xref: /freebsd/contrib/llvm-project/llvm/lib/Analysis/LoopCacheAnalysis.cpp (revision b1879975794772ee51f0b4865753364c7d7626c3)
1 //===- LoopCacheAnalysis.cpp - Loop Cache Analysis -------------------------==//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
6 // See https://llvm.org/LICENSE.txt for license information.
7 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
8 //
9 //===----------------------------------------------------------------------===//
10 ///
11 /// \file
12 /// This file defines the implementation for the loop cache analysis.
13 /// The implementation is largely based on the following paper:
14 ///
15 ///       Compiler Optimizations for Improving Data Locality
16 ///       By: Steve Carr, Katherine S. McKinley, Chau-Wen Tseng
17 ///       http://www.cs.utexas.edu/users/mckinley/papers/asplos-1994.pdf
18 ///
19 /// The general approach taken to estimate the number of cache lines used by the
20 /// memory references in an inner loop is:
21 ///    1. Partition memory references that exhibit temporal or spacial reuse
22 ///       into reference groups.
23 ///    2. For each loop L in the a loop nest LN:
24 ///       a. Compute the cost of the reference group
25 ///       b. Compute the loop cost by summing up the reference groups costs
26 //===----------------------------------------------------------------------===//
27 
28 #include "llvm/Analysis/LoopCacheAnalysis.h"
29 #include "llvm/ADT/BreadthFirstIterator.h"
30 #include "llvm/ADT/Sequence.h"
31 #include "llvm/ADT/SmallVector.h"
32 #include "llvm/Analysis/AliasAnalysis.h"
33 #include "llvm/Analysis/Delinearization.h"
34 #include "llvm/Analysis/DependenceAnalysis.h"
35 #include "llvm/Analysis/LoopInfo.h"
36 #include "llvm/Analysis/ScalarEvolutionExpressions.h"
37 #include "llvm/Analysis/TargetTransformInfo.h"
38 #include "llvm/Support/CommandLine.h"
39 #include "llvm/Support/Debug.h"
40 
41 using namespace llvm;
42 
43 #define DEBUG_TYPE "loop-cache-cost"
44 
45 static cl::opt<unsigned> DefaultTripCount(
46     "default-trip-count", cl::init(100), cl::Hidden,
47     cl::desc("Use this to specify the default trip count of a loop"));
48 
49 // In this analysis two array references are considered to exhibit temporal
50 // reuse if they access either the same memory location, or a memory location
51 // with distance smaller than a configurable threshold.
52 static cl::opt<unsigned> TemporalReuseThreshold(
53     "temporal-reuse-threshold", cl::init(2), cl::Hidden,
54     cl::desc("Use this to specify the max. distance between array elements "
55              "accessed in a loop so that the elements are classified to have "
56              "temporal reuse"));
57 
58 /// Retrieve the innermost loop in the given loop nest \p Loops. It returns a
59 /// nullptr if any loops in the loop vector supplied has more than one sibling.
60 /// The loop vector is expected to contain loops collected in breadth-first
61 /// order.
62 static Loop *getInnerMostLoop(const LoopVectorTy &Loops) {
63   assert(!Loops.empty() && "Expecting a non-empy loop vector");
64 
65   Loop *LastLoop = Loops.back();
66   Loop *ParentLoop = LastLoop->getParentLoop();
67 
68   if (ParentLoop == nullptr) {
69     assert(Loops.size() == 1 && "Expecting a single loop");
70     return LastLoop;
71   }
72 
73   return (llvm::is_sorted(Loops,
74                           [](const Loop *L1, const Loop *L2) {
75                             return L1->getLoopDepth() < L2->getLoopDepth();
76                           }))
77              ? LastLoop
78              : nullptr;
79 }
80 
81 static bool isOneDimensionalArray(const SCEV &AccessFn, const SCEV &ElemSize,
82                                   const Loop &L, ScalarEvolution &SE) {
83   const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(&AccessFn);
84   if (!AR || !AR->isAffine())
85     return false;
86 
87   assert(AR->getLoop() && "AR should have a loop");
88 
89   // Check that start and increment are not add recurrences.
90   const SCEV *Start = AR->getStart();
91   const SCEV *Step = AR->getStepRecurrence(SE);
92   if (isa<SCEVAddRecExpr>(Start) || isa<SCEVAddRecExpr>(Step))
93     return false;
94 
95   // Check that start and increment are both invariant in the loop.
96   if (!SE.isLoopInvariant(Start, &L) || !SE.isLoopInvariant(Step, &L))
97     return false;
98 
99   const SCEV *StepRec = AR->getStepRecurrence(SE);
100   if (StepRec && SE.isKnownNegative(StepRec))
101     StepRec = SE.getNegativeSCEV(StepRec);
102 
103   return StepRec == &ElemSize;
104 }
105 
106 /// Compute the trip count for the given loop \p L or assume a default value if
107 /// it is not a compile time constant. Return the SCEV expression for the trip
108 /// count.
109 static const SCEV *computeTripCount(const Loop &L, const SCEV &ElemSize,
110                                     ScalarEvolution &SE) {
111   const SCEV *BackedgeTakenCount = SE.getBackedgeTakenCount(&L);
112   const SCEV *TripCount = (!isa<SCEVCouldNotCompute>(BackedgeTakenCount) &&
113                            isa<SCEVConstant>(BackedgeTakenCount))
114                               ? SE.getTripCountFromExitCount(BackedgeTakenCount)
115                               : nullptr;
116 
117   if (!TripCount) {
118     LLVM_DEBUG(dbgs() << "Trip count of loop " << L.getName()
119                << " could not be computed, using DefaultTripCount\n");
120     TripCount = SE.getConstant(ElemSize.getType(), DefaultTripCount);
121   }
122 
123   return TripCount;
124 }
125 
126 //===----------------------------------------------------------------------===//
127 // IndexedReference implementation
128 //
129 raw_ostream &llvm::operator<<(raw_ostream &OS, const IndexedReference &R) {
130   if (!R.IsValid) {
131     OS << R.StoreOrLoadInst;
132     OS << ", IsValid=false.";
133     return OS;
134   }
135 
136   OS << *R.BasePointer;
137   for (const SCEV *Subscript : R.Subscripts)
138     OS << "[" << *Subscript << "]";
139 
140   OS << ", Sizes: ";
141   for (const SCEV *Size : R.Sizes)
142     OS << "[" << *Size << "]";
143 
144   return OS;
145 }
146 
147 IndexedReference::IndexedReference(Instruction &StoreOrLoadInst,
148                                    const LoopInfo &LI, ScalarEvolution &SE)
149     : StoreOrLoadInst(StoreOrLoadInst), SE(SE) {
150   assert((isa<StoreInst>(StoreOrLoadInst) || isa<LoadInst>(StoreOrLoadInst)) &&
151          "Expecting a load or store instruction");
152 
153   IsValid = delinearize(LI);
154   if (IsValid)
155     LLVM_DEBUG(dbgs().indent(2) << "Succesfully delinearized: " << *this
156                                 << "\n");
157 }
158 
159 std::optional<bool>
160 IndexedReference::hasSpacialReuse(const IndexedReference &Other, unsigned CLS,
161                                   AAResults &AA) const {
162   assert(IsValid && "Expecting a valid reference");
163 
164   if (BasePointer != Other.getBasePointer() && !isAliased(Other, AA)) {
165     LLVM_DEBUG(dbgs().indent(2)
166                << "No spacial reuse: different base pointers\n");
167     return false;
168   }
169 
170   unsigned NumSubscripts = getNumSubscripts();
171   if (NumSubscripts != Other.getNumSubscripts()) {
172     LLVM_DEBUG(dbgs().indent(2)
173                << "No spacial reuse: different number of subscripts\n");
174     return false;
175   }
176 
177   // all subscripts must be equal, except the leftmost one (the last one).
178   for (auto SubNum : seq<unsigned>(0, NumSubscripts - 1)) {
179     if (getSubscript(SubNum) != Other.getSubscript(SubNum)) {
180       LLVM_DEBUG(dbgs().indent(2) << "No spacial reuse, different subscripts: "
181                                   << "\n\t" << *getSubscript(SubNum) << "\n\t"
182                                   << *Other.getSubscript(SubNum) << "\n");
183       return false;
184     }
185   }
186 
187   // the difference between the last subscripts must be less than the cache line
188   // size.
189   const SCEV *LastSubscript = getLastSubscript();
190   const SCEV *OtherLastSubscript = Other.getLastSubscript();
191   const SCEVConstant *Diff = dyn_cast<SCEVConstant>(
192       SE.getMinusSCEV(LastSubscript, OtherLastSubscript));
193 
194   if (Diff == nullptr) {
195     LLVM_DEBUG(dbgs().indent(2)
196                << "No spacial reuse, difference between subscript:\n\t"
197                << *LastSubscript << "\n\t" << OtherLastSubscript
198                << "\nis not constant.\n");
199     return std::nullopt;
200   }
201 
202   bool InSameCacheLine = (Diff->getValue()->getSExtValue() < CLS);
203 
204   LLVM_DEBUG({
205     if (InSameCacheLine)
206       dbgs().indent(2) << "Found spacial reuse.\n";
207     else
208       dbgs().indent(2) << "No spacial reuse.\n";
209   });
210 
211   return InSameCacheLine;
212 }
213 
214 std::optional<bool>
215 IndexedReference::hasTemporalReuse(const IndexedReference &Other,
216                                    unsigned MaxDistance, const Loop &L,
217                                    DependenceInfo &DI, AAResults &AA) const {
218   assert(IsValid && "Expecting a valid reference");
219 
220   if (BasePointer != Other.getBasePointer() && !isAliased(Other, AA)) {
221     LLVM_DEBUG(dbgs().indent(2)
222                << "No temporal reuse: different base pointer\n");
223     return false;
224   }
225 
226   std::unique_ptr<Dependence> D =
227       DI.depends(&StoreOrLoadInst, &Other.StoreOrLoadInst, true);
228 
229   if (D == nullptr) {
230     LLVM_DEBUG(dbgs().indent(2) << "No temporal reuse: no dependence\n");
231     return false;
232   }
233 
234   if (D->isLoopIndependent()) {
235     LLVM_DEBUG(dbgs().indent(2) << "Found temporal reuse\n");
236     return true;
237   }
238 
239   // Check the dependence distance at every loop level. There is temporal reuse
240   // if the distance at the given loop's depth is small (|d| <= MaxDistance) and
241   // it is zero at every other loop level.
242   int LoopDepth = L.getLoopDepth();
243   int Levels = D->getLevels();
244   for (int Level = 1; Level <= Levels; ++Level) {
245     const SCEV *Distance = D->getDistance(Level);
246     const SCEVConstant *SCEVConst = dyn_cast_or_null<SCEVConstant>(Distance);
247 
248     if (SCEVConst == nullptr) {
249       LLVM_DEBUG(dbgs().indent(2) << "No temporal reuse: distance unknown\n");
250       return std::nullopt;
251     }
252 
253     const ConstantInt &CI = *SCEVConst->getValue();
254     if (Level != LoopDepth && !CI.isZero()) {
255       LLVM_DEBUG(dbgs().indent(2)
256                  << "No temporal reuse: distance is not zero at depth=" << Level
257                  << "\n");
258       return false;
259     } else if (Level == LoopDepth && CI.getSExtValue() > MaxDistance) {
260       LLVM_DEBUG(
261           dbgs().indent(2)
262           << "No temporal reuse: distance is greater than MaxDistance at depth="
263           << Level << "\n");
264       return false;
265     }
266   }
267 
268   LLVM_DEBUG(dbgs().indent(2) << "Found temporal reuse\n");
269   return true;
270 }
271 
272 CacheCostTy IndexedReference::computeRefCost(const Loop &L,
273                                              unsigned CLS) const {
274   assert(IsValid && "Expecting a valid reference");
275   LLVM_DEBUG({
276     dbgs().indent(2) << "Computing cache cost for:\n";
277     dbgs().indent(4) << *this << "\n";
278   });
279 
280   // If the indexed reference is loop invariant the cost is one.
281   if (isLoopInvariant(L)) {
282     LLVM_DEBUG(dbgs().indent(4) << "Reference is loop invariant: RefCost=1\n");
283     return 1;
284   }
285 
286   const SCEV *TripCount = computeTripCount(L, *Sizes.back(), SE);
287   assert(TripCount && "Expecting valid TripCount");
288   LLVM_DEBUG(dbgs() << "TripCount=" << *TripCount << "\n");
289 
290   const SCEV *RefCost = nullptr;
291   const SCEV *Stride = nullptr;
292   if (isConsecutive(L, Stride, CLS)) {
293     // If the indexed reference is 'consecutive' the cost is
294     // (TripCount*Stride)/CLS.
295     assert(Stride != nullptr &&
296            "Stride should not be null for consecutive access!");
297     Type *WiderType = SE.getWiderType(Stride->getType(), TripCount->getType());
298     const SCEV *CacheLineSize = SE.getConstant(WiderType, CLS);
299     Stride = SE.getNoopOrAnyExtend(Stride, WiderType);
300     TripCount = SE.getNoopOrZeroExtend(TripCount, WiderType);
301     const SCEV *Numerator = SE.getMulExpr(Stride, TripCount);
302     // Round the fractional cost up to the nearest integer number.
303     // The impact is the most significant when cost is calculated
304     // to be a number less than one, because it makes more sense
305     // to say one cache line is used rather than zero cache line
306     // is used.
307     RefCost = SE.getUDivCeilSCEV(Numerator, CacheLineSize);
308 
309     LLVM_DEBUG(dbgs().indent(4)
310                << "Access is consecutive: RefCost=(TripCount*Stride)/CLS="
311                << *RefCost << "\n");
312   } else {
313     // If the indexed reference is not 'consecutive' the cost is proportional to
314     // the trip count and the depth of the dimension which the subject loop
315     // subscript is accessing. We try to estimate this by multiplying the cost
316     // by the trip counts of loops corresponding to the inner dimensions. For
317     // example, given the indexed reference 'A[i][j][k]', and assuming the
318     // i-loop is in the innermost position, the cost would be equal to the
319     // iterations of the i-loop multiplied by iterations of the j-loop.
320     RefCost = TripCount;
321 
322     int Index = getSubscriptIndex(L);
323     assert(Index >= 0 && "Could not locate a valid Index");
324 
325     for (unsigned I = Index + 1; I < getNumSubscripts() - 1; ++I) {
326       const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(getSubscript(I));
327       assert(AR && AR->getLoop() && "Expecting valid loop");
328       const SCEV *TripCount =
329           computeTripCount(*AR->getLoop(), *Sizes.back(), SE);
330       Type *WiderType = SE.getWiderType(RefCost->getType(), TripCount->getType());
331       RefCost = SE.getMulExpr(SE.getNoopOrZeroExtend(RefCost, WiderType),
332                               SE.getNoopOrZeroExtend(TripCount, WiderType));
333     }
334 
335     LLVM_DEBUG(dbgs().indent(4)
336                << "Access is not consecutive: RefCost=" << *RefCost << "\n");
337   }
338   assert(RefCost && "Expecting a valid RefCost");
339 
340   // Attempt to fold RefCost into a constant.
341   if (auto ConstantCost = dyn_cast<SCEVConstant>(RefCost))
342     return ConstantCost->getValue()->getZExtValue();
343 
344   LLVM_DEBUG(dbgs().indent(4)
345              << "RefCost is not a constant! Setting to RefCost=InvalidCost "
346                 "(invalid value).\n");
347 
348   return CacheCost::InvalidCost;
349 }
350 
351 bool IndexedReference::tryDelinearizeFixedSize(
352     const SCEV *AccessFn, SmallVectorImpl<const SCEV *> &Subscripts) {
353   SmallVector<int, 4> ArraySizes;
354   if (!tryDelinearizeFixedSizeImpl(&SE, &StoreOrLoadInst, AccessFn, Subscripts,
355                                    ArraySizes))
356     return false;
357 
358   // Populate Sizes with scev expressions to be used in calculations later.
359   for (auto Idx : seq<unsigned>(1, Subscripts.size()))
360     Sizes.push_back(
361         SE.getConstant(Subscripts[Idx]->getType(), ArraySizes[Idx - 1]));
362 
363   LLVM_DEBUG({
364     dbgs() << "Delinearized subscripts of fixed-size array\n"
365            << "GEP:" << *getLoadStorePointerOperand(&StoreOrLoadInst)
366            << "\n";
367   });
368   return true;
369 }
370 
371 bool IndexedReference::delinearize(const LoopInfo &LI) {
372   assert(Subscripts.empty() && "Subscripts should be empty");
373   assert(Sizes.empty() && "Sizes should be empty");
374   assert(!IsValid && "Should be called once from the constructor");
375   LLVM_DEBUG(dbgs() << "Delinearizing: " << StoreOrLoadInst << "\n");
376 
377   const SCEV *ElemSize = SE.getElementSize(&StoreOrLoadInst);
378   const BasicBlock *BB = StoreOrLoadInst.getParent();
379 
380   if (Loop *L = LI.getLoopFor(BB)) {
381     const SCEV *AccessFn =
382         SE.getSCEVAtScope(getPointerOperand(&StoreOrLoadInst), L);
383 
384     BasePointer = dyn_cast<SCEVUnknown>(SE.getPointerBase(AccessFn));
385     if (BasePointer == nullptr) {
386       LLVM_DEBUG(
387           dbgs().indent(2)
388           << "ERROR: failed to delinearize, can't identify base pointer\n");
389       return false;
390     }
391 
392     bool IsFixedSize = false;
393     // Try to delinearize fixed-size arrays.
394     if (tryDelinearizeFixedSize(AccessFn, Subscripts)) {
395       IsFixedSize = true;
396       // The last element of Sizes is the element size.
397       Sizes.push_back(ElemSize);
398       LLVM_DEBUG(dbgs().indent(2) << "In Loop '" << L->getName()
399                                   << "', AccessFn: " << *AccessFn << "\n");
400     }
401 
402     AccessFn = SE.getMinusSCEV(AccessFn, BasePointer);
403 
404     // Try to delinearize parametric-size arrays.
405     if (!IsFixedSize) {
406       LLVM_DEBUG(dbgs().indent(2) << "In Loop '" << L->getName()
407                                   << "', AccessFn: " << *AccessFn << "\n");
408       llvm::delinearize(SE, AccessFn, Subscripts, Sizes,
409                         SE.getElementSize(&StoreOrLoadInst));
410     }
411 
412     if (Subscripts.empty() || Sizes.empty() ||
413         Subscripts.size() != Sizes.size()) {
414       // Attempt to determine whether we have a single dimensional array access.
415       // before giving up.
416       if (!isOneDimensionalArray(*AccessFn, *ElemSize, *L, SE)) {
417         LLVM_DEBUG(dbgs().indent(2)
418                    << "ERROR: failed to delinearize reference\n");
419         Subscripts.clear();
420         Sizes.clear();
421         return false;
422       }
423 
424       // The array may be accessed in reverse, for example:
425       //   for (i = N; i > 0; i--)
426       //     A[i] = 0;
427       // In this case, reconstruct the access function using the absolute value
428       // of the step recurrence.
429       const SCEVAddRecExpr *AccessFnAR = dyn_cast<SCEVAddRecExpr>(AccessFn);
430       const SCEV *StepRec = AccessFnAR ? AccessFnAR->getStepRecurrence(SE) : nullptr;
431 
432       if (StepRec && SE.isKnownNegative(StepRec))
433         AccessFn = SE.getAddRecExpr(AccessFnAR->getStart(),
434                                     SE.getNegativeSCEV(StepRec),
435                                     AccessFnAR->getLoop(),
436                                     AccessFnAR->getNoWrapFlags());
437       const SCEV *Div = SE.getUDivExactExpr(AccessFn, ElemSize);
438       Subscripts.push_back(Div);
439       Sizes.push_back(ElemSize);
440     }
441 
442     return all_of(Subscripts, [&](const SCEV *Subscript) {
443       return isSimpleAddRecurrence(*Subscript, *L);
444     });
445   }
446 
447   return false;
448 }
449 
450 bool IndexedReference::isLoopInvariant(const Loop &L) const {
451   Value *Addr = getPointerOperand(&StoreOrLoadInst);
452   assert(Addr != nullptr && "Expecting either a load or a store instruction");
453   assert(SE.isSCEVable(Addr->getType()) && "Addr should be SCEVable");
454 
455   if (SE.isLoopInvariant(SE.getSCEV(Addr), &L))
456     return true;
457 
458   // The indexed reference is loop invariant if none of the coefficients use
459   // the loop induction variable.
460   bool allCoeffForLoopAreZero = all_of(Subscripts, [&](const SCEV *Subscript) {
461     return isCoeffForLoopZeroOrInvariant(*Subscript, L);
462   });
463 
464   return allCoeffForLoopAreZero;
465 }
466 
467 bool IndexedReference::isConsecutive(const Loop &L, const SCEV *&Stride,
468                                      unsigned CLS) const {
469   // The indexed reference is 'consecutive' if the only coefficient that uses
470   // the loop induction variable is the last one...
471   const SCEV *LastSubscript = Subscripts.back();
472   for (const SCEV *Subscript : Subscripts) {
473     if (Subscript == LastSubscript)
474       continue;
475     if (!isCoeffForLoopZeroOrInvariant(*Subscript, L))
476       return false;
477   }
478 
479   // ...and the access stride is less than the cache line size.
480   const SCEV *Coeff = getLastCoefficient();
481   const SCEV *ElemSize = Sizes.back();
482   Type *WiderType = SE.getWiderType(Coeff->getType(), ElemSize->getType());
483   // FIXME: This assumes that all values are signed integers which may
484   // be incorrect in unusual codes and incorrectly use sext instead of zext.
485   // for (uint32_t i = 0; i < 512; ++i) {
486   //   uint8_t trunc = i;
487   //   A[trunc] = 42;
488   // }
489   // This consecutively iterates twice over A. If `trunc` is sign-extended,
490   // we would conclude that this may iterate backwards over the array.
491   // However, LoopCacheAnalysis is heuristic anyway and transformations must
492   // not result in wrong optimizations if the heuristic was incorrect.
493   Stride = SE.getMulExpr(SE.getNoopOrSignExtend(Coeff, WiderType),
494                          SE.getNoopOrSignExtend(ElemSize, WiderType));
495   const SCEV *CacheLineSize = SE.getConstant(Stride->getType(), CLS);
496 
497   Stride = SE.isKnownNegative(Stride) ? SE.getNegativeSCEV(Stride) : Stride;
498   return SE.isKnownPredicate(ICmpInst::ICMP_ULT, Stride, CacheLineSize);
499 }
500 
501 int IndexedReference::getSubscriptIndex(const Loop &L) const {
502   for (auto Idx : seq<int>(0, getNumSubscripts())) {
503     const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(getSubscript(Idx));
504     if (AR && AR->getLoop() == &L) {
505       return Idx;
506     }
507   }
508   return -1;
509 }
510 
511 const SCEV *IndexedReference::getLastCoefficient() const {
512   const SCEV *LastSubscript = getLastSubscript();
513   auto *AR = cast<SCEVAddRecExpr>(LastSubscript);
514   return AR->getStepRecurrence(SE);
515 }
516 
517 bool IndexedReference::isCoeffForLoopZeroOrInvariant(const SCEV &Subscript,
518                                                      const Loop &L) const {
519   const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(&Subscript);
520   return (AR != nullptr) ? AR->getLoop() != &L
521                          : SE.isLoopInvariant(&Subscript, &L);
522 }
523 
524 bool IndexedReference::isSimpleAddRecurrence(const SCEV &Subscript,
525                                              const Loop &L) const {
526   if (!isa<SCEVAddRecExpr>(Subscript))
527     return false;
528 
529   const SCEVAddRecExpr *AR = cast<SCEVAddRecExpr>(&Subscript);
530   assert(AR->getLoop() && "AR should have a loop");
531 
532   if (!AR->isAffine())
533     return false;
534 
535   const SCEV *Start = AR->getStart();
536   const SCEV *Step = AR->getStepRecurrence(SE);
537 
538   if (!SE.isLoopInvariant(Start, &L) || !SE.isLoopInvariant(Step, &L))
539     return false;
540 
541   return true;
542 }
543 
544 bool IndexedReference::isAliased(const IndexedReference &Other,
545                                  AAResults &AA) const {
546   const auto &Loc1 = MemoryLocation::get(&StoreOrLoadInst);
547   const auto &Loc2 = MemoryLocation::get(&Other.StoreOrLoadInst);
548   return AA.isMustAlias(Loc1, Loc2);
549 }
550 
551 //===----------------------------------------------------------------------===//
552 // CacheCost implementation
553 //
554 raw_ostream &llvm::operator<<(raw_ostream &OS, const CacheCost &CC) {
555   for (const auto &LC : CC.LoopCosts) {
556     const Loop *L = LC.first;
557     OS << "Loop '" << L->getName() << "' has cost = " << LC.second << "\n";
558   }
559   return OS;
560 }
561 
562 CacheCost::CacheCost(const LoopVectorTy &Loops, const LoopInfo &LI,
563                      ScalarEvolution &SE, TargetTransformInfo &TTI,
564                      AAResults &AA, DependenceInfo &DI,
565                      std::optional<unsigned> TRT)
566     : Loops(Loops), TRT(TRT.value_or(TemporalReuseThreshold)), LI(LI), SE(SE),
567       TTI(TTI), AA(AA), DI(DI) {
568   assert(!Loops.empty() && "Expecting a non-empty loop vector.");
569 
570   for (const Loop *L : Loops) {
571     unsigned TripCount = SE.getSmallConstantTripCount(L);
572     TripCount = (TripCount == 0) ? DefaultTripCount : TripCount;
573     TripCounts.push_back({L, TripCount});
574   }
575 
576   calculateCacheFootprint();
577 }
578 
579 std::unique_ptr<CacheCost>
580 CacheCost::getCacheCost(Loop &Root, LoopStandardAnalysisResults &AR,
581                         DependenceInfo &DI, std::optional<unsigned> TRT) {
582   if (!Root.isOutermost()) {
583     LLVM_DEBUG(dbgs() << "Expecting the outermost loop in a loop nest\n");
584     return nullptr;
585   }
586 
587   LoopVectorTy Loops;
588   append_range(Loops, breadth_first(&Root));
589 
590   if (!getInnerMostLoop(Loops)) {
591     LLVM_DEBUG(dbgs() << "Cannot compute cache cost of loop nest with more "
592                          "than one innermost loop\n");
593     return nullptr;
594   }
595 
596   return std::make_unique<CacheCost>(Loops, AR.LI, AR.SE, AR.TTI, AR.AA, DI, TRT);
597 }
598 
599 void CacheCost::calculateCacheFootprint() {
600   LLVM_DEBUG(dbgs() << "POPULATING REFERENCE GROUPS\n");
601   ReferenceGroupsTy RefGroups;
602   if (!populateReferenceGroups(RefGroups))
603     return;
604 
605   LLVM_DEBUG(dbgs() << "COMPUTING LOOP CACHE COSTS\n");
606   for (const Loop *L : Loops) {
607     assert(llvm::none_of(
608                LoopCosts,
609                [L](const LoopCacheCostTy &LCC) { return LCC.first == L; }) &&
610            "Should not add duplicate element");
611     CacheCostTy LoopCost = computeLoopCacheCost(*L, RefGroups);
612     LoopCosts.push_back(std::make_pair(L, LoopCost));
613   }
614 
615   sortLoopCosts();
616   RefGroups.clear();
617 }
618 
619 bool CacheCost::populateReferenceGroups(ReferenceGroupsTy &RefGroups) const {
620   assert(RefGroups.empty() && "Reference groups should be empty");
621 
622   unsigned CLS = TTI.getCacheLineSize();
623   Loop *InnerMostLoop = getInnerMostLoop(Loops);
624   assert(InnerMostLoop != nullptr && "Expecting a valid innermost loop");
625 
626   for (BasicBlock *BB : InnerMostLoop->getBlocks()) {
627     for (Instruction &I : *BB) {
628       if (!isa<StoreInst>(I) && !isa<LoadInst>(I))
629         continue;
630 
631       std::unique_ptr<IndexedReference> R(new IndexedReference(I, LI, SE));
632       if (!R->isValid())
633         continue;
634 
635       bool Added = false;
636       for (ReferenceGroupTy &RefGroup : RefGroups) {
637         const IndexedReference &Representative = *RefGroup.front();
638         LLVM_DEBUG({
639           dbgs() << "References:\n";
640           dbgs().indent(2) << *R << "\n";
641           dbgs().indent(2) << Representative << "\n";
642         });
643 
644 
645        // FIXME: Both positive and negative access functions will be placed
646        // into the same reference group, resulting in a bi-directional array
647        // access such as:
648        //   for (i = N; i > 0; i--)
649        //     A[i] = A[N - i];
650        // having the same cost calculation as a single dimention access pattern
651        //   for (i = 0; i < N; i++)
652        //     A[i] = A[i];
653        // when in actuality, depending on the array size, the first example
654        // should have a cost closer to 2x the second due to the two cache
655        // access per iteration from opposite ends of the array
656         std::optional<bool> HasTemporalReuse =
657             R->hasTemporalReuse(Representative, *TRT, *InnerMostLoop, DI, AA);
658         std::optional<bool> HasSpacialReuse =
659             R->hasSpacialReuse(Representative, CLS, AA);
660 
661         if ((HasTemporalReuse && *HasTemporalReuse) ||
662             (HasSpacialReuse && *HasSpacialReuse)) {
663           RefGroup.push_back(std::move(R));
664           Added = true;
665           break;
666         }
667       }
668 
669       if (!Added) {
670         ReferenceGroupTy RG;
671         RG.push_back(std::move(R));
672         RefGroups.push_back(std::move(RG));
673       }
674     }
675   }
676 
677   if (RefGroups.empty())
678     return false;
679 
680   LLVM_DEBUG({
681     dbgs() << "\nIDENTIFIED REFERENCE GROUPS:\n";
682     int n = 1;
683     for (const ReferenceGroupTy &RG : RefGroups) {
684       dbgs().indent(2) << "RefGroup " << n << ":\n";
685       for (const auto &IR : RG)
686         dbgs().indent(4) << *IR << "\n";
687       n++;
688     }
689     dbgs() << "\n";
690   });
691 
692   return true;
693 }
694 
695 CacheCostTy
696 CacheCost::computeLoopCacheCost(const Loop &L,
697                                 const ReferenceGroupsTy &RefGroups) const {
698   if (!L.isLoopSimplifyForm())
699     return InvalidCost;
700 
701   LLVM_DEBUG(dbgs() << "Considering loop '" << L.getName()
702                     << "' as innermost loop.\n");
703 
704   // Compute the product of the trip counts of each other loop in the nest.
705   CacheCostTy TripCountsProduct = 1;
706   for (const auto &TC : TripCounts) {
707     if (TC.first == &L)
708       continue;
709     TripCountsProduct *= TC.second;
710   }
711 
712   CacheCostTy LoopCost = 0;
713   for (const ReferenceGroupTy &RG : RefGroups) {
714     CacheCostTy RefGroupCost = computeRefGroupCacheCost(RG, L);
715     LoopCost += RefGroupCost * TripCountsProduct;
716   }
717 
718   LLVM_DEBUG(dbgs().indent(2) << "Loop '" << L.getName()
719                               << "' has cost=" << LoopCost << "\n");
720 
721   return LoopCost;
722 }
723 
724 CacheCostTy CacheCost::computeRefGroupCacheCost(const ReferenceGroupTy &RG,
725                                                 const Loop &L) const {
726   assert(!RG.empty() && "Reference group should have at least one member.");
727 
728   const IndexedReference *Representative = RG.front().get();
729   return Representative->computeRefCost(L, TTI.getCacheLineSize());
730 }
731 
732 //===----------------------------------------------------------------------===//
733 // LoopCachePrinterPass implementation
734 //
735 PreservedAnalyses LoopCachePrinterPass::run(Loop &L, LoopAnalysisManager &AM,
736                                             LoopStandardAnalysisResults &AR,
737                                             LPMUpdater &U) {
738   Function *F = L.getHeader()->getParent();
739   DependenceInfo DI(F, &AR.AA, &AR.SE, &AR.LI);
740 
741   if (auto CC = CacheCost::getCacheCost(L, AR, DI))
742     OS << *CC;
743 
744   return PreservedAnalyses::all();
745 }
746