C++ ScalarEvolution类代码示例

bool polly::isHoistableLoad(LoadInst *LInst, Region &R, LoopInfo &LI,
                            ScalarEvolution &SE, const DominatorTree &DT) {
  Loop *L = LI.getLoopFor(LInst->getParent());
  auto *Ptr = LInst->getPointerOperand();
  const SCEV *PtrSCEV = SE.getSCEVAtScope(Ptr, L);
  while (L && R.contains(L)) {
    if (!SE.isLoopInvariant(PtrSCEV, L))
      return false;
    L = L->getParentLoop();
  }

  for (auto *User : Ptr->users()) {
    auto *UserI = dyn_cast<Instruction>(User);
    if (!UserI || !R.contains(UserI))
      continue;
    if (!UserI->mayWriteToMemory())
      continue;

    auto &BB = *UserI->getParent();
    bool DominatesAllPredecessors = true;
    for (auto Pred : predecessors(R.getExit()))
      if (R.contains(Pred) && !DT.dominates(&BB, Pred))
        DominatesAllPredecessors = false;

    if (!DominatesAllPredecessors)
      continue;

    return false;
  }

  return true;
}

void IslNodeBuilder::createSubstitutions(isl_ast_expr *Expr, ScopStmt *Stmt,
                                         ValueMapT &VMap, LoopToScevMapT &LTS) {
  assert(isl_ast_expr_get_type(Expr) == isl_ast_expr_op &&
         "Expression of type 'op' expected");
  assert(isl_ast_expr_get_op_type(Expr) == isl_ast_op_call &&
         "Opertation of type 'call' expected");
  for (int i = 0; i < isl_ast_expr_get_op_n_arg(Expr) - 1; ++i) {
    isl_ast_expr *SubExpr;
    Value *V;

    SubExpr = isl_ast_expr_get_op_arg(Expr, i + 1);
    V = ExprBuilder.create(SubExpr);
    ScalarEvolution *SE = Stmt->getParent()->getSE();
    LTS[Stmt->getLoopForDimension(i)] = SE->getUnknown(V);

    // CreateIntCast can introduce trunc expressions. This is correct, as the
    // result will always fit into the type of the original induction variable
    // (because we calculate a value of the original induction variable).
    const Value *OldIV = Stmt->getInductionVariableForDimension(i);
    if (OldIV) {
      V = Builder.CreateIntCast(V, OldIV->getType(), true);
      VMap[OldIV] = V;
    }
  }

  isl_ast_expr_free(Expr);
}

void IslNodeBuilder::createSubstitutions(
    __isl_take isl_pw_multi_aff *PMA, __isl_take isl_ast_build *Context,
    ScopStmt *Stmt, ValueMapT &VMap, LoopToScevMapT &LTS) {
  for (unsigned i = 0; i < isl_pw_multi_aff_dim(PMA, isl_dim_out); ++i) {
    isl_pw_aff *Aff;
    isl_ast_expr *Expr;
    const Value *OldIV;
    Value *V;

    Aff = isl_pw_multi_aff_get_pw_aff(PMA, i);
    Expr = isl_ast_build_expr_from_pw_aff(Context, Aff);
    OldIV = Stmt->getInductionVariableForDimension(i);
    V = ExprBuilder.create(Expr);

    // CreateIntCast can introduce trunc expressions. This is correct, as the
    // result will always fit into the type of the original induction variable
    // (because we calculate a value of the original induction variable).
    V = Builder.CreateIntCast(V, OldIV->getType(), true);
    VMap[OldIV] = V;
    ScalarEvolution *SE = Stmt->getParent()->getSE();
    LTS[Stmt->getLoopForDimension(i)] = SE->getUnknown(V);
  }

  isl_pw_multi_aff_free(PMA);
  isl_ast_build_free(Context);
}

bool llvm::cannotBeMaxInLoop(const SCEV *S, const Loop *L, ScalarEvolution &SE,
                             bool Signed) {
  unsigned BitWidth = cast<IntegerType>(S->getType())->getBitWidth();
  APInt Max = Signed ? APInt::getSignedMaxValue(BitWidth) :
    APInt::getMaxValue(BitWidth);
  auto Predicate = Signed ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT;
  return SE.isAvailableAtLoopEntry(S, L) &&
         SE.isLoopEntryGuardedByCond(L, Predicate, S,
                                     SE.getConstant(Max));
}

static bool simplifyLoopCFG(Loop &L, DominatorTree &DT, LoopInfo &LI,
                            ScalarEvolution &SE) {
  bool Changed = false;
  // Copy blocks into a temporary array to avoid iterator invalidation issues
  // as we remove them.
  SmallVector<WeakTrackingVH, 16> Blocks(L.blocks());

  for (auto &Block : Blocks) {
    // Attempt to merge blocks in the trivial case. Don't modify blocks which
    // belong to other loops.
    BasicBlock *Succ = cast_or_null<BasicBlock>(Block);
    if (!Succ)
      continue;

    BasicBlock *Pred = Succ->getSinglePredecessor();
    if (!Pred || !Pred->getSingleSuccessor() || LI.getLoopFor(Pred) != &L)
      continue;

    // Merge Succ into Pred and delete it.
    MergeBlockIntoPredecessor(Succ, &DT, &LI);

    SE.forgetLoop(&L);
    Changed = true;
  }

  return Changed;
}

/// isLoopDead - Determined if a loop is dead.  This assumes that we've already
/// checked for unique exit and exiting blocks, and that the code is in LCSSA
/// form.
bool LoopDeletion::isLoopDead(Loop *L, ScalarEvolution &SE,
                              SmallVectorImpl<BasicBlock *> &exitingBlocks,
                              SmallVectorImpl<BasicBlock *> &exitBlocks,
                              bool &Changed, BasicBlock *Preheader) {
  BasicBlock *exitBlock = exitBlocks[0];

  // Make sure that all PHI entries coming from the loop are loop invariant.
  // Because the code is in LCSSA form, any values used outside of the loop
  // must pass through a PHI in the exit block, meaning that this check is
  // sufficient to guarantee that no loop-variant values are used outside
  // of the loop.
  BasicBlock::iterator BI = exitBlock->begin();
  bool AllEntriesInvariant = true;
  bool AllOutgoingValuesSame = true;
  while (PHINode *P = dyn_cast<PHINode>(BI)) {
    Value *incoming = P->getIncomingValueForBlock(exitingBlocks[0]);

    // Make sure all exiting blocks produce the same incoming value for the exit
    // block.  If there are different incoming values for different exiting
    // blocks, then it is impossible to statically determine which value should
    // be used.
    AllOutgoingValuesSame =
        all_of(makeArrayRef(exitingBlocks).slice(1), [&](BasicBlock *BB) {
          return incoming == P->getIncomingValueForBlock(BB);
        });

    if (!AllOutgoingValuesSame)
      break;

    if (Instruction *I = dyn_cast<Instruction>(incoming))
      if (!L->makeLoopInvariant(I, Changed, Preheader->getTerminator())) {
        AllEntriesInvariant = false;
        break;
      }

    ++BI;
  }

  if (Changed)
    SE.forgetLoopDispositions(L);

  if (!AllEntriesInvariant || !AllOutgoingValuesSame)
    return false;

  // Make sure that no instructions in the block have potential side-effects.
  // This includes instructions that could write to memory, and loads that are
  // marked volatile.  This could be made more aggressive by using aliasing
  // information to identify readonly and readnone calls.
  for (Loop::block_iterator LI = L->block_begin(), LE = L->block_end();
       LI != LE; ++LI) {
    for (BasicBlock::iterator BI = (*LI)->begin(), BE = (*LI)->end();
         BI != BE; ++BI) {
      if (BI->mayHaveSideEffects())
        return false;
    }
  }

  return true;
}

bool llvm::hasIterationCountInvariantInParent(Loop *InnerLoop,
                                              ScalarEvolution &SE) {
  Loop *OuterL = InnerLoop->getParentLoop();
  if (!OuterL)
    return true;

  // Get the backedge taken count for the inner loop
  BasicBlock *InnerLoopLatch = InnerLoop->getLoopLatch();
  const SCEV *InnerLoopBECountSC = SE.getExitCount(InnerLoop, InnerLoopLatch);
  if (isa<SCEVCouldNotCompute>(InnerLoopBECountSC) ||
      !InnerLoopBECountSC->getType()->isIntegerTy())
    return false;

  // Get whether count is invariant to the outer loop
  ScalarEvolution::LoopDisposition LD =
      SE.getLoopDisposition(InnerLoopBECountSC, OuterL);
  if (LD != ScalarEvolution::LoopInvariant)
    return false;

  return true;
}

/// Determines if a loop is dead.
///
/// This assumes that we've already checked for unique exit and exiting blocks,
/// and that the code is in LCSSA form.
static bool isLoopDead(Loop *L, ScalarEvolution &SE,
                       SmallVectorImpl<BasicBlock *> &ExitingBlocks,
                       BasicBlock *ExitBlock, bool &Changed,
                       BasicBlock *Preheader) {
  // Make sure that all PHI entries coming from the loop are loop invariant.
  // Because the code is in LCSSA form, any values used outside of the loop
  // must pass through a PHI in the exit block, meaning that this check is
  // sufficient to guarantee that no loop-variant values are used outside
  // of the loop.
  BasicBlock::iterator BI = ExitBlock->begin();
  bool AllEntriesInvariant = true;
  bool AllOutgoingValuesSame = true;
  while (PHINode *P = dyn_cast<PHINode>(BI)) {
    Value *incoming = P->getIncomingValueForBlock(ExitingBlocks[0]);

    // Make sure all exiting blocks produce the same incoming value for the exit
    // block.  If there are different incoming values for different exiting
    // blocks, then it is impossible to statically determine which value should
    // be used.
    AllOutgoingValuesSame =
        all_of(makeArrayRef(ExitingBlocks).slice(1), [&](BasicBlock *BB) {
          return incoming == P->getIncomingValueForBlock(BB);
        });

    if (!AllOutgoingValuesSame)
      break;

    if (Instruction *I = dyn_cast<Instruction>(incoming))
      if (!L->makeLoopInvariant(I, Changed, Preheader->getTerminator())) {
        AllEntriesInvariant = false;
        break;
      }

    ++BI;
  }

  if (Changed)
    SE.forgetLoopDispositions(L);

  if (!AllEntriesInvariant || !AllOutgoingValuesSame)
    return false;

  // Make sure that no instructions in the block have potential side-effects.
  // This includes instructions that could write to memory, and loads that are
  // marked volatile.
  for (auto &I : L->blocks())
    if (any_of(*I, [](Instruction &I) { return I.mayHaveSideEffects(); }))
      return false;
  return true;
}

static bool simplifyLoopCFG(Loop &L, DominatorTree &DT, LoopInfo &LI,
                            ScalarEvolution &SE, MemorySSAUpdater *MSSAU) {
  bool Changed = false;

  // Constant-fold terminators with known constant conditions.
  Changed |= constantFoldTerminators(L, DT, LI, SE, MSSAU);

  // Eliminate unconditional branches by merging blocks into their predecessors.
  Changed |= mergeBlocksIntoPredecessors(L, DT, LI, MSSAU);

  if (Changed)
    SE.forgetTopmostLoop(&L);

  return Changed;
}

      // If one loop has very large self trip count
      // we don't want to unroll it.
      // self trip count means trip count divide by the parent's trip count. for example
      // for (int i = 0; i < 16; i++) {
      //   for (int j = 0; j < 4; j++) {
      //     for (int k = 0; k < 2; k++) {
      //       ...
      //     }
      //     ...
      //   }
      // The inner loops j and k could be unrolled, but the loop i will not be unrolled.
      // The return value true means the L could be unrolled, otherwise, it could not
      // be unrolled.
      bool handleParentLoops(Loop *L, LPPassManager &LPM) {
        Loop *currL = L;
        ScalarEvolution *SE = &getAnalysis<ScalarEvolution>();
        BasicBlock *ExitBlock = currL->getLoopLatch();
        if (!ExitBlock || !L->isLoopExiting(ExitBlock))
          ExitBlock = currL->getExitingBlock();

        unsigned currTripCount = 0;
        bool shouldUnroll = true;
        if (ExitBlock)
          currTripCount = SE->getSmallConstantTripCount(L, ExitBlock);

        while(currL) {
          Loop *parentL = currL->getParentLoop();
          unsigned parentTripCount = 0;
          if (parentL) {
            BasicBlock *parentExitBlock = parentL->getLoopLatch();
            if (!parentExitBlock || !parentL->isLoopExiting(parentExitBlock))
              parentExitBlock = parentL->getExitingBlock();

            if (parentExitBlock)
              parentTripCount = SE->getSmallConstantTripCount(parentL, parentExitBlock);
          }
          if ((parentTripCount != 0 && currTripCount / parentTripCount > 16) ||
              (currTripCount > 32)) {
            if (currL == L)
              shouldUnroll = false;
            setUnrollID(currL, false);
            if (currL != L)
              LPM.deleteLoopFromQueue(currL);
          }
          currL = parentL;
          currTripCount = parentTripCount;
        }
        return shouldUnroll;
      }

std::tuple<std::vector<const SCEV *>, std::vector<int>>
polly::getIndexExpressionsFromGEP(GetElementPtrInst *GEP, ScalarEvolution &SE) {
  std::vector<const SCEV *> Subscripts;
  std::vector<int> Sizes;

  Type *Ty = GEP->getPointerOperandType();

  bool DroppedFirstDim = false;

  for (unsigned i = 1; i < GEP->getNumOperands(); i++) {

    const SCEV *Expr = SE.getSCEV(GEP->getOperand(i));

    if (i == 1) {
      if (auto *PtrTy = dyn_cast<PointerType>(Ty)) {
        Ty = PtrTy->getElementType();
      } else if (auto *ArrayTy = dyn_cast<ArrayType>(Ty)) {
        Ty = ArrayTy->getElementType();
      } else {
        Subscripts.clear();
        Sizes.clear();
        break;
      }
      if (auto *Const = dyn_cast<SCEVConstant>(Expr))
        if (Const->getValue()->isZero()) {
          DroppedFirstDim = true;
          continue;
        }
      Subscripts.push_back(Expr);
      continue;
    }

    auto *ArrayTy = dyn_cast<ArrayType>(Ty);
    if (!ArrayTy) {
      Subscripts.clear();
      Sizes.clear();
      break;
    }

    Subscripts.push_back(Expr);
    if (!(DroppedFirstDim && i == 2))
      Sizes.push_back(ArrayTy->getNumElements());

    Ty = ArrayTy->getElementType();
  }

  return std::make_tuple(Subscripts, Sizes);
}

// For Falkor, we want to avoid having too many strided loads in a loop since
// that can exhaust the HW prefetcher resources.  We adjust the unroller
// MaxCount preference below to attempt to ensure unrolling doesn't create too
// many strided loads.
static void
getFalkorUnrollingPreferences(Loop *L, ScalarEvolution &SE,
                              TargetTransformInfo::UnrollingPreferences &UP) {
  enum { MaxStridedLoads = 7 };
  auto countStridedLoads = [](Loop *L, ScalarEvolution &SE) {
    int StridedLoads = 0;
    // FIXME? We could make this more precise by looking at the CFG and
    // e.g. not counting loads in each side of an if-then-else diamond.
    for (const auto BB : L->blocks()) {
      for (auto &I : *BB) {
        LoadInst *LMemI = dyn_cast<LoadInst>(&I);
        if (!LMemI)
          continue;

        Value *PtrValue = LMemI->getPointerOperand();
        if (L->isLoopInvariant(PtrValue))
          continue;

        const SCEV *LSCEV = SE.getSCEV(PtrValue);
        const SCEVAddRecExpr *LSCEVAddRec = dyn_cast<SCEVAddRecExpr>(LSCEV);
        if (!LSCEVAddRec || !LSCEVAddRec->isAffine())
          continue;

        // FIXME? We could take pairing of unrolled load copies into account
        // by looking at the AddRec, but we would probably have to limit this
        // to loops with no stores or other memory optimization barriers.
        ++StridedLoads;
        // We've seen enough strided loads that seeing more won't make a
        // difference.
        if (StridedLoads > MaxStridedLoads / 2)
          return StridedLoads;
      }
    }
    return StridedLoads;
  };

  int StridedLoads = countStridedLoads(L, SE);
  LLVM_DEBUG(dbgs() << "falkor-hwpf: detected " << StridedLoads
                    << " strided loads\n");
  // Pick the largest power of 2 unroll count that won't result in too many
  // strided loads.
  if (StridedLoads) {
    UP.MaxCount = 1 << Log2_32(MaxStridedLoads / StridedLoads);
    LLVM_DEBUG(dbgs() << "falkor-hwpf: setting unroll MaxCount to "
                      << UP.MaxCount << '\n');
  }
}

/// Insert code in the prolog code when unrolling a loop with a
/// run-time trip-count.
///
/// This method assumes that the loop unroll factor is total number
/// of loop bodes in the loop after unrolling. (Some folks refer
/// to the unroll factor as the number of *extra* copies added).
/// We assume also that the loop unroll factor is a power-of-two. So, after
/// unrolling the loop, the number of loop bodies executed is 2,
/// 4, 8, etc.  Note - LLVM converts the if-then-sequence to a switch
/// instruction in SimplifyCFG.cpp.  Then, the backend decides how code for
/// the switch instruction is generated.
///
///    extraiters = tripcount % loopfactor
///    if (extraiters == 0) jump Loop:
///    if (extraiters == loopfactor) jump L1
///    if (extraiters == loopfactor-1) jump L2
///    ...
///    L1:  LoopBody;
///    L2:  LoopBody;
///    ...
///    if tripcount < loopfactor jump End
///    Loop:
///    ...
///    End:
///
bool llvm::UnrollRuntimeLoopProlog(Loop *L, unsigned Count, LoopInfo *LI,
                                   LPPassManager *LPM) {
  // for now, only unroll loops that contain a single exit
  if (!L->getExitingBlock())
    return false;

  // Make sure the loop is in canonical form, and there is a single
  // exit block only.
  if (!L->isLoopSimplifyForm() || !L->getUniqueExitBlock())
    return false;

  // Use Scalar Evolution to compute the trip count.  This allows more
  // loops to be unrolled than relying on induction var simplification
  if (!LPM)
    return false;
  ScalarEvolution *SE = LPM->getAnalysisIfAvailable<ScalarEvolution>();
  if (!SE)
    return false;

  // Only unroll loops with a computable trip count and the trip count needs
  // to be an int value (allowing a pointer type is a TODO item)
  const SCEV *BECount = SE->getBackedgeTakenCount(L);
  if (isa<SCEVCouldNotCompute>(BECount) || !BECount->getType()->isIntegerTy())
    return false;

  // Add 1 since the backedge count doesn't include the first loop iteration
  const SCEV *TripCountSC =
    SE->getAddExpr(BECount, SE->getConstant(BECount->getType(), 1));
  if (isa<SCEVCouldNotCompute>(TripCountSC))
    return false;

  // We only handle cases when the unroll factor is a power of 2.
  // Count is the loop unroll factor, the number of extra copies added + 1.
  if ((Count & (Count-1)) != 0)
    return false;

  // If this loop is nested, then the loop unroller changes the code in
  // parent loop, so the Scalar Evolution pass needs to be run again
  if (Loop *ParentLoop = L->getParentLoop())
    SE->forgetLoop(ParentLoop);

  BasicBlock *PH = L->getLoopPreheader();
  BasicBlock *Header = L->getHeader();
  BasicBlock *Latch = L->getLoopLatch();
  // It helps to splits the original preheader twice, one for the end of the
  // prolog code and one for a new loop preheader
  BasicBlock *PEnd = SplitEdge(PH, Header, LPM->getAsPass());
  BasicBlock *NewPH = SplitBlock(PEnd, PEnd->getTerminator(), LPM->getAsPass());
  BranchInst *PreHeaderBR = cast<BranchInst>(PH->getTerminator());

  // Compute the number of extra iterations required, which is:
  //  extra iterations = run-time trip count % (loop unroll factor + 1)
  SCEVExpander Expander(*SE, "loop-unroll");
  Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
                                            PreHeaderBR);
  Type *CountTy = TripCount->getType();
  BinaryOperator *ModVal =
    BinaryOperator::CreateURem(TripCount,
                               ConstantInt::get(CountTy, Count),
                               "xtraiter");
  ModVal->insertBefore(PreHeaderBR);

  // Check if for no extra iterations, then jump to unrolled loop
  Value *BranchVal = new ICmpInst(PreHeaderBR,
                                  ICmpInst::ICMP_NE, ModVal,
                                  ConstantInt::get(CountTy, 0), "lcmp");
  // Branch to either the extra iterations or the unrolled loop
  // We will fix up the true branch label when adding loop body copies
  BranchInst::Create(PEnd, PEnd, BranchVal, PreHeaderBR);
  assert(PreHeaderBR->isUnconditional() &&
         PreHeaderBR->getSuccessor(0) == PEnd &&
         "CFG edges in Preheader are not correct");
  PreHeaderBR->eraseFromParent();

  ValueToValueMapTy LVMap;
//.........这里部分代码省略.........

bool LoopUnroll::runOnLoop(Loop *L, LPPassManager &LPM) {
  if (skipOptnoneFunction(L))
    return false;

  LoopInfo *LI = &getAnalysis<LoopInfo>();
  ScalarEvolution *SE = &getAnalysis<ScalarEvolution>();
  const TargetTransformInfo &TTI = getAnalysis<TargetTransformInfo>();

  BasicBlock *Header = L->getHeader();
  DEBUG(dbgs() << "Loop Unroll: F[" << Header->getParent()->getName()
        << "] Loop %" << Header->getName() << "\n");
  (void)Header;

  TargetTransformInfo::UnrollingPreferences UP;
  UP.Threshold = CurrentThreshold;
  UP.OptSizeThreshold = OptSizeUnrollThreshold;
  UP.Count = CurrentCount;
  UP.Partial = CurrentAllowPartial;
  UP.Runtime = CurrentRuntime;
  TTI.getUnrollingPreferences(L, UP);

  // Determine the current unrolling threshold.  While this is normally set
  // from UnrollThreshold, it is overridden to a smaller value if the current
  // function is marked as optimize-for-size, and the unroll threshold was
  // not user specified.
  unsigned Threshold = UserThreshold ? CurrentThreshold : UP.Threshold;
  if (!UserThreshold &&
      Header->getParent()->getAttributes().
        hasAttribute(AttributeSet::FunctionIndex,
                     Attribute::OptimizeForSize))
    Threshold = UP.OptSizeThreshold;

  // Find trip count and trip multiple if count is not available
  unsigned TripCount = 0;
  unsigned TripMultiple = 1;
  // Find "latch trip count". UnrollLoop assumes that control cannot exit
  // via the loop latch on any iteration prior to TripCount. The loop may exit
  // early via an earlier branch.
  BasicBlock *LatchBlock = L->getLoopLatch();
  if (LatchBlock) {
    TripCount = SE->getSmallConstantTripCount(L, LatchBlock);
    TripMultiple = SE->getSmallConstantTripMultiple(L, LatchBlock);
  }

  bool Runtime = UserRuntime ? CurrentRuntime : UP.Runtime;

  // Use a default unroll-count if the user doesn't specify a value
  // and the trip count is a run-time value.  The default is different
  // for run-time or compile-time trip count loops.
  unsigned Count = UserCount ? CurrentCount : UP.Count;
  if (Runtime && Count == 0 && TripCount == 0)
    Count = UnrollRuntimeCount;

  if (Count == 0) {
    // Conservative heuristic: if we know the trip count, see if we can
    // completely unroll (subject to the threshold, checked below); otherwise
    // try to find greatest modulo of the trip count which is still under
    // threshold value.
    if (TripCount == 0)
      return false;
    Count = TripCount;
  }

  // Enforce the threshold.
  if (Threshold != NoThreshold) {
    unsigned NumInlineCandidates;
    bool notDuplicatable;
    unsigned LoopSize = ApproximateLoopSize(L, NumInlineCandidates,
                                            notDuplicatable, TTI);
    DEBUG(dbgs() << "  Loop Size = " << LoopSize << "\n");
    if (notDuplicatable) {
      DEBUG(dbgs() << "  Not unrolling loop which contains non-duplicatable"
            << " instructions.\n");
      return false;
    }
    if (NumInlineCandidates != 0) {
      DEBUG(dbgs() << "  Not unrolling loop with inlinable calls.\n");
      return false;
    }
    uint64_t Size = (uint64_t)LoopSize*Count;
    if (TripCount != 1 && Size > Threshold) {
      DEBUG(dbgs() << "  Too large to fully unroll with count: " << Count
            << " because size: " << Size << ">" << Threshold << "\n");
      bool AllowPartial = UserAllowPartial ? CurrentAllowPartial : UP.Partial;
      if (!AllowPartial && !(Runtime && TripCount == 0)) {
        DEBUG(dbgs() << "  will not try to unroll partially because "
              << "-unroll-allow-partial not given\n");
        return false;
      }
      if (TripCount) {
        // Reduce unroll count to be modulo of TripCount for partial unrolling
        Count = Threshold / LoopSize;
        while (Count != 0 && TripCount%Count != 0)
          Count--;
      }
      else if (Runtime) {
        // Reduce unroll count to be a lower power-of-two value
        while (Count != 0 && Size > Threshold) {
          Count >>= 1;
          Size = LoopSize*Count;
//.........这里部分代码省略.........

/// Unroll the given loop by Count. The loop must be in LCSSA form. Returns true
/// if unrolling was successful, or false if the loop was unmodified. Unrolling
/// can only fail when the loop's latch block is not terminated by a conditional
/// branch instruction. However, if the trip count (and multiple) are not known,
/// loop unrolling will mostly produce more code that is no faster.
///
/// TripCount is generally defined as the number of times the loop header
/// executes. UnrollLoop relaxes the definition to permit early exits: here
/// TripCount is the iteration on which control exits LatchBlock if no early
/// exits were taken. Note that UnrollLoop assumes that the loop counter test
/// terminates LatchBlock in order to remove unnecesssary instances of the
/// test. In other words, control may exit the loop prior to TripCount
/// iterations via an early branch, but control may not exit the loop from the
/// LatchBlock's terminator prior to TripCount iterations.
///
/// Similarly, TripMultiple divides the number of times that the LatchBlock may
/// execute without exiting the loop.
///
/// The LoopInfo Analysis that is passed will be kept consistent.
///
/// If a LoopPassManager is passed in, and the loop is fully removed, it will be
/// removed from the LoopPassManager as well. LPM can also be NULL.
///
/// This utility preserves LoopInfo. If DominatorTree or ScalarEvolution are
/// available from the Pass it must also preserve those analyses.
bool llvm::UnrollLoop(Loop *L, unsigned Count, unsigned TripCount,
                      bool AllowRuntime, unsigned TripMultiple,
                      LoopInfo *LI, Pass *PP, LPPassManager *LPM) {
  BasicBlock *Preheader = L->getLoopPreheader();
  if (!Preheader) {
    DEBUG(dbgs() << "  Can't unroll; loop preheader-insertion failed.\n");
    return false;
  }

  BasicBlock *LatchBlock = L->getLoopLatch();
  if (!LatchBlock) {
    DEBUG(dbgs() << "  Can't unroll; loop exit-block-insertion failed.\n");
    return false;
  }

  // Loops with indirectbr cannot be cloned.
  if (!L->isSafeToClone()) {
    DEBUG(dbgs() << "  Can't unroll; Loop body cannot be cloned.\n");
    return false;
  }

  BasicBlock *Header = L->getHeader();
  BranchInst *BI = dyn_cast<BranchInst>(LatchBlock->getTerminator());

  if (!BI || BI->isUnconditional()) {
    // The loop-rotate pass can be helpful to avoid this in many cases.
    DEBUG(dbgs() <<
             "  Can't unroll; loop not terminated by a conditional branch.\n");
    return false;
  }

  if (Header->hasAddressTaken()) {
    // The loop-rotate pass can be helpful to avoid this in many cases.
    DEBUG(dbgs() <<
          "  Won't unroll loop: address of header block is taken.\n");
    return false;
  }

  if (TripCount != 0)
    DEBUG(dbgs() << "  Trip Count = " << TripCount << "\n");
  if (TripMultiple != 1)
    DEBUG(dbgs() << "  Trip Multiple = " << TripMultiple << "\n");

  // Effectively "DCE" unrolled iterations that are beyond the tripcount
  // and will never be executed.
  if (TripCount != 0 && Count > TripCount)
    Count = TripCount;

  // Don't enter the unroll code if there is nothing to do. This way we don't
  // need to support "partial unrolling by 1".
  if (TripCount == 0 && Count < 2)
    return false;

  assert(Count > 0);
  assert(TripMultiple > 0);
  assert(TripCount == 0 || TripCount % TripMultiple == 0);

  // Are we eliminating the loop control altogether?
  bool CompletelyUnroll = Count == TripCount;

  // We assume a run-time trip count if the compiler cannot
  // figure out the loop trip count and the unroll-runtime
  // flag is specified.
  bool RuntimeTripCount = (TripCount == 0 && Count > 0 && AllowRuntime);

  if (RuntimeTripCount && !UnrollRuntimeLoopProlog(L, Count, LI, LPM))
    return false;

  // Notify ScalarEvolution that the loop will be substantially changed,
  // if not outright eliminated.
  if (PP) {
    ScalarEvolution *SE = PP->getAnalysisIfAvailable<ScalarEvolution>();
    if (SE)
      SE->forgetLoop(L);
  }
//.........这里部分代码省略.........