFP) { // if this is an FP divide, we finish up here and exit early if(isModulus) assert(0 && "Sorry, try another FORTRAN compiler."); SDOperand TmpE2, TmpY3, TmpQ0, TmpR0; TmpE2 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpE1, TmpE1, F0, TmpPR), 0); Chain = TmpE2.getValue(1); TmpY3 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpY2, TmpE2, TmpY2, TmpPR), 0); Chain = TmpY3.getValue(1); TmpQ0 = SDOperand(CurDAG->getTargetNode(IA64::CFMADS1, MVT::f64, // double prec! Tmp1, TmpY3, F0, TmpPR), 0); Chain = TmpQ0.getValue(1); TmpR0 = SDOperand(CurDAG->getTargetNode(IA64::CFNMADS1, MVT::f64, // double prec! Tmp2, TmpQ0, Tmp1, TmpPR), 0); Chain = TmpR0.getValue(1); // we want Result to have the same target register as the frcpa, so // we two-address hack it. See the comment "for this to work..." on // page 48 of Intel application note #245415 Result = CurDAG->getTargetNode(IA64::TCFMADS0, MVT::f64, // d.p. s0 rndg! TmpF5, TmpY3, TmpR0, TmpQ0, TmpPR); Chain = SDOperand(Result, 1); return SDOperand(Result, 0); // XXX: early exit! } else { // this is *not* an FP divide, so there's a bit left to do: SDOperand TmpQ2, TmpR2, TmpQ3, TmpQ; TmpQ2 = SDOperand(CurDAG->getTargetNode(IA64::CFMAS1, MVT::f64, TmpF3, TmpY2, F0, TmpPR), 0); Chain = TmpQ2.getValue(1); TmpR2 = SDOperand(CurDAG->getTargetNode(IA64::CFNMAS1, MVT::f64, TmpF4, TmpQ2, TmpF3, TmpPR), 0); Chain = TmpR2.getValue(1); // we want TmpQ3 to have the same target register as the frcpa? maybe we // should two-address hack it. See the comment "for this to work..." on page // 48 of Intel application note #245415 TmpQ3 = SDOperand(CurDAG->getTargetNode(IA64::TCFMAS1, MVT::f64, TmpF5, TmpR2, TmpY2, TmpQ2, TmpPR), 0); Chain = TmpQ3.getValue(1); // STORY: without these two-address instructions (TCFMAS1 and TCFMADS0) // the FPSWA won't be able to help out in the case of large/tiny // arguments. Other fun bugs may also appear, e.g. 0/x = x, not 0. if(isSigned) TmpQ = SDOperand(CurDAG->getTargetNode(IA64::FCVTFXTRUNCS1, MVT::f64, TmpQ3), 0); else TmpQ = SDOperand(CurDAG->getTargetNode(IA64::FCVTFXUTRUNCS1, MVT::f64, TmpQ3), 0); Chain = TmpQ.getValue(1); if(isModulus) { SDOperand FPminusB = SDOperand(CurDAG->getTargetNode(IA64::SETFSIG, MVT::f64, minusB), 0); Chain = FPminusB.getValue(1); SDOperand Remainder = SDOperand(CurDAG->getTargetNode(IA64::XMAL, MVT::f64, TmpQ, FPminusB, TmpF1), 0); Chain = Remainder.getValue(1); Result = CurDAG->getTargetNode(IA64::GETFSIG, MVT::i64, Remainder); Chain = SDOperand(Result, 1); } else { // just an integer divide Result = CurDAG->getTargetNode(IA64::GETFSIG, MVT::i64, TmpQ); Chain = SDOperand(Result, 1); } return SDOperand(Result, 0); } // wasn't an FP divide } // Select - Convert the specified operand from a target-independent to a // target-specific node if it hasn't already been changed. void IA64DAGToDAGISel::Select(SDOperand &Result, SDOperand Op) { SDNode *N = Op.Val; if (N->getOpcode() >= ISD::BUILTIN_OP_END && N->getOpcode() < IA64ISD::FIRST_NUMBER) { Result = Op; return; // Already selected. } // If this has already been converted, use it. std::map::iterator CGMI = CodeGenMap.find(Op); if (CGMI != CodeGenMap.end()) { Result = CGMI->second; return; } switch (N->getOpcode()) { default: break; case IA64ISD::BRCALL: { // XXX: this is also a hack! SDOperand Chain; SDOperand InFlag; // Null incoming flag value. Select(Chain, N->getOperand(0)); if(N->getNumOperands()==3) // we have an incoming chain, callee and flag Select(InFlag, N->getOperand(2)); unsigned CallOpcode; SDOperand CallOperand; // if we can call directly, do so if (GlobalAddressSDNode *GASD = dyn_cast(N->getOperand(1))) { CallOpcode = IA64::BRCALL_IPREL_GA; CallOperand = CurDAG->getTargetGlobalAddress(GASD->getGlobal(), MVT::i64); } else if (ExternalSymbolSDNode *ESSDN = // FIXME: we currently NEED this // case for correctness, to avoid // "non-pic code with imm reloc.n // against dynamic symbol" errors dyn_cast(N->getOperand(1))) { CallOpcode = IA64::BRCALL_IPREL_ES; CallOperand = N->getOperand(1); } else { // otherwise we need to load the function descriptor, // load the branch target (function)'s entry point and GP, // branch (call) then restore the GP SDOperand FnDescriptor; Select(FnDescriptor, N->getOperand(1)); // load the branch target's entry point [mem] and // GP value [mem+8] SDOperand targetEntryPoint= SDOperand(CurDAG->getTargetNode(IA64::LD8, MVT::i64, FnDescriptor), 0); Chain = targetEntryPoint.getValue(1); SDOperand targetGPAddr= SDOperand(CurDAG->getTargetNode(IA64::ADDS, MVT::i64, FnDescriptor, CurDAG->getConstant(8, MVT::i64)), 0); Chain = targetGPAddr.getValue(1); SDOperand targetGP= SDOperand(CurDAG->getTargetNode(IA64::LD8, MVT::i64, targetGPAddr), 0); Chain = targetGP.getValue(1); Chain = CurDAG->getCopyToReg(Chain, IA64::r1, targetGP, InFlag); InFlag = Chain.getValue(1); Chain = CurDAG->getCopyToReg(Chain, IA64::B6, targetEntryPoint, InFlag); // FLAG these? InFlag = Chain.getValue(1); CallOperand = CurDAG->getRegister(IA64::B6, MVT::i64); CallOpcode = IA64::BRCALL_INDIRECT; } // Finally, once everything is setup, emit the call itself if(InFlag.Val) Chain = SDOperand(CurDAG->getTargetNode(CallOpcode, MVT::Other, MVT::Flag, CallOperand, InFlag), 0); else // there might be no arguments Chain = SDOperand(CurDAG->getTargetNode(CallOpcode, MVT::Other, MVT::Flag, CallOperand, Chain), 0); InFlag = Chain.getValue(1); std::vector CallResults; CallResults.push_back(Chain); CallResults.push_back(InFlag); for (unsigned i = 0, e = CallResults.size(); i != e; ++i) CodeGenMap[Op.getValue(i)] = CallResults[i]; Result = CallResults[Op.ResNo]; return; } case IA64ISD::GETFD: { SDOperand Input; Select(Input, N->getOperand(0)); Result = SDOperand(CurDAG->getTargetNode(IA64::GETFD, MVT::i64, Input), 0); CodeGenMap[Op] = Result; return; } case ISD::FDIV: case ISD::SDIV: case ISD::UDIV: case ISD::SREM: case ISD::UREM: Result = SelectDIV(Op); return; case ISD::TargetConstantFP: { SDOperand Chain = CurDAG->getEntryNode(); // this is a constant, so.. if (cast(N)->isExactlyValue(+0.0)) { Result = CurDAG->getCopyFromReg(Chain, IA64::F0, MVT::f64); } else if (cast(N)->isExactlyValue(+1.0)) { Result = CurDAG->getCopyFromReg(Chain, IA64::F1, MVT::f64); } else assert(0 && "Unexpected FP constant!"); return; } case ISD::FrameIndex: { // TODO: reduce creepyness int FI = cast(N)->getIndex(); if (N->hasOneUse()) Result = CurDAG->SelectNodeTo(N, IA64::MOV, MVT::i64, CurDAG->getTargetFrameIndex(FI, MVT::i64)); else Result = CodeGenMap[Op] = SDOperand(CurDAG->getTargetNode(IA64::MOV, MVT::i64, CurDAG->getTargetFrameIndex(FI, MVT::i64)), 0); return; } case ISD::ConstantPool: { // TODO: nuke the constant pool // (ia64 doesn't need one) ConstantPoolSDNode *CP = cast(N); Constant *C = CP->get(); SDOperand CPI = CurDAG->getTargetConstantPool(C, MVT::i64, CP->getAlignment()); Result = SDOperand(CurDAG->getTargetNode(IA64::ADDL_GA, MVT::i64, // ? CurDAG->getRegister(IA64::r1, MVT::i64), CPI), 0); return; } case ISD::GlobalAddress: { GlobalValue *GV = cast(N)->getGlobal(); SDOperand GA = CurDAG->getTargetGlobalAddress(GV, MVT::i64); SDOperand Tmp = SDOperand(CurDAG->getTargetNode(IA64::ADDL_GA, MVT::i64, CurDAG->getRegister(IA64::r1, MVT::i64), GA), 0); Result = SDOperand(CurDAG->getTargetNode(IA64::LD8, MVT::i64, Tmp), 0); return; } /* XXX case ISD::ExternalSymbol: { SDOperand EA = CurDAG->getTargetExternalSymbol(cast(N)->getSymbol(), MVT::i64); SDOperand Tmp = CurDAG->getTargetNode(IA64::ADDL_EA, MVT::i64, CurDAG->getRegister(IA64::r1, MVT::i64), EA); return CurDAG->getTargetNode(IA64::LD8, MVT::i64, Tmp); } */ case ISD::LOAD: case ISD::EXTLOAD: // FIXME: load -1, not 1, for bools? case ISD::ZEXTLOAD: { SDOperand Chain, Address; Select(Chain, N->getOperand(0)); Select(Address, N->getOperand(1)); MVT::ValueType TypeBeingLoaded = (N->getOpcode() == ISD::LOAD) ? N->getValueType(0) : cast(N->getOperand(3))->getVT(); unsigned Opc; switch (TypeBeingLoaded) { default: #ifndef NDEBUG N->dump(); #endif assert(0 && "Cannot load this type!"); case MVT::i1: { // this is a bool Opc = IA64::LD1; // first we load a byte, then compare for != 0 if(N->getValueType(0) == MVT::i1) { // XXX: early exit! Result = CurDAG->SelectNodeTo(N, IA64::CMPNE, MVT::i1, MVT::Other, SDOperand(CurDAG->getTargetNode(Opc, MVT::i64, Address), 0), CurDAG->getRegister(IA64::r0, MVT::i64), Chain).getValue(Op.ResNo); return; } /* otherwise, we want to load a bool into something bigger: LD1 will do that for us, so we just fall through */ } case MVT::i8: Opc = IA64::LD1; break; case MVT::i16: Opc = IA64::LD2; break; case MVT::i32: Opc = IA64::LD4; break; case MVT::i64: Opc = IA64::LD8; break; case MVT::f32: Opc = IA64::LDF4; break; case MVT::f64: Opc = IA64::LDF8; break; } // TODO: comment this Result = CurDAG->SelectNodeTo(N, Opc, N->getValueType(0), MVT::Other, Address, Chain).getValue(Op.ResNo); return; } case ISD::TRUNCSTORE: case ISD::STORE: { SDOperand Address, Chain; Select(Address, N->getOperand(2)); Select(Chain, N->getOperand(0)); unsigned Opc; if (N->getOpcode() == ISD::STORE) { switch (N->getOperand(1).getValueType()) { default: assert(0 && "unknown type in store"); case MVT::i1: { // this is a bool Opc = IA64::ST1; // we store either 0 or 1 as a byte // first load zero! SDOperand Initial = CurDAG->getCopyFromReg(Chain, IA64::r0, MVT::i64); Chain = Initial.getValue(1); // then load 1 into the same reg iff the predicate to store is 1 SDOperand Tmp; Select(Tmp, N->getOperand(1)); Tmp = SDOperand(CurDAG->getTargetNode(IA64::TPCADDS, MVT::i64, Initial, CurDAG->getConstant(1, MVT::i64), Tmp), 0); Result = CurDAG->SelectNodeTo(N, Opc, MVT::Other, Address, Tmp, Chain); return; } case MVT::i64: Opc = IA64::ST8; break; case MVT::f64: Opc = IA64::STF8; break; } } else { //ISD::TRUNCSTORE switch(cast(N->getOperand(4))->getVT()) { default: assert(0 && "unknown type in truncstore"); case MVT::i8: Opc = IA64::ST1; break; case MVT::i16: Opc = IA64::ST2; break; case MVT::i32: Opc = IA64::ST4; break; case MVT::f32: Opc = IA64::STF4; break; } } SDOperand N1, N2; Select(N1, N->getOperand(1)); Select(N2, N->getOperand(2)); Result = CurDAG->SelectNodeTo(N, Opc, MVT::Other, N2, N1, Chain); return; } case ISD::BRCOND: { SDOperand Chain, CC; Select(Chain, N->getOperand(0)); Select(CC, N->getOperand(1)); MachineBasicBlock *Dest = cast(N->getOperand(2))->getBasicBlock(); //FIXME - we do NOT need long branches all the time Result = CurDAG->SelectNodeTo(N, IA64::BRLCOND_NOTCALL, MVT::Other, CC, CurDAG->getBasicBlock(Dest), Chain); return; } case ISD::CALLSEQ_START: case ISD::CALLSEQ_END: { int64_t Amt = cast(N->getOperand(1))->getValue(); unsigned Opc = N->getOpcode() == ISD::CALLSEQ_START ? IA64::ADJUSTCALLSTACKDOWN : IA64::ADJUSTCALLSTACKUP; SDOperand N0; Select(N0, N->getOperand(0)); Result = CurDAG->SelectNodeTo(N, Opc, MVT::Other, getI64Imm(Amt), N0); return; } case ISD::BR: // FIXME: we don't need long branches all the time! SDOperand N0; Select(N0, N->getOperand(0)); Result = CurDAG->SelectNodeTo(N, IA64::BRL_NOTCALL, MVT::Other, N->getOperand(1), N0); return; } SelectCode(Result, Op); } /// createIA64DAGToDAGInstructionSelector - This pass converts a legalized DAG /// into an IA64-specific DAG, ready for instruction scheduling. /// FunctionPass *llvm::createIA64DAGToDAGInstructionSelector(IA64TargetMachine &TM) { return new IA64DAGToDAGISel(TM); }