X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FScheduleDAGInstrs.cpp;h=11b246a8de250e159ea17bf59f3fec821afc62fd;hb=40b7a9f16ac30c8e0ee70a14e518ebe7e25aa1da;hp=ae4b935d719a8218f24f33a3cde74d29c3841331;hpb=cf0db29df20d9c665da7e82bb261bdd7cf7f1b2b;p=oota-llvm.git diff --git a/lib/CodeGen/ScheduleDAGInstrs.cpp b/lib/CodeGen/ScheduleDAGInstrs.cpp index ae4b935d719..11b246a8de2 100644 --- a/lib/CodeGen/ScheduleDAGInstrs.cpp +++ b/lib/CodeGen/ScheduleDAGInstrs.cpp @@ -13,12 +13,12 @@ //===----------------------------------------------------------------------===// #include "llvm/CodeGen/ScheduleDAGInstrs.h" +#include "llvm/ADT/IntEqClasses.h" #include "llvm/ADT/MapVector.h" #include "llvm/ADT/SmallPtrSet.h" #include "llvm/ADT/SmallSet.h" #include "llvm/Analysis/AliasAnalysis.h" #include "llvm/Analysis/ValueTracking.h" -#include "llvm/CodeGen/LiveIntervalAnalysis.h" #include "llvm/CodeGen/MachineFunctionPass.h" #include "llvm/CodeGen/MachineFrameInfo.h" #include "llvm/CodeGen/MachineInstrBuilder.h" @@ -51,15 +51,11 @@ static cl::opt UseTBAA("use-tbaa-in-sched-mi", cl::Hidden, ScheduleDAGInstrs::ScheduleDAGInstrs(MachineFunction &mf, const MachineLoopInfo *mli, - bool IsPostRAFlag, bool RemoveKillFlags, - LiveIntervals *lis) - : ScheduleDAG(mf), MLI(mli), MFI(mf.getFrameInfo()), LIS(lis), - IsPostRA(IsPostRAFlag), RemoveKillFlags(RemoveKillFlags), - CanHandleTerminators(false), FirstDbgValue(nullptr) { - assert((IsPostRA || LIS) && "PreRA scheduling requires LiveIntervals"); + bool RemoveKillFlags) + : ScheduleDAG(mf), MLI(mli), MFI(mf.getFrameInfo()), + RemoveKillFlags(RemoveKillFlags), CanHandleTerminators(false), + TrackLaneMasks(false), FirstDbgValue(nullptr) { DbgValues.clear(); - assert(!(IsPostRA && MRI.getNumVirtRegs()) && - "Virtual registers must be removed prior to PostRA scheduling"); const TargetSubtargetInfo &ST = mf.getSubtarget(); SchedModel.init(ST.getSchedModel(), &ST, TII); @@ -230,11 +226,8 @@ void ScheduleDAGInstrs::addSchedBarrierDeps() { if (TRI->isPhysicalRegister(Reg)) Uses.insert(PhysRegSUOper(&ExitSU, -1, Reg)); - else { - assert(!IsPostRA && "Virtual register encountered after regalloc."); - if (MO.readsReg()) // ignore undef operands - addVRegUseDeps(&ExitSU, i); - } + else if (MO.readsReg()) // ignore undef operands + addVRegUseDeps(&ExitSU, i); } } else { // For others, e.g. fallthrough, conditional branch, assume the exit @@ -242,11 +235,9 @@ void ScheduleDAGInstrs::addSchedBarrierDeps() { assert(Uses.empty() && "Uses in set before adding deps?"); for (MachineBasicBlock::succ_iterator SI = BB->succ_begin(), SE = BB->succ_end(); SI != SE; ++SI) - for (MachineBasicBlock::livein_iterator I = (*SI)->livein_begin(), - E = (*SI)->livein_end(); I != E; ++I) { - unsigned Reg = *I; - if (!Uses.contains(Reg)) - Uses.insert(PhysRegSUOper(&ExitSU, -1, Reg)); + for (const auto &LI : (*SI)->liveins()) { + if (!Uses.contains(LI.PhysReg)) + Uses.insert(PhysRegSUOper(&ExitSU, -1, LI.PhysReg)); } } } @@ -371,6 +362,20 @@ void ScheduleDAGInstrs::addPhysRegDeps(SUnit *SU, unsigned OperIdx) { } } +LaneBitmask ScheduleDAGInstrs::getLaneMaskForMO(const MachineOperand &MO) const +{ + unsigned Reg = MO.getReg(); + // No point in tracking lanemasks if we don't have interesting subregisters. + const TargetRegisterClass &RC = *MRI.getRegClass(Reg); + if (!RC.HasDisjunctSubRegs) + return ~0u; + + unsigned SubReg = MO.getSubReg(); + if (SubReg == 0) + return RC.getLaneMask(); + return TRI->getSubRegIndexLaneMask(SubReg); +} + /// addVRegDefDeps - Add register output and data dependencies from this SUnit /// to instructions that occur later in the same scheduling region if they read /// from or write to the virtual register defined at OperIdx. @@ -378,35 +383,106 @@ void ScheduleDAGInstrs::addPhysRegDeps(SUnit *SU, unsigned OperIdx) { /// TODO: Hoist loop induction variable increments. This has to be /// reevaluated. Generally, IV scheduling should be done before coalescing. void ScheduleDAGInstrs::addVRegDefDeps(SUnit *SU, unsigned OperIdx) { - const MachineInstr *MI = SU->getInstr(); - unsigned Reg = MI->getOperand(OperIdx).getReg(); + MachineInstr *MI = SU->getInstr(); + MachineOperand &MO = MI->getOperand(OperIdx); + unsigned Reg = MO.getReg(); + + LaneBitmask DefLaneMask; + LaneBitmask KillLaneMask; + if (TrackLaneMasks) { + bool IsKill = MO.getSubReg() == 0 || MO.isUndef(); + DefLaneMask = getLaneMaskForMO(MO); + // If we have a flag, none of the lane values comes from an + // earlier instruction. + KillLaneMask = IsKill ? ~0u : DefLaneMask; + + // Clear undef flag, we'll re-add it later once we know which subregister + // Def is first. + MO.setIsUndef(false); + } else { + DefLaneMask = ~0u; + KillLaneMask = ~0u; + } - // Singly defined vregs do not have output/anti dependencies. - // The current operand is a def, so we have at least one. - // Check here if there are any others... + if (MO.isDead()) { + assert(CurrentVRegUses.find(Reg) == CurrentVRegUses.end() && + "Dead defs should have no uses"); + } else { + // Add data dependence to all uses we found so far. + const TargetSubtargetInfo &ST = MF.getSubtarget(); + for (VReg2SUnitOperIdxMultiMap::iterator I = CurrentVRegUses.find(Reg), + E = CurrentVRegUses.end(); I != E; /*empty*/) { + LaneBitmask LaneMask = I->LaneMask; + // Ignore uses of other lanes. + if ((LaneMask & KillLaneMask) == 0) { + ++I; + continue; + } + + if ((LaneMask & DefLaneMask) != 0) { + SUnit *UseSU = I->SU; + MachineInstr *Use = UseSU->getInstr(); + SDep Dep(SU, SDep::Data, Reg); + Dep.setLatency(SchedModel.computeOperandLatency(MI, OperIdx, Use, + I->OperandIndex)); + ST.adjustSchedDependency(SU, UseSU, Dep); + UseSU->addPred(Dep); + } + + LaneMask &= ~KillLaneMask; + // If we found a Def for all lanes of this use, remove it from the list. + if (LaneMask != 0) { + I->LaneMask = LaneMask; + ++I; + } else + I = CurrentVRegUses.erase(I); + } + } + + // Shortcut: Singly defined vregs do not have output/anti dependencies. if (MRI.hasOneDef(Reg)) return; - // Add output dependence to the next nearest def of this vreg. + // Add output dependence to the next nearest defs of this vreg. // // Unless this definition is dead, the output dependence should be // transitively redundant with antidependencies from this definition's // uses. We're conservative for now until we have a way to guarantee the uses // are not eliminated sometime during scheduling. The output dependence edge // is also useful if output latency exceeds def-use latency. - VReg2SUnitMap::iterator DefI = VRegDefs.find(Reg); - if (DefI == VRegDefs.end()) - VRegDefs.insert(VReg2SUnit(Reg, SU)); - else { - SUnit *DefSU = DefI->SU; - if (DefSU != SU && DefSU != &ExitSU) { - SDep Dep(SU, SDep::Output, Reg); - Dep.setLatency( - SchedModel.computeOutputLatency(MI, OperIdx, DefSU->getInstr())); - DefSU->addPred(Dep); - } - DefI->SU = SU; + LaneBitmask LaneMask = DefLaneMask; + for (VReg2SUnit &V2SU : make_range(CurrentVRegDefs.find(Reg), + CurrentVRegDefs.end())) { + // Ignore defs for other lanes. + if ((V2SU.LaneMask & LaneMask) == 0) + continue; + // Add an output dependence. + SUnit *DefSU = V2SU.SU; + // Ignore additional defs of the same lanes in one instruction. This can + // happen because lanemasks are shared for targets with too many + // subregisters. We also use some representration tricks/hacks where we + // add super-register defs/uses, to imply that although we only access parts + // of the reg we care about the full one. + if (DefSU == SU) + continue; + SDep Dep(SU, SDep::Output, Reg); + Dep.setLatency( + SchedModel.computeOutputLatency(MI, OperIdx, DefSU->getInstr())); + DefSU->addPred(Dep); + + // Update current definition. This can get tricky if the def was about a + // bigger lanemask before. We then have to shrink it and create a new + // VReg2SUnit for the non-overlapping part. + LaneBitmask OverlapMask = V2SU.LaneMask & LaneMask; + LaneBitmask NonOverlapMask = V2SU.LaneMask & ~LaneMask; + if (NonOverlapMask != 0) + CurrentVRegDefs.insert(VReg2SUnit(Reg, NonOverlapMask, V2SU.SU)); + V2SU.SU = SU; + V2SU.LaneMask = OverlapMask; } + // If there was no CurrentVRegDefs entry for some lanes yet, create one. + if (LaneMask != 0) + CurrentVRegDefs.insert(VReg2SUnit(Reg, LaneMask, SU)); } /// addVRegUseDeps - Add a register data dependency if the instruction that @@ -416,59 +492,34 @@ void ScheduleDAGInstrs::addVRegDefDeps(SUnit *SU, unsigned OperIdx) { /// /// TODO: Handle ExitSU "uses" properly. void ScheduleDAGInstrs::addVRegUseDeps(SUnit *SU, unsigned OperIdx) { - MachineInstr *MI = SU->getInstr(); - unsigned Reg = MI->getOperand(OperIdx).getReg(); + const MachineInstr *MI = SU->getInstr(); + const MachineOperand &MO = MI->getOperand(OperIdx); + unsigned Reg = MO.getReg(); + + // Remember the use. Data dependencies will be added when we find the def. + LaneBitmask LaneMask = TrackLaneMasks ? getLaneMaskForMO(MO) : ~0u; + CurrentVRegUses.insert(VReg2SUnitOperIdx(Reg, LaneMask, OperIdx, SU)); + + // Add antidependences to the following defs of the vreg. + for (VReg2SUnit &V2SU : make_range(CurrentVRegDefs.find(Reg), + CurrentVRegDefs.end())) { + // Ignore defs for unrelated lanes. + LaneBitmask PrevDefLaneMask = V2SU.LaneMask; + if ((PrevDefLaneMask & LaneMask) == 0) + continue; + if (V2SU.SU == SU) + continue; - // Record this local VReg use. - VReg2UseMap::iterator UI = VRegUses.find(Reg); - for (; UI != VRegUses.end(); ++UI) { - if (UI->SU == SU) - break; + V2SU.SU->addPred(SDep(SU, SDep::Anti, Reg)); } - if (UI == VRegUses.end()) - VRegUses.insert(VReg2SUnit(Reg, SU)); - - // Lookup this operand's reaching definition. - assert(LIS && "vreg dependencies requires LiveIntervals"); - LiveQueryResult LRQ - = LIS->getInterval(Reg).Query(LIS->getInstructionIndex(MI)); - VNInfo *VNI = LRQ.valueIn(); - - // VNI will be valid because MachineOperand::readsReg() is checked by caller. - assert(VNI && "No value to read by operand"); - MachineInstr *Def = LIS->getInstructionFromIndex(VNI->def); - // Phis and other noninstructions (after coalescing) have a NULL Def. - if (Def) { - SUnit *DefSU = getSUnit(Def); - if (DefSU) { - // The reaching Def lives within this scheduling region. - // Create a data dependence. - SDep dep(DefSU, SDep::Data, Reg); - // Adjust the dependence latency using operand def/use information, then - // allow the target to perform its own adjustments. - int DefOp = Def->findRegisterDefOperandIdx(Reg); - dep.setLatency(SchedModel.computeOperandLatency(Def, DefOp, MI, OperIdx)); - - const TargetSubtargetInfo &ST = MF.getSubtarget(); - ST.adjustSchedDependency(DefSU, SU, const_cast(dep)); - SU->addPred(dep); - } - } - - // Add antidependence to the following def of the vreg it uses. - VReg2SUnitMap::iterator DefI = VRegDefs.find(Reg); - if (DefI != VRegDefs.end() && DefI->SU != SU) - DefI->SU->addPred(SDep(SU, SDep::Anti, Reg)); } /// Return true if MI is an instruction we are unable to reason about /// (like a call or something with unmodeled side effects). static inline bool isGlobalMemoryObject(AliasAnalysis *AA, MachineInstr *MI) { - if (MI->isCall() || MI->hasUnmodeledSideEffects() || - (MI->hasOrderedMemoryRef() && - (!MI->mayLoad() || !MI->isInvariantLoad(AA)))) - return true; - return false; + return MI->isCall() || MI->hasUnmodeledSideEffects() || + (MI->hasOrderedMemoryRef() && + (!MI->mayLoad() || !MI->isInvariantLoad(AA))); } // This MI might have either incomplete info, or known to be unsafe @@ -508,7 +559,7 @@ static inline bool isUnsafeMemoryObject(MachineInstr *MI, return false; } -/// This returns true if the two MIs need a chain edge betwee them. +/// This returns true if the two MIs need a chain edge between them. /// If these are not even memory operations, we still may need /// chain deps between them. The question really is - could /// these two MIs be reordered during scheduling from memory dependency @@ -574,13 +625,13 @@ static bool MIsNeedChainEdge(AliasAnalysis *AA, const MachineFrameInfo *MFI, int64_t Overlapa = MMOa->getSize() + MMOa->getOffset() - MinOffset; int64_t Overlapb = MMOb->getSize() + MMOb->getOffset() - MinOffset; - AliasAnalysis::AliasResult AAResult = + AliasResult AAResult = AA->alias(MemoryLocation(MMOa->getValue(), Overlapa, UseTBAA ? MMOa->getAAInfo() : AAMDNodes()), MemoryLocation(MMOb->getValue(), Overlapb, UseTBAA ? MMOb->getAAInfo() : AAMDNodes())); - return (AAResult != AliasAnalysis::NoAlias); + return (AAResult != NoAlias); } /// This recursive function iterates over chain deps of SUb looking for @@ -670,7 +721,7 @@ static inline void addChainDependency(AliasAnalysis *AA, unsigned TrueMemOrderLatency = 0, bool isNormalMemory = false) { // If this is a false dependency, - // do not add the edge, but rememeber the rejected node. + // do not add the edge, but remember the rejected node. if (MIsNeedChainEdge(AA, MFI, DL, SUa->getInstr(), SUb->getInstr())) { SDep Dep(SUa, isNormalMemory ? SDep::MayAliasMem : SDep::Barrier); Dep.setLatency(TrueMemOrderLatency); @@ -685,7 +736,7 @@ static inline void addChainDependency(AliasAnalysis *AA, } } -/// Create an SUnit for each real instruction, numbered in top-down toplological +/// Create an SUnit for each real instruction, numbered in top-down topological /// order. The instruction order A < B, implies that no edge exists from B to A. /// /// Map each real instruction to its SUnit. @@ -743,17 +794,44 @@ void ScheduleDAGInstrs::initSUnits() { } } +void ScheduleDAGInstrs::collectVRegUses(SUnit *SU) { + const MachineInstr *MI = SU->getInstr(); + for (const MachineOperand &MO : MI->operands()) { + if (!MO.isReg()) + continue; + if (!MO.readsReg()) + continue; + if (TrackLaneMasks && !MO.isUse()) + continue; + + unsigned Reg = MO.getReg(); + if (!TargetRegisterInfo::isVirtualRegister(Reg)) + continue; + + // Record this local VReg use. + VReg2SUnitMultiMap::iterator UI = VRegUses.find(Reg); + for (; UI != VRegUses.end(); ++UI) { + if (UI->SU == SU) + break; + } + if (UI == VRegUses.end()) + VRegUses.insert(VReg2SUnit(Reg, 0, SU)); + } +} + /// If RegPressure is non-null, compute register pressure as a side effect. The /// DAG builder is an efficient place to do it because it already visits /// operands. void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, RegPressureTracker *RPTracker, - PressureDiffs *PDiffs) { + PressureDiffs *PDiffs, + bool TrackLaneMasks) { const TargetSubtargetInfo &ST = MF.getSubtarget(); bool UseAA = EnableAASchedMI.getNumOccurrences() > 0 ? EnableAASchedMI : ST.useAA(); AliasAnalysis *AAForDep = UseAA ? AA : nullptr; + this->TrackLaneMasks = TrackLaneMasks; MISUnitMap.clear(); ScheduleDAG::clearDAG(); @@ -766,7 +844,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, // We build scheduling units by walking a block's instruction list from bottom // to top. - // Remember where a generic side-effecting instruction is as we procede. + // Remember where a generic side-effecting instruction is as we proceed. SUnit *BarrierChain = nullptr, *AliasChain = nullptr; // Memory references to specific known memory locations are tracked @@ -787,10 +865,14 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, Defs.setUniverse(TRI->getNumRegs()); Uses.setUniverse(TRI->getNumRegs()); - assert(VRegDefs.empty() && "Only BuildSchedGraph may access VRegDefs"); + assert(CurrentVRegDefs.empty() && "nobody else should use CurrentVRegDefs"); + assert(CurrentVRegUses.empty() && "nobody else should use CurrentVRegUses"); + unsigned NumVirtRegs = MRI.getNumVirtRegs(); + CurrentVRegDefs.setUniverse(NumVirtRegs); + CurrentVRegUses.setUniverse(NumVirtRegs); + VRegUses.clear(); - VRegDefs.setUniverse(MRI.getNumVirtRegs()); - VRegUses.setUniverse(MRI.getNumVirtRegs()); + VRegUses.setUniverse(NumVirtRegs); // Model data dependencies between instructions being scheduled and the // ExitSU. @@ -814,10 +896,16 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, assert(SU && "No SUnit mapped to this MI"); if (RPTracker) { - PressureDiff *PDiff = PDiffs ? &(*PDiffs)[SU->NodeNum] : nullptr; - RPTracker->recede(/*LiveUses=*/nullptr, PDiff); - assert(RPTracker->getPos() == std::prev(MII) && - "RPTracker can't find MI"); + collectVRegUses(SU); + + RegisterOperands RegOpers; + RegOpers.collect(*MI, *TRI, MRI); + if (PDiffs != nullptr) + PDiffs->addInstruction(SU->NodeNum, RegOpers, MRI); + + RPTracker->recedeSkipDebugValues(); + assert(&*RPTracker->getPos() == MI && "RPTracker in sync"); + RPTracker->recede(RegOpers); } assert( @@ -835,7 +923,6 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, if (TRI->isPhysicalRegister(Reg)) addPhysRegDeps(SU, j); else { - assert(!IsPostRA && "Virtual register encountered!"); if (MO.isDef()) { HasVRegDef = true; addVRegDefDeps(SU, j); @@ -890,7 +977,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, BarrierChain = SU; // This is a barrier event that acts as a pivotal node in the DAG, // so it is safe to clear list of exposed nodes. - adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes, + adjustChainDeps(AA, MFI, MF.getDataLayout(), SU, &ExitSU, RejectMemNodes, TrueMemOrderLatency); RejectMemNodes.clear(); NonAliasMemDefs.clear(); @@ -903,27 +990,30 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, unsigned ChainLatency = 0; if (AliasChain->getInstr()->mayLoad()) ChainLatency = TrueMemOrderLatency; - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, AliasChain, RejectMemNodes, ChainLatency); } AliasChain = SU; for (unsigned k = 0, m = PendingLoads.size(); k != m; ++k) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, PendingLoads[k], RejectMemNodes, TrueMemOrderLatency); for (MapVector >::iterator I = AliasMemDefs.begin(), E = AliasMemDefs.end(); I != E; ++I) { for (unsigned i = 0, e = I->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, I->second[i], RejectMemNodes); } for (MapVector >::iterator I = AliasMemUses.begin(), E = AliasMemUses.end(); I != E; ++I) { for (unsigned i = 0, e = I->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, I->second[i], RejectMemNodes, TrueMemOrderLatency); } - adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes, + // This call must come after calls to addChainDependency() since it + // consumes the 'RejectMemNodes' list that addChainDependency() possibly + // adds to. + adjustChainDeps(AA, MFI, MF.getDataLayout(), SU, &ExitSU, RejectMemNodes, TrueMemOrderLatency); PendingLoads.clear(); AliasMemDefs.clear(); @@ -937,7 +1027,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, BarrierChain->addPred(SDep(SU, SDep::Barrier)); UnderlyingObjectsVector Objs; - getUnderlyingObjectsForInstr(MI, MFI, Objs, *TM.getDataLayout()); + getUnderlyingObjectsForInstr(MI, MFI, Objs, MF.getDataLayout()); if (Objs.empty()) { // Treat all other stores conservatively. @@ -961,7 +1051,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, ((ThisMayAlias) ? AliasMemDefs.end() : NonAliasMemDefs.end()); if (I != IE) { for (unsigned i = 0, e = I->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, I->second[i], RejectMemNodes, 0, true); // If we're not using AA, then we only need one store per object. @@ -986,7 +1076,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, ((ThisMayAlias) ? AliasMemUses.end() : NonAliasMemUses.end()); if (J != JE) { for (unsigned i = 0, e = J->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, J->second[i], RejectMemNodes, TrueMemOrderLatency, true); J->second.clear(); @@ -996,15 +1086,18 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, // Add dependencies from all the PendingLoads, i.e. loads // with no underlying object. for (unsigned k = 0, m = PendingLoads.size(); k != m; ++k) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, PendingLoads[k], RejectMemNodes, TrueMemOrderLatency); // Add dependence on alias chain, if needed. if (AliasChain) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, AliasChain, RejectMemNodes); } - adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes, + // This call must come after calls to addChainDependency() since it + // consumes the 'RejectMemNodes' list that addChainDependency() possibly + // adds to. + adjustChainDeps(AA, MFI, MF.getDataLayout(), SU, &ExitSU, RejectMemNodes, TrueMemOrderLatency); } else if (MI->mayLoad()) { bool MayAlias = true; @@ -1012,7 +1105,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, // Invariant load, no chain dependencies needed! } else { UnderlyingObjectsVector Objs; - getUnderlyingObjectsForInstr(MI, MFI, Objs, *TM.getDataLayout()); + getUnderlyingObjectsForInstr(MI, MFI, Objs, MF.getDataLayout()); if (Objs.empty()) { // A load with no underlying object. Depend on all @@ -1020,7 +1113,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, for (MapVector >::iterator I = AliasMemDefs.begin(), E = AliasMemDefs.end(); I != E; ++I) for (unsigned i = 0, e = I->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, I->second[i], RejectMemNodes); PendingLoads.push_back(SU); @@ -1044,20 +1137,23 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, ((ThisMayAlias) ? AliasMemDefs.end() : NonAliasMemDefs.end()); if (I != IE) for (unsigned i = 0, e = I->second.size(); i != e; ++i) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, I->second[i], RejectMemNodes, 0, true); if (ThisMayAlias) AliasMemUses[V].push_back(SU); else NonAliasMemUses[V].push_back(SU); } - if (MayAlias) - adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, - RejectMemNodes, /*Latency=*/0); // Add dependencies on alias and barrier chains, if needed. if (MayAlias && AliasChain) - addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain, + addChainDependency(AAForDep, MFI, MF.getDataLayout(), SU, AliasChain, RejectMemNodes); + if (MayAlias) + // This call must come after calls to addChainDependency() since it + // consumes the 'RejectMemNodes' list that addChainDependency() + // possibly adds to. + adjustChainDeps(AA, MFI, MF.getDataLayout(), SU, &ExitSU, + RejectMemNodes, /*Latency=*/0); if (BarrierChain) BarrierChain->addPred(SDep(SU, SDep::Barrier)); } @@ -1068,7 +1164,8 @@ void ScheduleDAGInstrs::buildSchedGraph(AliasAnalysis *AA, Defs.clear(); Uses.clear(); - VRegDefs.clear(); + CurrentVRegDefs.clear(); + CurrentVRegUses.clear(); PendingLoads.clear(); } @@ -1080,11 +1177,9 @@ void ScheduleDAGInstrs::startBlockForKills(MachineBasicBlock *BB) { // Examine the live-in regs of all successors. for (MachineBasicBlock::succ_iterator SI = BB->succ_begin(), SE = BB->succ_end(); SI != SE; ++SI) { - for (MachineBasicBlock::livein_iterator I = (*SI)->livein_begin(), - E = (*SI)->livein_end(); I != E; ++I) { - unsigned Reg = *I; + for (const auto &LI : (*SI)->liveins()) { // Repeat, for reg and all subregs. - for (MCSubRegIterator SubRegs(Reg, TRI, /*IncludeSelf=*/true); + for (MCSubRegIterator SubRegs(LI.PhysReg, TRI, /*IncludeSelf=*/true); SubRegs.isValid(); ++SubRegs) LiveRegs.set(*SubRegs); } @@ -1103,7 +1198,7 @@ static void toggleBundleKillFlag(MachineInstr *MI, unsigned Reg, // Once we set a kill flag on an instruction, we bail out, as otherwise we // might set it on too many operands. We will clear as many flags as we // can though. - MachineBasicBlock::instr_iterator Begin = MI; + MachineBasicBlock::instr_iterator Begin = MI->getIterator(); MachineBasicBlock::instr_iterator End = getBundleEnd(MI); while (Begin != End) { for (MachineOperand &MO : (--End)->operands()) { @@ -1237,7 +1332,7 @@ void ScheduleDAGInstrs::fixupKills(MachineBasicBlock *MBB) { toggleKillFlag(MI, MO); DEBUG(MI->dump()); DEBUG(if (MI->getOpcode() == TargetOpcode::BUNDLE) { - MachineBasicBlock::instr_iterator Begin = MI; + MachineBasicBlock::instr_iterator Begin = MI->getIterator(); MachineBasicBlock::instr_iterator End = getBundleEnd(MI); while (++Begin != End) DEBUG(Begin->dump()); @@ -1508,7 +1603,7 @@ public: return getCurr()->Preds.end(); } }; -} // namespace +} // anonymous static bool hasDataSucc(const SUnit *SU) { for (SUnit::const_succ_iterator