[NFC][InlineCost] Factor cost modeling out of CallAnalyzer traversal.

[mtrofin]

Summary

The goal is to simplify experimentation on the cost model. Today,
CallAnalyzer decides 2 things: legality, and benefit. The refactoring
keeps legality assessment in CallAnalyzer, and factors benefit
evaluation out, as an extension.

[davidxl]

llvm/lib/Analysis/InlineCost.cpp

161

Can you document each handler in source?

Also does it make more sense to provide a dummy empty body so that users are not forced to implement all of them?

640

InlineCostCallAnalyzer?

[mtrofin]

virtuals have implementation; renamed derived class; made OnJumpTable return bool, like OnCaseCluster does.

[davidxl]

The SROA handling code should also be abstracted away and let the derived class handling cost accumulation. In particular, the common code pattern like:

if (lookupSROAArgAndCost(I.getPointerOperand(), SROAArg, CostIt)) {

  if (I.isSimple()) {
    accumulateSROACost(CostIt, InlineConstants::InstrCost);
    return true;
  }

  disableSROA(CostIt);
}

Can be abtracted with interface CallAnalyzer::handleSROA(..), and the derived class implements this pattern.

[mtrofin]

• renamed to camelCase
• factored the sroa costing out

[mtrofin]

In D71733#1792219, @davidxl wrote:

The SROA handling code should also be abstracted away and let the derived class handling cost accumulation. In particular, the common code pattern like:

if (lookupSROAArgAndCost(I.getPointerOperand(), SROAArg, CostIt)) {

  if (I.isSimple()) {
    accumulateSROACost(CostIt, InlineConstants::InstrCost);
    return true;
  }

  disableSROA(CostIt);
}

Can be abtracted with interface CallAnalyzer::handleSROA(..), and the derived class implements this pattern.

Done - that also pulls out the sroa-related costing stuff out, which I've missed originally.

[davidxl]

llvm/lib/Analysis/InlineCost.cpp

179

Document -- similarly for other APIs overrider may need to define.

180

document.

186

Should it be named 'onInitializeSROAArg' or something?

187

Document semantics of the API. Also should it be renamed 'onAggregateSROAUse'?

[mtrofin]

Comments, some renames, and more clarity around the SROA APIs' arguments.

[davidxl]

llvm/lib/Analysis/InlineCost.cpp

164

Document what to return.

171

Called before the analysis of the callee body (with callsite contexts propagate) starts. It checks callsites specific information.

ALso document what to return.

173

Called when the analysis engine decides that SROA can not be done.

175

Called by the analysis when load elimination can not happen.

182

Make a note that 'Call' is a callsite in the callee body -- not to be confused with the callsite being analyzed.

197

onFinalizeSwitch

206

analyzeIndirectCall

[mtrofin]

incorporated feedback

[davidxl]

this looks good to me. Easwaran, do you have any comments on the refactoring?

[davidxl]

The intention is to make the base CallAnalysis becomes a symbolic execution engine (virtual optimizations) what can be reused. The cost tracking is extracted into the derived class.

[hiraditya]

Thanks for refactoring LGTM.

[eraman]

llvm/lib/Analysis/InlineCost.cpp

129–131

This should go to InlineCostCallAnalyzer as it is not needed for legality check or symbolic evaluation.

136

Please leave an empty line before the doc comment to make this more readable (here and other places below)

676

Let's say you are using a different flavor of InlineCostCallAnalyzer (one which has different implemnetation of the virtual onXYZ methods), you should be instantiating an instance of that class here and call analyze here.

1519

Comment should go to callee (please check all places where you call onXYZ methods if the comment belong in the caller or callee)

1545

Why not move the entire block into a single function?

1673–1674

Comments should go into the callee. Also, it's better to merge onCaseCluster and onFinalizeSwitch into one function and just do

onFinalizeSwitch();
return false;

(Revisiting this - I think even onJumpTable should be folded into onFinalizeSwitch. From simplification/legality perspective, there is notthing left to do here after checking if the condition folds into a constant)

1767

onCommonInstructionSimplification is more meaningful.

1967–1968

The comments are not in-sync with the code. shouldStop could potentially use other ways to check if the analysis should stop - not just cost exceeds the threshold

2038–2039

Shouldn't this entire block be moved to InlineCostCallAnalyzer as this isn't performing symbolic evaluation or legality check?

[mtrofin]

Incorporated feedback.

[mtrofin]

llvm/lib/Analysis/InlineCost.cpp

676

Maybe, but I'd prefer addressing that problem when the scenario arises, and avoid the added complexity meanwhile, wdyt?

1967–1968

True, and in this case, the code is kind of self-evident. I moved the comment to the shouldStop implementation.

2038–2039

Possibly, but all that prep work - getting a DominatorTree, LoopInfo, going over loops, making sure we ignore those that aren't executed - would be necessary if we want to account the number of loops somehow.

[eraman]

llvm/lib/Analysis/InlineCost.cpp

676

I still think this is better handled now, but if you think otherwise, please add a comment.

2038–2039

Well, all this prep work is for cost computation right? Again, since the goal is to keep only symbolic analysis + legality check here, why keep these here?

[davidxl]

llvm/lib/Analysis/InlineCost.cpp

2038–2039

This looks like cost computation specific and should probably be moved to the derived class. It can query the base class about deadblocks.

[mtrofin]

Removed 'onLoops'

[eraman]

LGTM

llvm/lib/Analysis/InlineCost.cpp

457

Please add a FIXME here saying we should be using the same derived class instance

[mtrofin]

added fixmes

[mtrofin]

llvm/lib/Analysis/InlineCost.cpp

457

done, also marked InlineCostCallAnalyzer as final and referenced this fixme there, it'll be more visible should deriving InlineCostCallAnalyzer happen.

[mtrofin]

fixed bugs

[mtrofin]

fixes. Validated building a few binaries are identical before/after this change

[davidxl]

lgtm

[eraman]

LGTM

llvm/lib/Analysis/InlineCost.cpp

241

nit on naming: Perhaps use EnabledSROAAllocas ? Also, I feel the comment lines starting from We don't delete... is not required.

2226

Looks like unintended change.

[mtrofin]

llvm/lib/Analysis/InlineCost.cpp

2226

It's unrelated, but intended: clang-format kept indenting the original comments. I looked closer at what was happening, and it looks like the comments characterize the 'case' they used to be above, but clang-format wants to indent them as a case statement body. So I moved them as shown, to avoid future churn with clang-format.

[mtrofin]

llvm/lib/Analysis/InlineCost.cpp

241

Agreed, I'll send a followup to rename.

llvm/lib/Analysis/InlineCost.cpp

cl::desc("Minimum block frequency, expressed as a multiple of caller's "
              "profile information."));
 
 static cl::opt<bool> OptComputeFullInlineCost(
     "inline-cost-full", cl::Hidden, cl::init(false), cl::ZeroOrMore,
     cl::desc("Compute the full inline cost of a call site even when the cost "
              "exceeds the threshold."));
 
 namespace {
-
+class InlineCostCallAnalyzer;
 class CallAnalyzer : public InstVisitor<CallAnalyzer, bool> {
   typedef InstVisitor<CallAnalyzer, bool> Base;
   friend class InstVisitor<CallAnalyzer, bool>;
 
+protected:
+  virtual ~CallAnalyzer() {}
   /// The TargetTransformInfo available for this compilation.
   const TargetTransformInfo &TTI;
 
   /// Getter for the cache of @llvm.assume intrinsics.
   std::function<AssumptionCache &(Function &)> &GetAssumptionCache;
 
   /// Getter for BlockFrequencyInfo
   Optional<function_ref<BlockFrequencyInfo &(Function &)>> &GetBFI;
protected:
   /// The OptimizationRemarkEmitter available for this compilation.
   OptimizationRemarkEmitter *ORE;
 
   /// The candidate callsite being analyzed. Please do not use this to do
   /// analysis in the caller function; we want the inline cost query to be
   /// easily cacheable. Instead, use the cover function paramHasAttr.
   CallBase &CandidateCall;
 
-  /// Tunable parameters that control the analysis.
-  const InlineParams &Params;
+  /// Extension points for handling callsite features.
+  /// Called after a basic block was analyzed.
+  virtual void onBlockAnalyzed(const BasicBlock *BB) {}
 
-  /// Upper bound for the inlining cost. Bonuses are being applied to account
-  /// for speculative "expected profit" of the inlining decision.
-  int Threshold;
+  /// Called at the end of the analysis of the callsite. Return the outcome of
+  /// the analysis, i.e. 'InlineResult(true)' if the inlining may happen, or
+  /// the reason it can't.
+  virtual InlineResult finalizeAnalysis() { return true; }
 
-  /// Inlining cost measured in abstract units, accounts for all the
-  /// instructions expected to be executed for a given function invocation.
-  /// Instructions that are statically proven to be dead based on call-site
-  /// arguments are not counted here.
-  int Cost = 0;
+  /// Called when we're about to start processing a basic block, and every time
+  /// we are done processing an instruction. Return true if there is no point in
+  /// continuing the analysis (e.g. we've determined already the call site is
+  /// too expensive to inline)
+  virtual bool shouldStop() { return false; }
+
+  /// Called before the analysis of the callee body starts (with callsite
+  /// contexts propagated).  It checks callsite-specific information. Return a
+  /// reason analysis can't continue if that's the case, or 'true' if it may
+  /// continue.
+  virtual InlineResult onAnalysisStart() { return true; }
+
+  /// Called if the analysis engine decides SROA cannot be done for the given
+  /// alloca.
+  virtual void onDisableSROA(AllocaInst *Arg) {}
+
+  /// Called the analysis engine determines load elimination won't happen.
+  virtual void onDisableLoadElimination() {}
+
+  /// Called to account for a call.
+  virtual void onCallPenalty() {}
+
+  /// Called to account for the expectation the inlining would result in a load
+  /// elimination.
+  virtual void onLoadEliminationOpportunity() {}
 
-  bool ComputeFullInlineCost;
+  /// Called to account for the cost of argument setup for the Call in the
+  /// callee's body (not the callsite currently under analysis).
+  virtual void onCallArgumentSetup(const CallBase &Call) {}
+
+  /// Called to account for a load relative intrinsic.
+  virtual void onLoadRelativeIntrinsic() {}
+
+  /// Called to account for a lowered call.
+  virtual void onLoweredCall(Function *F, CallBase &Call, bool IsIndirectCall) {
+  }
+
+  /// Account for a jump table of given size. Return false to stop further
+  /// processing the switch instruction
+  virtual bool onJumpTable(unsigned JumpTableSize) { return true; }
+
+  /// Account for a case cluster of given size. Return false to stop further
+  /// processing of the instruction.
+  virtual bool onCaseCluster(unsigned NumCaseCluster) { return true; }
+
+  /// Called at the end of processing a switch instruction, with the given
+  /// number of case clusters.
+  virtual void onFinalizeSwitch(unsigned JumpTableSize,
+                                unsigned NumCaseCluster) {}
+
+  /// Called to account for any other instruction not specifically accounted
+  /// for.
+  virtual void onCommonInstructionSimplification() {}
+
+  /// Start accounting potential benefits due to SROA for the given alloca.
+  virtual void onInitializeSROAArg(AllocaInst *Arg) {}
+
+  /// Account SROA savings for the AllocaInst value.
+  virtual void onAggregateSROAUse(AllocaInst *V) {}
+
+  bool handleSROA(Value *V, bool DoNotDisable) {
+    // Check for SROA candidates in comparisons.
+    if (auto *SROAArg = getSROAArgForValueOrNull(V)) {
+      if (DoNotDisable) {
+        onAggregateSROAUse(SROAArg);
+        return true;
+      }
+      disableSROAForArg(SROAArg);
+    }
+    return false;
+  }
 
   bool IsCallerRecursive = false;
   bool IsRecursiveCall = false;
   bool ExposesReturnsTwice = false;
   bool HasDynamicAlloca = false;
   bool ContainsNoDuplicateCall = false;
   bool HasReturn = false;
   bool HasIndirectBr = false;
   bool HasUninlineableIntrinsic = false;
   bool InitsVargArgs = false;
 
-  /// Attempt to evaluate indirect calls to boost its inline cost.
-  bool BoostIndirectCalls;
-
   /// Number of bytes allocated statically by the callee.
   uint64_t AllocatedSize = 0;
   unsigned NumInstructions = 0;
   unsigned NumVectorInstructions = 0;
 
-  /// Bonus to be applied when percentage of vector instructions in callee is
-  /// high (see more details in updateThreshold).
-  int VectorBonus = 0;
-  /// Bonus to be applied when the callee has only one reachable basic block.
-  int SingleBBBonus = 0;
-
   /// While we walk the potentially-inlined instructions, we build up and
   /// maintain a mapping of simplified values specific to this callsite. The
   /// idea is to propagate any special information we have about arguments to
   /// this call through the inlinable section of the function, and account for
   /// likely simplifications post-inlining. The most important aspect we track
   /// is CFG altering simplifications -- when we prove a basic block dead, that
   /// can cause dramatic shifts in the cost of inlining a function.
   DenseMap<Value *, Constant *> SimplifiedValues;
 
   /// Keep track of the values which map back (through function arguments) to
   /// allocas on the caller stack which could be simplified through SROA.
-  DenseMap<Value *, Value *> SROAArgValues;
+  DenseMap<Value *, AllocaInst *> SROAArgValues;
 
-  /// The mapping of caller Alloca values to their accumulated cost savings. If
-  /// we have to disable SROA for one of the allocas, this tells us how much
-  /// cost must be added.
-  DenseMap<Value *, int> SROAArgCosts;
+  /// Keep track of Allocas for which we believe we may get SROA optimization.
+  /// We don't delete entries in SROAArgValue because we still want
+  /// isAllocaDerivedArg to function correctly.
+  DenseSet<AllocaInst *> EnabledSROAArgValues;
 
   /// Keep track of values which map to a pointer base and constant offset.
   DenseMap<Value *, std::pair<Value *, APInt>> ConstantOffsetPtrs;
 
   /// Keep track of dead blocks due to the constant arguments.
   SetVector<BasicBlock *> DeadBlocks;
 
   /// The mapping of the blocks to their known unique successors due to the
   /// constant arguments.
   DenseMap<BasicBlock *, BasicBlock *> KnownSuccessors;
 
   /// Model the elimination of repeated loads that is expected to happen
   /// whenever we simplify away the stores that would otherwise cause them to be
   /// loads.
   bool EnableLoadElimination;
   SmallPtrSet<Value *, 16> LoadAddrSet;
-  int LoadEliminationCost = 0;
+
+  AllocaInst *getSROAArgForValueOrNull(Value *V) const {
+    auto It = SROAArgValues.find(V);
+    if (It == SROAArgValues.end() ||
+        EnabledSROAArgValues.count(It->second) == 0)
+      return nullptr;
+    return It->second;
+  }
 
   // Custom simplification helper routines.
   bool isAllocaDerivedArg(Value *V);
-  bool lookupSROAArgAndCost(Value *V, Value *&Arg,
-                            DenseMap<Value *, int>::iterator &CostIt);
-  void disableSROA(DenseMap<Value *, int>::iterator CostIt);
+  void disableSROAForArg(AllocaInst *SROAArg);
   void disableSROA(Value *V);
   void findDeadBlocks(BasicBlock *CurrBB, BasicBlock *NextBB);
-  void accumulateSROACost(DenseMap<Value *, int>::iterator CostIt,
-                          int InstructionCost);
   void disableLoadElimination();
   bool isGEPFree(GetElementPtrInst &GEP);
   bool canFoldInboundsGEP(GetElementPtrInst &I);
   bool accumulateGEPOffset(GEPOperator &GEP, APInt &Offset);
   bool simplifyCallSite(Function *F, CallBase &Call);
   template <typename Callable>
   bool simplifyInstruction(Instruction &I, Callable Evaluate);
   ConstantInt *stripAndComputeInBoundsConstantOffsets(Value *&V);
 
   /// Return true if the given argument to the function being considered for
   /// inlining has the given attribute set either at the call site or the
   /// function declaration.  Primarily used to inspect call site specific
   /// attributes since these can be more precise than the ones on the callee
   /// itself.
   bool paramHasAttr(Argument *A, Attribute::AttrKind Attr);
 
   /// Return true if the given value is known non null within the callee if
   /// inlined through this particular callsite.
   bool isKnownNonNullInCallee(Value *V);
 
-  /// Update Threshold based on callsite properties such as callee
-  /// attributes and callee hotness for PGO builds. The Callee is explicitly
-  /// passed to support analyzing indirect calls whose target is inferred by
-  /// analysis.
-  void updateThreshold(CallBase &Call, Function &Callee);
-
   /// Return true if size growth is allowed when inlining the callee at \p Call.
   bool allowSizeGrowth(CallBase &Call);
 
-  /// Return true if \p Call is a cold callsite.
-  bool isColdCallSite(CallBase &Call, BlockFrequencyInfo *CallerBFI);
-
-  /// Return a higher threshold if \p Call is a hot callsite.
-  Optional<int> getHotCallSiteThreshold(CallBase &Call,
-                                        BlockFrequencyInfo *CallerBFI);
-
   // Custom analysis routines.
   InlineResult analyzeBlock(BasicBlock *BB,
                             SmallPtrSetImpl<const Value *> &EphValues);
 
-  /// Handle a capped 'int' increment for Cost.
-  void addCost(int64_t Inc, int64_t UpperBound = INT_MAX) {
-    assert(UpperBound > 0 && UpperBound <= INT_MAX && "invalid upper bound");
-    Cost = (int)std::min(UpperBound, Cost + Inc);
-  }
-
   // Disable several entry points to the visitor so we don't accidentally use
   // them by declaring but not defining them here.
   void visit(Module *);
   void visit(Module &);
   void visit(Function *);
   void visit(Function &);
   void visit(BasicBlock *);
   void visit(BasicBlock &);
protected:
   bool visitCatchReturnInst(CatchReturnInst &RI);
   bool visitUnreachableInst(UnreachableInst &I);
 
 public:
   CallAnalyzer(const TargetTransformInfo &TTI,
                std::function<AssumptionCache &(Function &)> &GetAssumptionCache,
                Optional<function_ref<BlockFrequencyInfo &(Function &)>> &GetBFI,
                ProfileSummaryInfo *PSI, OptimizationRemarkEmitter *ORE,
-               Function &Callee, CallBase &Call, const InlineParams &Params,
-               bool BoostIndirect = true)
+               Function &Callee, CallBase &Call)
       : TTI(TTI), GetAssumptionCache(GetAssumptionCache), GetBFI(GetBFI),
         PSI(PSI), F(Callee), DL(F.getParent()->getDataLayout()), ORE(ORE),
-        CandidateCall(Call), Params(Params), Threshold(Params.DefaultThreshold),
-        ComputeFullInlineCost(OptComputeFullInlineCost ||
-                              Params.ComputeFullInlineCost || ORE),
-        BoostIndirectCalls(BoostIndirect), EnableLoadElimination(true) {}
+        CandidateCall(Call), EnableLoadElimination(true) {}
 
   InlineResult analyze();
 
-  int getThreshold() { return Threshold; }
-  int getCost() { return Cost; }
-
   // Keep a bunch of stats about the cost savings found so we can print them
   // out when debugging.
   unsigned NumConstantArgs = 0;
   unsigned NumConstantOffsetPtrArgs = 0;
   unsigned NumAllocaArgs = 0;
   unsigned NumConstantPtrCmps = 0;
   unsigned NumConstantPtrDiffs = 0;
   unsigned NumInstructionsSimplified = 0;
-  unsigned SROACostSavings = 0;
-  unsigned SROACostSavingsLost = 0;
 
   void dump();
 };
 
-} // namespace
+/// FIXME: if it is necessary to derive from InlineCostCallAnalyzer, note
+/// the FIXME in onLoweredCall, when instantiating an InlineCostCallAnalyzer
+class InlineCostCallAnalyzer final : public CallAnalyzer {
+  const int CostUpperBound = INT_MAX - InlineConstants::InstrCost - 1;
+  const bool ComputeFullInlineCost;
+  int LoadEliminationCost = 0;
+  /// Bonus to be applied when percentage of vector instructions in callee is
+  /// high (see more details in updateThreshold).
+  int VectorBonus = 0;
+  /// Bonus to be applied when the callee has only one reachable basic block.
+  int SingleBBBonus = 0;
 
-/// Test whether the given value is an Alloca-derived function argument.
-bool CallAnalyzer::isAllocaDerivedArg(Value *V) {
-  return SROAArgValues.count(V);
-}
+  /// Tunable parameters that control the analysis.
+  const InlineParams &Params;
 
-/// Lookup the SROA-candidate argument and cost iterator which V maps to.
-/// Returns false if V does not map to a SROA-candidate.
-bool CallAnalyzer::lookupSROAArgAndCost(
-    Value *V, Value *&Arg, DenseMap<Value *, int>::iterator &CostIt) {
-  if (SROAArgValues.empty() || SROAArgCosts.empty())
-    return false;
+  /// Upper bound for the inlining cost. Bonuses are being applied to account
+  /// for speculative "expected profit" of the inlining decision.
+  int Threshold = 0;
 
-  DenseMap<Value *, Value *>::iterator ArgIt = SROAArgValues.find(V);
-  if (ArgIt == SROAArgValues.end())
-    return false;
+  /// Attempt to evaluate indirect calls to boost its inline cost.
+  const bool BoostIndirectCalls;
+
+  /// Inlining cost measured in abstract units, accounts for all the
+  /// instructions expected to be executed for a given function invocation.
+  /// Instructions that are statically proven to be dead based on call-site
+  /// arguments are not counted here.
+  int Cost = 0;
+
+  bool SingleBB = true;
+
+  unsigned SROACostSavings = 0;
+  unsigned SROACostSavingsLost = 0;
+
+  /// The mapping of caller Alloca values to their accumulated cost savings. If
+  /// we have to disable SROA for one of the allocas, this tells us how much
+  /// cost must be added.
+  DenseMap<AllocaInst *, int> SROAArgCosts;
+
+  /// Return true if \p Call is a cold callsite.
+  bool isColdCallSite(CallBase &Call, BlockFrequencyInfo *CallerBFI);
+
+  /// Update Threshold based on callsite properties such as callee
+  /// attributes and callee hotness for PGO builds. The Callee is explicitly
+  /// passed to support analyzing indirect calls whose target is inferred by
+  /// analysis.
+  void updateThreshold(CallBase &Call, Function &Callee);
+  /// Return a higher threshold if \p Call is a hot callsite.
+  Optional<int> getHotCallSiteThreshold(CallBase &Call,
+                                        BlockFrequencyInfo *CallerBFI);
 
-  Arg = ArgIt->second;
-  CostIt = SROAArgCosts.find(Arg);
-  return CostIt != SROAArgCosts.end();
+  /// Handle a capped 'int' increment for Cost.
+  void addCost(int64_t Inc, int64_t UpperBound = INT_MAX) {
+    assert(UpperBound > 0 && UpperBound <= INT_MAX && "invalid upper bound");
+    Cost = (int)std::min(UpperBound, Cost + Inc);
   }
 
-/// Disable SROA for the candidate marked by this cost iterator.
-///
-/// This marks the candidate as no longer viable for SROA, and adds the cost
-/// savings associated with it back into the inline cost measurement.
-void CallAnalyzer::disableSROA(DenseMap<Value *, int>::iterator CostIt) {
-  // If we're no longer able to perform SROA we need to undo its cost savings
-  // and prevent subsequent analysis.
+  void onDisableSROA(AllocaInst *Arg) override {
+    auto CostIt = SROAArgCosts.find(Arg);
+    if (CostIt == SROAArgCosts.end())
+      return;
     addCost(CostIt->second);
     SROACostSavings -= CostIt->second;
     SROACostSavingsLost += CostIt->second;
     SROAArgCosts.erase(CostIt);
-  disableLoadElimination();
   }
 
+  void onDisableLoadElimination() override {
+    addCost(LoadEliminationCost);
+    LoadEliminationCost = 0;
+  }
+  void onCallPenalty() override { addCost(InlineConstants::CallPenalty); }
+  void onCallArgumentSetup(const CallBase &Call) override {
+    // Pay the price of the argument setup. We account for the average 1
+    // instruction per call argument setup here.
+    addCost(Call.arg_size() * InlineConstants::InstrCost);
+  }
+  void onLoadRelativeIntrinsic() override {
+    // This is normally lowered to 4 LLVM instructions.
+    addCost(3 * InlineConstants::InstrCost);
+  }
+  void onLoweredCall(Function *F, CallBase &Call,
+                     bool IsIndirectCall) override {
+    // We account for the average 1 instruction per call argument setup here.
+    addCost(Call.arg_size() * InlineConstants::InstrCost);
+
+    // If we have a constant that we are calling as a function, we can peer
+    // through it and see the function target. This happens not infrequently
+    // during devirtualization and so we want to give it a hefty bonus for
+    // inlining, but cap that bonus in the event that inlining wouldn't pan out.
+    // Pretend to inline the function, with a custom threshold.
+    if (IsIndirectCall && BoostIndirectCalls) {
+      auto IndirectCallParams = Params;
+      IndirectCallParams.DefaultThreshold =
+          InlineConstants::IndirectCallThreshold;
+      /// FIXME: if InlineCostCallAnalyzer is derived from, this may need
+      /// to instantiate the derived class.
+      InlineCostCallAnalyzer CA(TTI, GetAssumptionCache, GetBFI, PSI, ORE, *F,
+                                Call, IndirectCallParams, false);
+      if (CA.analyze()) {
+        // We were able to inline the indirect call! Subtract the cost from the
+        // threshold to get the bonus we want to apply, but don't go below zero.
+        Cost -= std::max(0, CA.getThreshold() - CA.getCost());
+      }
+    } else
+      // Otherwise simply add the cost for merely making the call.
+      addCost(InlineConstants::CallPenalty);
+  }
+
+  void onFinalizeSwitch(unsigned JumpTableSize,
+                        unsigned NumCaseCluster) override {
+    // If suitable for a jump table, consider the cost for the table size and
+    // branch to destination.
+    // Maximum valid cost increased in this function.
+    if (JumpTableSize) {
+      int64_t JTCost = (int64_t)JumpTableSize * InlineConstants::InstrCost +
+                       4 * InlineConstants::InstrCost;
+
+      addCost(JTCost, (int64_t)CostUpperBound);
+      return;
+    }
+    // Considering forming a binary search, we should find the number of nodes
+    // which is same as the number of comparisons when lowered. For a given
+    // number of clusters, n, we can define a recursive function, f(n), to find
+    // the number of nodes in the tree. The recursion is :
+    // f(n) = 1 + f(n/2) + f (n - n/2), when n > 3,
+    // and f(n) = n, when n <= 3.
+    // This will lead a binary tree where the leaf should be either f(2) or f(3)
+    // when n > 3.  So, the number of comparisons from leaves should be n, while
+    // the number of non-leaf should be :
+    //   2^(log2(n) - 1) - 1
+    //   = 2^log2(n) * 2^-1 - 1
+    //   = n / 2 - 1.
+    // Considering comparisons from leaf and non-leaf nodes, we can estimate the
+    // number of comparisons in a simple closed form :
+    //   n + n / 2 - 1 = n * 3 / 2 - 1
+    if (NumCaseCluster <= 3) {
+      // Suppose a comparison includes one compare and one conditional branch.
+      addCost(NumCaseCluster * 2 * InlineConstants::InstrCost);
+      return;
+    }
+
+    int64_t ExpectedNumberOfCompare = 3 * (int64_t)NumCaseCluster / 2 - 1;
+    int64_t SwitchCost =
+        ExpectedNumberOfCompare * 2 * InlineConstants::InstrCost;
+
+    addCost(SwitchCost, (int64_t)CostUpperBound);
+  }
+  void onCommonInstructionSimplification() override {
+    addCost(InlineConstants::InstrCost);
+  }
+
+  void onInitializeSROAArg(AllocaInst *Arg) override {
+    assert(Arg != nullptr &&
+           "Should not initialize SROA costs for null value.");
+    SROAArgCosts[Arg] = 0;
+    EnabledSROAArgValues.insert(Arg);
+  }
+
+  void onAggregateSROAUse(AllocaInst *SROAArg) override {
+    auto CostIt = SROAArgCosts.find(SROAArg);
+    assert(CostIt != SROAArgCosts.end() &&
+           "expected this argument to have a cost");
+    CostIt->second += InlineConstants::InstrCost;
+    SROACostSavings += InlineConstants::InstrCost;
+  }
+
+  void onBlockAnalyzed(const BasicBlock *BB) override {
+    auto *TI = BB->getTerminator();
+    // If we had any successors at this point, than post-inlining is likely to
+    // have them as well. Note that we assume any basic blocks which existed
+    // due to branches or switches which folded above will also fold after
+    // inlining.
+    if (SingleBB && TI->getNumSuccessors() > 1) {
+      // Take off the bonus we applied to the threshold.
+      Threshold -= SingleBBBonus;
+      SingleBB = false;
+    }
+  }
+  InlineResult finalizeAnalysis() override {
+    // Loops generally act a lot like calls in that they act like barriers to
+    // movement, require a certain amount of setup, etc. So when optimising for
+    // size, we penalise any call sites that perform loops. We do this after all
+    // other costs here, so will likely only be dealing with relatively small
+    // functions (and hence DT and LI will hopefully be cheap).
+    auto *Caller = CandidateCall.getFunction();
+    if (Caller->hasMinSize()) {
+      DominatorTree DT(F);
+      LoopInfo LI(DT);
+      int NumLoops = 0;
+      for (Loop *L : LI) {
+        // Ignore loops that will not be executed
+        if (DeadBlocks.count(L->getHeader()))
+          continue;
+        NumLoops++;
+      }
+      addCost(NumLoops * InlineConstants::CallPenalty);
+    }
+
+    // We applied the maximum possible vector bonus at the beginning. Now,
+    // subtract the excess bonus, if any, from the Threshold before
+    // comparing against Cost.
+    if (NumVectorInstructions <= NumInstructions / 10)
+      Threshold -= VectorBonus;
+    else if (NumVectorInstructions <= NumInstructions / 2)
+      Threshold -= VectorBonus / 2;
+
+    return Cost < std::max(1, Threshold);
+  }
+  bool shouldStop() override {
+    // Bail out the moment we cross the threshold. This means we'll under-count
+    // the cost, but only when undercounting doesn't matter.
+    return Cost >= Threshold && !ComputeFullInlineCost;
+  }
+
+  void onLoadEliminationOpportunity() override {
+    LoadEliminationCost += InlineConstants::InstrCost;
+  }
+
+  InlineResult onAnalysisStart() override {
+    // Perform some tweaks to the cost and threshold based on the direct
+    // callsite information.
+
+    // We want to more aggressively inline vector-dense kernels, so up the
+    // threshold, and we'll lower it if the % of vector instructions gets too
+    // low. Note that these bonuses are some what arbitrary and evolved over
+    // time by accident as much as because they are principled bonuses.
+    //
+    // FIXME: It would be nice to remove all such bonuses. At least it would be
+    // nice to base the bonus values on something more scientific.
+    assert(NumInstructions == 0);
+    assert(NumVectorInstructions == 0);
+
+    // Update the threshold based on callsite properties
+    updateThreshold(CandidateCall, F);
+
+    // While Threshold depends on commandline options that can take negative
+    // values, we want to enforce the invariant that the computed threshold and
+    // bonuses are non-negative.
+    assert(Threshold >= 0);
+    assert(SingleBBBonus >= 0);
+    assert(VectorBonus >= 0);
+
+    // Speculatively apply all possible bonuses to Threshold. If cost exceeds
+    // this Threshold any time, and cost cannot decrease, we can stop processing
+    // the rest of the function body.
+    Threshold += (SingleBBBonus + VectorBonus);
+
+    // Give out bonuses for the callsite, as the instructions setting them up
+    // will be gone after inlining.
+    addCost(-getCallsiteCost(this->CandidateCall, DL));
+
+    // If this function uses the coldcc calling convention, prefer not to inline
+    // it.
+    if (F.getCallingConv() == CallingConv::Cold)
+      Cost += InlineConstants::ColdccPenalty;
+
+    // Check if we're done. This can happen due to bonuses and penalties.
+    if (Cost >= Threshold && !ComputeFullInlineCost)
+      return "high cost";
+
+    return true;
+  }
+
+public:
+  InlineCostCallAnalyzer(
+      const TargetTransformInfo &TTI,
+      std::function<AssumptionCache &(Function &)> &GetAssumptionCache,
+      Optional<function_ref<BlockFrequencyInfo &(Function &)>> &GetBFI,
+      ProfileSummaryInfo *PSI, OptimizationRemarkEmitter *ORE, Function &Callee,
+      CallBase &Call, const InlineParams &Params, bool BoostIndirect = true)
+      : CallAnalyzer(TTI, GetAssumptionCache, GetBFI, PSI, ORE, Callee, Call),
+        ComputeFullInlineCost(OptComputeFullInlineCost ||
+                              Params.ComputeFullInlineCost || ORE),
+        Params(Params), Threshold(Params.DefaultThreshold),
+        BoostIndirectCalls(BoostIndirect) {}
+  void dump();
+
+  virtual ~InlineCostCallAnalyzer() {}
+  int getThreshold() { return Threshold; }
+  int getCost() { return Cost; }
+};
+} // namespace
+
+/// Test whether the given value is an Alloca-derived function argument.
+bool CallAnalyzer::isAllocaDerivedArg(Value *V) {
+  return SROAArgValues.count(V);
+}
+
+void CallAnalyzer::disableSROAForArg(AllocaInst *SROAArg) {
+  onDisableSROA(SROAArg);
+  EnabledSROAArgValues.erase(SROAArg);
+  disableLoadElimination();
+}
 /// If 'V' maps to a SROA candidate, disable SROA for it.
 void CallAnalyzer::disableSROA(Value *V) {
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(V, SROAArg, CostIt))
-    disableSROA(CostIt);
+  if (auto *SROAArg = getSROAArgForValueOrNull(V)) {
+    disableSROAForArg(SROAArg);
   }
-
-/// Accumulate the given cost for a particular SROA candidate.
-void CallAnalyzer::accumulateSROACost(DenseMap<Value *, int>::iterator CostIt,
-                                      int InstructionCost) {
-  CostIt->second += InstructionCost;
-  SROACostSavings += InstructionCost;
 }
 
 void CallAnalyzer::disableLoadElimination() {
   if (EnableLoadElimination) {
-    addCost(LoadEliminationCost);
-    LoadEliminationCost = 0;
+    onDisableLoadElimination();
     EnableLoadElimination = false;
   }
 }
 
 /// Accumulate a constant GEP offset into an APInt if possible.
 ///
 /// Returns false if unable to compute the offset for any reason. Respects any
 /// simplified values known during the analysis of this callsite.
 bool CallAnalyzer::accumulateGEPOffset(GEPOperator &GEP, APInt &Offset) {
   unsigned IntPtrWidth = DL.getIndexTypeSizeInBits(GEP.getType());
   assert(IntPtrWidth == Offset.getBitWidth());
 
   for (gep_type_iterator GTI = gep_type_begin(GEP), GTE = gep_type_end(GEP);
        GTI != GTE; ++GTI) {
     ConstantInt *OpC = dyn_cast<ConstantInt>(GTI.getOperand());
     if (!OpC)
       if (Constant *SimpleOp = SimplifiedValues.lookup(GTI.getOperand()))
         OpC = dyn_cast<ConstantInt>(SimpleOp);
     if (!OpC)
if (FirstC) {
     SimplifiedValues[&I] = FirstC;
     return true;
   }
 
   // Check if we can map phi to a pointer with constant offset.
   if (FirstBaseAndOffset.first) {
     ConstantOffsetPtrs[&I] = FirstBaseAndOffset;
 
-    Value *SROAArg;
-    DenseMap<Value *, int>::iterator CostIt;
-    if (lookupSROAArgAndCost(FirstV, SROAArg, CostIt))
+    if (auto *SROAArg = getSROAArgForValueOrNull(FirstV))
       SROAArgValues[&I] = SROAArg;
   }
 
   return true;
 }
 
 /// Check we can fold GEPs of constant-offset call site argument pointers.
 /// This requires target data and inbounds GEPs.
bool CallAnalyzer::canFoldInboundsGEP(GetElementPtrInst &I) {
 
   // Add the result as a new mapping to Base + Offset.
   ConstantOffsetPtrs[&I] = BaseAndOffset;
 
   return true;
 }
 
 bool CallAnalyzer::visitGetElementPtr(GetElementPtrInst &I) {
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  bool SROACandidate =
-      lookupSROAArgAndCost(I.getPointerOperand(), SROAArg, CostIt);
+  auto *SROAArg = getSROAArgForValueOrNull(I.getPointerOperand());
 
   // Lambda to check whether a GEP's indices are all constant.
   auto IsGEPOffsetConstant = [&](GetElementPtrInst &GEP) {
     for (User::op_iterator I = GEP.idx_begin(), E = GEP.idx_end(); I != E; ++I)
       if (!isa<Constant>(*I) && !SimplifiedValues.lookup(*I))
         return false;
     return true;
   };
 
   if ((I.isInBounds() && canFoldInboundsGEP(I)) || IsGEPOffsetConstant(I)) {
-    if (SROACandidate)
+    if (SROAArg)
       SROAArgValues[&I] = SROAArg;
 
     // Constant GEPs are modeled as free.
     return true;
   }
 
   // Variable GEPs will require math and will disable SROA.
-  if (SROACandidate)
-    disableSROA(CostIt);
+  if (SROAArg)
+    disableSROAForArg(SROAArg);
   return isGEPFree(I);
 }
 
 /// Simplify \p I if its operands are constants and update SimplifiedValues.
 /// \p Evaluate is a callable specific to instruction type that evaluates the
 /// instruction when all the operands are constants.
 template <typename Callable>
 bool CallAnalyzer::simplifyInstruction(Instruction &I, Callable Evaluate) {
bool CallAnalyzer::visitBitCast(BitCastInst &I) {
   // Track base/offsets through casts
   std::pair<Value *, APInt> BaseAndOffset =
       ConstantOffsetPtrs.lookup(I.getOperand(0));
   // Casts don't change the offset, just wrap it up.
   if (BaseAndOffset.first)
     ConstantOffsetPtrs[&I] = BaseAndOffset;
 
   // Also look for SROA candidates here.
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(I.getOperand(0), SROAArg, CostIt))
+  if (auto *SROAArg = getSROAArgForValueOrNull(I.getOperand(0)))
     SROAArgValues[&I] = SROAArg;
 
   // Bitcasts are always zero cost.
   return true;
 }
 
 bool CallAnalyzer::visitPtrToInt(PtrToIntInst &I) {
   // Propagate constants through ptrtoint.
bool CallAnalyzer::visitPtrToInt(PtrToIntInst &I) {
 
   // This is really weird. Technically, ptrtoint will disable SROA. However,
   // unless that ptrtoint is *used* somewhere in the live basic blocks after
   // inlining, it will be nuked, and SROA should proceed. All of the uses which
   // would block SROA would also block SROA if applied directly to a pointer,
   // and so we can just add the integer in here. The only places where SROA is
   // preserved either cannot fire on an integer, or won't in-and-of themselves
   // disable SROA (ext) w/o some later use that we would see and disable.
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(I.getOperand(0), SROAArg, CostIt))
+  if (auto *SROAArg = getSROAArgForValueOrNull(I.getOperand(0)))
     SROAArgValues[&I] = SROAArg;
 
   return TargetTransformInfo::TCC_Free == TTI.getUserCost(&I);
 }
 
 bool CallAnalyzer::visitIntToPtr(IntToPtrInst &I) {
   // Propagate constants through ptrtoint.
   if (simplifyInstruction(I, [&](SmallVectorImpl<Constant *> &COps) {
         return ConstantExpr::getIntToPtr(COps[0], I.getType());
       }))
     return true;
 
   // Track base/offset pairs when round-tripped through a pointer without
   // modifications provided the integer is not too large.
   Value *Op = I.getOperand(0);
   unsigned IntegerSize = Op->getType()->getScalarSizeInBits();
   if (IntegerSize <= DL.getPointerTypeSizeInBits(I.getType())) {
     std::pair<Value *, APInt> BaseAndOffset = ConstantOffsetPtrs.lookup(Op);
     if (BaseAndOffset.first)
       ConstantOffsetPtrs[&I] = BaseAndOffset;
   }
 
   // "Propagate" SROA here in the same manner as we do for ptrtoint above.
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(Op, SROAArg, CostIt))
+  if (auto *SROAArg = getSROAArgForValueOrNull(Op))
     SROAArgValues[&I] = SROAArg;
 
   return TargetTransformInfo::TCC_Free == TTI.getUserCost(&I);
 }
 
 bool CallAnalyzer::visitCastInst(CastInst &I) {
   // Propagate constants through casts.
   if (simplifyInstruction(I, [&](SmallVectorImpl<Constant *> &COps) {
bool CallAnalyzer::visitCastInst(CastInst &I) {
   switch (I.getOpcode()) {
   case Instruction::FPTrunc:
   case Instruction::FPExt:
   case Instruction::UIToFP:
   case Instruction::SIToFP:
   case Instruction::FPToUI:
   case Instruction::FPToSI:
     if (TTI.getFPOpCost(I.getType()) == TargetTransformInfo::TCC_Expensive)
-      addCost(InlineConstants::CallPenalty);
+      onCallPenalty();
     break;
   default:
     break;
   }
 
   return TargetTransformInfo::TCC_Free == TTI.getUserCost(&I);
 }
 
if (InvokeInst *II = dyn_cast<InvokeInst>(&Call)) {
     if (isa<UnreachableInst>(II->getNormalDest()->getTerminator()))
       return false;
   } else if (isa<UnreachableInst>(Call.getParent()->getTerminator()))
     return false;
 
   return true;
 }
 
-bool CallAnalyzer::isColdCallSite(CallBase &Call,
+bool InlineCostCallAnalyzer::isColdCallSite(CallBase &Call,
                                             BlockFrequencyInfo *CallerBFI) {
   // If global profile summary is available, then callsite's coldness is
   // determined based on that.
   if (PSI && PSI->hasProfileSummary())
     return PSI->isColdCallSite(CallSite(&Call), CallerBFI);
 
   // Otherwise we need BFI to be available.
   if (!CallerBFI)
     return false;
 
   // Determine if the callsite is cold relative to caller's entry. We could
   // potentially cache the computation of scaled entry frequency, but the added
   // complexity is not worth it unless this scaling shows up high in the
   // profiles.
   const BranchProbability ColdProb(ColdCallSiteRelFreq, 100);
   auto CallSiteBB = Call.getParent();
   auto CallSiteFreq = CallerBFI->getBlockFreq(CallSiteBB);
   auto CallerEntryFreq =
       CallerBFI->getBlockFreq(&(Call.getCaller()->getEntryBlock()));
   return CallSiteFreq < CallerEntryFreq * ColdProb;
 }
 
 Optional<int>
-CallAnalyzer::getHotCallSiteThreshold(CallBase &Call,
+InlineCostCallAnalyzer::getHotCallSiteThreshold(CallBase &Call,
                                                 BlockFrequencyInfo *CallerBFI) {
 
   // If global profile summary is available, then callsite's hotness is
   // determined based on that.
   if (PSI && PSI->hasProfileSummary() &&
       PSI->isHotCallSite(CallSite(&Call), CallerBFI))
     return Params.HotCallSiteThreshold;
 
   // Otherwise we need BFI to be available and to have a locally hot callsite
InlineCostCallAnalyzer::getHotCallSiteThreshold(CallBase &Call,
   auto CallerEntryFreq = CallerBFI->getEntryFreq();
   if (CallSiteFreq >= CallerEntryFreq * HotCallSiteRelFreq)
     return Params.LocallyHotCallSiteThreshold;
 
   // Otherwise treat it normally.
   return None;
 }
 
-void CallAnalyzer::updateThreshold(CallBase &Call, Function &Callee) {
+void InlineCostCallAnalyzer::updateThreshold(CallBase &Call, Function &Callee) {
   // If no size growth is allowed for this inlining, set Threshold to 0.
   if (!allowSizeGrowth(Call)) {
     Threshold = 0;
     return;
   }
 
   Function *Caller = Call.getCaller();
 
bool CallAnalyzer::visitCmpInst(CmpInst &I) {
   // if we know the value (argument) can't be null
   if (I.isEquality() && isa<ConstantPointerNull>(I.getOperand(1)) &&
       isKnownNonNullInCallee(I.getOperand(0))) {
     bool IsNotEqual = I.getPredicate() == CmpInst::ICMP_NE;
     SimplifiedValues[&I] = IsNotEqual ? ConstantInt::getTrue(I.getType())
                                       : ConstantInt::getFalse(I.getType());
     return true;
   }
-  // Finally check for SROA candidates in comparisons.
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(I.getOperand(0), SROAArg, CostIt)) {
-    if (isa<ConstantPointerNull>(I.getOperand(1))) {
-      accumulateSROACost(CostIt, InlineConstants::InstrCost);
-      return true;
-    }
-
-    disableSROA(CostIt);
-  }
-
-  return false;
+  return handleSROA(I.getOperand(0), isa<ConstantPointerNull>(I.getOperand(1)));
 }
 
 bool CallAnalyzer::visitSub(BinaryOperator &I) {
   // Try to handle a special case: we can fold computing the difference of two
   // constant-related pointers.
   Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
   Value *LHSBase, *RHSBase;
   APInt LHSOffset, RHSOffset;
bool CallAnalyzer::visitBinaryOperator(BinaryOperator &I) {
 
   // If the instruction is floating point, and the target says this operation
   // is expensive, this may eventually become a library call. Treat the cost
   // as such. Unless it's fneg which can be implemented with an xor.
   using namespace llvm::PatternMatch;
   if (I.getType()->isFloatingPointTy() &&
       TTI.getFPOpCost(I.getType()) == TargetTransformInfo::TCC_Expensive &&
       !match(&I, m_FNeg(m_Value())))
-    addCost(InlineConstants::CallPenalty);
+    onCallPenalty();
 
   return false;
 }
 
 bool CallAnalyzer::visitFNeg(UnaryOperator &I) {
   Value *Op = I.getOperand(0);
   Constant *COp = dyn_cast<Constant>(Op);
   if (!COp)
bool CallAnalyzer::visitFNeg(UnaryOperator &I) {
 
   // Disable any SROA on arguments to arbitrary, unsimplified fneg.
   disableSROA(Op);
 
   return false;
 }
 
 bool CallAnalyzer::visitLoad(LoadInst &I) {
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(I.getPointerOperand(), SROAArg, CostIt)) {
-    if (I.isSimple()) {
-      accumulateSROACost(CostIt, InlineConstants::InstrCost);
-      return true;
-    }
-
-    disableSROA(CostIt);
-  }
+  if (handleSROA(I.getPointerOperand(), I.isSimple()))
 
   // If the data is already loaded from this address and hasn't been clobbered
   // by any stores or calls, this load is likely to be redundant and can be
   // eliminated.
   if (EnableLoadElimination &&
       !LoadAddrSet.insert(I.getPointerOperand()).second && I.isUnordered()) {
-    LoadEliminationCost += InlineConstants::InstrCost;
+    onLoadEliminationOpportunity();
     return true;
   }
 
   return false;
 }
 
 bool CallAnalyzer::visitStore(StoreInst &I) {
-  Value *SROAArg;
-  DenseMap<Value *, int>::iterator CostIt;
-  if (lookupSROAArgAndCost(I.getPointerOperand(), SROAArg, CostIt)) {
-    if (I.isSimple()) {
-      accumulateSROACost(CostIt, InlineConstants::InstrCost);
-      return true;
-    }
-
-    disableSROA(CostIt);
-  }
+  if (handleSROA(I.getPointerOperand(), I.isSimple()))
 
   // The store can potentially clobber loads and prevent repeated loads from
   // being eliminated.
   // FIXME:
   // 1. We can probably keep an initial set of eliminatable loads substracted
   // from the cost even when we finally see a store. We just need to disable
   // *further* accumulation of elimination savings.
   // 2. We should probably at some point thread MemorySSA for the callee into
bool CallAnalyzer::visitCallBase(CallBase &Call) {
   Value *Callee = Call.getCalledOperand();
   Function *F = dyn_cast_or_null<Function>(Callee);
   bool IsIndirectCall = !F;
   if (IsIndirectCall) {
     // Check if this happens to be an indirect function call to a known function
     // in this inline context. If not, we've done all we can.
     F = dyn_cast_or_null<Function>(SimplifiedValues.lookup(Callee));
     if (!F) {
-      // Pay the price of the argument setup. We account for the average 1
-      // instruction per call argument setup here.
-      addCost(Call.arg_size() * InlineConstants::InstrCost);
+      onCallArgumentSetup(Call);
 
       if (!Call.onlyReadsMemory())
         disableLoadElimination();
       return Base::visitCallBase(Call);
     }
   }
 
   assert(F && "Expected a call to a known function");
 
   // When we have a concrete function, first try to simplify it directly.
   if (simplifyCallSite(F, Call))
     return true;
 
   // Next check if it is an intrinsic we know about.
   // FIXME: Lift this into part of the InstVisitor.
   if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(&Call)) {
     switch (II->getIntrinsicID()) {
     default:
       if (!Call.onlyReadsMemory() && !isAssumeLikeIntrinsic(II))
         disableLoadElimination();
       return Base::visitCallBase(Call);
 
     case Intrinsic::load_relative:
-      // This is normally lowered to 4 LLVM instructions.
-      addCost(3 * InlineConstants::InstrCost);
+      onLoadRelativeIntrinsic();
       return false;
 
     case Intrinsic::memset:
     case Intrinsic::memcpy:
     case Intrinsic::memmove:
       disableLoadElimination();
       // SROA can usually chew through these intrinsics, but they aren't free.
       return false;
bool CallAnalyzer::visitCallBase(CallBase &Call) {
 
   if (F == Call.getFunction()) {
     // This flag will fully abort the analysis, so don't bother with anything
     // else.
     IsRecursiveCall = true;
     return false;
   }
 
   if (TTI.isLoweredToCall(F)) {
-    // We account for the average 1 instruction per call argument setup here.
-    addCost(Call.arg_size() * InlineConstants::InstrCost);
-
-    // If we have a constant that we are calling as a function, we can peer
-    // through it and see the function target. This happens not infrequently
-    // during devirtualization and so we want to give it a hefty bonus for
-    // inlining, but cap that bonus in the event that inlining wouldn't pan out.
-    // Pretend to inline the function, with a custom threshold.
-    if (IsIndirectCall && BoostIndirectCalls) {
-      auto IndirectCallParams = Params;
-      IndirectCallParams.DefaultThreshold =
-          InlineConstants::IndirectCallThreshold;
-      CallAnalyzer CA(TTI, GetAssumptionCache, GetBFI, PSI, ORE, *F, Call,
-                      IndirectCallParams, false);
-      if (CA.analyze()) {
-        // We were able to inline the indirect call! Subtract the cost from the
-        // threshold to get the bonus we want to apply, but don't go below zero.
-        Cost -= std::max(0, CA.getThreshold() - CA.getCost());
-      }
-    } else
-      // Otherwise simply add the cost for merely making the call.
-      addCost(InlineConstants::CallPenalty);
+    onLoweredCall(F, Call, IsIndirectCall);
   }
 
   if (!(Call.onlyReadsMemory() || (IsIndirectCall && F->onlyReadsMemory())))
     disableLoadElimination();
   return Base::visitCallBase(Call);
 }
 
 bool CallAnalyzer::visitReturnInst(ReturnInst &RI) {
if (!CondC) {
 
     std::pair<Value *, APInt> TrueBaseAndOffset =
         ConstantOffsetPtrs.lookup(TrueVal);
     std::pair<Value *, APInt> FalseBaseAndOffset =
         ConstantOffsetPtrs.lookup(FalseVal);
     if (TrueBaseAndOffset == FalseBaseAndOffset && TrueBaseAndOffset.first) {
       ConstantOffsetPtrs[&SI] = TrueBaseAndOffset;
 
-      Value *SROAArg;
-      DenseMap<Value *, int>::iterator CostIt;
-      if (lookupSROAArgAndCost(TrueVal, SROAArg, CostIt))
+      if (auto *SROAArg = getSROAArgForValueOrNull(TrueVal))
         SROAArgValues[&SI] = SROAArg;
       return true;
     }
 
     return Base::visitSelectInst(SI);
   }
 
   // Select condition is a constant.
bool CallAnalyzer::visitSelectInst(SelectInst &SI) {
   if (!CheckSROA)
     return true;
 
   std::pair<Value *, APInt> BaseAndOffset =
       ConstantOffsetPtrs.lookup(SelectedV);
   if (BaseAndOffset.first) {
     ConstantOffsetPtrs[&SI] = BaseAndOffset;
 
-    Value *SROAArg;
-    DenseMap<Value *, int>::iterator CostIt;
-    if (lookupSROAArgAndCost(SelectedV, SROAArg, CostIt))
+    if (auto *SROAArg = getSROAArgForValueOrNull(SelectedV))
       SROAArgValues[&SI] = SROAArg;
   }
 
   return true;
 }
 
 bool CallAnalyzer::visitSwitchInst(SwitchInst &SI) {
   // We model unconditional switches as free, see the comments on handling
bool CallAnalyzer::visitSwitchInst(SwitchInst &SI) {
   // of jump table/bit test/binary search tree. The cost of the switch is
   // proportional to the size of the tree or the size of jump table range.
   //
   // NB: We convert large switches which are just used to initialize large phi
   // nodes to lookup tables instead in simplify-cfg, so this shouldn't prevent
   // inlining those. It will prevent inlining in cases where the optimization
   // does not (yet) fire.
 
-  // Maximum valid cost increased in this function.
-  int CostUpperBound = INT_MAX - InlineConstants::InstrCost - 1;
-
   unsigned JumpTableSize = 0;
   BlockFrequencyInfo *BFI = GetBFI ? &((*GetBFI)(F)) : nullptr;
   unsigned NumCaseCluster =
       TTI.getEstimatedNumberOfCaseClusters(SI, JumpTableSize, PSI, BFI);
 
-  // If suitable for a jump table, consider the cost for the table size and
-  // branch to destination.
-  if (JumpTableSize) {
-    int64_t JTCost = (int64_t)JumpTableSize * InlineConstants::InstrCost +
-                     4 * InlineConstants::InstrCost;
-
-    addCost(JTCost, (int64_t)CostUpperBound);
-    return false;
-  }
-
-  // Considering forming a binary search, we should find the number of nodes
-  // which is same as the number of comparisons when lowered. For a given
-  // number of clusters, n, we can define a recursive function, f(n), to find
-  // the number of nodes in the tree. The recursion is :
-  // f(n) = 1 + f(n/2) + f (n - n/2), when n > 3,
-  // and f(n) = n, when n <= 3.
-  // This will lead a binary tree where the leaf should be either f(2) or f(3)
-  // when n > 3.  So, the number of comparisons from leaves should be n, while
-  // the number of non-leaf should be :
-  //   2^(log2(n) - 1) - 1
-  //   = 2^log2(n) * 2^-1 - 1
-  //   = n / 2 - 1.
-  // Considering comparisons from leaf and non-leaf nodes, we can estimate the
-  // number of comparisons in a simple closed form :
-  //   n + n / 2 - 1 = n * 3 / 2 - 1
-  if (NumCaseCluster <= 3) {
-    // Suppose a comparison includes one compare and one conditional branch.
-    addCost(NumCaseCluster * 2 * InlineConstants::InstrCost);
-    return false;
-  }
-
-  int64_t ExpectedNumberOfCompare = 3 * (int64_t)NumCaseCluster / 2 - 1;
-  int64_t SwitchCost = ExpectedNumberOfCompare * 2 * InlineConstants::InstrCost;
-
-  addCost(SwitchCost, (int64_t)CostUpperBound);
+  onFinalizeSwitch(JumpTableSize, NumCaseCluster);
   return false;
 }
 
 bool CallAnalyzer::visitIndirectBrInst(IndirectBrInst &IBI) {
   // We never want to inline functions that contain an indirectbr.  This is
   // incorrect because all the blockaddress's (in static global initializers
   // for example) would be referring to the original function, and this
   // indirect jump would jump from the inlined copy of the function into the
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
     // If the instruction simplified to a constant, there is no cost to this
     // instruction. Visit the instructions using our InstVisitor to account for
     // all of the per-instruction logic. The visit tree returns true if we
     // consumed the instruction in any way, and false if the instruction's base
     // cost should count against inlining.
     if (Base::visit(&*I))
       ++NumInstructionsSimplified;
     else
-      addCost(InlineConstants::InstrCost);
+      onCommonInstructionSimplification();
 
     using namespace ore;
     // If the visit this instruction detected an uninlinable pattern, abort.
     InlineResult IR;
     if (IsRecursiveCall)
       IR = "recursive";
     else if (ExposesReturnsTwice)
       IR = "exposes returns twice";
if (IsCallerRecursive &&
           return OptimizationRemarkMissed(DEBUG_TYPE, "NeverInline",
                                           &CandidateCall)
                  << NV("Callee", &F) << " is " << NV("InlineResult", IR.message)
                  << ". Cost is not fully computed";
         });
       return IR;
     }
 
-    // Check if we've passed the maximum possible threshold so we don't spin in
-    // huge basic blocks that will never inline.
-    if (Cost >= Threshold && !ComputeFullInlineCost)
+    if (shouldStop())
       return false;
   }
 
   return true;
 }
 
 /// Compute the base pointer and cumulative constant offsets for V.
 ///

 /// Returns true if inlining this call is viable, and false if it is not
 /// viable. It computes the cost and adjusts the threshold based on numerous
 /// factors and heuristics. If this method returns false but the computed cost
 /// is below the computed threshold, then inlining was forcibly disabled by
 /// some artifact of the routine.
 InlineResult CallAnalyzer::analyze() {
   ++NumCallsAnalyzed;
 
-  // Perform some tweaks to the cost and threshold based on the direct
-  // callsite information.
-
-  // We want to more aggressively inline vector-dense kernels, so up the
-  // threshold, and we'll lower it if the % of vector instructions gets too
-  // low. Note that these bonuses are some what arbitrary and evolved over time
-  // by accident as much as because they are principled bonuses.
-  //
-  // FIXME: It would be nice to remove all such bonuses. At least it would be
-  // nice to base the bonus values on something more scientific.
-  assert(NumInstructions == 0);
-  assert(NumVectorInstructions == 0);
-
-  // Update the threshold based on callsite properties
-  updateThreshold(CandidateCall, F);
-
-  // While Threshold depends on commandline options that can take negative
-  // values, we want to enforce the invariant that the computed threshold and
-  // bonuses are non-negative.
-  assert(Threshold >= 0);
-  assert(SingleBBBonus >= 0);
-  assert(VectorBonus >= 0);
-
-  // Speculatively apply all possible bonuses to Threshold. If cost exceeds
-  // this Threshold any time, and cost cannot decrease, we can stop processing
-  // the rest of the function body.
-  Threshold += (SingleBBBonus + VectorBonus);
-
-  // Give out bonuses for the callsite, as the instructions setting them up
-  // will be gone after inlining.
-  addCost(-getCallsiteCost(CandidateCall, DL));
-
-  // If this function uses the coldcc calling convention, prefer not to inline
-  // it.
-  if (F.getCallingConv() == CallingConv::Cold)
-    Cost += InlineConstants::ColdccPenalty;
-
-  // Check if we're done. This can happen due to bonuses and penalties.
-  if (Cost >= Threshold && !ComputeFullInlineCost)
-    return "high cost";
+  auto Result = onAnalysisStart();
+  if (!Result)
+    return Result;
 
   if (F.empty())
     return true;
 
   Function *Caller = CandidateCall.getFunction();
   // Check if the caller function is recursive itself.
   for (User *U : Caller->users()) {
     CallBase *Call = dyn_cast<CallBase>(U);
for (Function::arg_iterator FAI = F.arg_begin(), FAE = F.arg_end();
     if (Constant *C = dyn_cast<Constant>(CAI))
       SimplifiedValues[&*FAI] = C;
 
     Value *PtrArg = *CAI;
     if (ConstantInt *C = stripAndComputeInBoundsConstantOffsets(PtrArg)) {
       ConstantOffsetPtrs[&*FAI] = std::make_pair(PtrArg, C->getValue());
 
       // We can SROA any pointer arguments derived from alloca instructions.
-      if (isa<AllocaInst>(PtrArg)) {
-        SROAArgValues[&*FAI] = PtrArg;
-        SROAArgCosts[PtrArg] = 0;
+      if (auto *SROAArg = dyn_cast<AllocaInst>(PtrArg)) {
+        SROAArgValues[&*FAI] = SROAArg;
+        onInitializeSROAArg(SROAArg);
       }
     }
   }
   NumConstantArgs = SimplifiedValues.size();
   NumConstantOffsetPtrArgs = ConstantOffsetPtrs.size();
   NumAllocaArgs = SROAArgValues.size();
 
   // FIXME: If a caller has multiple calls to a callee, we end up recomputing
InlineResult CallAnalyzer::analyze() {
   // basic blocks in a breadth-first order as we insert live successors. To
   // accomplish this, prioritizing for small iterations because we exit after
   // crossing our threshold, we use a small-size optimized SetVector.
   typedef SetVector<BasicBlock *, SmallVector<BasicBlock *, 16>,
                     SmallPtrSet<BasicBlock *, 16>>
       BBSetVector;
   BBSetVector BBWorklist;
   BBWorklist.insert(&F.getEntryBlock());
-  bool SingleBB = true;
+
   // Note that we *must not* cache the size, this loop grows the worklist.
   for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
-    // Bail out the moment we cross the threshold. This means we'll under-count
-    // the cost, but only when undercounting doesn't matter.
-    if (Cost >= Threshold && !ComputeFullInlineCost)
+    if (shouldStop())
       break;
 
     BasicBlock *BB = BBWorklist[Idx];
     if (BB->empty())
       continue;
 
     // Disallow inlining a blockaddress with uses other than strictly callbr.
     // A blockaddress only has defined behavior for an indirect branch in the
for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
     }
 
     // If we're unable to select a particular successor, just count all of
     // them.
     for (unsigned TIdx = 0, TSize = TI->getNumSuccessors(); TIdx != TSize;
          ++TIdx)
       BBWorklist.insert(TI->getSuccessor(TIdx));
 
-    // If we had any successors at this point, than post-inlining is likely to
-    // have them as well. Note that we assume any basic blocks which existed
-    // due to branches or switches which folded above will also fold after
-    // inlining.
-    if (SingleBB && TI->getNumSuccessors() > 1) {
-      // Take off the bonus we applied to the threshold.
-      Threshold -= SingleBBBonus;
-      SingleBB = false;
-    }
+    onBlockAnalyzed(BB);
   }
 
   bool OnlyOneCallAndLocalLinkage = F.hasLocalLinkage() && F.hasOneUse() &&
                                     &F == CandidateCall.getCalledFunction();
   // If this is a noduplicate call, we can still inline as long as
   // inlining this would cause the removal of the caller (so the instruction
   // is not actually duplicated, just moved).
   if (!OnlyOneCallAndLocalLinkage && ContainsNoDuplicateCall)
     return "noduplicate";
 
-  // Loops generally act a lot like calls in that they act like barriers to
-  // movement, require a certain amount of setup, etc. So when optimising for
-  // size, we penalise any call sites that perform loops. We do this after all
-  // other costs here, so will likely only be dealing with relatively small
-  // functions (and hence DT and LI will hopefully be cheap).
-  if (Caller->hasMinSize()) {
-    DominatorTree DT(F);
-    LoopInfo LI(DT);
-    int NumLoops = 0;
-    for (Loop *L : LI) {
-      // Ignore loops that will not be executed
-      if (DeadBlocks.count(L->getHeader()))
-        continue;
-      NumLoops++;
-    }
-    addCost(NumLoops * InlineConstants::CallPenalty);
-  }
-
-  // We applied the maximum possible vector bonus at the beginning. Now,
-  // subtract the excess bonus, if any, from the Threshold before
-  // comparing against Cost.
-  if (NumVectorInstructions <= NumInstructions / 10)
-    Threshold -= VectorBonus;
-  else if (NumVectorInstructions <= NumInstructions / 2)
-    Threshold -= VectorBonus / 2;
-
-  return Cost < std::max(1, Threshold);
+  return finalizeAnalysis();
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
 /// Dump stats about this call's analysis.
-LLVM_DUMP_METHOD void CallAnalyzer::dump() {
+LLVM_DUMP_METHOD void InlineCostCallAnalyzer::dump() {
 #define DEBUG_PRINT_STAT(x) dbgs() << "      " #x ": " << x << "\n"
   DEBUG_PRINT_STAT(NumConstantArgs);
   DEBUG_PRINT_STAT(NumConstantOffsetPtrArgs);
   DEBUG_PRINT_STAT(NumAllocaArgs);
   DEBUG_PRINT_STAT(NumConstantPtrCmps);
   DEBUG_PRINT_STAT(NumConstantPtrDiffs);
   DEBUG_PRINT_STAT(NumInstructionsSimplified);
   DEBUG_PRINT_STAT(NumInstructions);
InlineCost llvm::getInlineCost(
 
   // Don't inline call sites marked noinline.
   if (Call.isNoInline())
     return llvm::InlineCost::getNever("noinline call site attribute");
 
   LLVM_DEBUG(llvm::dbgs() << "      Analyzing call of " << Callee->getName()
                           << "... (caller:" << Caller->getName() << ")\n");
 
-  CallAnalyzer CA(CalleeTTI, GetAssumptionCache, GetBFI, PSI, ORE, *Callee,
-                  Call, Params);
+  InlineCostCallAnalyzer CA(CalleeTTI, GetAssumptionCache, GetBFI, PSI, ORE,
+                            *Callee, Call, Params);
   InlineResult ShouldInline = CA.analyze();
 
   LLVM_DEBUG(CA.dump());
 
   // Check if there was a reason to force inlining or no inlining.
   if (!ShouldInline && CA.getCost() < CA.getThreshold())
     return InlineCost::getNever(ShouldInline.message);
   if (ShouldInline && CA.getCost() >= CA.getThreshold())
for (auto &II : *BI) {
       if (!ReturnsTwice && isa<CallInst>(Call) &&
           cast<CallInst>(Call)->canReturnTwice())
         return "exposes returns-twice attribute";
 
       if (Call->getCalledFunction())
         switch (Call->getCalledFunction()->getIntrinsicID()) {
         default:
           break;
+        case llvm::Intrinsic::icall_branch_funnel:
           // Disallow inlining of @llvm.icall.branch.funnel because current
           // backend can't separate call targets from call arguments.
-        case llvm::Intrinsic::icall_branch_funnel:
           return "disallowed inlining of @llvm.icall.branch.funnel";
+        case llvm::Intrinsic::localescape:
           // Disallow inlining functions that call @llvm.localescape. Doing this
           // correctly would require major changes to the inliner.
-        case llvm::Intrinsic::localescape:
           return "disallowed inlining of @llvm.localescape";
-        // Disallow inlining of functions that initialize VarArgs with va_start.
         case llvm::Intrinsic::vastart:
+          // Disallow inlining of functions that initialize VarArgs with
+          // va_start.
           return "contains VarArgs initialized with va_start";
         }
     }
   }
 
   return true;
 }