1 //===- llvm/CodeGen/GlobalISel/Utils.cpp -------------------------*- C++ -*-==// 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 This file implements the utility functions used by the GlobalISel 9 /// pipeline. 10 //===----------------------------------------------------------------------===// 11 12 #include "llvm/CodeGen/GlobalISel/Utils.h" 13 #include "llvm/ADT/APFloat.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/Optional.h" 16 #include "llvm/CodeGen/GlobalISel/GISelChangeObserver.h" 17 #include "llvm/CodeGen/GlobalISel/GISelKnownBits.h" 18 #include "llvm/CodeGen/GlobalISel/GenericMachineInstrs.h" 19 #include "llvm/CodeGen/GlobalISel/LostDebugLocObserver.h" 20 #include "llvm/CodeGen/GlobalISel/MIPatternMatch.h" 21 #include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h" 22 #include "llvm/CodeGen/MachineInstr.h" 23 #include "llvm/CodeGen/MachineInstrBuilder.h" 24 #include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h" 25 #include "llvm/CodeGen/MachineRegisterInfo.h" 26 #include "llvm/CodeGen/MachineSizeOpts.h" 27 #include "llvm/CodeGen/RegisterBankInfo.h" 28 #include "llvm/CodeGen/StackProtector.h" 29 #include "llvm/CodeGen/TargetInstrInfo.h" 30 #include "llvm/CodeGen/TargetLowering.h" 31 #include "llvm/CodeGen/TargetPassConfig.h" 32 #include "llvm/CodeGen/TargetRegisterInfo.h" 33 #include "llvm/IR/Constants.h" 34 #include "llvm/Target/TargetMachine.h" 35 #include "llvm/Transforms/Utils/SizeOpts.h" 36 37 #define DEBUG_TYPE "globalisel-utils" 38 39 using namespace llvm; 40 using namespace MIPatternMatch; 41 42 Register llvm::constrainRegToClass(MachineRegisterInfo &MRI, 43 const TargetInstrInfo &TII, 44 const RegisterBankInfo &RBI, Register Reg, 45 const TargetRegisterClass &RegClass) { 46 if (!RBI.constrainGenericRegister(Reg, RegClass, MRI)) 47 return MRI.createVirtualRegister(&RegClass); 48 49 return Reg; 50 } 51 52 Register llvm::constrainOperandRegClass( 53 const MachineFunction &MF, const TargetRegisterInfo &TRI, 54 MachineRegisterInfo &MRI, const TargetInstrInfo &TII, 55 const RegisterBankInfo &RBI, MachineInstr &InsertPt, 56 const TargetRegisterClass &RegClass, MachineOperand &RegMO) { 57 Register Reg = RegMO.getReg(); 58 // Assume physical registers are properly constrained. 59 assert(Register::isVirtualRegister(Reg) && "PhysReg not implemented"); 60 61 Register ConstrainedReg = constrainRegToClass(MRI, TII, RBI, Reg, RegClass); 62 // If we created a new virtual register because the class is not compatible 63 // then create a copy between the new and the old register. 64 if (ConstrainedReg != Reg) { 65 MachineBasicBlock::iterator InsertIt(&InsertPt); 66 MachineBasicBlock &MBB = *InsertPt.getParent(); 67 // FIXME: The copy needs to have the classes constrained for its operands. 68 // Use operand's regbank to get the class for old register (Reg). 69 if (RegMO.isUse()) { 70 BuildMI(MBB, InsertIt, InsertPt.getDebugLoc(), 71 TII.get(TargetOpcode::COPY), ConstrainedReg) 72 .addReg(Reg); 73 } else { 74 assert(RegMO.isDef() && "Must be a definition"); 75 BuildMI(MBB, std::next(InsertIt), InsertPt.getDebugLoc(), 76 TII.get(TargetOpcode::COPY), Reg) 77 .addReg(ConstrainedReg); 78 } 79 if (GISelChangeObserver *Observer = MF.getObserver()) { 80 Observer->changingInstr(*RegMO.getParent()); 81 } 82 RegMO.setReg(ConstrainedReg); 83 if (GISelChangeObserver *Observer = MF.getObserver()) { 84 Observer->changedInstr(*RegMO.getParent()); 85 } 86 } else { 87 if (GISelChangeObserver *Observer = MF.getObserver()) { 88 if (!RegMO.isDef()) { 89 MachineInstr *RegDef = MRI.getVRegDef(Reg); 90 Observer->changedInstr(*RegDef); 91 } 92 Observer->changingAllUsesOfReg(MRI, Reg); 93 Observer->finishedChangingAllUsesOfReg(); 94 } 95 } 96 return ConstrainedReg; 97 } 98 99 Register llvm::constrainOperandRegClass( 100 const MachineFunction &MF, const TargetRegisterInfo &TRI, 101 MachineRegisterInfo &MRI, const TargetInstrInfo &TII, 102 const RegisterBankInfo &RBI, MachineInstr &InsertPt, const MCInstrDesc &II, 103 MachineOperand &RegMO, unsigned OpIdx) { 104 Register Reg = RegMO.getReg(); 105 // Assume physical registers are properly constrained. 106 assert(Register::isVirtualRegister(Reg) && "PhysReg not implemented"); 107 108 const TargetRegisterClass *OpRC = TII.getRegClass(II, OpIdx, &TRI, MF); 109 // Some of the target independent instructions, like COPY, may not impose any 110 // register class constraints on some of their operands: If it's a use, we can 111 // skip constraining as the instruction defining the register would constrain 112 // it. 113 114 if (OpRC) { 115 // Obtain the RC from incoming regbank if it is a proper sub-class. Operands 116 // can have multiple regbanks for a superclass that combine different 117 // register types (E.g., AMDGPU's VGPR and AGPR). The regbank ambiguity 118 // resolved by targets during regbankselect should not be overridden. 119 if (const auto *SubRC = TRI.getCommonSubClass( 120 OpRC, TRI.getConstrainedRegClassForOperand(RegMO, MRI))) 121 OpRC = SubRC; 122 123 OpRC = TRI.getAllocatableClass(OpRC); 124 } 125 126 if (!OpRC) { 127 assert((!isTargetSpecificOpcode(II.getOpcode()) || RegMO.isUse()) && 128 "Register class constraint is required unless either the " 129 "instruction is target independent or the operand is a use"); 130 // FIXME: Just bailing out like this here could be not enough, unless we 131 // expect the users of this function to do the right thing for PHIs and 132 // COPY: 133 // v1 = COPY v0 134 // v2 = COPY v1 135 // v1 here may end up not being constrained at all. Please notice that to 136 // reproduce the issue we likely need a destination pattern of a selection 137 // rule producing such extra copies, not just an input GMIR with them as 138 // every existing target using selectImpl handles copies before calling it 139 // and they never reach this function. 140 return Reg; 141 } 142 return constrainOperandRegClass(MF, TRI, MRI, TII, RBI, InsertPt, *OpRC, 143 RegMO); 144 } 145 146 bool llvm::constrainSelectedInstRegOperands(MachineInstr &I, 147 const TargetInstrInfo &TII, 148 const TargetRegisterInfo &TRI, 149 const RegisterBankInfo &RBI) { 150 assert(!isPreISelGenericOpcode(I.getOpcode()) && 151 "A selected instruction is expected"); 152 MachineBasicBlock &MBB = *I.getParent(); 153 MachineFunction &MF = *MBB.getParent(); 154 MachineRegisterInfo &MRI = MF.getRegInfo(); 155 156 for (unsigned OpI = 0, OpE = I.getNumExplicitOperands(); OpI != OpE; ++OpI) { 157 MachineOperand &MO = I.getOperand(OpI); 158 159 // There's nothing to be done on non-register operands. 160 if (!MO.isReg()) 161 continue; 162 163 LLVM_DEBUG(dbgs() << "Converting operand: " << MO << '\n'); 164 assert(MO.isReg() && "Unsupported non-reg operand"); 165 166 Register Reg = MO.getReg(); 167 // Physical registers don't need to be constrained. 168 if (Register::isPhysicalRegister(Reg)) 169 continue; 170 171 // Register operands with a value of 0 (e.g. predicate operands) don't need 172 // to be constrained. 173 if (Reg == 0) 174 continue; 175 176 // If the operand is a vreg, we should constrain its regclass, and only 177 // insert COPYs if that's impossible. 178 // constrainOperandRegClass does that for us. 179 constrainOperandRegClass(MF, TRI, MRI, TII, RBI, I, I.getDesc(), MO, OpI); 180 181 // Tie uses to defs as indicated in MCInstrDesc if this hasn't already been 182 // done. 183 if (MO.isUse()) { 184 int DefIdx = I.getDesc().getOperandConstraint(OpI, MCOI::TIED_TO); 185 if (DefIdx != -1 && !I.isRegTiedToUseOperand(DefIdx)) 186 I.tieOperands(DefIdx, OpI); 187 } 188 } 189 return true; 190 } 191 192 bool llvm::canReplaceReg(Register DstReg, Register SrcReg, 193 MachineRegisterInfo &MRI) { 194 // Give up if either DstReg or SrcReg is a physical register. 195 if (DstReg.isPhysical() || SrcReg.isPhysical()) 196 return false; 197 // Give up if the types don't match. 198 if (MRI.getType(DstReg) != MRI.getType(SrcReg)) 199 return false; 200 // Replace if either DstReg has no constraints or the register 201 // constraints match. 202 return !MRI.getRegClassOrRegBank(DstReg) || 203 MRI.getRegClassOrRegBank(DstReg) == MRI.getRegClassOrRegBank(SrcReg); 204 } 205 206 bool llvm::isTriviallyDead(const MachineInstr &MI, 207 const MachineRegisterInfo &MRI) { 208 // FIXME: This logical is mostly duplicated with 209 // DeadMachineInstructionElim::isDead. Why is LOCAL_ESCAPE not considered in 210 // MachineInstr::isLabel? 211 212 // Don't delete frame allocation labels. 213 if (MI.getOpcode() == TargetOpcode::LOCAL_ESCAPE) 214 return false; 215 // LIFETIME markers should be preserved even if they seem dead. 216 if (MI.getOpcode() == TargetOpcode::LIFETIME_START || 217 MI.getOpcode() == TargetOpcode::LIFETIME_END) 218 return false; 219 220 // If we can move an instruction, we can remove it. Otherwise, it has 221 // a side-effect of some sort. 222 bool SawStore = false; 223 if (!MI.isSafeToMove(/*AA=*/nullptr, SawStore) && !MI.isPHI()) 224 return false; 225 226 // Instructions without side-effects are dead iff they only define dead vregs. 227 for (auto &MO : MI.operands()) { 228 if (!MO.isReg() || !MO.isDef()) 229 continue; 230 231 Register Reg = MO.getReg(); 232 if (Register::isPhysicalRegister(Reg) || !MRI.use_nodbg_empty(Reg)) 233 return false; 234 } 235 return true; 236 } 237 238 static void reportGISelDiagnostic(DiagnosticSeverity Severity, 239 MachineFunction &MF, 240 const TargetPassConfig &TPC, 241 MachineOptimizationRemarkEmitter &MORE, 242 MachineOptimizationRemarkMissed &R) { 243 bool IsFatal = Severity == DS_Error && 244 TPC.isGlobalISelAbortEnabled(); 245 // Print the function name explicitly if we don't have a debug location (which 246 // makes the diagnostic less useful) or if we're going to emit a raw error. 247 if (!R.getLocation().isValid() || IsFatal) 248 R << (" (in function: " + MF.getName() + ")").str(); 249 250 if (IsFatal) 251 report_fatal_error(Twine(R.getMsg())); 252 else 253 MORE.emit(R); 254 } 255 256 void llvm::reportGISelWarning(MachineFunction &MF, const TargetPassConfig &TPC, 257 MachineOptimizationRemarkEmitter &MORE, 258 MachineOptimizationRemarkMissed &R) { 259 reportGISelDiagnostic(DS_Warning, MF, TPC, MORE, R); 260 } 261 262 void llvm::reportGISelFailure(MachineFunction &MF, const TargetPassConfig &TPC, 263 MachineOptimizationRemarkEmitter &MORE, 264 MachineOptimizationRemarkMissed &R) { 265 MF.getProperties().set(MachineFunctionProperties::Property::FailedISel); 266 reportGISelDiagnostic(DS_Error, MF, TPC, MORE, R); 267 } 268 269 void llvm::reportGISelFailure(MachineFunction &MF, const TargetPassConfig &TPC, 270 MachineOptimizationRemarkEmitter &MORE, 271 const char *PassName, StringRef Msg, 272 const MachineInstr &MI) { 273 MachineOptimizationRemarkMissed R(PassName, "GISelFailure: ", 274 MI.getDebugLoc(), MI.getParent()); 275 R << Msg; 276 // Printing MI is expensive; only do it if expensive remarks are enabled. 277 if (TPC.isGlobalISelAbortEnabled() || MORE.allowExtraAnalysis(PassName)) 278 R << ": " << ore::MNV("Inst", MI); 279 reportGISelFailure(MF, TPC, MORE, R); 280 } 281 282 Optional<APInt> llvm::getIConstantVRegVal(Register VReg, 283 const MachineRegisterInfo &MRI) { 284 Optional<ValueAndVReg> ValAndVReg = getIConstantVRegValWithLookThrough( 285 VReg, MRI, /*LookThroughInstrs*/ false); 286 assert((!ValAndVReg || ValAndVReg->VReg == VReg) && 287 "Value found while looking through instrs"); 288 if (!ValAndVReg) 289 return None; 290 return ValAndVReg->Value; 291 } 292 293 Optional<int64_t> 294 llvm::getIConstantVRegSExtVal(Register VReg, const MachineRegisterInfo &MRI) { 295 Optional<APInt> Val = getIConstantVRegVal(VReg, MRI); 296 if (Val && Val->getBitWidth() <= 64) 297 return Val->getSExtValue(); 298 return None; 299 } 300 301 namespace { 302 303 typedef std::function<bool(const MachineInstr *)> IsOpcodeFn; 304 typedef std::function<Optional<APInt>(const MachineInstr *MI)> GetAPCstFn; 305 306 Optional<ValueAndVReg> getConstantVRegValWithLookThrough( 307 Register VReg, const MachineRegisterInfo &MRI, IsOpcodeFn IsConstantOpcode, 308 GetAPCstFn getAPCstValue, bool LookThroughInstrs = true, 309 bool LookThroughAnyExt = false) { 310 SmallVector<std::pair<unsigned, unsigned>, 4> SeenOpcodes; 311 MachineInstr *MI; 312 313 while ((MI = MRI.getVRegDef(VReg)) && !IsConstantOpcode(MI) && 314 LookThroughInstrs) { 315 switch (MI->getOpcode()) { 316 case TargetOpcode::G_ANYEXT: 317 if (!LookThroughAnyExt) 318 return None; 319 LLVM_FALLTHROUGH; 320 case TargetOpcode::G_TRUNC: 321 case TargetOpcode::G_SEXT: 322 case TargetOpcode::G_ZEXT: 323 SeenOpcodes.push_back(std::make_pair( 324 MI->getOpcode(), 325 MRI.getType(MI->getOperand(0).getReg()).getSizeInBits())); 326 VReg = MI->getOperand(1).getReg(); 327 break; 328 case TargetOpcode::COPY: 329 VReg = MI->getOperand(1).getReg(); 330 if (Register::isPhysicalRegister(VReg)) 331 return None; 332 break; 333 case TargetOpcode::G_INTTOPTR: 334 VReg = MI->getOperand(1).getReg(); 335 break; 336 default: 337 return None; 338 } 339 } 340 if (!MI || !IsConstantOpcode(MI)) 341 return None; 342 343 Optional<APInt> MaybeVal = getAPCstValue(MI); 344 if (!MaybeVal) 345 return None; 346 APInt &Val = *MaybeVal; 347 while (!SeenOpcodes.empty()) { 348 std::pair<unsigned, unsigned> OpcodeAndSize = SeenOpcodes.pop_back_val(); 349 switch (OpcodeAndSize.first) { 350 case TargetOpcode::G_TRUNC: 351 Val = Val.trunc(OpcodeAndSize.second); 352 break; 353 case TargetOpcode::G_ANYEXT: 354 case TargetOpcode::G_SEXT: 355 Val = Val.sext(OpcodeAndSize.second); 356 break; 357 case TargetOpcode::G_ZEXT: 358 Val = Val.zext(OpcodeAndSize.second); 359 break; 360 } 361 } 362 363 return ValueAndVReg{Val, VReg}; 364 } 365 366 bool isIConstant(const MachineInstr *MI) { 367 if (!MI) 368 return false; 369 return MI->getOpcode() == TargetOpcode::G_CONSTANT; 370 } 371 372 bool isFConstant(const MachineInstr *MI) { 373 if (!MI) 374 return false; 375 return MI->getOpcode() == TargetOpcode::G_FCONSTANT; 376 } 377 378 bool isAnyConstant(const MachineInstr *MI) { 379 if (!MI) 380 return false; 381 unsigned Opc = MI->getOpcode(); 382 return Opc == TargetOpcode::G_CONSTANT || Opc == TargetOpcode::G_FCONSTANT; 383 } 384 385 Optional<APInt> getCImmAsAPInt(const MachineInstr *MI) { 386 const MachineOperand &CstVal = MI->getOperand(1); 387 if (CstVal.isCImm()) 388 return CstVal.getCImm()->getValue(); 389 return None; 390 } 391 392 Optional<APInt> getCImmOrFPImmAsAPInt(const MachineInstr *MI) { 393 const MachineOperand &CstVal = MI->getOperand(1); 394 if (CstVal.isCImm()) 395 return CstVal.getCImm()->getValue(); 396 if (CstVal.isFPImm()) 397 return CstVal.getFPImm()->getValueAPF().bitcastToAPInt(); 398 return None; 399 } 400 401 } // end anonymous namespace 402 403 Optional<ValueAndVReg> llvm::getIConstantVRegValWithLookThrough( 404 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs) { 405 return getConstantVRegValWithLookThrough(VReg, MRI, isIConstant, 406 getCImmAsAPInt, LookThroughInstrs); 407 } 408 409 Optional<ValueAndVReg> llvm::getAnyConstantVRegValWithLookThrough( 410 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs, 411 bool LookThroughAnyExt) { 412 return getConstantVRegValWithLookThrough( 413 VReg, MRI, isAnyConstant, getCImmOrFPImmAsAPInt, LookThroughInstrs, 414 LookThroughAnyExt); 415 } 416 417 Optional<FPValueAndVReg> llvm::getFConstantVRegValWithLookThrough( 418 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs) { 419 auto Reg = getConstantVRegValWithLookThrough( 420 VReg, MRI, isFConstant, getCImmOrFPImmAsAPInt, LookThroughInstrs); 421 if (!Reg) 422 return None; 423 return FPValueAndVReg{getConstantFPVRegVal(Reg->VReg, MRI)->getValueAPF(), 424 Reg->VReg}; 425 } 426 427 const ConstantFP * 428 llvm::getConstantFPVRegVal(Register VReg, const MachineRegisterInfo &MRI) { 429 MachineInstr *MI = MRI.getVRegDef(VReg); 430 if (TargetOpcode::G_FCONSTANT != MI->getOpcode()) 431 return nullptr; 432 return MI->getOperand(1).getFPImm(); 433 } 434 435 Optional<DefinitionAndSourceRegister> 436 llvm::getDefSrcRegIgnoringCopies(Register Reg, const MachineRegisterInfo &MRI) { 437 Register DefSrcReg = Reg; 438 auto *DefMI = MRI.getVRegDef(Reg); 439 auto DstTy = MRI.getType(DefMI->getOperand(0).getReg()); 440 if (!DstTy.isValid()) 441 return None; 442 unsigned Opc = DefMI->getOpcode(); 443 while (Opc == TargetOpcode::COPY || isPreISelGenericOptimizationHint(Opc)) { 444 Register SrcReg = DefMI->getOperand(1).getReg(); 445 auto SrcTy = MRI.getType(SrcReg); 446 if (!SrcTy.isValid()) 447 break; 448 DefMI = MRI.getVRegDef(SrcReg); 449 DefSrcReg = SrcReg; 450 Opc = DefMI->getOpcode(); 451 } 452 return DefinitionAndSourceRegister{DefMI, DefSrcReg}; 453 } 454 455 MachineInstr *llvm::getDefIgnoringCopies(Register Reg, 456 const MachineRegisterInfo &MRI) { 457 Optional<DefinitionAndSourceRegister> DefSrcReg = 458 getDefSrcRegIgnoringCopies(Reg, MRI); 459 return DefSrcReg ? DefSrcReg->MI : nullptr; 460 } 461 462 Register llvm::getSrcRegIgnoringCopies(Register Reg, 463 const MachineRegisterInfo &MRI) { 464 Optional<DefinitionAndSourceRegister> DefSrcReg = 465 getDefSrcRegIgnoringCopies(Reg, MRI); 466 return DefSrcReg ? DefSrcReg->Reg : Register(); 467 } 468 469 MachineInstr *llvm::getOpcodeDef(unsigned Opcode, Register Reg, 470 const MachineRegisterInfo &MRI) { 471 MachineInstr *DefMI = getDefIgnoringCopies(Reg, MRI); 472 return DefMI && DefMI->getOpcode() == Opcode ? DefMI : nullptr; 473 } 474 475 APFloat llvm::getAPFloatFromSize(double Val, unsigned Size) { 476 if (Size == 32) 477 return APFloat(float(Val)); 478 if (Size == 64) 479 return APFloat(Val); 480 if (Size != 16) 481 llvm_unreachable("Unsupported FPConstant size"); 482 bool Ignored; 483 APFloat APF(Val); 484 APF.convert(APFloat::IEEEhalf(), APFloat::rmNearestTiesToEven, &Ignored); 485 return APF; 486 } 487 488 Optional<APInt> llvm::ConstantFoldBinOp(unsigned Opcode, const Register Op1, 489 const Register Op2, 490 const MachineRegisterInfo &MRI) { 491 auto MaybeOp2Cst = getAnyConstantVRegValWithLookThrough(Op2, MRI, false); 492 if (!MaybeOp2Cst) 493 return None; 494 495 auto MaybeOp1Cst = getAnyConstantVRegValWithLookThrough(Op1, MRI, false); 496 if (!MaybeOp1Cst) 497 return None; 498 499 const APInt &C1 = MaybeOp1Cst->Value; 500 const APInt &C2 = MaybeOp2Cst->Value; 501 switch (Opcode) { 502 default: 503 break; 504 case TargetOpcode::G_ADD: 505 case TargetOpcode::G_PTR_ADD: 506 return C1 + C2; 507 case TargetOpcode::G_AND: 508 return C1 & C2; 509 case TargetOpcode::G_ASHR: 510 return C1.ashr(C2); 511 case TargetOpcode::G_LSHR: 512 return C1.lshr(C2); 513 case TargetOpcode::G_MUL: 514 return C1 * C2; 515 case TargetOpcode::G_OR: 516 return C1 | C2; 517 case TargetOpcode::G_SHL: 518 return C1 << C2; 519 case TargetOpcode::G_SUB: 520 return C1 - C2; 521 case TargetOpcode::G_XOR: 522 return C1 ^ C2; 523 case TargetOpcode::G_UDIV: 524 if (!C2.getBoolValue()) 525 break; 526 return C1.udiv(C2); 527 case TargetOpcode::G_SDIV: 528 if (!C2.getBoolValue()) 529 break; 530 return C1.sdiv(C2); 531 case TargetOpcode::G_UREM: 532 if (!C2.getBoolValue()) 533 break; 534 return C1.urem(C2); 535 case TargetOpcode::G_SREM: 536 if (!C2.getBoolValue()) 537 break; 538 return C1.srem(C2); 539 case TargetOpcode::G_SMIN: 540 return APIntOps::smin(C1, C2); 541 case TargetOpcode::G_SMAX: 542 return APIntOps::smax(C1, C2); 543 case TargetOpcode::G_UMIN: 544 return APIntOps::umin(C1, C2); 545 case TargetOpcode::G_UMAX: 546 return APIntOps::umax(C1, C2); 547 } 548 549 return None; 550 } 551 552 Optional<APFloat> llvm::ConstantFoldFPBinOp(unsigned Opcode, const Register Op1, 553 const Register Op2, 554 const MachineRegisterInfo &MRI) { 555 const ConstantFP *Op2Cst = getConstantFPVRegVal(Op2, MRI); 556 if (!Op2Cst) 557 return None; 558 559 const ConstantFP *Op1Cst = getConstantFPVRegVal(Op1, MRI); 560 if (!Op1Cst) 561 return None; 562 563 APFloat C1 = Op1Cst->getValueAPF(); 564 const APFloat &C2 = Op2Cst->getValueAPF(); 565 switch (Opcode) { 566 case TargetOpcode::G_FADD: 567 C1.add(C2, APFloat::rmNearestTiesToEven); 568 return C1; 569 case TargetOpcode::G_FSUB: 570 C1.subtract(C2, APFloat::rmNearestTiesToEven); 571 return C1; 572 case TargetOpcode::G_FMUL: 573 C1.multiply(C2, APFloat::rmNearestTiesToEven); 574 return C1; 575 case TargetOpcode::G_FDIV: 576 C1.divide(C2, APFloat::rmNearestTiesToEven); 577 return C1; 578 case TargetOpcode::G_FREM: 579 C1.mod(C2); 580 return C1; 581 case TargetOpcode::G_FCOPYSIGN: 582 C1.copySign(C2); 583 return C1; 584 case TargetOpcode::G_FMINNUM: 585 return minnum(C1, C2); 586 case TargetOpcode::G_FMAXNUM: 587 return maxnum(C1, C2); 588 case TargetOpcode::G_FMINIMUM: 589 return minimum(C1, C2); 590 case TargetOpcode::G_FMAXIMUM: 591 return maximum(C1, C2); 592 case TargetOpcode::G_FMINNUM_IEEE: 593 case TargetOpcode::G_FMAXNUM_IEEE: 594 // FIXME: These operations were unfortunately named. fminnum/fmaxnum do not 595 // follow the IEEE behavior for signaling nans and follow libm's fmin/fmax, 596 // and currently there isn't a nice wrapper in APFloat for the version with 597 // correct snan handling. 598 break; 599 default: 600 break; 601 } 602 603 return None; 604 } 605 606 Register llvm::ConstantFoldVectorBinop(unsigned Opcode, const Register Op1, 607 const Register Op2, 608 const MachineRegisterInfo &MRI, 609 MachineIRBuilder &MIB) { 610 auto *SrcVec2 = getOpcodeDef<GBuildVector>(Op2, MRI); 611 if (!SrcVec2) 612 return Register(); 613 614 auto *SrcVec1 = getOpcodeDef<GBuildVector>(Op1, MRI); 615 if (!SrcVec1) 616 return Register(); 617 618 const LLT EltTy = MRI.getType(SrcVec1->getSourceReg(0)); 619 620 SmallVector<Register, 16> FoldedElements; 621 for (unsigned Idx = 0, E = SrcVec1->getNumSources(); Idx < E; ++Idx) { 622 auto MaybeCst = ConstantFoldBinOp(Opcode, SrcVec1->getSourceReg(Idx), 623 SrcVec2->getSourceReg(Idx), MRI); 624 if (!MaybeCst) 625 return Register(); 626 auto FoldedCstReg = MIB.buildConstant(EltTy, *MaybeCst).getReg(0); 627 FoldedElements.emplace_back(FoldedCstReg); 628 } 629 // Create the new vector constant. 630 auto CstVec = 631 MIB.buildBuildVector(MRI.getType(SrcVec1->getReg(0)), FoldedElements); 632 return CstVec.getReg(0); 633 } 634 635 bool llvm::isKnownNeverNaN(Register Val, const MachineRegisterInfo &MRI, 636 bool SNaN) { 637 const MachineInstr *DefMI = MRI.getVRegDef(Val); 638 if (!DefMI) 639 return false; 640 641 const TargetMachine& TM = DefMI->getMF()->getTarget(); 642 if (DefMI->getFlag(MachineInstr::FmNoNans) || TM.Options.NoNaNsFPMath) 643 return true; 644 645 // If the value is a constant, we can obviously see if it is a NaN or not. 646 if (const ConstantFP *FPVal = getConstantFPVRegVal(Val, MRI)) { 647 return !FPVal->getValueAPF().isNaN() || 648 (SNaN && !FPVal->getValueAPF().isSignaling()); 649 } 650 651 if (DefMI->getOpcode() == TargetOpcode::G_BUILD_VECTOR) { 652 for (const auto &Op : DefMI->uses()) 653 if (!isKnownNeverNaN(Op.getReg(), MRI, SNaN)) 654 return false; 655 return true; 656 } 657 658 switch (DefMI->getOpcode()) { 659 default: 660 break; 661 case TargetOpcode::G_FMINNUM_IEEE: 662 case TargetOpcode::G_FMAXNUM_IEEE: { 663 if (SNaN) 664 return true; 665 // This can return a NaN if either operand is an sNaN, or if both operands 666 // are NaN. 667 return (isKnownNeverNaN(DefMI->getOperand(1).getReg(), MRI) && 668 isKnownNeverSNaN(DefMI->getOperand(2).getReg(), MRI)) || 669 (isKnownNeverSNaN(DefMI->getOperand(1).getReg(), MRI) && 670 isKnownNeverNaN(DefMI->getOperand(2).getReg(), MRI)); 671 } 672 case TargetOpcode::G_FMINNUM: 673 case TargetOpcode::G_FMAXNUM: { 674 // Only one needs to be known not-nan, since it will be returned if the 675 // other ends up being one. 676 return isKnownNeverNaN(DefMI->getOperand(1).getReg(), MRI, SNaN) || 677 isKnownNeverNaN(DefMI->getOperand(2).getReg(), MRI, SNaN); 678 } 679 } 680 681 if (SNaN) { 682 // FP operations quiet. For now, just handle the ones inserted during 683 // legalization. 684 switch (DefMI->getOpcode()) { 685 case TargetOpcode::G_FPEXT: 686 case TargetOpcode::G_FPTRUNC: 687 case TargetOpcode::G_FCANONICALIZE: 688 return true; 689 default: 690 return false; 691 } 692 } 693 694 return false; 695 } 696 697 Align llvm::inferAlignFromPtrInfo(MachineFunction &MF, 698 const MachinePointerInfo &MPO) { 699 auto PSV = MPO.V.dyn_cast<const PseudoSourceValue *>(); 700 if (auto FSPV = dyn_cast_or_null<FixedStackPseudoSourceValue>(PSV)) { 701 MachineFrameInfo &MFI = MF.getFrameInfo(); 702 return commonAlignment(MFI.getObjectAlign(FSPV->getFrameIndex()), 703 MPO.Offset); 704 } 705 706 if (const Value *V = MPO.V.dyn_cast<const Value *>()) { 707 const Module *M = MF.getFunction().getParent(); 708 return V->getPointerAlignment(M->getDataLayout()); 709 } 710 711 return Align(1); 712 } 713 714 Register llvm::getFunctionLiveInPhysReg(MachineFunction &MF, 715 const TargetInstrInfo &TII, 716 MCRegister PhysReg, 717 const TargetRegisterClass &RC, 718 const DebugLoc &DL, LLT RegTy) { 719 MachineBasicBlock &EntryMBB = MF.front(); 720 MachineRegisterInfo &MRI = MF.getRegInfo(); 721 Register LiveIn = MRI.getLiveInVirtReg(PhysReg); 722 if (LiveIn) { 723 MachineInstr *Def = MRI.getVRegDef(LiveIn); 724 if (Def) { 725 // FIXME: Should the verifier check this is in the entry block? 726 assert(Def->getParent() == &EntryMBB && "live-in copy not in entry block"); 727 return LiveIn; 728 } 729 730 // It's possible the incoming argument register and copy was added during 731 // lowering, but later deleted due to being/becoming dead. If this happens, 732 // re-insert the copy. 733 } else { 734 // The live in register was not present, so add it. 735 LiveIn = MF.addLiveIn(PhysReg, &RC); 736 if (RegTy.isValid()) 737 MRI.setType(LiveIn, RegTy); 738 } 739 740 BuildMI(EntryMBB, EntryMBB.begin(), DL, TII.get(TargetOpcode::COPY), LiveIn) 741 .addReg(PhysReg); 742 if (!EntryMBB.isLiveIn(PhysReg)) 743 EntryMBB.addLiveIn(PhysReg); 744 return LiveIn; 745 } 746 747 Optional<APInt> llvm::ConstantFoldExtOp(unsigned Opcode, const Register Op1, 748 uint64_t Imm, 749 const MachineRegisterInfo &MRI) { 750 auto MaybeOp1Cst = getIConstantVRegVal(Op1, MRI); 751 if (MaybeOp1Cst) { 752 switch (Opcode) { 753 default: 754 break; 755 case TargetOpcode::G_SEXT_INREG: { 756 LLT Ty = MRI.getType(Op1); 757 return MaybeOp1Cst->trunc(Imm).sext(Ty.getScalarSizeInBits()); 758 } 759 } 760 } 761 return None; 762 } 763 764 Optional<APFloat> llvm::ConstantFoldIntToFloat(unsigned Opcode, LLT DstTy, 765 Register Src, 766 const MachineRegisterInfo &MRI) { 767 assert(Opcode == TargetOpcode::G_SITOFP || Opcode == TargetOpcode::G_UITOFP); 768 if (auto MaybeSrcVal = getIConstantVRegVal(Src, MRI)) { 769 APFloat DstVal(getFltSemanticForLLT(DstTy)); 770 DstVal.convertFromAPInt(*MaybeSrcVal, Opcode == TargetOpcode::G_SITOFP, 771 APFloat::rmNearestTiesToEven); 772 return DstVal; 773 } 774 return None; 775 } 776 777 Optional<SmallVector<unsigned>> 778 llvm::ConstantFoldCTLZ(Register Src, const MachineRegisterInfo &MRI) { 779 LLT Ty = MRI.getType(Src); 780 SmallVector<unsigned> FoldedCTLZs; 781 auto tryFoldScalar = [&](Register R) -> Optional<unsigned> { 782 auto MaybeCst = getIConstantVRegVal(R, MRI); 783 if (!MaybeCst) 784 return None; 785 return MaybeCst->countLeadingZeros(); 786 }; 787 if (Ty.isVector()) { 788 // Try to constant fold each element. 789 auto *BV = getOpcodeDef<GBuildVector>(Src, MRI); 790 if (!BV) 791 return None; 792 for (unsigned SrcIdx = 0; SrcIdx < BV->getNumSources(); ++SrcIdx) { 793 if (auto MaybeFold = tryFoldScalar(BV->getSourceReg(SrcIdx))) { 794 FoldedCTLZs.emplace_back(*MaybeFold); 795 continue; 796 } 797 return None; 798 } 799 return FoldedCTLZs; 800 } 801 if (auto MaybeCst = tryFoldScalar(Src)) { 802 FoldedCTLZs.emplace_back(*MaybeCst); 803 return FoldedCTLZs; 804 } 805 return None; 806 } 807 808 bool llvm::isKnownToBeAPowerOfTwo(Register Reg, const MachineRegisterInfo &MRI, 809 GISelKnownBits *KB) { 810 Optional<DefinitionAndSourceRegister> DefSrcReg = 811 getDefSrcRegIgnoringCopies(Reg, MRI); 812 if (!DefSrcReg) 813 return false; 814 815 const MachineInstr &MI = *DefSrcReg->MI; 816 const LLT Ty = MRI.getType(Reg); 817 818 switch (MI.getOpcode()) { 819 case TargetOpcode::G_CONSTANT: { 820 unsigned BitWidth = Ty.getScalarSizeInBits(); 821 const ConstantInt *CI = MI.getOperand(1).getCImm(); 822 return CI->getValue().zextOrTrunc(BitWidth).isPowerOf2(); 823 } 824 case TargetOpcode::G_SHL: { 825 // A left-shift of a constant one will have exactly one bit set because 826 // shifting the bit off the end is undefined. 827 828 // TODO: Constant splat 829 if (auto ConstLHS = getIConstantVRegVal(MI.getOperand(1).getReg(), MRI)) { 830 if (*ConstLHS == 1) 831 return true; 832 } 833 834 break; 835 } 836 case TargetOpcode::G_LSHR: { 837 if (auto ConstLHS = getIConstantVRegVal(MI.getOperand(1).getReg(), MRI)) { 838 if (ConstLHS->isSignMask()) 839 return true; 840 } 841 842 break; 843 } 844 case TargetOpcode::G_BUILD_VECTOR: { 845 // TODO: Probably should have a recursion depth guard since you could have 846 // bitcasted vector elements. 847 for (const MachineOperand &MO : llvm::drop_begin(MI.operands())) 848 if (!isKnownToBeAPowerOfTwo(MO.getReg(), MRI, KB)) 849 return false; 850 851 return true; 852 } 853 case TargetOpcode::G_BUILD_VECTOR_TRUNC: { 854 // Only handle constants since we would need to know if number of leading 855 // zeros is greater than the truncation amount. 856 const unsigned BitWidth = Ty.getScalarSizeInBits(); 857 for (const MachineOperand &MO : llvm::drop_begin(MI.operands())) { 858 auto Const = getIConstantVRegVal(MO.getReg(), MRI); 859 if (!Const || !Const->zextOrTrunc(BitWidth).isPowerOf2()) 860 return false; 861 } 862 863 return true; 864 } 865 default: 866 break; 867 } 868 869 if (!KB) 870 return false; 871 872 // More could be done here, though the above checks are enough 873 // to handle some common cases. 874 875 // Fall back to computeKnownBits to catch other known cases. 876 KnownBits Known = KB->getKnownBits(Reg); 877 return (Known.countMaxPopulation() == 1) && (Known.countMinPopulation() == 1); 878 } 879 880 void llvm::getSelectionDAGFallbackAnalysisUsage(AnalysisUsage &AU) { 881 AU.addPreserved<StackProtector>(); 882 } 883 884 static unsigned getLCMSize(unsigned OrigSize, unsigned TargetSize) { 885 unsigned Mul = OrigSize * TargetSize; 886 unsigned GCDSize = greatestCommonDivisor(OrigSize, TargetSize); 887 return Mul / GCDSize; 888 } 889 890 LLT llvm::getLCMType(LLT OrigTy, LLT TargetTy) { 891 const unsigned OrigSize = OrigTy.getSizeInBits(); 892 const unsigned TargetSize = TargetTy.getSizeInBits(); 893 894 if (OrigSize == TargetSize) 895 return OrigTy; 896 897 if (OrigTy.isVector()) { 898 const LLT OrigElt = OrigTy.getElementType(); 899 900 if (TargetTy.isVector()) { 901 const LLT TargetElt = TargetTy.getElementType(); 902 903 if (OrigElt.getSizeInBits() == TargetElt.getSizeInBits()) { 904 int GCDElts = greatestCommonDivisor(OrigTy.getNumElements(), 905 TargetTy.getNumElements()); 906 // Prefer the original element type. 907 ElementCount Mul = OrigTy.getElementCount() * TargetTy.getNumElements(); 908 return LLT::vector(Mul.divideCoefficientBy(GCDElts), 909 OrigTy.getElementType()); 910 } 911 } else { 912 if (OrigElt.getSizeInBits() == TargetSize) 913 return OrigTy; 914 } 915 916 unsigned LCMSize = getLCMSize(OrigSize, TargetSize); 917 return LLT::fixed_vector(LCMSize / OrigElt.getSizeInBits(), OrigElt); 918 } 919 920 if (TargetTy.isVector()) { 921 unsigned LCMSize = getLCMSize(OrigSize, TargetSize); 922 return LLT::fixed_vector(LCMSize / OrigSize, OrigTy); 923 } 924 925 unsigned LCMSize = getLCMSize(OrigSize, TargetSize); 926 927 // Preserve pointer types. 928 if (LCMSize == OrigSize) 929 return OrigTy; 930 if (LCMSize == TargetSize) 931 return TargetTy; 932 933 return LLT::scalar(LCMSize); 934 } 935 936 LLT llvm::getCoverTy(LLT OrigTy, LLT TargetTy) { 937 if (!OrigTy.isVector() || !TargetTy.isVector() || OrigTy == TargetTy || 938 (OrigTy.getScalarSizeInBits() != TargetTy.getScalarSizeInBits())) 939 return getLCMType(OrigTy, TargetTy); 940 941 unsigned OrigTyNumElts = OrigTy.getNumElements(); 942 unsigned TargetTyNumElts = TargetTy.getNumElements(); 943 if (OrigTyNumElts % TargetTyNumElts == 0) 944 return OrigTy; 945 946 unsigned NumElts = alignTo(OrigTyNumElts, TargetTyNumElts); 947 return LLT::scalarOrVector(ElementCount::getFixed(NumElts), 948 OrigTy.getElementType()); 949 } 950 951 LLT llvm::getGCDType(LLT OrigTy, LLT TargetTy) { 952 const unsigned OrigSize = OrigTy.getSizeInBits(); 953 const unsigned TargetSize = TargetTy.getSizeInBits(); 954 955 if (OrigSize == TargetSize) 956 return OrigTy; 957 958 if (OrigTy.isVector()) { 959 LLT OrigElt = OrigTy.getElementType(); 960 if (TargetTy.isVector()) { 961 LLT TargetElt = TargetTy.getElementType(); 962 if (OrigElt.getSizeInBits() == TargetElt.getSizeInBits()) { 963 int GCD = greatestCommonDivisor(OrigTy.getNumElements(), 964 TargetTy.getNumElements()); 965 return LLT::scalarOrVector(ElementCount::getFixed(GCD), OrigElt); 966 } 967 } else { 968 // If the source is a vector of pointers, return a pointer element. 969 if (OrigElt.getSizeInBits() == TargetSize) 970 return OrigElt; 971 } 972 973 unsigned GCD = greatestCommonDivisor(OrigSize, TargetSize); 974 if (GCD == OrigElt.getSizeInBits()) 975 return OrigElt; 976 977 // If we can't produce the original element type, we have to use a smaller 978 // scalar. 979 if (GCD < OrigElt.getSizeInBits()) 980 return LLT::scalar(GCD); 981 return LLT::fixed_vector(GCD / OrigElt.getSizeInBits(), OrigElt); 982 } 983 984 if (TargetTy.isVector()) { 985 // Try to preserve the original element type. 986 LLT TargetElt = TargetTy.getElementType(); 987 if (TargetElt.getSizeInBits() == OrigSize) 988 return OrigTy; 989 } 990 991 unsigned GCD = greatestCommonDivisor(OrigSize, TargetSize); 992 return LLT::scalar(GCD); 993 } 994 995 Optional<int> llvm::getSplatIndex(MachineInstr &MI) { 996 assert(MI.getOpcode() == TargetOpcode::G_SHUFFLE_VECTOR && 997 "Only G_SHUFFLE_VECTOR can have a splat index!"); 998 ArrayRef<int> Mask = MI.getOperand(3).getShuffleMask(); 999 auto FirstDefinedIdx = find_if(Mask, [](int Elt) { return Elt >= 0; }); 1000 1001 // If all elements are undefined, this shuffle can be considered a splat. 1002 // Return 0 for better potential for callers to simplify. 1003 if (FirstDefinedIdx == Mask.end()) 1004 return 0; 1005 1006 // Make sure all remaining elements are either undef or the same 1007 // as the first non-undef value. 1008 int SplatValue = *FirstDefinedIdx; 1009 if (any_of(make_range(std::next(FirstDefinedIdx), Mask.end()), 1010 [&SplatValue](int Elt) { return Elt >= 0 && Elt != SplatValue; })) 1011 return None; 1012 1013 return SplatValue; 1014 } 1015 1016 static bool isBuildVectorOp(unsigned Opcode) { 1017 return Opcode == TargetOpcode::G_BUILD_VECTOR || 1018 Opcode == TargetOpcode::G_BUILD_VECTOR_TRUNC; 1019 } 1020 1021 namespace { 1022 1023 Optional<ValueAndVReg> getAnyConstantSplat(Register VReg, 1024 const MachineRegisterInfo &MRI, 1025 bool AllowUndef) { 1026 MachineInstr *MI = getDefIgnoringCopies(VReg, MRI); 1027 if (!MI) 1028 return None; 1029 1030 if (!isBuildVectorOp(MI->getOpcode())) 1031 return None; 1032 1033 Optional<ValueAndVReg> SplatValAndReg = None; 1034 for (MachineOperand &Op : MI->uses()) { 1035 Register Element = Op.getReg(); 1036 auto ElementValAndReg = 1037 getAnyConstantVRegValWithLookThrough(Element, MRI, true, true); 1038 1039 // If AllowUndef, treat undef as value that will result in a constant splat. 1040 if (!ElementValAndReg) { 1041 if (AllowUndef && isa<GImplicitDef>(MRI.getVRegDef(Element))) 1042 continue; 1043 return None; 1044 } 1045 1046 // Record splat value 1047 if (!SplatValAndReg) 1048 SplatValAndReg = ElementValAndReg; 1049 1050 // Different constant then the one already recorded, not a constant splat. 1051 if (SplatValAndReg->Value != ElementValAndReg->Value) 1052 return None; 1053 } 1054 1055 return SplatValAndReg; 1056 } 1057 1058 } // end anonymous namespace 1059 1060 bool llvm::isBuildVectorConstantSplat(const Register Reg, 1061 const MachineRegisterInfo &MRI, 1062 int64_t SplatValue, bool AllowUndef) { 1063 if (auto SplatValAndReg = getAnyConstantSplat(Reg, MRI, AllowUndef)) 1064 return mi_match(SplatValAndReg->VReg, MRI, m_SpecificICst(SplatValue)); 1065 return false; 1066 } 1067 1068 bool llvm::isBuildVectorConstantSplat(const MachineInstr &MI, 1069 const MachineRegisterInfo &MRI, 1070 int64_t SplatValue, bool AllowUndef) { 1071 return isBuildVectorConstantSplat(MI.getOperand(0).getReg(), MRI, SplatValue, 1072 AllowUndef); 1073 } 1074 1075 Optional<int64_t> 1076 llvm::getBuildVectorConstantSplat(const MachineInstr &MI, 1077 const MachineRegisterInfo &MRI) { 1078 if (auto SplatValAndReg = 1079 getAnyConstantSplat(MI.getOperand(0).getReg(), MRI, false)) 1080 return getIConstantVRegSExtVal(SplatValAndReg->VReg, MRI); 1081 return None; 1082 } 1083 1084 Optional<FPValueAndVReg> llvm::getFConstantSplat(Register VReg, 1085 const MachineRegisterInfo &MRI, 1086 bool AllowUndef) { 1087 if (auto SplatValAndReg = getAnyConstantSplat(VReg, MRI, AllowUndef)) 1088 return getFConstantVRegValWithLookThrough(SplatValAndReg->VReg, MRI); 1089 return None; 1090 } 1091 1092 bool llvm::isBuildVectorAllZeros(const MachineInstr &MI, 1093 const MachineRegisterInfo &MRI, 1094 bool AllowUndef) { 1095 return isBuildVectorConstantSplat(MI, MRI, 0, AllowUndef); 1096 } 1097 1098 bool llvm::isBuildVectorAllOnes(const MachineInstr &MI, 1099 const MachineRegisterInfo &MRI, 1100 bool AllowUndef) { 1101 return isBuildVectorConstantSplat(MI, MRI, -1, AllowUndef); 1102 } 1103 1104 Optional<RegOrConstant> llvm::getVectorSplat(const MachineInstr &MI, 1105 const MachineRegisterInfo &MRI) { 1106 unsigned Opc = MI.getOpcode(); 1107 if (!isBuildVectorOp(Opc)) 1108 return None; 1109 if (auto Splat = getBuildVectorConstantSplat(MI, MRI)) 1110 return RegOrConstant(*Splat); 1111 auto Reg = MI.getOperand(1).getReg(); 1112 if (any_of(make_range(MI.operands_begin() + 2, MI.operands_end()), 1113 [&Reg](const MachineOperand &Op) { return Op.getReg() != Reg; })) 1114 return None; 1115 return RegOrConstant(Reg); 1116 } 1117 1118 static bool isConstantScalar(const MachineInstr &MI, 1119 const MachineRegisterInfo &MRI, 1120 bool AllowFP = true, 1121 bool AllowOpaqueConstants = true) { 1122 switch (MI.getOpcode()) { 1123 case TargetOpcode::G_CONSTANT: 1124 case TargetOpcode::G_IMPLICIT_DEF: 1125 return true; 1126 case TargetOpcode::G_FCONSTANT: 1127 return AllowFP; 1128 case TargetOpcode::G_GLOBAL_VALUE: 1129 case TargetOpcode::G_FRAME_INDEX: 1130 case TargetOpcode::G_BLOCK_ADDR: 1131 case TargetOpcode::G_JUMP_TABLE: 1132 return AllowOpaqueConstants; 1133 default: 1134 return false; 1135 } 1136 } 1137 1138 bool llvm::isConstantOrConstantVector(MachineInstr &MI, 1139 const MachineRegisterInfo &MRI) { 1140 Register Def = MI.getOperand(0).getReg(); 1141 if (auto C = getIConstantVRegValWithLookThrough(Def, MRI)) 1142 return true; 1143 GBuildVector *BV = dyn_cast<GBuildVector>(&MI); 1144 if (!BV) 1145 return false; 1146 for (unsigned SrcIdx = 0; SrcIdx < BV->getNumSources(); ++SrcIdx) { 1147 if (getIConstantVRegValWithLookThrough(BV->getSourceReg(SrcIdx), MRI) || 1148 getOpcodeDef<GImplicitDef>(BV->getSourceReg(SrcIdx), MRI)) 1149 continue; 1150 return false; 1151 } 1152 return true; 1153 } 1154 1155 bool llvm::isConstantOrConstantVector(const MachineInstr &MI, 1156 const MachineRegisterInfo &MRI, 1157 bool AllowFP, bool AllowOpaqueConstants) { 1158 if (isConstantScalar(MI, MRI, AllowFP, AllowOpaqueConstants)) 1159 return true; 1160 1161 if (!isBuildVectorOp(MI.getOpcode())) 1162 return false; 1163 1164 const unsigned NumOps = MI.getNumOperands(); 1165 for (unsigned I = 1; I != NumOps; ++I) { 1166 const MachineInstr *ElementDef = MRI.getVRegDef(MI.getOperand(I).getReg()); 1167 if (!isConstantScalar(*ElementDef, MRI, AllowFP, AllowOpaqueConstants)) 1168 return false; 1169 } 1170 1171 return true; 1172 } 1173 1174 Optional<APInt> 1175 llvm::isConstantOrConstantSplatVector(MachineInstr &MI, 1176 const MachineRegisterInfo &MRI) { 1177 Register Def = MI.getOperand(0).getReg(); 1178 if (auto C = getIConstantVRegValWithLookThrough(Def, MRI)) 1179 return C->Value; 1180 auto MaybeCst = getBuildVectorConstantSplat(MI, MRI); 1181 if (!MaybeCst) 1182 return None; 1183 const unsigned ScalarSize = MRI.getType(Def).getScalarSizeInBits(); 1184 return APInt(ScalarSize, *MaybeCst, true); 1185 } 1186 1187 bool llvm::isNullOrNullSplat(const MachineInstr &MI, 1188 const MachineRegisterInfo &MRI, bool AllowUndefs) { 1189 switch (MI.getOpcode()) { 1190 case TargetOpcode::G_IMPLICIT_DEF: 1191 return AllowUndefs; 1192 case TargetOpcode::G_CONSTANT: 1193 return MI.getOperand(1).getCImm()->isNullValue(); 1194 case TargetOpcode::G_FCONSTANT: { 1195 const ConstantFP *FPImm = MI.getOperand(1).getFPImm(); 1196 return FPImm->isZero() && !FPImm->isNegative(); 1197 } 1198 default: 1199 if (!AllowUndefs) // TODO: isBuildVectorAllZeros assumes undef is OK already 1200 return false; 1201 return isBuildVectorAllZeros(MI, MRI); 1202 } 1203 } 1204 1205 bool llvm::isAllOnesOrAllOnesSplat(const MachineInstr &MI, 1206 const MachineRegisterInfo &MRI, 1207 bool AllowUndefs) { 1208 switch (MI.getOpcode()) { 1209 case TargetOpcode::G_IMPLICIT_DEF: 1210 return AllowUndefs; 1211 case TargetOpcode::G_CONSTANT: 1212 return MI.getOperand(1).getCImm()->isAllOnesValue(); 1213 default: 1214 if (!AllowUndefs) // TODO: isBuildVectorAllOnes assumes undef is OK already 1215 return false; 1216 return isBuildVectorAllOnes(MI, MRI); 1217 } 1218 } 1219 1220 bool llvm::matchUnaryPredicate( 1221 const MachineRegisterInfo &MRI, Register Reg, 1222 std::function<bool(const Constant *ConstVal)> Match, bool AllowUndefs) { 1223 1224 const MachineInstr *Def = getDefIgnoringCopies(Reg, MRI); 1225 if (AllowUndefs && Def->getOpcode() == TargetOpcode::G_IMPLICIT_DEF) 1226 return Match(nullptr); 1227 1228 // TODO: Also handle fconstant 1229 if (Def->getOpcode() == TargetOpcode::G_CONSTANT) 1230 return Match(Def->getOperand(1).getCImm()); 1231 1232 if (Def->getOpcode() != TargetOpcode::G_BUILD_VECTOR) 1233 return false; 1234 1235 for (unsigned I = 1, E = Def->getNumOperands(); I != E; ++I) { 1236 Register SrcElt = Def->getOperand(I).getReg(); 1237 const MachineInstr *SrcDef = getDefIgnoringCopies(SrcElt, MRI); 1238 if (AllowUndefs && SrcDef->getOpcode() == TargetOpcode::G_IMPLICIT_DEF) { 1239 if (!Match(nullptr)) 1240 return false; 1241 continue; 1242 } 1243 1244 if (SrcDef->getOpcode() != TargetOpcode::G_CONSTANT || 1245 !Match(SrcDef->getOperand(1).getCImm())) 1246 return false; 1247 } 1248 1249 return true; 1250 } 1251 1252 bool llvm::isConstTrueVal(const TargetLowering &TLI, int64_t Val, bool IsVector, 1253 bool IsFP) { 1254 switch (TLI.getBooleanContents(IsVector, IsFP)) { 1255 case TargetLowering::UndefinedBooleanContent: 1256 return Val & 0x1; 1257 case TargetLowering::ZeroOrOneBooleanContent: 1258 return Val == 1; 1259 case TargetLowering::ZeroOrNegativeOneBooleanContent: 1260 return Val == -1; 1261 } 1262 llvm_unreachable("Invalid boolean contents"); 1263 } 1264 1265 int64_t llvm::getICmpTrueVal(const TargetLowering &TLI, bool IsVector, 1266 bool IsFP) { 1267 switch (TLI.getBooleanContents(IsVector, IsFP)) { 1268 case TargetLowering::UndefinedBooleanContent: 1269 case TargetLowering::ZeroOrOneBooleanContent: 1270 return 1; 1271 case TargetLowering::ZeroOrNegativeOneBooleanContent: 1272 return -1; 1273 } 1274 llvm_unreachable("Invalid boolean contents"); 1275 } 1276 1277 bool llvm::shouldOptForSize(const MachineBasicBlock &MBB, 1278 ProfileSummaryInfo *PSI, BlockFrequencyInfo *BFI) { 1279 const auto &F = MBB.getParent()->getFunction(); 1280 return F.hasOptSize() || F.hasMinSize() || 1281 llvm::shouldOptimizeForSize(MBB.getBasicBlock(), PSI, BFI); 1282 } 1283 1284 void llvm::saveUsesAndErase(MachineInstr &MI, MachineRegisterInfo &MRI, 1285 LostDebugLocObserver *LocObserver, 1286 SmallInstListTy &DeadInstChain) { 1287 for (MachineOperand &Op : MI.uses()) { 1288 if (Op.isReg() && Op.getReg().isVirtual()) 1289 DeadInstChain.insert(MRI.getVRegDef(Op.getReg())); 1290 } 1291 LLVM_DEBUG(dbgs() << MI << "Is dead; erasing.\n"); 1292 DeadInstChain.remove(&MI); 1293 MI.eraseFromParent(); 1294 if (LocObserver) 1295 LocObserver->checkpoint(false); 1296 } 1297 1298 void llvm::eraseInstrs(ArrayRef<MachineInstr *> DeadInstrs, 1299 MachineRegisterInfo &MRI, 1300 LostDebugLocObserver *LocObserver) { 1301 SmallInstListTy DeadInstChain; 1302 for (MachineInstr *MI : DeadInstrs) 1303 saveUsesAndErase(*MI, MRI, LocObserver, DeadInstChain); 1304 1305 while (!DeadInstChain.empty()) { 1306 MachineInstr *Inst = DeadInstChain.pop_back_val(); 1307 if (!isTriviallyDead(*Inst, MRI)) 1308 continue; 1309 saveUsesAndErase(*Inst, MRI, LocObserver, DeadInstChain); 1310 } 1311 } 1312 1313 void llvm::eraseInstr(MachineInstr &MI, MachineRegisterInfo &MRI, 1314 LostDebugLocObserver *LocObserver) { 1315 return eraseInstrs({&MI}, MRI, LocObserver); 1316 } 1317