brintos

brintos / llvm-project-archived public Read only

0
0
Text · 33.8 KiB · 048cd21 Raw
956 lines · cpp
1//===- RegAllocPBQP.cpp ---- PBQP Register Allocator ----------------------===//2//3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.4// See https://llvm.org/LICENSE.txt for license information.5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception6//7//===----------------------------------------------------------------------===//8//9// This file contains a Partitioned Boolean Quadratic Programming (PBQP) based10// register allocator for LLVM. This allocator works by constructing a PBQP11// problem representing the register allocation problem under consideration,12// solving this using a PBQP solver, and mapping the solution back to a13// register assignment. If any variables are selected for spilling then spill14// code is inserted and the process repeated.15//16// The PBQP solver (pbqp.c) provided for this allocator uses a heuristic tuned17// for register allocation. For more information on PBQP for register18// allocation, see the following papers:19//20//   (1) Hames, L. and Scholz, B. 2006. Nearly optimal register allocation with21//   PBQP. In Proceedings of the 7th Joint Modular Languages Conference22//   (JMLC'06). LNCS, vol. 4228. Springer, New York, NY, USA. 346-361.23//24//   (2) Scholz, B., Eckstein, E. 2002. Register allocation for irregular25//   architectures. In Proceedings of the Joint Conference on Languages,26//   Compilers and Tools for Embedded Systems (LCTES'02), ACM Press, New York,27//   NY, USA, 139-148.28//29//===----------------------------------------------------------------------===//30 31#include "llvm/CodeGen/RegAllocPBQP.h"32#include "RegisterCoalescer.h"33#include "llvm/ADT/ArrayRef.h"34#include "llvm/ADT/BitVector.h"35#include "llvm/ADT/DenseMap.h"36#include "llvm/ADT/DenseSet.h"37#include "llvm/ADT/STLExtras.h"38#include "llvm/ADT/SmallPtrSet.h"39#include "llvm/ADT/SmallVector.h"40#include "llvm/ADT/StringRef.h"41#include "llvm/Analysis/AliasAnalysis.h"42#include "llvm/CodeGen/CalcSpillWeights.h"43#include "llvm/CodeGen/LiveInterval.h"44#include "llvm/CodeGen/LiveIntervals.h"45#include "llvm/CodeGen/LiveRangeEdit.h"46#include "llvm/CodeGen/LiveStacks.h"47#include "llvm/CodeGen/MachineBlockFrequencyInfo.h"48#include "llvm/CodeGen/MachineDominators.h"49#include "llvm/CodeGen/MachineFunction.h"50#include "llvm/CodeGen/MachineFunctionPass.h"51#include "llvm/CodeGen/MachineInstr.h"52#include "llvm/CodeGen/MachineLoopInfo.h"53#include "llvm/CodeGen/MachineRegisterInfo.h"54#include "llvm/CodeGen/PBQP/Graph.h"55#include "llvm/CodeGen/PBQP/Math.h"56#include "llvm/CodeGen/PBQP/Solution.h"57#include "llvm/CodeGen/PBQPRAConstraint.h"58#include "llvm/CodeGen/RegAllocRegistry.h"59#include "llvm/CodeGen/SlotIndexes.h"60#include "llvm/CodeGen/Spiller.h"61#include "llvm/CodeGen/TargetRegisterInfo.h"62#include "llvm/CodeGen/TargetSubtargetInfo.h"63#include "llvm/CodeGen/VirtRegMap.h"64#include "llvm/Config/llvm-config.h"65#include "llvm/IR/Function.h"66#include "llvm/IR/Module.h"67#include "llvm/Pass.h"68#include "llvm/Support/CommandLine.h"69#include "llvm/Support/Compiler.h"70#include "llvm/Support/Debug.h"71#include "llvm/Support/FileSystem.h"72#include "llvm/Support/Printable.h"73#include "llvm/Support/raw_ostream.h"74#include <algorithm>75#include <cassert>76#include <cstddef>77#include <limits>78#include <map>79#include <memory>80#include <queue>81#include <set>82#include <sstream>83#include <string>84#include <system_error>85#include <tuple>86#include <utility>87#include <vector>88 89using namespace llvm;90 91#define DEBUG_TYPE "regalloc"92 93static RegisterRegAlloc94RegisterPBQPRepAlloc("pbqp", "PBQP register allocator",95                       createDefaultPBQPRegisterAllocator);96 97static cl::opt<bool>98PBQPCoalescing("pbqp-coalescing",99                cl::desc("Attempt coalescing during PBQP register allocation."),100                cl::init(false), cl::Hidden);101 102#ifndef NDEBUG103static cl::opt<bool>104PBQPDumpGraphs("pbqp-dump-graphs",105               cl::desc("Dump graphs for each function/round in the compilation unit."),106               cl::init(false), cl::Hidden);107#endif108 109namespace {110 111///112/// PBQP based allocators solve the register allocation problem by mapping113/// register allocation problems to Partitioned Boolean Quadratic114/// Programming problems.115class RegAllocPBQP : public MachineFunctionPass {116public:117  static char ID;118 119  /// Construct a PBQP register allocator.120  RegAllocPBQP(char *cPassID = nullptr)121      : MachineFunctionPass(ID), customPassID(cPassID) {122    initializeSlotIndexesWrapperPassPass(*PassRegistry::getPassRegistry());123    initializeLiveIntervalsWrapperPassPass(*PassRegistry::getPassRegistry());124    initializeLiveStacksWrapperLegacyPass(*PassRegistry::getPassRegistry());125    initializeVirtRegMapWrapperLegacyPass(*PassRegistry::getPassRegistry());126  }127 128  /// Return the pass name.129  StringRef getPassName() const override { return "PBQP Register Allocator"; }130 131  /// PBQP analysis usage.132  void getAnalysisUsage(AnalysisUsage &au) const override;133 134  /// Perform register allocation135  bool runOnMachineFunction(MachineFunction &MF) override;136 137  MachineFunctionProperties getRequiredProperties() const override {138    return MachineFunctionProperties().setNoPHIs();139  }140 141  MachineFunctionProperties getClearedProperties() const override {142    return MachineFunctionProperties().setIsSSA();143  }144 145private:146  using RegSet = std::set<Register>;147 148  char *customPassID;149 150  RegSet VRegsToAlloc, EmptyIntervalVRegs;151 152  /// Inst which is a def of an original reg and whose defs are already all153  /// dead after remat is saved in DeadRemats. The deletion of such inst is154  /// postponed till all the allocations are done, so its remat expr is155  /// always available for the remat of all the siblings of the original reg.156  SmallPtrSet<MachineInstr *, 32> DeadRemats;157 158  /// Finds the initial set of vreg intervals to allocate.159  void findVRegIntervalsToAlloc(const MachineFunction &MF, LiveIntervals &LIS);160 161  /// Constructs an initial graph.162  void initializeGraph(PBQPRAGraph &G, VirtRegMap &VRM, Spiller &VRegSpiller);163 164  /// Spill the given VReg.165  void spillVReg(Register VReg, SmallVectorImpl<Register> &NewIntervals,166                 MachineFunction &MF, LiveIntervals &LIS, VirtRegMap &VRM,167                 Spiller &VRegSpiller);168 169  /// Given a solved PBQP problem maps this solution back to a register170  /// assignment.171  bool mapPBQPToRegAlloc(const PBQPRAGraph &G,172                         const PBQP::Solution &Solution,173                         VirtRegMap &VRM,174                         Spiller &VRegSpiller);175 176  /// Postprocessing before final spilling. Sets basic block "live in"177  /// variables.178  void finalizeAlloc(MachineFunction &MF, LiveIntervals &LIS,179                     VirtRegMap &VRM) const;180 181  void postOptimization(Spiller &VRegSpiller, LiveIntervals &LIS);182};183 184char RegAllocPBQP::ID = 0;185 186/// Set spill costs for each node in the PBQP reg-alloc graph.187class SpillCosts : public PBQPRAConstraint {188public:189  void apply(PBQPRAGraph &G) override {190    LiveIntervals &LIS = G.getMetadata().LIS;191 192    // A minimum spill costs, so that register constraints can be set193    // without normalization in the [0.0:MinSpillCost( interval.194    const PBQP::PBQPNum MinSpillCost = 10.0;195 196    for (auto NId : G.nodeIds()) {197      PBQP::PBQPNum SpillCost =198          LIS.getInterval(G.getNodeMetadata(NId).getVReg()).weight();199      if (SpillCost == 0.0)200        SpillCost = std::numeric_limits<PBQP::PBQPNum>::min();201      else202        SpillCost += MinSpillCost;203      PBQPRAGraph::RawVector NodeCosts(G.getNodeCosts(NId));204      NodeCosts[PBQP::RegAlloc::getSpillOptionIdx()] = SpillCost;205      G.setNodeCosts(NId, std::move(NodeCosts));206    }207  }208};209 210/// Add interference edges between overlapping vregs.211class Interference : public PBQPRAConstraint {212private:213  using AllowedRegVecPtr = const PBQP::RegAlloc::AllowedRegVector *;214  using IKey = std::pair<AllowedRegVecPtr, AllowedRegVecPtr>;215  using IMatrixCache = DenseMap<IKey, PBQPRAGraph::MatrixPtr>;216  using DisjointAllowedRegsCache = DenseSet<IKey>;217  using IEdgeKey = std::pair<PBQP::GraphBase::NodeId, PBQP::GraphBase::NodeId>;218  using IEdgeCache = DenseSet<IEdgeKey>;219 220  bool haveDisjointAllowedRegs(const PBQPRAGraph &G, PBQPRAGraph::NodeId NId,221                               PBQPRAGraph::NodeId MId,222                               const DisjointAllowedRegsCache &D) const {223    const auto *NRegs = &G.getNodeMetadata(NId).getAllowedRegs();224    const auto *MRegs = &G.getNodeMetadata(MId).getAllowedRegs();225 226    if (NRegs == MRegs)227      return false;228 229    if (NRegs < MRegs)230      return D.contains(IKey(NRegs, MRegs));231 232    return D.contains(IKey(MRegs, NRegs));233  }234 235  void setDisjointAllowedRegs(const PBQPRAGraph &G, PBQPRAGraph::NodeId NId,236                              PBQPRAGraph::NodeId MId,237                              DisjointAllowedRegsCache &D) {238    const auto *NRegs = &G.getNodeMetadata(NId).getAllowedRegs();239    const auto *MRegs = &G.getNodeMetadata(MId).getAllowedRegs();240 241    assert(NRegs != MRegs && "AllowedRegs can not be disjoint with itself");242 243    if (NRegs < MRegs)244      D.insert(IKey(NRegs, MRegs));245    else246      D.insert(IKey(MRegs, NRegs));247  }248 249  // Holds (Interval, CurrentSegmentID, and NodeId). The first two are required250  // for the fast interference graph construction algorithm. The last is there251  // to save us from looking up node ids via the VRegToNode map in the graph252  // metadata.253  using IntervalInfo =254      std::tuple<LiveInterval*, size_t, PBQP::GraphBase::NodeId>;255 256  static SlotIndex getStartPoint(const IntervalInfo &I) {257    return std::get<0>(I)->segments[std::get<1>(I)].start;258  }259 260  static SlotIndex getEndPoint(const IntervalInfo &I) {261    return std::get<0>(I)->segments[std::get<1>(I)].end;262  }263 264  static PBQP::GraphBase::NodeId getNodeId(const IntervalInfo &I) {265    return std::get<2>(I);266  }267 268  static bool lowestStartPoint(const IntervalInfo &I1,269                               const IntervalInfo &I2) {270    // Condition reversed because priority queue has the *highest* element at271    // the front, rather than the lowest.272    return getStartPoint(I1) > getStartPoint(I2);273  }274 275  static bool lowestEndPoint(const IntervalInfo &I1,276                             const IntervalInfo &I2) {277    SlotIndex E1 = getEndPoint(I1);278    SlotIndex E2 = getEndPoint(I2);279 280    if (E1 < E2)281      return true;282 283    if (E1 > E2)284      return false;285 286    // If two intervals end at the same point, we need a way to break the tie or287    // the set will assume they're actually equal and refuse to insert a288    // "duplicate". Just compare the vregs - fast and guaranteed unique.289    return std::get<0>(I1)->reg() < std::get<0>(I2)->reg();290  }291 292  static bool isAtLastSegment(const IntervalInfo &I) {293    return std::get<1>(I) == std::get<0>(I)->size() - 1;294  }295 296  static IntervalInfo nextSegment(const IntervalInfo &I) {297    return std::make_tuple(std::get<0>(I), std::get<1>(I) + 1, std::get<2>(I));298  }299 300public:301  void apply(PBQPRAGraph &G) override {302    // The following is loosely based on the linear scan algorithm introduced in303    // "Linear Scan Register Allocation" by Poletto and Sarkar. This version304    // isn't linear, because the size of the active set isn't bound by the305    // number of registers, but rather the size of the largest clique in the306    // graph. Still, we expect this to be better than N^2.307    LiveIntervals &LIS = G.getMetadata().LIS;308 309    // Interferenc matrices are incredibly regular - they're only a function of310    // the allowed sets, so we cache them to avoid the overhead of constructing311    // and uniquing them.312    IMatrixCache C;313 314    // Finding an edge is expensive in the worst case (O(max_clique(G))). So315    // cache locally edges we have already seen.316    IEdgeCache EC;317 318    // Cache known disjoint allowed registers pairs319    DisjointAllowedRegsCache D;320 321    using IntervalSet = std::set<IntervalInfo, decltype(&lowestEndPoint)>;322    using IntervalQueue =323        std::priority_queue<IntervalInfo, std::vector<IntervalInfo>,324                            decltype(&lowestStartPoint)>;325    IntervalSet Active(lowestEndPoint);326    IntervalQueue Inactive(lowestStartPoint);327 328    // Start by building the inactive set.329    for (auto NId : G.nodeIds()) {330      Register VReg = G.getNodeMetadata(NId).getVReg();331      LiveInterval &LI = LIS.getInterval(VReg);332      assert(!LI.empty() && "PBQP graph contains node for empty interval");333      Inactive.push(std::make_tuple(&LI, 0, NId));334    }335 336    while (!Inactive.empty()) {337      // Tentatively grab the "next" interval - this choice may be overriden338      // below.339      IntervalInfo Cur = Inactive.top();340 341      // Retire any active intervals that end before Cur starts.342      IntervalSet::iterator RetireItr = Active.begin();343      while (RetireItr != Active.end() &&344             (getEndPoint(*RetireItr) <= getStartPoint(Cur))) {345        // If this interval has subsequent segments, add the next one to the346        // inactive list.347        if (!isAtLastSegment(*RetireItr))348          Inactive.push(nextSegment(*RetireItr));349 350        ++RetireItr;351      }352      Active.erase(Active.begin(), RetireItr);353 354      // One of the newly retired segments may actually start before the355      // Cur segment, so re-grab the front of the inactive list.356      Cur = Inactive.top();357      Inactive.pop();358 359      // At this point we know that Cur overlaps all active intervals. Add the360      // interference edges.361      PBQP::GraphBase::NodeId NId = getNodeId(Cur);362      for (const auto &A : Active) {363        PBQP::GraphBase::NodeId MId = getNodeId(A);364 365        // Do not add an edge when the nodes' allowed registers do not366        // intersect: there is obviously no interference.367        if (haveDisjointAllowedRegs(G, NId, MId, D))368          continue;369 370        // Check that we haven't already added this edge371        IEdgeKey EK(std::min(NId, MId), std::max(NId, MId));372        if (EC.count(EK))373          continue;374 375        // This is a new edge - add it to the graph.376        if (!createInterferenceEdge(G, NId, MId, C))377          setDisjointAllowedRegs(G, NId, MId, D);378        else379          EC.insert(EK);380      }381 382      // Finally, add Cur to the Active set.383      Active.insert(Cur);384    }385  }386 387private:388  // Create an Interference edge and add it to the graph, unless it is389  // a null matrix, meaning the nodes' allowed registers do not have any390  // interference. This case occurs frequently between integer and floating391  // point registers for example.392  // return true iff both nodes interferes.393  bool createInterferenceEdge(PBQPRAGraph &G,394                              PBQPRAGraph::NodeId NId, PBQPRAGraph::NodeId MId,395                              IMatrixCache &C) {396    const TargetRegisterInfo &TRI =397        *G.getMetadata().MF.getSubtarget().getRegisterInfo();398    const auto &NRegs = G.getNodeMetadata(NId).getAllowedRegs();399    const auto &MRegs = G.getNodeMetadata(MId).getAllowedRegs();400 401    // Try looking the edge costs up in the IMatrixCache first.402    IKey K(&NRegs, &MRegs);403    IMatrixCache::iterator I = C.find(K);404    if (I != C.end()) {405      G.addEdgeBypassingCostAllocator(NId, MId, I->second);406      return true;407    }408 409    PBQPRAGraph::RawMatrix M(NRegs.size() + 1, MRegs.size() + 1, 0);410    bool NodesInterfere = false;411    for (unsigned I = 0; I != NRegs.size(); ++I) {412      MCRegister PRegN = NRegs[I];413      for (unsigned J = 0; J != MRegs.size(); ++J) {414        MCRegister PRegM = MRegs[J];415        if (TRI.regsOverlap(PRegN, PRegM)) {416          M[I + 1][J + 1] = std::numeric_limits<PBQP::PBQPNum>::infinity();417          NodesInterfere = true;418        }419      }420    }421 422    if (!NodesInterfere)423      return false;424 425    PBQPRAGraph::EdgeId EId = G.addEdge(NId, MId, std::move(M));426    C[K] = G.getEdgeCostsPtr(EId);427 428    return true;429  }430};431 432class Coalescing : public PBQPRAConstraint {433public:434  void apply(PBQPRAGraph &G) override {435    MachineFunction &MF = G.getMetadata().MF;436    MachineBlockFrequencyInfo &MBFI = G.getMetadata().MBFI;437    CoalescerPair CP(*MF.getSubtarget().getRegisterInfo());438 439    // Scan the machine function and add a coalescing cost whenever CoalescerPair440    // gives the Ok.441    for (const auto &MBB : MF) {442      for (const auto &MI : MBB) {443        // Skip not-coalescable or already coalesced copies.444        if (!CP.setRegisters(&MI) || CP.getSrcReg() == CP.getDstReg())445          continue;446 447        Register DstReg = CP.getDstReg();448        Register SrcReg = CP.getSrcReg();449 450        PBQP::PBQPNum CBenefit = MBFI.getBlockFreqRelativeToEntryBlock(&MBB);451 452        if (CP.isPhys()) {453          if (!MF.getRegInfo().isAllocatable(DstReg))454            continue;455 456          PBQPRAGraph::NodeId NId = G.getMetadata().getNodeIdForVReg(SrcReg);457 458          const PBQPRAGraph::NodeMetadata::AllowedRegVector &Allowed =459            G.getNodeMetadata(NId).getAllowedRegs();460 461          unsigned PRegOpt = 0;462          while (PRegOpt < Allowed.size() && Allowed[PRegOpt].id() != DstReg)463            ++PRegOpt;464 465          if (PRegOpt < Allowed.size()) {466            PBQPRAGraph::RawVector NewCosts(G.getNodeCosts(NId));467            NewCosts[PRegOpt + 1] -= CBenefit;468            G.setNodeCosts(NId, std::move(NewCosts));469          }470        } else {471          PBQPRAGraph::NodeId N1Id = G.getMetadata().getNodeIdForVReg(DstReg);472          PBQPRAGraph::NodeId N2Id = G.getMetadata().getNodeIdForVReg(SrcReg);473          const PBQPRAGraph::NodeMetadata::AllowedRegVector *Allowed1 =474            &G.getNodeMetadata(N1Id).getAllowedRegs();475          const PBQPRAGraph::NodeMetadata::AllowedRegVector *Allowed2 =476            &G.getNodeMetadata(N2Id).getAllowedRegs();477 478          PBQPRAGraph::EdgeId EId = G.findEdge(N1Id, N2Id);479          if (EId == G.invalidEdgeId()) {480            PBQPRAGraph::RawMatrix Costs(Allowed1->size() + 1,481                                         Allowed2->size() + 1, 0);482            addVirtRegCoalesce(Costs, *Allowed1, *Allowed2, CBenefit);483            G.addEdge(N1Id, N2Id, std::move(Costs));484          } else {485            if (G.getEdgeNode1Id(EId) == N2Id) {486              std::swap(N1Id, N2Id);487              std::swap(Allowed1, Allowed2);488            }489            PBQPRAGraph::RawMatrix Costs(G.getEdgeCosts(EId));490            addVirtRegCoalesce(Costs, *Allowed1, *Allowed2, CBenefit);491            G.updateEdgeCosts(EId, std::move(Costs));492          }493        }494      }495    }496  }497 498private:499  void addVirtRegCoalesce(500                    PBQPRAGraph::RawMatrix &CostMat,501                    const PBQPRAGraph::NodeMetadata::AllowedRegVector &Allowed1,502                    const PBQPRAGraph::NodeMetadata::AllowedRegVector &Allowed2,503                    PBQP::PBQPNum Benefit) {504    assert(CostMat.getRows() == Allowed1.size() + 1 && "Size mismatch.");505    assert(CostMat.getCols() == Allowed2.size() + 1 && "Size mismatch.");506    for (unsigned I = 0; I != Allowed1.size(); ++I) {507      MCRegister PReg1 = Allowed1[I];508      for (unsigned J = 0; J != Allowed2.size(); ++J) {509        MCRegister PReg2 = Allowed2[J];510        if (PReg1 == PReg2)511          CostMat[I + 1][J + 1] -= Benefit;512      }513    }514  }515};516 517/// PBQP-specific implementation of weight normalization.518class PBQPVirtRegAuxInfo final : public VirtRegAuxInfo {519  float normalize(float UseDefFreq, unsigned Size, unsigned NumInstr) override {520    // All intervals have a spill weight that is mostly proportional to the521    // number of uses, with uses in loops having a bigger weight.522    return NumInstr * VirtRegAuxInfo::normalize(UseDefFreq, Size, 1);523  }524 525public:526  PBQPVirtRegAuxInfo(MachineFunction &MF, LiveIntervals &LIS, VirtRegMap &VRM,527                     const MachineLoopInfo &Loops,528                     const MachineBlockFrequencyInfo &MBFI)529      : VirtRegAuxInfo(MF, LIS, VRM, Loops, MBFI) {}530};531} // end anonymous namespace532 533// Out-of-line destructor/anchor for PBQPRAConstraint.534PBQPRAConstraint::~PBQPRAConstraint() = default;535 536void PBQPRAConstraint::anchor() {}537 538void PBQPRAConstraintList::anchor() {}539 540void RegAllocPBQP::getAnalysisUsage(AnalysisUsage &au) const {541  au.setPreservesCFG();542  au.addRequired<AAResultsWrapperPass>();543  au.addPreserved<AAResultsWrapperPass>();544  au.addRequired<SlotIndexesWrapperPass>();545  au.addPreserved<SlotIndexesWrapperPass>();546  au.addRequired<LiveIntervalsWrapperPass>();547  au.addPreserved<LiveIntervalsWrapperPass>();548  //au.addRequiredID(SplitCriticalEdgesID);549  if (customPassID)550    au.addRequiredID(*customPassID);551  au.addRequired<LiveStacksWrapperLegacy>();552  au.addPreserved<LiveStacksWrapperLegacy>();553  au.addRequired<MachineBlockFrequencyInfoWrapperPass>();554  au.addPreserved<MachineBlockFrequencyInfoWrapperPass>();555  au.addRequired<MachineLoopInfoWrapperPass>();556  au.addPreserved<MachineLoopInfoWrapperPass>();557  au.addRequired<MachineDominatorTreeWrapperPass>();558  au.addPreserved<MachineDominatorTreeWrapperPass>();559  au.addRequired<VirtRegMapWrapperLegacy>();560  au.addPreserved<VirtRegMapWrapperLegacy>();561  MachineFunctionPass::getAnalysisUsage(au);562}563 564void RegAllocPBQP::findVRegIntervalsToAlloc(const MachineFunction &MF,565                                            LiveIntervals &LIS) {566  const MachineRegisterInfo &MRI = MF.getRegInfo();567 568  // Iterate over all live ranges.569  for (unsigned I = 0, E = MRI.getNumVirtRegs(); I != E; ++I) {570    Register Reg = Register::index2VirtReg(I);571    if (MRI.reg_nodbg_empty(Reg))572      continue;573    VRegsToAlloc.insert(Reg);574  }575}576 577static bool isACalleeSavedRegister(MCRegister Reg,578                                   const TargetRegisterInfo &TRI,579                                   const MachineFunction &MF) {580  const MCPhysReg *CSR = MF.getRegInfo().getCalleeSavedRegs();581  for (unsigned i = 0; CSR[i] != 0; ++i)582    if (TRI.regsOverlap(Reg, CSR[i]))583      return true;584  return false;585}586 587void RegAllocPBQP::initializeGraph(PBQPRAGraph &G, VirtRegMap &VRM,588                                   Spiller &VRegSpiller) {589  MachineFunction &MF = G.getMetadata().MF;590 591  LiveIntervals &LIS = G.getMetadata().LIS;592  const MachineRegisterInfo &MRI = G.getMetadata().MF.getRegInfo();593  const TargetRegisterInfo &TRI =594      *G.getMetadata().MF.getSubtarget().getRegisterInfo();595 596  std::vector<Register> Worklist(VRegsToAlloc.begin(), VRegsToAlloc.end());597 598  std::map<Register, std::vector<MCRegister>> VRegAllowedMap;599 600  while (!Worklist.empty()) {601    Register VReg = Worklist.back();602    Worklist.pop_back();603 604    LiveInterval &VRegLI = LIS.getInterval(VReg);605 606    // If this is an empty interval move it to the EmptyIntervalVRegs set then607    // continue.608    if (VRegLI.empty()) {609      EmptyIntervalVRegs.insert(VRegLI.reg());610      VRegsToAlloc.erase(VRegLI.reg());611      continue;612    }613 614    const TargetRegisterClass *TRC = MRI.getRegClass(VReg);615 616    // Record any overlaps with regmask operands.617    BitVector RegMaskOverlaps;618    LIS.checkRegMaskInterference(VRegLI, RegMaskOverlaps);619 620    // Compute an initial allowed set for the current vreg.621    std::vector<MCRegister> VRegAllowed;622    ArrayRef<MCPhysReg> RawPRegOrder = TRC->getRawAllocationOrder(MF);623    for (MCPhysReg R : RawPRegOrder) {624      MCRegister PReg(R);625      if (MRI.isReserved(PReg))626        continue;627 628      // vregLI crosses a regmask operand that clobbers preg.629      if (!RegMaskOverlaps.empty() && !RegMaskOverlaps.test(PReg))630        continue;631 632      // vregLI overlaps fixed regunit interference.633      bool Interference = false;634      for (MCRegUnit Unit : TRI.regunits(PReg)) {635        if (VRegLI.overlaps(LIS.getRegUnit(Unit))) {636          Interference = true;637          break;638        }639      }640      if (Interference)641        continue;642 643      // preg is usable for this virtual register.644      VRegAllowed.push_back(PReg);645    }646 647    // Check for vregs that have no allowed registers. These should be648    // pre-spilled and the new vregs added to the worklist.649    if (VRegAllowed.empty()) {650      SmallVector<Register, 8> NewVRegs;651      spillVReg(VReg, NewVRegs, MF, LIS, VRM, VRegSpiller);652      llvm::append_range(Worklist, NewVRegs);653      continue;654    }655 656    VRegAllowedMap[VReg.id()] = std::move(VRegAllowed);657  }658 659  for (auto &KV : VRegAllowedMap) {660    auto VReg = KV.first;661 662    // Move empty intervals to the EmptyIntervalVReg set.663    if (LIS.getInterval(VReg).empty()) {664      EmptyIntervalVRegs.insert(VReg);665      VRegsToAlloc.erase(VReg);666      continue;667    }668 669    auto &VRegAllowed = KV.second;670 671    PBQPRAGraph::RawVector NodeCosts(VRegAllowed.size() + 1, 0);672 673    // Tweak cost of callee saved registers, as using then force spilling and674    // restoring them. This would only happen in the prologue / epilogue though.675    for (unsigned i = 0; i != VRegAllowed.size(); ++i)676      if (isACalleeSavedRegister(VRegAllowed[i], TRI, MF))677        NodeCosts[1 + i] += 1.0;678 679    PBQPRAGraph::NodeId NId = G.addNode(std::move(NodeCosts));680    G.getNodeMetadata(NId).setVReg(VReg);681    G.getNodeMetadata(NId).setAllowedRegs(682      G.getMetadata().getAllowedRegs(std::move(VRegAllowed)));683    G.getMetadata().setNodeIdForVReg(VReg, NId);684  }685}686 687void RegAllocPBQP::spillVReg(Register VReg,688                             SmallVectorImpl<Register> &NewIntervals,689                             MachineFunction &MF, LiveIntervals &LIS,690                             VirtRegMap &VRM, Spiller &VRegSpiller) {691  VRegsToAlloc.erase(VReg);692  LiveRangeEdit LRE(&LIS.getInterval(VReg), NewIntervals, MF, LIS, &VRM,693                    nullptr, &DeadRemats);694  VRegSpiller.spill(LRE);695 696  const TargetRegisterInfo &TRI = *MF.getSubtarget().getRegisterInfo();697  (void)TRI;698  LLVM_DEBUG(dbgs() << "VREG " << printReg(VReg, &TRI) << " -> SPILLED (Cost: "699                    << LRE.getParent().weight() << ", New vregs: ");700 701  // Copy any newly inserted live intervals into the list of regs to702  // allocate.703  for (const Register &R : LRE) {704    const LiveInterval &LI = LIS.getInterval(R);705    assert(!LI.empty() && "Empty spill range.");706    LLVM_DEBUG(dbgs() << printReg(LI.reg(), &TRI) << " ");707    VRegsToAlloc.insert(LI.reg());708  }709 710  LLVM_DEBUG(dbgs() << ")\n");711}712 713bool RegAllocPBQP::mapPBQPToRegAlloc(const PBQPRAGraph &G,714                                     const PBQP::Solution &Solution,715                                     VirtRegMap &VRM,716                                     Spiller &VRegSpiller) {717  MachineFunction &MF = G.getMetadata().MF;718  LiveIntervals &LIS = G.getMetadata().LIS;719  const TargetRegisterInfo &TRI = *MF.getSubtarget().getRegisterInfo();720  (void)TRI;721 722  // Set to true if we have any spills723  bool AnotherRoundNeeded = false;724 725  // Clear the existing allocation.726  VRM.clearAllVirt();727 728  // Iterate over the nodes mapping the PBQP solution to a register729  // assignment.730  for (auto NId : G.nodeIds()) {731    Register VReg = G.getNodeMetadata(NId).getVReg();732    unsigned AllocOpt = Solution.getSelection(NId);733 734    if (AllocOpt != PBQP::RegAlloc::getSpillOptionIdx()) {735      MCRegister PReg = G.getNodeMetadata(NId).getAllowedRegs()[AllocOpt - 1];736      LLVM_DEBUG(dbgs() << "VREG " << printReg(VReg, &TRI) << " -> "737                        << TRI.getName(PReg) << "\n");738      assert(PReg != 0 && "Invalid preg selected.");739      VRM.assignVirt2Phys(VReg, PReg);740    } else {741      // Spill VReg. If this introduces new intervals we'll need another round742      // of allocation.743      SmallVector<Register, 8> NewVRegs;744      spillVReg(VReg, NewVRegs, MF, LIS, VRM, VRegSpiller);745      AnotherRoundNeeded |= !NewVRegs.empty();746    }747  }748 749  return !AnotherRoundNeeded;750}751 752void RegAllocPBQP::finalizeAlloc(MachineFunction &MF,753                                 LiveIntervals &LIS,754                                 VirtRegMap &VRM) const {755  MachineRegisterInfo &MRI = MF.getRegInfo();756 757  // First allocate registers for the empty intervals.758  for (const Register &R : EmptyIntervalVRegs) {759    LiveInterval &LI = LIS.getInterval(R);760 761    Register PReg = MRI.getSimpleHint(LI.reg());762 763    if (PReg == 0) {764      const TargetRegisterClass &RC = *MRI.getRegClass(LI.reg());765      const ArrayRef<MCPhysReg> RawPRegOrder = RC.getRawAllocationOrder(MF);766      for (MCRegister CandidateReg : RawPRegOrder) {767        if (!VRM.getRegInfo().isReserved(CandidateReg)) {768          PReg = CandidateReg;769          break;770        }771      }772      assert(PReg &&773             "No un-reserved physical registers in this register class");774    }775 776    VRM.assignVirt2Phys(LI.reg(), PReg);777  }778}779 780void RegAllocPBQP::postOptimization(Spiller &VRegSpiller, LiveIntervals &LIS) {781  VRegSpiller.postOptimization();782  /// Remove dead defs because of rematerialization.783  for (auto *DeadInst : DeadRemats) {784    LIS.RemoveMachineInstrFromMaps(*DeadInst);785    DeadInst->eraseFromParent();786  }787  DeadRemats.clear();788}789 790bool RegAllocPBQP::runOnMachineFunction(MachineFunction &MF) {791  LiveIntervals &LIS = getAnalysis<LiveIntervalsWrapperPass>().getLIS();792  MachineBlockFrequencyInfo &MBFI =793      getAnalysis<MachineBlockFrequencyInfoWrapperPass>().getMBFI();794 795  auto &LiveStks = getAnalysis<LiveStacksWrapperLegacy>().getLS();796  auto &MDT = getAnalysis<MachineDominatorTreeWrapperPass>().getDomTree();797 798  VirtRegMap &VRM = getAnalysis<VirtRegMapWrapperLegacy>().getVRM();799 800  PBQPVirtRegAuxInfo VRAI(801      MF, LIS, VRM, getAnalysis<MachineLoopInfoWrapperPass>().getLI(), MBFI);802  VRAI.calculateSpillWeightsAndHints();803 804  // FIXME: we create DefaultVRAI here to match existing behavior pre-passing805  // the VRAI through the spiller to the live range editor. However, it probably806  // makes more sense to pass the PBQP VRAI. The existing behavior had807  // LiveRangeEdit make its own VirtRegAuxInfo object.808  VirtRegAuxInfo DefaultVRAI(809      MF, LIS, VRM, getAnalysis<MachineLoopInfoWrapperPass>().getLI(), MBFI);810  std::unique_ptr<Spiller> VRegSpiller(811      createInlineSpiller({LIS, LiveStks, MDT, MBFI}, MF, VRM, DefaultVRAI));812 813  MF.getRegInfo().freezeReservedRegs();814 815  LLVM_DEBUG(dbgs() << "PBQP Register Allocating for " << MF.getName() << "\n");816 817  // Allocator main loop:818  //819  // * Map current regalloc problem to a PBQP problem820  // * Solve the PBQP problem821  // * Map the solution back to a register allocation822  // * Spill if necessary823  //824  // This process is continued till no more spills are generated.825 826  // Find the vreg intervals in need of allocation.827  findVRegIntervalsToAlloc(MF, LIS);828 829#ifndef NDEBUG830  const Function &F = MF.getFunction();831  std::string FullyQualifiedName =832    F.getParent()->getModuleIdentifier() + "." + F.getName().str();833#endif834 835  // If there are non-empty intervals allocate them using pbqp.836  if (!VRegsToAlloc.empty()) {837    const TargetSubtargetInfo &Subtarget = MF.getSubtarget();838    std::unique_ptr<PBQPRAConstraintList> ConstraintsRoot =839      std::make_unique<PBQPRAConstraintList>();840    ConstraintsRoot->addConstraint(std::make_unique<SpillCosts>());841    ConstraintsRoot->addConstraint(std::make_unique<Interference>());842    if (PBQPCoalescing)843      ConstraintsRoot->addConstraint(std::make_unique<Coalescing>());844    ConstraintsRoot->addConstraint(Subtarget.getCustomPBQPConstraints());845 846    bool PBQPAllocComplete = false;847    unsigned Round = 0;848 849    while (!PBQPAllocComplete) {850      LLVM_DEBUG(dbgs() << "  PBQP Regalloc round " << Round << ":\n");851      (void) Round;852 853      PBQPRAGraph G(PBQPRAGraph::GraphMetadata(MF, LIS, MBFI));854      initializeGraph(G, VRM, *VRegSpiller);855      ConstraintsRoot->apply(G);856 857#ifndef NDEBUG858      if (PBQPDumpGraphs) {859        std::ostringstream RS;860        RS << Round;861        std::string GraphFileName = FullyQualifiedName + "." + RS.str() +862                                    ".pbqpgraph";863        std::error_code EC;864        raw_fd_ostream OS(GraphFileName, EC, sys::fs::OF_TextWithCRLF);865        LLVM_DEBUG(dbgs() << "Dumping graph for round " << Round << " to \""866                          << GraphFileName << "\"\n");867        G.dump(OS);868      }869#endif870 871      PBQP::Solution Solution = PBQP::RegAlloc::solve(G);872      PBQPAllocComplete = mapPBQPToRegAlloc(G, Solution, VRM, *VRegSpiller);873      ++Round;874    }875  }876 877  // Finalise allocation, allocate empty ranges.878  finalizeAlloc(MF, LIS, VRM);879  postOptimization(*VRegSpiller, LIS);880  VRegsToAlloc.clear();881  EmptyIntervalVRegs.clear();882 883  LLVM_DEBUG(dbgs() << "Post alloc VirtRegMap:\n" << VRM << "\n");884 885  return true;886}887 888/// Create Printable object for node and register info.889static Printable PrintNodeInfo(PBQP::RegAlloc::PBQPRAGraph::NodeId NId,890                               const PBQP::RegAlloc::PBQPRAGraph &G) {891  return Printable([NId, &G](raw_ostream &OS) {892    const MachineRegisterInfo &MRI = G.getMetadata().MF.getRegInfo();893    const TargetRegisterInfo *TRI = MRI.getTargetRegisterInfo();894    Register VReg = G.getNodeMetadata(NId).getVReg();895    const char *RegClassName = TRI->getRegClassName(MRI.getRegClass(VReg));896    OS << NId << " (" << RegClassName << ':' << printReg(VReg, TRI) << ')';897  });898}899 900#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)901LLVM_DUMP_METHOD void PBQP::RegAlloc::PBQPRAGraph::dump(raw_ostream &OS) const {902  for (auto NId : nodeIds()) {903    const Vector &Costs = getNodeCosts(NId);904    assert(Costs.getLength() != 0 && "Empty vector in graph.");905    OS << PrintNodeInfo(NId, *this) << ": " << Costs << '\n';906  }907  OS << '\n';908 909  for (auto EId : edgeIds()) {910    NodeId N1Id = getEdgeNode1Id(EId);911    NodeId N2Id = getEdgeNode2Id(EId);912    assert(N1Id != N2Id && "PBQP graphs should not have self-edges.");913    const Matrix &M = getEdgeCosts(EId);914    assert(M.getRows() != 0 && "No rows in matrix.");915    assert(M.getCols() != 0 && "No cols in matrix.");916    OS << PrintNodeInfo(N1Id, *this) << ' ' << M.getRows() << " rows / ";917    OS << PrintNodeInfo(N2Id, *this) << ' ' << M.getCols() << " cols:\n";918    OS << M << '\n';919  }920}921 922LLVM_DUMP_METHOD void PBQP::RegAlloc::PBQPRAGraph::dump() const {923  dump(dbgs());924}925#endif926 927void PBQP::RegAlloc::PBQPRAGraph::printDot(raw_ostream &OS) const {928  OS << "graph {\n";929  for (auto NId : nodeIds()) {930    OS << "  node" << NId << " [ label=\""931       << PrintNodeInfo(NId, *this) << "\\n"932       << getNodeCosts(NId) << "\" ]\n";933  }934 935  OS << "  edge [ len=" << nodeIds().size() << " ]\n";936  for (auto EId : edgeIds()) {937    OS << "  node" << getEdgeNode1Id(EId)938       << " -- node" << getEdgeNode2Id(EId)939       << " [ label=\"";940    const Matrix &EdgeCosts = getEdgeCosts(EId);941    for (unsigned i = 0; i < EdgeCosts.getRows(); ++i) {942      OS << EdgeCosts.getRowAsVector(i) << "\\n";943    }944    OS << "\" ]\n";945  }946  OS << "}\n";947}948 949FunctionPass *llvm::createPBQPRegisterAllocator(char *customPassID) {950  return new RegAllocPBQP(customPassID);951}952 953FunctionPass* llvm::createDefaultPBQPRegisterAllocator() {954  return createPBQPRegisterAllocator();955}956