1 //===- BBVectorize.cpp - A Basic-Block Vectorizer -------------------------===// 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 a basic-block vectorization pass. The algorithm was 11 // inspired by that used by the Vienna MAP Vectorizor by Franchetti and Kral, 12 // et al. It works by looking for chains of pairable operations and then 13 // pairing them. 14 // 15 //===----------------------------------------------------------------------===// 16 17 #define BBV_NAME "bb-vectorize" 18 #include "llvm/Transforms/Vectorize.h" 19 #include "llvm/ADT/DenseMap.h" 20 #include "llvm/ADT/DenseSet.h" 21 #include "llvm/ADT/STLExtras.h" 22 #include "llvm/ADT/SmallSet.h" 23 #include "llvm/ADT/SmallVector.h" 24 #include "llvm/ADT/Statistic.h" 25 #include "llvm/ADT/StringExtras.h" 26 #include "llvm/Analysis/AliasAnalysis.h" 27 #include "llvm/Analysis/AliasSetTracker.h" 28 #include "llvm/Analysis/ScalarEvolution.h" 29 #include "llvm/Analysis/ScalarEvolutionExpressions.h" 30 #include "llvm/Analysis/TargetTransformInfo.h" 31 #include "llvm/Analysis/ValueTracking.h" 32 #include "llvm/IR/Constants.h" 33 #include "llvm/IR/DataLayout.h" 34 #include "llvm/IR/DerivedTypes.h" 35 #include "llvm/IR/Dominators.h" 36 #include "llvm/IR/Function.h" 37 #include "llvm/IR/Instructions.h" 38 #include "llvm/IR/IntrinsicInst.h" 39 #include "llvm/IR/Intrinsics.h" 40 #include "llvm/IR/LLVMContext.h" 41 #include "llvm/IR/Metadata.h" 42 #include "llvm/IR/Type.h" 43 #include "llvm/IR/ValueHandle.h" 44 #include "llvm/Pass.h" 45 #include "llvm/Support/CommandLine.h" 46 #include "llvm/Support/Debug.h" 47 #include "llvm/Support/raw_ostream.h" 48 #include "llvm/Transforms/Utils/Local.h" 49 #include <algorithm> 50 using namespace llvm; 51 52 #define DEBUG_TYPE BBV_NAME 53 54 static cl::opt<bool> 55 IgnoreTargetInfo("bb-vectorize-ignore-target-info", cl::init(false), 56 cl::Hidden, cl::desc("Ignore target information")); 57 58 static cl::opt<unsigned> 59 ReqChainDepth("bb-vectorize-req-chain-depth", cl::init(6), cl::Hidden, 60 cl::desc("The required chain depth for vectorization")); 61 62 static cl::opt<bool> 63 UseChainDepthWithTI("bb-vectorize-use-chain-depth", cl::init(false), 64 cl::Hidden, cl::desc("Use the chain depth requirement with" 65 " target information")); 66 67 static cl::opt<unsigned> 68 SearchLimit("bb-vectorize-search-limit", cl::init(400), cl::Hidden, 69 cl::desc("The maximum search distance for instruction pairs")); 70 71 static cl::opt<bool> 72 SplatBreaksChain("bb-vectorize-splat-breaks-chain", cl::init(false), cl::Hidden, 73 cl::desc("Replicating one element to a pair breaks the chain")); 74 75 static cl::opt<unsigned> 76 VectorBits("bb-vectorize-vector-bits", cl::init(128), cl::Hidden, 77 cl::desc("The size of the native vector registers")); 78 79 static cl::opt<unsigned> 80 MaxIter("bb-vectorize-max-iter", cl::init(0), cl::Hidden, 81 cl::desc("The maximum number of pairing iterations")); 82 83 static cl::opt<bool> 84 Pow2LenOnly("bb-vectorize-pow2-len-only", cl::init(false), cl::Hidden, 85 cl::desc("Don't try to form non-2^n-length vectors")); 86 87 static cl::opt<unsigned> 88 MaxInsts("bb-vectorize-max-instr-per-group", cl::init(500), cl::Hidden, 89 cl::desc("The maximum number of pairable instructions per group")); 90 91 static cl::opt<unsigned> 92 MaxPairs("bb-vectorize-max-pairs-per-group", cl::init(3000), cl::Hidden, 93 cl::desc("The maximum number of candidate instruction pairs per group")); 94 95 static cl::opt<unsigned> 96 MaxCandPairsForCycleCheck("bb-vectorize-max-cycle-check-pairs", cl::init(200), 97 cl::Hidden, cl::desc("The maximum number of candidate pairs with which to use" 98 " a full cycle check")); 99 100 static cl::opt<bool> 101 NoBools("bb-vectorize-no-bools", cl::init(false), cl::Hidden, 102 cl::desc("Don't try to vectorize boolean (i1) values")); 103 104 static cl::opt<bool> 105 NoInts("bb-vectorize-no-ints", cl::init(false), cl::Hidden, 106 cl::desc("Don't try to vectorize integer values")); 107 108 static cl::opt<bool> 109 NoFloats("bb-vectorize-no-floats", cl::init(false), cl::Hidden, 110 cl::desc("Don't try to vectorize floating-point values")); 111 112 // FIXME: This should default to false once pointer vector support works. 113 static cl::opt<bool> 114 NoPointers("bb-vectorize-no-pointers", cl::init(/*false*/ true), cl::Hidden, 115 cl::desc("Don't try to vectorize pointer values")); 116 117 static cl::opt<bool> 118 NoCasts("bb-vectorize-no-casts", cl::init(false), cl::Hidden, 119 cl::desc("Don't try to vectorize casting (conversion) operations")); 120 121 static cl::opt<bool> 122 NoMath("bb-vectorize-no-math", cl::init(false), cl::Hidden, 123 cl::desc("Don't try to vectorize floating-point math intrinsics")); 124 125 static cl::opt<bool> 126 NoBitManipulation("bb-vectorize-no-bitmanip", cl::init(false), cl::Hidden, 127 cl::desc("Don't try to vectorize BitManipulation intrinsics")); 128 129 static cl::opt<bool> 130 NoFMA("bb-vectorize-no-fma", cl::init(false), cl::Hidden, 131 cl::desc("Don't try to vectorize the fused-multiply-add intrinsic")); 132 133 static cl::opt<bool> 134 NoSelect("bb-vectorize-no-select", cl::init(false), cl::Hidden, 135 cl::desc("Don't try to vectorize select instructions")); 136 137 static cl::opt<bool> 138 NoCmp("bb-vectorize-no-cmp", cl::init(false), cl::Hidden, 139 cl::desc("Don't try to vectorize comparison instructions")); 140 141 static cl::opt<bool> 142 NoGEP("bb-vectorize-no-gep", cl::init(false), cl::Hidden, 143 cl::desc("Don't try to vectorize getelementptr instructions")); 144 145 static cl::opt<bool> 146 NoMemOps("bb-vectorize-no-mem-ops", cl::init(false), cl::Hidden, 147 cl::desc("Don't try to vectorize loads and stores")); 148 149 static cl::opt<bool> 150 AlignedOnly("bb-vectorize-aligned-only", cl::init(false), cl::Hidden, 151 cl::desc("Only generate aligned loads and stores")); 152 153 static cl::opt<bool> 154 NoMemOpBoost("bb-vectorize-no-mem-op-boost", 155 cl::init(false), cl::Hidden, 156 cl::desc("Don't boost the chain-depth contribution of loads and stores")); 157 158 static cl::opt<bool> 159 FastDep("bb-vectorize-fast-dep", cl::init(false), cl::Hidden, 160 cl::desc("Use a fast instruction dependency analysis")); 161 162 #ifndef NDEBUG 163 static cl::opt<bool> 164 DebugInstructionExamination("bb-vectorize-debug-instruction-examination", 165 cl::init(false), cl::Hidden, 166 cl::desc("When debugging is enabled, output information on the" 167 " instruction-examination process")); 168 static cl::opt<bool> 169 DebugCandidateSelection("bb-vectorize-debug-candidate-selection", 170 cl::init(false), cl::Hidden, 171 cl::desc("When debugging is enabled, output information on the" 172 " candidate-selection process")); 173 static cl::opt<bool> 174 DebugPairSelection("bb-vectorize-debug-pair-selection", 175 cl::init(false), cl::Hidden, 176 cl::desc("When debugging is enabled, output information on the" 177 " pair-selection process")); 178 static cl::opt<bool> 179 DebugCycleCheck("bb-vectorize-debug-cycle-check", 180 cl::init(false), cl::Hidden, 181 cl::desc("When debugging is enabled, output information on the" 182 " cycle-checking process")); 183 184 static cl::opt<bool> 185 PrintAfterEveryPair("bb-vectorize-debug-print-after-every-pair", 186 cl::init(false), cl::Hidden, 187 cl::desc("When debugging is enabled, dump the basic block after" 188 " every pair is fused")); 189 #endif 190 191 STATISTIC(NumFusedOps, "Number of operations fused by bb-vectorize"); 192 193 namespace { 194 struct BBVectorize : public BasicBlockPass { 195 static char ID; // Pass identification, replacement for typeid 196 197 const VectorizeConfig Config; 198 199 BBVectorize(const VectorizeConfig &C = VectorizeConfig()) 200 : BasicBlockPass(ID), Config(C) { 201 initializeBBVectorizePass(*PassRegistry::getPassRegistry()); 202 } 203 204 BBVectorize(Pass *P, const VectorizeConfig &C) 205 : BasicBlockPass(ID), Config(C) { 206 AA = &P->getAnalysis<AliasAnalysis>(); 207 DT = &P->getAnalysis<DominatorTreeWrapperPass>().getDomTree(); 208 SE = &P->getAnalysis<ScalarEvolution>(); 209 DataLayoutPass *DLP = P->getAnalysisIfAvailable<DataLayoutPass>(); 210 DL = DLP ? &DLP->getDataLayout() : nullptr; 211 TTI = IgnoreTargetInfo ? nullptr : &P->getAnalysis<TargetTransformInfo>(); 212 } 213 214 typedef std::pair<Value *, Value *> ValuePair; 215 typedef std::pair<ValuePair, int> ValuePairWithCost; 216 typedef std::pair<ValuePair, size_t> ValuePairWithDepth; 217 typedef std::pair<ValuePair, ValuePair> VPPair; // A ValuePair pair 218 typedef std::pair<VPPair, unsigned> VPPairWithType; 219 220 AliasAnalysis *AA; 221 DominatorTree *DT; 222 ScalarEvolution *SE; 223 const DataLayout *DL; 224 const TargetTransformInfo *TTI; 225 226 // FIXME: const correct? 227 228 bool vectorizePairs(BasicBlock &BB, bool NonPow2Len = false); 229 230 bool getCandidatePairs(BasicBlock &BB, 231 BasicBlock::iterator &Start, 232 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 233 DenseSet<ValuePair> &FixedOrderPairs, 234 DenseMap<ValuePair, int> &CandidatePairCostSavings, 235 std::vector<Value *> &PairableInsts, bool NonPow2Len); 236 237 // FIXME: The current implementation does not account for pairs that 238 // are connected in multiple ways. For example: 239 // C1 = A1 / A2; C2 = A2 / A1 (which may be both direct and a swap) 240 enum PairConnectionType { 241 PairConnectionDirect, 242 PairConnectionSwap, 243 PairConnectionSplat 244 }; 245 246 void computeConnectedPairs( 247 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 248 DenseSet<ValuePair> &CandidatePairsSet, 249 std::vector<Value *> &PairableInsts, 250 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 251 DenseMap<VPPair, unsigned> &PairConnectionTypes); 252 253 void buildDepMap(BasicBlock &BB, 254 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 255 std::vector<Value *> &PairableInsts, 256 DenseSet<ValuePair> &PairableInstUsers); 257 258 void choosePairs(DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 259 DenseSet<ValuePair> &CandidatePairsSet, 260 DenseMap<ValuePair, int> &CandidatePairCostSavings, 261 std::vector<Value *> &PairableInsts, 262 DenseSet<ValuePair> &FixedOrderPairs, 263 DenseMap<VPPair, unsigned> &PairConnectionTypes, 264 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 265 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps, 266 DenseSet<ValuePair> &PairableInstUsers, 267 DenseMap<Value *, Value *>& ChosenPairs); 268 269 void fuseChosenPairs(BasicBlock &BB, 270 std::vector<Value *> &PairableInsts, 271 DenseMap<Value *, Value *>& ChosenPairs, 272 DenseSet<ValuePair> &FixedOrderPairs, 273 DenseMap<VPPair, unsigned> &PairConnectionTypes, 274 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 275 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps); 276 277 278 bool isInstVectorizable(Instruction *I, bool &IsSimpleLoadStore); 279 280 bool areInstsCompatible(Instruction *I, Instruction *J, 281 bool IsSimpleLoadStore, bool NonPow2Len, 282 int &CostSavings, int &FixedOrder); 283 284 bool trackUsesOfI(DenseSet<Value *> &Users, 285 AliasSetTracker &WriteSet, Instruction *I, 286 Instruction *J, bool UpdateUsers = true, 287 DenseSet<ValuePair> *LoadMoveSetPairs = nullptr); 288 289 void computePairsConnectedTo( 290 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 291 DenseSet<ValuePair> &CandidatePairsSet, 292 std::vector<Value *> &PairableInsts, 293 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 294 DenseMap<VPPair, unsigned> &PairConnectionTypes, 295 ValuePair P); 296 297 bool pairsConflict(ValuePair P, ValuePair Q, 298 DenseSet<ValuePair> &PairableInstUsers, 299 DenseMap<ValuePair, std::vector<ValuePair> > 300 *PairableInstUserMap = nullptr, 301 DenseSet<VPPair> *PairableInstUserPairSet = nullptr); 302 303 bool pairWillFormCycle(ValuePair P, 304 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUsers, 305 DenseSet<ValuePair> &CurrentPairs); 306 307 void pruneDAGFor( 308 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 309 std::vector<Value *> &PairableInsts, 310 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 311 DenseSet<ValuePair> &PairableInstUsers, 312 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUserMap, 313 DenseSet<VPPair> &PairableInstUserPairSet, 314 DenseMap<Value *, Value *> &ChosenPairs, 315 DenseMap<ValuePair, size_t> &DAG, 316 DenseSet<ValuePair> &PrunedDAG, ValuePair J, 317 bool UseCycleCheck); 318 319 void buildInitialDAGFor( 320 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 321 DenseSet<ValuePair> &CandidatePairsSet, 322 std::vector<Value *> &PairableInsts, 323 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 324 DenseSet<ValuePair> &PairableInstUsers, 325 DenseMap<Value *, Value *> &ChosenPairs, 326 DenseMap<ValuePair, size_t> &DAG, ValuePair J); 327 328 void findBestDAGFor( 329 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 330 DenseSet<ValuePair> &CandidatePairsSet, 331 DenseMap<ValuePair, int> &CandidatePairCostSavings, 332 std::vector<Value *> &PairableInsts, 333 DenseSet<ValuePair> &FixedOrderPairs, 334 DenseMap<VPPair, unsigned> &PairConnectionTypes, 335 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 336 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps, 337 DenseSet<ValuePair> &PairableInstUsers, 338 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUserMap, 339 DenseSet<VPPair> &PairableInstUserPairSet, 340 DenseMap<Value *, Value *> &ChosenPairs, 341 DenseSet<ValuePair> &BestDAG, size_t &BestMaxDepth, 342 int &BestEffSize, Value *II, std::vector<Value *>&JJ, 343 bool UseCycleCheck); 344 345 Value *getReplacementPointerInput(LLVMContext& Context, Instruction *I, 346 Instruction *J, unsigned o); 347 348 void fillNewShuffleMask(LLVMContext& Context, Instruction *J, 349 unsigned MaskOffset, unsigned NumInElem, 350 unsigned NumInElem1, unsigned IdxOffset, 351 std::vector<Constant*> &Mask); 352 353 Value *getReplacementShuffleMask(LLVMContext& Context, Instruction *I, 354 Instruction *J); 355 356 bool expandIEChain(LLVMContext& Context, Instruction *I, Instruction *J, 357 unsigned o, Value *&LOp, unsigned numElemL, 358 Type *ArgTypeL, Type *ArgTypeR, bool IBeforeJ, 359 unsigned IdxOff = 0); 360 361 Value *getReplacementInput(LLVMContext& Context, Instruction *I, 362 Instruction *J, unsigned o, bool IBeforeJ); 363 364 void getReplacementInputsForPair(LLVMContext& Context, Instruction *I, 365 Instruction *J, SmallVectorImpl<Value *> &ReplacedOperands, 366 bool IBeforeJ); 367 368 void replaceOutputsOfPair(LLVMContext& Context, Instruction *I, 369 Instruction *J, Instruction *K, 370 Instruction *&InsertionPt, Instruction *&K1, 371 Instruction *&K2); 372 373 void collectPairLoadMoveSet(BasicBlock &BB, 374 DenseMap<Value *, Value *> &ChosenPairs, 375 DenseMap<Value *, std::vector<Value *> > &LoadMoveSet, 376 DenseSet<ValuePair> &LoadMoveSetPairs, 377 Instruction *I); 378 379 void collectLoadMoveSet(BasicBlock &BB, 380 std::vector<Value *> &PairableInsts, 381 DenseMap<Value *, Value *> &ChosenPairs, 382 DenseMap<Value *, std::vector<Value *> > &LoadMoveSet, 383 DenseSet<ValuePair> &LoadMoveSetPairs); 384 385 bool canMoveUsesOfIAfterJ(BasicBlock &BB, 386 DenseSet<ValuePair> &LoadMoveSetPairs, 387 Instruction *I, Instruction *J); 388 389 void moveUsesOfIAfterJ(BasicBlock &BB, 390 DenseSet<ValuePair> &LoadMoveSetPairs, 391 Instruction *&InsertionPt, 392 Instruction *I, Instruction *J); 393 394 void combineMetadata(Instruction *K, const Instruction *J); 395 396 bool vectorizeBB(BasicBlock &BB) { 397 if (skipOptnoneFunction(BB)) 398 return false; 399 if (!DT->isReachableFromEntry(&BB)) { 400 DEBUG(dbgs() << "BBV: skipping unreachable " << BB.getName() << 401 " in " << BB.getParent()->getName() << "\n"); 402 return false; 403 } 404 405 DEBUG(if (TTI) dbgs() << "BBV: using target information\n"); 406 407 bool changed = false; 408 // Iterate a sufficient number of times to merge types of size 1 bit, 409 // then 2 bits, then 4, etc. up to half of the target vector width of the 410 // target vector register. 411 unsigned n = 1; 412 for (unsigned v = 2; 413 (TTI || v <= Config.VectorBits) && 414 (!Config.MaxIter || n <= Config.MaxIter); 415 v *= 2, ++n) { 416 DEBUG(dbgs() << "BBV: fusing loop #" << n << 417 " for " << BB.getName() << " in " << 418 BB.getParent()->getName() << "...\n"); 419 if (vectorizePairs(BB)) 420 changed = true; 421 else 422 break; 423 } 424 425 if (changed && !Pow2LenOnly) { 426 ++n; 427 for (; !Config.MaxIter || n <= Config.MaxIter; ++n) { 428 DEBUG(dbgs() << "BBV: fusing for non-2^n-length vectors loop #: " << 429 n << " for " << BB.getName() << " in " << 430 BB.getParent()->getName() << "...\n"); 431 if (!vectorizePairs(BB, true)) break; 432 } 433 } 434 435 DEBUG(dbgs() << "BBV: done!\n"); 436 return changed; 437 } 438 439 bool runOnBasicBlock(BasicBlock &BB) override { 440 // OptimizeNone check deferred to vectorizeBB(). 441 442 AA = &getAnalysis<AliasAnalysis>(); 443 DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree(); 444 SE = &getAnalysis<ScalarEvolution>(); 445 DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>(); 446 DL = DLP ? &DLP->getDataLayout() : nullptr; 447 TTI = IgnoreTargetInfo ? nullptr : &getAnalysis<TargetTransformInfo>(); 448 449 return vectorizeBB(BB); 450 } 451 452 void getAnalysisUsage(AnalysisUsage &AU) const override { 453 BasicBlockPass::getAnalysisUsage(AU); 454 AU.addRequired<AliasAnalysis>(); 455 AU.addRequired<DominatorTreeWrapperPass>(); 456 AU.addRequired<ScalarEvolution>(); 457 AU.addRequired<TargetTransformInfo>(); 458 AU.addPreserved<AliasAnalysis>(); 459 AU.addPreserved<DominatorTreeWrapperPass>(); 460 AU.addPreserved<ScalarEvolution>(); 461 AU.setPreservesCFG(); 462 } 463 464 static inline VectorType *getVecTypeForPair(Type *ElemTy, Type *Elem2Ty) { 465 assert(ElemTy->getScalarType() == Elem2Ty->getScalarType() && 466 "Cannot form vector from incompatible scalar types"); 467 Type *STy = ElemTy->getScalarType(); 468 469 unsigned numElem; 470 if (VectorType *VTy = dyn_cast<VectorType>(ElemTy)) { 471 numElem = VTy->getNumElements(); 472 } else { 473 numElem = 1; 474 } 475 476 if (VectorType *VTy = dyn_cast<VectorType>(Elem2Ty)) { 477 numElem += VTy->getNumElements(); 478 } else { 479 numElem += 1; 480 } 481 482 return VectorType::get(STy, numElem); 483 } 484 485 static inline void getInstructionTypes(Instruction *I, 486 Type *&T1, Type *&T2) { 487 if (StoreInst *SI = dyn_cast<StoreInst>(I)) { 488 // For stores, it is the value type, not the pointer type that matters 489 // because the value is what will come from a vector register. 490 491 Value *IVal = SI->getValueOperand(); 492 T1 = IVal->getType(); 493 } else { 494 T1 = I->getType(); 495 } 496 497 if (CastInst *CI = dyn_cast<CastInst>(I)) 498 T2 = CI->getSrcTy(); 499 else 500 T2 = T1; 501 502 if (SelectInst *SI = dyn_cast<SelectInst>(I)) { 503 T2 = SI->getCondition()->getType(); 504 } else if (ShuffleVectorInst *SI = dyn_cast<ShuffleVectorInst>(I)) { 505 T2 = SI->getOperand(0)->getType(); 506 } else if (CmpInst *CI = dyn_cast<CmpInst>(I)) { 507 T2 = CI->getOperand(0)->getType(); 508 } 509 } 510 511 // Returns the weight associated with the provided value. A chain of 512 // candidate pairs has a length given by the sum of the weights of its 513 // members (one weight per pair; the weight of each member of the pair 514 // is assumed to be the same). This length is then compared to the 515 // chain-length threshold to determine if a given chain is significant 516 // enough to be vectorized. The length is also used in comparing 517 // candidate chains where longer chains are considered to be better. 518 // Note: when this function returns 0, the resulting instructions are 519 // not actually fused. 520 inline size_t getDepthFactor(Value *V) { 521 // InsertElement and ExtractElement have a depth factor of zero. This is 522 // for two reasons: First, they cannot be usefully fused. Second, because 523 // the pass generates a lot of these, they can confuse the simple metric 524 // used to compare the dags in the next iteration. Thus, giving them a 525 // weight of zero allows the pass to essentially ignore them in 526 // subsequent iterations when looking for vectorization opportunities 527 // while still tracking dependency chains that flow through those 528 // instructions. 529 if (isa<InsertElementInst>(V) || isa<ExtractElementInst>(V)) 530 return 0; 531 532 // Give a load or store half of the required depth so that load/store 533 // pairs will vectorize. 534 if (!Config.NoMemOpBoost && (isa<LoadInst>(V) || isa<StoreInst>(V))) 535 return Config.ReqChainDepth/2; 536 537 return 1; 538 } 539 540 // Returns the cost of the provided instruction using TTI. 541 // This does not handle loads and stores. 542 unsigned getInstrCost(unsigned Opcode, Type *T1, Type *T2, 543 TargetTransformInfo::OperandValueKind Op1VK = 544 TargetTransformInfo::OK_AnyValue, 545 TargetTransformInfo::OperandValueKind Op2VK = 546 TargetTransformInfo::OK_AnyValue) { 547 switch (Opcode) { 548 default: break; 549 case Instruction::GetElementPtr: 550 // We mark this instruction as zero-cost because scalar GEPs are usually 551 // lowered to the instruction addressing mode. At the moment we don't 552 // generate vector GEPs. 553 return 0; 554 case Instruction::Br: 555 return TTI->getCFInstrCost(Opcode); 556 case Instruction::PHI: 557 return 0; 558 case Instruction::Add: 559 case Instruction::FAdd: 560 case Instruction::Sub: 561 case Instruction::FSub: 562 case Instruction::Mul: 563 case Instruction::FMul: 564 case Instruction::UDiv: 565 case Instruction::SDiv: 566 case Instruction::FDiv: 567 case Instruction::URem: 568 case Instruction::SRem: 569 case Instruction::FRem: 570 case Instruction::Shl: 571 case Instruction::LShr: 572 case Instruction::AShr: 573 case Instruction::And: 574 case Instruction::Or: 575 case Instruction::Xor: 576 return TTI->getArithmeticInstrCost(Opcode, T1, Op1VK, Op2VK); 577 case Instruction::Select: 578 case Instruction::ICmp: 579 case Instruction::FCmp: 580 return TTI->getCmpSelInstrCost(Opcode, T1, T2); 581 case Instruction::ZExt: 582 case Instruction::SExt: 583 case Instruction::FPToUI: 584 case Instruction::FPToSI: 585 case Instruction::FPExt: 586 case Instruction::PtrToInt: 587 case Instruction::IntToPtr: 588 case Instruction::SIToFP: 589 case Instruction::UIToFP: 590 case Instruction::Trunc: 591 case Instruction::FPTrunc: 592 case Instruction::BitCast: 593 case Instruction::ShuffleVector: 594 return TTI->getCastInstrCost(Opcode, T1, T2); 595 } 596 597 return 1; 598 } 599 600 // This determines the relative offset of two loads or stores, returning 601 // true if the offset could be determined to be some constant value. 602 // For example, if OffsetInElmts == 1, then J accesses the memory directly 603 // after I; if OffsetInElmts == -1 then I accesses the memory 604 // directly after J. 605 bool getPairPtrInfo(Instruction *I, Instruction *J, 606 Value *&IPtr, Value *&JPtr, unsigned &IAlignment, unsigned &JAlignment, 607 unsigned &IAddressSpace, unsigned &JAddressSpace, 608 int64_t &OffsetInElmts, bool ComputeOffset = true) { 609 OffsetInElmts = 0; 610 if (LoadInst *LI = dyn_cast<LoadInst>(I)) { 611 LoadInst *LJ = cast<LoadInst>(J); 612 IPtr = LI->getPointerOperand(); 613 JPtr = LJ->getPointerOperand(); 614 IAlignment = LI->getAlignment(); 615 JAlignment = LJ->getAlignment(); 616 IAddressSpace = LI->getPointerAddressSpace(); 617 JAddressSpace = LJ->getPointerAddressSpace(); 618 } else { 619 StoreInst *SI = cast<StoreInst>(I), *SJ = cast<StoreInst>(J); 620 IPtr = SI->getPointerOperand(); 621 JPtr = SJ->getPointerOperand(); 622 IAlignment = SI->getAlignment(); 623 JAlignment = SJ->getAlignment(); 624 IAddressSpace = SI->getPointerAddressSpace(); 625 JAddressSpace = SJ->getPointerAddressSpace(); 626 } 627 628 if (!ComputeOffset) 629 return true; 630 631 const SCEV *IPtrSCEV = SE->getSCEV(IPtr); 632 const SCEV *JPtrSCEV = SE->getSCEV(JPtr); 633 634 // If this is a trivial offset, then we'll get something like 635 // 1*sizeof(type). With target data, which we need anyway, this will get 636 // constant folded into a number. 637 const SCEV *OffsetSCEV = SE->getMinusSCEV(JPtrSCEV, IPtrSCEV); 638 if (const SCEVConstant *ConstOffSCEV = 639 dyn_cast<SCEVConstant>(OffsetSCEV)) { 640 ConstantInt *IntOff = ConstOffSCEV->getValue(); 641 int64_t Offset = IntOff->getSExtValue(); 642 643 Type *VTy = IPtr->getType()->getPointerElementType(); 644 int64_t VTyTSS = (int64_t) DL->getTypeStoreSize(VTy); 645 646 Type *VTy2 = JPtr->getType()->getPointerElementType(); 647 if (VTy != VTy2 && Offset < 0) { 648 int64_t VTy2TSS = (int64_t) DL->getTypeStoreSize(VTy2); 649 OffsetInElmts = Offset/VTy2TSS; 650 return (abs64(Offset) % VTy2TSS) == 0; 651 } 652 653 OffsetInElmts = Offset/VTyTSS; 654 return (abs64(Offset) % VTyTSS) == 0; 655 } 656 657 return false; 658 } 659 660 // Returns true if the provided CallInst represents an intrinsic that can 661 // be vectorized. 662 bool isVectorizableIntrinsic(CallInst* I) { 663 Function *F = I->getCalledFunction(); 664 if (!F) return false; 665 666 Intrinsic::ID IID = (Intrinsic::ID) F->getIntrinsicID(); 667 if (!IID) return false; 668 669 switch(IID) { 670 default: 671 return false; 672 case Intrinsic::sqrt: 673 case Intrinsic::powi: 674 case Intrinsic::sin: 675 case Intrinsic::cos: 676 case Intrinsic::log: 677 case Intrinsic::log2: 678 case Intrinsic::log10: 679 case Intrinsic::exp: 680 case Intrinsic::exp2: 681 case Intrinsic::pow: 682 case Intrinsic::round: 683 case Intrinsic::copysign: 684 case Intrinsic::ceil: 685 case Intrinsic::nearbyint: 686 case Intrinsic::rint: 687 case Intrinsic::trunc: 688 case Intrinsic::floor: 689 case Intrinsic::fabs: 690 return Config.VectorizeMath; 691 case Intrinsic::bswap: 692 case Intrinsic::ctpop: 693 case Intrinsic::ctlz: 694 case Intrinsic::cttz: 695 return Config.VectorizeBitManipulations; 696 case Intrinsic::fma: 697 case Intrinsic::fmuladd: 698 return Config.VectorizeFMA; 699 } 700 } 701 702 bool isPureIEChain(InsertElementInst *IE) { 703 InsertElementInst *IENext = IE; 704 do { 705 if (!isa<UndefValue>(IENext->getOperand(0)) && 706 !isa<InsertElementInst>(IENext->getOperand(0))) { 707 return false; 708 } 709 } while ((IENext = 710 dyn_cast<InsertElementInst>(IENext->getOperand(0)))); 711 712 return true; 713 } 714 }; 715 716 // This function implements one vectorization iteration on the provided 717 // basic block. It returns true if the block is changed. 718 bool BBVectorize::vectorizePairs(BasicBlock &BB, bool NonPow2Len) { 719 bool ShouldContinue; 720 BasicBlock::iterator Start = BB.getFirstInsertionPt(); 721 722 std::vector<Value *> AllPairableInsts; 723 DenseMap<Value *, Value *> AllChosenPairs; 724 DenseSet<ValuePair> AllFixedOrderPairs; 725 DenseMap<VPPair, unsigned> AllPairConnectionTypes; 726 DenseMap<ValuePair, std::vector<ValuePair> > AllConnectedPairs, 727 AllConnectedPairDeps; 728 729 do { 730 std::vector<Value *> PairableInsts; 731 DenseMap<Value *, std::vector<Value *> > CandidatePairs; 732 DenseSet<ValuePair> FixedOrderPairs; 733 DenseMap<ValuePair, int> CandidatePairCostSavings; 734 ShouldContinue = getCandidatePairs(BB, Start, CandidatePairs, 735 FixedOrderPairs, 736 CandidatePairCostSavings, 737 PairableInsts, NonPow2Len); 738 if (PairableInsts.empty()) continue; 739 740 // Build the candidate pair set for faster lookups. 741 DenseSet<ValuePair> CandidatePairsSet; 742 for (DenseMap<Value *, std::vector<Value *> >::iterator I = 743 CandidatePairs.begin(), E = CandidatePairs.end(); I != E; ++I) 744 for (std::vector<Value *>::iterator J = I->second.begin(), 745 JE = I->second.end(); J != JE; ++J) 746 CandidatePairsSet.insert(ValuePair(I->first, *J)); 747 748 // Now we have a map of all of the pairable instructions and we need to 749 // select the best possible pairing. A good pairing is one such that the 750 // users of the pair are also paired. This defines a (directed) forest 751 // over the pairs such that two pairs are connected iff the second pair 752 // uses the first. 753 754 // Note that it only matters that both members of the second pair use some 755 // element of the first pair (to allow for splatting). 756 757 DenseMap<ValuePair, std::vector<ValuePair> > ConnectedPairs, 758 ConnectedPairDeps; 759 DenseMap<VPPair, unsigned> PairConnectionTypes; 760 computeConnectedPairs(CandidatePairs, CandidatePairsSet, 761 PairableInsts, ConnectedPairs, PairConnectionTypes); 762 if (ConnectedPairs.empty()) continue; 763 764 for (DenseMap<ValuePair, std::vector<ValuePair> >::iterator 765 I = ConnectedPairs.begin(), IE = ConnectedPairs.end(); 766 I != IE; ++I) 767 for (std::vector<ValuePair>::iterator J = I->second.begin(), 768 JE = I->second.end(); J != JE; ++J) 769 ConnectedPairDeps[*J].push_back(I->first); 770 771 // Build the pairable-instruction dependency map 772 DenseSet<ValuePair> PairableInstUsers; 773 buildDepMap(BB, CandidatePairs, PairableInsts, PairableInstUsers); 774 775 // There is now a graph of the connected pairs. For each variable, pick 776 // the pairing with the largest dag meeting the depth requirement on at 777 // least one branch. Then select all pairings that are part of that dag 778 // and remove them from the list of available pairings and pairable 779 // variables. 780 781 DenseMap<Value *, Value *> ChosenPairs; 782 choosePairs(CandidatePairs, CandidatePairsSet, 783 CandidatePairCostSavings, 784 PairableInsts, FixedOrderPairs, PairConnectionTypes, 785 ConnectedPairs, ConnectedPairDeps, 786 PairableInstUsers, ChosenPairs); 787 788 if (ChosenPairs.empty()) continue; 789 AllPairableInsts.insert(AllPairableInsts.end(), PairableInsts.begin(), 790 PairableInsts.end()); 791 AllChosenPairs.insert(ChosenPairs.begin(), ChosenPairs.end()); 792 793 // Only for the chosen pairs, propagate information on fixed-order pairs, 794 // pair connections, and their types to the data structures used by the 795 // pair fusion procedures. 796 for (DenseMap<Value *, Value *>::iterator I = ChosenPairs.begin(), 797 IE = ChosenPairs.end(); I != IE; ++I) { 798 if (FixedOrderPairs.count(*I)) 799 AllFixedOrderPairs.insert(*I); 800 else if (FixedOrderPairs.count(ValuePair(I->second, I->first))) 801 AllFixedOrderPairs.insert(ValuePair(I->second, I->first)); 802 803 for (DenseMap<Value *, Value *>::iterator J = ChosenPairs.begin(); 804 J != IE; ++J) { 805 DenseMap<VPPair, unsigned>::iterator K = 806 PairConnectionTypes.find(VPPair(*I, *J)); 807 if (K != PairConnectionTypes.end()) { 808 AllPairConnectionTypes.insert(*K); 809 } else { 810 K = PairConnectionTypes.find(VPPair(*J, *I)); 811 if (K != PairConnectionTypes.end()) 812 AllPairConnectionTypes.insert(*K); 813 } 814 } 815 } 816 817 for (DenseMap<ValuePair, std::vector<ValuePair> >::iterator 818 I = ConnectedPairs.begin(), IE = ConnectedPairs.end(); 819 I != IE; ++I) 820 for (std::vector<ValuePair>::iterator J = I->second.begin(), 821 JE = I->second.end(); J != JE; ++J) 822 if (AllPairConnectionTypes.count(VPPair(I->first, *J))) { 823 AllConnectedPairs[I->first].push_back(*J); 824 AllConnectedPairDeps[*J].push_back(I->first); 825 } 826 } while (ShouldContinue); 827 828 if (AllChosenPairs.empty()) return false; 829 NumFusedOps += AllChosenPairs.size(); 830 831 // A set of pairs has now been selected. It is now necessary to replace the 832 // paired instructions with vector instructions. For this procedure each 833 // operand must be replaced with a vector operand. This vector is formed 834 // by using build_vector on the old operands. The replaced values are then 835 // replaced with a vector_extract on the result. Subsequent optimization 836 // passes should coalesce the build/extract combinations. 837 838 fuseChosenPairs(BB, AllPairableInsts, AllChosenPairs, AllFixedOrderPairs, 839 AllPairConnectionTypes, 840 AllConnectedPairs, AllConnectedPairDeps); 841 842 // It is important to cleanup here so that future iterations of this 843 // function have less work to do. 844 (void) SimplifyInstructionsInBlock(&BB, DL, AA->getTargetLibraryInfo()); 845 return true; 846 } 847 848 // This function returns true if the provided instruction is capable of being 849 // fused into a vector instruction. This determination is based only on the 850 // type and other attributes of the instruction. 851 bool BBVectorize::isInstVectorizable(Instruction *I, 852 bool &IsSimpleLoadStore) { 853 IsSimpleLoadStore = false; 854 855 if (CallInst *C = dyn_cast<CallInst>(I)) { 856 if (!isVectorizableIntrinsic(C)) 857 return false; 858 } else if (LoadInst *L = dyn_cast<LoadInst>(I)) { 859 // Vectorize simple loads if possbile: 860 IsSimpleLoadStore = L->isSimple(); 861 if (!IsSimpleLoadStore || !Config.VectorizeMemOps) 862 return false; 863 } else if (StoreInst *S = dyn_cast<StoreInst>(I)) { 864 // Vectorize simple stores if possbile: 865 IsSimpleLoadStore = S->isSimple(); 866 if (!IsSimpleLoadStore || !Config.VectorizeMemOps) 867 return false; 868 } else if (CastInst *C = dyn_cast<CastInst>(I)) { 869 // We can vectorize casts, but not casts of pointer types, etc. 870 if (!Config.VectorizeCasts) 871 return false; 872 873 Type *SrcTy = C->getSrcTy(); 874 if (!SrcTy->isSingleValueType()) 875 return false; 876 877 Type *DestTy = C->getDestTy(); 878 if (!DestTy->isSingleValueType()) 879 return false; 880 } else if (isa<SelectInst>(I)) { 881 if (!Config.VectorizeSelect) 882 return false; 883 } else if (isa<CmpInst>(I)) { 884 if (!Config.VectorizeCmp) 885 return false; 886 } else if (GetElementPtrInst *G = dyn_cast<GetElementPtrInst>(I)) { 887 if (!Config.VectorizeGEP) 888 return false; 889 890 // Currently, vector GEPs exist only with one index. 891 if (G->getNumIndices() != 1) 892 return false; 893 } else if (!(I->isBinaryOp() || isa<ShuffleVectorInst>(I) || 894 isa<ExtractElementInst>(I) || isa<InsertElementInst>(I))) { 895 return false; 896 } 897 898 // We can't vectorize memory operations without target data 899 if (!DL && IsSimpleLoadStore) 900 return false; 901 902 Type *T1, *T2; 903 getInstructionTypes(I, T1, T2); 904 905 // Not every type can be vectorized... 906 if (!(VectorType::isValidElementType(T1) || T1->isVectorTy()) || 907 !(VectorType::isValidElementType(T2) || T2->isVectorTy())) 908 return false; 909 910 if (T1->getScalarSizeInBits() == 1) { 911 if (!Config.VectorizeBools) 912 return false; 913 } else { 914 if (!Config.VectorizeInts && T1->isIntOrIntVectorTy()) 915 return false; 916 } 917 918 if (T2->getScalarSizeInBits() == 1) { 919 if (!Config.VectorizeBools) 920 return false; 921 } else { 922 if (!Config.VectorizeInts && T2->isIntOrIntVectorTy()) 923 return false; 924 } 925 926 if (!Config.VectorizeFloats 927 && (T1->isFPOrFPVectorTy() || T2->isFPOrFPVectorTy())) 928 return false; 929 930 // Don't vectorize target-specific types. 931 if (T1->isX86_FP80Ty() || T1->isPPC_FP128Ty() || T1->isX86_MMXTy()) 932 return false; 933 if (T2->isX86_FP80Ty() || T2->isPPC_FP128Ty() || T2->isX86_MMXTy()) 934 return false; 935 936 if ((!Config.VectorizePointers || !DL) && 937 (T1->getScalarType()->isPointerTy() || 938 T2->getScalarType()->isPointerTy())) 939 return false; 940 941 if (!TTI && (T1->getPrimitiveSizeInBits() >= Config.VectorBits || 942 T2->getPrimitiveSizeInBits() >= Config.VectorBits)) 943 return false; 944 945 return true; 946 } 947 948 // This function returns true if the two provided instructions are compatible 949 // (meaning that they can be fused into a vector instruction). This assumes 950 // that I has already been determined to be vectorizable and that J is not 951 // in the use dag of I. 952 bool BBVectorize::areInstsCompatible(Instruction *I, Instruction *J, 953 bool IsSimpleLoadStore, bool NonPow2Len, 954 int &CostSavings, int &FixedOrder) { 955 DEBUG(if (DebugInstructionExamination) dbgs() << "BBV: looking at " << *I << 956 " <-> " << *J << "\n"); 957 958 CostSavings = 0; 959 FixedOrder = 0; 960 961 // Loads and stores can be merged if they have different alignments, 962 // but are otherwise the same. 963 if (!J->isSameOperationAs(I, Instruction::CompareIgnoringAlignment | 964 (NonPow2Len ? Instruction::CompareUsingScalarTypes : 0))) 965 return false; 966 967 Type *IT1, *IT2, *JT1, *JT2; 968 getInstructionTypes(I, IT1, IT2); 969 getInstructionTypes(J, JT1, JT2); 970 unsigned MaxTypeBits = std::max( 971 IT1->getPrimitiveSizeInBits() + JT1->getPrimitiveSizeInBits(), 972 IT2->getPrimitiveSizeInBits() + JT2->getPrimitiveSizeInBits()); 973 if (!TTI && MaxTypeBits > Config.VectorBits) 974 return false; 975 976 // FIXME: handle addsub-type operations! 977 978 if (IsSimpleLoadStore) { 979 Value *IPtr, *JPtr; 980 unsigned IAlignment, JAlignment, IAddressSpace, JAddressSpace; 981 int64_t OffsetInElmts = 0; 982 if (getPairPtrInfo(I, J, IPtr, JPtr, IAlignment, JAlignment, 983 IAddressSpace, JAddressSpace, 984 OffsetInElmts) && abs64(OffsetInElmts) == 1) { 985 FixedOrder = (int) OffsetInElmts; 986 unsigned BottomAlignment = IAlignment; 987 if (OffsetInElmts < 0) BottomAlignment = JAlignment; 988 989 Type *aTypeI = isa<StoreInst>(I) ? 990 cast<StoreInst>(I)->getValueOperand()->getType() : I->getType(); 991 Type *aTypeJ = isa<StoreInst>(J) ? 992 cast<StoreInst>(J)->getValueOperand()->getType() : J->getType(); 993 Type *VType = getVecTypeForPair(aTypeI, aTypeJ); 994 995 if (Config.AlignedOnly) { 996 // An aligned load or store is possible only if the instruction 997 // with the lower offset has an alignment suitable for the 998 // vector type. 999 1000 unsigned VecAlignment = DL->getPrefTypeAlignment(VType); 1001 if (BottomAlignment < VecAlignment) 1002 return false; 1003 } 1004 1005 if (TTI) { 1006 unsigned ICost = TTI->getMemoryOpCost(I->getOpcode(), aTypeI, 1007 IAlignment, IAddressSpace); 1008 unsigned JCost = TTI->getMemoryOpCost(J->getOpcode(), aTypeJ, 1009 JAlignment, JAddressSpace); 1010 unsigned VCost = TTI->getMemoryOpCost(I->getOpcode(), VType, 1011 BottomAlignment, 1012 IAddressSpace); 1013 1014 ICost += TTI->getAddressComputationCost(aTypeI); 1015 JCost += TTI->getAddressComputationCost(aTypeJ); 1016 VCost += TTI->getAddressComputationCost(VType); 1017 1018 if (VCost > ICost + JCost) 1019 return false; 1020 1021 // We don't want to fuse to a type that will be split, even 1022 // if the two input types will also be split and there is no other 1023 // associated cost. 1024 unsigned VParts = TTI->getNumberOfParts(VType); 1025 if (VParts > 1) 1026 return false; 1027 else if (!VParts && VCost == ICost + JCost) 1028 return false; 1029 1030 CostSavings = ICost + JCost - VCost; 1031 } 1032 } else { 1033 return false; 1034 } 1035 } else if (TTI) { 1036 unsigned ICost = getInstrCost(I->getOpcode(), IT1, IT2); 1037 unsigned JCost = getInstrCost(J->getOpcode(), JT1, JT2); 1038 Type *VT1 = getVecTypeForPair(IT1, JT1), 1039 *VT2 = getVecTypeForPair(IT2, JT2); 1040 TargetTransformInfo::OperandValueKind Op1VK = 1041 TargetTransformInfo::OK_AnyValue; 1042 TargetTransformInfo::OperandValueKind Op2VK = 1043 TargetTransformInfo::OK_AnyValue; 1044 1045 // On some targets (example X86) the cost of a vector shift may vary 1046 // depending on whether the second operand is a Uniform or 1047 // NonUniform Constant. 1048 switch (I->getOpcode()) { 1049 default : break; 1050 case Instruction::Shl: 1051 case Instruction::LShr: 1052 case Instruction::AShr: 1053 1054 // If both I and J are scalar shifts by constant, then the 1055 // merged vector shift count would be either a constant splat value 1056 // or a non-uniform vector of constants. 1057 if (ConstantInt *CII = dyn_cast<ConstantInt>(I->getOperand(1))) { 1058 if (ConstantInt *CIJ = dyn_cast<ConstantInt>(J->getOperand(1))) 1059 Op2VK = CII == CIJ ? TargetTransformInfo::OK_UniformConstantValue : 1060 TargetTransformInfo::OK_NonUniformConstantValue; 1061 } else { 1062 // Check for a splat of a constant or for a non uniform vector 1063 // of constants. 1064 Value *IOp = I->getOperand(1); 1065 Value *JOp = J->getOperand(1); 1066 if ((isa<ConstantVector>(IOp) || isa<ConstantDataVector>(IOp)) && 1067 (isa<ConstantVector>(JOp) || isa<ConstantDataVector>(JOp))) { 1068 Op2VK = TargetTransformInfo::OK_NonUniformConstantValue; 1069 Constant *SplatValue = cast<Constant>(IOp)->getSplatValue(); 1070 if (SplatValue != nullptr && 1071 SplatValue == cast<Constant>(JOp)->getSplatValue()) 1072 Op2VK = TargetTransformInfo::OK_UniformConstantValue; 1073 } 1074 } 1075 } 1076 1077 // Note that this procedure is incorrect for insert and extract element 1078 // instructions (because combining these often results in a shuffle), 1079 // but this cost is ignored (because insert and extract element 1080 // instructions are assigned a zero depth factor and are not really 1081 // fused in general). 1082 unsigned VCost = getInstrCost(I->getOpcode(), VT1, VT2, Op1VK, Op2VK); 1083 1084 if (VCost > ICost + JCost) 1085 return false; 1086 1087 // We don't want to fuse to a type that will be split, even 1088 // if the two input types will also be split and there is no other 1089 // associated cost. 1090 unsigned VParts1 = TTI->getNumberOfParts(VT1), 1091 VParts2 = TTI->getNumberOfParts(VT2); 1092 if (VParts1 > 1 || VParts2 > 1) 1093 return false; 1094 else if ((!VParts1 || !VParts2) && VCost == ICost + JCost) 1095 return false; 1096 1097 CostSavings = ICost + JCost - VCost; 1098 } 1099 1100 // The powi,ctlz,cttz intrinsics are special because only the first 1101 // argument is vectorized, the second arguments must be equal. 1102 CallInst *CI = dyn_cast<CallInst>(I); 1103 Function *FI; 1104 if (CI && (FI = CI->getCalledFunction())) { 1105 Intrinsic::ID IID = (Intrinsic::ID) FI->getIntrinsicID(); 1106 if (IID == Intrinsic::powi || IID == Intrinsic::ctlz || 1107 IID == Intrinsic::cttz) { 1108 Value *A1I = CI->getArgOperand(1), 1109 *A1J = cast<CallInst>(J)->getArgOperand(1); 1110 const SCEV *A1ISCEV = SE->getSCEV(A1I), 1111 *A1JSCEV = SE->getSCEV(A1J); 1112 return (A1ISCEV == A1JSCEV); 1113 } 1114 1115 if (IID && TTI) { 1116 SmallVector<Type*, 4> Tys; 1117 for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i) 1118 Tys.push_back(CI->getArgOperand(i)->getType()); 1119 unsigned ICost = TTI->getIntrinsicInstrCost(IID, IT1, Tys); 1120 1121 Tys.clear(); 1122 CallInst *CJ = cast<CallInst>(J); 1123 for (unsigned i = 0, ie = CJ->getNumArgOperands(); i != ie; ++i) 1124 Tys.push_back(CJ->getArgOperand(i)->getType()); 1125 unsigned JCost = TTI->getIntrinsicInstrCost(IID, JT1, Tys); 1126 1127 Tys.clear(); 1128 assert(CI->getNumArgOperands() == CJ->getNumArgOperands() && 1129 "Intrinsic argument counts differ"); 1130 for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i) { 1131 if ((IID == Intrinsic::powi || IID == Intrinsic::ctlz || 1132 IID == Intrinsic::cttz) && i == 1) 1133 Tys.push_back(CI->getArgOperand(i)->getType()); 1134 else 1135 Tys.push_back(getVecTypeForPair(CI->getArgOperand(i)->getType(), 1136 CJ->getArgOperand(i)->getType())); 1137 } 1138 1139 Type *RetTy = getVecTypeForPair(IT1, JT1); 1140 unsigned VCost = TTI->getIntrinsicInstrCost(IID, RetTy, Tys); 1141 1142 if (VCost > ICost + JCost) 1143 return false; 1144 1145 // We don't want to fuse to a type that will be split, even 1146 // if the two input types will also be split and there is no other 1147 // associated cost. 1148 unsigned RetParts = TTI->getNumberOfParts(RetTy); 1149 if (RetParts > 1) 1150 return false; 1151 else if (!RetParts && VCost == ICost + JCost) 1152 return false; 1153 1154 for (unsigned i = 0, ie = CI->getNumArgOperands(); i != ie; ++i) { 1155 if (!Tys[i]->isVectorTy()) 1156 continue; 1157 1158 unsigned NumParts = TTI->getNumberOfParts(Tys[i]); 1159 if (NumParts > 1) 1160 return false; 1161 else if (!NumParts && VCost == ICost + JCost) 1162 return false; 1163 } 1164 1165 CostSavings = ICost + JCost - VCost; 1166 } 1167 } 1168 1169 return true; 1170 } 1171 1172 // Figure out whether or not J uses I and update the users and write-set 1173 // structures associated with I. Specifically, Users represents the set of 1174 // instructions that depend on I. WriteSet represents the set 1175 // of memory locations that are dependent on I. If UpdateUsers is true, 1176 // and J uses I, then Users is updated to contain J and WriteSet is updated 1177 // to contain any memory locations to which J writes. The function returns 1178 // true if J uses I. By default, alias analysis is used to determine 1179 // whether J reads from memory that overlaps with a location in WriteSet. 1180 // If LoadMoveSet is not null, then it is a previously-computed map 1181 // where the key is the memory-based user instruction and the value is 1182 // the instruction to be compared with I. So, if LoadMoveSet is provided, 1183 // then the alias analysis is not used. This is necessary because this 1184 // function is called during the process of moving instructions during 1185 // vectorization and the results of the alias analysis are not stable during 1186 // that process. 1187 bool BBVectorize::trackUsesOfI(DenseSet<Value *> &Users, 1188 AliasSetTracker &WriteSet, Instruction *I, 1189 Instruction *J, bool UpdateUsers, 1190 DenseSet<ValuePair> *LoadMoveSetPairs) { 1191 bool UsesI = false; 1192 1193 // This instruction may already be marked as a user due, for example, to 1194 // being a member of a selected pair. 1195 if (Users.count(J)) 1196 UsesI = true; 1197 1198 if (!UsesI) 1199 for (User::op_iterator JU = J->op_begin(), JE = J->op_end(); 1200 JU != JE; ++JU) { 1201 Value *V = *JU; 1202 if (I == V || Users.count(V)) { 1203 UsesI = true; 1204 break; 1205 } 1206 } 1207 if (!UsesI && J->mayReadFromMemory()) { 1208 if (LoadMoveSetPairs) { 1209 UsesI = LoadMoveSetPairs->count(ValuePair(J, I)); 1210 } else { 1211 for (AliasSetTracker::iterator W = WriteSet.begin(), 1212 WE = WriteSet.end(); W != WE; ++W) { 1213 if (W->aliasesUnknownInst(J, *AA)) { 1214 UsesI = true; 1215 break; 1216 } 1217 } 1218 } 1219 } 1220 1221 if (UsesI && UpdateUsers) { 1222 if (J->mayWriteToMemory()) WriteSet.add(J); 1223 Users.insert(J); 1224 } 1225 1226 return UsesI; 1227 } 1228 1229 // This function iterates over all instruction pairs in the provided 1230 // basic block and collects all candidate pairs for vectorization. 1231 bool BBVectorize::getCandidatePairs(BasicBlock &BB, 1232 BasicBlock::iterator &Start, 1233 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1234 DenseSet<ValuePair> &FixedOrderPairs, 1235 DenseMap<ValuePair, int> &CandidatePairCostSavings, 1236 std::vector<Value *> &PairableInsts, bool NonPow2Len) { 1237 size_t TotalPairs = 0; 1238 BasicBlock::iterator E = BB.end(); 1239 if (Start == E) return false; 1240 1241 bool ShouldContinue = false, IAfterStart = false; 1242 for (BasicBlock::iterator I = Start++; I != E; ++I) { 1243 if (I == Start) IAfterStart = true; 1244 1245 bool IsSimpleLoadStore; 1246 if (!isInstVectorizable(I, IsSimpleLoadStore)) continue; 1247 1248 // Look for an instruction with which to pair instruction *I... 1249 DenseSet<Value *> Users; 1250 AliasSetTracker WriteSet(*AA); 1251 if (I->mayWriteToMemory()) WriteSet.add(I); 1252 1253 bool JAfterStart = IAfterStart; 1254 BasicBlock::iterator J = std::next(I); 1255 for (unsigned ss = 0; J != E && ss <= Config.SearchLimit; ++J, ++ss) { 1256 if (J == Start) JAfterStart = true; 1257 1258 // Determine if J uses I, if so, exit the loop. 1259 bool UsesI = trackUsesOfI(Users, WriteSet, I, J, !Config.FastDep); 1260 if (Config.FastDep) { 1261 // Note: For this heuristic to be effective, independent operations 1262 // must tend to be intermixed. This is likely to be true from some 1263 // kinds of grouped loop unrolling (but not the generic LLVM pass), 1264 // but otherwise may require some kind of reordering pass. 1265 1266 // When using fast dependency analysis, 1267 // stop searching after first use: 1268 if (UsesI) break; 1269 } else { 1270 if (UsesI) continue; 1271 } 1272 1273 // J does not use I, and comes before the first use of I, so it can be 1274 // merged with I if the instructions are compatible. 1275 int CostSavings, FixedOrder; 1276 if (!areInstsCompatible(I, J, IsSimpleLoadStore, NonPow2Len, 1277 CostSavings, FixedOrder)) continue; 1278 1279 // J is a candidate for merging with I. 1280 if (!PairableInsts.size() || 1281 PairableInsts[PairableInsts.size()-1] != I) { 1282 PairableInsts.push_back(I); 1283 } 1284 1285 CandidatePairs[I].push_back(J); 1286 ++TotalPairs; 1287 if (TTI) 1288 CandidatePairCostSavings.insert(ValuePairWithCost(ValuePair(I, J), 1289 CostSavings)); 1290 1291 if (FixedOrder == 1) 1292 FixedOrderPairs.insert(ValuePair(I, J)); 1293 else if (FixedOrder == -1) 1294 FixedOrderPairs.insert(ValuePair(J, I)); 1295 1296 // The next call to this function must start after the last instruction 1297 // selected during this invocation. 1298 if (JAfterStart) { 1299 Start = std::next(J); 1300 IAfterStart = JAfterStart = false; 1301 } 1302 1303 DEBUG(if (DebugCandidateSelection) dbgs() << "BBV: candidate pair " 1304 << *I << " <-> " << *J << " (cost savings: " << 1305 CostSavings << ")\n"); 1306 1307 // If we have already found too many pairs, break here and this function 1308 // will be called again starting after the last instruction selected 1309 // during this invocation. 1310 if (PairableInsts.size() >= Config.MaxInsts || 1311 TotalPairs >= Config.MaxPairs) { 1312 ShouldContinue = true; 1313 break; 1314 } 1315 } 1316 1317 if (ShouldContinue) 1318 break; 1319 } 1320 1321 DEBUG(dbgs() << "BBV: found " << PairableInsts.size() 1322 << " instructions with candidate pairs\n"); 1323 1324 return ShouldContinue; 1325 } 1326 1327 // Finds candidate pairs connected to the pair P = <PI, PJ>. This means that 1328 // it looks for pairs such that both members have an input which is an 1329 // output of PI or PJ. 1330 void BBVectorize::computePairsConnectedTo( 1331 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1332 DenseSet<ValuePair> &CandidatePairsSet, 1333 std::vector<Value *> &PairableInsts, 1334 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 1335 DenseMap<VPPair, unsigned> &PairConnectionTypes, 1336 ValuePair P) { 1337 StoreInst *SI, *SJ; 1338 1339 // For each possible pairing for this variable, look at the uses of 1340 // the first value... 1341 for (Value::user_iterator I = P.first->user_begin(), 1342 E = P.first->user_end(); 1343 I != E; ++I) { 1344 User *UI = *I; 1345 if (isa<LoadInst>(UI)) { 1346 // A pair cannot be connected to a load because the load only takes one 1347 // operand (the address) and it is a scalar even after vectorization. 1348 continue; 1349 } else if ((SI = dyn_cast<StoreInst>(UI)) && 1350 P.first == SI->getPointerOperand()) { 1351 // Similarly, a pair cannot be connected to a store through its 1352 // pointer operand. 1353 continue; 1354 } 1355 1356 // For each use of the first variable, look for uses of the second 1357 // variable... 1358 for (User *UJ : P.second->users()) { 1359 if ((SJ = dyn_cast<StoreInst>(UJ)) && 1360 P.second == SJ->getPointerOperand()) 1361 continue; 1362 1363 // Look for <I, J>: 1364 if (CandidatePairsSet.count(ValuePair(UI, UJ))) { 1365 VPPair VP(P, ValuePair(UI, UJ)); 1366 ConnectedPairs[VP.first].push_back(VP.second); 1367 PairConnectionTypes.insert(VPPairWithType(VP, PairConnectionDirect)); 1368 } 1369 1370 // Look for <J, I>: 1371 if (CandidatePairsSet.count(ValuePair(UJ, UI))) { 1372 VPPair VP(P, ValuePair(UJ, UI)); 1373 ConnectedPairs[VP.first].push_back(VP.second); 1374 PairConnectionTypes.insert(VPPairWithType(VP, PairConnectionSwap)); 1375 } 1376 } 1377 1378 if (Config.SplatBreaksChain) continue; 1379 // Look for cases where just the first value in the pair is used by 1380 // both members of another pair (splatting). 1381 for (Value::user_iterator J = P.first->user_begin(); J != E; ++J) { 1382 User *UJ = *J; 1383 if ((SJ = dyn_cast<StoreInst>(UJ)) && 1384 P.first == SJ->getPointerOperand()) 1385 continue; 1386 1387 if (CandidatePairsSet.count(ValuePair(UI, UJ))) { 1388 VPPair VP(P, ValuePair(UI, UJ)); 1389 ConnectedPairs[VP.first].push_back(VP.second); 1390 PairConnectionTypes.insert(VPPairWithType(VP, PairConnectionSplat)); 1391 } 1392 } 1393 } 1394 1395 if (Config.SplatBreaksChain) return; 1396 // Look for cases where just the second value in the pair is used by 1397 // both members of another pair (splatting). 1398 for (Value::user_iterator I = P.second->user_begin(), 1399 E = P.second->user_end(); 1400 I != E; ++I) { 1401 User *UI = *I; 1402 if (isa<LoadInst>(UI)) 1403 continue; 1404 else if ((SI = dyn_cast<StoreInst>(UI)) && 1405 P.second == SI->getPointerOperand()) 1406 continue; 1407 1408 for (Value::user_iterator J = P.second->user_begin(); J != E; ++J) { 1409 User *UJ = *J; 1410 if ((SJ = dyn_cast<StoreInst>(UJ)) && 1411 P.second == SJ->getPointerOperand()) 1412 continue; 1413 1414 if (CandidatePairsSet.count(ValuePair(UI, UJ))) { 1415 VPPair VP(P, ValuePair(UI, UJ)); 1416 ConnectedPairs[VP.first].push_back(VP.second); 1417 PairConnectionTypes.insert(VPPairWithType(VP, PairConnectionSplat)); 1418 } 1419 } 1420 } 1421 } 1422 1423 // This function figures out which pairs are connected. Two pairs are 1424 // connected if some output of the first pair forms an input to both members 1425 // of the second pair. 1426 void BBVectorize::computeConnectedPairs( 1427 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1428 DenseSet<ValuePair> &CandidatePairsSet, 1429 std::vector<Value *> &PairableInsts, 1430 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 1431 DenseMap<VPPair, unsigned> &PairConnectionTypes) { 1432 for (std::vector<Value *>::iterator PI = PairableInsts.begin(), 1433 PE = PairableInsts.end(); PI != PE; ++PI) { 1434 DenseMap<Value *, std::vector<Value *> >::iterator PP = 1435 CandidatePairs.find(*PI); 1436 if (PP == CandidatePairs.end()) 1437 continue; 1438 1439 for (std::vector<Value *>::iterator P = PP->second.begin(), 1440 E = PP->second.end(); P != E; ++P) 1441 computePairsConnectedTo(CandidatePairs, CandidatePairsSet, 1442 PairableInsts, ConnectedPairs, 1443 PairConnectionTypes, ValuePair(*PI, *P)); 1444 } 1445 1446 DEBUG(size_t TotalPairs = 0; 1447 for (DenseMap<ValuePair, std::vector<ValuePair> >::iterator I = 1448 ConnectedPairs.begin(), IE = ConnectedPairs.end(); I != IE; ++I) 1449 TotalPairs += I->second.size(); 1450 dbgs() << "BBV: found " << TotalPairs 1451 << " pair connections.\n"); 1452 } 1453 1454 // This function builds a set of use tuples such that <A, B> is in the set 1455 // if B is in the use dag of A. If B is in the use dag of A, then B 1456 // depends on the output of A. 1457 void BBVectorize::buildDepMap( 1458 BasicBlock &BB, 1459 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1460 std::vector<Value *> &PairableInsts, 1461 DenseSet<ValuePair> &PairableInstUsers) { 1462 DenseSet<Value *> IsInPair; 1463 for (DenseMap<Value *, std::vector<Value *> >::iterator C = 1464 CandidatePairs.begin(), E = CandidatePairs.end(); C != E; ++C) { 1465 IsInPair.insert(C->first); 1466 IsInPair.insert(C->second.begin(), C->second.end()); 1467 } 1468 1469 // Iterate through the basic block, recording all users of each 1470 // pairable instruction. 1471 1472 BasicBlock::iterator E = BB.end(), EL = 1473 BasicBlock::iterator(cast<Instruction>(PairableInsts.back())); 1474 for (BasicBlock::iterator I = BB.getFirstInsertionPt(); I != E; ++I) { 1475 if (IsInPair.find(I) == IsInPair.end()) continue; 1476 1477 DenseSet<Value *> Users; 1478 AliasSetTracker WriteSet(*AA); 1479 if (I->mayWriteToMemory()) WriteSet.add(I); 1480 1481 for (BasicBlock::iterator J = std::next(I); J != E; ++J) { 1482 (void) trackUsesOfI(Users, WriteSet, I, J); 1483 1484 if (J == EL) 1485 break; 1486 } 1487 1488 for (DenseSet<Value *>::iterator U = Users.begin(), E = Users.end(); 1489 U != E; ++U) { 1490 if (IsInPair.find(*U) == IsInPair.end()) continue; 1491 PairableInstUsers.insert(ValuePair(I, *U)); 1492 } 1493 1494 if (I == EL) 1495 break; 1496 } 1497 } 1498 1499 // Returns true if an input to pair P is an output of pair Q and also an 1500 // input of pair Q is an output of pair P. If this is the case, then these 1501 // two pairs cannot be simultaneously fused. 1502 bool BBVectorize::pairsConflict(ValuePair P, ValuePair Q, 1503 DenseSet<ValuePair> &PairableInstUsers, 1504 DenseMap<ValuePair, std::vector<ValuePair> > *PairableInstUserMap, 1505 DenseSet<VPPair> *PairableInstUserPairSet) { 1506 // Two pairs are in conflict if they are mutual Users of eachother. 1507 bool QUsesP = PairableInstUsers.count(ValuePair(P.first, Q.first)) || 1508 PairableInstUsers.count(ValuePair(P.first, Q.second)) || 1509 PairableInstUsers.count(ValuePair(P.second, Q.first)) || 1510 PairableInstUsers.count(ValuePair(P.second, Q.second)); 1511 bool PUsesQ = PairableInstUsers.count(ValuePair(Q.first, P.first)) || 1512 PairableInstUsers.count(ValuePair(Q.first, P.second)) || 1513 PairableInstUsers.count(ValuePair(Q.second, P.first)) || 1514 PairableInstUsers.count(ValuePair(Q.second, P.second)); 1515 if (PairableInstUserMap) { 1516 // FIXME: The expensive part of the cycle check is not so much the cycle 1517 // check itself but this edge insertion procedure. This needs some 1518 // profiling and probably a different data structure. 1519 if (PUsesQ) { 1520 if (PairableInstUserPairSet->insert(VPPair(Q, P)).second) 1521 (*PairableInstUserMap)[Q].push_back(P); 1522 } 1523 if (QUsesP) { 1524 if (PairableInstUserPairSet->insert(VPPair(P, Q)).second) 1525 (*PairableInstUserMap)[P].push_back(Q); 1526 } 1527 } 1528 1529 return (QUsesP && PUsesQ); 1530 } 1531 1532 // This function walks the use graph of current pairs to see if, starting 1533 // from P, the walk returns to P. 1534 bool BBVectorize::pairWillFormCycle(ValuePair P, 1535 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUserMap, 1536 DenseSet<ValuePair> &CurrentPairs) { 1537 DEBUG(if (DebugCycleCheck) 1538 dbgs() << "BBV: starting cycle check for : " << *P.first << " <-> " 1539 << *P.second << "\n"); 1540 // A lookup table of visisted pairs is kept because the PairableInstUserMap 1541 // contains non-direct associations. 1542 DenseSet<ValuePair> Visited; 1543 SmallVector<ValuePair, 32> Q; 1544 // General depth-first post-order traversal: 1545 Q.push_back(P); 1546 do { 1547 ValuePair QTop = Q.pop_back_val(); 1548 Visited.insert(QTop); 1549 1550 DEBUG(if (DebugCycleCheck) 1551 dbgs() << "BBV: cycle check visiting: " << *QTop.first << " <-> " 1552 << *QTop.second << "\n"); 1553 DenseMap<ValuePair, std::vector<ValuePair> >::iterator QQ = 1554 PairableInstUserMap.find(QTop); 1555 if (QQ == PairableInstUserMap.end()) 1556 continue; 1557 1558 for (std::vector<ValuePair>::iterator C = QQ->second.begin(), 1559 CE = QQ->second.end(); C != CE; ++C) { 1560 if (*C == P) { 1561 DEBUG(dbgs() 1562 << "BBV: rejected to prevent non-trivial cycle formation: " 1563 << QTop.first << " <-> " << C->second << "\n"); 1564 return true; 1565 } 1566 1567 if (CurrentPairs.count(*C) && !Visited.count(*C)) 1568 Q.push_back(*C); 1569 } 1570 } while (!Q.empty()); 1571 1572 return false; 1573 } 1574 1575 // This function builds the initial dag of connected pairs with the 1576 // pair J at the root. 1577 void BBVectorize::buildInitialDAGFor( 1578 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1579 DenseSet<ValuePair> &CandidatePairsSet, 1580 std::vector<Value *> &PairableInsts, 1581 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 1582 DenseSet<ValuePair> &PairableInstUsers, 1583 DenseMap<Value *, Value *> &ChosenPairs, 1584 DenseMap<ValuePair, size_t> &DAG, ValuePair J) { 1585 // Each of these pairs is viewed as the root node of a DAG. The DAG 1586 // is then walked (depth-first). As this happens, we keep track of 1587 // the pairs that compose the DAG and the maximum depth of the DAG. 1588 SmallVector<ValuePairWithDepth, 32> Q; 1589 // General depth-first post-order traversal: 1590 Q.push_back(ValuePairWithDepth(J, getDepthFactor(J.first))); 1591 do { 1592 ValuePairWithDepth QTop = Q.back(); 1593 1594 // Push each child onto the queue: 1595 bool MoreChildren = false; 1596 size_t MaxChildDepth = QTop.second; 1597 DenseMap<ValuePair, std::vector<ValuePair> >::iterator QQ = 1598 ConnectedPairs.find(QTop.first); 1599 if (QQ != ConnectedPairs.end()) 1600 for (std::vector<ValuePair>::iterator k = QQ->second.begin(), 1601 ke = QQ->second.end(); k != ke; ++k) { 1602 // Make sure that this child pair is still a candidate: 1603 if (CandidatePairsSet.count(*k)) { 1604 DenseMap<ValuePair, size_t>::iterator C = DAG.find(*k); 1605 if (C == DAG.end()) { 1606 size_t d = getDepthFactor(k->first); 1607 Q.push_back(ValuePairWithDepth(*k, QTop.second+d)); 1608 MoreChildren = true; 1609 } else { 1610 MaxChildDepth = std::max(MaxChildDepth, C->second); 1611 } 1612 } 1613 } 1614 1615 if (!MoreChildren) { 1616 // Record the current pair as part of the DAG: 1617 DAG.insert(ValuePairWithDepth(QTop.first, MaxChildDepth)); 1618 Q.pop_back(); 1619 } 1620 } while (!Q.empty()); 1621 } 1622 1623 // Given some initial dag, prune it by removing conflicting pairs (pairs 1624 // that cannot be simultaneously chosen for vectorization). 1625 void BBVectorize::pruneDAGFor( 1626 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1627 std::vector<Value *> &PairableInsts, 1628 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 1629 DenseSet<ValuePair> &PairableInstUsers, 1630 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUserMap, 1631 DenseSet<VPPair> &PairableInstUserPairSet, 1632 DenseMap<Value *, Value *> &ChosenPairs, 1633 DenseMap<ValuePair, size_t> &DAG, 1634 DenseSet<ValuePair> &PrunedDAG, ValuePair J, 1635 bool UseCycleCheck) { 1636 SmallVector<ValuePairWithDepth, 32> Q; 1637 // General depth-first post-order traversal: 1638 Q.push_back(ValuePairWithDepth(J, getDepthFactor(J.first))); 1639 do { 1640 ValuePairWithDepth QTop = Q.pop_back_val(); 1641 PrunedDAG.insert(QTop.first); 1642 1643 // Visit each child, pruning as necessary... 1644 SmallVector<ValuePairWithDepth, 8> BestChildren; 1645 DenseMap<ValuePair, std::vector<ValuePair> >::iterator QQ = 1646 ConnectedPairs.find(QTop.first); 1647 if (QQ == ConnectedPairs.end()) 1648 continue; 1649 1650 for (std::vector<ValuePair>::iterator K = QQ->second.begin(), 1651 KE = QQ->second.end(); K != KE; ++K) { 1652 DenseMap<ValuePair, size_t>::iterator C = DAG.find(*K); 1653 if (C == DAG.end()) continue; 1654 1655 // This child is in the DAG, now we need to make sure it is the 1656 // best of any conflicting children. There could be multiple 1657 // conflicting children, so first, determine if we're keeping 1658 // this child, then delete conflicting children as necessary. 1659 1660 // It is also necessary to guard against pairing-induced 1661 // dependencies. Consider instructions a .. x .. y .. b 1662 // such that (a,b) are to be fused and (x,y) are to be fused 1663 // but a is an input to x and b is an output from y. This 1664 // means that y cannot be moved after b but x must be moved 1665 // after b for (a,b) to be fused. In other words, after 1666 // fusing (a,b) we have y .. a/b .. x where y is an input 1667 // to a/b and x is an output to a/b: x and y can no longer 1668 // be legally fused. To prevent this condition, we must 1669 // make sure that a child pair added to the DAG is not 1670 // both an input and output of an already-selected pair. 1671 1672 // Pairing-induced dependencies can also form from more complicated 1673 // cycles. The pair vs. pair conflicts are easy to check, and so 1674 // that is done explicitly for "fast rejection", and because for 1675 // child vs. child conflicts, we may prefer to keep the current 1676 // pair in preference to the already-selected child. 1677 DenseSet<ValuePair> CurrentPairs; 1678 1679 bool CanAdd = true; 1680 for (SmallVectorImpl<ValuePairWithDepth>::iterator C2 1681 = BestChildren.begin(), E2 = BestChildren.end(); 1682 C2 != E2; ++C2) { 1683 if (C2->first.first == C->first.first || 1684 C2->first.first == C->first.second || 1685 C2->first.second == C->first.first || 1686 C2->first.second == C->first.second || 1687 pairsConflict(C2->first, C->first, PairableInstUsers, 1688 UseCycleCheck ? &PairableInstUserMap : nullptr, 1689 UseCycleCheck ? &PairableInstUserPairSet 1690 : nullptr)) { 1691 if (C2->second >= C->second) { 1692 CanAdd = false; 1693 break; 1694 } 1695 1696 CurrentPairs.insert(C2->first); 1697 } 1698 } 1699 if (!CanAdd) continue; 1700 1701 // Even worse, this child could conflict with another node already 1702 // selected for the DAG. If that is the case, ignore this child. 1703 for (DenseSet<ValuePair>::iterator T = PrunedDAG.begin(), 1704 E2 = PrunedDAG.end(); T != E2; ++T) { 1705 if (T->first == C->first.first || 1706 T->first == C->first.second || 1707 T->second == C->first.first || 1708 T->second == C->first.second || 1709 pairsConflict(*T, C->first, PairableInstUsers, 1710 UseCycleCheck ? &PairableInstUserMap : nullptr, 1711 UseCycleCheck ? &PairableInstUserPairSet 1712 : nullptr)) { 1713 CanAdd = false; 1714 break; 1715 } 1716 1717 CurrentPairs.insert(*T); 1718 } 1719 if (!CanAdd) continue; 1720 1721 // And check the queue too... 1722 for (SmallVectorImpl<ValuePairWithDepth>::iterator C2 = Q.begin(), 1723 E2 = Q.end(); C2 != E2; ++C2) { 1724 if (C2->first.first == C->first.first || 1725 C2->first.first == C->first.second || 1726 C2->first.second == C->first.first || 1727 C2->first.second == C->first.second || 1728 pairsConflict(C2->first, C->first, PairableInstUsers, 1729 UseCycleCheck ? &PairableInstUserMap : nullptr, 1730 UseCycleCheck ? &PairableInstUserPairSet 1731 : nullptr)) { 1732 CanAdd = false; 1733 break; 1734 } 1735 1736 CurrentPairs.insert(C2->first); 1737 } 1738 if (!CanAdd) continue; 1739 1740 // Last but not least, check for a conflict with any of the 1741 // already-chosen pairs. 1742 for (DenseMap<Value *, Value *>::iterator C2 = 1743 ChosenPairs.begin(), E2 = ChosenPairs.end(); 1744 C2 != E2; ++C2) { 1745 if (pairsConflict(*C2, C->first, PairableInstUsers, 1746 UseCycleCheck ? &PairableInstUserMap : nullptr, 1747 UseCycleCheck ? &PairableInstUserPairSet 1748 : nullptr)) { 1749 CanAdd = false; 1750 break; 1751 } 1752 1753 CurrentPairs.insert(*C2); 1754 } 1755 if (!CanAdd) continue; 1756 1757 // To check for non-trivial cycles formed by the addition of the 1758 // current pair we've formed a list of all relevant pairs, now use a 1759 // graph walk to check for a cycle. We start from the current pair and 1760 // walk the use dag to see if we again reach the current pair. If we 1761 // do, then the current pair is rejected. 1762 1763 // FIXME: It may be more efficient to use a topological-ordering 1764 // algorithm to improve the cycle check. This should be investigated. 1765 if (UseCycleCheck && 1766 pairWillFormCycle(C->first, PairableInstUserMap, CurrentPairs)) 1767 continue; 1768 1769 // This child can be added, but we may have chosen it in preference 1770 // to an already-selected child. Check for this here, and if a 1771 // conflict is found, then remove the previously-selected child 1772 // before adding this one in its place. 1773 for (SmallVectorImpl<ValuePairWithDepth>::iterator C2 1774 = BestChildren.begin(); C2 != BestChildren.end();) { 1775 if (C2->first.first == C->first.first || 1776 C2->first.first == C->first.second || 1777 C2->first.second == C->first.first || 1778 C2->first.second == C->first.second || 1779 pairsConflict(C2->first, C->first, PairableInstUsers)) 1780 C2 = BestChildren.erase(C2); 1781 else 1782 ++C2; 1783 } 1784 1785 BestChildren.push_back(ValuePairWithDepth(C->first, C->second)); 1786 } 1787 1788 for (SmallVectorImpl<ValuePairWithDepth>::iterator C 1789 = BestChildren.begin(), E2 = BestChildren.end(); 1790 C != E2; ++C) { 1791 size_t DepthF = getDepthFactor(C->first.first); 1792 Q.push_back(ValuePairWithDepth(C->first, QTop.second+DepthF)); 1793 } 1794 } while (!Q.empty()); 1795 } 1796 1797 // This function finds the best dag of mututally-compatible connected 1798 // pairs, given the choice of root pairs as an iterator range. 1799 void BBVectorize::findBestDAGFor( 1800 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 1801 DenseSet<ValuePair> &CandidatePairsSet, 1802 DenseMap<ValuePair, int> &CandidatePairCostSavings, 1803 std::vector<Value *> &PairableInsts, 1804 DenseSet<ValuePair> &FixedOrderPairs, 1805 DenseMap<VPPair, unsigned> &PairConnectionTypes, 1806 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 1807 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps, 1808 DenseSet<ValuePair> &PairableInstUsers, 1809 DenseMap<ValuePair, std::vector<ValuePair> > &PairableInstUserMap, 1810 DenseSet<VPPair> &PairableInstUserPairSet, 1811 DenseMap<Value *, Value *> &ChosenPairs, 1812 DenseSet<ValuePair> &BestDAG, size_t &BestMaxDepth, 1813 int &BestEffSize, Value *II, std::vector<Value *>&JJ, 1814 bool UseCycleCheck) { 1815 for (std::vector<Value *>::iterator J = JJ.begin(), JE = JJ.end(); 1816 J != JE; ++J) { 1817 ValuePair IJ(II, *J); 1818 if (!CandidatePairsSet.count(IJ)) 1819 continue; 1820 1821 // Before going any further, make sure that this pair does not 1822 // conflict with any already-selected pairs (see comment below 1823 // near the DAG pruning for more details). 1824 DenseSet<ValuePair> ChosenPairSet; 1825 bool DoesConflict = false; 1826 for (DenseMap<Value *, Value *>::iterator C = ChosenPairs.begin(), 1827 E = ChosenPairs.end(); C != E; ++C) { 1828 if (pairsConflict(*C, IJ, PairableInstUsers, 1829 UseCycleCheck ? &PairableInstUserMap : nullptr, 1830 UseCycleCheck ? &PairableInstUserPairSet : nullptr)) { 1831 DoesConflict = true; 1832 break; 1833 } 1834 1835 ChosenPairSet.insert(*C); 1836 } 1837 if (DoesConflict) continue; 1838 1839 if (UseCycleCheck && 1840 pairWillFormCycle(IJ, PairableInstUserMap, ChosenPairSet)) 1841 continue; 1842 1843 DenseMap<ValuePair, size_t> DAG; 1844 buildInitialDAGFor(CandidatePairs, CandidatePairsSet, 1845 PairableInsts, ConnectedPairs, 1846 PairableInstUsers, ChosenPairs, DAG, IJ); 1847 1848 // Because we'll keep the child with the largest depth, the largest 1849 // depth is still the same in the unpruned DAG. 1850 size_t MaxDepth = DAG.lookup(IJ); 1851 1852 DEBUG(if (DebugPairSelection) dbgs() << "BBV: found DAG for pair {" 1853 << *IJ.first << " <-> " << *IJ.second << "} of depth " << 1854 MaxDepth << " and size " << DAG.size() << "\n"); 1855 1856 // At this point the DAG has been constructed, but, may contain 1857 // contradictory children (meaning that different children of 1858 // some dag node may be attempting to fuse the same instruction). 1859 // So now we walk the dag again, in the case of a conflict, 1860 // keep only the child with the largest depth. To break a tie, 1861 // favor the first child. 1862 1863 DenseSet<ValuePair> PrunedDAG; 1864 pruneDAGFor(CandidatePairs, PairableInsts, ConnectedPairs, 1865 PairableInstUsers, PairableInstUserMap, 1866 PairableInstUserPairSet, 1867 ChosenPairs, DAG, PrunedDAG, IJ, UseCycleCheck); 1868 1869 int EffSize = 0; 1870 if (TTI) { 1871 DenseSet<Value *> PrunedDAGInstrs; 1872 for (DenseSet<ValuePair>::iterator S = PrunedDAG.begin(), 1873 E = PrunedDAG.end(); S != E; ++S) { 1874 PrunedDAGInstrs.insert(S->first); 1875 PrunedDAGInstrs.insert(S->second); 1876 } 1877 1878 // The set of pairs that have already contributed to the total cost. 1879 DenseSet<ValuePair> IncomingPairs; 1880 1881 // If the cost model were perfect, this might not be necessary; but we 1882 // need to make sure that we don't get stuck vectorizing our own 1883 // shuffle chains. 1884 bool HasNontrivialInsts = false; 1885 1886 // The node weights represent the cost savings associated with 1887 // fusing the pair of instructions. 1888 for (DenseSet<ValuePair>::iterator S = PrunedDAG.begin(), 1889 E = PrunedDAG.end(); S != E; ++S) { 1890 if (!isa<ShuffleVectorInst>(S->first) && 1891 !isa<InsertElementInst>(S->first) && 1892 !isa<ExtractElementInst>(S->first)) 1893 HasNontrivialInsts = true; 1894 1895 bool FlipOrder = false; 1896 1897 if (getDepthFactor(S->first)) { 1898 int ESContrib = CandidatePairCostSavings.find(*S)->second; 1899 DEBUG(if (DebugPairSelection) dbgs() << "\tweight {" 1900 << *S->first << " <-> " << *S->second << "} = " << 1901 ESContrib << "\n"); 1902 EffSize += ESContrib; 1903 } 1904 1905 // The edge weights contribute in a negative sense: they represent 1906 // the cost of shuffles. 1907 DenseMap<ValuePair, std::vector<ValuePair> >::iterator SS = 1908 ConnectedPairDeps.find(*S); 1909 if (SS != ConnectedPairDeps.end()) { 1910 unsigned NumDepsDirect = 0, NumDepsSwap = 0; 1911 for (std::vector<ValuePair>::iterator T = SS->second.begin(), 1912 TE = SS->second.end(); T != TE; ++T) { 1913 VPPair Q(*S, *T); 1914 if (!PrunedDAG.count(Q.second)) 1915 continue; 1916 DenseMap<VPPair, unsigned>::iterator R = 1917 PairConnectionTypes.find(VPPair(Q.second, Q.first)); 1918 assert(R != PairConnectionTypes.end() && 1919 "Cannot find pair connection type"); 1920 if (R->second == PairConnectionDirect) 1921 ++NumDepsDirect; 1922 else if (R->second == PairConnectionSwap) 1923 ++NumDepsSwap; 1924 } 1925 1926 // If there are more swaps than direct connections, then 1927 // the pair order will be flipped during fusion. So the real 1928 // number of swaps is the minimum number. 1929 FlipOrder = !FixedOrderPairs.count(*S) && 1930 ((NumDepsSwap > NumDepsDirect) || 1931 FixedOrderPairs.count(ValuePair(S->second, S->first))); 1932 1933 for (std::vector<ValuePair>::iterator T = SS->second.begin(), 1934 TE = SS->second.end(); T != TE; ++T) { 1935 VPPair Q(*S, *T); 1936 if (!PrunedDAG.count(Q.second)) 1937 continue; 1938 DenseMap<VPPair, unsigned>::iterator R = 1939 PairConnectionTypes.find(VPPair(Q.second, Q.first)); 1940 assert(R != PairConnectionTypes.end() && 1941 "Cannot find pair connection type"); 1942 Type *Ty1 = Q.second.first->getType(), 1943 *Ty2 = Q.second.second->getType(); 1944 Type *VTy = getVecTypeForPair(Ty1, Ty2); 1945 if ((R->second == PairConnectionDirect && FlipOrder) || 1946 (R->second == PairConnectionSwap && !FlipOrder) || 1947 R->second == PairConnectionSplat) { 1948 int ESContrib = (int) getInstrCost(Instruction::ShuffleVector, 1949 VTy, VTy); 1950 1951 if (VTy->getVectorNumElements() == 2) { 1952 if (R->second == PairConnectionSplat) 1953 ESContrib = std::min(ESContrib, (int) TTI->getShuffleCost( 1954 TargetTransformInfo::SK_Broadcast, VTy)); 1955 else 1956 ESContrib = std::min(ESContrib, (int) TTI->getShuffleCost( 1957 TargetTransformInfo::SK_Reverse, VTy)); 1958 } 1959 1960 DEBUG(if (DebugPairSelection) dbgs() << "\tcost {" << 1961 *Q.second.first << " <-> " << *Q.second.second << 1962 "} -> {" << 1963 *S->first << " <-> " << *S->second << "} = " << 1964 ESContrib << "\n"); 1965 EffSize -= ESContrib; 1966 } 1967 } 1968 } 1969 1970 // Compute the cost of outgoing edges. We assume that edges outgoing 1971 // to shuffles, inserts or extracts can be merged, and so contribute 1972 // no additional cost. 1973 if (!S->first->getType()->isVoidTy()) { 1974 Type *Ty1 = S->first->getType(), 1975 *Ty2 = S->second->getType(); 1976 Type *VTy = getVecTypeForPair(Ty1, Ty2); 1977 1978 bool NeedsExtraction = false; 1979 for (User *U : S->first->users()) { 1980 if (ShuffleVectorInst *SI = dyn_cast<ShuffleVectorInst>(U)) { 1981 // Shuffle can be folded if it has no other input 1982 if (isa<UndefValue>(SI->getOperand(1))) 1983 continue; 1984 } 1985 if (isa<ExtractElementInst>(U)) 1986 continue; 1987 if (PrunedDAGInstrs.count(U)) 1988 continue; 1989 NeedsExtraction = true; 1990 break; 1991 } 1992 1993 if (NeedsExtraction) { 1994 int ESContrib; 1995 if (Ty1->isVectorTy()) { 1996 ESContrib = (int) getInstrCost(Instruction::ShuffleVector, 1997 Ty1, VTy); 1998 ESContrib = std::min(ESContrib, (int) TTI->getShuffleCost( 1999 TargetTransformInfo::SK_ExtractSubvector, VTy, 0, Ty1)); 2000 } else 2001 ESContrib = (int) TTI->getVectorInstrCost( 2002 Instruction::ExtractElement, VTy, 0); 2003 2004 DEBUG(if (DebugPairSelection) dbgs() << "\tcost {" << 2005 *S->first << "} = " << ESContrib << "\n"); 2006 EffSize -= ESContrib; 2007 } 2008 2009 NeedsExtraction = false; 2010 for (User *U : S->second->users()) { 2011 if (ShuffleVectorInst *SI = dyn_cast<ShuffleVectorInst>(U)) { 2012 // Shuffle can be folded if it has no other input 2013 if (isa<UndefValue>(SI->getOperand(1))) 2014 continue; 2015 } 2016 if (isa<ExtractElementInst>(U)) 2017 continue; 2018 if (PrunedDAGInstrs.count(U)) 2019 continue; 2020 NeedsExtraction = true; 2021 break; 2022 } 2023 2024 if (NeedsExtraction) { 2025 int ESContrib; 2026 if (Ty2->isVectorTy()) { 2027 ESContrib = (int) getInstrCost(Instruction::ShuffleVector, 2028 Ty2, VTy); 2029 ESContrib = std::min(ESContrib, (int) TTI->getShuffleCost( 2030 TargetTransformInfo::SK_ExtractSubvector, VTy, 2031 Ty1->isVectorTy() ? Ty1->getVectorNumElements() : 1, Ty2)); 2032 } else 2033 ESContrib = (int) TTI->getVectorInstrCost( 2034 Instruction::ExtractElement, VTy, 1); 2035 DEBUG(if (DebugPairSelection) dbgs() << "\tcost {" << 2036 *S->second << "} = " << ESContrib << "\n"); 2037 EffSize -= ESContrib; 2038 } 2039 } 2040 2041 // Compute the cost of incoming edges. 2042 if (!isa<LoadInst>(S->first) && !isa<StoreInst>(S->first)) { 2043 Instruction *S1 = cast<Instruction>(S->first), 2044 *S2 = cast<Instruction>(S->second); 2045 for (unsigned o = 0; o < S1->getNumOperands(); ++o) { 2046 Value *O1 = S1->getOperand(o), *O2 = S2->getOperand(o); 2047 2048 // Combining constants into vector constants (or small vector 2049 // constants into larger ones are assumed free). 2050 if (isa<Constant>(O1) && isa<Constant>(O2)) 2051 continue; 2052 2053 if (FlipOrder) 2054 std::swap(O1, O2); 2055 2056 ValuePair VP = ValuePair(O1, O2); 2057 ValuePair VPR = ValuePair(O2, O1); 2058 2059 // Internal edges are not handled here. 2060 if (PrunedDAG.count(VP) || PrunedDAG.count(VPR)) 2061 continue; 2062 2063 Type *Ty1 = O1->getType(), 2064 *Ty2 = O2->getType(); 2065 Type *VTy = getVecTypeForPair(Ty1, Ty2); 2066 2067 // Combining vector operations of the same type is also assumed 2068 // folded with other operations. 2069 if (Ty1 == Ty2) { 2070 // If both are insert elements, then both can be widened. 2071 InsertElementInst *IEO1 = dyn_cast<InsertElementInst>(O1), 2072 *IEO2 = dyn_cast<InsertElementInst>(O2); 2073 if (IEO1 && IEO2 && isPureIEChain(IEO1) && isPureIEChain(IEO2)) 2074 continue; 2075 // If both are extract elements, and both have the same input 2076 // type, then they can be replaced with a shuffle 2077 ExtractElementInst *EIO1 = dyn_cast<ExtractElementInst>(O1), 2078 *EIO2 = dyn_cast<ExtractElementInst>(O2); 2079 if (EIO1 && EIO2 && 2080 EIO1->getOperand(0)->getType() == 2081 EIO2->getOperand(0)->getType()) 2082 continue; 2083 // If both are a shuffle with equal operand types and only two 2084 // unqiue operands, then they can be replaced with a single 2085 // shuffle 2086 ShuffleVectorInst *SIO1 = dyn_cast<ShuffleVectorInst>(O1), 2087 *SIO2 = dyn_cast<ShuffleVectorInst>(O2); 2088 if (SIO1 && SIO2 && 2089 SIO1->getOperand(0)->getType() == 2090 SIO2->getOperand(0)->getType()) { 2091 SmallSet<Value *, 4> SIOps; 2092 SIOps.insert(SIO1->getOperand(0)); 2093 SIOps.insert(SIO1->getOperand(1)); 2094 SIOps.insert(SIO2->getOperand(0)); 2095 SIOps.insert(SIO2->getOperand(1)); 2096 if (SIOps.size() <= 2) 2097 continue; 2098 } 2099 } 2100 2101 int ESContrib; 2102 // This pair has already been formed. 2103 if (IncomingPairs.count(VP)) { 2104 continue; 2105 } else if (IncomingPairs.count(VPR)) { 2106 ESContrib = (int) getInstrCost(Instruction::ShuffleVector, 2107 VTy, VTy); 2108 2109 if (VTy->getVectorNumElements() == 2) 2110 ESContrib = std::min(ESContrib, (int) TTI->getShuffleCost( 2111 TargetTransformInfo::SK_Reverse, VTy)); 2112 } else if (!Ty1->isVectorTy() && !Ty2->isVectorTy()) { 2113 ESContrib = (int) TTI->getVectorInstrCost( 2114 Instruction::InsertElement, VTy, 0); 2115 ESContrib += (int) TTI->getVectorInstrCost( 2116 Instruction::InsertElement, VTy, 1); 2117 } else if (!Ty1->isVectorTy()) { 2118 // O1 needs to be inserted into a vector of size O2, and then 2119 // both need to be shuffled together. 2120 ESContrib = (int) TTI->getVectorInstrCost( 2121 Instruction::InsertElement, Ty2, 0); 2122 ESContrib += (int) getInstrCost(Instruction::ShuffleVector, 2123 VTy, Ty2); 2124 } else if (!Ty2->isVectorTy()) { 2125 // O2 needs to be inserted into a vector of size O1, and then 2126 // both need to be shuffled together. 2127 ESContrib = (int) TTI->getVectorInstrCost( 2128 Instruction::InsertElement, Ty1, 0); 2129 ESContrib += (int) getInstrCost(Instruction::ShuffleVector, 2130 VTy, Ty1); 2131 } else { 2132 Type *TyBig = Ty1, *TySmall = Ty2; 2133 if (Ty2->getVectorNumElements() > Ty1->getVectorNumElements()) 2134 std::swap(TyBig, TySmall); 2135 2136 ESContrib = (int) getInstrCost(Instruction::ShuffleVector, 2137 VTy, TyBig); 2138 if (TyBig != TySmall) 2139 ESContrib += (int) getInstrCost(Instruction::ShuffleVector, 2140 TyBig, TySmall); 2141 } 2142 2143 DEBUG(if (DebugPairSelection) dbgs() << "\tcost {" 2144 << *O1 << " <-> " << *O2 << "} = " << 2145 ESContrib << "\n"); 2146 EffSize -= ESContrib; 2147 IncomingPairs.insert(VP); 2148 } 2149 } 2150 } 2151 2152 if (!HasNontrivialInsts) { 2153 DEBUG(if (DebugPairSelection) dbgs() << 2154 "\tNo non-trivial instructions in DAG;" 2155 " override to zero effective size\n"); 2156 EffSize = 0; 2157 } 2158 } else { 2159 for (DenseSet<ValuePair>::iterator S = PrunedDAG.begin(), 2160 E = PrunedDAG.end(); S != E; ++S) 2161 EffSize += (int) getDepthFactor(S->first); 2162 } 2163 2164 DEBUG(if (DebugPairSelection) 2165 dbgs() << "BBV: found pruned DAG for pair {" 2166 << *IJ.first << " <-> " << *IJ.second << "} of depth " << 2167 MaxDepth << " and size " << PrunedDAG.size() << 2168 " (effective size: " << EffSize << ")\n"); 2169 if (((TTI && !UseChainDepthWithTI) || 2170 MaxDepth >= Config.ReqChainDepth) && 2171 EffSize > 0 && EffSize > BestEffSize) { 2172 BestMaxDepth = MaxDepth; 2173 BestEffSize = EffSize; 2174 BestDAG = PrunedDAG; 2175 } 2176 } 2177 } 2178 2179 // Given the list of candidate pairs, this function selects those 2180 // that will be fused into vector instructions. 2181 void BBVectorize::choosePairs( 2182 DenseMap<Value *, std::vector<Value *> > &CandidatePairs, 2183 DenseSet<ValuePair> &CandidatePairsSet, 2184 DenseMap<ValuePair, int> &CandidatePairCostSavings, 2185 std::vector<Value *> &PairableInsts, 2186 DenseSet<ValuePair> &FixedOrderPairs, 2187 DenseMap<VPPair, unsigned> &PairConnectionTypes, 2188 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 2189 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps, 2190 DenseSet<ValuePair> &PairableInstUsers, 2191 DenseMap<Value *, Value *>& ChosenPairs) { 2192 bool UseCycleCheck = 2193 CandidatePairsSet.size() <= Config.MaxCandPairsForCycleCheck; 2194 2195 DenseMap<Value *, std::vector<Value *> > CandidatePairs2; 2196 for (DenseSet<ValuePair>::iterator I = CandidatePairsSet.begin(), 2197 E = CandidatePairsSet.end(); I != E; ++I) { 2198 std::vector<Value *> &JJ = CandidatePairs2[I->second]; 2199 if (JJ.empty()) JJ.reserve(32); 2200 JJ.push_back(I->first); 2201 } 2202 2203 DenseMap<ValuePair, std::vector<ValuePair> > PairableInstUserMap; 2204 DenseSet<VPPair> PairableInstUserPairSet; 2205 for (std::vector<Value *>::iterator I = PairableInsts.begin(), 2206 E = PairableInsts.end(); I != E; ++I) { 2207 // The number of possible pairings for this variable: 2208 size_t NumChoices = CandidatePairs.lookup(*I).size(); 2209 if (!NumChoices) continue; 2210 2211 std::vector<Value *> &JJ = CandidatePairs[*I]; 2212 2213 // The best pair to choose and its dag: 2214 size_t BestMaxDepth = 0; 2215 int BestEffSize = 0; 2216 DenseSet<ValuePair> BestDAG; 2217 findBestDAGFor(CandidatePairs, CandidatePairsSet, 2218 CandidatePairCostSavings, 2219 PairableInsts, FixedOrderPairs, PairConnectionTypes, 2220 ConnectedPairs, ConnectedPairDeps, 2221 PairableInstUsers, PairableInstUserMap, 2222 PairableInstUserPairSet, ChosenPairs, 2223 BestDAG, BestMaxDepth, BestEffSize, *I, JJ, 2224 UseCycleCheck); 2225 2226 if (BestDAG.empty()) 2227 continue; 2228 2229 // A dag has been chosen (or not) at this point. If no dag was 2230 // chosen, then this instruction, I, cannot be paired (and is no longer 2231 // considered). 2232 2233 DEBUG(dbgs() << "BBV: selected pairs in the best DAG for: " 2234 << *cast<Instruction>(*I) << "\n"); 2235 2236 for (DenseSet<ValuePair>::iterator S = BestDAG.begin(), 2237 SE2 = BestDAG.end(); S != SE2; ++S) { 2238 // Insert the members of this dag into the list of chosen pairs. 2239 ChosenPairs.insert(ValuePair(S->first, S->second)); 2240 DEBUG(dbgs() << "BBV: selected pair: " << *S->first << " <-> " << 2241 *S->second << "\n"); 2242 2243 // Remove all candidate pairs that have values in the chosen dag. 2244 std::vector<Value *> &KK = CandidatePairs[S->first]; 2245 for (std::vector<Value *>::iterator K = KK.begin(), KE = KK.end(); 2246 K != KE; ++K) { 2247 if (*K == S->second) 2248 continue; 2249 2250 CandidatePairsSet.erase(ValuePair(S->first, *K)); 2251 } 2252 2253 std::vector<Value *> &LL = CandidatePairs2[S->second]; 2254 for (std::vector<Value *>::iterator L = LL.begin(), LE = LL.end(); 2255 L != LE; ++L) { 2256 if (*L == S->first) 2257 continue; 2258 2259 CandidatePairsSet.erase(ValuePair(*L, S->second)); 2260 } 2261 2262 std::vector<Value *> &MM = CandidatePairs[S->second]; 2263 for (std::vector<Value *>::iterator M = MM.begin(), ME = MM.end(); 2264 M != ME; ++M) { 2265 assert(*M != S->first && "Flipped pair in candidate list?"); 2266 CandidatePairsSet.erase(ValuePair(S->second, *M)); 2267 } 2268 2269 std::vector<Value *> &NN = CandidatePairs2[S->first]; 2270 for (std::vector<Value *>::iterator N = NN.begin(), NE = NN.end(); 2271 N != NE; ++N) { 2272 assert(*N != S->second && "Flipped pair in candidate list?"); 2273 CandidatePairsSet.erase(ValuePair(*N, S->first)); 2274 } 2275 } 2276 } 2277 2278 DEBUG(dbgs() << "BBV: selected " << ChosenPairs.size() << " pairs.\n"); 2279 } 2280 2281 std::string getReplacementName(Instruction *I, bool IsInput, unsigned o, 2282 unsigned n = 0) { 2283 if (!I->hasName()) 2284 return ""; 2285 2286 return (I->getName() + (IsInput ? ".v.i" : ".v.r") + utostr(o) + 2287 (n > 0 ? "." + utostr(n) : "")).str(); 2288 } 2289 2290 // Returns the value that is to be used as the pointer input to the vector 2291 // instruction that fuses I with J. 2292 Value *BBVectorize::getReplacementPointerInput(LLVMContext& Context, 2293 Instruction *I, Instruction *J, unsigned o) { 2294 Value *IPtr, *JPtr; 2295 unsigned IAlignment, JAlignment, IAddressSpace, JAddressSpace; 2296 int64_t OffsetInElmts; 2297 2298 // Note: the analysis might fail here, that is why the pair order has 2299 // been precomputed (OffsetInElmts must be unused here). 2300 (void) getPairPtrInfo(I, J, IPtr, JPtr, IAlignment, JAlignment, 2301 IAddressSpace, JAddressSpace, 2302 OffsetInElmts, false); 2303 2304 // The pointer value is taken to be the one with the lowest offset. 2305 Value *VPtr = IPtr; 2306 2307 Type *ArgTypeI = IPtr->getType()->getPointerElementType(); 2308 Type *ArgTypeJ = JPtr->getType()->getPointerElementType(); 2309 Type *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ); 2310 Type *VArgPtrType 2311 = PointerType::get(VArgType, 2312 IPtr->getType()->getPointerAddressSpace()); 2313 return new BitCastInst(VPtr, VArgPtrType, getReplacementName(I, true, o), 2314 /* insert before */ I); 2315 } 2316 2317 void BBVectorize::fillNewShuffleMask(LLVMContext& Context, Instruction *J, 2318 unsigned MaskOffset, unsigned NumInElem, 2319 unsigned NumInElem1, unsigned IdxOffset, 2320 std::vector<Constant*> &Mask) { 2321 unsigned NumElem1 = J->getType()->getVectorNumElements(); 2322 for (unsigned v = 0; v < NumElem1; ++v) { 2323 int m = cast<ShuffleVectorInst>(J)->getMaskValue(v); 2324 if (m < 0) { 2325 Mask[v+MaskOffset] = UndefValue::get(Type::getInt32Ty(Context)); 2326 } else { 2327 unsigned mm = m + (int) IdxOffset; 2328 if (m >= (int) NumInElem1) 2329 mm += (int) NumInElem; 2330 2331 Mask[v+MaskOffset] = 2332 ConstantInt::get(Type::getInt32Ty(Context), mm); 2333 } 2334 } 2335 } 2336 2337 // Returns the value that is to be used as the vector-shuffle mask to the 2338 // vector instruction that fuses I with J. 2339 Value *BBVectorize::getReplacementShuffleMask(LLVMContext& Context, 2340 Instruction *I, Instruction *J) { 2341 // This is the shuffle mask. We need to append the second 2342 // mask to the first, and the numbers need to be adjusted. 2343 2344 Type *ArgTypeI = I->getType(); 2345 Type *ArgTypeJ = J->getType(); 2346 Type *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ); 2347 2348 unsigned NumElemI = ArgTypeI->getVectorNumElements(); 2349 2350 // Get the total number of elements in the fused vector type. 2351 // By definition, this must equal the number of elements in 2352 // the final mask. 2353 unsigned NumElem = VArgType->getVectorNumElements(); 2354 std::vector<Constant*> Mask(NumElem); 2355 2356 Type *OpTypeI = I->getOperand(0)->getType(); 2357 unsigned NumInElemI = OpTypeI->getVectorNumElements(); 2358 Type *OpTypeJ = J->getOperand(0)->getType(); 2359 unsigned NumInElemJ = OpTypeJ->getVectorNumElements(); 2360 2361 // The fused vector will be: 2362 // ----------------------------------------------------- 2363 // | NumInElemI | NumInElemJ | NumInElemI | NumInElemJ | 2364 // ----------------------------------------------------- 2365 // from which we'll extract NumElem total elements (where the first NumElemI 2366 // of them come from the mask in I and the remainder come from the mask 2367 // in J. 2368 2369 // For the mask from the first pair... 2370 fillNewShuffleMask(Context, I, 0, NumInElemJ, NumInElemI, 2371 0, Mask); 2372 2373 // For the mask from the second pair... 2374 fillNewShuffleMask(Context, J, NumElemI, NumInElemI, NumInElemJ, 2375 NumInElemI, Mask); 2376 2377 return ConstantVector::get(Mask); 2378 } 2379 2380 bool BBVectorize::expandIEChain(LLVMContext& Context, Instruction *I, 2381 Instruction *J, unsigned o, Value *&LOp, 2382 unsigned numElemL, 2383 Type *ArgTypeL, Type *ArgTypeH, 2384 bool IBeforeJ, unsigned IdxOff) { 2385 bool ExpandedIEChain = false; 2386 if (InsertElementInst *LIE = dyn_cast<InsertElementInst>(LOp)) { 2387 // If we have a pure insertelement chain, then this can be rewritten 2388 // into a chain that directly builds the larger type. 2389 if (isPureIEChain(LIE)) { 2390 SmallVector<Value *, 8> VectElemts(numElemL, 2391 UndefValue::get(ArgTypeL->getScalarType())); 2392 InsertElementInst *LIENext = LIE; 2393 do { 2394 unsigned Idx = 2395 cast<ConstantInt>(LIENext->getOperand(2))->getSExtValue(); 2396 VectElemts[Idx] = LIENext->getOperand(1); 2397 } while ((LIENext = 2398 dyn_cast<InsertElementInst>(LIENext->getOperand(0)))); 2399 2400 LIENext = nullptr; 2401 Value *LIEPrev = UndefValue::get(ArgTypeH); 2402 for (unsigned i = 0; i < numElemL; ++i) { 2403 if (isa<UndefValue>(VectElemts[i])) continue; 2404 LIENext = InsertElementInst::Create(LIEPrev, VectElemts[i], 2405 ConstantInt::get(Type::getInt32Ty(Context), 2406 i + IdxOff), 2407 getReplacementName(IBeforeJ ? I : J, 2408 true, o, i+1)); 2409 LIENext->insertBefore(IBeforeJ ? J : I); 2410 LIEPrev = LIENext; 2411 } 2412 2413 LOp = LIENext ? (Value*) LIENext : UndefValue::get(ArgTypeH); 2414 ExpandedIEChain = true; 2415 } 2416 } 2417 2418 return ExpandedIEChain; 2419 } 2420 2421 static unsigned getNumScalarElements(Type *Ty) { 2422 if (VectorType *VecTy = dyn_cast<VectorType>(Ty)) 2423 return VecTy->getNumElements(); 2424 return 1; 2425 } 2426 2427 // Returns the value to be used as the specified operand of the vector 2428 // instruction that fuses I with J. 2429 Value *BBVectorize::getReplacementInput(LLVMContext& Context, Instruction *I, 2430 Instruction *J, unsigned o, bool IBeforeJ) { 2431 Value *CV0 = ConstantInt::get(Type::getInt32Ty(Context), 0); 2432 Value *CV1 = ConstantInt::get(Type::getInt32Ty(Context), 1); 2433 2434 // Compute the fused vector type for this operand 2435 Type *ArgTypeI = I->getOperand(o)->getType(); 2436 Type *ArgTypeJ = J->getOperand(o)->getType(); 2437 VectorType *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ); 2438 2439 Instruction *L = I, *H = J; 2440 Type *ArgTypeL = ArgTypeI, *ArgTypeH = ArgTypeJ; 2441 2442 unsigned numElemL = getNumScalarElements(ArgTypeL); 2443 unsigned numElemH = getNumScalarElements(ArgTypeH); 2444 2445 Value *LOp = L->getOperand(o); 2446 Value *HOp = H->getOperand(o); 2447 unsigned numElem = VArgType->getNumElements(); 2448 2449 // First, we check if we can reuse the "original" vector outputs (if these 2450 // exist). We might need a shuffle. 2451 ExtractElementInst *LEE = dyn_cast<ExtractElementInst>(LOp); 2452 ExtractElementInst *HEE = dyn_cast<ExtractElementInst>(HOp); 2453 ShuffleVectorInst *LSV = dyn_cast<ShuffleVectorInst>(LOp); 2454 ShuffleVectorInst *HSV = dyn_cast<ShuffleVectorInst>(HOp); 2455 2456 // FIXME: If we're fusing shuffle instructions, then we can't apply this 2457 // optimization. The input vectors to the shuffle might be a different 2458 // length from the shuffle outputs. Unfortunately, the replacement 2459 // shuffle mask has already been formed, and the mask entries are sensitive 2460 // to the sizes of the inputs. 2461 bool IsSizeChangeShuffle = 2462 isa<ShuffleVectorInst>(L) && 2463 (LOp->getType() != L->getType() || HOp->getType() != H->getType()); 2464 2465 if ((LEE || LSV) && (HEE || HSV) && !IsSizeChangeShuffle) { 2466 // We can have at most two unique vector inputs. 2467 bool CanUseInputs = true; 2468 Value *I1, *I2 = nullptr; 2469 if (LEE) { 2470 I1 = LEE->getOperand(0); 2471 } else { 2472 I1 = LSV->getOperand(0); 2473 I2 = LSV->getOperand(1); 2474 if (I2 == I1 || isa<UndefValue>(I2)) 2475 I2 = nullptr; 2476 } 2477 2478 if (HEE) { 2479 Value *I3 = HEE->getOperand(0); 2480 if (!I2 && I3 != I1) 2481 I2 = I3; 2482 else if (I3 != I1 && I3 != I2) 2483 CanUseInputs = false; 2484 } else { 2485 Value *I3 = HSV->getOperand(0); 2486 if (!I2 && I3 != I1) 2487 I2 = I3; 2488 else if (I3 != I1 && I3 != I2) 2489 CanUseInputs = false; 2490 2491 if (CanUseInputs) { 2492 Value *I4 = HSV->getOperand(1); 2493 if (!isa<UndefValue>(I4)) { 2494 if (!I2 && I4 != I1) 2495 I2 = I4; 2496 else if (I4 != I1 && I4 != I2) 2497 CanUseInputs = false; 2498 } 2499 } 2500 } 2501 2502 if (CanUseInputs) { 2503 unsigned LOpElem = 2504 cast<Instruction>(LOp)->getOperand(0)->getType() 2505 ->getVectorNumElements(); 2506 2507 unsigned HOpElem = 2508 cast<Instruction>(HOp)->getOperand(0)->getType() 2509 ->getVectorNumElements(); 2510 2511 // We have one or two input vectors. We need to map each index of the 2512 // operands to the index of the original vector. 2513 SmallVector<std::pair<int, int>, 8> II(numElem); 2514 for (unsigned i = 0; i < numElemL; ++i) { 2515 int Idx, INum; 2516 if (LEE) { 2517 Idx = 2518 cast<ConstantInt>(LEE->getOperand(1))->getSExtValue(); 2519 INum = LEE->getOperand(0) == I1 ? 0 : 1; 2520 } else { 2521 Idx = LSV->getMaskValue(i); 2522 if (Idx < (int) LOpElem) { 2523 INum = LSV->getOperand(0) == I1 ? 0 : 1; 2524 } else { 2525 Idx -= LOpElem; 2526 INum = LSV->getOperand(1) == I1 ? 0 : 1; 2527 } 2528 } 2529 2530 II[i] = std::pair<int, int>(Idx, INum); 2531 } 2532 for (unsigned i = 0; i < numElemH; ++i) { 2533 int Idx, INum; 2534 if (HEE) { 2535 Idx = 2536 cast<ConstantInt>(HEE->getOperand(1))->getSExtValue(); 2537 INum = HEE->getOperand(0) == I1 ? 0 : 1; 2538 } else { 2539 Idx = HSV->getMaskValue(i); 2540 if (Idx < (int) HOpElem) { 2541 INum = HSV->getOperand(0) == I1 ? 0 : 1; 2542 } else { 2543 Idx -= HOpElem; 2544 INum = HSV->getOperand(1) == I1 ? 0 : 1; 2545 } 2546 } 2547 2548 II[i + numElemL] = std::pair<int, int>(Idx, INum); 2549 } 2550 2551 // We now have an array which tells us from which index of which 2552 // input vector each element of the operand comes. 2553 VectorType *I1T = cast<VectorType>(I1->getType()); 2554 unsigned I1Elem = I1T->getNumElements(); 2555 2556 if (!I2) { 2557 // In this case there is only one underlying vector input. Check for 2558 // the trivial case where we can use the input directly. 2559 if (I1Elem == numElem) { 2560 bool ElemInOrder = true; 2561 for (unsigned i = 0; i < numElem; ++i) { 2562 if (II[i].first != (int) i && II[i].first != -1) { 2563 ElemInOrder = false; 2564 break; 2565 } 2566 } 2567 2568 if (ElemInOrder) 2569 return I1; 2570 } 2571 2572 // A shuffle is needed. 2573 std::vector<Constant *> Mask(numElem); 2574 for (unsigned i = 0; i < numElem; ++i) { 2575 int Idx = II[i].first; 2576 if (Idx == -1) 2577 Mask[i] = UndefValue::get(Type::getInt32Ty(Context)); 2578 else 2579 Mask[i] = ConstantInt::get(Type::getInt32Ty(Context), Idx); 2580 } 2581 2582 Instruction *S = 2583 new ShuffleVectorInst(I1, UndefValue::get(I1T), 2584 ConstantVector::get(Mask), 2585 getReplacementName(IBeforeJ ? I : J, 2586 true, o)); 2587 S->insertBefore(IBeforeJ ? J : I); 2588 return S; 2589 } 2590 2591 VectorType *I2T = cast<VectorType>(I2->getType()); 2592 unsigned I2Elem = I2T->getNumElements(); 2593 2594 // This input comes from two distinct vectors. The first step is to 2595 // make sure that both vectors are the same length. If not, the 2596 // smaller one will need to grow before they can be shuffled together. 2597 if (I1Elem < I2Elem) { 2598 std::vector<Constant *> Mask(I2Elem); 2599 unsigned v = 0; 2600 for (; v < I1Elem; ++v) 2601 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2602 for (; v < I2Elem; ++v) 2603 Mask[v] = UndefValue::get(Type::getInt32Ty(Context)); 2604 2605 Instruction *NewI1 = 2606 new ShuffleVectorInst(I1, UndefValue::get(I1T), 2607 ConstantVector::get(Mask), 2608 getReplacementName(IBeforeJ ? I : J, 2609 true, o, 1)); 2610 NewI1->insertBefore(IBeforeJ ? J : I); 2611 I1 = NewI1; 2612 I1T = I2T; 2613 I1Elem = I2Elem; 2614 } else if (I1Elem > I2Elem) { 2615 std::vector<Constant *> Mask(I1Elem); 2616 unsigned v = 0; 2617 for (; v < I2Elem; ++v) 2618 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2619 for (; v < I1Elem; ++v) 2620 Mask[v] = UndefValue::get(Type::getInt32Ty(Context)); 2621 2622 Instruction *NewI2 = 2623 new ShuffleVectorInst(I2, UndefValue::get(I2T), 2624 ConstantVector::get(Mask), 2625 getReplacementName(IBeforeJ ? I : J, 2626 true, o, 1)); 2627 NewI2->insertBefore(IBeforeJ ? J : I); 2628 I2 = NewI2; 2629 I2T = I1T; 2630 I2Elem = I1Elem; 2631 } 2632 2633 // Now that both I1 and I2 are the same length we can shuffle them 2634 // together (and use the result). 2635 std::vector<Constant *> Mask(numElem); 2636 for (unsigned v = 0; v < numElem; ++v) { 2637 if (II[v].first == -1) { 2638 Mask[v] = UndefValue::get(Type::getInt32Ty(Context)); 2639 } else { 2640 int Idx = II[v].first + II[v].second * I1Elem; 2641 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), Idx); 2642 } 2643 } 2644 2645 Instruction *NewOp = 2646 new ShuffleVectorInst(I1, I2, ConstantVector::get(Mask), 2647 getReplacementName(IBeforeJ ? I : J, true, o)); 2648 NewOp->insertBefore(IBeforeJ ? J : I); 2649 return NewOp; 2650 } 2651 } 2652 2653 Type *ArgType = ArgTypeL; 2654 if (numElemL < numElemH) { 2655 if (numElemL == 1 && expandIEChain(Context, I, J, o, HOp, numElemH, 2656 ArgTypeL, VArgType, IBeforeJ, 1)) { 2657 // This is another short-circuit case: we're combining a scalar into 2658 // a vector that is formed by an IE chain. We've just expanded the IE 2659 // chain, now insert the scalar and we're done. 2660 2661 Instruction *S = InsertElementInst::Create(HOp, LOp, CV0, 2662 getReplacementName(IBeforeJ ? I : J, true, o)); 2663 S->insertBefore(IBeforeJ ? J : I); 2664 return S; 2665 } else if (!expandIEChain(Context, I, J, o, LOp, numElemL, ArgTypeL, 2666 ArgTypeH, IBeforeJ)) { 2667 // The two vector inputs to the shuffle must be the same length, 2668 // so extend the smaller vector to be the same length as the larger one. 2669 Instruction *NLOp; 2670 if (numElemL > 1) { 2671 2672 std::vector<Constant *> Mask(numElemH); 2673 unsigned v = 0; 2674 for (; v < numElemL; ++v) 2675 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2676 for (; v < numElemH; ++v) 2677 Mask[v] = UndefValue::get(Type::getInt32Ty(Context)); 2678 2679 NLOp = new ShuffleVectorInst(LOp, UndefValue::get(ArgTypeL), 2680 ConstantVector::get(Mask), 2681 getReplacementName(IBeforeJ ? I : J, 2682 true, o, 1)); 2683 } else { 2684 NLOp = InsertElementInst::Create(UndefValue::get(ArgTypeH), LOp, CV0, 2685 getReplacementName(IBeforeJ ? I : J, 2686 true, o, 1)); 2687 } 2688 2689 NLOp->insertBefore(IBeforeJ ? J : I); 2690 LOp = NLOp; 2691 } 2692 2693 ArgType = ArgTypeH; 2694 } else if (numElemL > numElemH) { 2695 if (numElemH == 1 && expandIEChain(Context, I, J, o, LOp, numElemL, 2696 ArgTypeH, VArgType, IBeforeJ)) { 2697 Instruction *S = 2698 InsertElementInst::Create(LOp, HOp, 2699 ConstantInt::get(Type::getInt32Ty(Context), 2700 numElemL), 2701 getReplacementName(IBeforeJ ? I : J, 2702 true, o)); 2703 S->insertBefore(IBeforeJ ? J : I); 2704 return S; 2705 } else if (!expandIEChain(Context, I, J, o, HOp, numElemH, ArgTypeH, 2706 ArgTypeL, IBeforeJ)) { 2707 Instruction *NHOp; 2708 if (numElemH > 1) { 2709 std::vector<Constant *> Mask(numElemL); 2710 unsigned v = 0; 2711 for (; v < numElemH; ++v) 2712 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2713 for (; v < numElemL; ++v) 2714 Mask[v] = UndefValue::get(Type::getInt32Ty(Context)); 2715 2716 NHOp = new ShuffleVectorInst(HOp, UndefValue::get(ArgTypeH), 2717 ConstantVector::get(Mask), 2718 getReplacementName(IBeforeJ ? I : J, 2719 true, o, 1)); 2720 } else { 2721 NHOp = InsertElementInst::Create(UndefValue::get(ArgTypeL), HOp, CV0, 2722 getReplacementName(IBeforeJ ? I : J, 2723 true, o, 1)); 2724 } 2725 2726 NHOp->insertBefore(IBeforeJ ? J : I); 2727 HOp = NHOp; 2728 } 2729 } 2730 2731 if (ArgType->isVectorTy()) { 2732 unsigned numElem = VArgType->getVectorNumElements(); 2733 std::vector<Constant*> Mask(numElem); 2734 for (unsigned v = 0; v < numElem; ++v) { 2735 unsigned Idx = v; 2736 // If the low vector was expanded, we need to skip the extra 2737 // undefined entries. 2738 if (v >= numElemL && numElemH > numElemL) 2739 Idx += (numElemH - numElemL); 2740 Mask[v] = ConstantInt::get(Type::getInt32Ty(Context), Idx); 2741 } 2742 2743 Instruction *BV = new ShuffleVectorInst(LOp, HOp, 2744 ConstantVector::get(Mask), 2745 getReplacementName(IBeforeJ ? I : J, true, o)); 2746 BV->insertBefore(IBeforeJ ? J : I); 2747 return BV; 2748 } 2749 2750 Instruction *BV1 = InsertElementInst::Create( 2751 UndefValue::get(VArgType), LOp, CV0, 2752 getReplacementName(IBeforeJ ? I : J, 2753 true, o, 1)); 2754 BV1->insertBefore(IBeforeJ ? J : I); 2755 Instruction *BV2 = InsertElementInst::Create(BV1, HOp, CV1, 2756 getReplacementName(IBeforeJ ? I : J, 2757 true, o, 2)); 2758 BV2->insertBefore(IBeforeJ ? J : I); 2759 return BV2; 2760 } 2761 2762 // This function creates an array of values that will be used as the inputs 2763 // to the vector instruction that fuses I with J. 2764 void BBVectorize::getReplacementInputsForPair(LLVMContext& Context, 2765 Instruction *I, Instruction *J, 2766 SmallVectorImpl<Value *> &ReplacedOperands, 2767 bool IBeforeJ) { 2768 unsigned NumOperands = I->getNumOperands(); 2769 2770 for (unsigned p = 0, o = NumOperands-1; p < NumOperands; ++p, --o) { 2771 // Iterate backward so that we look at the store pointer 2772 // first and know whether or not we need to flip the inputs. 2773 2774 if (isa<LoadInst>(I) || (o == 1 && isa<StoreInst>(I))) { 2775 // This is the pointer for a load/store instruction. 2776 ReplacedOperands[o] = getReplacementPointerInput(Context, I, J, o); 2777 continue; 2778 } else if (isa<CallInst>(I)) { 2779 Function *F = cast<CallInst>(I)->getCalledFunction(); 2780 Intrinsic::ID IID = (Intrinsic::ID) F->getIntrinsicID(); 2781 if (o == NumOperands-1) { 2782 BasicBlock &BB = *I->getParent(); 2783 2784 Module *M = BB.getParent()->getParent(); 2785 Type *ArgTypeI = I->getType(); 2786 Type *ArgTypeJ = J->getType(); 2787 Type *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ); 2788 2789 ReplacedOperands[o] = Intrinsic::getDeclaration(M, IID, VArgType); 2790 continue; 2791 } else if ((IID == Intrinsic::powi || IID == Intrinsic::ctlz || 2792 IID == Intrinsic::cttz) && o == 1) { 2793 // The second argument of powi/ctlz/cttz is a single integer/constant 2794 // and we've already checked that both arguments are equal. 2795 // As a result, we just keep I's second argument. 2796 ReplacedOperands[o] = I->getOperand(o); 2797 continue; 2798 } 2799 } else if (isa<ShuffleVectorInst>(I) && o == NumOperands-1) { 2800 ReplacedOperands[o] = getReplacementShuffleMask(Context, I, J); 2801 continue; 2802 } 2803 2804 ReplacedOperands[o] = getReplacementInput(Context, I, J, o, IBeforeJ); 2805 } 2806 } 2807 2808 // This function creates two values that represent the outputs of the 2809 // original I and J instructions. These are generally vector shuffles 2810 // or extracts. In many cases, these will end up being unused and, thus, 2811 // eliminated by later passes. 2812 void BBVectorize::replaceOutputsOfPair(LLVMContext& Context, Instruction *I, 2813 Instruction *J, Instruction *K, 2814 Instruction *&InsertionPt, 2815 Instruction *&K1, Instruction *&K2) { 2816 if (isa<StoreInst>(I)) { 2817 AA->replaceWithNewValue(I, K); 2818 AA->replaceWithNewValue(J, K); 2819 } else { 2820 Type *IType = I->getType(); 2821 Type *JType = J->getType(); 2822 2823 VectorType *VType = getVecTypeForPair(IType, JType); 2824 unsigned numElem = VType->getNumElements(); 2825 2826 unsigned numElemI = getNumScalarElements(IType); 2827 unsigned numElemJ = getNumScalarElements(JType); 2828 2829 if (IType->isVectorTy()) { 2830 std::vector<Constant*> Mask1(numElemI), Mask2(numElemI); 2831 for (unsigned v = 0; v < numElemI; ++v) { 2832 Mask1[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2833 Mask2[v] = ConstantInt::get(Type::getInt32Ty(Context), numElemJ+v); 2834 } 2835 2836 K1 = new ShuffleVectorInst(K, UndefValue::get(VType), 2837 ConstantVector::get( Mask1), 2838 getReplacementName(K, false, 1)); 2839 } else { 2840 Value *CV0 = ConstantInt::get(Type::getInt32Ty(Context), 0); 2841 K1 = ExtractElementInst::Create(K, CV0, 2842 getReplacementName(K, false, 1)); 2843 } 2844 2845 if (JType->isVectorTy()) { 2846 std::vector<Constant*> Mask1(numElemJ), Mask2(numElemJ); 2847 for (unsigned v = 0; v < numElemJ; ++v) { 2848 Mask1[v] = ConstantInt::get(Type::getInt32Ty(Context), v); 2849 Mask2[v] = ConstantInt::get(Type::getInt32Ty(Context), numElemI+v); 2850 } 2851 2852 K2 = new ShuffleVectorInst(K, UndefValue::get(VType), 2853 ConstantVector::get( Mask2), 2854 getReplacementName(K, false, 2)); 2855 } else { 2856 Value *CV1 = ConstantInt::get(Type::getInt32Ty(Context), numElem-1); 2857 K2 = ExtractElementInst::Create(K, CV1, 2858 getReplacementName(K, false, 2)); 2859 } 2860 2861 K1->insertAfter(K); 2862 K2->insertAfter(K1); 2863 InsertionPt = K2; 2864 } 2865 } 2866 2867 // Move all uses of the function I (including pairing-induced uses) after J. 2868 bool BBVectorize::canMoveUsesOfIAfterJ(BasicBlock &BB, 2869 DenseSet<ValuePair> &LoadMoveSetPairs, 2870 Instruction *I, Instruction *J) { 2871 // Skip to the first instruction past I. 2872 BasicBlock::iterator L = std::next(BasicBlock::iterator(I)); 2873 2874 DenseSet<Value *> Users; 2875 AliasSetTracker WriteSet(*AA); 2876 if (I->mayWriteToMemory()) WriteSet.add(I); 2877 2878 for (; cast<Instruction>(L) != J; ++L) 2879 (void) trackUsesOfI(Users, WriteSet, I, L, true, &LoadMoveSetPairs); 2880 2881 assert(cast<Instruction>(L) == J && 2882 "Tracking has not proceeded far enough to check for dependencies"); 2883 // If J is now in the use set of I, then trackUsesOfI will return true 2884 // and we have a dependency cycle (and the fusing operation must abort). 2885 return !trackUsesOfI(Users, WriteSet, I, J, true, &LoadMoveSetPairs); 2886 } 2887 2888 // Move all uses of the function I (including pairing-induced uses) after J. 2889 void BBVectorize::moveUsesOfIAfterJ(BasicBlock &BB, 2890 DenseSet<ValuePair> &LoadMoveSetPairs, 2891 Instruction *&InsertionPt, 2892 Instruction *I, Instruction *J) { 2893 // Skip to the first instruction past I. 2894 BasicBlock::iterator L = std::next(BasicBlock::iterator(I)); 2895 2896 DenseSet<Value *> Users; 2897 AliasSetTracker WriteSet(*AA); 2898 if (I->mayWriteToMemory()) WriteSet.add(I); 2899 2900 for (; cast<Instruction>(L) != J;) { 2901 if (trackUsesOfI(Users, WriteSet, I, L, true, &LoadMoveSetPairs)) { 2902 // Move this instruction 2903 Instruction *InstToMove = L; ++L; 2904 2905 DEBUG(dbgs() << "BBV: moving: " << *InstToMove << 2906 " to after " << *InsertionPt << "\n"); 2907 InstToMove->removeFromParent(); 2908 InstToMove->insertAfter(InsertionPt); 2909 InsertionPt = InstToMove; 2910 } else { 2911 ++L; 2912 } 2913 } 2914 } 2915 2916 // Collect all load instruction that are in the move set of a given first 2917 // pair member. These loads depend on the first instruction, I, and so need 2918 // to be moved after J (the second instruction) when the pair is fused. 2919 void BBVectorize::collectPairLoadMoveSet(BasicBlock &BB, 2920 DenseMap<Value *, Value *> &ChosenPairs, 2921 DenseMap<Value *, std::vector<Value *> > &LoadMoveSet, 2922 DenseSet<ValuePair> &LoadMoveSetPairs, 2923 Instruction *I) { 2924 // Skip to the first instruction past I. 2925 BasicBlock::iterator L = std::next(BasicBlock::iterator(I)); 2926 2927 DenseSet<Value *> Users; 2928 AliasSetTracker WriteSet(*AA); 2929 if (I->mayWriteToMemory()) WriteSet.add(I); 2930 2931 // Note: We cannot end the loop when we reach J because J could be moved 2932 // farther down the use chain by another instruction pairing. Also, J 2933 // could be before I if this is an inverted input. 2934 for (BasicBlock::iterator E = BB.end(); cast<Instruction>(L) != E; ++L) { 2935 if (trackUsesOfI(Users, WriteSet, I, L)) { 2936 if (L->mayReadFromMemory()) { 2937 LoadMoveSet[L].push_back(I); 2938 LoadMoveSetPairs.insert(ValuePair(L, I)); 2939 } 2940 } 2941 } 2942 } 2943 2944 // In cases where both load/stores and the computation of their pointers 2945 // are chosen for vectorization, we can end up in a situation where the 2946 // aliasing analysis starts returning different query results as the 2947 // process of fusing instruction pairs continues. Because the algorithm 2948 // relies on finding the same use dags here as were found earlier, we'll 2949 // need to precompute the necessary aliasing information here and then 2950 // manually update it during the fusion process. 2951 void BBVectorize::collectLoadMoveSet(BasicBlock &BB, 2952 std::vector<Value *> &PairableInsts, 2953 DenseMap<Value *, Value *> &ChosenPairs, 2954 DenseMap<Value *, std::vector<Value *> > &LoadMoveSet, 2955 DenseSet<ValuePair> &LoadMoveSetPairs) { 2956 for (std::vector<Value *>::iterator PI = PairableInsts.begin(), 2957 PIE = PairableInsts.end(); PI != PIE; ++PI) { 2958 DenseMap<Value *, Value *>::iterator P = ChosenPairs.find(*PI); 2959 if (P == ChosenPairs.end()) continue; 2960 2961 Instruction *I = cast<Instruction>(P->first); 2962 collectPairLoadMoveSet(BB, ChosenPairs, LoadMoveSet, 2963 LoadMoveSetPairs, I); 2964 } 2965 } 2966 2967 // When the first instruction in each pair is cloned, it will inherit its 2968 // parent's metadata. This metadata must be combined with that of the other 2969 // instruction in a safe way. 2970 void BBVectorize::combineMetadata(Instruction *K, const Instruction *J) { 2971 SmallVector<std::pair<unsigned, MDNode*>, 4> Metadata; 2972 K->getAllMetadataOtherThanDebugLoc(Metadata); 2973 for (unsigned i = 0, n = Metadata.size(); i < n; ++i) { 2974 unsigned Kind = Metadata[i].first; 2975 MDNode *JMD = J->getMetadata(Kind); 2976 MDNode *KMD = Metadata[i].second; 2977 2978 switch (Kind) { 2979 default: 2980 K->setMetadata(Kind, nullptr); // Remove unknown metadata 2981 break; 2982 case LLVMContext::MD_tbaa: 2983 K->setMetadata(Kind, MDNode::getMostGenericTBAA(JMD, KMD)); 2984 break; 2985 case LLVMContext::MD_fpmath: 2986 K->setMetadata(Kind, MDNode::getMostGenericFPMath(JMD, KMD)); 2987 break; 2988 } 2989 } 2990 } 2991 2992 // This function fuses the chosen instruction pairs into vector instructions, 2993 // taking care preserve any needed scalar outputs and, then, it reorders the 2994 // remaining instructions as needed (users of the first member of the pair 2995 // need to be moved to after the location of the second member of the pair 2996 // because the vector instruction is inserted in the location of the pair's 2997 // second member). 2998 void BBVectorize::fuseChosenPairs(BasicBlock &BB, 2999 std::vector<Value *> &PairableInsts, 3000 DenseMap<Value *, Value *> &ChosenPairs, 3001 DenseSet<ValuePair> &FixedOrderPairs, 3002 DenseMap<VPPair, unsigned> &PairConnectionTypes, 3003 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairs, 3004 DenseMap<ValuePair, std::vector<ValuePair> > &ConnectedPairDeps) { 3005 LLVMContext& Context = BB.getContext(); 3006 3007 // During the vectorization process, the order of the pairs to be fused 3008 // could be flipped. So we'll add each pair, flipped, into the ChosenPairs 3009 // list. After a pair is fused, the flipped pair is removed from the list. 3010 DenseSet<ValuePair> FlippedPairs; 3011 for (DenseMap<Value *, Value *>::iterator P = ChosenPairs.begin(), 3012 E = ChosenPairs.end(); P != E; ++P) 3013 FlippedPairs.insert(ValuePair(P->second, P->first)); 3014 for (DenseSet<ValuePair>::iterator P = FlippedPairs.begin(), 3015 E = FlippedPairs.end(); P != E; ++P) 3016 ChosenPairs.insert(*P); 3017 3018 DenseMap<Value *, std::vector<Value *> > LoadMoveSet; 3019 DenseSet<ValuePair> LoadMoveSetPairs; 3020 collectLoadMoveSet(BB, PairableInsts, ChosenPairs, 3021 LoadMoveSet, LoadMoveSetPairs); 3022 3023 DEBUG(dbgs() << "BBV: initial: \n" << BB << "\n"); 3024 3025 for (BasicBlock::iterator PI = BB.getFirstInsertionPt(); PI != BB.end();) { 3026 DenseMap<Value *, Value *>::iterator P = ChosenPairs.find(PI); 3027 if (P == ChosenPairs.end()) { 3028 ++PI; 3029 continue; 3030 } 3031 3032 if (getDepthFactor(P->first) == 0) { 3033 // These instructions are not really fused, but are tracked as though 3034 // they are. Any case in which it would be interesting to fuse them 3035 // will be taken care of by InstCombine. 3036 --NumFusedOps; 3037 ++PI; 3038 continue; 3039 } 3040 3041 Instruction *I = cast<Instruction>(P->first), 3042 *J = cast<Instruction>(P->second); 3043 3044 DEBUG(dbgs() << "BBV: fusing: " << *I << 3045 " <-> " << *J << "\n"); 3046 3047 // Remove the pair and flipped pair from the list. 3048 DenseMap<Value *, Value *>::iterator FP = ChosenPairs.find(P->second); 3049 assert(FP != ChosenPairs.end() && "Flipped pair not found in list"); 3050 ChosenPairs.erase(FP); 3051 ChosenPairs.erase(P); 3052 3053 if (!canMoveUsesOfIAfterJ(BB, LoadMoveSetPairs, I, J)) { 3054 DEBUG(dbgs() << "BBV: fusion of: " << *I << 3055 " <-> " << *J << 3056 " aborted because of non-trivial dependency cycle\n"); 3057 --NumFusedOps; 3058 ++PI; 3059 continue; 3060 } 3061 3062 // If the pair must have the other order, then flip it. 3063 bool FlipPairOrder = FixedOrderPairs.count(ValuePair(J, I)); 3064 if (!FlipPairOrder && !FixedOrderPairs.count(ValuePair(I, J))) { 3065 // This pair does not have a fixed order, and so we might want to 3066 // flip it if that will yield fewer shuffles. We count the number 3067 // of dependencies connected via swaps, and those directly connected, 3068 // and flip the order if the number of swaps is greater. 3069 bool OrigOrder = true; 3070 DenseMap<ValuePair, std::vector<ValuePair> >::iterator IJ = 3071 ConnectedPairDeps.find(ValuePair(I, J)); 3072 if (IJ == ConnectedPairDeps.end()) { 3073 IJ = ConnectedPairDeps.find(ValuePair(J, I)); 3074 OrigOrder = false; 3075 } 3076 3077 if (IJ != ConnectedPairDeps.end()) { 3078 unsigned NumDepsDirect = 0, NumDepsSwap = 0; 3079 for (std::vector<ValuePair>::iterator T = IJ->second.begin(), 3080 TE = IJ->second.end(); T != TE; ++T) { 3081 VPPair Q(IJ->first, *T); 3082 DenseMap<VPPair, unsigned>::iterator R = 3083 PairConnectionTypes.find(VPPair(Q.second, Q.first)); 3084 assert(R != PairConnectionTypes.end() && 3085 "Cannot find pair connection type"); 3086 if (R->second == PairConnectionDirect) 3087 ++NumDepsDirect; 3088 else if (R->second == PairConnectionSwap) 3089 ++NumDepsSwap; 3090 } 3091 3092 if (!OrigOrder) 3093 std::swap(NumDepsDirect, NumDepsSwap); 3094 3095 if (NumDepsSwap > NumDepsDirect) { 3096 FlipPairOrder = true; 3097 DEBUG(dbgs() << "BBV: reordering pair: " << *I << 3098 " <-> " << *J << "\n"); 3099 } 3100 } 3101 } 3102 3103 Instruction *L = I, *H = J; 3104 if (FlipPairOrder) 3105 std::swap(H, L); 3106 3107 // If the pair being fused uses the opposite order from that in the pair 3108 // connection map, then we need to flip the types. 3109 DenseMap<ValuePair, std::vector<ValuePair> >::iterator HL = 3110 ConnectedPairs.find(ValuePair(H, L)); 3111 if (HL != ConnectedPairs.end()) 3112 for (std::vector<ValuePair>::iterator T = HL->second.begin(), 3113 TE = HL->second.end(); T != TE; ++T) { 3114 VPPair Q(HL->first, *T); 3115 DenseMap<VPPair, unsigned>::iterator R = PairConnectionTypes.find(Q); 3116 assert(R != PairConnectionTypes.end() && 3117 "Cannot find pair connection type"); 3118 if (R->second == PairConnectionDirect) 3119 R->second = PairConnectionSwap; 3120 else if (R->second == PairConnectionSwap) 3121 R->second = PairConnectionDirect; 3122 } 3123 3124 bool LBeforeH = !FlipPairOrder; 3125 unsigned NumOperands = I->getNumOperands(); 3126 SmallVector<Value *, 3> ReplacedOperands(NumOperands); 3127 getReplacementInputsForPair(Context, L, H, ReplacedOperands, 3128 LBeforeH); 3129 3130 // Make a copy of the original operation, change its type to the vector 3131 // type and replace its operands with the vector operands. 3132 Instruction *K = L->clone(); 3133 if (L->hasName()) 3134 K->takeName(L); 3135 else if (H->hasName()) 3136 K->takeName(H); 3137 3138 if (!isa<StoreInst>(K)) 3139 K->mutateType(getVecTypeForPair(L->getType(), H->getType())); 3140 3141 combineMetadata(K, H); 3142 K->intersectOptionalDataWith(H); 3143 3144 for (unsigned o = 0; o < NumOperands; ++o) 3145 K->setOperand(o, ReplacedOperands[o]); 3146 3147 K->insertAfter(J); 3148 3149 // Instruction insertion point: 3150 Instruction *InsertionPt = K; 3151 Instruction *K1 = nullptr, *K2 = nullptr; 3152 replaceOutputsOfPair(Context, L, H, K, InsertionPt, K1, K2); 3153 3154 // The use dag of the first original instruction must be moved to after 3155 // the location of the second instruction. The entire use dag of the 3156 // first instruction is disjoint from the input dag of the second 3157 // (by definition), and so commutes with it. 3158 3159 moveUsesOfIAfterJ(BB, LoadMoveSetPairs, InsertionPt, I, J); 3160 3161 if (!isa<StoreInst>(I)) { 3162 L->replaceAllUsesWith(K1); 3163 H->replaceAllUsesWith(K2); 3164 AA->replaceWithNewValue(L, K1); 3165 AA->replaceWithNewValue(H, K2); 3166 } 3167 3168 // Instructions that may read from memory may be in the load move set. 3169 // Once an instruction is fused, we no longer need its move set, and so 3170 // the values of the map never need to be updated. However, when a load 3171 // is fused, we need to merge the entries from both instructions in the 3172 // pair in case those instructions were in the move set of some other 3173 // yet-to-be-fused pair. The loads in question are the keys of the map. 3174 if (I->mayReadFromMemory()) { 3175 std::vector<ValuePair> NewSetMembers; 3176 DenseMap<Value *, std::vector<Value *> >::iterator II = 3177 LoadMoveSet.find(I); 3178 if (II != LoadMoveSet.end()) 3179 for (std::vector<Value *>::iterator N = II->second.begin(), 3180 NE = II->second.end(); N != NE; ++N) 3181 NewSetMembers.push_back(ValuePair(K, *N)); 3182 DenseMap<Value *, std::vector<Value *> >::iterator JJ = 3183 LoadMoveSet.find(J); 3184 if (JJ != LoadMoveSet.end()) 3185 for (std::vector<Value *>::iterator N = JJ->second.begin(), 3186 NE = JJ->second.end(); N != NE; ++N) 3187 NewSetMembers.push_back(ValuePair(K, *N)); 3188 for (std::vector<ValuePair>::iterator A = NewSetMembers.begin(), 3189 AE = NewSetMembers.end(); A != AE; ++A) { 3190 LoadMoveSet[A->first].push_back(A->second); 3191 LoadMoveSetPairs.insert(*A); 3192 } 3193 } 3194 3195 // Before removing I, set the iterator to the next instruction. 3196 PI = std::next(BasicBlock::iterator(I)); 3197 if (cast<Instruction>(PI) == J) 3198 ++PI; 3199 3200 SE->forgetValue(I); 3201 SE->forgetValue(J); 3202 I->eraseFromParent(); 3203 J->eraseFromParent(); 3204 3205 DEBUG(if (PrintAfterEveryPair) dbgs() << "BBV: block is now: \n" << 3206 BB << "\n"); 3207 } 3208 3209 DEBUG(dbgs() << "BBV: final: \n" << BB << "\n"); 3210 } 3211 } 3212 3213 char BBVectorize::ID = 0; 3214 static const char bb_vectorize_name[] = "Basic-Block Vectorization"; 3215 INITIALIZE_PASS_BEGIN(BBVectorize, BBV_NAME, bb_vectorize_name, false, false) 3216 INITIALIZE_AG_DEPENDENCY(AliasAnalysis) 3217 INITIALIZE_AG_DEPENDENCY(TargetTransformInfo) 3218 INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) 3219 INITIALIZE_PASS_DEPENDENCY(ScalarEvolution) 3220 INITIALIZE_PASS_END(BBVectorize, BBV_NAME, bb_vectorize_name, false, false) 3221 3222 BasicBlockPass *llvm::createBBVectorizePass(const VectorizeConfig &C) { 3223 return new BBVectorize(C); 3224 } 3225 3226 bool 3227 llvm::vectorizeBasicBlock(Pass *P, BasicBlock &BB, const VectorizeConfig &C) { 3228 BBVectorize BBVectorizer(P, C); 3229 return BBVectorizer.vectorizeBB(BB); 3230 } 3231 3232 //===----------------------------------------------------------------------===// 3233 VectorizeConfig::VectorizeConfig() { 3234 VectorBits = ::VectorBits; 3235 VectorizeBools = !::NoBools; 3236 VectorizeInts = !::NoInts; 3237 VectorizeFloats = !::NoFloats; 3238 VectorizePointers = !::NoPointers; 3239 VectorizeCasts = !::NoCasts; 3240 VectorizeMath = !::NoMath; 3241 VectorizeBitManipulations = !::NoBitManipulation; 3242 VectorizeFMA = !::NoFMA; 3243 VectorizeSelect = !::NoSelect; 3244 VectorizeCmp = !::NoCmp; 3245 VectorizeGEP = !::NoGEP; 3246 VectorizeMemOps = !::NoMemOps; 3247 AlignedOnly = ::AlignedOnly; 3248 ReqChainDepth= ::ReqChainDepth; 3249 SearchLimit = ::SearchLimit; 3250 MaxCandPairsForCycleCheck = ::MaxCandPairsForCycleCheck; 3251 SplatBreaksChain = ::SplatBreaksChain; 3252 MaxInsts = ::MaxInsts; 3253 MaxPairs = ::MaxPairs; 3254 MaxIter = ::MaxIter; 3255 Pow2LenOnly = ::Pow2LenOnly; 3256 NoMemOpBoost = ::NoMemOpBoost; 3257 FastDep = ::FastDep; 3258 } 3259