xref: /freebsd/contrib/llvm-project/llvm/lib/CodeGen/MachineCSE.cpp (revision 60d0871dbfbebe20a9e24c1b4c659cfe841154c1)
1  //===- MachineCSE.cpp - Machine Common Subexpression Elimination Pass -----===//
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-exception
6  //
7  //===----------------------------------------------------------------------===//
8  //
9  // This pass performs global common subexpression elimination on machine
10  // instructions using a scoped hash table based value numbering scheme. It
11  // must be run while the machine function is still in SSA form.
12  //
13  //===----------------------------------------------------------------------===//
14  
15  #include "llvm/ADT/DenseMap.h"
16  #include "llvm/ADT/ScopedHashTable.h"
17  #include "llvm/ADT/SmallPtrSet.h"
18  #include "llvm/ADT/SmallSet.h"
19  #include "llvm/ADT/SmallVector.h"
20  #include "llvm/ADT/Statistic.h"
21  #include "llvm/Analysis/AliasAnalysis.h"
22  #include "llvm/Analysis/CFG.h"
23  #include "llvm/CodeGen/MachineBasicBlock.h"
24  #include "llvm/CodeGen/MachineBlockFrequencyInfo.h"
25  #include "llvm/CodeGen/MachineDominators.h"
26  #include "llvm/CodeGen/MachineFunction.h"
27  #include "llvm/CodeGen/MachineFunctionPass.h"
28  #include "llvm/CodeGen/MachineInstr.h"
29  #include "llvm/CodeGen/MachineOperand.h"
30  #include "llvm/CodeGen/MachineRegisterInfo.h"
31  #include "llvm/CodeGen/Passes.h"
32  #include "llvm/CodeGen/TargetInstrInfo.h"
33  #include "llvm/CodeGen/TargetOpcodes.h"
34  #include "llvm/CodeGen/TargetRegisterInfo.h"
35  #include "llvm/CodeGen/TargetSubtargetInfo.h"
36  #include "llvm/InitializePasses.h"
37  #include "llvm/MC/MCRegister.h"
38  #include "llvm/MC/MCRegisterInfo.h"
39  #include "llvm/Pass.h"
40  #include "llvm/Support/Allocator.h"
41  #include "llvm/Support/Debug.h"
42  #include "llvm/Support/RecyclingAllocator.h"
43  #include "llvm/Support/raw_ostream.h"
44  #include <cassert>
45  #include <iterator>
46  #include <utility>
47  #include <vector>
48  
49  using namespace llvm;
50  
51  #define DEBUG_TYPE "machine-cse"
52  
53  STATISTIC(NumCoalesces, "Number of copies coalesced");
54  STATISTIC(NumCSEs,      "Number of common subexpression eliminated");
55  STATISTIC(NumPREs,      "Number of partial redundant expression"
56                          " transformed to fully redundant");
57  STATISTIC(NumPhysCSEs,
58            "Number of physreg referencing common subexpr eliminated");
59  STATISTIC(NumCrossBBCSEs,
60            "Number of cross-MBB physreg referencing CS eliminated");
61  STATISTIC(NumCommutes,  "Number of copies coalesced after commuting");
62  
63  namespace {
64  
65    class MachineCSE : public MachineFunctionPass {
66      const TargetInstrInfo *TII;
67      const TargetRegisterInfo *TRI;
68      AliasAnalysis *AA;
69      MachineDominatorTree *DT;
70      MachineRegisterInfo *MRI;
71      MachineBlockFrequencyInfo *MBFI;
72  
73    public:
74      static char ID; // Pass identification
75  
76      MachineCSE() : MachineFunctionPass(ID) {
77        initializeMachineCSEPass(*PassRegistry::getPassRegistry());
78      }
79  
80      bool runOnMachineFunction(MachineFunction &MF) override;
81  
82      void getAnalysisUsage(AnalysisUsage &AU) const override {
83        AU.setPreservesCFG();
84        MachineFunctionPass::getAnalysisUsage(AU);
85        AU.addRequired<AAResultsWrapperPass>();
86        AU.addPreservedID(MachineLoopInfoID);
87        AU.addRequired<MachineDominatorTree>();
88        AU.addPreserved<MachineDominatorTree>();
89        AU.addRequired<MachineBlockFrequencyInfo>();
90        AU.addPreserved<MachineBlockFrequencyInfo>();
91      }
92  
93      MachineFunctionProperties getRequiredProperties() const override {
94        return MachineFunctionProperties()
95          .set(MachineFunctionProperties::Property::IsSSA);
96      }
97  
98      void releaseMemory() override {
99        ScopeMap.clear();
100        PREMap.clear();
101        Exps.clear();
102      }
103  
104    private:
105      using AllocatorTy = RecyclingAllocator<BumpPtrAllocator,
106                              ScopedHashTableVal<MachineInstr *, unsigned>>;
107      using ScopedHTType =
108          ScopedHashTable<MachineInstr *, unsigned, MachineInstrExpressionTrait,
109                          AllocatorTy>;
110      using ScopeType = ScopedHTType::ScopeTy;
111      using PhysDefVector = SmallVector<std::pair<unsigned, unsigned>, 2>;
112  
113      unsigned LookAheadLimit = 0;
114      DenseMap<MachineBasicBlock *, ScopeType *> ScopeMap;
115      DenseMap<MachineInstr *, MachineBasicBlock *, MachineInstrExpressionTrait>
116          PREMap;
117      ScopedHTType VNT;
118      SmallVector<MachineInstr *, 64> Exps;
119      unsigned CurrVN = 0;
120  
121      bool PerformTrivialCopyPropagation(MachineInstr *MI,
122                                         MachineBasicBlock *MBB);
123      bool isPhysDefTriviallyDead(MCRegister Reg,
124                                  MachineBasicBlock::const_iterator I,
125                                  MachineBasicBlock::const_iterator E) const;
126      bool hasLivePhysRegDefUses(const MachineInstr *MI,
127                                 const MachineBasicBlock *MBB,
128                                 SmallSet<MCRegister, 8> &PhysRefs,
129                                 PhysDefVector &PhysDefs, bool &PhysUseDef) const;
130      bool PhysRegDefsReach(MachineInstr *CSMI, MachineInstr *MI,
131                            SmallSet<MCRegister, 8> &PhysRefs,
132                            PhysDefVector &PhysDefs, bool &NonLocal) const;
133      bool isCSECandidate(MachineInstr *MI);
134      bool isProfitableToCSE(Register CSReg, Register Reg,
135                             MachineBasicBlock *CSBB, MachineInstr *MI);
136      void EnterScope(MachineBasicBlock *MBB);
137      void ExitScope(MachineBasicBlock *MBB);
138      bool ProcessBlockCSE(MachineBasicBlock *MBB);
139      void ExitScopeIfDone(MachineDomTreeNode *Node,
140                           DenseMap<MachineDomTreeNode*, unsigned> &OpenChildren);
141      bool PerformCSE(MachineDomTreeNode *Node);
142  
143      bool isPRECandidate(MachineInstr *MI);
144      bool ProcessBlockPRE(MachineDominatorTree *MDT, MachineBasicBlock *MBB);
145      bool PerformSimplePRE(MachineDominatorTree *DT);
146      /// Heuristics to see if it's profitable to move common computations of MBB
147      /// and MBB1 to CandidateBB.
148      bool isProfitableToHoistInto(MachineBasicBlock *CandidateBB,
149                                   MachineBasicBlock *MBB,
150                                   MachineBasicBlock *MBB1);
151    };
152  
153  } // end anonymous namespace
154  
155  char MachineCSE::ID = 0;
156  
157  char &llvm::MachineCSEID = MachineCSE::ID;
158  
159  INITIALIZE_PASS_BEGIN(MachineCSE, DEBUG_TYPE,
160                        "Machine Common Subexpression Elimination", false, false)
161  INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree)
162  INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass)
163  INITIALIZE_PASS_END(MachineCSE, DEBUG_TYPE,
164                      "Machine Common Subexpression Elimination", false, false)
165  
166  /// The source register of a COPY machine instruction can be propagated to all
167  /// its users, and this propagation could increase the probability of finding
168  /// common subexpressions. If the COPY has only one user, the COPY itself can
169  /// be removed.
170  bool MachineCSE::PerformTrivialCopyPropagation(MachineInstr *MI,
171                                                 MachineBasicBlock *MBB) {
172    bool Changed = false;
173    for (MachineOperand &MO : MI->operands()) {
174      if (!MO.isReg() || !MO.isUse())
175        continue;
176      Register Reg = MO.getReg();
177      if (!Register::isVirtualRegister(Reg))
178        continue;
179      bool OnlyOneUse = MRI->hasOneNonDBGUse(Reg);
180      MachineInstr *DefMI = MRI->getVRegDef(Reg);
181      if (!DefMI->isCopy())
182        continue;
183      Register SrcReg = DefMI->getOperand(1).getReg();
184      if (!Register::isVirtualRegister(SrcReg))
185        continue;
186      if (DefMI->getOperand(0).getSubReg())
187        continue;
188      // FIXME: We should trivially coalesce subregister copies to expose CSE
189      // opportunities on instructions with truncated operands (see
190      // cse-add-with-overflow.ll). This can be done here as follows:
191      // if (SrcSubReg)
192      //  RC = TRI->getMatchingSuperRegClass(MRI->getRegClass(SrcReg), RC,
193      //                                     SrcSubReg);
194      // MO.substVirtReg(SrcReg, SrcSubReg, *TRI);
195      //
196      // The 2-addr pass has been updated to handle coalesced subregs. However,
197      // some machine-specific code still can't handle it.
198      // To handle it properly we also need a way find a constrained subregister
199      // class given a super-reg class and subreg index.
200      if (DefMI->getOperand(1).getSubReg())
201        continue;
202      if (!MRI->constrainRegAttrs(SrcReg, Reg))
203        continue;
204      LLVM_DEBUG(dbgs() << "Coalescing: " << *DefMI);
205      LLVM_DEBUG(dbgs() << "***     to: " << *MI);
206  
207      // Propagate SrcReg of copies to MI.
208      MO.setReg(SrcReg);
209      MRI->clearKillFlags(SrcReg);
210      // Coalesce single use copies.
211      if (OnlyOneUse) {
212        // If (and only if) we've eliminated all uses of the copy, also
213        // copy-propagate to any debug-users of MI, or they'll be left using
214        // an undefined value.
215        DefMI->changeDebugValuesDefReg(SrcReg);
216  
217        DefMI->eraseFromParent();
218        ++NumCoalesces;
219      }
220      Changed = true;
221    }
222  
223    return Changed;
224  }
225  
226  bool MachineCSE::isPhysDefTriviallyDead(
227      MCRegister Reg, MachineBasicBlock::const_iterator I,
228      MachineBasicBlock::const_iterator E) const {
229    unsigned LookAheadLeft = LookAheadLimit;
230    while (LookAheadLeft) {
231      // Skip over dbg_value's.
232      I = skipDebugInstructionsForward(I, E);
233  
234      if (I == E)
235        // Reached end of block, we don't know if register is dead or not.
236        return false;
237  
238      bool SeenDef = false;
239      for (const MachineOperand &MO : I->operands()) {
240        if (MO.isRegMask() && MO.clobbersPhysReg(Reg))
241          SeenDef = true;
242        if (!MO.isReg() || !MO.getReg())
243          continue;
244        if (!TRI->regsOverlap(MO.getReg(), Reg))
245          continue;
246        if (MO.isUse())
247          // Found a use!
248          return false;
249        SeenDef = true;
250      }
251      if (SeenDef)
252        // See a def of Reg (or an alias) before encountering any use, it's
253        // trivially dead.
254        return true;
255  
256      --LookAheadLeft;
257      ++I;
258    }
259    return false;
260  }
261  
262  static bool isCallerPreservedOrConstPhysReg(MCRegister Reg,
263                                              const MachineFunction &MF,
264                                              const TargetRegisterInfo &TRI) {
265    // MachineRegisterInfo::isConstantPhysReg directly called by
266    // MachineRegisterInfo::isCallerPreservedOrConstPhysReg expects the
267    // reserved registers to be frozen. That doesn't cause a problem  post-ISel as
268    // most (if not all) targets freeze reserved registers right after ISel.
269    //
270    // It does cause issues mid-GlobalISel, however, hence the additional
271    // reservedRegsFrozen check.
272    const MachineRegisterInfo &MRI = MF.getRegInfo();
273    return TRI.isCallerPreservedPhysReg(Reg, MF) ||
274           (MRI.reservedRegsFrozen() && MRI.isConstantPhysReg(Reg));
275  }
276  
277  /// hasLivePhysRegDefUses - Return true if the specified instruction read/write
278  /// physical registers (except for dead defs of physical registers). It also
279  /// returns the physical register def by reference if it's the only one and the
280  /// instruction does not uses a physical register.
281  bool MachineCSE::hasLivePhysRegDefUses(const MachineInstr *MI,
282                                         const MachineBasicBlock *MBB,
283                                         SmallSet<MCRegister, 8> &PhysRefs,
284                                         PhysDefVector &PhysDefs,
285                                         bool &PhysUseDef) const {
286    // First, add all uses to PhysRefs.
287    for (const MachineOperand &MO : MI->operands()) {
288      if (!MO.isReg() || MO.isDef())
289        continue;
290      Register Reg = MO.getReg();
291      if (!Reg)
292        continue;
293      if (Register::isVirtualRegister(Reg))
294        continue;
295      // Reading either caller preserved or constant physregs is ok.
296      if (!isCallerPreservedOrConstPhysReg(Reg.asMCReg(), *MI->getMF(), *TRI))
297        for (MCRegAliasIterator AI(Reg, TRI, true); AI.isValid(); ++AI)
298          PhysRefs.insert(*AI);
299    }
300  
301    // Next, collect all defs into PhysDefs.  If any is already in PhysRefs
302    // (which currently contains only uses), set the PhysUseDef flag.
303    PhysUseDef = false;
304    MachineBasicBlock::const_iterator I = MI; I = std::next(I);
305    for (const auto &MOP : llvm::enumerate(MI->operands())) {
306      const MachineOperand &MO = MOP.value();
307      if (!MO.isReg() || !MO.isDef())
308        continue;
309      Register Reg = MO.getReg();
310      if (!Reg)
311        continue;
312      if (Register::isVirtualRegister(Reg))
313        continue;
314      // Check against PhysRefs even if the def is "dead".
315      if (PhysRefs.count(Reg.asMCReg()))
316        PhysUseDef = true;
317      // If the def is dead, it's ok. But the def may not marked "dead". That's
318      // common since this pass is run before livevariables. We can scan
319      // forward a few instructions and check if it is obviously dead.
320      if (!MO.isDead() && !isPhysDefTriviallyDead(Reg.asMCReg(), I, MBB->end()))
321        PhysDefs.push_back(std::make_pair(MOP.index(), Reg));
322    }
323  
324    // Finally, add all defs to PhysRefs as well.
325    for (unsigned i = 0, e = PhysDefs.size(); i != e; ++i)
326      for (MCRegAliasIterator AI(PhysDefs[i].second, TRI, true); AI.isValid();
327           ++AI)
328        PhysRefs.insert(*AI);
329  
330    return !PhysRefs.empty();
331  }
332  
333  bool MachineCSE::PhysRegDefsReach(MachineInstr *CSMI, MachineInstr *MI,
334                                    SmallSet<MCRegister, 8> &PhysRefs,
335                                    PhysDefVector &PhysDefs,
336                                    bool &NonLocal) const {
337    // For now conservatively returns false if the common subexpression is
338    // not in the same basic block as the given instruction. The only exception
339    // is if the common subexpression is in the sole predecessor block.
340    const MachineBasicBlock *MBB = MI->getParent();
341    const MachineBasicBlock *CSMBB = CSMI->getParent();
342  
343    bool CrossMBB = false;
344    if (CSMBB != MBB) {
345      if (MBB->pred_size() != 1 || *MBB->pred_begin() != CSMBB)
346        return false;
347  
348      for (unsigned i = 0, e = PhysDefs.size(); i != e; ++i) {
349        if (MRI->isAllocatable(PhysDefs[i].second) ||
350            MRI->isReserved(PhysDefs[i].second))
351          // Avoid extending live range of physical registers if they are
352          //allocatable or reserved.
353          return false;
354      }
355      CrossMBB = true;
356    }
357    MachineBasicBlock::const_iterator I = CSMI; I = std::next(I);
358    MachineBasicBlock::const_iterator E = MI;
359    MachineBasicBlock::const_iterator EE = CSMBB->end();
360    unsigned LookAheadLeft = LookAheadLimit;
361    while (LookAheadLeft) {
362      // Skip over dbg_value's.
363      while (I != E && I != EE && I->isDebugInstr())
364        ++I;
365  
366      if (I == EE) {
367        assert(CrossMBB && "Reaching end-of-MBB without finding MI?");
368        (void)CrossMBB;
369        CrossMBB = false;
370        NonLocal = true;
371        I = MBB->begin();
372        EE = MBB->end();
373        continue;
374      }
375  
376      if (I == E)
377        return true;
378  
379      for (const MachineOperand &MO : I->operands()) {
380        // RegMasks go on instructions like calls that clobber lots of physregs.
381        // Don't attempt to CSE across such an instruction.
382        if (MO.isRegMask())
383          return false;
384        if (!MO.isReg() || !MO.isDef())
385          continue;
386        Register MOReg = MO.getReg();
387        if (Register::isVirtualRegister(MOReg))
388          continue;
389        if (PhysRefs.count(MOReg.asMCReg()))
390          return false;
391      }
392  
393      --LookAheadLeft;
394      ++I;
395    }
396  
397    return false;
398  }
399  
400  bool MachineCSE::isCSECandidate(MachineInstr *MI) {
401    if (MI->isPosition() || MI->isPHI() || MI->isImplicitDef() || MI->isKill() ||
402        MI->isInlineAsm() || MI->isDebugInstr())
403      return false;
404  
405    // Ignore copies.
406    if (MI->isCopyLike())
407      return false;
408  
409    // Ignore stuff that we obviously can't move.
410    if (MI->mayStore() || MI->isCall() || MI->isTerminator() ||
411        MI->mayRaiseFPException() || MI->hasUnmodeledSideEffects())
412      return false;
413  
414    if (MI->mayLoad()) {
415      // Okay, this instruction does a load. As a refinement, we allow the target
416      // to decide whether the loaded value is actually a constant. If so, we can
417      // actually use it as a load.
418      if (!MI->isDereferenceableInvariantLoad())
419        // FIXME: we should be able to hoist loads with no other side effects if
420        // there are no other instructions which can change memory in this loop.
421        // This is a trivial form of alias analysis.
422        return false;
423    }
424  
425    // Ignore stack guard loads, otherwise the register that holds CSEed value may
426    // be spilled and get loaded back with corrupted data.
427    if (MI->getOpcode() == TargetOpcode::LOAD_STACK_GUARD)
428      return false;
429  
430    return true;
431  }
432  
433  /// isProfitableToCSE - Return true if it's profitable to eliminate MI with a
434  /// common expression that defines Reg. CSBB is basic block where CSReg is
435  /// defined.
436  bool MachineCSE::isProfitableToCSE(Register CSReg, Register Reg,
437                                     MachineBasicBlock *CSBB, MachineInstr *MI) {
438    // FIXME: Heuristics that works around the lack the live range splitting.
439  
440    // If CSReg is used at all uses of Reg, CSE should not increase register
441    // pressure of CSReg.
442    bool MayIncreasePressure = true;
443    if (Register::isVirtualRegister(CSReg) && Register::isVirtualRegister(Reg)) {
444      MayIncreasePressure = false;
445      SmallPtrSet<MachineInstr*, 8> CSUses;
446      for (MachineInstr &MI : MRI->use_nodbg_instructions(CSReg)) {
447        CSUses.insert(&MI);
448      }
449      for (MachineInstr &MI : MRI->use_nodbg_instructions(Reg)) {
450        if (!CSUses.count(&MI)) {
451          MayIncreasePressure = true;
452          break;
453        }
454      }
455    }
456    if (!MayIncreasePressure) return true;
457  
458    // Heuristics #1: Don't CSE "cheap" computation if the def is not local or in
459    // an immediate predecessor. We don't want to increase register pressure and
460    // end up causing other computation to be spilled.
461    if (TII->isAsCheapAsAMove(*MI)) {
462      MachineBasicBlock *BB = MI->getParent();
463      if (CSBB != BB && !CSBB->isSuccessor(BB))
464        return false;
465    }
466  
467    // Heuristics #2: If the expression doesn't not use a vr and the only use
468    // of the redundant computation are copies, do not cse.
469    bool HasVRegUse = false;
470    for (const MachineOperand &MO : MI->operands()) {
471      if (MO.isReg() && MO.isUse() && Register::isVirtualRegister(MO.getReg())) {
472        HasVRegUse = true;
473        break;
474      }
475    }
476    if (!HasVRegUse) {
477      bool HasNonCopyUse = false;
478      for (MachineInstr &MI : MRI->use_nodbg_instructions(Reg)) {
479        // Ignore copies.
480        if (!MI.isCopyLike()) {
481          HasNonCopyUse = true;
482          break;
483        }
484      }
485      if (!HasNonCopyUse)
486        return false;
487    }
488  
489    // Heuristics #3: If the common subexpression is used by PHIs, do not reuse
490    // it unless the defined value is already used in the BB of the new use.
491    bool HasPHI = false;
492    for (MachineInstr &UseMI : MRI->use_nodbg_instructions(CSReg)) {
493      HasPHI |= UseMI.isPHI();
494      if (UseMI.getParent() == MI->getParent())
495        return true;
496    }
497  
498    return !HasPHI;
499  }
500  
501  void MachineCSE::EnterScope(MachineBasicBlock *MBB) {
502    LLVM_DEBUG(dbgs() << "Entering: " << MBB->getName() << '\n');
503    ScopeType *Scope = new ScopeType(VNT);
504    ScopeMap[MBB] = Scope;
505  }
506  
507  void MachineCSE::ExitScope(MachineBasicBlock *MBB) {
508    LLVM_DEBUG(dbgs() << "Exiting: " << MBB->getName() << '\n');
509    DenseMap<MachineBasicBlock*, ScopeType*>::iterator SI = ScopeMap.find(MBB);
510    assert(SI != ScopeMap.end());
511    delete SI->second;
512    ScopeMap.erase(SI);
513  }
514  
515  bool MachineCSE::ProcessBlockCSE(MachineBasicBlock *MBB) {
516    bool Changed = false;
517  
518    SmallVector<std::pair<unsigned, unsigned>, 8> CSEPairs;
519    SmallVector<unsigned, 2> ImplicitDefsToUpdate;
520    SmallVector<unsigned, 2> ImplicitDefs;
521    for (MachineInstr &MI : llvm::make_early_inc_range(*MBB)) {
522      if (!isCSECandidate(&MI))
523        continue;
524  
525      bool FoundCSE = VNT.count(&MI);
526      if (!FoundCSE) {
527        // Using trivial copy propagation to find more CSE opportunities.
528        if (PerformTrivialCopyPropagation(&MI, MBB)) {
529          Changed = true;
530  
531          // After coalescing MI itself may become a copy.
532          if (MI.isCopyLike())
533            continue;
534  
535          // Try again to see if CSE is possible.
536          FoundCSE = VNT.count(&MI);
537        }
538      }
539  
540      // Commute commutable instructions.
541      bool Commuted = false;
542      if (!FoundCSE && MI.isCommutable()) {
543        if (MachineInstr *NewMI = TII->commuteInstruction(MI)) {
544          Commuted = true;
545          FoundCSE = VNT.count(NewMI);
546          if (NewMI != &MI) {
547            // New instruction. It doesn't need to be kept.
548            NewMI->eraseFromParent();
549            Changed = true;
550          } else if (!FoundCSE)
551            // MI was changed but it didn't help, commute it back!
552            (void)TII->commuteInstruction(MI);
553        }
554      }
555  
556      // If the instruction defines physical registers and the values *may* be
557      // used, then it's not safe to replace it with a common subexpression.
558      // It's also not safe if the instruction uses physical registers.
559      bool CrossMBBPhysDef = false;
560      SmallSet<MCRegister, 8> PhysRefs;
561      PhysDefVector PhysDefs;
562      bool PhysUseDef = false;
563      if (FoundCSE &&
564          hasLivePhysRegDefUses(&MI, MBB, PhysRefs, PhysDefs, PhysUseDef)) {
565        FoundCSE = false;
566  
567        // ... Unless the CS is local or is in the sole predecessor block
568        // and it also defines the physical register which is not clobbered
569        // in between and the physical register uses were not clobbered.
570        // This can never be the case if the instruction both uses and
571        // defines the same physical register, which was detected above.
572        if (!PhysUseDef) {
573          unsigned CSVN = VNT.lookup(&MI);
574          MachineInstr *CSMI = Exps[CSVN];
575          if (PhysRegDefsReach(CSMI, &MI, PhysRefs, PhysDefs, CrossMBBPhysDef))
576            FoundCSE = true;
577        }
578      }
579  
580      if (!FoundCSE) {
581        VNT.insert(&MI, CurrVN++);
582        Exps.push_back(&MI);
583        continue;
584      }
585  
586      // Found a common subexpression, eliminate it.
587      unsigned CSVN = VNT.lookup(&MI);
588      MachineInstr *CSMI = Exps[CSVN];
589      LLVM_DEBUG(dbgs() << "Examining: " << MI);
590      LLVM_DEBUG(dbgs() << "*** Found a common subexpression: " << *CSMI);
591  
592      // Prevent CSE-ing non-local convergent instructions.
593      // LLVM's current definition of `isConvergent` does not necessarily prove
594      // that non-local CSE is illegal. The following check extends the definition
595      // of `isConvergent` to assume a convergent instruction is dependent not
596      // only on additional conditions, but also on fewer conditions. LLVM does
597      // not have a MachineInstr attribute which expresses this extended
598      // definition, so it's necessary to use `isConvergent` to prevent illegally
599      // CSE-ing the subset of `isConvergent` instructions which do fall into this
600      // extended definition.
601      if (MI.isConvergent() && MI.getParent() != CSMI->getParent()) {
602        LLVM_DEBUG(dbgs() << "*** Convergent MI and subexpression exist in "
603                             "different BBs, avoid CSE!\n");
604        VNT.insert(&MI, CurrVN++);
605        Exps.push_back(&MI);
606        continue;
607      }
608  
609      // Check if it's profitable to perform this CSE.
610      bool DoCSE = true;
611      unsigned NumDefs = MI.getNumDefs();
612  
613      for (unsigned i = 0, e = MI.getNumOperands(); NumDefs && i != e; ++i) {
614        MachineOperand &MO = MI.getOperand(i);
615        if (!MO.isReg() || !MO.isDef())
616          continue;
617        Register OldReg = MO.getReg();
618        Register NewReg = CSMI->getOperand(i).getReg();
619  
620        // Go through implicit defs of CSMI and MI, if a def is not dead at MI,
621        // we should make sure it is not dead at CSMI.
622        if (MO.isImplicit() && !MO.isDead() && CSMI->getOperand(i).isDead())
623          ImplicitDefsToUpdate.push_back(i);
624  
625        // Keep track of implicit defs of CSMI and MI, to clear possibly
626        // made-redundant kill flags.
627        if (MO.isImplicit() && !MO.isDead() && OldReg == NewReg)
628          ImplicitDefs.push_back(OldReg);
629  
630        if (OldReg == NewReg) {
631          --NumDefs;
632          continue;
633        }
634  
635        assert(Register::isVirtualRegister(OldReg) &&
636               Register::isVirtualRegister(NewReg) &&
637               "Do not CSE physical register defs!");
638  
639        if (!isProfitableToCSE(NewReg, OldReg, CSMI->getParent(), &MI)) {
640          LLVM_DEBUG(dbgs() << "*** Not profitable, avoid CSE!\n");
641          DoCSE = false;
642          break;
643        }
644  
645        // Don't perform CSE if the result of the new instruction cannot exist
646        // within the constraints (register class, bank, or low-level type) of
647        // the old instruction.
648        if (!MRI->constrainRegAttrs(NewReg, OldReg)) {
649          LLVM_DEBUG(
650              dbgs() << "*** Not the same register constraints, avoid CSE!\n");
651          DoCSE = false;
652          break;
653        }
654  
655        CSEPairs.push_back(std::make_pair(OldReg, NewReg));
656        --NumDefs;
657      }
658  
659      // Actually perform the elimination.
660      if (DoCSE) {
661        for (const std::pair<unsigned, unsigned> &CSEPair : CSEPairs) {
662          unsigned OldReg = CSEPair.first;
663          unsigned NewReg = CSEPair.second;
664          // OldReg may have been unused but is used now, clear the Dead flag
665          MachineInstr *Def = MRI->getUniqueVRegDef(NewReg);
666          assert(Def != nullptr && "CSEd register has no unique definition?");
667          Def->clearRegisterDeads(NewReg);
668          // Replace with NewReg and clear kill flags which may be wrong now.
669          MRI->replaceRegWith(OldReg, NewReg);
670          MRI->clearKillFlags(NewReg);
671        }
672  
673        // Go through implicit defs of CSMI and MI, if a def is not dead at MI,
674        // we should make sure it is not dead at CSMI.
675        for (unsigned ImplicitDefToUpdate : ImplicitDefsToUpdate)
676          CSMI->getOperand(ImplicitDefToUpdate).setIsDead(false);
677        for (const auto &PhysDef : PhysDefs)
678          if (!MI.getOperand(PhysDef.first).isDead())
679            CSMI->getOperand(PhysDef.first).setIsDead(false);
680  
681        // Go through implicit defs of CSMI and MI, and clear the kill flags on
682        // their uses in all the instructions between CSMI and MI.
683        // We might have made some of the kill flags redundant, consider:
684        //   subs  ... implicit-def %nzcv    <- CSMI
685        //   csinc ... implicit killed %nzcv <- this kill flag isn't valid anymore
686        //   subs  ... implicit-def %nzcv    <- MI, to be eliminated
687        //   csinc ... implicit killed %nzcv
688        // Since we eliminated MI, and reused a register imp-def'd by CSMI
689        // (here %nzcv), that register, if it was killed before MI, should have
690        // that kill flag removed, because it's lifetime was extended.
691        if (CSMI->getParent() == MI.getParent()) {
692          for (MachineBasicBlock::iterator II = CSMI, IE = &MI; II != IE; ++II)
693            for (auto ImplicitDef : ImplicitDefs)
694              if (MachineOperand *MO = II->findRegisterUseOperand(
695                      ImplicitDef, /*isKill=*/true, TRI))
696                MO->setIsKill(false);
697        } else {
698          // If the instructions aren't in the same BB, bail out and clear the
699          // kill flag on all uses of the imp-def'd register.
700          for (auto ImplicitDef : ImplicitDefs)
701            MRI->clearKillFlags(ImplicitDef);
702        }
703  
704        if (CrossMBBPhysDef) {
705          // Add physical register defs now coming in from a predecessor to MBB
706          // livein list.
707          while (!PhysDefs.empty()) {
708            auto LiveIn = PhysDefs.pop_back_val();
709            if (!MBB->isLiveIn(LiveIn.second))
710              MBB->addLiveIn(LiveIn.second);
711          }
712          ++NumCrossBBCSEs;
713        }
714  
715        MI.eraseFromParent();
716        ++NumCSEs;
717        if (!PhysRefs.empty())
718          ++NumPhysCSEs;
719        if (Commuted)
720          ++NumCommutes;
721        Changed = true;
722      } else {
723        VNT.insert(&MI, CurrVN++);
724        Exps.push_back(&MI);
725      }
726      CSEPairs.clear();
727      ImplicitDefsToUpdate.clear();
728      ImplicitDefs.clear();
729    }
730  
731    return Changed;
732  }
733  
734  /// ExitScopeIfDone - Destroy scope for the MBB that corresponds to the given
735  /// dominator tree node if its a leaf or all of its children are done. Walk
736  /// up the dominator tree to destroy ancestors which are now done.
737  void
738  MachineCSE::ExitScopeIfDone(MachineDomTreeNode *Node,
739                          DenseMap<MachineDomTreeNode*, unsigned> &OpenChildren) {
740    if (OpenChildren[Node])
741      return;
742  
743    // Pop scope.
744    ExitScope(Node->getBlock());
745  
746    // Now traverse upwards to pop ancestors whose offsprings are all done.
747    while (MachineDomTreeNode *Parent = Node->getIDom()) {
748      unsigned Left = --OpenChildren[Parent];
749      if (Left != 0)
750        break;
751      ExitScope(Parent->getBlock());
752      Node = Parent;
753    }
754  }
755  
756  bool MachineCSE::PerformCSE(MachineDomTreeNode *Node) {
757    SmallVector<MachineDomTreeNode*, 32> Scopes;
758    SmallVector<MachineDomTreeNode*, 8> WorkList;
759    DenseMap<MachineDomTreeNode*, unsigned> OpenChildren;
760  
761    CurrVN = 0;
762  
763    // Perform a DFS walk to determine the order of visit.
764    WorkList.push_back(Node);
765    do {
766      Node = WorkList.pop_back_val();
767      Scopes.push_back(Node);
768      OpenChildren[Node] = Node->getNumChildren();
769      append_range(WorkList, Node->children());
770    } while (!WorkList.empty());
771  
772    // Now perform CSE.
773    bool Changed = false;
774    for (MachineDomTreeNode *Node : Scopes) {
775      MachineBasicBlock *MBB = Node->getBlock();
776      EnterScope(MBB);
777      Changed |= ProcessBlockCSE(MBB);
778      // If it's a leaf node, it's done. Traverse upwards to pop ancestors.
779      ExitScopeIfDone(Node, OpenChildren);
780    }
781  
782    return Changed;
783  }
784  
785  // We use stronger checks for PRE candidate rather than for CSE ones to embrace
786  // checks inside ProcessBlockCSE(), not only inside isCSECandidate(). This helps
787  // to exclude instrs created by PRE that won't be CSEed later.
788  bool MachineCSE::isPRECandidate(MachineInstr *MI) {
789    if (!isCSECandidate(MI) ||
790        MI->isNotDuplicable() ||
791        MI->mayLoad() ||
792        MI->isAsCheapAsAMove() ||
793        MI->getNumDefs() != 1 ||
794        MI->getNumExplicitDefs() != 1)
795      return false;
796  
797    for (const auto &def : MI->defs())
798      if (!Register::isVirtualRegister(def.getReg()))
799        return false;
800  
801    for (const auto &use : MI->uses())
802      if (use.isReg() && !Register::isVirtualRegister(use.getReg()))
803        return false;
804  
805    return true;
806  }
807  
808  bool MachineCSE::ProcessBlockPRE(MachineDominatorTree *DT,
809                                   MachineBasicBlock *MBB) {
810    bool Changed = false;
811    for (MachineInstr &MI : llvm::make_early_inc_range(*MBB)) {
812      if (!isPRECandidate(&MI))
813        continue;
814  
815      if (!PREMap.count(&MI)) {
816        PREMap[&MI] = MBB;
817        continue;
818      }
819  
820      auto MBB1 = PREMap[&MI];
821      assert(
822          !DT->properlyDominates(MBB, MBB1) &&
823          "MBB cannot properly dominate MBB1 while DFS through dominators tree!");
824      auto CMBB = DT->findNearestCommonDominator(MBB, MBB1);
825      if (!CMBB->isLegalToHoistInto())
826        continue;
827  
828      if (!isProfitableToHoistInto(CMBB, MBB, MBB1))
829        continue;
830  
831      // Two instrs are partial redundant if their basic blocks are reachable
832      // from one to another but one doesn't dominate another.
833      if (CMBB != MBB1) {
834        auto BB = MBB->getBasicBlock(), BB1 = MBB1->getBasicBlock();
835        if (BB != nullptr && BB1 != nullptr &&
836            (isPotentiallyReachable(BB1, BB) ||
837             isPotentiallyReachable(BB, BB1))) {
838          // The following check extends the definition of `isConvergent` to
839          // assume a convergent instruction is dependent not only on additional
840          // conditions, but also on fewer conditions. LLVM does not have a
841          // MachineInstr attribute which expresses this extended definition, so
842          // it's necessary to use `isConvergent` to prevent illegally PRE-ing the
843          // subset of `isConvergent` instructions which do fall into this
844          // extended definition.
845          if (MI.isConvergent() && CMBB != MBB)
846            continue;
847  
848          assert(MI.getOperand(0).isDef() &&
849                 "First operand of instr with one explicit def must be this def");
850          Register VReg = MI.getOperand(0).getReg();
851          Register NewReg = MRI->cloneVirtualRegister(VReg);
852          if (!isProfitableToCSE(NewReg, VReg, CMBB, &MI))
853            continue;
854          MachineInstr &NewMI =
855              TII->duplicate(*CMBB, CMBB->getFirstTerminator(), MI);
856  
857          // When hoisting, make sure we don't carry the debug location of
858          // the original instruction, as that's not correct and can cause
859          // unexpected jumps when debugging optimized code.
860          auto EmptyDL = DebugLoc();
861          NewMI.setDebugLoc(EmptyDL);
862  
863          NewMI.getOperand(0).setReg(NewReg);
864  
865          PREMap[&MI] = CMBB;
866          ++NumPREs;
867          Changed = true;
868        }
869      }
870    }
871    return Changed;
872  }
873  
874  // This simple PRE (partial redundancy elimination) pass doesn't actually
875  // eliminate partial redundancy but transforms it to full redundancy,
876  // anticipating that the next CSE step will eliminate this created redundancy.
877  // If CSE doesn't eliminate this, than created instruction will remain dead
878  // and eliminated later by Remove Dead Machine Instructions pass.
879  bool MachineCSE::PerformSimplePRE(MachineDominatorTree *DT) {
880    SmallVector<MachineDomTreeNode *, 32> BBs;
881  
882    PREMap.clear();
883    bool Changed = false;
884    BBs.push_back(DT->getRootNode());
885    do {
886      auto Node = BBs.pop_back_val();
887      append_range(BBs, Node->children());
888  
889      MachineBasicBlock *MBB = Node->getBlock();
890      Changed |= ProcessBlockPRE(DT, MBB);
891  
892    } while (!BBs.empty());
893  
894    return Changed;
895  }
896  
897  bool MachineCSE::isProfitableToHoistInto(MachineBasicBlock *CandidateBB,
898                                           MachineBasicBlock *MBB,
899                                           MachineBasicBlock *MBB1) {
900    if (CandidateBB->getParent()->getFunction().hasMinSize())
901      return true;
902    assert(DT->dominates(CandidateBB, MBB) && "CandidateBB should dominate MBB");
903    assert(DT->dominates(CandidateBB, MBB1) &&
904           "CandidateBB should dominate MBB1");
905    return MBFI->getBlockFreq(CandidateBB) <=
906           MBFI->getBlockFreq(MBB) + MBFI->getBlockFreq(MBB1);
907  }
908  
909  bool MachineCSE::runOnMachineFunction(MachineFunction &MF) {
910    if (skipFunction(MF.getFunction()))
911      return false;
912  
913    TII = MF.getSubtarget().getInstrInfo();
914    TRI = MF.getSubtarget().getRegisterInfo();
915    MRI = &MF.getRegInfo();
916    AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();
917    DT = &getAnalysis<MachineDominatorTree>();
918    MBFI = &getAnalysis<MachineBlockFrequencyInfo>();
919    LookAheadLimit = TII->getMachineCSELookAheadLimit();
920    bool ChangedPRE, ChangedCSE;
921    ChangedPRE = PerformSimplePRE(DT);
922    ChangedCSE = PerformCSE(DT->getRootNode());
923    return ChangedPRE || ChangedCSE;
924  }
925