1 //===----------- VectorUtils.cpp - Vectorizer utility functions -----------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file defines vectorizer utilities.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "llvm/Analysis/VectorUtils.h"
14 #include "llvm/ADT/EquivalenceClasses.h"
15 #include "llvm/Analysis/DemandedBits.h"
16 #include "llvm/Analysis/LoopInfo.h"
17 #include "llvm/Analysis/LoopIterator.h"
18 #include "llvm/Analysis/ScalarEvolution.h"
19 #include "llvm/Analysis/ScalarEvolutionExpressions.h"
20 #include "llvm/Analysis/TargetTransformInfo.h"
21 #include "llvm/Analysis/ValueTracking.h"
22 #include "llvm/IR/Constants.h"
23 #include "llvm/IR/GetElementPtrTypeIterator.h"
24 #include "llvm/IR/IRBuilder.h"
25 #include "llvm/IR/PatternMatch.h"
26 #include "llvm/IR/Value.h"
27 #include "llvm/Support/CommandLine.h"
28 
29 #define DEBUG_TYPE "vectorutils"
30 
31 using namespace llvm;
32 using namespace llvm::PatternMatch;
33 
34 /// Maximum factor for an interleaved memory access.
35 static cl::opt<unsigned> MaxInterleaveGroupFactor(
36     "max-interleave-group-factor", cl::Hidden,
37     cl::desc("Maximum factor for an interleaved access group (default = 8)"),
38     cl::init(8));
39 
40 /// Return true if all of the intrinsic's arguments and return type are scalars
41 /// for the scalar form of the intrinsic, and vectors for the vector form of the
42 /// intrinsic (except operands that are marked as always being scalar by
43 /// hasVectorInstrinsicScalarOpd).
isTriviallyVectorizable(Intrinsic::ID ID)44 bool llvm::isTriviallyVectorizable(Intrinsic::ID ID) {
45   switch (ID) {
46   case Intrinsic::abs:   // Begin integer bit-manipulation.
47   case Intrinsic::bswap:
48   case Intrinsic::bitreverse:
49   case Intrinsic::ctpop:
50   case Intrinsic::ctlz:
51   case Intrinsic::cttz:
52   case Intrinsic::fshl:
53   case Intrinsic::fshr:
54   case Intrinsic::smax:
55   case Intrinsic::smin:
56   case Intrinsic::umax:
57   case Intrinsic::umin:
58   case Intrinsic::sadd_sat:
59   case Intrinsic::ssub_sat:
60   case Intrinsic::uadd_sat:
61   case Intrinsic::usub_sat:
62   case Intrinsic::smul_fix:
63   case Intrinsic::smul_fix_sat:
64   case Intrinsic::umul_fix:
65   case Intrinsic::umul_fix_sat:
66   case Intrinsic::sqrt: // Begin floating-point.
67   case Intrinsic::sin:
68   case Intrinsic::cos:
69   case Intrinsic::exp:
70   case Intrinsic::exp2:
71   case Intrinsic::log:
72   case Intrinsic::log10:
73   case Intrinsic::log2:
74   case Intrinsic::fabs:
75   case Intrinsic::minnum:
76   case Intrinsic::maxnum:
77   case Intrinsic::minimum:
78   case Intrinsic::maximum:
79   case Intrinsic::copysign:
80   case Intrinsic::floor:
81   case Intrinsic::ceil:
82   case Intrinsic::trunc:
83   case Intrinsic::rint:
84   case Intrinsic::nearbyint:
85   case Intrinsic::round:
86   case Intrinsic::roundeven:
87   case Intrinsic::pow:
88   case Intrinsic::fma:
89   case Intrinsic::fmuladd:
90   case Intrinsic::powi:
91   case Intrinsic::canonicalize:
92     return true;
93   default:
94     return false;
95   }
96 }
97 
98 /// Identifies if the vector form of the intrinsic has a scalar operand.
hasVectorInstrinsicScalarOpd(Intrinsic::ID ID,unsigned ScalarOpdIdx)99 bool llvm::hasVectorInstrinsicScalarOpd(Intrinsic::ID ID,
100                                         unsigned ScalarOpdIdx) {
101   switch (ID) {
102   case Intrinsic::abs:
103   case Intrinsic::ctlz:
104   case Intrinsic::cttz:
105   case Intrinsic::powi:
106     return (ScalarOpdIdx == 1);
107   case Intrinsic::smul_fix:
108   case Intrinsic::smul_fix_sat:
109   case Intrinsic::umul_fix:
110   case Intrinsic::umul_fix_sat:
111     return (ScalarOpdIdx == 2);
112   default:
113     return false;
114   }
115 }
116 
hasVectorInstrinsicOverloadedScalarOpd(Intrinsic::ID ID,unsigned ScalarOpdIdx)117 bool llvm::hasVectorInstrinsicOverloadedScalarOpd(Intrinsic::ID ID,
118                                                   unsigned ScalarOpdIdx) {
119   switch (ID) {
120   case Intrinsic::powi:
121     return (ScalarOpdIdx == 1);
122   default:
123     return false;
124   }
125 }
126 
127 /// Returns intrinsic ID for call.
128 /// For the input call instruction it finds mapping intrinsic and returns
129 /// its ID, in case it does not found it return not_intrinsic.
getVectorIntrinsicIDForCall(const CallInst * CI,const TargetLibraryInfo * TLI)130 Intrinsic::ID llvm::getVectorIntrinsicIDForCall(const CallInst *CI,
131                                                 const TargetLibraryInfo *TLI) {
132   Intrinsic::ID ID = getIntrinsicForCallSite(*CI, TLI);
133   if (ID == Intrinsic::not_intrinsic)
134     return Intrinsic::not_intrinsic;
135 
136   if (isTriviallyVectorizable(ID) || ID == Intrinsic::lifetime_start ||
137       ID == Intrinsic::lifetime_end || ID == Intrinsic::assume ||
138       ID == Intrinsic::experimental_noalias_scope_decl ||
139       ID == Intrinsic::sideeffect || ID == Intrinsic::pseudoprobe)
140     return ID;
141   return Intrinsic::not_intrinsic;
142 }
143 
144 /// Find the operand of the GEP that should be checked for consecutive
145 /// stores. This ignores trailing indices that have no effect on the final
146 /// pointer.
getGEPInductionOperand(const GetElementPtrInst * Gep)147 unsigned llvm::getGEPInductionOperand(const GetElementPtrInst *Gep) {
148   const DataLayout &DL = Gep->getModule()->getDataLayout();
149   unsigned LastOperand = Gep->getNumOperands() - 1;
150   TypeSize GEPAllocSize = DL.getTypeAllocSize(Gep->getResultElementType());
151 
152   // Walk backwards and try to peel off zeros.
153   while (LastOperand > 1 && match(Gep->getOperand(LastOperand), m_Zero())) {
154     // Find the type we're currently indexing into.
155     gep_type_iterator GEPTI = gep_type_begin(Gep);
156     std::advance(GEPTI, LastOperand - 2);
157 
158     // If it's a type with the same allocation size as the result of the GEP we
159     // can peel off the zero index.
160     if (DL.getTypeAllocSize(GEPTI.getIndexedType()) != GEPAllocSize)
161       break;
162     --LastOperand;
163   }
164 
165   return LastOperand;
166 }
167 
168 /// If the argument is a GEP, then returns the operand identified by
169 /// getGEPInductionOperand. However, if there is some other non-loop-invariant
170 /// operand, it returns that instead.
stripGetElementPtr(Value * Ptr,ScalarEvolution * SE,Loop * Lp)171 Value *llvm::stripGetElementPtr(Value *Ptr, ScalarEvolution *SE, Loop *Lp) {
172   GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr);
173   if (!GEP)
174     return Ptr;
175 
176   unsigned InductionOperand = getGEPInductionOperand(GEP);
177 
178   // Check that all of the gep indices are uniform except for our induction
179   // operand.
180   for (unsigned i = 0, e = GEP->getNumOperands(); i != e; ++i)
181     if (i != InductionOperand &&
182         !SE->isLoopInvariant(SE->getSCEV(GEP->getOperand(i)), Lp))
183       return Ptr;
184   return GEP->getOperand(InductionOperand);
185 }
186 
187 /// If a value has only one user that is a CastInst, return it.
getUniqueCastUse(Value * Ptr,Loop * Lp,Type * Ty)188 Value *llvm::getUniqueCastUse(Value *Ptr, Loop *Lp, Type *Ty) {
189   Value *UniqueCast = nullptr;
190   for (User *U : Ptr->users()) {
191     CastInst *CI = dyn_cast<CastInst>(U);
192     if (CI && CI->getType() == Ty) {
193       if (!UniqueCast)
194         UniqueCast = CI;
195       else
196         return nullptr;
197     }
198   }
199   return UniqueCast;
200 }
201 
202 /// Get the stride of a pointer access in a loop. Looks for symbolic
203 /// strides "a[i*stride]". Returns the symbolic stride, or null otherwise.
getStrideFromPointer(Value * Ptr,ScalarEvolution * SE,Loop * Lp)204 Value *llvm::getStrideFromPointer(Value *Ptr, ScalarEvolution *SE, Loop *Lp) {
205   auto *PtrTy = dyn_cast<PointerType>(Ptr->getType());
206   if (!PtrTy || PtrTy->isAggregateType())
207     return nullptr;
208 
209   // Try to remove a gep instruction to make the pointer (actually index at this
210   // point) easier analyzable. If OrigPtr is equal to Ptr we are analyzing the
211   // pointer, otherwise, we are analyzing the index.
212   Value *OrigPtr = Ptr;
213 
214   // The size of the pointer access.
215   int64_t PtrAccessSize = 1;
216 
217   Ptr = stripGetElementPtr(Ptr, SE, Lp);
218   const SCEV *V = SE->getSCEV(Ptr);
219 
220   if (Ptr != OrigPtr)
221     // Strip off casts.
222     while (const SCEVIntegralCastExpr *C = dyn_cast<SCEVIntegralCastExpr>(V))
223       V = C->getOperand();
224 
225   const SCEVAddRecExpr *S = dyn_cast<SCEVAddRecExpr>(V);
226   if (!S)
227     return nullptr;
228 
229   V = S->getStepRecurrence(*SE);
230   if (!V)
231     return nullptr;
232 
233   // Strip off the size of access multiplication if we are still analyzing the
234   // pointer.
235   if (OrigPtr == Ptr) {
236     if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(V)) {
237       if (M->getOperand(0)->getSCEVType() != scConstant)
238         return nullptr;
239 
240       const APInt &APStepVal = cast<SCEVConstant>(M->getOperand(0))->getAPInt();
241 
242       // Huge step value - give up.
243       if (APStepVal.getBitWidth() > 64)
244         return nullptr;
245 
246       int64_t StepVal = APStepVal.getSExtValue();
247       if (PtrAccessSize != StepVal)
248         return nullptr;
249       V = M->getOperand(1);
250     }
251   }
252 
253   // Strip off casts.
254   Type *StripedOffRecurrenceCast = nullptr;
255   if (const SCEVIntegralCastExpr *C = dyn_cast<SCEVIntegralCastExpr>(V)) {
256     StripedOffRecurrenceCast = C->getType();
257     V = C->getOperand();
258   }
259 
260   // Look for the loop invariant symbolic value.
261   const SCEVUnknown *U = dyn_cast<SCEVUnknown>(V);
262   if (!U)
263     return nullptr;
264 
265   Value *Stride = U->getValue();
266   if (!Lp->isLoopInvariant(Stride))
267     return nullptr;
268 
269   // If we have stripped off the recurrence cast we have to make sure that we
270   // return the value that is used in this loop so that we can replace it later.
271   if (StripedOffRecurrenceCast)
272     Stride = getUniqueCastUse(Stride, Lp, StripedOffRecurrenceCast);
273 
274   return Stride;
275 }
276 
277 /// Given a vector and an element number, see if the scalar value is
278 /// already around as a register, for example if it were inserted then extracted
279 /// from the vector.
findScalarElement(Value * V,unsigned EltNo)280 Value *llvm::findScalarElement(Value *V, unsigned EltNo) {
281   assert(V->getType()->isVectorTy() && "Not looking at a vector?");
282   VectorType *VTy = cast<VectorType>(V->getType());
283   // For fixed-length vector, return undef for out of range access.
284   if (auto *FVTy = dyn_cast<FixedVectorType>(VTy)) {
285     unsigned Width = FVTy->getNumElements();
286     if (EltNo >= Width)
287       return UndefValue::get(FVTy->getElementType());
288   }
289 
290   if (Constant *C = dyn_cast<Constant>(V))
291     return C->getAggregateElement(EltNo);
292 
293   if (InsertElementInst *III = dyn_cast<InsertElementInst>(V)) {
294     // If this is an insert to a variable element, we don't know what it is.
295     if (!isa<ConstantInt>(III->getOperand(2)))
296       return nullptr;
297     unsigned IIElt = cast<ConstantInt>(III->getOperand(2))->getZExtValue();
298 
299     // If this is an insert to the element we are looking for, return the
300     // inserted value.
301     if (EltNo == IIElt)
302       return III->getOperand(1);
303 
304     // Guard against infinite loop on malformed, unreachable IR.
305     if (III == III->getOperand(0))
306       return nullptr;
307 
308     // Otherwise, the insertelement doesn't modify the value, recurse on its
309     // vector input.
310     return findScalarElement(III->getOperand(0), EltNo);
311   }
312 
313   ShuffleVectorInst *SVI = dyn_cast<ShuffleVectorInst>(V);
314   // Restrict the following transformation to fixed-length vector.
315   if (SVI && isa<FixedVectorType>(SVI->getType())) {
316     unsigned LHSWidth =
317         cast<FixedVectorType>(SVI->getOperand(0)->getType())->getNumElements();
318     int InEl = SVI->getMaskValue(EltNo);
319     if (InEl < 0)
320       return UndefValue::get(VTy->getElementType());
321     if (InEl < (int)LHSWidth)
322       return findScalarElement(SVI->getOperand(0), InEl);
323     return findScalarElement(SVI->getOperand(1), InEl - LHSWidth);
324   }
325 
326   // Extract a value from a vector add operation with a constant zero.
327   // TODO: Use getBinOpIdentity() to generalize this.
328   Value *Val; Constant *C;
329   if (match(V, m_Add(m_Value(Val), m_Constant(C))))
330     if (Constant *Elt = C->getAggregateElement(EltNo))
331       if (Elt->isNullValue())
332         return findScalarElement(Val, EltNo);
333 
334   // Otherwise, we don't know.
335   return nullptr;
336 }
337 
getSplatIndex(ArrayRef<int> Mask)338 int llvm::getSplatIndex(ArrayRef<int> Mask) {
339   int SplatIndex = -1;
340   for (int M : Mask) {
341     // Ignore invalid (undefined) mask elements.
342     if (M < 0)
343       continue;
344 
345     // There can be only 1 non-negative mask element value if this is a splat.
346     if (SplatIndex != -1 && SplatIndex != M)
347       return -1;
348 
349     // Initialize the splat index to the 1st non-negative mask element.
350     SplatIndex = M;
351   }
352   assert((SplatIndex == -1 || SplatIndex >= 0) && "Negative index?");
353   return SplatIndex;
354 }
355 
356 /// Get splat value if the input is a splat vector or return nullptr.
357 /// This function is not fully general. It checks only 2 cases:
358 /// the input value is (1) a splat constant vector or (2) a sequence
359 /// of instructions that broadcasts a scalar at element 0.
getSplatValue(const Value * V)360 Value *llvm::getSplatValue(const Value *V) {
361   if (isa<VectorType>(V->getType()))
362     if (auto *C = dyn_cast<Constant>(V))
363       return C->getSplatValue();
364 
365   // shuf (inselt ?, Splat, 0), ?, <0, undef, 0, ...>
366   Value *Splat;
367   if (match(V,
368             m_Shuffle(m_InsertElt(m_Value(), m_Value(Splat), m_ZeroInt()),
369                       m_Value(), m_ZeroMask())))
370     return Splat;
371 
372   return nullptr;
373 }
374 
isSplatValue(const Value * V,int Index,unsigned Depth)375 bool llvm::isSplatValue(const Value *V, int Index, unsigned Depth) {
376   assert(Depth <= MaxAnalysisRecursionDepth && "Limit Search Depth");
377 
378   if (isa<VectorType>(V->getType())) {
379     if (isa<UndefValue>(V))
380       return true;
381     // FIXME: We can allow undefs, but if Index was specified, we may want to
382     //        check that the constant is defined at that index.
383     if (auto *C = dyn_cast<Constant>(V))
384       return C->getSplatValue() != nullptr;
385   }
386 
387   if (auto *Shuf = dyn_cast<ShuffleVectorInst>(V)) {
388     // FIXME: We can safely allow undefs here. If Index was specified, we will
389     //        check that the mask elt is defined at the required index.
390     if (!is_splat(Shuf->getShuffleMask()))
391       return false;
392 
393     // Match any index.
394     if (Index == -1)
395       return true;
396 
397     // Match a specific element. The mask should be defined at and match the
398     // specified index.
399     return Shuf->getMaskValue(Index) == Index;
400   }
401 
402   // The remaining tests are all recursive, so bail out if we hit the limit.
403   if (Depth++ == MaxAnalysisRecursionDepth)
404     return false;
405 
406   // If both operands of a binop are splats, the result is a splat.
407   Value *X, *Y, *Z;
408   if (match(V, m_BinOp(m_Value(X), m_Value(Y))))
409     return isSplatValue(X, Index, Depth) && isSplatValue(Y, Index, Depth);
410 
411   // If all operands of a select are splats, the result is a splat.
412   if (match(V, m_Select(m_Value(X), m_Value(Y), m_Value(Z))))
413     return isSplatValue(X, Index, Depth) && isSplatValue(Y, Index, Depth) &&
414            isSplatValue(Z, Index, Depth);
415 
416   // TODO: Add support for unary ops (fneg), casts, intrinsics (overflow ops).
417 
418   return false;
419 }
420 
narrowShuffleMaskElts(int Scale,ArrayRef<int> Mask,SmallVectorImpl<int> & ScaledMask)421 void llvm::narrowShuffleMaskElts(int Scale, ArrayRef<int> Mask,
422                                  SmallVectorImpl<int> &ScaledMask) {
423   assert(Scale > 0 && "Unexpected scaling factor");
424 
425   // Fast-path: if no scaling, then it is just a copy.
426   if (Scale == 1) {
427     ScaledMask.assign(Mask.begin(), Mask.end());
428     return;
429   }
430 
431   ScaledMask.clear();
432   for (int MaskElt : Mask) {
433     if (MaskElt >= 0) {
434       assert(((uint64_t)Scale * MaskElt + (Scale - 1)) <= INT32_MAX &&
435              "Overflowed 32-bits");
436     }
437     for (int SliceElt = 0; SliceElt != Scale; ++SliceElt)
438       ScaledMask.push_back(MaskElt < 0 ? MaskElt : Scale * MaskElt + SliceElt);
439   }
440 }
441 
widenShuffleMaskElts(int Scale,ArrayRef<int> Mask,SmallVectorImpl<int> & ScaledMask)442 bool llvm::widenShuffleMaskElts(int Scale, ArrayRef<int> Mask,
443                                 SmallVectorImpl<int> &ScaledMask) {
444   assert(Scale > 0 && "Unexpected scaling factor");
445 
446   // Fast-path: if no scaling, then it is just a copy.
447   if (Scale == 1) {
448     ScaledMask.assign(Mask.begin(), Mask.end());
449     return true;
450   }
451 
452   // We must map the original elements down evenly to a type with less elements.
453   int NumElts = Mask.size();
454   if (NumElts % Scale != 0)
455     return false;
456 
457   ScaledMask.clear();
458   ScaledMask.reserve(NumElts / Scale);
459 
460   // Step through the input mask by splitting into Scale-sized slices.
461   do {
462     ArrayRef<int> MaskSlice = Mask.take_front(Scale);
463     assert((int)MaskSlice.size() == Scale && "Expected Scale-sized slice.");
464 
465     // The first element of the slice determines how we evaluate this slice.
466     int SliceFront = MaskSlice.front();
467     if (SliceFront < 0) {
468       // Negative values (undef or other "sentinel" values) must be equal across
469       // the entire slice.
470       if (!is_splat(MaskSlice))
471         return false;
472       ScaledMask.push_back(SliceFront);
473     } else {
474       // A positive mask element must be cleanly divisible.
475       if (SliceFront % Scale != 0)
476         return false;
477       // Elements of the slice must be consecutive.
478       for (int i = 1; i < Scale; ++i)
479         if (MaskSlice[i] != SliceFront + i)
480           return false;
481       ScaledMask.push_back(SliceFront / Scale);
482     }
483     Mask = Mask.drop_front(Scale);
484   } while (!Mask.empty());
485 
486   assert((int)ScaledMask.size() * Scale == NumElts && "Unexpected scaled mask");
487 
488   // All elements of the original mask can be scaled down to map to the elements
489   // of a mask with wider elements.
490   return true;
491 }
492 
493 MapVector<Instruction *, uint64_t>
computeMinimumValueSizes(ArrayRef<BasicBlock * > Blocks,DemandedBits & DB,const TargetTransformInfo * TTI)494 llvm::computeMinimumValueSizes(ArrayRef<BasicBlock *> Blocks, DemandedBits &DB,
495                                const TargetTransformInfo *TTI) {
496 
497   // DemandedBits will give us every value's live-out bits. But we want
498   // to ensure no extra casts would need to be inserted, so every DAG
499   // of connected values must have the same minimum bitwidth.
500   EquivalenceClasses<Value *> ECs;
501   SmallVector<Value *, 16> Worklist;
502   SmallPtrSet<Value *, 4> Roots;
503   SmallPtrSet<Value *, 16> Visited;
504   DenseMap<Value *, uint64_t> DBits;
505   SmallPtrSet<Instruction *, 4> InstructionSet;
506   MapVector<Instruction *, uint64_t> MinBWs;
507 
508   // Determine the roots. We work bottom-up, from truncs or icmps.
509   bool SeenExtFromIllegalType = false;
510   for (auto *BB : Blocks)
511     for (auto &I : *BB) {
512       InstructionSet.insert(&I);
513 
514       if (TTI && (isa<ZExtInst>(&I) || isa<SExtInst>(&I)) &&
515           !TTI->isTypeLegal(I.getOperand(0)->getType()))
516         SeenExtFromIllegalType = true;
517 
518       // Only deal with non-vector integers up to 64-bits wide.
519       if ((isa<TruncInst>(&I) || isa<ICmpInst>(&I)) &&
520           !I.getType()->isVectorTy() &&
521           I.getOperand(0)->getType()->getScalarSizeInBits() <= 64) {
522         // Don't make work for ourselves. If we know the loaded type is legal,
523         // don't add it to the worklist.
524         if (TTI && isa<TruncInst>(&I) && TTI->isTypeLegal(I.getType()))
525           continue;
526 
527         Worklist.push_back(&I);
528         Roots.insert(&I);
529       }
530     }
531   // Early exit.
532   if (Worklist.empty() || (TTI && !SeenExtFromIllegalType))
533     return MinBWs;
534 
535   // Now proceed breadth-first, unioning values together.
536   while (!Worklist.empty()) {
537     Value *Val = Worklist.pop_back_val();
538     Value *Leader = ECs.getOrInsertLeaderValue(Val);
539 
540     if (Visited.count(Val))
541       continue;
542     Visited.insert(Val);
543 
544     // Non-instructions terminate a chain successfully.
545     if (!isa<Instruction>(Val))
546       continue;
547     Instruction *I = cast<Instruction>(Val);
548 
549     // If we encounter a type that is larger than 64 bits, we can't represent
550     // it so bail out.
551     if (DB.getDemandedBits(I).getBitWidth() > 64)
552       return MapVector<Instruction *, uint64_t>();
553 
554     uint64_t V = DB.getDemandedBits(I).getZExtValue();
555     DBits[Leader] |= V;
556     DBits[I] = V;
557 
558     // Casts, loads and instructions outside of our range terminate a chain
559     // successfully.
560     if (isa<SExtInst>(I) || isa<ZExtInst>(I) || isa<LoadInst>(I) ||
561         !InstructionSet.count(I))
562       continue;
563 
564     // Unsafe casts terminate a chain unsuccessfully. We can't do anything
565     // useful with bitcasts, ptrtoints or inttoptrs and it'd be unsafe to
566     // transform anything that relies on them.
567     if (isa<BitCastInst>(I) || isa<PtrToIntInst>(I) || isa<IntToPtrInst>(I) ||
568         !I->getType()->isIntegerTy()) {
569       DBits[Leader] |= ~0ULL;
570       continue;
571     }
572 
573     // We don't modify the types of PHIs. Reductions will already have been
574     // truncated if possible, and inductions' sizes will have been chosen by
575     // indvars.
576     if (isa<PHINode>(I))
577       continue;
578 
579     if (DBits[Leader] == ~0ULL)
580       // All bits demanded, no point continuing.
581       continue;
582 
583     for (Value *O : cast<User>(I)->operands()) {
584       ECs.unionSets(Leader, O);
585       Worklist.push_back(O);
586     }
587   }
588 
589   // Now we've discovered all values, walk them to see if there are
590   // any users we didn't see. If there are, we can't optimize that
591   // chain.
592   for (auto &I : DBits)
593     for (auto *U : I.first->users())
594       if (U->getType()->isIntegerTy() && DBits.count(U) == 0)
595         DBits[ECs.getOrInsertLeaderValue(I.first)] |= ~0ULL;
596 
597   for (auto I = ECs.begin(), E = ECs.end(); I != E; ++I) {
598     uint64_t LeaderDemandedBits = 0;
599     for (Value *M : llvm::make_range(ECs.member_begin(I), ECs.member_end()))
600       LeaderDemandedBits |= DBits[M];
601 
602     uint64_t MinBW = (sizeof(LeaderDemandedBits) * 8) -
603                      llvm::countLeadingZeros(LeaderDemandedBits);
604     // Round up to a power of 2
605     if (!isPowerOf2_64((uint64_t)MinBW))
606       MinBW = NextPowerOf2(MinBW);
607 
608     // We don't modify the types of PHIs. Reductions will already have been
609     // truncated if possible, and inductions' sizes will have been chosen by
610     // indvars.
611     // If we are required to shrink a PHI, abandon this entire equivalence class.
612     bool Abort = false;
613     for (Value *M : llvm::make_range(ECs.member_begin(I), ECs.member_end()))
614       if (isa<PHINode>(M) && MinBW < M->getType()->getScalarSizeInBits()) {
615         Abort = true;
616         break;
617       }
618     if (Abort)
619       continue;
620 
621     for (Value *M : llvm::make_range(ECs.member_begin(I), ECs.member_end())) {
622       if (!isa<Instruction>(M))
623         continue;
624       Type *Ty = M->getType();
625       if (Roots.count(M))
626         Ty = cast<Instruction>(M)->getOperand(0)->getType();
627       if (MinBW < Ty->getScalarSizeInBits())
628         MinBWs[cast<Instruction>(M)] = MinBW;
629     }
630   }
631 
632   return MinBWs;
633 }
634 
635 /// Add all access groups in @p AccGroups to @p List.
636 template <typename ListT>
addToAccessGroupList(ListT & List,MDNode * AccGroups)637 static void addToAccessGroupList(ListT &List, MDNode *AccGroups) {
638   // Interpret an access group as a list containing itself.
639   if (AccGroups->getNumOperands() == 0) {
640     assert(isValidAsAccessGroup(AccGroups) && "Node must be an access group");
641     List.insert(AccGroups);
642     return;
643   }
644 
645   for (auto &AccGroupListOp : AccGroups->operands()) {
646     auto *Item = cast<MDNode>(AccGroupListOp.get());
647     assert(isValidAsAccessGroup(Item) && "List item must be an access group");
648     List.insert(Item);
649   }
650 }
651 
uniteAccessGroups(MDNode * AccGroups1,MDNode * AccGroups2)652 MDNode *llvm::uniteAccessGroups(MDNode *AccGroups1, MDNode *AccGroups2) {
653   if (!AccGroups1)
654     return AccGroups2;
655   if (!AccGroups2)
656     return AccGroups1;
657   if (AccGroups1 == AccGroups2)
658     return AccGroups1;
659 
660   SmallSetVector<Metadata *, 4> Union;
661   addToAccessGroupList(Union, AccGroups1);
662   addToAccessGroupList(Union, AccGroups2);
663 
664   if (Union.size() == 0)
665     return nullptr;
666   if (Union.size() == 1)
667     return cast<MDNode>(Union.front());
668 
669   LLVMContext &Ctx = AccGroups1->getContext();
670   return MDNode::get(Ctx, Union.getArrayRef());
671 }
672 
intersectAccessGroups(const Instruction * Inst1,const Instruction * Inst2)673 MDNode *llvm::intersectAccessGroups(const Instruction *Inst1,
674                                     const Instruction *Inst2) {
675   bool MayAccessMem1 = Inst1->mayReadOrWriteMemory();
676   bool MayAccessMem2 = Inst2->mayReadOrWriteMemory();
677 
678   if (!MayAccessMem1 && !MayAccessMem2)
679     return nullptr;
680   if (!MayAccessMem1)
681     return Inst2->getMetadata(LLVMContext::MD_access_group);
682   if (!MayAccessMem2)
683     return Inst1->getMetadata(LLVMContext::MD_access_group);
684 
685   MDNode *MD1 = Inst1->getMetadata(LLVMContext::MD_access_group);
686   MDNode *MD2 = Inst2->getMetadata(LLVMContext::MD_access_group);
687   if (!MD1 || !MD2)
688     return nullptr;
689   if (MD1 == MD2)
690     return MD1;
691 
692   // Use set for scalable 'contains' check.
693   SmallPtrSet<Metadata *, 4> AccGroupSet2;
694   addToAccessGroupList(AccGroupSet2, MD2);
695 
696   SmallVector<Metadata *, 4> Intersection;
697   if (MD1->getNumOperands() == 0) {
698     assert(isValidAsAccessGroup(MD1) && "Node must be an access group");
699     if (AccGroupSet2.count(MD1))
700       Intersection.push_back(MD1);
701   } else {
702     for (const MDOperand &Node : MD1->operands()) {
703       auto *Item = cast<MDNode>(Node.get());
704       assert(isValidAsAccessGroup(Item) && "List item must be an access group");
705       if (AccGroupSet2.count(Item))
706         Intersection.push_back(Item);
707     }
708   }
709 
710   if (Intersection.size() == 0)
711     return nullptr;
712   if (Intersection.size() == 1)
713     return cast<MDNode>(Intersection.front());
714 
715   LLVMContext &Ctx = Inst1->getContext();
716   return MDNode::get(Ctx, Intersection);
717 }
718 
719 /// \returns \p I after propagating metadata from \p VL.
propagateMetadata(Instruction * Inst,ArrayRef<Value * > VL)720 Instruction *llvm::propagateMetadata(Instruction *Inst, ArrayRef<Value *> VL) {
721   if (VL.empty())
722     return Inst;
723   Instruction *I0 = cast<Instruction>(VL[0]);
724   SmallVector<std::pair<unsigned, MDNode *>, 4> Metadata;
725   I0->getAllMetadataOtherThanDebugLoc(Metadata);
726 
727   for (auto Kind : {LLVMContext::MD_tbaa, LLVMContext::MD_alias_scope,
728                     LLVMContext::MD_noalias, LLVMContext::MD_fpmath,
729                     LLVMContext::MD_nontemporal, LLVMContext::MD_invariant_load,
730                     LLVMContext::MD_access_group}) {
731     MDNode *MD = I0->getMetadata(Kind);
732 
733     for (int J = 1, E = VL.size(); MD && J != E; ++J) {
734       const Instruction *IJ = cast<Instruction>(VL[J]);
735       MDNode *IMD = IJ->getMetadata(Kind);
736       switch (Kind) {
737       case LLVMContext::MD_tbaa:
738         MD = MDNode::getMostGenericTBAA(MD, IMD);
739         break;
740       case LLVMContext::MD_alias_scope:
741         MD = MDNode::getMostGenericAliasScope(MD, IMD);
742         break;
743       case LLVMContext::MD_fpmath:
744         MD = MDNode::getMostGenericFPMath(MD, IMD);
745         break;
746       case LLVMContext::MD_noalias:
747       case LLVMContext::MD_nontemporal:
748       case LLVMContext::MD_invariant_load:
749         MD = MDNode::intersect(MD, IMD);
750         break;
751       case LLVMContext::MD_access_group:
752         MD = intersectAccessGroups(Inst, IJ);
753         break;
754       default:
755         llvm_unreachable("unhandled metadata");
756       }
757     }
758 
759     Inst->setMetadata(Kind, MD);
760   }
761 
762   return Inst;
763 }
764 
765 Constant *
createBitMaskForGaps(IRBuilderBase & Builder,unsigned VF,const InterleaveGroup<Instruction> & Group)766 llvm::createBitMaskForGaps(IRBuilderBase &Builder, unsigned VF,
767                            const InterleaveGroup<Instruction> &Group) {
768   // All 1's means mask is not needed.
769   if (Group.getNumMembers() == Group.getFactor())
770     return nullptr;
771 
772   // TODO: support reversed access.
773   assert(!Group.isReverse() && "Reversed group not supported.");
774 
775   SmallVector<Constant *, 16> Mask;
776   for (unsigned i = 0; i < VF; i++)
777     for (unsigned j = 0; j < Group.getFactor(); ++j) {
778       unsigned HasMember = Group.getMember(j) ? 1 : 0;
779       Mask.push_back(Builder.getInt1(HasMember));
780     }
781 
782   return ConstantVector::get(Mask);
783 }
784 
785 llvm::SmallVector<int, 16>
createReplicatedMask(unsigned ReplicationFactor,unsigned VF)786 llvm::createReplicatedMask(unsigned ReplicationFactor, unsigned VF) {
787   SmallVector<int, 16> MaskVec;
788   for (unsigned i = 0; i < VF; i++)
789     for (unsigned j = 0; j < ReplicationFactor; j++)
790       MaskVec.push_back(i);
791 
792   return MaskVec;
793 }
794 
createInterleaveMask(unsigned VF,unsigned NumVecs)795 llvm::SmallVector<int, 16> llvm::createInterleaveMask(unsigned VF,
796                                                       unsigned NumVecs) {
797   SmallVector<int, 16> Mask;
798   for (unsigned i = 0; i < VF; i++)
799     for (unsigned j = 0; j < NumVecs; j++)
800       Mask.push_back(j * VF + i);
801 
802   return Mask;
803 }
804 
805 llvm::SmallVector<int, 16>
createStrideMask(unsigned Start,unsigned Stride,unsigned VF)806 llvm::createStrideMask(unsigned Start, unsigned Stride, unsigned VF) {
807   SmallVector<int, 16> Mask;
808   for (unsigned i = 0; i < VF; i++)
809     Mask.push_back(Start + i * Stride);
810 
811   return Mask;
812 }
813 
createSequentialMask(unsigned Start,unsigned NumInts,unsigned NumUndefs)814 llvm::SmallVector<int, 16> llvm::createSequentialMask(unsigned Start,
815                                                       unsigned NumInts,
816                                                       unsigned NumUndefs) {
817   SmallVector<int, 16> Mask;
818   for (unsigned i = 0; i < NumInts; i++)
819     Mask.push_back(Start + i);
820 
821   for (unsigned i = 0; i < NumUndefs; i++)
822     Mask.push_back(-1);
823 
824   return Mask;
825 }
826 
827 /// A helper function for concatenating vectors. This function concatenates two
828 /// vectors having the same element type. If the second vector has fewer
829 /// elements than the first, it is padded with undefs.
concatenateTwoVectors(IRBuilderBase & Builder,Value * V1,Value * V2)830 static Value *concatenateTwoVectors(IRBuilderBase &Builder, Value *V1,
831                                     Value *V2) {
832   VectorType *VecTy1 = dyn_cast<VectorType>(V1->getType());
833   VectorType *VecTy2 = dyn_cast<VectorType>(V2->getType());
834   assert(VecTy1 && VecTy2 &&
835          VecTy1->getScalarType() == VecTy2->getScalarType() &&
836          "Expect two vectors with the same element type");
837 
838   unsigned NumElts1 = cast<FixedVectorType>(VecTy1)->getNumElements();
839   unsigned NumElts2 = cast<FixedVectorType>(VecTy2)->getNumElements();
840   assert(NumElts1 >= NumElts2 && "Unexpect the first vector has less elements");
841 
842   if (NumElts1 > NumElts2) {
843     // Extend with UNDEFs.
844     V2 = Builder.CreateShuffleVector(
845         V2, createSequentialMask(0, NumElts2, NumElts1 - NumElts2));
846   }
847 
848   return Builder.CreateShuffleVector(
849       V1, V2, createSequentialMask(0, NumElts1 + NumElts2, 0));
850 }
851 
concatenateVectors(IRBuilderBase & Builder,ArrayRef<Value * > Vecs)852 Value *llvm::concatenateVectors(IRBuilderBase &Builder,
853                                 ArrayRef<Value *> Vecs) {
854   unsigned NumVecs = Vecs.size();
855   assert(NumVecs > 1 && "Should be at least two vectors");
856 
857   SmallVector<Value *, 8> ResList;
858   ResList.append(Vecs.begin(), Vecs.end());
859   do {
860     SmallVector<Value *, 8> TmpList;
861     for (unsigned i = 0; i < NumVecs - 1; i += 2) {
862       Value *V0 = ResList[i], *V1 = ResList[i + 1];
863       assert((V0->getType() == V1->getType() || i == NumVecs - 2) &&
864              "Only the last vector may have a different type");
865 
866       TmpList.push_back(concatenateTwoVectors(Builder, V0, V1));
867     }
868 
869     // Push the last vector if the total number of vectors is odd.
870     if (NumVecs % 2 != 0)
871       TmpList.push_back(ResList[NumVecs - 1]);
872 
873     ResList = TmpList;
874     NumVecs = ResList.size();
875   } while (NumVecs > 1);
876 
877   return ResList[0];
878 }
879 
maskIsAllZeroOrUndef(Value * Mask)880 bool llvm::maskIsAllZeroOrUndef(Value *Mask) {
881   assert(isa<VectorType>(Mask->getType()) &&
882          isa<IntegerType>(Mask->getType()->getScalarType()) &&
883          cast<IntegerType>(Mask->getType()->getScalarType())->getBitWidth() ==
884              1 &&
885          "Mask must be a vector of i1");
886 
887   auto *ConstMask = dyn_cast<Constant>(Mask);
888   if (!ConstMask)
889     return false;
890   if (ConstMask->isNullValue() || isa<UndefValue>(ConstMask))
891     return true;
892   if (isa<ScalableVectorType>(ConstMask->getType()))
893     return false;
894   for (unsigned
895            I = 0,
896            E = cast<FixedVectorType>(ConstMask->getType())->getNumElements();
897        I != E; ++I) {
898     if (auto *MaskElt = ConstMask->getAggregateElement(I))
899       if (MaskElt->isNullValue() || isa<UndefValue>(MaskElt))
900         continue;
901     return false;
902   }
903   return true;
904 }
905 
maskIsAllOneOrUndef(Value * Mask)906 bool llvm::maskIsAllOneOrUndef(Value *Mask) {
907   assert(isa<VectorType>(Mask->getType()) &&
908          isa<IntegerType>(Mask->getType()->getScalarType()) &&
909          cast<IntegerType>(Mask->getType()->getScalarType())->getBitWidth() ==
910              1 &&
911          "Mask must be a vector of i1");
912 
913   auto *ConstMask = dyn_cast<Constant>(Mask);
914   if (!ConstMask)
915     return false;
916   if (ConstMask->isAllOnesValue() || isa<UndefValue>(ConstMask))
917     return true;
918   if (isa<ScalableVectorType>(ConstMask->getType()))
919     return false;
920   for (unsigned
921            I = 0,
922            E = cast<FixedVectorType>(ConstMask->getType())->getNumElements();
923        I != E; ++I) {
924     if (auto *MaskElt = ConstMask->getAggregateElement(I))
925       if (MaskElt->isAllOnesValue() || isa<UndefValue>(MaskElt))
926         continue;
927     return false;
928   }
929   return true;
930 }
931 
932 /// TODO: This is a lot like known bits, but for
933 /// vectors.  Is there something we can common this with?
possiblyDemandedEltsInMask(Value * Mask)934 APInt llvm::possiblyDemandedEltsInMask(Value *Mask) {
935   assert(isa<FixedVectorType>(Mask->getType()) &&
936          isa<IntegerType>(Mask->getType()->getScalarType()) &&
937          cast<IntegerType>(Mask->getType()->getScalarType())->getBitWidth() ==
938              1 &&
939          "Mask must be a fixed width vector of i1");
940 
941   const unsigned VWidth =
942       cast<FixedVectorType>(Mask->getType())->getNumElements();
943   APInt DemandedElts = APInt::getAllOnesValue(VWidth);
944   if (auto *CV = dyn_cast<ConstantVector>(Mask))
945     for (unsigned i = 0; i < VWidth; i++)
946       if (CV->getAggregateElement(i)->isNullValue())
947         DemandedElts.clearBit(i);
948   return DemandedElts;
949 }
950 
isStrided(int Stride)951 bool InterleavedAccessInfo::isStrided(int Stride) {
952   unsigned Factor = std::abs(Stride);
953   return Factor >= 2 && Factor <= MaxInterleaveGroupFactor;
954 }
955 
collectConstStrideAccesses(MapVector<Instruction *,StrideDescriptor> & AccessStrideInfo,const ValueToValueMap & Strides)956 void InterleavedAccessInfo::collectConstStrideAccesses(
957     MapVector<Instruction *, StrideDescriptor> &AccessStrideInfo,
958     const ValueToValueMap &Strides) {
959   auto &DL = TheLoop->getHeader()->getModule()->getDataLayout();
960 
961   // Since it's desired that the load/store instructions be maintained in
962   // "program order" for the interleaved access analysis, we have to visit the
963   // blocks in the loop in reverse postorder (i.e., in a topological order).
964   // Such an ordering will ensure that any load/store that may be executed
965   // before a second load/store will precede the second load/store in
966   // AccessStrideInfo.
967   LoopBlocksDFS DFS(TheLoop);
968   DFS.perform(LI);
969   for (BasicBlock *BB : make_range(DFS.beginRPO(), DFS.endRPO()))
970     for (auto &I : *BB) {
971       Value *Ptr = getLoadStorePointerOperand(&I);
972       if (!Ptr)
973         continue;
974       Type *ElementTy = getLoadStoreType(&I);
975 
976       // We don't check wrapping here because we don't know yet if Ptr will be
977       // part of a full group or a group with gaps. Checking wrapping for all
978       // pointers (even those that end up in groups with no gaps) will be overly
979       // conservative. For full groups, wrapping should be ok since if we would
980       // wrap around the address space we would do a memory access at nullptr
981       // even without the transformation. The wrapping checks are therefore
982       // deferred until after we've formed the interleaved groups.
983       int64_t Stride = getPtrStride(PSE, Ptr, TheLoop, Strides,
984                                     /*Assume=*/true, /*ShouldCheckWrap=*/false);
985 
986       const SCEV *Scev = replaceSymbolicStrideSCEV(PSE, Strides, Ptr);
987       uint64_t Size = DL.getTypeAllocSize(ElementTy);
988       AccessStrideInfo[&I] = StrideDescriptor(Stride, Scev, Size,
989                                               getLoadStoreAlignment(&I));
990     }
991 }
992 
993 // Analyze interleaved accesses and collect them into interleaved load and
994 // store groups.
995 //
996 // When generating code for an interleaved load group, we effectively hoist all
997 // loads in the group to the location of the first load in program order. When
998 // generating code for an interleaved store group, we sink all stores to the
999 // location of the last store. This code motion can change the order of load
1000 // and store instructions and may break dependences.
1001 //
1002 // The code generation strategy mentioned above ensures that we won't violate
1003 // any write-after-read (WAR) dependences.
1004 //
1005 // E.g., for the WAR dependence:  a = A[i];      // (1)
1006 //                                A[i] = b;      // (2)
1007 //
1008 // The store group of (2) is always inserted at or below (2), and the load
1009 // group of (1) is always inserted at or above (1). Thus, the instructions will
1010 // never be reordered. All other dependences are checked to ensure the
1011 // correctness of the instruction reordering.
1012 //
1013 // The algorithm visits all memory accesses in the loop in bottom-up program
1014 // order. Program order is established by traversing the blocks in the loop in
1015 // reverse postorder when collecting the accesses.
1016 //
1017 // We visit the memory accesses in bottom-up order because it can simplify the
1018 // construction of store groups in the presence of write-after-write (WAW)
1019 // dependences.
1020 //
1021 // E.g., for the WAW dependence:  A[i] = a;      // (1)
1022 //                                A[i] = b;      // (2)
1023 //                                A[i + 1] = c;  // (3)
1024 //
1025 // We will first create a store group with (3) and (2). (1) can't be added to
1026 // this group because it and (2) are dependent. However, (1) can be grouped
1027 // with other accesses that may precede it in program order. Note that a
1028 // bottom-up order does not imply that WAW dependences should not be checked.
analyzeInterleaving(bool EnablePredicatedInterleavedMemAccesses)1029 void InterleavedAccessInfo::analyzeInterleaving(
1030                                  bool EnablePredicatedInterleavedMemAccesses) {
1031   LLVM_DEBUG(dbgs() << "LV: Analyzing interleaved accesses...\n");
1032   const ValueToValueMap &Strides = LAI->getSymbolicStrides();
1033 
1034   // Holds all accesses with a constant stride.
1035   MapVector<Instruction *, StrideDescriptor> AccessStrideInfo;
1036   collectConstStrideAccesses(AccessStrideInfo, Strides);
1037 
1038   if (AccessStrideInfo.empty())
1039     return;
1040 
1041   // Collect the dependences in the loop.
1042   collectDependences();
1043 
1044   // Holds all interleaved store groups temporarily.
1045   SmallSetVector<InterleaveGroup<Instruction> *, 4> StoreGroups;
1046   // Holds all interleaved load groups temporarily.
1047   SmallSetVector<InterleaveGroup<Instruction> *, 4> LoadGroups;
1048 
1049   // Search in bottom-up program order for pairs of accesses (A and B) that can
1050   // form interleaved load or store groups. In the algorithm below, access A
1051   // precedes access B in program order. We initialize a group for B in the
1052   // outer loop of the algorithm, and then in the inner loop, we attempt to
1053   // insert each A into B's group if:
1054   //
1055   //  1. A and B have the same stride,
1056   //  2. A and B have the same memory object size, and
1057   //  3. A belongs in B's group according to its distance from B.
1058   //
1059   // Special care is taken to ensure group formation will not break any
1060   // dependences.
1061   for (auto BI = AccessStrideInfo.rbegin(), E = AccessStrideInfo.rend();
1062        BI != E; ++BI) {
1063     Instruction *B = BI->first;
1064     StrideDescriptor DesB = BI->second;
1065 
1066     // Initialize a group for B if it has an allowable stride. Even if we don't
1067     // create a group for B, we continue with the bottom-up algorithm to ensure
1068     // we don't break any of B's dependences.
1069     InterleaveGroup<Instruction> *Group = nullptr;
1070     if (isStrided(DesB.Stride) &&
1071         (!isPredicated(B->getParent()) || EnablePredicatedInterleavedMemAccesses)) {
1072       Group = getInterleaveGroup(B);
1073       if (!Group) {
1074         LLVM_DEBUG(dbgs() << "LV: Creating an interleave group with:" << *B
1075                           << '\n');
1076         Group = createInterleaveGroup(B, DesB.Stride, DesB.Alignment);
1077       }
1078       if (B->mayWriteToMemory())
1079         StoreGroups.insert(Group);
1080       else
1081         LoadGroups.insert(Group);
1082     }
1083 
1084     for (auto AI = std::next(BI); AI != E; ++AI) {
1085       Instruction *A = AI->first;
1086       StrideDescriptor DesA = AI->second;
1087 
1088       // Our code motion strategy implies that we can't have dependences
1089       // between accesses in an interleaved group and other accesses located
1090       // between the first and last member of the group. Note that this also
1091       // means that a group can't have more than one member at a given offset.
1092       // The accesses in a group can have dependences with other accesses, but
1093       // we must ensure we don't extend the boundaries of the group such that
1094       // we encompass those dependent accesses.
1095       //
1096       // For example, assume we have the sequence of accesses shown below in a
1097       // stride-2 loop:
1098       //
1099       //  (1, 2) is a group | A[i]   = a;  // (1)
1100       //                    | A[i-1] = b;  // (2) |
1101       //                      A[i-3] = c;  // (3)
1102       //                      A[i]   = d;  // (4) | (2, 4) is not a group
1103       //
1104       // Because accesses (2) and (3) are dependent, we can group (2) with (1)
1105       // but not with (4). If we did, the dependent access (3) would be within
1106       // the boundaries of the (2, 4) group.
1107       if (!canReorderMemAccessesForInterleavedGroups(&*AI, &*BI)) {
1108         // If a dependence exists and A is already in a group, we know that A
1109         // must be a store since A precedes B and WAR dependences are allowed.
1110         // Thus, A would be sunk below B. We release A's group to prevent this
1111         // illegal code motion. A will then be free to form another group with
1112         // instructions that precede it.
1113         if (isInterleaved(A)) {
1114           InterleaveGroup<Instruction> *StoreGroup = getInterleaveGroup(A);
1115 
1116           LLVM_DEBUG(dbgs() << "LV: Invalidated store group due to "
1117                                "dependence between " << *A << " and "<< *B << '\n');
1118 
1119           StoreGroups.remove(StoreGroup);
1120           releaseGroup(StoreGroup);
1121         }
1122 
1123         // If a dependence exists and A is not already in a group (or it was
1124         // and we just released it), B might be hoisted above A (if B is a
1125         // load) or another store might be sunk below A (if B is a store). In
1126         // either case, we can't add additional instructions to B's group. B
1127         // will only form a group with instructions that it precedes.
1128         break;
1129       }
1130 
1131       // At this point, we've checked for illegal code motion. If either A or B
1132       // isn't strided, there's nothing left to do.
1133       if (!isStrided(DesA.Stride) || !isStrided(DesB.Stride))
1134         continue;
1135 
1136       // Ignore A if it's already in a group or isn't the same kind of memory
1137       // operation as B.
1138       // Note that mayReadFromMemory() isn't mutually exclusive to
1139       // mayWriteToMemory in the case of atomic loads. We shouldn't see those
1140       // here, canVectorizeMemory() should have returned false - except for the
1141       // case we asked for optimization remarks.
1142       if (isInterleaved(A) ||
1143           (A->mayReadFromMemory() != B->mayReadFromMemory()) ||
1144           (A->mayWriteToMemory() != B->mayWriteToMemory()))
1145         continue;
1146 
1147       // Check rules 1 and 2. Ignore A if its stride or size is different from
1148       // that of B.
1149       if (DesA.Stride != DesB.Stride || DesA.Size != DesB.Size)
1150         continue;
1151 
1152       // Ignore A if the memory object of A and B don't belong to the same
1153       // address space
1154       if (getLoadStoreAddressSpace(A) != getLoadStoreAddressSpace(B))
1155         continue;
1156 
1157       // Calculate the distance from A to B.
1158       const SCEVConstant *DistToB = dyn_cast<SCEVConstant>(
1159           PSE.getSE()->getMinusSCEV(DesA.Scev, DesB.Scev));
1160       if (!DistToB)
1161         continue;
1162       int64_t DistanceToB = DistToB->getAPInt().getSExtValue();
1163 
1164       // Check rule 3. Ignore A if its distance to B is not a multiple of the
1165       // size.
1166       if (DistanceToB % static_cast<int64_t>(DesB.Size))
1167         continue;
1168 
1169       // All members of a predicated interleave-group must have the same predicate,
1170       // and currently must reside in the same BB.
1171       BasicBlock *BlockA = A->getParent();
1172       BasicBlock *BlockB = B->getParent();
1173       if ((isPredicated(BlockA) || isPredicated(BlockB)) &&
1174           (!EnablePredicatedInterleavedMemAccesses || BlockA != BlockB))
1175         continue;
1176 
1177       // The index of A is the index of B plus A's distance to B in multiples
1178       // of the size.
1179       int IndexA =
1180           Group->getIndex(B) + DistanceToB / static_cast<int64_t>(DesB.Size);
1181 
1182       // Try to insert A into B's group.
1183       if (Group->insertMember(A, IndexA, DesA.Alignment)) {
1184         LLVM_DEBUG(dbgs() << "LV: Inserted:" << *A << '\n'
1185                           << "    into the interleave group with" << *B
1186                           << '\n');
1187         InterleaveGroupMap[A] = Group;
1188 
1189         // Set the first load in program order as the insert position.
1190         if (A->mayReadFromMemory())
1191           Group->setInsertPos(A);
1192       }
1193     } // Iteration over A accesses.
1194   }   // Iteration over B accesses.
1195 
1196   // Remove interleaved store groups with gaps.
1197   for (auto *Group : StoreGroups)
1198     if (Group->getNumMembers() != Group->getFactor()) {
1199       LLVM_DEBUG(
1200           dbgs() << "LV: Invalidate candidate interleaved store group due "
1201                     "to gaps.\n");
1202       releaseGroup(Group);
1203     }
1204   // Remove interleaved groups with gaps (currently only loads) whose memory
1205   // accesses may wrap around. We have to revisit the getPtrStride analysis,
1206   // this time with ShouldCheckWrap=true, since collectConstStrideAccesses does
1207   // not check wrapping (see documentation there).
1208   // FORNOW we use Assume=false;
1209   // TODO: Change to Assume=true but making sure we don't exceed the threshold
1210   // of runtime SCEV assumptions checks (thereby potentially failing to
1211   // vectorize altogether).
1212   // Additional optional optimizations:
1213   // TODO: If we are peeling the loop and we know that the first pointer doesn't
1214   // wrap then we can deduce that all pointers in the group don't wrap.
1215   // This means that we can forcefully peel the loop in order to only have to
1216   // check the first pointer for no-wrap. When we'll change to use Assume=true
1217   // we'll only need at most one runtime check per interleaved group.
1218   for (auto *Group : LoadGroups) {
1219     // Case 1: A full group. Can Skip the checks; For full groups, if the wide
1220     // load would wrap around the address space we would do a memory access at
1221     // nullptr even without the transformation.
1222     if (Group->getNumMembers() == Group->getFactor())
1223       continue;
1224 
1225     // Case 2: If first and last members of the group don't wrap this implies
1226     // that all the pointers in the group don't wrap.
1227     // So we check only group member 0 (which is always guaranteed to exist),
1228     // and group member Factor - 1; If the latter doesn't exist we rely on
1229     // peeling (if it is a non-reversed accsess -- see Case 3).
1230     Value *FirstMemberPtr = getLoadStorePointerOperand(Group->getMember(0));
1231     if (!getPtrStride(PSE, FirstMemberPtr, TheLoop, Strides, /*Assume=*/false,
1232                       /*ShouldCheckWrap=*/true)) {
1233       LLVM_DEBUG(
1234           dbgs() << "LV: Invalidate candidate interleaved group due to "
1235                     "first group member potentially pointer-wrapping.\n");
1236       releaseGroup(Group);
1237       continue;
1238     }
1239     Instruction *LastMember = Group->getMember(Group->getFactor() - 1);
1240     if (LastMember) {
1241       Value *LastMemberPtr = getLoadStorePointerOperand(LastMember);
1242       if (!getPtrStride(PSE, LastMemberPtr, TheLoop, Strides, /*Assume=*/false,
1243                         /*ShouldCheckWrap=*/true)) {
1244         LLVM_DEBUG(
1245             dbgs() << "LV: Invalidate candidate interleaved group due to "
1246                       "last group member potentially pointer-wrapping.\n");
1247         releaseGroup(Group);
1248       }
1249     } else {
1250       // Case 3: A non-reversed interleaved load group with gaps: We need
1251       // to execute at least one scalar epilogue iteration. This will ensure
1252       // we don't speculatively access memory out-of-bounds. We only need
1253       // to look for a member at index factor - 1, since every group must have
1254       // a member at index zero.
1255       if (Group->isReverse()) {
1256         LLVM_DEBUG(
1257             dbgs() << "LV: Invalidate candidate interleaved group due to "
1258                       "a reverse access with gaps.\n");
1259         releaseGroup(Group);
1260         continue;
1261       }
1262       LLVM_DEBUG(
1263           dbgs() << "LV: Interleaved group requires epilogue iteration.\n");
1264       RequiresScalarEpilogue = true;
1265     }
1266   }
1267 }
1268 
invalidateGroupsRequiringScalarEpilogue()1269 void InterleavedAccessInfo::invalidateGroupsRequiringScalarEpilogue() {
1270   // If no group had triggered the requirement to create an epilogue loop,
1271   // there is nothing to do.
1272   if (!requiresScalarEpilogue())
1273     return;
1274 
1275   bool ReleasedGroup = false;
1276   // Release groups requiring scalar epilogues. Note that this also removes them
1277   // from InterleaveGroups.
1278   for (auto *Group : make_early_inc_range(InterleaveGroups)) {
1279     if (!Group->requiresScalarEpilogue())
1280       continue;
1281     LLVM_DEBUG(
1282         dbgs()
1283         << "LV: Invalidate candidate interleaved group due to gaps that "
1284            "require a scalar epilogue (not allowed under optsize) and cannot "
1285            "be masked (not enabled). \n");
1286     releaseGroup(Group);
1287     ReleasedGroup = true;
1288   }
1289   assert(ReleasedGroup && "At least one group must be invalidated, as a "
1290                           "scalar epilogue was required");
1291   (void)ReleasedGroup;
1292   RequiresScalarEpilogue = false;
1293 }
1294 
1295 template <typename InstT>
addMetadata(InstT * NewInst) const1296 void InterleaveGroup<InstT>::addMetadata(InstT *NewInst) const {
1297   llvm_unreachable("addMetadata can only be used for Instruction");
1298 }
1299 
1300 namespace llvm {
1301 template <>
addMetadata(Instruction * NewInst) const1302 void InterleaveGroup<Instruction>::addMetadata(Instruction *NewInst) const {
1303   SmallVector<Value *, 4> VL;
1304   std::transform(Members.begin(), Members.end(), std::back_inserter(VL),
1305                  [](std::pair<int, Instruction *> p) { return p.second; });
1306   propagateMetadata(NewInst, VL);
1307 }
1308 }
1309 
mangleTLIVectorName(StringRef VectorName,StringRef ScalarName,unsigned numArgs,ElementCount VF)1310 std::string VFABI::mangleTLIVectorName(StringRef VectorName,
1311                                        StringRef ScalarName, unsigned numArgs,
1312                                        ElementCount VF) {
1313   SmallString<256> Buffer;
1314   llvm::raw_svector_ostream Out(Buffer);
1315   Out << "_ZGV" << VFABI::_LLVM_ << "N";
1316   if (VF.isScalable())
1317     Out << 'x';
1318   else
1319     Out << VF.getFixedValue();
1320   for (unsigned I = 0; I < numArgs; ++I)
1321     Out << "v";
1322   Out << "_" << ScalarName << "(" << VectorName << ")";
1323   return std::string(Out.str());
1324 }
1325 
getVectorVariantNames(const CallInst & CI,SmallVectorImpl<std::string> & VariantMappings)1326 void VFABI::getVectorVariantNames(
1327     const CallInst &CI, SmallVectorImpl<std::string> &VariantMappings) {
1328   const StringRef S =
1329       CI.getAttribute(AttributeList::FunctionIndex, VFABI::MappingsAttrName)
1330           .getValueAsString();
1331   if (S.empty())
1332     return;
1333 
1334   SmallVector<StringRef, 8> ListAttr;
1335   S.split(ListAttr, ",");
1336 
1337   for (auto &S : SetVector<StringRef>(ListAttr.begin(), ListAttr.end())) {
1338 #ifndef NDEBUG
1339     LLVM_DEBUG(dbgs() << "VFABI: adding mapping '" << S << "'\n");
1340     Optional<VFInfo> Info = VFABI::tryDemangleForVFABI(S, *(CI.getModule()));
1341     assert(Info.hasValue() && "Invalid name for a VFABI variant.");
1342     assert(CI.getModule()->getFunction(Info.getValue().VectorName) &&
1343            "Vector function is missing.");
1344 #endif
1345     VariantMappings.push_back(std::string(S));
1346   }
1347 }
1348 
hasValidParameterList() const1349 bool VFShape::hasValidParameterList() const {
1350   for (unsigned Pos = 0, NumParams = Parameters.size(); Pos < NumParams;
1351        ++Pos) {
1352     assert(Parameters[Pos].ParamPos == Pos && "Broken parameter list.");
1353 
1354     switch (Parameters[Pos].ParamKind) {
1355     default: // Nothing to check.
1356       break;
1357     case VFParamKind::OMP_Linear:
1358     case VFParamKind::OMP_LinearRef:
1359     case VFParamKind::OMP_LinearVal:
1360     case VFParamKind::OMP_LinearUVal:
1361       // Compile time linear steps must be non-zero.
1362       if (Parameters[Pos].LinearStepOrPos == 0)
1363         return false;
1364       break;
1365     case VFParamKind::OMP_LinearPos:
1366     case VFParamKind::OMP_LinearRefPos:
1367     case VFParamKind::OMP_LinearValPos:
1368     case VFParamKind::OMP_LinearUValPos:
1369       // The runtime linear step must be referring to some other
1370       // parameters in the signature.
1371       if (Parameters[Pos].LinearStepOrPos >= int(NumParams))
1372         return false;
1373       // The linear step parameter must be marked as uniform.
1374       if (Parameters[Parameters[Pos].LinearStepOrPos].ParamKind !=
1375           VFParamKind::OMP_Uniform)
1376         return false;
1377       // The linear step parameter can't point at itself.
1378       if (Parameters[Pos].LinearStepOrPos == int(Pos))
1379         return false;
1380       break;
1381     case VFParamKind::GlobalPredicate:
1382       // The global predicate must be the unique. Can be placed anywhere in the
1383       // signature.
1384       for (unsigned NextPos = Pos + 1; NextPos < NumParams; ++NextPos)
1385         if (Parameters[NextPos].ParamKind == VFParamKind::GlobalPredicate)
1386           return false;
1387       break;
1388     }
1389   }
1390   return true;
1391 }
1392