• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===- InlineCost.cpp - Cost analysis for inliner -------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements inline cost analysis.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "llvm/Analysis/InlineCost.h"
15 #include "llvm/Support/CallSite.h"
16 #include "llvm/CallingConv.h"
17 #include "llvm/IntrinsicInst.h"
18 #include "llvm/Target/TargetData.h"
19 #include "llvm/ADT/SmallPtrSet.h"
20 
21 using namespace llvm;
22 
23 /// callIsSmall - If a call is likely to lower to a single target instruction,
24 /// or is otherwise deemed small return true.
25 /// TODO: Perhaps calls like memcpy, strcpy, etc?
callIsSmall(const Function * F)26 bool llvm::callIsSmall(const Function *F) {
27   if (!F) return false;
28 
29   if (F->hasLocalLinkage()) return false;
30 
31   if (!F->hasName()) return false;
32 
33   StringRef Name = F->getName();
34 
35   // These will all likely lower to a single selection DAG node.
36   if (Name == "copysign" || Name == "copysignf" || Name == "copysignl" ||
37       Name == "fabs" || Name == "fabsf" || Name == "fabsl" ||
38       Name == "sin" || Name == "sinf" || Name == "sinl" ||
39       Name == "cos" || Name == "cosf" || Name == "cosl" ||
40       Name == "sqrt" || Name == "sqrtf" || Name == "sqrtl" )
41     return true;
42 
43   // These are all likely to be optimized into something smaller.
44   if (Name == "pow" || Name == "powf" || Name == "powl" ||
45       Name == "exp2" || Name == "exp2l" || Name == "exp2f" ||
46       Name == "floor" || Name == "floorf" || Name == "ceil" ||
47       Name == "round" || Name == "ffs" || Name == "ffsl" ||
48       Name == "abs" || Name == "labs" || Name == "llabs")
49     return true;
50 
51   return false;
52 }
53 
54 /// analyzeBasicBlock - Fill in the current structure with information gleaned
55 /// from the specified block.
analyzeBasicBlock(const BasicBlock * BB,const TargetData * TD)56 void CodeMetrics::analyzeBasicBlock(const BasicBlock *BB,
57                                     const TargetData *TD) {
58   ++NumBlocks;
59   unsigned NumInstsBeforeThisBB = NumInsts;
60   for (BasicBlock::const_iterator II = BB->begin(), E = BB->end();
61        II != E; ++II) {
62     if (isa<PHINode>(II)) continue;           // PHI nodes don't count.
63 
64     // Special handling for calls.
65     if (isa<CallInst>(II) || isa<InvokeInst>(II)) {
66       if (isa<DbgInfoIntrinsic>(II))
67         continue;  // Debug intrinsics don't count as size.
68 
69       ImmutableCallSite CS(cast<Instruction>(II));
70 
71       if (const Function *F = CS.getCalledFunction()) {
72         // If a function is both internal and has a single use, then it is
73         // extremely likely to get inlined in the future (it was probably
74         // exposed by an interleaved devirtualization pass).
75         if (F->hasInternalLinkage() && F->hasOneUse())
76           ++NumInlineCandidates;
77 
78         // If this call is to function itself, then the function is recursive.
79         // Inlining it into other functions is a bad idea, because this is
80         // basically just a form of loop peeling, and our metrics aren't useful
81         // for that case.
82         if (F == BB->getParent())
83           isRecursive = true;
84       }
85 
86       if (!isa<IntrinsicInst>(II) && !callIsSmall(CS.getCalledFunction())) {
87         // Each argument to a call takes on average one instruction to set up.
88         NumInsts += CS.arg_size();
89 
90         // We don't want inline asm to count as a call - that would prevent loop
91         // unrolling. The argument setup cost is still real, though.
92         if (!isa<InlineAsm>(CS.getCalledValue()))
93           ++NumCalls;
94       }
95     }
96 
97     if (const AllocaInst *AI = dyn_cast<AllocaInst>(II)) {
98       if (!AI->isStaticAlloca())
99         this->usesDynamicAlloca = true;
100     }
101 
102     if (isa<ExtractElementInst>(II) || II->getType()->isVectorTy())
103       ++NumVectorInsts;
104 
105     if (const CastInst *CI = dyn_cast<CastInst>(II)) {
106       // Noop casts, including ptr <-> int,  don't count.
107       if (CI->isLosslessCast() || isa<IntToPtrInst>(CI) ||
108           isa<PtrToIntInst>(CI))
109         continue;
110       // trunc to a native type is free (assuming the target has compare and
111       // shift-right of the same width).
112       if (isa<TruncInst>(CI) && TD &&
113           TD->isLegalInteger(TD->getTypeSizeInBits(CI->getType())))
114         continue;
115       // Result of a cmp instruction is often extended (to be used by other
116       // cmp instructions, logical or return instructions). These are usually
117       // nop on most sane targets.
118       if (isa<CmpInst>(CI->getOperand(0)))
119         continue;
120     } else if (const GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(II)){
121       // If a GEP has all constant indices, it will probably be folded with
122       // a load/store.
123       if (GEPI->hasAllConstantIndices())
124         continue;
125     }
126 
127     ++NumInsts;
128   }
129 
130   if (isa<ReturnInst>(BB->getTerminator()))
131     ++NumRets;
132 
133   // We never want to inline functions that contain an indirectbr.  This is
134   // incorrect because all the blockaddress's (in static global initializers
135   // for example) would be referring to the original function, and this indirect
136   // jump would jump from the inlined copy of the function into the original
137   // function which is extremely undefined behavior.
138   if (isa<IndirectBrInst>(BB->getTerminator()))
139     containsIndirectBr = true;
140 
141   // Remember NumInsts for this BB.
142   NumBBInsts[BB] = NumInsts - NumInstsBeforeThisBB;
143 }
144 
145 // CountCodeReductionForConstant - Figure out an approximation for how many
146 // instructions will be constant folded if the specified value is constant.
147 //
CountCodeReductionForConstant(Value * V)148 unsigned CodeMetrics::CountCodeReductionForConstant(Value *V) {
149   unsigned Reduction = 0;
150   for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E;++UI){
151     User *U = *UI;
152     if (isa<BranchInst>(U) || isa<SwitchInst>(U)) {
153       // We will be able to eliminate all but one of the successors.
154       const TerminatorInst &TI = cast<TerminatorInst>(*U);
155       const unsigned NumSucc = TI.getNumSuccessors();
156       unsigned Instrs = 0;
157       for (unsigned I = 0; I != NumSucc; ++I)
158         Instrs += NumBBInsts[TI.getSuccessor(I)];
159       // We don't know which blocks will be eliminated, so use the average size.
160       Reduction += InlineConstants::InstrCost*Instrs*(NumSucc-1)/NumSucc;
161     } else {
162       // Figure out if this instruction will be removed due to simple constant
163       // propagation.
164       Instruction &Inst = cast<Instruction>(*U);
165 
166       // We can't constant propagate instructions which have effects or
167       // read memory.
168       //
169       // FIXME: It would be nice to capture the fact that a load from a
170       // pointer-to-constant-global is actually a *really* good thing to zap.
171       // Unfortunately, we don't know the pointer that may get propagated here,
172       // so we can't make this decision.
173       if (Inst.mayReadFromMemory() || Inst.mayHaveSideEffects() ||
174           isa<AllocaInst>(Inst))
175         continue;
176 
177       bool AllOperandsConstant = true;
178       for (unsigned i = 0, e = Inst.getNumOperands(); i != e; ++i)
179         if (!isa<Constant>(Inst.getOperand(i)) && Inst.getOperand(i) != V) {
180           AllOperandsConstant = false;
181           break;
182         }
183 
184       if (AllOperandsConstant) {
185         // We will get to remove this instruction...
186         Reduction += InlineConstants::InstrCost;
187 
188         // And any other instructions that use it which become constants
189         // themselves.
190         Reduction += CountCodeReductionForConstant(&Inst);
191       }
192     }
193   }
194   return Reduction;
195 }
196 
197 // CountCodeReductionForAlloca - Figure out an approximation of how much smaller
198 // the function will be if it is inlined into a context where an argument
199 // becomes an alloca.
200 //
CountCodeReductionForAlloca(Value * V)201 unsigned CodeMetrics::CountCodeReductionForAlloca(Value *V) {
202   if (!V->getType()->isPointerTy()) return 0;  // Not a pointer
203   unsigned Reduction = 0;
204   for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E;++UI){
205     Instruction *I = cast<Instruction>(*UI);
206     if (isa<LoadInst>(I) || isa<StoreInst>(I))
207       Reduction += InlineConstants::InstrCost;
208     else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(I)) {
209       // If the GEP has variable indices, we won't be able to do much with it.
210       if (GEP->hasAllConstantIndices())
211         Reduction += CountCodeReductionForAlloca(GEP);
212     } else if (BitCastInst *BCI = dyn_cast<BitCastInst>(I)) {
213       // Track pointer through bitcasts.
214       Reduction += CountCodeReductionForAlloca(BCI);
215     } else {
216       // If there is some other strange instruction, we're not going to be able
217       // to do much if we inline this.
218       return 0;
219     }
220   }
221 
222   return Reduction;
223 }
224 
225 /// analyzeFunction - Fill in the current structure with information gleaned
226 /// from the specified function.
analyzeFunction(Function * F,const TargetData * TD)227 void CodeMetrics::analyzeFunction(Function *F, const TargetData *TD) {
228   // If this function contains a call to setjmp or _setjmp, never inline
229   // it.  This is a hack because we depend on the user marking their local
230   // variables as volatile if they are live across a setjmp call, and they
231   // probably won't do this in callers.
232   if (F->callsFunctionThatReturnsTwice())
233     callsSetJmp = true;
234 
235   // Look at the size of the callee.
236   for (Function::const_iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
237     analyzeBasicBlock(&*BB, TD);
238 }
239 
240 /// analyzeFunction - Fill in the current structure with information gleaned
241 /// from the specified function.
analyzeFunction(Function * F,const TargetData * TD)242 void InlineCostAnalyzer::FunctionInfo::analyzeFunction(Function *F,
243                                                        const TargetData *TD) {
244   Metrics.analyzeFunction(F, TD);
245 
246   // A function with exactly one return has it removed during the inlining
247   // process (see InlineFunction), so don't count it.
248   // FIXME: This knowledge should really be encoded outside of FunctionInfo.
249   if (Metrics.NumRets==1)
250     --Metrics.NumInsts;
251 
252   // Check out all of the arguments to the function, figuring out how much
253   // code can be eliminated if one of the arguments is a constant.
254   ArgumentWeights.reserve(F->arg_size());
255   for (Function::arg_iterator I = F->arg_begin(), E = F->arg_end(); I != E; ++I)
256     ArgumentWeights.push_back(ArgInfo(Metrics.CountCodeReductionForConstant(I),
257                                       Metrics.CountCodeReductionForAlloca(I)));
258 }
259 
260 /// NeverInline - returns true if the function should never be inlined into
261 /// any caller
NeverInline()262 bool InlineCostAnalyzer::FunctionInfo::NeverInline() {
263   return (Metrics.callsSetJmp || Metrics.isRecursive ||
264           Metrics.containsIndirectBr);
265 }
266 // getSpecializationBonus - The heuristic used to determine the per-call
267 // performance boost for using a specialization of Callee with argument
268 // specializedArgNo replaced by a constant.
getSpecializationBonus(Function * Callee,SmallVectorImpl<unsigned> & SpecializedArgNos)269 int InlineCostAnalyzer::getSpecializationBonus(Function *Callee,
270          SmallVectorImpl<unsigned> &SpecializedArgNos)
271 {
272   if (Callee->mayBeOverridden())
273     return 0;
274 
275   int Bonus = 0;
276   // If this function uses the coldcc calling convention, prefer not to
277   // specialize it.
278   if (Callee->getCallingConv() == CallingConv::Cold)
279     Bonus -= InlineConstants::ColdccPenalty;
280 
281   // Get information about the callee.
282   FunctionInfo *CalleeFI = &CachedFunctionInfo[Callee];
283 
284   // If we haven't calculated this information yet, do so now.
285   if (CalleeFI->Metrics.NumBlocks == 0)
286     CalleeFI->analyzeFunction(Callee, TD);
287 
288   unsigned ArgNo = 0;
289   unsigned i = 0;
290   for (Function::arg_iterator I = Callee->arg_begin(), E = Callee->arg_end();
291        I != E; ++I, ++ArgNo)
292     if (ArgNo == SpecializedArgNos[i]) {
293       ++i;
294       Bonus += CountBonusForConstant(I);
295     }
296 
297   // Calls usually take a long time, so they make the specialization gain
298   // smaller.
299   Bonus -= CalleeFI->Metrics.NumCalls * InlineConstants::CallPenalty;
300 
301   return Bonus;
302 }
303 
304 // ConstantFunctionBonus - Figure out how much of a bonus we can get for
305 // possibly devirtualizing a function. We'll subtract the size of the function
306 // we may wish to inline from the indirect call bonus providing a limit on
307 // growth. Leave an upper limit of 0 for the bonus - we don't want to penalize
308 // inlining because we decide we don't want to give a bonus for
309 // devirtualizing.
ConstantFunctionBonus(CallSite CS,Constant * C)310 int InlineCostAnalyzer::ConstantFunctionBonus(CallSite CS, Constant *C) {
311 
312   // This could just be NULL.
313   if (!C) return 0;
314 
315   Function *F = dyn_cast<Function>(C);
316   if (!F) return 0;
317 
318   int Bonus = InlineConstants::IndirectCallBonus + getInlineSize(CS, F);
319   return (Bonus > 0) ? 0 : Bonus;
320 }
321 
322 // CountBonusForConstant - Figure out an approximation for how much per-call
323 // performance boost we can expect if the specified value is constant.
CountBonusForConstant(Value * V,Constant * C)324 int InlineCostAnalyzer::CountBonusForConstant(Value *V, Constant *C) {
325   unsigned Bonus = 0;
326   for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E;++UI){
327     User *U = *UI;
328     if (CallInst *CI = dyn_cast<CallInst>(U)) {
329       // Turning an indirect call into a direct call is a BIG win
330       if (CI->getCalledValue() == V)
331         Bonus += ConstantFunctionBonus(CallSite(CI), C);
332     } else if (InvokeInst *II = dyn_cast<InvokeInst>(U)) {
333       // Turning an indirect call into a direct call is a BIG win
334       if (II->getCalledValue() == V)
335         Bonus += ConstantFunctionBonus(CallSite(II), C);
336     }
337     // FIXME: Eliminating conditional branches and switches should
338     // also yield a per-call performance boost.
339     else {
340       // Figure out the bonuses that wll accrue due to simple constant
341       // propagation.
342       Instruction &Inst = cast<Instruction>(*U);
343 
344       // We can't constant propagate instructions which have effects or
345       // read memory.
346       //
347       // FIXME: It would be nice to capture the fact that a load from a
348       // pointer-to-constant-global is actually a *really* good thing to zap.
349       // Unfortunately, we don't know the pointer that may get propagated here,
350       // so we can't make this decision.
351       if (Inst.mayReadFromMemory() || Inst.mayHaveSideEffects() ||
352           isa<AllocaInst>(Inst))
353         continue;
354 
355       bool AllOperandsConstant = true;
356       for (unsigned i = 0, e = Inst.getNumOperands(); i != e; ++i)
357         if (!isa<Constant>(Inst.getOperand(i)) && Inst.getOperand(i) != V) {
358           AllOperandsConstant = false;
359           break;
360         }
361 
362       if (AllOperandsConstant)
363         Bonus += CountBonusForConstant(&Inst);
364     }
365   }
366 
367   return Bonus;
368 }
369 
getInlineSize(CallSite CS,Function * Callee)370 int InlineCostAnalyzer::getInlineSize(CallSite CS, Function *Callee) {
371   // Get information about the callee.
372   FunctionInfo *CalleeFI = &CachedFunctionInfo[Callee];
373 
374   // If we haven't calculated this information yet, do so now.
375   if (CalleeFI->Metrics.NumBlocks == 0)
376     CalleeFI->analyzeFunction(Callee, TD);
377 
378   // InlineCost - This value measures how good of an inline candidate this call
379   // site is to inline.  A lower inline cost make is more likely for the call to
380   // be inlined.  This value may go negative.
381   //
382   int InlineCost = 0;
383 
384   // Compute any size reductions we can expect due to arguments being passed into
385   // the function.
386   //
387   unsigned ArgNo = 0;
388   CallSite::arg_iterator I = CS.arg_begin();
389   for (Function::arg_iterator FI = Callee->arg_begin(), FE = Callee->arg_end();
390        FI != FE; ++I, ++FI, ++ArgNo) {
391 
392     // If an alloca is passed in, inlining this function is likely to allow
393     // significant future optimization possibilities (like scalar promotion, and
394     // scalarization), so encourage the inlining of the function.
395     //
396     if (isa<AllocaInst>(I))
397       InlineCost -= CalleeFI->ArgumentWeights[ArgNo].AllocaWeight;
398 
399     // If this is a constant being passed into the function, use the argument
400     // weights calculated for the callee to determine how much will be folded
401     // away with this information.
402     else if (isa<Constant>(I))
403       InlineCost -= CalleeFI->ArgumentWeights[ArgNo].ConstantWeight;
404   }
405 
406   // Each argument passed in has a cost at both the caller and the callee
407   // sides.  Measurements show that each argument costs about the same as an
408   // instruction.
409   InlineCost -= (CS.arg_size() * InlineConstants::InstrCost);
410 
411   // Now that we have considered all of the factors that make the call site more
412   // likely to be inlined, look at factors that make us not want to inline it.
413 
414   // Calls usually take a long time, so they make the inlining gain smaller.
415   InlineCost += CalleeFI->Metrics.NumCalls * InlineConstants::CallPenalty;
416 
417   // Look at the size of the callee. Each instruction counts as 5.
418   InlineCost += CalleeFI->Metrics.NumInsts*InlineConstants::InstrCost;
419 
420   return InlineCost;
421 }
422 
getInlineBonuses(CallSite CS,Function * Callee)423 int InlineCostAnalyzer::getInlineBonuses(CallSite CS, Function *Callee) {
424   // Get information about the callee.
425   FunctionInfo *CalleeFI = &CachedFunctionInfo[Callee];
426 
427   // If we haven't calculated this information yet, do so now.
428   if (CalleeFI->Metrics.NumBlocks == 0)
429     CalleeFI->analyzeFunction(Callee, TD);
430 
431   bool isDirectCall = CS.getCalledFunction() == Callee;
432   Instruction *TheCall = CS.getInstruction();
433   int Bonus = 0;
434 
435   // If there is only one call of the function, and it has internal linkage,
436   // make it almost guaranteed to be inlined.
437   //
438   if (Callee->hasLocalLinkage() && Callee->hasOneUse() && isDirectCall)
439     Bonus += InlineConstants::LastCallToStaticBonus;
440 
441   // If the instruction after the call, or if the normal destination of the
442   // invoke is an unreachable instruction, the function is noreturn.  As such,
443   // there is little point in inlining this.
444   if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
445     if (isa<UnreachableInst>(II->getNormalDest()->begin()))
446       Bonus += InlineConstants::NoreturnPenalty;
447   } else if (isa<UnreachableInst>(++BasicBlock::iterator(TheCall)))
448     Bonus += InlineConstants::NoreturnPenalty;
449 
450   // If this function uses the coldcc calling convention, prefer not to inline
451   // it.
452   if (Callee->getCallingConv() == CallingConv::Cold)
453     Bonus += InlineConstants::ColdccPenalty;
454 
455   // Add to the inline quality for properties that make the call valuable to
456   // inline.  This includes factors that indicate that the result of inlining
457   // the function will be optimizable.  Currently this just looks at arguments
458   // passed into the function.
459   //
460   CallSite::arg_iterator I = CS.arg_begin();
461   for (Function::arg_iterator FI = Callee->arg_begin(), FE = Callee->arg_end();
462        FI != FE; ++I, ++FI)
463     // Compute any constant bonus due to inlining we want to give here.
464     if (isa<Constant>(I))
465       Bonus += CountBonusForConstant(FI, cast<Constant>(I));
466 
467   return Bonus;
468 }
469 
470 // getInlineCost - The heuristic used to determine if we should inline the
471 // function call or not.
472 //
getInlineCost(CallSite CS,SmallPtrSet<const Function *,16> & NeverInline)473 InlineCost InlineCostAnalyzer::getInlineCost(CallSite CS,
474                                SmallPtrSet<const Function*, 16> &NeverInline) {
475   return getInlineCost(CS, CS.getCalledFunction(), NeverInline);
476 }
477 
getInlineCost(CallSite CS,Function * Callee,SmallPtrSet<const Function *,16> & NeverInline)478 InlineCost InlineCostAnalyzer::getInlineCost(CallSite CS,
479                                Function *Callee,
480                                SmallPtrSet<const Function*, 16> &NeverInline) {
481   Instruction *TheCall = CS.getInstruction();
482   Function *Caller = TheCall->getParent()->getParent();
483 
484   // Don't inline functions which can be redefined at link-time to mean
485   // something else.  Don't inline functions marked noinline or call sites
486   // marked noinline.
487   if (Callee->mayBeOverridden() ||
488       Callee->hasFnAttr(Attribute::NoInline) || NeverInline.count(Callee) ||
489       CS.isNoInline())
490     return llvm::InlineCost::getNever();
491 
492   // Get information about the callee.
493   FunctionInfo *CalleeFI = &CachedFunctionInfo[Callee];
494 
495   // If we haven't calculated this information yet, do so now.
496   if (CalleeFI->Metrics.NumBlocks == 0)
497     CalleeFI->analyzeFunction(Callee, TD);
498 
499   // If we should never inline this, return a huge cost.
500   if (CalleeFI->NeverInline())
501     return InlineCost::getNever();
502 
503   // FIXME: It would be nice to kill off CalleeFI->NeverInline. Then we
504   // could move this up and avoid computing the FunctionInfo for
505   // things we are going to just return always inline for. This
506   // requires handling setjmp somewhere else, however.
507   if (!Callee->isDeclaration() && Callee->hasFnAttr(Attribute::AlwaysInline))
508     return InlineCost::getAlways();
509 
510   if (CalleeFI->Metrics.usesDynamicAlloca) {
511     // Get information about the caller.
512     FunctionInfo &CallerFI = CachedFunctionInfo[Caller];
513 
514     // If we haven't calculated this information yet, do so now.
515     if (CallerFI.Metrics.NumBlocks == 0) {
516       CallerFI.analyzeFunction(Caller, TD);
517 
518       // Recompute the CalleeFI pointer, getting Caller could have invalidated
519       // it.
520       CalleeFI = &CachedFunctionInfo[Callee];
521     }
522 
523     // Don't inline a callee with dynamic alloca into a caller without them.
524     // Functions containing dynamic alloca's are inefficient in various ways;
525     // don't create more inefficiency.
526     if (!CallerFI.Metrics.usesDynamicAlloca)
527       return InlineCost::getNever();
528   }
529 
530   // InlineCost - This value measures how good of an inline candidate this call
531   // site is to inline.  A lower inline cost make is more likely for the call to
532   // be inlined.  This value may go negative due to the fact that bonuses
533   // are negative numbers.
534   //
535   int InlineCost = getInlineSize(CS, Callee) + getInlineBonuses(CS, Callee);
536   return llvm::InlineCost::get(InlineCost);
537 }
538 
539 // getSpecializationCost - The heuristic used to determine the code-size
540 // impact of creating a specialized version of Callee with argument
541 // SpecializedArgNo replaced by a constant.
getSpecializationCost(Function * Callee,SmallVectorImpl<unsigned> & SpecializedArgNos)542 InlineCost InlineCostAnalyzer::getSpecializationCost(Function *Callee,
543                                SmallVectorImpl<unsigned> &SpecializedArgNos)
544 {
545   // Don't specialize functions which can be redefined at link-time to mean
546   // something else.
547   if (Callee->mayBeOverridden())
548     return llvm::InlineCost::getNever();
549 
550   // Get information about the callee.
551   FunctionInfo *CalleeFI = &CachedFunctionInfo[Callee];
552 
553   // If we haven't calculated this information yet, do so now.
554   if (CalleeFI->Metrics.NumBlocks == 0)
555     CalleeFI->analyzeFunction(Callee, TD);
556 
557   int Cost = 0;
558 
559   // Look at the original size of the callee.  Each instruction counts as 5.
560   Cost += CalleeFI->Metrics.NumInsts * InlineConstants::InstrCost;
561 
562   // Offset that with the amount of code that can be constant-folded
563   // away with the given arguments replaced by constants.
564   for (SmallVectorImpl<unsigned>::iterator an = SpecializedArgNos.begin(),
565        ae = SpecializedArgNos.end(); an != ae; ++an)
566     Cost -= CalleeFI->ArgumentWeights[*an].ConstantWeight;
567 
568   return llvm::InlineCost::get(Cost);
569 }
570 
571 // getInlineFudgeFactor - Return a > 1.0 factor if the inliner should use a
572 // higher threshold to determine if the function call should be inlined.
getInlineFudgeFactor(CallSite CS)573 float InlineCostAnalyzer::getInlineFudgeFactor(CallSite CS) {
574   Function *Callee = CS.getCalledFunction();
575 
576   // Get information about the callee.
577   FunctionInfo &CalleeFI = CachedFunctionInfo[Callee];
578 
579   // If we haven't calculated this information yet, do so now.
580   if (CalleeFI.Metrics.NumBlocks == 0)
581     CalleeFI.analyzeFunction(Callee, TD);
582 
583   float Factor = 1.0f;
584   // Single BB functions are often written to be inlined.
585   if (CalleeFI.Metrics.NumBlocks == 1)
586     Factor += 0.5f;
587 
588   // Be more aggressive if the function contains a good chunk (if it mades up
589   // at least 10% of the instructions) of vector instructions.
590   if (CalleeFI.Metrics.NumVectorInsts > CalleeFI.Metrics.NumInsts/2)
591     Factor += 2.0f;
592   else if (CalleeFI.Metrics.NumVectorInsts > CalleeFI.Metrics.NumInsts/10)
593     Factor += 1.5f;
594   return Factor;
595 }
596 
597 /// growCachedCostInfo - update the cached cost info for Caller after Callee has
598 /// been inlined.
599 void
growCachedCostInfo(Function * Caller,Function * Callee)600 InlineCostAnalyzer::growCachedCostInfo(Function *Caller, Function *Callee) {
601   CodeMetrics &CallerMetrics = CachedFunctionInfo[Caller].Metrics;
602 
603   // For small functions we prefer to recalculate the cost for better accuracy.
604   if (CallerMetrics.NumBlocks < 10 && CallerMetrics.NumInsts < 1000) {
605     resetCachedCostInfo(Caller);
606     return;
607   }
608 
609   // For large functions, we can save a lot of computation time by skipping
610   // recalculations.
611   if (CallerMetrics.NumCalls > 0)
612     --CallerMetrics.NumCalls;
613 
614   if (Callee == 0) return;
615 
616   CodeMetrics &CalleeMetrics = CachedFunctionInfo[Callee].Metrics;
617 
618   // If we don't have metrics for the callee, don't recalculate them just to
619   // update an approximation in the caller.  Instead, just recalculate the
620   // caller info from scratch.
621   if (CalleeMetrics.NumBlocks == 0) {
622     resetCachedCostInfo(Caller);
623     return;
624   }
625 
626   // Since CalleeMetrics were already calculated, we know that the CallerMetrics
627   // reference isn't invalidated: both were in the DenseMap.
628   CallerMetrics.usesDynamicAlloca |= CalleeMetrics.usesDynamicAlloca;
629 
630   // FIXME: If any of these three are true for the callee, the callee was
631   // not inlined into the caller, so I think they're redundant here.
632   CallerMetrics.callsSetJmp |= CalleeMetrics.callsSetJmp;
633   CallerMetrics.isRecursive |= CalleeMetrics.isRecursive;
634   CallerMetrics.containsIndirectBr |= CalleeMetrics.containsIndirectBr;
635 
636   CallerMetrics.NumInsts += CalleeMetrics.NumInsts;
637   CallerMetrics.NumBlocks += CalleeMetrics.NumBlocks;
638   CallerMetrics.NumCalls += CalleeMetrics.NumCalls;
639   CallerMetrics.NumVectorInsts += CalleeMetrics.NumVectorInsts;
640   CallerMetrics.NumRets += CalleeMetrics.NumRets;
641 
642   // analyzeBasicBlock counts each function argument as an inst.
643   if (CallerMetrics.NumInsts >= Callee->arg_size())
644     CallerMetrics.NumInsts -= Callee->arg_size();
645   else
646     CallerMetrics.NumInsts = 0;
647 
648   // We are not updating the argument weights. We have already determined that
649   // Caller is a fairly large function, so we accept the loss of precision.
650 }
651 
652 /// clear - empty the cache of inline costs
clear()653 void InlineCostAnalyzer::clear() {
654   CachedFunctionInfo.clear();
655 }
656