1//===--- Scalarizer.cpp - Scalarize vector operations ---------------------===// 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 pass converts vector operations into scalar operations, in order 11// to expose optimization opportunities on the individual scalar operations. 12// It is mainly intended for targets that do not have vector units, but it 13// may also be useful for revectorizing code to different vector widths. 14// 15//===----------------------------------------------------------------------===// 16 17#include "llvm/ADT/STLExtras.h" 18#include "llvm/IR/IRBuilder.h" 19#include "llvm/IR/InstVisitor.h" 20#include "llvm/Pass.h" 21#include "llvm/Support/CommandLine.h" 22#include "llvm/Transforms/Scalar.h" 23#include "llvm/Transforms/Utils/BasicBlockUtils.h" 24 25using namespace llvm; 26 27#define DEBUG_TYPE "scalarizer" 28 29namespace { 30// Used to store the scattered form of a vector. 31typedef SmallVector<Value *, 8> ValueVector; 32 33// Used to map a vector Value to its scattered form. We use std::map 34// because we want iterators to persist across insertion and because the 35// values are relatively large. 36typedef std::map<Value *, ValueVector> ScatterMap; 37 38// Lists Instructions that have been replaced with scalar implementations, 39// along with a pointer to their scattered forms. 40typedef SmallVector<std::pair<Instruction *, ValueVector *>, 16> GatherList; 41 42// Provides a very limited vector-like interface for lazily accessing one 43// component of a scattered vector or vector pointer. 44class Scatterer { 45public: 46 Scatterer() {} 47 48 // Scatter V into Size components. If new instructions are needed, 49 // insert them before BBI in BB. If Cache is nonnull, use it to cache 50 // the results. 51 Scatterer(BasicBlock *bb, BasicBlock::iterator bbi, Value *v, 52 ValueVector *cachePtr = nullptr); 53 54 // Return component I, creating a new Value for it if necessary. 55 Value *operator[](unsigned I); 56 57 // Return the number of components. 58 unsigned size() const { return Size; } 59 60private: 61 BasicBlock *BB; 62 BasicBlock::iterator BBI; 63 Value *V; 64 ValueVector *CachePtr; 65 PointerType *PtrTy; 66 ValueVector Tmp; 67 unsigned Size; 68}; 69 70// FCmpSpliiter(FCI)(Builder, X, Y, Name) uses Builder to create an FCmp 71// called Name that compares X and Y in the same way as FCI. 72struct FCmpSplitter { 73 FCmpSplitter(FCmpInst &fci) : FCI(fci) {} 74 Value *operator()(IRBuilder<> &Builder, Value *Op0, Value *Op1, 75 const Twine &Name) const { 76 return Builder.CreateFCmp(FCI.getPredicate(), Op0, Op1, Name); 77 } 78 FCmpInst &FCI; 79}; 80 81// ICmpSpliiter(ICI)(Builder, X, Y, Name) uses Builder to create an ICmp 82// called Name that compares X and Y in the same way as ICI. 83struct ICmpSplitter { 84 ICmpSplitter(ICmpInst &ici) : ICI(ici) {} 85 Value *operator()(IRBuilder<> &Builder, Value *Op0, Value *Op1, 86 const Twine &Name) const { 87 return Builder.CreateICmp(ICI.getPredicate(), Op0, Op1, Name); 88 } 89 ICmpInst &ICI; 90}; 91 92// BinarySpliiter(BO)(Builder, X, Y, Name) uses Builder to create 93// a binary operator like BO called Name with operands X and Y. 94struct BinarySplitter { 95 BinarySplitter(BinaryOperator &bo) : BO(bo) {} 96 Value *operator()(IRBuilder<> &Builder, Value *Op0, Value *Op1, 97 const Twine &Name) const { 98 return Builder.CreateBinOp(BO.getOpcode(), Op0, Op1, Name); 99 } 100 BinaryOperator &BO; 101}; 102 103// Information about a load or store that we're scalarizing. 104struct VectorLayout { 105 VectorLayout() : VecTy(nullptr), ElemTy(nullptr), VecAlign(0), ElemSize(0) {} 106 107 // Return the alignment of element I. 108 uint64_t getElemAlign(unsigned I) { 109 return MinAlign(VecAlign, I * ElemSize); 110 } 111 112 // The type of the vector. 113 VectorType *VecTy; 114 115 // The type of each element. 116 Type *ElemTy; 117 118 // The alignment of the vector. 119 uint64_t VecAlign; 120 121 // The size of each element. 122 uint64_t ElemSize; 123}; 124 125class Scalarizer : public FunctionPass, 126 public InstVisitor<Scalarizer, bool> { 127public: 128 static char ID; 129 130 Scalarizer() : 131 FunctionPass(ID) { 132 initializeScalarizerPass(*PassRegistry::getPassRegistry()); 133 } 134 135 bool doInitialization(Module &M) override; 136 bool runOnFunction(Function &F) override; 137 138 // InstVisitor methods. They return true if the instruction was scalarized, 139 // false if nothing changed. 140 bool visitInstruction(Instruction &) { return false; } 141 bool visitSelectInst(SelectInst &SI); 142 bool visitICmpInst(ICmpInst &); 143 bool visitFCmpInst(FCmpInst &); 144 bool visitBinaryOperator(BinaryOperator &); 145 bool visitGetElementPtrInst(GetElementPtrInst &); 146 bool visitCastInst(CastInst &); 147 bool visitBitCastInst(BitCastInst &); 148 bool visitShuffleVectorInst(ShuffleVectorInst &); 149 bool visitPHINode(PHINode &); 150 bool visitLoadInst(LoadInst &); 151 bool visitStoreInst(StoreInst &); 152 153 static void registerOptions() { 154 // This is disabled by default because having separate loads and stores 155 // makes it more likely that the -combiner-alias-analysis limits will be 156 // reached. 157 OptionRegistry::registerOption<bool, Scalarizer, 158 &Scalarizer::ScalarizeLoadStore>( 159 "scalarize-load-store", 160 "Allow the scalarizer pass to scalarize loads and store", false); 161 } 162 163private: 164 Scatterer scatter(Instruction *, Value *); 165 void gather(Instruction *, const ValueVector &); 166 bool canTransferMetadata(unsigned Kind); 167 void transferMetadata(Instruction *, const ValueVector &); 168 bool getVectorLayout(Type *, unsigned, VectorLayout &, const DataLayout &); 169 bool finish(); 170 171 template<typename T> bool splitBinary(Instruction &, const T &); 172 173 ScatterMap Scattered; 174 GatherList Gathered; 175 unsigned ParallelLoopAccessMDKind; 176 bool ScalarizeLoadStore; 177}; 178 179char Scalarizer::ID = 0; 180} // end anonymous namespace 181 182INITIALIZE_PASS_WITH_OPTIONS(Scalarizer, "scalarizer", 183 "Scalarize vector operations", false, false) 184 185Scatterer::Scatterer(BasicBlock *bb, BasicBlock::iterator bbi, Value *v, 186 ValueVector *cachePtr) 187 : BB(bb), BBI(bbi), V(v), CachePtr(cachePtr) { 188 Type *Ty = V->getType(); 189 PtrTy = dyn_cast<PointerType>(Ty); 190 if (PtrTy) 191 Ty = PtrTy->getElementType(); 192 Size = Ty->getVectorNumElements(); 193 if (!CachePtr) 194 Tmp.resize(Size, nullptr); 195 else if (CachePtr->empty()) 196 CachePtr->resize(Size, nullptr); 197 else 198 assert(Size == CachePtr->size() && "Inconsistent vector sizes"); 199} 200 201// Return component I, creating a new Value for it if necessary. 202Value *Scatterer::operator[](unsigned I) { 203 ValueVector &CV = (CachePtr ? *CachePtr : Tmp); 204 // Try to reuse a previous value. 205 if (CV[I]) 206 return CV[I]; 207 IRBuilder<> Builder(BB, BBI); 208 if (PtrTy) { 209 if (!CV[0]) { 210 Type *Ty = 211 PointerType::get(PtrTy->getElementType()->getVectorElementType(), 212 PtrTy->getAddressSpace()); 213 CV[0] = Builder.CreateBitCast(V, Ty, V->getName() + ".i0"); 214 } 215 if (I != 0) 216 CV[I] = Builder.CreateConstGEP1_32(nullptr, CV[0], I, 217 V->getName() + ".i" + Twine(I)); 218 } else { 219 // Search through a chain of InsertElementInsts looking for element I. 220 // Record other elements in the cache. The new V is still suitable 221 // for all uncached indices. 222 for (;;) { 223 InsertElementInst *Insert = dyn_cast<InsertElementInst>(V); 224 if (!Insert) 225 break; 226 ConstantInt *Idx = dyn_cast<ConstantInt>(Insert->getOperand(2)); 227 if (!Idx) 228 break; 229 unsigned J = Idx->getZExtValue(); 230 V = Insert->getOperand(0); 231 if (I == J) { 232 CV[J] = Insert->getOperand(1); 233 return CV[J]; 234 } else if (!CV[J]) { 235 // Only cache the first entry we find for each index we're not actively 236 // searching for. This prevents us from going too far up the chain and 237 // caching incorrect entries. 238 CV[J] = Insert->getOperand(1); 239 } 240 } 241 CV[I] = Builder.CreateExtractElement(V, Builder.getInt32(I), 242 V->getName() + ".i" + Twine(I)); 243 } 244 return CV[I]; 245} 246 247bool Scalarizer::doInitialization(Module &M) { 248 ParallelLoopAccessMDKind = 249 M.getContext().getMDKindID("llvm.mem.parallel_loop_access"); 250 ScalarizeLoadStore = 251 M.getContext().getOption<bool, Scalarizer, &Scalarizer::ScalarizeLoadStore>(); 252 return false; 253} 254 255bool Scalarizer::runOnFunction(Function &F) { 256 assert(Gathered.empty() && Scattered.empty()); 257 for (BasicBlock &BB : F) { 258 for (BasicBlock::iterator II = BB.begin(), IE = BB.end(); II != IE;) { 259 Instruction *I = &*II; 260 bool Done = visit(I); 261 ++II; 262 if (Done && I->getType()->isVoidTy()) 263 I->eraseFromParent(); 264 } 265 } 266 return finish(); 267} 268 269// Return a scattered form of V that can be accessed by Point. V must be a 270// vector or a pointer to a vector. 271Scatterer Scalarizer::scatter(Instruction *Point, Value *V) { 272 if (Argument *VArg = dyn_cast<Argument>(V)) { 273 // Put the scattered form of arguments in the entry block, 274 // so that it can be used everywhere. 275 Function *F = VArg->getParent(); 276 BasicBlock *BB = &F->getEntryBlock(); 277 return Scatterer(BB, BB->begin(), V, &Scattered[V]); 278 } 279 if (Instruction *VOp = dyn_cast<Instruction>(V)) { 280 // Put the scattered form of an instruction directly after the 281 // instruction. 282 BasicBlock *BB = VOp->getParent(); 283 return Scatterer(BB, std::next(BasicBlock::iterator(VOp)), 284 V, &Scattered[V]); 285 } 286 // In the fallback case, just put the scattered before Point and 287 // keep the result local to Point. 288 return Scatterer(Point->getParent(), Point->getIterator(), V); 289} 290 291// Replace Op with the gathered form of the components in CV. Defer the 292// deletion of Op and creation of the gathered form to the end of the pass, 293// so that we can avoid creating the gathered form if all uses of Op are 294// replaced with uses of CV. 295void Scalarizer::gather(Instruction *Op, const ValueVector &CV) { 296 // Since we're not deleting Op yet, stub out its operands, so that it 297 // doesn't make anything live unnecessarily. 298 for (unsigned I = 0, E = Op->getNumOperands(); I != E; ++I) 299 Op->setOperand(I, UndefValue::get(Op->getOperand(I)->getType())); 300 301 transferMetadata(Op, CV); 302 303 // If we already have a scattered form of Op (created from ExtractElements 304 // of Op itself), replace them with the new form. 305 ValueVector &SV = Scattered[Op]; 306 if (!SV.empty()) { 307 for (unsigned I = 0, E = SV.size(); I != E; ++I) { 308 Instruction *Old = cast<Instruction>(SV[I]); 309 CV[I]->takeName(Old); 310 Old->replaceAllUsesWith(CV[I]); 311 Old->eraseFromParent(); 312 } 313 } 314 SV = CV; 315 Gathered.push_back(GatherList::value_type(Op, &SV)); 316} 317 318// Return true if it is safe to transfer the given metadata tag from 319// vector to scalar instructions. 320bool Scalarizer::canTransferMetadata(unsigned Tag) { 321 return (Tag == LLVMContext::MD_tbaa 322 || Tag == LLVMContext::MD_fpmath 323 || Tag == LLVMContext::MD_tbaa_struct 324 || Tag == LLVMContext::MD_invariant_load 325 || Tag == LLVMContext::MD_alias_scope 326 || Tag == LLVMContext::MD_noalias 327 || Tag == ParallelLoopAccessMDKind); 328} 329 330// Transfer metadata from Op to the instructions in CV if it is known 331// to be safe to do so. 332void Scalarizer::transferMetadata(Instruction *Op, const ValueVector &CV) { 333 SmallVector<std::pair<unsigned, MDNode *>, 4> MDs; 334 Op->getAllMetadataOtherThanDebugLoc(MDs); 335 for (unsigned I = 0, E = CV.size(); I != E; ++I) { 336 if (Instruction *New = dyn_cast<Instruction>(CV[I])) { 337 for (SmallVectorImpl<std::pair<unsigned, MDNode *>>::iterator 338 MI = MDs.begin(), 339 ME = MDs.end(); 340 MI != ME; ++MI) 341 if (canTransferMetadata(MI->first)) 342 New->setMetadata(MI->first, MI->second); 343 New->setDebugLoc(Op->getDebugLoc()); 344 } 345 } 346} 347 348// Try to fill in Layout from Ty, returning true on success. Alignment is 349// the alignment of the vector, or 0 if the ABI default should be used. 350bool Scalarizer::getVectorLayout(Type *Ty, unsigned Alignment, 351 VectorLayout &Layout, const DataLayout &DL) { 352 // Make sure we're dealing with a vector. 353 Layout.VecTy = dyn_cast<VectorType>(Ty); 354 if (!Layout.VecTy) 355 return false; 356 357 // Check that we're dealing with full-byte elements. 358 Layout.ElemTy = Layout.VecTy->getElementType(); 359 if (DL.getTypeSizeInBits(Layout.ElemTy) != 360 DL.getTypeStoreSizeInBits(Layout.ElemTy)) 361 return false; 362 363 if (Alignment) 364 Layout.VecAlign = Alignment; 365 else 366 Layout.VecAlign = DL.getABITypeAlignment(Layout.VecTy); 367 Layout.ElemSize = DL.getTypeStoreSize(Layout.ElemTy); 368 return true; 369} 370 371// Scalarize two-operand instruction I, using Split(Builder, X, Y, Name) 372// to create an instruction like I with operands X and Y and name Name. 373template<typename Splitter> 374bool Scalarizer::splitBinary(Instruction &I, const Splitter &Split) { 375 VectorType *VT = dyn_cast<VectorType>(I.getType()); 376 if (!VT) 377 return false; 378 379 unsigned NumElems = VT->getNumElements(); 380 IRBuilder<> Builder(&I); 381 Scatterer Op0 = scatter(&I, I.getOperand(0)); 382 Scatterer Op1 = scatter(&I, I.getOperand(1)); 383 assert(Op0.size() == NumElems && "Mismatched binary operation"); 384 assert(Op1.size() == NumElems && "Mismatched binary operation"); 385 ValueVector Res; 386 Res.resize(NumElems); 387 for (unsigned Elem = 0; Elem < NumElems; ++Elem) 388 Res[Elem] = Split(Builder, Op0[Elem], Op1[Elem], 389 I.getName() + ".i" + Twine(Elem)); 390 gather(&I, Res); 391 return true; 392} 393 394bool Scalarizer::visitSelectInst(SelectInst &SI) { 395 VectorType *VT = dyn_cast<VectorType>(SI.getType()); 396 if (!VT) 397 return false; 398 399 unsigned NumElems = VT->getNumElements(); 400 IRBuilder<> Builder(&SI); 401 Scatterer Op1 = scatter(&SI, SI.getOperand(1)); 402 Scatterer Op2 = scatter(&SI, SI.getOperand(2)); 403 assert(Op1.size() == NumElems && "Mismatched select"); 404 assert(Op2.size() == NumElems && "Mismatched select"); 405 ValueVector Res; 406 Res.resize(NumElems); 407 408 if (SI.getOperand(0)->getType()->isVectorTy()) { 409 Scatterer Op0 = scatter(&SI, SI.getOperand(0)); 410 assert(Op0.size() == NumElems && "Mismatched select"); 411 for (unsigned I = 0; I < NumElems; ++I) 412 Res[I] = Builder.CreateSelect(Op0[I], Op1[I], Op2[I], 413 SI.getName() + ".i" + Twine(I)); 414 } else { 415 Value *Op0 = SI.getOperand(0); 416 for (unsigned I = 0; I < NumElems; ++I) 417 Res[I] = Builder.CreateSelect(Op0, Op1[I], Op2[I], 418 SI.getName() + ".i" + Twine(I)); 419 } 420 gather(&SI, Res); 421 return true; 422} 423 424bool Scalarizer::visitICmpInst(ICmpInst &ICI) { 425 return splitBinary(ICI, ICmpSplitter(ICI)); 426} 427 428bool Scalarizer::visitFCmpInst(FCmpInst &FCI) { 429 return splitBinary(FCI, FCmpSplitter(FCI)); 430} 431 432bool Scalarizer::visitBinaryOperator(BinaryOperator &BO) { 433 return splitBinary(BO, BinarySplitter(BO)); 434} 435 436bool Scalarizer::visitGetElementPtrInst(GetElementPtrInst &GEPI) { 437 VectorType *VT = dyn_cast<VectorType>(GEPI.getType()); 438 if (!VT) 439 return false; 440 441 IRBuilder<> Builder(&GEPI); 442 unsigned NumElems = VT->getNumElements(); 443 unsigned NumIndices = GEPI.getNumIndices(); 444 445 Scatterer Base = scatter(&GEPI, GEPI.getOperand(0)); 446 447 SmallVector<Scatterer, 8> Ops; 448 Ops.resize(NumIndices); 449 for (unsigned I = 0; I < NumIndices; ++I) 450 Ops[I] = scatter(&GEPI, GEPI.getOperand(I + 1)); 451 452 ValueVector Res; 453 Res.resize(NumElems); 454 for (unsigned I = 0; I < NumElems; ++I) { 455 SmallVector<Value *, 8> Indices; 456 Indices.resize(NumIndices); 457 for (unsigned J = 0; J < NumIndices; ++J) 458 Indices[J] = Ops[J][I]; 459 Res[I] = Builder.CreateGEP(GEPI.getSourceElementType(), Base[I], Indices, 460 GEPI.getName() + ".i" + Twine(I)); 461 if (GEPI.isInBounds()) 462 if (GetElementPtrInst *NewGEPI = dyn_cast<GetElementPtrInst>(Res[I])) 463 NewGEPI->setIsInBounds(); 464 } 465 gather(&GEPI, Res); 466 return true; 467} 468 469bool Scalarizer::visitCastInst(CastInst &CI) { 470 VectorType *VT = dyn_cast<VectorType>(CI.getDestTy()); 471 if (!VT) 472 return false; 473 474 unsigned NumElems = VT->getNumElements(); 475 IRBuilder<> Builder(&CI); 476 Scatterer Op0 = scatter(&CI, CI.getOperand(0)); 477 assert(Op0.size() == NumElems && "Mismatched cast"); 478 ValueVector Res; 479 Res.resize(NumElems); 480 for (unsigned I = 0; I < NumElems; ++I) 481 Res[I] = Builder.CreateCast(CI.getOpcode(), Op0[I], VT->getElementType(), 482 CI.getName() + ".i" + Twine(I)); 483 gather(&CI, Res); 484 return true; 485} 486 487bool Scalarizer::visitBitCastInst(BitCastInst &BCI) { 488 VectorType *DstVT = dyn_cast<VectorType>(BCI.getDestTy()); 489 VectorType *SrcVT = dyn_cast<VectorType>(BCI.getSrcTy()); 490 if (!DstVT || !SrcVT) 491 return false; 492 493 unsigned DstNumElems = DstVT->getNumElements(); 494 unsigned SrcNumElems = SrcVT->getNumElements(); 495 IRBuilder<> Builder(&BCI); 496 Scatterer Op0 = scatter(&BCI, BCI.getOperand(0)); 497 ValueVector Res; 498 Res.resize(DstNumElems); 499 500 if (DstNumElems == SrcNumElems) { 501 for (unsigned I = 0; I < DstNumElems; ++I) 502 Res[I] = Builder.CreateBitCast(Op0[I], DstVT->getElementType(), 503 BCI.getName() + ".i" + Twine(I)); 504 } else if (DstNumElems > SrcNumElems) { 505 // <M x t1> -> <N*M x t2>. Convert each t1 to <N x t2> and copy the 506 // individual elements to the destination. 507 unsigned FanOut = DstNumElems / SrcNumElems; 508 Type *MidTy = VectorType::get(DstVT->getElementType(), FanOut); 509 unsigned ResI = 0; 510 for (unsigned Op0I = 0; Op0I < SrcNumElems; ++Op0I) { 511 Value *V = Op0[Op0I]; 512 Instruction *VI; 513 // Look through any existing bitcasts before converting to <N x t2>. 514 // In the best case, the resulting conversion might be a no-op. 515 while ((VI = dyn_cast<Instruction>(V)) && 516 VI->getOpcode() == Instruction::BitCast) 517 V = VI->getOperand(0); 518 V = Builder.CreateBitCast(V, MidTy, V->getName() + ".cast"); 519 Scatterer Mid = scatter(&BCI, V); 520 for (unsigned MidI = 0; MidI < FanOut; ++MidI) 521 Res[ResI++] = Mid[MidI]; 522 } 523 } else { 524 // <N*M x t1> -> <M x t2>. Convert each group of <N x t1> into a t2. 525 unsigned FanIn = SrcNumElems / DstNumElems; 526 Type *MidTy = VectorType::get(SrcVT->getElementType(), FanIn); 527 unsigned Op0I = 0; 528 for (unsigned ResI = 0; ResI < DstNumElems; ++ResI) { 529 Value *V = UndefValue::get(MidTy); 530 for (unsigned MidI = 0; MidI < FanIn; ++MidI) 531 V = Builder.CreateInsertElement(V, Op0[Op0I++], Builder.getInt32(MidI), 532 BCI.getName() + ".i" + Twine(ResI) 533 + ".upto" + Twine(MidI)); 534 Res[ResI] = Builder.CreateBitCast(V, DstVT->getElementType(), 535 BCI.getName() + ".i" + Twine(ResI)); 536 } 537 } 538 gather(&BCI, Res); 539 return true; 540} 541 542bool Scalarizer::visitShuffleVectorInst(ShuffleVectorInst &SVI) { 543 VectorType *VT = dyn_cast<VectorType>(SVI.getType()); 544 if (!VT) 545 return false; 546 547 unsigned NumElems = VT->getNumElements(); 548 Scatterer Op0 = scatter(&SVI, SVI.getOperand(0)); 549 Scatterer Op1 = scatter(&SVI, SVI.getOperand(1)); 550 ValueVector Res; 551 Res.resize(NumElems); 552 553 for (unsigned I = 0; I < NumElems; ++I) { 554 int Selector = SVI.getMaskValue(I); 555 if (Selector < 0) 556 Res[I] = UndefValue::get(VT->getElementType()); 557 else if (unsigned(Selector) < Op0.size()) 558 Res[I] = Op0[Selector]; 559 else 560 Res[I] = Op1[Selector - Op0.size()]; 561 } 562 gather(&SVI, Res); 563 return true; 564} 565 566bool Scalarizer::visitPHINode(PHINode &PHI) { 567 VectorType *VT = dyn_cast<VectorType>(PHI.getType()); 568 if (!VT) 569 return false; 570 571 unsigned NumElems = VT->getNumElements(); 572 IRBuilder<> Builder(&PHI); 573 ValueVector Res; 574 Res.resize(NumElems); 575 576 unsigned NumOps = PHI.getNumOperands(); 577 for (unsigned I = 0; I < NumElems; ++I) 578 Res[I] = Builder.CreatePHI(VT->getElementType(), NumOps, 579 PHI.getName() + ".i" + Twine(I)); 580 581 for (unsigned I = 0; I < NumOps; ++I) { 582 Scatterer Op = scatter(&PHI, PHI.getIncomingValue(I)); 583 BasicBlock *IncomingBlock = PHI.getIncomingBlock(I); 584 for (unsigned J = 0; J < NumElems; ++J) 585 cast<PHINode>(Res[J])->addIncoming(Op[J], IncomingBlock); 586 } 587 gather(&PHI, Res); 588 return true; 589} 590 591bool Scalarizer::visitLoadInst(LoadInst &LI) { 592 if (!ScalarizeLoadStore) 593 return false; 594 if (!LI.isSimple()) 595 return false; 596 597 VectorLayout Layout; 598 if (!getVectorLayout(LI.getType(), LI.getAlignment(), Layout, 599 LI.getModule()->getDataLayout())) 600 return false; 601 602 unsigned NumElems = Layout.VecTy->getNumElements(); 603 IRBuilder<> Builder(&LI); 604 Scatterer Ptr = scatter(&LI, LI.getPointerOperand()); 605 ValueVector Res; 606 Res.resize(NumElems); 607 608 for (unsigned I = 0; I < NumElems; ++I) 609 Res[I] = Builder.CreateAlignedLoad(Ptr[I], Layout.getElemAlign(I), 610 LI.getName() + ".i" + Twine(I)); 611 gather(&LI, Res); 612 return true; 613} 614 615bool Scalarizer::visitStoreInst(StoreInst &SI) { 616 if (!ScalarizeLoadStore) 617 return false; 618 if (!SI.isSimple()) 619 return false; 620 621 VectorLayout Layout; 622 Value *FullValue = SI.getValueOperand(); 623 if (!getVectorLayout(FullValue->getType(), SI.getAlignment(), Layout, 624 SI.getModule()->getDataLayout())) 625 return false; 626 627 unsigned NumElems = Layout.VecTy->getNumElements(); 628 IRBuilder<> Builder(&SI); 629 Scatterer Ptr = scatter(&SI, SI.getPointerOperand()); 630 Scatterer Val = scatter(&SI, FullValue); 631 632 ValueVector Stores; 633 Stores.resize(NumElems); 634 for (unsigned I = 0; I < NumElems; ++I) { 635 unsigned Align = Layout.getElemAlign(I); 636 Stores[I] = Builder.CreateAlignedStore(Val[I], Ptr[I], Align); 637 } 638 transferMetadata(&SI, Stores); 639 return true; 640} 641 642// Delete the instructions that we scalarized. If a full vector result 643// is still needed, recreate it using InsertElements. 644bool Scalarizer::finish() { 645 // The presence of data in Gathered or Scattered indicates changes 646 // made to the Function. 647 if (Gathered.empty() && Scattered.empty()) 648 return false; 649 for (GatherList::iterator GMI = Gathered.begin(), GME = Gathered.end(); 650 GMI != GME; ++GMI) { 651 Instruction *Op = GMI->first; 652 ValueVector &CV = *GMI->second; 653 if (!Op->use_empty()) { 654 // The value is still needed, so recreate it using a series of 655 // InsertElements. 656 Type *Ty = Op->getType(); 657 Value *Res = UndefValue::get(Ty); 658 BasicBlock *BB = Op->getParent(); 659 unsigned Count = Ty->getVectorNumElements(); 660 IRBuilder<> Builder(Op); 661 if (isa<PHINode>(Op)) 662 Builder.SetInsertPoint(BB, BB->getFirstInsertionPt()); 663 for (unsigned I = 0; I < Count; ++I) 664 Res = Builder.CreateInsertElement(Res, CV[I], Builder.getInt32(I), 665 Op->getName() + ".upto" + Twine(I)); 666 Res->takeName(Op); 667 Op->replaceAllUsesWith(Res); 668 } 669 Op->eraseFromParent(); 670 } 671 Gathered.clear(); 672 Scattered.clear(); 673 return true; 674} 675 676FunctionPass *llvm::createScalarizerPass() { 677 return new Scalarizer(); 678} 679