1 //=== AMDGPUPrintfRuntimeBinding.cpp - OpenCL printf implementation -------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // \file 9 // 10 // The pass bind printfs to a kernel arg pointer that will be bound to a buffer 11 // later by the runtime. 12 // 13 // This pass traverses the functions in the module and converts 14 // each call to printf to a sequence of operations that 15 // store the following into the printf buffer: 16 // - format string (passed as a module's metadata unique ID) 17 // - bitwise copies of printf arguments 18 // The backend passes will need to store metadata in the kernel 19 //===----------------------------------------------------------------------===// 20 21 #include "AMDGPU.h" 22 #include "llvm/ADT/SmallString.h" 23 #include "llvm/ADT/StringExtras.h" 24 #include "llvm/ADT/Triple.h" 25 #include "llvm/Analysis/InstructionSimplify.h" 26 #include "llvm/Analysis/TargetLibraryInfo.h" 27 #include "llvm/CodeGen/Passes.h" 28 #include "llvm/IR/Constants.h" 29 #include "llvm/IR/DataLayout.h" 30 #include "llvm/IR/Dominators.h" 31 #include "llvm/IR/GlobalVariable.h" 32 #include "llvm/IR/IRBuilder.h" 33 #include "llvm/IR/InstVisitor.h" 34 #include "llvm/IR/Instructions.h" 35 #include "llvm/IR/Module.h" 36 #include "llvm/IR/Type.h" 37 #include "llvm/Support/CommandLine.h" 38 #include "llvm/Support/Debug.h" 39 #include "llvm/Support/raw_ostream.h" 40 #include "llvm/Transforms/Utils/BasicBlockUtils.h" 41 using namespace llvm; 42 43 #define DEBUG_TYPE "printfToRuntime" 44 #define DWORD_ALIGN 4 45 46 namespace { 47 class LLVM_LIBRARY_VISIBILITY AMDGPUPrintfRuntimeBinding final 48 : public ModulePass, 49 public InstVisitor<AMDGPUPrintfRuntimeBinding> { 50 51 public: 52 static char ID; 53 54 explicit AMDGPUPrintfRuntimeBinding(); 55 56 void visitCallSite(CallSite CS) { 57 Function *F = CS.getCalledFunction(); 58 if (F && F->hasName() && F->getName() == "printf") 59 Printfs.push_back(CS.getInstruction()); 60 } 61 62 private: 63 bool runOnModule(Module &M) override; 64 void getConversionSpecifiers(SmallVectorImpl<char> &OpConvSpecifiers, 65 StringRef fmt, size_t num_ops) const; 66 67 bool shouldPrintAsStr(char Specifier, Type *OpType) const; 68 bool 69 lowerPrintfForGpu(Module &M, 70 function_ref<const TargetLibraryInfo &(Function &)> GetTLI); 71 72 void getAnalysisUsage(AnalysisUsage &AU) const override { 73 AU.addRequired<TargetLibraryInfoWrapperPass>(); 74 AU.addRequired<DominatorTreeWrapperPass>(); 75 } 76 77 Value *simplify(Instruction *I, const TargetLibraryInfo *TLI) { 78 return SimplifyInstruction(I, {*TD, TLI, DT}); 79 } 80 81 const DataLayout *TD; 82 const DominatorTree *DT; 83 SmallVector<Value *, 32> Printfs; 84 }; 85 } // namespace 86 87 char AMDGPUPrintfRuntimeBinding::ID = 0; 88 89 INITIALIZE_PASS_BEGIN(AMDGPUPrintfRuntimeBinding, 90 "amdgpu-printf-runtime-binding", "AMDGPU Printf lowering", 91 false, false) 92 INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass) 93 INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) 94 INITIALIZE_PASS_END(AMDGPUPrintfRuntimeBinding, "amdgpu-printf-runtime-binding", 95 "AMDGPU Printf lowering", false, false) 96 97 char &llvm::AMDGPUPrintfRuntimeBindingID = AMDGPUPrintfRuntimeBinding::ID; 98 99 namespace llvm { 100 ModulePass *createAMDGPUPrintfRuntimeBinding() { 101 return new AMDGPUPrintfRuntimeBinding(); 102 } 103 } // namespace llvm 104 105 AMDGPUPrintfRuntimeBinding::AMDGPUPrintfRuntimeBinding() 106 : ModulePass(ID), TD(nullptr), DT(nullptr) { 107 initializeAMDGPUPrintfRuntimeBindingPass(*PassRegistry::getPassRegistry()); 108 } 109 110 void AMDGPUPrintfRuntimeBinding::getConversionSpecifiers( 111 SmallVectorImpl<char> &OpConvSpecifiers, StringRef Fmt, 112 size_t NumOps) const { 113 // not all format characters are collected. 114 // At this time the format characters of interest 115 // are %p and %s, which use to know if we 116 // are either storing a literal string or a 117 // pointer to the printf buffer. 118 static const char ConvSpecifiers[] = "cdieEfgGaosuxXp"; 119 size_t CurFmtSpecifierIdx = 0; 120 size_t PrevFmtSpecifierIdx = 0; 121 122 while ((CurFmtSpecifierIdx = Fmt.find_first_of( 123 ConvSpecifiers, CurFmtSpecifierIdx)) != StringRef::npos) { 124 bool ArgDump = false; 125 StringRef CurFmt = Fmt.substr(PrevFmtSpecifierIdx, 126 CurFmtSpecifierIdx - PrevFmtSpecifierIdx); 127 size_t pTag = CurFmt.find_last_of("%"); 128 if (pTag != StringRef::npos) { 129 ArgDump = true; 130 while (pTag && CurFmt[--pTag] == '%') { 131 ArgDump = !ArgDump; 132 } 133 } 134 135 if (ArgDump) 136 OpConvSpecifiers.push_back(Fmt[CurFmtSpecifierIdx]); 137 138 PrevFmtSpecifierIdx = ++CurFmtSpecifierIdx; 139 } 140 } 141 142 bool AMDGPUPrintfRuntimeBinding::shouldPrintAsStr(char Specifier, 143 Type *OpType) const { 144 if (Specifier != 's') 145 return false; 146 const PointerType *PT = dyn_cast<PointerType>(OpType); 147 if (!PT || PT->getAddressSpace() != AMDGPUAS::CONSTANT_ADDRESS) 148 return false; 149 Type *ElemType = PT->getContainedType(0); 150 if (ElemType->getTypeID() != Type::IntegerTyID) 151 return false; 152 IntegerType *ElemIType = cast<IntegerType>(ElemType); 153 return ElemIType->getBitWidth() == 8; 154 } 155 156 bool AMDGPUPrintfRuntimeBinding::lowerPrintfForGpu( 157 Module &M, function_ref<const TargetLibraryInfo &(Function &)> GetTLI) { 158 LLVMContext &Ctx = M.getContext(); 159 IRBuilder<> Builder(Ctx); 160 Type *I32Ty = Type::getInt32Ty(Ctx); 161 unsigned UniqID = 0; 162 // NB: This is important for this string size to be divizable by 4 163 const char NonLiteralStr[4] = "???"; 164 165 for (auto P : Printfs) { 166 auto CI = cast<CallInst>(P); 167 unsigned NumOps = CI->getNumArgOperands(); 168 169 SmallString<16> OpConvSpecifiers; 170 Value *Op = CI->getArgOperand(0); 171 172 if (auto LI = dyn_cast<LoadInst>(Op)) { 173 Op = LI->getPointerOperand(); 174 for (auto Use : Op->users()) { 175 if (auto SI = dyn_cast<StoreInst>(Use)) { 176 Op = SI->getValueOperand(); 177 break; 178 } 179 } 180 } 181 182 if (auto I = dyn_cast<Instruction>(Op)) { 183 Value *Op_simplified = simplify(I, &GetTLI(*I->getFunction())); 184 if (Op_simplified) 185 Op = Op_simplified; 186 } 187 188 ConstantExpr *ConstExpr = dyn_cast<ConstantExpr>(Op); 189 190 if (ConstExpr) { 191 GlobalVariable *GVar = dyn_cast<GlobalVariable>(ConstExpr->getOperand(0)); 192 193 StringRef Str("unknown"); 194 if (GVar && GVar->hasInitializer()) { 195 auto Init = GVar->getInitializer(); 196 if (auto CA = dyn_cast<ConstantDataArray>(Init)) { 197 if (CA->isString()) 198 Str = CA->getAsCString(); 199 } else if (isa<ConstantAggregateZero>(Init)) { 200 Str = ""; 201 } 202 // 203 // we need this call to ascertain 204 // that we are printing a string 205 // or a pointer. It takes out the 206 // specifiers and fills up the first 207 // arg 208 getConversionSpecifiers(OpConvSpecifiers, Str, NumOps - 1); 209 } 210 // Add metadata for the string 211 std::string AStreamHolder; 212 raw_string_ostream Sizes(AStreamHolder); 213 int Sum = DWORD_ALIGN; 214 Sizes << CI->getNumArgOperands() - 1; 215 Sizes << ':'; 216 for (unsigned ArgCount = 1; ArgCount < CI->getNumArgOperands() && 217 ArgCount <= OpConvSpecifiers.size(); 218 ArgCount++) { 219 Value *Arg = CI->getArgOperand(ArgCount); 220 Type *ArgType = Arg->getType(); 221 unsigned ArgSize = TD->getTypeAllocSizeInBits(ArgType); 222 ArgSize = ArgSize / 8; 223 // 224 // ArgSize by design should be a multiple of DWORD_ALIGN, 225 // expand the arguments that do not follow this rule. 226 // 227 if (ArgSize % DWORD_ALIGN != 0) { 228 llvm::Type *ResType = llvm::Type::getInt32Ty(Ctx); 229 VectorType *LLVMVecType = llvm::dyn_cast<llvm::VectorType>(ArgType); 230 int NumElem = LLVMVecType ? LLVMVecType->getNumElements() : 1; 231 if (LLVMVecType && NumElem > 1) 232 ResType = llvm::VectorType::get(ResType, NumElem); 233 Builder.SetInsertPoint(CI); 234 Builder.SetCurrentDebugLocation(CI->getDebugLoc()); 235 if (OpConvSpecifiers[ArgCount - 1] == 'x' || 236 OpConvSpecifiers[ArgCount - 1] == 'X' || 237 OpConvSpecifiers[ArgCount - 1] == 'u' || 238 OpConvSpecifiers[ArgCount - 1] == 'o') 239 Arg = Builder.CreateZExt(Arg, ResType); 240 else 241 Arg = Builder.CreateSExt(Arg, ResType); 242 ArgType = Arg->getType(); 243 ArgSize = TD->getTypeAllocSizeInBits(ArgType); 244 ArgSize = ArgSize / 8; 245 CI->setOperand(ArgCount, Arg); 246 } 247 if (OpConvSpecifiers[ArgCount - 1] == 'f') { 248 ConstantFP *FpCons = dyn_cast<ConstantFP>(Arg); 249 if (FpCons) 250 ArgSize = 4; 251 else { 252 FPExtInst *FpExt = dyn_cast<FPExtInst>(Arg); 253 if (FpExt && FpExt->getType()->isDoubleTy() && 254 FpExt->getOperand(0)->getType()->isFloatTy()) 255 ArgSize = 4; 256 } 257 } 258 if (shouldPrintAsStr(OpConvSpecifiers[ArgCount - 1], ArgType)) { 259 if (ConstantExpr *ConstExpr = dyn_cast<ConstantExpr>(Arg)) { 260 GlobalVariable *GV = 261 dyn_cast<GlobalVariable>(ConstExpr->getOperand(0)); 262 if (GV && GV->hasInitializer()) { 263 Constant *Init = GV->getInitializer(); 264 ConstantDataArray *CA = dyn_cast<ConstantDataArray>(Init); 265 if (Init->isZeroValue() || CA->isString()) { 266 size_t SizeStr = Init->isZeroValue() 267 ? 1 268 : (strlen(CA->getAsCString().data()) + 1); 269 size_t Rem = SizeStr % DWORD_ALIGN; 270 size_t NSizeStr = 0; 271 LLVM_DEBUG(dbgs() << "Printf string original size = " << SizeStr 272 << '\n'); 273 if (Rem) { 274 NSizeStr = SizeStr + (DWORD_ALIGN - Rem); 275 } else { 276 NSizeStr = SizeStr; 277 } 278 ArgSize = NSizeStr; 279 } 280 } else { 281 ArgSize = sizeof(NonLiteralStr); 282 } 283 } else { 284 ArgSize = sizeof(NonLiteralStr); 285 } 286 } 287 LLVM_DEBUG(dbgs() << "Printf ArgSize (in buffer) = " << ArgSize 288 << " for type: " << *ArgType << '\n'); 289 Sizes << ArgSize << ':'; 290 Sum += ArgSize; 291 } 292 LLVM_DEBUG(dbgs() << "Printf format string in source = " << Str.str() 293 << '\n'); 294 for (size_t I = 0; I < Str.size(); ++I) { 295 // Rest of the C escape sequences (e.g. \') are handled correctly 296 // by the MDParser 297 switch (Str[I]) { 298 case '\a': 299 Sizes << "\\a"; 300 break; 301 case '\b': 302 Sizes << "\\b"; 303 break; 304 case '\f': 305 Sizes << "\\f"; 306 break; 307 case '\n': 308 Sizes << "\\n"; 309 break; 310 case '\r': 311 Sizes << "\\r"; 312 break; 313 case '\v': 314 Sizes << "\\v"; 315 break; 316 case ':': 317 // ':' cannot be scanned by Flex, as it is defined as a delimiter 318 // Replace it with it's octal representation \72 319 Sizes << "\\72"; 320 break; 321 default: 322 Sizes << Str[I]; 323 break; 324 } 325 } 326 327 // Insert the printf_alloc call 328 Builder.SetInsertPoint(CI); 329 Builder.SetCurrentDebugLocation(CI->getDebugLoc()); 330 331 AttributeList Attr = AttributeList::get(Ctx, AttributeList::FunctionIndex, 332 Attribute::NoUnwind); 333 334 Type *SizetTy = Type::getInt32Ty(Ctx); 335 336 Type *Tys_alloc[1] = {SizetTy}; 337 Type *I8Ptr = PointerType::get(Type::getInt8Ty(Ctx), 1); 338 FunctionType *FTy_alloc = FunctionType::get(I8Ptr, Tys_alloc, false); 339 FunctionCallee PrintfAllocFn = 340 M.getOrInsertFunction(StringRef("__printf_alloc"), FTy_alloc, Attr); 341 342 LLVM_DEBUG(dbgs() << "Printf metadata = " << Sizes.str() << '\n'); 343 std::string fmtstr = itostr(++UniqID) + ":" + Sizes.str().c_str(); 344 MDString *fmtStrArray = MDString::get(Ctx, fmtstr); 345 346 // Instead of creating global variables, the 347 // printf format strings are extracted 348 // and passed as metadata. This avoids 349 // polluting llvm's symbol tables in this module. 350 // Metadata is going to be extracted 351 // by the backend passes and inserted 352 // into the OpenCL binary as appropriate. 353 StringRef amd("llvm.printf.fmts"); 354 NamedMDNode *metaD = M.getOrInsertNamedMetadata(amd); 355 MDNode *myMD = MDNode::get(Ctx, fmtStrArray); 356 metaD->addOperand(myMD); 357 Value *sumC = ConstantInt::get(SizetTy, Sum, false); 358 SmallVector<Value *, 1> alloc_args; 359 alloc_args.push_back(sumC); 360 CallInst *pcall = 361 CallInst::Create(PrintfAllocFn, alloc_args, "printf_alloc_fn", CI); 362 363 // 364 // Insert code to split basicblock with a 365 // piece of hammock code. 366 // basicblock splits after buffer overflow check 367 // 368 ConstantPointerNull *zeroIntPtr = 369 ConstantPointerNull::get(PointerType::get(Type::getInt8Ty(Ctx), 1)); 370 ICmpInst *cmp = 371 dyn_cast<ICmpInst>(Builder.CreateICmpNE(pcall, zeroIntPtr, "")); 372 if (!CI->use_empty()) { 373 Value *result = 374 Builder.CreateSExt(Builder.CreateNot(cmp), I32Ty, "printf_res"); 375 CI->replaceAllUsesWith(result); 376 } 377 SplitBlock(CI->getParent(), cmp); 378 Instruction *Brnch = 379 SplitBlockAndInsertIfThen(cmp, cmp->getNextNode(), false); 380 381 Builder.SetInsertPoint(Brnch); 382 383 // store unique printf id in the buffer 384 // 385 SmallVector<Value *, 1> ZeroIdxList; 386 ConstantInt *zeroInt = 387 ConstantInt::get(Ctx, APInt(32, StringRef("0"), 10)); 388 ZeroIdxList.push_back(zeroInt); 389 390 GetElementPtrInst *BufferIdx = 391 dyn_cast<GetElementPtrInst>(GetElementPtrInst::Create( 392 nullptr, pcall, ZeroIdxList, "PrintBuffID", Brnch)); 393 394 Type *idPointer = PointerType::get(I32Ty, AMDGPUAS::GLOBAL_ADDRESS); 395 Value *id_gep_cast = 396 new BitCastInst(BufferIdx, idPointer, "PrintBuffIdCast", Brnch); 397 398 StoreInst *stbuff = 399 new StoreInst(ConstantInt::get(I32Ty, UniqID), id_gep_cast); 400 stbuff->insertBefore(Brnch); // to Remove unused variable warning 401 402 SmallVector<Value *, 2> FourthIdxList; 403 ConstantInt *fourInt = 404 ConstantInt::get(Ctx, APInt(32, StringRef("4"), 10)); 405 406 FourthIdxList.push_back(fourInt); // 1st 4 bytes hold the printf_id 407 // the following GEP is the buffer pointer 408 BufferIdx = cast<GetElementPtrInst>(GetElementPtrInst::Create( 409 nullptr, pcall, FourthIdxList, "PrintBuffGep", Brnch)); 410 411 Type *Int32Ty = Type::getInt32Ty(Ctx); 412 Type *Int64Ty = Type::getInt64Ty(Ctx); 413 for (unsigned ArgCount = 1; ArgCount < CI->getNumArgOperands() && 414 ArgCount <= OpConvSpecifiers.size(); 415 ArgCount++) { 416 Value *Arg = CI->getArgOperand(ArgCount); 417 Type *ArgType = Arg->getType(); 418 SmallVector<Value *, 32> WhatToStore; 419 if (ArgType->isFPOrFPVectorTy() && 420 (ArgType->getTypeID() != Type::VectorTyID)) { 421 Type *IType = (ArgType->isFloatTy()) ? Int32Ty : Int64Ty; 422 if (OpConvSpecifiers[ArgCount - 1] == 'f') { 423 ConstantFP *fpCons = dyn_cast<ConstantFP>(Arg); 424 if (fpCons) { 425 APFloat Val(fpCons->getValueAPF()); 426 bool Lost = false; 427 Val.convert(APFloat::IEEEsingle(), APFloat::rmNearestTiesToEven, 428 &Lost); 429 Arg = ConstantFP::get(Ctx, Val); 430 IType = Int32Ty; 431 } else { 432 FPExtInst *FpExt = dyn_cast<FPExtInst>(Arg); 433 if (FpExt && FpExt->getType()->isDoubleTy() && 434 FpExt->getOperand(0)->getType()->isFloatTy()) { 435 Arg = FpExt->getOperand(0); 436 IType = Int32Ty; 437 } 438 } 439 } 440 Arg = new BitCastInst(Arg, IType, "PrintArgFP", Brnch); 441 WhatToStore.push_back(Arg); 442 } else if (ArgType->getTypeID() == Type::PointerTyID) { 443 if (shouldPrintAsStr(OpConvSpecifiers[ArgCount - 1], ArgType)) { 444 const char *S = NonLiteralStr; 445 if (ConstantExpr *ConstExpr = dyn_cast<ConstantExpr>(Arg)) { 446 GlobalVariable *GV = 447 dyn_cast<GlobalVariable>(ConstExpr->getOperand(0)); 448 if (GV && GV->hasInitializer()) { 449 Constant *Init = GV->getInitializer(); 450 ConstantDataArray *CA = dyn_cast<ConstantDataArray>(Init); 451 if (Init->isZeroValue() || CA->isString()) { 452 S = Init->isZeroValue() ? "" : CA->getAsCString().data(); 453 } 454 } 455 } 456 size_t SizeStr = strlen(S) + 1; 457 size_t Rem = SizeStr % DWORD_ALIGN; 458 size_t NSizeStr = 0; 459 if (Rem) { 460 NSizeStr = SizeStr + (DWORD_ALIGN - Rem); 461 } else { 462 NSizeStr = SizeStr; 463 } 464 if (S[0]) { 465 char *MyNewStr = new char[NSizeStr](); 466 strcpy(MyNewStr, S); 467 int NumInts = NSizeStr / 4; 468 int CharC = 0; 469 while (NumInts) { 470 int ANum = *(int *)(MyNewStr + CharC); 471 CharC += 4; 472 NumInts--; 473 Value *ANumV = ConstantInt::get(Int32Ty, ANum, false); 474 WhatToStore.push_back(ANumV); 475 } 476 delete[] MyNewStr; 477 } else { 478 // Empty string, give a hint to RT it is no NULL 479 Value *ANumV = ConstantInt::get(Int32Ty, 0xFFFFFF00, false); 480 WhatToStore.push_back(ANumV); 481 } 482 } else { 483 uint64_t Size = TD->getTypeAllocSizeInBits(ArgType); 484 assert((Size == 32 || Size == 64) && "unsupported size"); 485 Type *DstType = (Size == 32) ? Int32Ty : Int64Ty; 486 Arg = new PtrToIntInst(Arg, DstType, "PrintArgPtr", Brnch); 487 WhatToStore.push_back(Arg); 488 } 489 } else if (ArgType->getTypeID() == Type::VectorTyID) { 490 Type *IType = NULL; 491 uint32_t EleCount = cast<VectorType>(ArgType)->getNumElements(); 492 uint32_t EleSize = ArgType->getScalarSizeInBits(); 493 uint32_t TotalSize = EleCount * EleSize; 494 if (EleCount == 3) { 495 IntegerType *Int32Ty = Type::getInt32Ty(ArgType->getContext()); 496 Constant *Indices[4] = { 497 ConstantInt::get(Int32Ty, 0), ConstantInt::get(Int32Ty, 1), 498 ConstantInt::get(Int32Ty, 2), ConstantInt::get(Int32Ty, 2)}; 499 Constant *Mask = ConstantVector::get(Indices); 500 ShuffleVectorInst *Shuffle = new ShuffleVectorInst(Arg, Arg, Mask); 501 Shuffle->insertBefore(Brnch); 502 Arg = Shuffle; 503 ArgType = Arg->getType(); 504 TotalSize += EleSize; 505 } 506 switch (EleSize) { 507 default: 508 EleCount = TotalSize / 64; 509 IType = dyn_cast<Type>(Type::getInt64Ty(ArgType->getContext())); 510 break; 511 case 8: 512 if (EleCount >= 8) { 513 EleCount = TotalSize / 64; 514 IType = dyn_cast<Type>(Type::getInt64Ty(ArgType->getContext())); 515 } else if (EleCount >= 3) { 516 EleCount = 1; 517 IType = dyn_cast<Type>(Type::getInt32Ty(ArgType->getContext())); 518 } else { 519 EleCount = 1; 520 IType = dyn_cast<Type>(Type::getInt16Ty(ArgType->getContext())); 521 } 522 break; 523 case 16: 524 if (EleCount >= 3) { 525 EleCount = TotalSize / 64; 526 IType = dyn_cast<Type>(Type::getInt64Ty(ArgType->getContext())); 527 } else { 528 EleCount = 1; 529 IType = dyn_cast<Type>(Type::getInt32Ty(ArgType->getContext())); 530 } 531 break; 532 } 533 if (EleCount > 1) { 534 IType = dyn_cast<Type>(VectorType::get(IType, EleCount)); 535 } 536 Arg = new BitCastInst(Arg, IType, "PrintArgVect", Brnch); 537 WhatToStore.push_back(Arg); 538 } else { 539 WhatToStore.push_back(Arg); 540 } 541 for (unsigned I = 0, E = WhatToStore.size(); I != E; ++I) { 542 Value *TheBtCast = WhatToStore[I]; 543 unsigned ArgSize = 544 TD->getTypeAllocSizeInBits(TheBtCast->getType()) / 8; 545 SmallVector<Value *, 1> BuffOffset; 546 BuffOffset.push_back(ConstantInt::get(I32Ty, ArgSize)); 547 548 Type *ArgPointer = PointerType::get(TheBtCast->getType(), 1); 549 Value *CastedGEP = 550 new BitCastInst(BufferIdx, ArgPointer, "PrintBuffPtrCast", Brnch); 551 StoreInst *StBuff = new StoreInst(TheBtCast, CastedGEP, Brnch); 552 LLVM_DEBUG(dbgs() << "inserting store to printf buffer:\n" 553 << *StBuff << '\n'); 554 (void)StBuff; 555 if (I + 1 == E && ArgCount + 1 == CI->getNumArgOperands()) 556 break; 557 BufferIdx = dyn_cast<GetElementPtrInst>(GetElementPtrInst::Create( 558 nullptr, BufferIdx, BuffOffset, "PrintBuffNextPtr", Brnch)); 559 LLVM_DEBUG(dbgs() << "inserting gep to the printf buffer:\n" 560 << *BufferIdx << '\n'); 561 } 562 } 563 } 564 } 565 566 // erase the printf calls 567 for (auto P : Printfs) { 568 auto CI = cast<CallInst>(P); 569 CI->eraseFromParent(); 570 } 571 572 Printfs.clear(); 573 return true; 574 } 575 576 bool AMDGPUPrintfRuntimeBinding::runOnModule(Module &M) { 577 Triple TT(M.getTargetTriple()); 578 if (TT.getArch() == Triple::r600) 579 return false; 580 581 visit(M); 582 583 if (Printfs.empty()) 584 return false; 585 586 TD = &M.getDataLayout(); 587 auto DTWP = getAnalysisIfAvailable<DominatorTreeWrapperPass>(); 588 DT = DTWP ? &DTWP->getDomTree() : nullptr; 589 auto GetTLI = [this](Function &F) -> TargetLibraryInfo & { 590 return this->getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F); 591 }; 592 593 return lowerPrintfForGpu(M, GetTLI); 594 } 595