1 //===- AMDGPULibCalls.cpp -------------------------------------------------===// 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 /// \file 11 /// \brief This file does AMD library function optimizations. 12 // 13 //===----------------------------------------------------------------------===// 14 15 #define DEBUG_TYPE "amdgpu-simplifylib" 16 17 #include "AMDGPU.h" 18 #include "AMDGPULibFunc.h" 19 #include "llvm/Analysis/AliasAnalysis.h" 20 #include "llvm/Analysis/Loads.h" 21 #include "llvm/ADT/StringSet.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/IR/Constants.h" 24 #include "llvm/IR/DerivedTypes.h" 25 #include "llvm/IR/Instructions.h" 26 #include "llvm/IR/IRBuilder.h" 27 #include "llvm/IR/Function.h" 28 #include "llvm/IR/LLVMContext.h" 29 #include "llvm/IR/Module.h" 30 #include "llvm/IR/ValueSymbolTable.h" 31 #include "llvm/Support/Debug.h" 32 #include "llvm/Support/raw_ostream.h" 33 #include <vector> 34 #include <cmath> 35 36 using namespace llvm; 37 38 static cl::opt<bool> EnablePreLink("amdgpu-prelink", 39 cl::desc("Enable pre-link mode optimizations"), 40 cl::init(false), 41 cl::Hidden); 42 43 static cl::list<std::string> UseNative("amdgpu-use-native", 44 cl::desc("Comma separated list of functions to replace with native, or all"), 45 cl::CommaSeparated, cl::ValueOptional, 46 cl::Hidden); 47 48 #define MATH_PI 3.14159265358979323846264338327950288419716939937511 49 #define MATH_E 2.71828182845904523536028747135266249775724709369996 50 #define MATH_SQRT2 1.41421356237309504880168872420969807856967187537695 51 52 #define MATH_LOG2E 1.4426950408889634073599246810018921374266459541529859 53 #define MATH_LOG10E 0.4342944819032518276511289189166050822943970058036665 54 // Value of log2(10) 55 #define MATH_LOG2_10 3.3219280948873623478703194294893901758648313930245806 56 // Value of 1 / log2(10) 57 #define MATH_RLOG2_10 0.3010299956639811952137388947244930267681898814621085 58 // Value of 1 / M_LOG2E_F = 1 / log2(e) 59 #define MATH_RLOG2_E 0.6931471805599453094172321214581765680755001343602552 60 61 namespace llvm { 62 63 class AMDGPULibCalls { 64 private: 65 66 typedef llvm::AMDGPULibFunc FuncInfo; 67 68 // -fuse-native. 69 bool AllNative = false; 70 71 bool useNativeFunc(const StringRef F) const; 72 73 // Return a pointer (pointer expr) to the function if function defintion with 74 // "FuncName" exists. It may create a new function prototype in pre-link mode. 75 Constant *getFunction(Module *M, const FuncInfo& fInfo); 76 77 // Replace a normal function with its native version. 78 bool replaceWithNative(CallInst *CI, const FuncInfo &FInfo); 79 80 bool parseFunctionName(const StringRef& FMangledName, 81 FuncInfo *FInfo=nullptr /*out*/); 82 83 bool TDOFold(CallInst *CI, const FuncInfo &FInfo); 84 85 /* Specialized optimizations */ 86 87 // recip (half or native) 88 bool fold_recip(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 89 90 // divide (half or native) 91 bool fold_divide(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 92 93 // pow/powr/pown 94 bool fold_pow(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 95 96 // rootn 97 bool fold_rootn(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 98 99 // fma/mad 100 bool fold_fma_mad(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 101 102 // -fuse-native for sincos 103 bool sincosUseNative(CallInst *aCI, const FuncInfo &FInfo); 104 105 // evaluate calls if calls' arguments are constants. 106 bool evaluateScalarMathFunc(FuncInfo &FInfo, double& Res0, 107 double& Res1, Constant *copr0, Constant *copr1, Constant *copr2); 108 bool evaluateCall(CallInst *aCI, FuncInfo &FInfo); 109 110 // exp 111 bool fold_exp(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 112 113 // exp2 114 bool fold_exp2(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 115 116 // exp10 117 bool fold_exp10(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 118 119 // log 120 bool fold_log(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 121 122 // log2 123 bool fold_log2(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 124 125 // log10 126 bool fold_log10(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 127 128 // sqrt 129 bool fold_sqrt(CallInst *CI, IRBuilder<> &B, const FuncInfo &FInfo); 130 131 // sin/cos 132 bool fold_sincos(CallInst * CI, IRBuilder<> &B, AliasAnalysis * AA); 133 134 // Get insertion point at entry. 135 BasicBlock::iterator getEntryIns(CallInst * UI); 136 // Insert an Alloc instruction. 137 AllocaInst* insertAlloca(CallInst * UI, IRBuilder<> &B, const char *prefix); 138 // Get a scalar native builtin signle argument FP function 139 Constant* getNativeFunction(Module* M, const FuncInfo &FInfo); 140 141 protected: 142 CallInst *CI; 143 144 bool isUnsafeMath(const CallInst *CI) const; 145 146 void replaceCall(Value *With) { 147 CI->replaceAllUsesWith(With); 148 CI->eraseFromParent(); 149 } 150 151 public: 152 bool fold(CallInst *CI, AliasAnalysis *AA = nullptr); 153 154 void initNativeFuncs(); 155 156 // Replace a normal math function call with that native version 157 bool useNative(CallInst *CI); 158 }; 159 160 } // end llvm namespace 161 162 namespace { 163 164 class AMDGPUSimplifyLibCalls : public FunctionPass { 165 166 AMDGPULibCalls Simplifier; 167 168 public: 169 static char ID; // Pass identification 170 171 AMDGPUSimplifyLibCalls() : FunctionPass(ID) { 172 initializeAMDGPUSimplifyLibCallsPass(*PassRegistry::getPassRegistry()); 173 } 174 175 void getAnalysisUsage(AnalysisUsage &AU) const override { 176 AU.addRequired<AAResultsWrapperPass>(); 177 } 178 179 bool runOnFunction(Function &M) override; 180 }; 181 182 class AMDGPUUseNativeCalls : public FunctionPass { 183 184 AMDGPULibCalls Simplifier; 185 186 public: 187 static char ID; // Pass identification 188 189 AMDGPUUseNativeCalls() : FunctionPass(ID) { 190 initializeAMDGPUUseNativeCallsPass(*PassRegistry::getPassRegistry()); 191 Simplifier.initNativeFuncs(); 192 } 193 194 bool runOnFunction(Function &F) override; 195 }; 196 197 } // end anonymous namespace. 198 199 char AMDGPUSimplifyLibCalls::ID = 0; 200 char AMDGPUUseNativeCalls::ID = 0; 201 202 INITIALIZE_PASS_BEGIN(AMDGPUSimplifyLibCalls, "amdgpu-simplifylib", 203 "Simplify well-known AMD library calls", false, false) 204 INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass) 205 INITIALIZE_PASS_END(AMDGPUSimplifyLibCalls, "amdgpu-simplifylib", 206 "Simplify well-known AMD library calls", false, false) 207 208 INITIALIZE_PASS(AMDGPUUseNativeCalls, "amdgpu-usenative", 209 "Replace builtin math calls with that native versions.", 210 false, false) 211 212 template <typename IRB> 213 CallInst *CreateCallEx(IRB &B, Value *Callee, Value *Arg, const Twine &Name="") 214 { 215 CallInst *R = B.CreateCall(Callee, Arg, Name); 216 if (Function* F = dyn_cast<Function>(Callee)) 217 R->setCallingConv(F->getCallingConv()); 218 return R; 219 } 220 221 template <typename IRB> 222 CallInst *CreateCallEx2(IRB &B, Value *Callee, Value *Arg1, Value *Arg2, 223 const Twine &Name="") { 224 CallInst *R = B.CreateCall(Callee, {Arg1, Arg2}, Name); 225 if (Function* F = dyn_cast<Function>(Callee)) 226 R->setCallingConv(F->getCallingConv()); 227 return R; 228 } 229 230 // Data structures for table-driven optimizations. 231 // FuncTbl works for both f32 and f64 functions with 1 input argument 232 233 struct TableEntry { 234 double result; 235 double input; 236 }; 237 238 /* a list of {result, input} */ 239 static const TableEntry tbl_acos[] = { 240 {MATH_PI/2.0, 0.0}, 241 {MATH_PI/2.0, -0.0}, 242 {0.0, 1.0}, 243 {MATH_PI, -1.0} 244 }; 245 static const TableEntry tbl_acosh[] = { 246 {0.0, 1.0} 247 }; 248 static const TableEntry tbl_acospi[] = { 249 {0.5, 0.0}, 250 {0.5, -0.0}, 251 {0.0, 1.0}, 252 {1.0, -1.0} 253 }; 254 static const TableEntry tbl_asin[] = { 255 {0.0, 0.0}, 256 {-0.0, -0.0}, 257 {MATH_PI/2.0, 1.0}, 258 {-MATH_PI/2.0, -1.0} 259 }; 260 static const TableEntry tbl_asinh[] = { 261 {0.0, 0.0}, 262 {-0.0, -0.0} 263 }; 264 static const TableEntry tbl_asinpi[] = { 265 {0.0, 0.0}, 266 {-0.0, -0.0}, 267 {0.5, 1.0}, 268 {-0.5, -1.0} 269 }; 270 static const TableEntry tbl_atan[] = { 271 {0.0, 0.0}, 272 {-0.0, -0.0}, 273 {MATH_PI/4.0, 1.0}, 274 {-MATH_PI/4.0, -1.0} 275 }; 276 static const TableEntry tbl_atanh[] = { 277 {0.0, 0.0}, 278 {-0.0, -0.0} 279 }; 280 static const TableEntry tbl_atanpi[] = { 281 {0.0, 0.0}, 282 {-0.0, -0.0}, 283 {0.25, 1.0}, 284 {-0.25, -1.0} 285 }; 286 static const TableEntry tbl_cbrt[] = { 287 {0.0, 0.0}, 288 {-0.0, -0.0}, 289 {1.0, 1.0}, 290 {-1.0, -1.0}, 291 }; 292 static const TableEntry tbl_cos[] = { 293 {1.0, 0.0}, 294 {1.0, -0.0} 295 }; 296 static const TableEntry tbl_cosh[] = { 297 {1.0, 0.0}, 298 {1.0, -0.0} 299 }; 300 static const TableEntry tbl_cospi[] = { 301 {1.0, 0.0}, 302 {1.0, -0.0} 303 }; 304 static const TableEntry tbl_erfc[] = { 305 {1.0, 0.0}, 306 {1.0, -0.0} 307 }; 308 static const TableEntry tbl_erf[] = { 309 {0.0, 0.0}, 310 {-0.0, -0.0} 311 }; 312 static const TableEntry tbl_exp[] = { 313 {1.0, 0.0}, 314 {1.0, -0.0}, 315 {MATH_E, 1.0} 316 }; 317 static const TableEntry tbl_exp2[] = { 318 {1.0, 0.0}, 319 {1.0, -0.0}, 320 {2.0, 1.0} 321 }; 322 static const TableEntry tbl_exp10[] = { 323 {1.0, 0.0}, 324 {1.0, -0.0}, 325 {10.0, 1.0} 326 }; 327 static const TableEntry tbl_expm1[] = { 328 {0.0, 0.0}, 329 {-0.0, -0.0} 330 }; 331 static const TableEntry tbl_log[] = { 332 {0.0, 1.0}, 333 {1.0, MATH_E} 334 }; 335 static const TableEntry tbl_log2[] = { 336 {0.0, 1.0}, 337 {1.0, 2.0} 338 }; 339 static const TableEntry tbl_log10[] = { 340 {0.0, 1.0}, 341 {1.0, 10.0} 342 }; 343 static const TableEntry tbl_rsqrt[] = { 344 {1.0, 1.0}, 345 {1.0/MATH_SQRT2, 2.0} 346 }; 347 static const TableEntry tbl_sin[] = { 348 {0.0, 0.0}, 349 {-0.0, -0.0} 350 }; 351 static const TableEntry tbl_sinh[] = { 352 {0.0, 0.0}, 353 {-0.0, -0.0} 354 }; 355 static const TableEntry tbl_sinpi[] = { 356 {0.0, 0.0}, 357 {-0.0, -0.0} 358 }; 359 static const TableEntry tbl_sqrt[] = { 360 {0.0, 0.0}, 361 {1.0, 1.0}, 362 {MATH_SQRT2, 2.0} 363 }; 364 static const TableEntry tbl_tan[] = { 365 {0.0, 0.0}, 366 {-0.0, -0.0} 367 }; 368 static const TableEntry tbl_tanh[] = { 369 {0.0, 0.0}, 370 {-0.0, -0.0} 371 }; 372 static const TableEntry tbl_tanpi[] = { 373 {0.0, 0.0}, 374 {-0.0, -0.0} 375 }; 376 static const TableEntry tbl_tgamma[] = { 377 {1.0, 1.0}, 378 {1.0, 2.0}, 379 {2.0, 3.0}, 380 {6.0, 4.0} 381 }; 382 383 static bool HasNative(AMDGPULibFunc::EFuncId id) { 384 switch(id) { 385 case AMDGPULibFunc::EI_DIVIDE: 386 case AMDGPULibFunc::EI_COS: 387 case AMDGPULibFunc::EI_EXP: 388 case AMDGPULibFunc::EI_EXP2: 389 case AMDGPULibFunc::EI_EXP10: 390 case AMDGPULibFunc::EI_LOG: 391 case AMDGPULibFunc::EI_LOG2: 392 case AMDGPULibFunc::EI_LOG10: 393 case AMDGPULibFunc::EI_POWR: 394 case AMDGPULibFunc::EI_RECIP: 395 case AMDGPULibFunc::EI_RSQRT: 396 case AMDGPULibFunc::EI_SIN: 397 case AMDGPULibFunc::EI_SINCOS: 398 case AMDGPULibFunc::EI_SQRT: 399 case AMDGPULibFunc::EI_TAN: 400 return true; 401 default:; 402 } 403 return false; 404 } 405 406 struct TableRef { 407 size_t size; 408 const TableEntry *table; // variable size: from 0 to (size - 1) 409 410 TableRef() : size(0), table(nullptr) {} 411 412 template <size_t N> 413 TableRef(const TableEntry (&tbl)[N]) : size(N), table(&tbl[0]) {} 414 }; 415 416 static TableRef getOptTable(AMDGPULibFunc::EFuncId id) { 417 switch(id) { 418 case AMDGPULibFunc::EI_ACOS: return TableRef(tbl_acos); 419 case AMDGPULibFunc::EI_ACOSH: return TableRef(tbl_acosh); 420 case AMDGPULibFunc::EI_ACOSPI: return TableRef(tbl_acospi); 421 case AMDGPULibFunc::EI_ASIN: return TableRef(tbl_asin); 422 case AMDGPULibFunc::EI_ASINH: return TableRef(tbl_asinh); 423 case AMDGPULibFunc::EI_ASINPI: return TableRef(tbl_asinpi); 424 case AMDGPULibFunc::EI_ATAN: return TableRef(tbl_atan); 425 case AMDGPULibFunc::EI_ATANH: return TableRef(tbl_atanh); 426 case AMDGPULibFunc::EI_ATANPI: return TableRef(tbl_atanpi); 427 case AMDGPULibFunc::EI_CBRT: return TableRef(tbl_cbrt); 428 case AMDGPULibFunc::EI_NCOS: 429 case AMDGPULibFunc::EI_COS: return TableRef(tbl_cos); 430 case AMDGPULibFunc::EI_COSH: return TableRef(tbl_cosh); 431 case AMDGPULibFunc::EI_COSPI: return TableRef(tbl_cospi); 432 case AMDGPULibFunc::EI_ERFC: return TableRef(tbl_erfc); 433 case AMDGPULibFunc::EI_ERF: return TableRef(tbl_erf); 434 case AMDGPULibFunc::EI_EXP: return TableRef(tbl_exp); 435 case AMDGPULibFunc::EI_NEXP2: 436 case AMDGPULibFunc::EI_EXP2: return TableRef(tbl_exp2); 437 case AMDGPULibFunc::EI_EXP10: return TableRef(tbl_exp10); 438 case AMDGPULibFunc::EI_EXPM1: return TableRef(tbl_expm1); 439 case AMDGPULibFunc::EI_LOG: return TableRef(tbl_log); 440 case AMDGPULibFunc::EI_NLOG2: 441 case AMDGPULibFunc::EI_LOG2: return TableRef(tbl_log2); 442 case AMDGPULibFunc::EI_LOG10: return TableRef(tbl_log10); 443 case AMDGPULibFunc::EI_NRSQRT: 444 case AMDGPULibFunc::EI_RSQRT: return TableRef(tbl_rsqrt); 445 case AMDGPULibFunc::EI_NSIN: 446 case AMDGPULibFunc::EI_SIN: return TableRef(tbl_sin); 447 case AMDGPULibFunc::EI_SINH: return TableRef(tbl_sinh); 448 case AMDGPULibFunc::EI_SINPI: return TableRef(tbl_sinpi); 449 case AMDGPULibFunc::EI_NSQRT: 450 case AMDGPULibFunc::EI_SQRT: return TableRef(tbl_sqrt); 451 case AMDGPULibFunc::EI_TAN: return TableRef(tbl_tan); 452 case AMDGPULibFunc::EI_TANH: return TableRef(tbl_tanh); 453 case AMDGPULibFunc::EI_TANPI: return TableRef(tbl_tanpi); 454 case AMDGPULibFunc::EI_TGAMMA: return TableRef(tbl_tgamma); 455 default:; 456 } 457 return TableRef(); 458 } 459 460 static inline int getVecSize(const AMDGPULibFunc& FInfo) { 461 return FInfo.Leads[0].VectorSize; 462 } 463 464 static inline AMDGPULibFunc::EType getArgType(const AMDGPULibFunc& FInfo) { 465 return (AMDGPULibFunc::EType)FInfo.Leads[0].ArgType; 466 } 467 468 Constant *AMDGPULibCalls::getFunction(Module *M, const FuncInfo& fInfo) { 469 // If we are doing PreLinkOpt, the function is external. So it is safe to 470 // use getOrInsertFunction() at this stage. 471 472 return EnablePreLink ? AMDGPULibFunc::getOrInsertFunction(M, fInfo) 473 : AMDGPULibFunc::getFunction(M, fInfo); 474 } 475 476 bool AMDGPULibCalls::parseFunctionName(const StringRef& FMangledName, 477 FuncInfo *FInfo) { 478 return AMDGPULibFunc::parse(FMangledName, *FInfo); 479 } 480 481 bool AMDGPULibCalls::isUnsafeMath(const CallInst *CI) const { 482 if (auto Op = dyn_cast<FPMathOperator>(CI)) 483 if (Op->hasUnsafeAlgebra()) 484 return true; 485 const Function *F = CI->getParent()->getParent(); 486 Attribute Attr = F->getFnAttribute("unsafe-fp-math"); 487 return Attr.getValueAsString() == "true"; 488 } 489 490 bool AMDGPULibCalls::useNativeFunc(const StringRef F) const { 491 return AllNative || 492 std::find(UseNative.begin(), UseNative.end(), F) != UseNative.end(); 493 } 494 495 void AMDGPULibCalls::initNativeFuncs() { 496 AllNative = useNativeFunc("all") || 497 (UseNative.getNumOccurrences() && UseNative.size() == 1 && 498 UseNative.begin()->empty()); 499 } 500 501 bool AMDGPULibCalls::sincosUseNative(CallInst *aCI, const FuncInfo &FInfo) { 502 bool native_sin = useNativeFunc("sin"); 503 bool native_cos = useNativeFunc("cos"); 504 505 if (native_sin && native_cos) { 506 Module *M = aCI->getModule(); 507 Value *opr0 = aCI->getArgOperand(0); 508 509 AMDGPULibFunc nf; 510 nf.Leads[0].ArgType = FInfo.Leads[0].ArgType; 511 nf.Leads[0].VectorSize = FInfo.Leads[0].VectorSize; 512 513 nf.setPrefix(AMDGPULibFunc::NATIVE); 514 nf.setId(AMDGPULibFunc::EI_SIN); 515 Constant *sinExpr = getFunction(M, nf); 516 517 nf.setPrefix(AMDGPULibFunc::NATIVE); 518 nf.setId(AMDGPULibFunc::EI_COS); 519 Constant *cosExpr = getFunction(M, nf); 520 if (sinExpr && cosExpr) { 521 Value *sinval = CallInst::Create(sinExpr, opr0, "splitsin", aCI); 522 Value *cosval = CallInst::Create(cosExpr, opr0, "splitcos", aCI); 523 new StoreInst(cosval, aCI->getArgOperand(1), aCI); 524 525 DEBUG_WITH_TYPE("usenative", dbgs() << "<useNative> replace " << *aCI 526 << " with native version of sin/cos"); 527 528 replaceCall(sinval); 529 return true; 530 } 531 } 532 return false; 533 } 534 535 bool AMDGPULibCalls::useNative(CallInst *aCI) { 536 CI = aCI; 537 Function *Callee = aCI->getCalledFunction(); 538 539 FuncInfo FInfo; 540 if (!parseFunctionName(Callee->getName(), &FInfo) || 541 FInfo.getPrefix() != AMDGPULibFunc::NOPFX || 542 getArgType(FInfo) == AMDGPULibFunc::F64 || 543 !HasNative(FInfo.getId()) || 544 !(AllNative || useNativeFunc(FInfo.getName())) ) { 545 return false; 546 } 547 548 if (FInfo.getId() == AMDGPULibFunc::EI_SINCOS) 549 return sincosUseNative(aCI, FInfo); 550 551 FInfo.setPrefix(AMDGPULibFunc::NATIVE); 552 Constant *F = getFunction(aCI->getModule(), FInfo); 553 if (!F) 554 return false; 555 556 aCI->setCalledFunction(F); 557 DEBUG_WITH_TYPE("usenative", dbgs() << "<useNative> replace " << *aCI 558 << " with native version"); 559 return true; 560 } 561 562 // This function returns false if no change; return true otherwise. 563 bool AMDGPULibCalls::fold(CallInst *CI, AliasAnalysis *AA) { 564 this->CI = CI; 565 Function *Callee = CI->getCalledFunction(); 566 567 // Ignore indirect calls. 568 if (Callee == 0) return false; 569 570 FuncInfo FInfo; 571 if (!parseFunctionName(Callee->getName(), &FInfo)) 572 return false; 573 574 // Further check the number of arguments to see if they match. 575 if (CI->getNumArgOperands() != FInfo.getNumArgs()) 576 return false; 577 578 BasicBlock *BB = CI->getParent(); 579 LLVMContext &Context = CI->getParent()->getContext(); 580 IRBuilder<> B(Context); 581 582 // Set the builder to the instruction after the call. 583 B.SetInsertPoint(BB, CI->getIterator()); 584 585 // Copy fast flags from the original call. 586 if (const FPMathOperator *FPOp = dyn_cast<const FPMathOperator>(CI)) 587 B.setFastMathFlags(FPOp->getFastMathFlags()); 588 589 if (TDOFold(CI, FInfo)) 590 return true; 591 592 // Under unsafe-math, evaluate calls if possible. 593 // According to Brian Sumner, we can do this for all f32 function calls 594 // using host's double function calls. 595 if (isUnsafeMath(CI) && evaluateCall(CI, FInfo)) 596 return true; 597 598 // Specilized optimizations for each function call 599 switch (FInfo.getId()) { 600 case AMDGPULibFunc::EI_RECIP: 601 // skip vector function 602 assert ((FInfo.getPrefix() == AMDGPULibFunc::NATIVE || 603 FInfo.getPrefix() == AMDGPULibFunc::HALF) && 604 "recip must be an either native or half function"); 605 return (getVecSize(FInfo) != 1) ? false : fold_recip(CI, B, FInfo); 606 607 case AMDGPULibFunc::EI_DIVIDE: 608 // skip vector function 609 assert ((FInfo.getPrefix() == AMDGPULibFunc::NATIVE || 610 FInfo.getPrefix() == AMDGPULibFunc::HALF) && 611 "divide must be an either native or half function"); 612 return (getVecSize(FInfo) != 1) ? false : fold_divide(CI, B, FInfo); 613 614 case AMDGPULibFunc::EI_POW: 615 case AMDGPULibFunc::EI_POWR: 616 case AMDGPULibFunc::EI_POWN: 617 return fold_pow(CI, B, FInfo); 618 619 case AMDGPULibFunc::EI_ROOTN: 620 // skip vector function 621 return (getVecSize(FInfo) != 1) ? false : fold_rootn(CI, B, FInfo); 622 623 case AMDGPULibFunc::EI_FMA: 624 case AMDGPULibFunc::EI_MAD: 625 case AMDGPULibFunc::EI_NFMA: 626 // skip vector function 627 return (getVecSize(FInfo) != 1) ? false : fold_fma_mad(CI, B, FInfo); 628 629 case AMDGPULibFunc::EI_SQRT: 630 return isUnsafeMath(CI) && fold_sqrt(CI, B, FInfo); 631 case AMDGPULibFunc::EI_COS: 632 case AMDGPULibFunc::EI_SIN: 633 if ((getArgType(FInfo) == AMDGPULibFunc::F32 || 634 getArgType(FInfo) == AMDGPULibFunc::F64) 635 && (FInfo.getPrefix() == AMDGPULibFunc::NOPFX)) 636 return fold_sincos(CI, B, AA); 637 638 break; 639 640 default: 641 break; 642 } 643 644 return false; 645 } 646 647 bool AMDGPULibCalls::TDOFold(CallInst *CI, const FuncInfo &FInfo) { 648 // Table-Driven optimization 649 const TableRef tr = getOptTable(FInfo.getId()); 650 if (tr.size==0) 651 return false; 652 653 int const sz = (int)tr.size; 654 const TableEntry * const ftbl = tr.table; 655 Value *opr0 = CI->getArgOperand(0); 656 657 if (getVecSize(FInfo) > 1) { 658 if (ConstantDataVector *CV = dyn_cast<ConstantDataVector>(opr0)) { 659 SmallVector<double, 0> DVal; 660 for (int eltNo = 0; eltNo < getVecSize(FInfo); ++eltNo) { 661 ConstantFP *eltval = dyn_cast<ConstantFP>( 662 CV->getElementAsConstant((unsigned)eltNo)); 663 assert(eltval && "Non-FP arguments in math function!"); 664 bool found = false; 665 for (int i=0; i < sz; ++i) { 666 if (eltval->isExactlyValue(ftbl[i].input)) { 667 DVal.push_back(ftbl[i].result); 668 found = true; 669 break; 670 } 671 } 672 if (!found) { 673 // This vector constants not handled yet. 674 return false; 675 } 676 } 677 LLVMContext &context = CI->getParent()->getParent()->getContext(); 678 Constant *nval; 679 if (getArgType(FInfo) == AMDGPULibFunc::F32) { 680 SmallVector<float, 0> FVal; 681 for (unsigned i = 0; i < DVal.size(); ++i) { 682 FVal.push_back((float)DVal[i]); 683 } 684 ArrayRef<float> tmp(FVal); 685 nval = ConstantDataVector::get(context, tmp); 686 } else { // F64 687 ArrayRef<double> tmp(DVal); 688 nval = ConstantDataVector::get(context, tmp); 689 } 690 DEBUG(errs() << "AMDIC: " << *CI 691 << " ---> " << *nval << "\n"); 692 replaceCall(nval); 693 return true; 694 } 695 } else { 696 // Scalar version 697 if (ConstantFP *CF = dyn_cast<ConstantFP>(opr0)) { 698 for (int i = 0; i < sz; ++i) { 699 if (CF->isExactlyValue(ftbl[i].input)) { 700 Value *nval = ConstantFP::get(CF->getType(), ftbl[i].result); 701 DEBUG(errs() << "AMDIC: " << *CI 702 << " ---> " << *nval << "\n"); 703 replaceCall(nval); 704 return true; 705 } 706 } 707 } 708 } 709 710 return false; 711 } 712 713 bool AMDGPULibCalls::replaceWithNative(CallInst *CI, const FuncInfo &FInfo) { 714 Module *M = CI->getModule(); 715 if (getArgType(FInfo) != AMDGPULibFunc::F32 || 716 FInfo.getPrefix() != AMDGPULibFunc::NOPFX || 717 !HasNative(FInfo.getId())) 718 return false; 719 720 AMDGPULibFunc nf = FInfo; 721 nf.setPrefix(AMDGPULibFunc::NATIVE); 722 if (Constant *FPExpr = getFunction(M, nf)) { 723 DEBUG(dbgs() << "AMDIC: " << *CI << " ---> "); 724 725 CI->setCalledFunction(FPExpr); 726 727 DEBUG(dbgs() << *CI << '\n'); 728 729 return true; 730 } 731 return false; 732 } 733 734 // [native_]half_recip(c) ==> 1.0/c 735 bool AMDGPULibCalls::fold_recip(CallInst *CI, IRBuilder<> &B, 736 const FuncInfo &FInfo) { 737 Value *opr0 = CI->getArgOperand(0); 738 if (ConstantFP *CF = dyn_cast<ConstantFP>(opr0)) { 739 // Just create a normal div. Later, InstCombine will be able 740 // to compute the divide into a constant (avoid check float infinity 741 // or subnormal at this point). 742 Value *nval = B.CreateFDiv(ConstantFP::get(CF->getType(), 1.0), 743 opr0, 744 "recip2div"); 745 DEBUG(errs() << "AMDIC: " << *CI 746 << " ---> " << *nval << "\n"); 747 replaceCall(nval); 748 return true; 749 } 750 return false; 751 } 752 753 // [native_]half_divide(x, c) ==> x/c 754 bool AMDGPULibCalls::fold_divide(CallInst *CI, IRBuilder<> &B, 755 const FuncInfo &FInfo) { 756 Value *opr0 = CI->getArgOperand(0); 757 Value *opr1 = CI->getArgOperand(1); 758 ConstantFP *CF0 = dyn_cast<ConstantFP>(opr0); 759 ConstantFP *CF1 = dyn_cast<ConstantFP>(opr1); 760 761 if ((CF0 && CF1) || // both are constants 762 (CF1 && (getArgType(FInfo) == AMDGPULibFunc::F32))) 763 // CF1 is constant && f32 divide 764 { 765 Value *nval1 = B.CreateFDiv(ConstantFP::get(opr1->getType(), 1.0), 766 opr1, "__div2recip"); 767 Value *nval = B.CreateFMul(opr0, nval1, "__div2mul"); 768 replaceCall(nval); 769 return true; 770 } 771 return false; 772 } 773 774 namespace llvm { 775 static double log2(double V) { 776 #if _XOPEN_SOURCE >= 600 || _ISOC99_SOURCE || _POSIX_C_SOURCE >= 200112L 777 return ::log2(V); 778 #else 779 return log(V) / 0.693147180559945309417; 780 #endif 781 } 782 } 783 784 bool AMDGPULibCalls::fold_pow(CallInst *CI, IRBuilder<> &B, 785 const FuncInfo &FInfo) { 786 assert((FInfo.getId() == AMDGPULibFunc::EI_POW || 787 FInfo.getId() == AMDGPULibFunc::EI_POWR || 788 FInfo.getId() == AMDGPULibFunc::EI_POWN) && 789 "fold_pow: encounter a wrong function call"); 790 791 Value *opr0, *opr1; 792 ConstantFP *CF; 793 ConstantInt *CINT; 794 ConstantAggregateZero *CZero; 795 Type *eltType; 796 797 opr0 = CI->getArgOperand(0); 798 opr1 = CI->getArgOperand(1); 799 CZero = dyn_cast<ConstantAggregateZero>(opr1); 800 if (getVecSize(FInfo) == 1) { 801 eltType = opr0->getType(); 802 CF = dyn_cast<ConstantFP>(opr1); 803 CINT = dyn_cast<ConstantInt>(opr1); 804 } else { 805 VectorType *VTy = dyn_cast<VectorType>(opr0->getType()); 806 assert(VTy && "Oprand of vector function should be of vectortype"); 807 eltType = VTy->getElementType(); 808 ConstantDataVector *CDV = dyn_cast<ConstantDataVector>(opr1); 809 810 // Now, only Handle vector const whose elements have the same value. 811 CF = CDV ? dyn_cast_or_null<ConstantFP>(CDV->getSplatValue()) : nullptr; 812 CINT = CDV ? dyn_cast_or_null<ConstantInt>(CDV->getSplatValue()) : nullptr; 813 } 814 815 // No unsafe math , no constant argument, do nothing 816 if (!isUnsafeMath(CI) && !CF && !CINT && !CZero) 817 return false; 818 819 // 0x1111111 means that we don't do anything for this call. 820 int ci_opr1 = (CINT ? (int)CINT->getSExtValue() : 0x1111111); 821 822 if ((CF && CF->isZero()) || (CINT && ci_opr1 == 0) || CZero) { 823 // pow/powr/pown(x, 0) == 1 824 DEBUG(errs() << "AMDIC: " << *CI << " ---> 1\n"); 825 Constant *cnval = ConstantFP::get(eltType, 1.0); 826 if (getVecSize(FInfo) > 1) { 827 cnval = ConstantDataVector::getSplat(getVecSize(FInfo), cnval); 828 } 829 replaceCall(cnval); 830 return true; 831 } 832 if ((CF && CF->isExactlyValue(1.0)) || (CINT && ci_opr1 == 1)) { 833 // pow/powr/pown(x, 1.0) = x 834 DEBUG(errs() << "AMDIC: " << *CI 835 << " ---> " << *opr0 << "\n"); 836 replaceCall(opr0); 837 return true; 838 } 839 if ((CF && CF->isExactlyValue(2.0)) || (CINT && ci_opr1 == 2)) { 840 // pow/powr/pown(x, 2.0) = x*x 841 DEBUG(errs() << "AMDIC: " << *CI 842 << " ---> " << *opr0 << " * " << *opr0 << "\n"); 843 Value *nval = B.CreateFMul(opr0, opr0, "__pow2"); 844 replaceCall(nval); 845 return true; 846 } 847 if ((CF && CF->isExactlyValue(-1.0)) || (CINT && ci_opr1 == -1)) { 848 // pow/powr/pown(x, -1.0) = 1.0/x 849 DEBUG(errs() << "AMDIC: " << *CI 850 << " ---> 1 / " << *opr0 << "\n"); 851 Constant *cnval = ConstantFP::get(eltType, 1.0); 852 if (getVecSize(FInfo) > 1) { 853 cnval = ConstantDataVector::getSplat(getVecSize(FInfo), cnval); 854 } 855 Value *nval = B.CreateFDiv(cnval, opr0, "__powrecip"); 856 replaceCall(nval); 857 return true; 858 } 859 860 Module *M = CI->getModule(); 861 if (CF && (CF->isExactlyValue(0.5) || CF->isExactlyValue(-0.5))) { 862 // pow[r](x, [-]0.5) = sqrt(x) 863 bool issqrt = CF->isExactlyValue(0.5); 864 if (Constant *FPExpr = getFunction(M, 865 AMDGPULibFunc(issqrt ? AMDGPULibFunc::EI_SQRT 866 : AMDGPULibFunc::EI_RSQRT, FInfo))) { 867 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 868 << FInfo.getName().c_str() << "(" << *opr0 << ")\n"); 869 Value *nval = CreateCallEx(B,FPExpr, opr0, issqrt ? "__pow2sqrt" 870 : "__pow2rsqrt"); 871 replaceCall(nval); 872 return true; 873 } 874 } 875 876 if (!isUnsafeMath(CI)) 877 return false; 878 879 // Unsafe Math optimization 880 881 // Remember that ci_opr1 is set if opr1 is integral 882 if (CF) { 883 double dval = (getArgType(FInfo) == AMDGPULibFunc::F32) 884 ? (double)CF->getValueAPF().convertToFloat() 885 : CF->getValueAPF().convertToDouble(); 886 int ival = (int)dval; 887 if ((double)ival == dval) { 888 ci_opr1 = ival; 889 } else 890 ci_opr1 = 0x11111111; 891 } 892 893 // pow/powr/pown(x, c) = [1/](x*x*..x); where 894 // trunc(c) == c && the number of x == c && |c| <= 12 895 unsigned abs_opr1 = (ci_opr1 < 0) ? -ci_opr1 : ci_opr1; 896 if (abs_opr1 <= 12) { 897 Constant *cnval; 898 Value *nval; 899 if (abs_opr1 == 0) { 900 cnval = ConstantFP::get(eltType, 1.0); 901 if (getVecSize(FInfo) > 1) { 902 cnval = ConstantDataVector::getSplat(getVecSize(FInfo), cnval); 903 } 904 nval = cnval; 905 } else { 906 Value *valx2 = nullptr; 907 nval = nullptr; 908 while (abs_opr1 > 0) { 909 valx2 = valx2 ? B.CreateFMul(valx2, valx2, "__powx2") : opr0; 910 if (abs_opr1 & 1) { 911 nval = nval ? B.CreateFMul(nval, valx2, "__powprod") : valx2; 912 } 913 abs_opr1 >>= 1; 914 } 915 } 916 917 if (ci_opr1 < 0) { 918 cnval = ConstantFP::get(eltType, 1.0); 919 if (getVecSize(FInfo) > 1) { 920 cnval = ConstantDataVector::getSplat(getVecSize(FInfo), cnval); 921 } 922 nval = B.CreateFDiv(cnval, nval, "__1powprod"); 923 } 924 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 925 << ((ci_opr1 < 0) ? "1/prod(" : "prod(") << *opr0 << ")\n"); 926 replaceCall(nval); 927 return true; 928 } 929 930 // powr ---> exp2(y * log2(x)) 931 // pown/pow ---> powr(fabs(x), y) | (x & ((int)y << 31)) 932 Constant *ExpExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_EXP2, 933 FInfo)); 934 if (!ExpExpr) 935 return false; 936 937 bool needlog = false; 938 bool needabs = false; 939 bool needcopysign = false; 940 Constant *cnval = nullptr; 941 if (getVecSize(FInfo) == 1) { 942 CF = dyn_cast<ConstantFP>(opr0); 943 944 if (CF) { 945 double V = (getArgType(FInfo) == AMDGPULibFunc::F32) 946 ? (double)CF->getValueAPF().convertToFloat() 947 : CF->getValueAPF().convertToDouble(); 948 949 V = log2(std::abs(V)); 950 cnval = ConstantFP::get(eltType, V); 951 needcopysign = (FInfo.getId() != AMDGPULibFunc::EI_POWR) && 952 CF->isNegative(); 953 } else { 954 needlog = true; 955 needcopysign = needabs = FInfo.getId() != AMDGPULibFunc::EI_POWR && 956 (!CF || CF->isNegative()); 957 } 958 } else { 959 ConstantDataVector *CDV = dyn_cast<ConstantDataVector>(opr0); 960 961 if (!CDV) { 962 needlog = true; 963 needcopysign = needabs = FInfo.getId() != AMDGPULibFunc::EI_POWR; 964 } else { 965 assert ((int)CDV->getNumElements() == getVecSize(FInfo) && 966 "Wrong vector size detected"); 967 968 SmallVector<double, 0> DVal; 969 for (int i=0; i < getVecSize(FInfo); ++i) { 970 double V = (getArgType(FInfo) == AMDGPULibFunc::F32) 971 ? (double)CDV->getElementAsFloat(i) 972 : CDV->getElementAsDouble(i); 973 if (V < 0.0) needcopysign = true; 974 V = log2(std::abs(V)); 975 DVal.push_back(V); 976 } 977 if (getArgType(FInfo) == AMDGPULibFunc::F32) { 978 SmallVector<float, 0> FVal; 979 for (unsigned i=0; i < DVal.size(); ++i) { 980 FVal.push_back((float)DVal[i]); 981 } 982 ArrayRef<float> tmp(FVal); 983 cnval = ConstantDataVector::get(M->getContext(), tmp); 984 } else { 985 ArrayRef<double> tmp(DVal); 986 cnval = ConstantDataVector::get(M->getContext(), tmp); 987 } 988 } 989 } 990 991 if (needcopysign && (FInfo.getId() == AMDGPULibFunc::EI_POW)) { 992 // We cannot handle corner cases for a general pow() function, give up 993 // unless y is a constant integral value. Then proceed as if it were pown. 994 if (getVecSize(FInfo) == 1) { 995 if (const ConstantFP *CF = dyn_cast<ConstantFP>(opr1)) { 996 double y = (getArgType(FInfo) == AMDGPULibFunc::F32) 997 ? (double)CF->getValueAPF().convertToFloat() 998 : CF->getValueAPF().convertToDouble(); 999 if (y != (double)(int64_t)y) 1000 return false; 1001 } else 1002 return false; 1003 } else { 1004 if (const ConstantDataVector *CDV = dyn_cast<ConstantDataVector>(opr1)) { 1005 for (int i=0; i < getVecSize(FInfo); ++i) { 1006 double y = (getArgType(FInfo) == AMDGPULibFunc::F32) 1007 ? (double)CDV->getElementAsFloat(i) 1008 : CDV->getElementAsDouble(i); 1009 if (y != (double)(int64_t)y) 1010 return false; 1011 } 1012 } else 1013 return false; 1014 } 1015 } 1016 1017 Value *nval; 1018 if (needabs) { 1019 Constant *AbsExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_FABS, 1020 FInfo)); 1021 if (!AbsExpr) 1022 return false; 1023 nval = CreateCallEx(B, AbsExpr, opr0, "__fabs"); 1024 } else { 1025 nval = cnval ? cnval : opr0; 1026 } 1027 if (needlog) { 1028 Constant *LogExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_LOG2, 1029 FInfo)); 1030 if (!LogExpr) 1031 return false; 1032 nval = CreateCallEx(B,LogExpr, nval, "__log2"); 1033 } 1034 1035 if (FInfo.getId() == AMDGPULibFunc::EI_POWN) { 1036 // convert int(32) to fp(f32 or f64) 1037 opr1 = B.CreateSIToFP(opr1, nval->getType(), "pownI2F"); 1038 } 1039 nval = B.CreateFMul(opr1, nval, "__ylogx"); 1040 nval = CreateCallEx(B,ExpExpr, nval, "__exp2"); 1041 1042 if (needcopysign) { 1043 Value *opr_n; 1044 Type* rTy = opr0->getType(); 1045 Type* nTyS = eltType->isDoubleTy() ? B.getInt64Ty() : B.getInt32Ty(); 1046 Type *nTy = nTyS; 1047 if (const VectorType *vTy = dyn_cast<VectorType>(rTy)) 1048 nTy = VectorType::get(nTyS, vTy->getNumElements()); 1049 unsigned size = nTy->getScalarSizeInBits(); 1050 opr_n = CI->getArgOperand(1); 1051 if (opr_n->getType()->isIntegerTy()) 1052 opr_n = B.CreateZExtOrBitCast(opr_n, nTy, "__ytou"); 1053 else 1054 opr_n = B.CreateFPToSI(opr1, nTy, "__ytou"); 1055 1056 Value *sign = B.CreateShl(opr_n, size-1, "__yeven"); 1057 sign = B.CreateAnd(B.CreateBitCast(opr0, nTy), sign, "__pow_sign"); 1058 nval = B.CreateOr(B.CreateBitCast(nval, nTy), sign); 1059 nval = B.CreateBitCast(nval, opr0->getType()); 1060 } 1061 1062 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 1063 << "exp2(" << *opr1 << " * log2(" << *opr0 << "))\n"); 1064 replaceCall(nval); 1065 1066 return true; 1067 } 1068 1069 bool AMDGPULibCalls::fold_rootn(CallInst *CI, IRBuilder<> &B, 1070 const FuncInfo &FInfo) { 1071 Value *opr0 = CI->getArgOperand(0); 1072 Value *opr1 = CI->getArgOperand(1); 1073 1074 ConstantInt *CINT = dyn_cast<ConstantInt>(opr1); 1075 if (!CINT) { 1076 return false; 1077 } 1078 int ci_opr1 = (int)CINT->getSExtValue(); 1079 if (ci_opr1 == 1) { // rootn(x, 1) = x 1080 DEBUG(errs() << "AMDIC: " << *CI 1081 << " ---> " << *opr0 << "\n"); 1082 replaceCall(opr0); 1083 return true; 1084 } 1085 if (ci_opr1 == 2) { // rootn(x, 2) = sqrt(x) 1086 std::vector<const Type*> ParamsTys; 1087 ParamsTys.push_back(opr0->getType()); 1088 Module *M = CI->getModule(); 1089 if (Constant *FPExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_SQRT, 1090 FInfo))) { 1091 DEBUG(errs() << "AMDIC: " << *CI << " ---> sqrt(" << *opr0 << ")\n"); 1092 Value *nval = CreateCallEx(B,FPExpr, opr0, "__rootn2sqrt"); 1093 replaceCall(nval); 1094 return true; 1095 } 1096 } else if (ci_opr1 == 3) { // rootn(x, 3) = cbrt(x) 1097 Module *M = CI->getModule(); 1098 if (Constant *FPExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_CBRT, 1099 FInfo))) { 1100 DEBUG(errs() << "AMDIC: " << *CI << " ---> cbrt(" << *opr0 << ")\n"); 1101 Value *nval = CreateCallEx(B,FPExpr, opr0, "__rootn2cbrt"); 1102 replaceCall(nval); 1103 return true; 1104 } 1105 } else if (ci_opr1 == -1) { // rootn(x, -1) = 1.0/x 1106 DEBUG(errs() << "AMDIC: " << *CI << " ---> 1.0 / " << *opr0 << "\n"); 1107 Value *nval = B.CreateFDiv(ConstantFP::get(opr0->getType(), 1.0), 1108 opr0, 1109 "__rootn2div"); 1110 replaceCall(nval); 1111 return true; 1112 } else if (ci_opr1 == -2) { // rootn(x, -2) = rsqrt(x) 1113 std::vector<const Type*> ParamsTys; 1114 ParamsTys.push_back(opr0->getType()); 1115 Module *M = CI->getModule(); 1116 if (Constant *FPExpr = getFunction(M, AMDGPULibFunc(AMDGPULibFunc::EI_RSQRT, 1117 FInfo))) { 1118 DEBUG(errs() << "AMDIC: " << *CI << " ---> rsqrt(" << *opr0 << ")\n"); 1119 Value *nval = CreateCallEx(B,FPExpr, opr0, "__rootn2rsqrt"); 1120 replaceCall(nval); 1121 return true; 1122 } 1123 } 1124 return false; 1125 } 1126 1127 bool AMDGPULibCalls::fold_fma_mad(CallInst *CI, IRBuilder<> &B, 1128 const FuncInfo &FInfo) { 1129 Value *opr0 = CI->getArgOperand(0); 1130 Value *opr1 = CI->getArgOperand(1); 1131 Value *opr2 = CI->getArgOperand(2); 1132 1133 ConstantFP *CF0 = dyn_cast<ConstantFP>(opr0); 1134 ConstantFP *CF1 = dyn_cast<ConstantFP>(opr1); 1135 if ((CF0 && CF0->isZero()) || (CF1 && CF1->isZero())) { 1136 // fma/mad(a, b, c) = c if a=0 || b=0 1137 DEBUG(errs() << "AMDIC: " << *CI << " ---> " << *opr2 << "\n"); 1138 replaceCall(opr2); 1139 return true; 1140 } 1141 if (CF0 && CF0->isExactlyValue(1.0f)) { 1142 // fma/mad(a, b, c) = b+c if a=1 1143 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 1144 << *opr1 << " + " << *opr2 << "\n"); 1145 Value *nval = B.CreateFAdd(opr1, opr2, "fmaadd"); 1146 replaceCall(nval); 1147 return true; 1148 } 1149 if (CF1 && CF1->isExactlyValue(1.0f)) { 1150 // fma/mad(a, b, c) = a+c if b=1 1151 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 1152 << *opr0 << " + " << *opr2 << "\n"); 1153 Value *nval = B.CreateFAdd(opr0, opr2, "fmaadd"); 1154 replaceCall(nval); 1155 return true; 1156 } 1157 if (ConstantFP *CF = dyn_cast<ConstantFP>(opr2)) { 1158 if (CF->isZero()) { 1159 // fma/mad(a, b, c) = a*b if c=0 1160 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 1161 << *opr0 << " * " << *opr1 << "\n"); 1162 Value *nval = B.CreateFMul(opr0, opr1, "fmamul"); 1163 replaceCall(nval); 1164 return true; 1165 } 1166 } 1167 1168 return false; 1169 } 1170 1171 // Get a scalar native builtin signle argument FP function 1172 Constant* AMDGPULibCalls::getNativeFunction(Module* M, const FuncInfo& FInfo) { 1173 FuncInfo nf = FInfo; 1174 nf.setPrefix(AMDGPULibFunc::NATIVE); 1175 return getFunction(M, nf); 1176 } 1177 1178 // fold sqrt -> native_sqrt (x) 1179 bool AMDGPULibCalls::fold_sqrt(CallInst *CI, IRBuilder<> &B, 1180 const FuncInfo &FInfo) { 1181 if ((getArgType(FInfo) == AMDGPULibFunc::F32 || 1182 getArgType(FInfo) == AMDGPULibFunc::F64) && 1183 (getVecSize(FInfo) == 1) && 1184 (FInfo.getPrefix() != AMDGPULibFunc::NATIVE)) { 1185 if (Constant *FPExpr = getNativeFunction( 1186 CI->getModule(), AMDGPULibFunc(AMDGPULibFunc::EI_SQRT, FInfo))) { 1187 Value *opr0 = CI->getArgOperand(0); 1188 DEBUG(errs() << "AMDIC: " << *CI << " ---> " 1189 << "sqrt(" << *opr0 << ")\n"); 1190 Value *nval = CreateCallEx(B,FPExpr, opr0, "__sqrt"); 1191 replaceCall(nval); 1192 return true; 1193 } 1194 } 1195 return false; 1196 } 1197 1198 // fold sin, cos -> sincos. 1199 bool AMDGPULibCalls::fold_sincos(CallInst *CI, IRBuilder<> &B, 1200 AliasAnalysis *AA) { 1201 AMDGPULibFunc fInfo; 1202 if (!AMDGPULibFunc::parse(CI->getCalledFunction()->getName(), fInfo)) 1203 return false; 1204 1205 assert(fInfo.getId() == AMDGPULibFunc::EI_SIN || 1206 fInfo.getId() == AMDGPULibFunc::EI_COS); 1207 bool const isSin = fInfo.getId() == AMDGPULibFunc::EI_SIN; 1208 1209 Value *CArgVal = CI->getArgOperand(0); 1210 BasicBlock * const CBB = CI->getParent(); 1211 1212 int const MaxScan = 30; 1213 1214 { // fold in load value. 1215 LoadInst *LI = dyn_cast<LoadInst>(CArgVal); 1216 if (LI && LI->getParent() == CBB) { 1217 BasicBlock::iterator BBI = LI->getIterator(); 1218 Value *AvailableVal = FindAvailableLoadedValue(LI, CBB, BBI, MaxScan, AA); 1219 if (AvailableVal) { 1220 CArgVal->replaceAllUsesWith(AvailableVal); 1221 if (CArgVal->getNumUses() == 0) 1222 LI->eraseFromParent(); 1223 CArgVal = CI->getArgOperand(0); 1224 } 1225 } 1226 } 1227 1228 Module *M = CI->getModule(); 1229 fInfo.setId(isSin ? AMDGPULibFunc::EI_COS : AMDGPULibFunc::EI_SIN); 1230 std::string const PairName = fInfo.mangle(); 1231 1232 CallInst *UI = nullptr; 1233 for (User* U : CArgVal->users()) { 1234 CallInst *XI = dyn_cast_or_null<CallInst>(U); 1235 if (!XI || XI == CI || XI->getParent() != CBB) 1236 continue; 1237 1238 Function *UCallee = XI->getCalledFunction(); 1239 if (!UCallee || !UCallee->getName().equals(PairName)) 1240 continue; 1241 1242 BasicBlock::iterator BBI = CI->getIterator(); 1243 if (BBI == CI->getParent()->begin()) 1244 break; 1245 --BBI; 1246 for (int I = MaxScan; I > 0 && BBI != CBB->begin(); --BBI, --I) { 1247 if (cast<Instruction>(BBI) == XI) { 1248 UI = XI; 1249 break; 1250 } 1251 } 1252 if (UI) break; 1253 } 1254 1255 if (!UI) return false; 1256 1257 // Merge the sin and cos. 1258 1259 // for OpenCL 2.0 we have only generic implementation of sincos 1260 // function. 1261 AMDGPULibFunc nf(AMDGPULibFunc::EI_SINCOS, fInfo); 1262 nf.Leads[0].PtrKind = AMDGPULibFunc::GENERIC; 1263 Function *Fsincos = dyn_cast_or_null<Function>(getFunction(M, nf)); 1264 if (!Fsincos) return false; 1265 1266 BasicBlock::iterator ItOld = B.GetInsertPoint(); 1267 AllocaInst *Alloc = insertAlloca(UI, B, "__sincos_"); 1268 B.SetInsertPoint(UI); 1269 1270 Value *P = Alloc; 1271 Type *PTy = Fsincos->getFunctionType()->getParamType(1); 1272 // The allocaInst allocates the memory in private address space. This need 1273 // to be bitcasted to point to the address space of cos pointer type. 1274 // In OpenCL 2.0 this is generic, while in 1.2 that is private. 1275 const AMDGPUAS AS = AMDGPU::getAMDGPUAS(*M); 1276 if (PTy->getPointerAddressSpace() != AS.PRIVATE_ADDRESS) 1277 P = B.CreateAddrSpaceCast(Alloc, PTy); 1278 CallInst *Call = CreateCallEx2(B, Fsincos, UI->getArgOperand(0), P); 1279 1280 DEBUG(errs() << "AMDIC: fold_sincos (" << *CI << ", " << *UI 1281 << ") with " << *Call << "\n"); 1282 1283 if (!isSin) { // CI->cos, UI->sin 1284 B.SetInsertPoint(&*ItOld); 1285 UI->replaceAllUsesWith(&*Call); 1286 Instruction *Reload = B.CreateLoad(Alloc); 1287 CI->replaceAllUsesWith(Reload); 1288 UI->eraseFromParent(); 1289 CI->eraseFromParent(); 1290 } else { // CI->sin, UI->cos 1291 Instruction *Reload = B.CreateLoad(Alloc); 1292 UI->replaceAllUsesWith(Reload); 1293 CI->replaceAllUsesWith(Call); 1294 UI->eraseFromParent(); 1295 CI->eraseFromParent(); 1296 } 1297 return true; 1298 } 1299 1300 // Get insertion point at entry. 1301 BasicBlock::iterator AMDGPULibCalls::getEntryIns(CallInst * UI) { 1302 Function * Func = UI->getParent()->getParent(); 1303 BasicBlock * BB = &Func->getEntryBlock(); 1304 assert(BB && "Entry block not found!"); 1305 BasicBlock::iterator ItNew = BB->begin(); 1306 return ItNew; 1307 } 1308 1309 // Insert a AllocsInst at the beginning of function entry block. 1310 AllocaInst* AMDGPULibCalls::insertAlloca(CallInst *UI, IRBuilder<> &B, 1311 const char *prefix) { 1312 BasicBlock::iterator ItNew = getEntryIns(UI); 1313 Function *UCallee = UI->getCalledFunction(); 1314 Type *RetType = UCallee->getReturnType(); 1315 B.SetInsertPoint(&*ItNew); 1316 AllocaInst *Alloc = B.CreateAlloca(RetType, 0, 1317 std::string(prefix) + UI->getName()); 1318 Alloc->setAlignment(UCallee->getParent()->getDataLayout() 1319 .getTypeAllocSize(RetType)); 1320 return Alloc; 1321 } 1322 1323 bool AMDGPULibCalls::evaluateScalarMathFunc(FuncInfo &FInfo, 1324 double& Res0, double& Res1, 1325 Constant *copr0, Constant *copr1, 1326 Constant *copr2) { 1327 // By default, opr0/opr1/opr3 holds values of float/double type. 1328 // If they are not float/double, each function has to its 1329 // operand separately. 1330 double opr0=0.0, opr1=0.0, opr2=0.0; 1331 ConstantFP *fpopr0 = dyn_cast_or_null<ConstantFP>(copr0); 1332 ConstantFP *fpopr1 = dyn_cast_or_null<ConstantFP>(copr1); 1333 ConstantFP *fpopr2 = dyn_cast_or_null<ConstantFP>(copr2); 1334 if (fpopr0) { 1335 opr0 = (getArgType(FInfo) == AMDGPULibFunc::F64) 1336 ? fpopr0->getValueAPF().convertToDouble() 1337 : (double)fpopr0->getValueAPF().convertToFloat(); 1338 } 1339 1340 if (fpopr1) { 1341 opr1 = (getArgType(FInfo) == AMDGPULibFunc::F64) 1342 ? fpopr1->getValueAPF().convertToDouble() 1343 : (double)fpopr1->getValueAPF().convertToFloat(); 1344 } 1345 1346 if (fpopr2) { 1347 opr2 = (getArgType(FInfo) == AMDGPULibFunc::F64) 1348 ? fpopr2->getValueAPF().convertToDouble() 1349 : (double)fpopr2->getValueAPF().convertToFloat(); 1350 } 1351 1352 switch (FInfo.getId()) { 1353 default : return false; 1354 1355 case AMDGPULibFunc::EI_ACOS: 1356 Res0 = acos(opr0); 1357 return true; 1358 1359 case AMDGPULibFunc::EI_ACOSH: 1360 // acosh(x) == log(x + sqrt(x*x - 1)) 1361 Res0 = log(opr0 + sqrt(opr0*opr0 - 1.0)); 1362 return true; 1363 1364 case AMDGPULibFunc::EI_ACOSPI: 1365 Res0 = acos(opr0) / MATH_PI; 1366 return true; 1367 1368 case AMDGPULibFunc::EI_ASIN: 1369 Res0 = asin(opr0); 1370 return true; 1371 1372 case AMDGPULibFunc::EI_ASINH: 1373 // asinh(x) == log(x + sqrt(x*x + 1)) 1374 Res0 = log(opr0 + sqrt(opr0*opr0 + 1.0)); 1375 return true; 1376 1377 case AMDGPULibFunc::EI_ASINPI: 1378 Res0 = asin(opr0) / MATH_PI; 1379 return true; 1380 1381 case AMDGPULibFunc::EI_ATAN: 1382 Res0 = atan(opr0); 1383 return true; 1384 1385 case AMDGPULibFunc::EI_ATANH: 1386 // atanh(x) == (log(x+1) - log(x-1))/2; 1387 Res0 = (log(opr0 + 1.0) - log(opr0 - 1.0))/2.0; 1388 return true; 1389 1390 case AMDGPULibFunc::EI_ATANPI: 1391 Res0 = atan(opr0) / MATH_PI; 1392 return true; 1393 1394 case AMDGPULibFunc::EI_CBRT: 1395 Res0 = (opr0 < 0.0) ? -pow(-opr0, 1.0/3.0) : pow(opr0, 1.0/3.0); 1396 return true; 1397 1398 case AMDGPULibFunc::EI_COS: 1399 Res0 = cos(opr0); 1400 return true; 1401 1402 case AMDGPULibFunc::EI_COSH: 1403 Res0 = cosh(opr0); 1404 return true; 1405 1406 case AMDGPULibFunc::EI_COSPI: 1407 Res0 = cos(MATH_PI * opr0); 1408 return true; 1409 1410 case AMDGPULibFunc::EI_EXP: 1411 Res0 = exp(opr0); 1412 return true; 1413 1414 case AMDGPULibFunc::EI_EXP2: 1415 Res0 = pow(2.0, opr0); 1416 return true; 1417 1418 case AMDGPULibFunc::EI_EXP10: 1419 Res0 = pow(10.0, opr0); 1420 return true; 1421 1422 case AMDGPULibFunc::EI_EXPM1: 1423 Res0 = exp(opr0) - 1.0; 1424 return true; 1425 1426 case AMDGPULibFunc::EI_LOG: 1427 Res0 = log(opr0); 1428 return true; 1429 1430 case AMDGPULibFunc::EI_LOG2: 1431 Res0 = log(opr0) / log(2.0); 1432 return true; 1433 1434 case AMDGPULibFunc::EI_LOG10: 1435 Res0 = log(opr0) / log(10.0); 1436 return true; 1437 1438 case AMDGPULibFunc::EI_RSQRT: 1439 Res0 = 1.0 / sqrt(opr0); 1440 return true; 1441 1442 case AMDGPULibFunc::EI_SIN: 1443 Res0 = sin(opr0); 1444 return true; 1445 1446 case AMDGPULibFunc::EI_SINH: 1447 Res0 = sinh(opr0); 1448 return true; 1449 1450 case AMDGPULibFunc::EI_SINPI: 1451 Res0 = sin(MATH_PI * opr0); 1452 return true; 1453 1454 case AMDGPULibFunc::EI_SQRT: 1455 Res0 = sqrt(opr0); 1456 return true; 1457 1458 case AMDGPULibFunc::EI_TAN: 1459 Res0 = tan(opr0); 1460 return true; 1461 1462 case AMDGPULibFunc::EI_TANH: 1463 Res0 = tanh(opr0); 1464 return true; 1465 1466 case AMDGPULibFunc::EI_TANPI: 1467 Res0 = tan(MATH_PI * opr0); 1468 return true; 1469 1470 case AMDGPULibFunc::EI_RECIP: 1471 Res0 = 1.0 / opr0; 1472 return true; 1473 1474 // two-arg functions 1475 case AMDGPULibFunc::EI_DIVIDE: 1476 Res0 = opr0 / opr1; 1477 return true; 1478 1479 case AMDGPULibFunc::EI_POW: 1480 case AMDGPULibFunc::EI_POWR: 1481 Res0 = pow(opr0, opr1); 1482 return true; 1483 1484 case AMDGPULibFunc::EI_POWN: { 1485 if (ConstantInt *iopr1 = dyn_cast_or_null<ConstantInt>(copr1)) { 1486 double val = (double)iopr1->getSExtValue(); 1487 Res0 = pow(opr0, val); 1488 return true; 1489 } 1490 return false; 1491 } 1492 1493 case AMDGPULibFunc::EI_ROOTN: { 1494 if (ConstantInt *iopr1 = dyn_cast_or_null<ConstantInt>(copr1)) { 1495 double val = (double)iopr1->getSExtValue(); 1496 Res0 = pow(opr0, 1.0 / val); 1497 return true; 1498 } 1499 return false; 1500 } 1501 1502 // with ptr arg 1503 case AMDGPULibFunc::EI_SINCOS: 1504 Res0 = sin(opr0); 1505 Res1 = cos(opr0); 1506 return true; 1507 1508 // three-arg functions 1509 case AMDGPULibFunc::EI_FMA: 1510 case AMDGPULibFunc::EI_MAD: 1511 Res0 = opr0 * opr1 + opr2; 1512 return true; 1513 } 1514 1515 return false; 1516 } 1517 1518 bool AMDGPULibCalls::evaluateCall(CallInst *aCI, FuncInfo &FInfo) { 1519 int numArgs = (int)aCI->getNumArgOperands(); 1520 if (numArgs > 3) 1521 return false; 1522 1523 Constant *copr0 = nullptr; 1524 Constant *copr1 = nullptr; 1525 Constant *copr2 = nullptr; 1526 if (numArgs > 0) { 1527 if ((copr0 = dyn_cast<Constant>(aCI->getArgOperand(0))) == nullptr) 1528 return false; 1529 } 1530 1531 if (numArgs > 1) { 1532 if ((copr1 = dyn_cast<Constant>(aCI->getArgOperand(1))) == nullptr) { 1533 if (FInfo.getId() != AMDGPULibFunc::EI_SINCOS) 1534 return false; 1535 } 1536 } 1537 1538 if (numArgs > 2) { 1539 if ((copr2 = dyn_cast<Constant>(aCI->getArgOperand(2))) == nullptr) 1540 return false; 1541 } 1542 1543 // At this point, all arguments to aCI are constants. 1544 1545 // max vector size is 16, and sincos will generate two results. 1546 double DVal0[16], DVal1[16]; 1547 bool hasTwoResults = (FInfo.getId() == AMDGPULibFunc::EI_SINCOS); 1548 if (getVecSize(FInfo) == 1) { 1549 if (!evaluateScalarMathFunc(FInfo, DVal0[0], 1550 DVal1[0], copr0, copr1, copr2)) { 1551 return false; 1552 } 1553 } else { 1554 ConstantDataVector *CDV0 = dyn_cast_or_null<ConstantDataVector>(copr0); 1555 ConstantDataVector *CDV1 = dyn_cast_or_null<ConstantDataVector>(copr1); 1556 ConstantDataVector *CDV2 = dyn_cast_or_null<ConstantDataVector>(copr2); 1557 for (int i=0; i < getVecSize(FInfo); ++i) { 1558 Constant *celt0 = CDV0 ? CDV0->getElementAsConstant(i) : nullptr; 1559 Constant *celt1 = CDV1 ? CDV1->getElementAsConstant(i) : nullptr; 1560 Constant *celt2 = CDV2 ? CDV2->getElementAsConstant(i) : nullptr; 1561 if (!evaluateScalarMathFunc(FInfo, DVal0[i], 1562 DVal1[i], celt0, celt1, celt2)) { 1563 return false; 1564 } 1565 } 1566 } 1567 1568 LLVMContext &context = CI->getParent()->getParent()->getContext(); 1569 Constant *nval0, *nval1; 1570 if (getVecSize(FInfo) == 1) { 1571 nval0 = ConstantFP::get(CI->getType(), DVal0[0]); 1572 if (hasTwoResults) 1573 nval1 = ConstantFP::get(CI->getType(), DVal1[0]); 1574 } else { 1575 if (getArgType(FInfo) == AMDGPULibFunc::F32) { 1576 SmallVector <float, 0> FVal0, FVal1; 1577 for (int i=0; i < getVecSize(FInfo); ++i) 1578 FVal0.push_back((float)DVal0[i]); 1579 ArrayRef<float> tmp0(FVal0); 1580 nval0 = ConstantDataVector::get(context, tmp0); 1581 if (hasTwoResults) { 1582 for (int i=0; i < getVecSize(FInfo); ++i) 1583 FVal1.push_back((float)DVal1[i]); 1584 ArrayRef<float> tmp1(FVal1); 1585 nval1 = ConstantDataVector::get(context, tmp1); 1586 } 1587 } else { 1588 ArrayRef<double> tmp0(DVal0); 1589 nval0 = ConstantDataVector::get(context, tmp0); 1590 if (hasTwoResults) { 1591 ArrayRef<double> tmp1(DVal1); 1592 nval1 = ConstantDataVector::get(context, tmp1); 1593 } 1594 } 1595 } 1596 1597 if (hasTwoResults) { 1598 // sincos 1599 assert(FInfo.getId() == AMDGPULibFunc::EI_SINCOS && 1600 "math function with ptr arg not supported yet"); 1601 new StoreInst(nval1, aCI->getArgOperand(1), aCI); 1602 } 1603 1604 replaceCall(nval0); 1605 return true; 1606 } 1607 1608 // Public interface to the Simplify LibCalls pass. 1609 FunctionPass *llvm::createAMDGPUSimplifyLibCallsPass() { 1610 return new AMDGPUSimplifyLibCalls(); 1611 } 1612 1613 FunctionPass *llvm::createAMDGPUUseNativeCallsPass() { 1614 return new AMDGPUUseNativeCalls(); 1615 } 1616 1617 bool AMDGPUSimplifyLibCalls::runOnFunction(Function &F) { 1618 if (skipFunction(F)) 1619 return false; 1620 1621 bool Changed = false; 1622 auto AA = &getAnalysis<AAResultsWrapperPass>().getAAResults(); 1623 1624 DEBUG(dbgs() << "AMDIC: process function "; 1625 F.printAsOperand(dbgs(), false, F.getParent()); 1626 dbgs() << '\n';); 1627 1628 for (auto &BB : F) { 1629 for (BasicBlock::iterator I = BB.begin(), E = BB.end(); I != E; ) { 1630 // Ignore non-calls. 1631 CallInst *CI = dyn_cast<CallInst>(I); 1632 ++I; 1633 if (!CI) continue; 1634 1635 // Ignore indirect calls. 1636 Function *Callee = CI->getCalledFunction(); 1637 if (Callee == 0) continue; 1638 1639 DEBUG(dbgs() << "AMDIC: try folding " << *CI << "\n"; 1640 dbgs().flush()); 1641 if(Simplifier.fold(CI, AA)) 1642 Changed = true; 1643 } 1644 } 1645 return Changed; 1646 } 1647 1648 bool AMDGPUUseNativeCalls::runOnFunction(Function &F) { 1649 if (skipFunction(F) || UseNative.empty()) 1650 return false; 1651 1652 bool Changed = false; 1653 for (auto &BB : F) { 1654 for (BasicBlock::iterator I = BB.begin(), E = BB.end(); I != E; ) { 1655 // Ignore non-calls. 1656 CallInst *CI = dyn_cast<CallInst>(I); 1657 ++I; 1658 if (!CI) continue; 1659 1660 // Ignore indirect calls. 1661 Function *Callee = CI->getCalledFunction(); 1662 if (Callee == 0) continue; 1663 1664 if(Simplifier.useNative(CI)) 1665 Changed = true; 1666 } 1667 } 1668 return Changed; 1669 } 1670