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/CodeGenCommonISel.h" 17 #include "llvm/CodeGen/GlobalISel/GISelChangeObserver.h" 18 #include "llvm/CodeGen/GlobalISel/GISelKnownBits.h" 19 #include "llvm/CodeGen/GlobalISel/GenericMachineInstrs.h" 20 #include "llvm/CodeGen/GlobalISel/LostDebugLocObserver.h" 21 #include "llvm/CodeGen/GlobalISel/MIPatternMatch.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 #include <numeric> 37 #include <optional> 38 39 #define DEBUG_TYPE "globalisel-utils" 40 41 using namespace llvm; 42 using namespace MIPatternMatch; 43 44 Register llvm::constrainRegToClass(MachineRegisterInfo &MRI, 45 const TargetInstrInfo &TII, 46 const RegisterBankInfo &RBI, Register Reg, 47 const TargetRegisterClass &RegClass) { 48 if (!RBI.constrainGenericRegister(Reg, RegClass, MRI)) 49 return MRI.createVirtualRegister(&RegClass); 50 51 return Reg; 52 } 53 54 Register llvm::constrainOperandRegClass( 55 const MachineFunction &MF, const TargetRegisterInfo &TRI, 56 MachineRegisterInfo &MRI, const TargetInstrInfo &TII, 57 const RegisterBankInfo &RBI, MachineInstr &InsertPt, 58 const TargetRegisterClass &RegClass, MachineOperand &RegMO) { 59 Register Reg = RegMO.getReg(); 60 // Assume physical registers are properly constrained. 61 assert(Register::isVirtualRegister(Reg) && "PhysReg not implemented"); 62 63 // Save the old register class to check whether 64 // the change notifications will be required. 65 // TODO: A better approach would be to pass 66 // the observers to constrainRegToClass(). 67 auto *OldRegClass = MRI.getRegClassOrNull(Reg); 68 Register ConstrainedReg = constrainRegToClass(MRI, TII, RBI, Reg, RegClass); 69 // If we created a new virtual register because the class is not compatible 70 // then create a copy between the new and the old register. 71 if (ConstrainedReg != Reg) { 72 MachineBasicBlock::iterator InsertIt(&InsertPt); 73 MachineBasicBlock &MBB = *InsertPt.getParent(); 74 // FIXME: The copy needs to have the classes constrained for its operands. 75 // Use operand's regbank to get the class for old register (Reg). 76 if (RegMO.isUse()) { 77 BuildMI(MBB, InsertIt, InsertPt.getDebugLoc(), 78 TII.get(TargetOpcode::COPY), ConstrainedReg) 79 .addReg(Reg); 80 } else { 81 assert(RegMO.isDef() && "Must be a definition"); 82 BuildMI(MBB, std::next(InsertIt), InsertPt.getDebugLoc(), 83 TII.get(TargetOpcode::COPY), Reg) 84 .addReg(ConstrainedReg); 85 } 86 if (GISelChangeObserver *Observer = MF.getObserver()) { 87 Observer->changingInstr(*RegMO.getParent()); 88 } 89 RegMO.setReg(ConstrainedReg); 90 if (GISelChangeObserver *Observer = MF.getObserver()) { 91 Observer->changedInstr(*RegMO.getParent()); 92 } 93 } else if (OldRegClass != MRI.getRegClassOrNull(Reg)) { 94 if (GISelChangeObserver *Observer = MF.getObserver()) { 95 if (!RegMO.isDef()) { 96 MachineInstr *RegDef = MRI.getVRegDef(Reg); 97 Observer->changedInstr(*RegDef); 98 } 99 Observer->changingAllUsesOfReg(MRI, Reg); 100 Observer->finishedChangingAllUsesOfReg(); 101 } 102 } 103 return ConstrainedReg; 104 } 105 106 Register llvm::constrainOperandRegClass( 107 const MachineFunction &MF, const TargetRegisterInfo &TRI, 108 MachineRegisterInfo &MRI, const TargetInstrInfo &TII, 109 const RegisterBankInfo &RBI, MachineInstr &InsertPt, const MCInstrDesc &II, 110 MachineOperand &RegMO, unsigned OpIdx) { 111 Register Reg = RegMO.getReg(); 112 // Assume physical registers are properly constrained. 113 assert(Register::isVirtualRegister(Reg) && "PhysReg not implemented"); 114 115 const TargetRegisterClass *OpRC = TII.getRegClass(II, OpIdx, &TRI, MF); 116 // Some of the target independent instructions, like COPY, may not impose any 117 // register class constraints on some of their operands: If it's a use, we can 118 // skip constraining as the instruction defining the register would constrain 119 // it. 120 121 if (OpRC) { 122 // Obtain the RC from incoming regbank if it is a proper sub-class. Operands 123 // can have multiple regbanks for a superclass that combine different 124 // register types (E.g., AMDGPU's VGPR and AGPR). The regbank ambiguity 125 // resolved by targets during regbankselect should not be overridden. 126 if (const auto *SubRC = TRI.getCommonSubClass( 127 OpRC, TRI.getConstrainedRegClassForOperand(RegMO, MRI))) 128 OpRC = SubRC; 129 130 OpRC = TRI.getAllocatableClass(OpRC); 131 } 132 133 if (!OpRC) { 134 assert((!isTargetSpecificOpcode(II.getOpcode()) || RegMO.isUse()) && 135 "Register class constraint is required unless either the " 136 "instruction is target independent or the operand is a use"); 137 // FIXME: Just bailing out like this here could be not enough, unless we 138 // expect the users of this function to do the right thing for PHIs and 139 // COPY: 140 // v1 = COPY v0 141 // v2 = COPY v1 142 // v1 here may end up not being constrained at all. Please notice that to 143 // reproduce the issue we likely need a destination pattern of a selection 144 // rule producing such extra copies, not just an input GMIR with them as 145 // every existing target using selectImpl handles copies before calling it 146 // and they never reach this function. 147 return Reg; 148 } 149 return constrainOperandRegClass(MF, TRI, MRI, TII, RBI, InsertPt, *OpRC, 150 RegMO); 151 } 152 153 bool llvm::constrainSelectedInstRegOperands(MachineInstr &I, 154 const TargetInstrInfo &TII, 155 const TargetRegisterInfo &TRI, 156 const RegisterBankInfo &RBI) { 157 assert(!isPreISelGenericOpcode(I.getOpcode()) && 158 "A selected instruction is expected"); 159 MachineBasicBlock &MBB = *I.getParent(); 160 MachineFunction &MF = *MBB.getParent(); 161 MachineRegisterInfo &MRI = MF.getRegInfo(); 162 163 for (unsigned OpI = 0, OpE = I.getNumExplicitOperands(); OpI != OpE; ++OpI) { 164 MachineOperand &MO = I.getOperand(OpI); 165 166 // There's nothing to be done on non-register operands. 167 if (!MO.isReg()) 168 continue; 169 170 LLVM_DEBUG(dbgs() << "Converting operand: " << MO << '\n'); 171 assert(MO.isReg() && "Unsupported non-reg operand"); 172 173 Register Reg = MO.getReg(); 174 // Physical registers don't need to be constrained. 175 if (Register::isPhysicalRegister(Reg)) 176 continue; 177 178 // Register operands with a value of 0 (e.g. predicate operands) don't need 179 // to be constrained. 180 if (Reg == 0) 181 continue; 182 183 // If the operand is a vreg, we should constrain its regclass, and only 184 // insert COPYs if that's impossible. 185 // constrainOperandRegClass does that for us. 186 constrainOperandRegClass(MF, TRI, MRI, TII, RBI, I, I.getDesc(), MO, OpI); 187 188 // Tie uses to defs as indicated in MCInstrDesc if this hasn't already been 189 // done. 190 if (MO.isUse()) { 191 int DefIdx = I.getDesc().getOperandConstraint(OpI, MCOI::TIED_TO); 192 if (DefIdx != -1 && !I.isRegTiedToUseOperand(DefIdx)) 193 I.tieOperands(DefIdx, OpI); 194 } 195 } 196 return true; 197 } 198 199 bool llvm::canReplaceReg(Register DstReg, Register SrcReg, 200 MachineRegisterInfo &MRI) { 201 // Give up if either DstReg or SrcReg is a physical register. 202 if (DstReg.isPhysical() || SrcReg.isPhysical()) 203 return false; 204 // Give up if the types don't match. 205 if (MRI.getType(DstReg) != MRI.getType(SrcReg)) 206 return false; 207 // Replace if either DstReg has no constraints or the register 208 // constraints match. 209 return !MRI.getRegClassOrRegBank(DstReg) || 210 MRI.getRegClassOrRegBank(DstReg) == MRI.getRegClassOrRegBank(SrcReg); 211 } 212 213 bool llvm::isTriviallyDead(const MachineInstr &MI, 214 const MachineRegisterInfo &MRI) { 215 // FIXME: This logical is mostly duplicated with 216 // DeadMachineInstructionElim::isDead. Why is LOCAL_ESCAPE not considered in 217 // MachineInstr::isLabel? 218 219 // Don't delete frame allocation labels. 220 if (MI.getOpcode() == TargetOpcode::LOCAL_ESCAPE) 221 return false; 222 // LIFETIME markers should be preserved even if they seem dead. 223 if (MI.getOpcode() == TargetOpcode::LIFETIME_START || 224 MI.getOpcode() == TargetOpcode::LIFETIME_END) 225 return false; 226 227 // If we can move an instruction, we can remove it. Otherwise, it has 228 // a side-effect of some sort. 229 bool SawStore = false; 230 if (!MI.isSafeToMove(/*AA=*/nullptr, SawStore) && !MI.isPHI()) 231 return false; 232 233 // Instructions without side-effects are dead iff they only define dead vregs. 234 for (const auto &MO : MI.operands()) { 235 if (!MO.isReg() || !MO.isDef()) 236 continue; 237 238 Register Reg = MO.getReg(); 239 if (Register::isPhysicalRegister(Reg) || !MRI.use_nodbg_empty(Reg)) 240 return false; 241 } 242 return true; 243 } 244 245 static void reportGISelDiagnostic(DiagnosticSeverity Severity, 246 MachineFunction &MF, 247 const TargetPassConfig &TPC, 248 MachineOptimizationRemarkEmitter &MORE, 249 MachineOptimizationRemarkMissed &R) { 250 bool IsFatal = Severity == DS_Error && 251 TPC.isGlobalISelAbortEnabled(); 252 // Print the function name explicitly if we don't have a debug location (which 253 // makes the diagnostic less useful) or if we're going to emit a raw error. 254 if (!R.getLocation().isValid() || IsFatal) 255 R << (" (in function: " + MF.getName() + ")").str(); 256 257 if (IsFatal) 258 report_fatal_error(Twine(R.getMsg())); 259 else 260 MORE.emit(R); 261 } 262 263 void llvm::reportGISelWarning(MachineFunction &MF, const TargetPassConfig &TPC, 264 MachineOptimizationRemarkEmitter &MORE, 265 MachineOptimizationRemarkMissed &R) { 266 reportGISelDiagnostic(DS_Warning, MF, TPC, MORE, R); 267 } 268 269 void llvm::reportGISelFailure(MachineFunction &MF, const TargetPassConfig &TPC, 270 MachineOptimizationRemarkEmitter &MORE, 271 MachineOptimizationRemarkMissed &R) { 272 MF.getProperties().set(MachineFunctionProperties::Property::FailedISel); 273 reportGISelDiagnostic(DS_Error, MF, TPC, MORE, R); 274 } 275 276 void llvm::reportGISelFailure(MachineFunction &MF, const TargetPassConfig &TPC, 277 MachineOptimizationRemarkEmitter &MORE, 278 const char *PassName, StringRef Msg, 279 const MachineInstr &MI) { 280 MachineOptimizationRemarkMissed R(PassName, "GISelFailure: ", 281 MI.getDebugLoc(), MI.getParent()); 282 R << Msg; 283 // Printing MI is expensive; only do it if expensive remarks are enabled. 284 if (TPC.isGlobalISelAbortEnabled() || MORE.allowExtraAnalysis(PassName)) 285 R << ": " << ore::MNV("Inst", MI); 286 reportGISelFailure(MF, TPC, MORE, R); 287 } 288 289 std::optional<APInt> llvm::getIConstantVRegVal(Register VReg, 290 const MachineRegisterInfo &MRI) { 291 std::optional<ValueAndVReg> ValAndVReg = getIConstantVRegValWithLookThrough( 292 VReg, MRI, /*LookThroughInstrs*/ false); 293 assert((!ValAndVReg || ValAndVReg->VReg == VReg) && 294 "Value found while looking through instrs"); 295 if (!ValAndVReg) 296 return std::nullopt; 297 return ValAndVReg->Value; 298 } 299 300 std::optional<int64_t> 301 llvm::getIConstantVRegSExtVal(Register VReg, const MachineRegisterInfo &MRI) { 302 std::optional<APInt> Val = getIConstantVRegVal(VReg, MRI); 303 if (Val && Val->getBitWidth() <= 64) 304 return Val->getSExtValue(); 305 return std::nullopt; 306 } 307 308 namespace { 309 310 typedef std::function<bool(const MachineInstr *)> IsOpcodeFn; 311 typedef std::function<std::optional<APInt>(const MachineInstr *MI)> GetAPCstFn; 312 313 std::optional<ValueAndVReg> getConstantVRegValWithLookThrough( 314 Register VReg, const MachineRegisterInfo &MRI, IsOpcodeFn IsConstantOpcode, 315 GetAPCstFn getAPCstValue, bool LookThroughInstrs = true, 316 bool LookThroughAnyExt = false) { 317 SmallVector<std::pair<unsigned, unsigned>, 4> SeenOpcodes; 318 MachineInstr *MI; 319 320 while ((MI = MRI.getVRegDef(VReg)) && !IsConstantOpcode(MI) && 321 LookThroughInstrs) { 322 switch (MI->getOpcode()) { 323 case TargetOpcode::G_ANYEXT: 324 if (!LookThroughAnyExt) 325 return std::nullopt; 326 [[fallthrough]]; 327 case TargetOpcode::G_TRUNC: 328 case TargetOpcode::G_SEXT: 329 case TargetOpcode::G_ZEXT: 330 SeenOpcodes.push_back(std::make_pair( 331 MI->getOpcode(), 332 MRI.getType(MI->getOperand(0).getReg()).getSizeInBits())); 333 VReg = MI->getOperand(1).getReg(); 334 break; 335 case TargetOpcode::COPY: 336 VReg = MI->getOperand(1).getReg(); 337 if (Register::isPhysicalRegister(VReg)) 338 return std::nullopt; 339 break; 340 case TargetOpcode::G_INTTOPTR: 341 VReg = MI->getOperand(1).getReg(); 342 break; 343 default: 344 return std::nullopt; 345 } 346 } 347 if (!MI || !IsConstantOpcode(MI)) 348 return std::nullopt; 349 350 std::optional<APInt> MaybeVal = getAPCstValue(MI); 351 if (!MaybeVal) 352 return std::nullopt; 353 APInt &Val = *MaybeVal; 354 while (!SeenOpcodes.empty()) { 355 std::pair<unsigned, unsigned> OpcodeAndSize = SeenOpcodes.pop_back_val(); 356 switch (OpcodeAndSize.first) { 357 case TargetOpcode::G_TRUNC: 358 Val = Val.trunc(OpcodeAndSize.second); 359 break; 360 case TargetOpcode::G_ANYEXT: 361 case TargetOpcode::G_SEXT: 362 Val = Val.sext(OpcodeAndSize.second); 363 break; 364 case TargetOpcode::G_ZEXT: 365 Val = Val.zext(OpcodeAndSize.second); 366 break; 367 } 368 } 369 370 return ValueAndVReg{Val, VReg}; 371 } 372 373 bool isIConstant(const MachineInstr *MI) { 374 if (!MI) 375 return false; 376 return MI->getOpcode() == TargetOpcode::G_CONSTANT; 377 } 378 379 bool isFConstant(const MachineInstr *MI) { 380 if (!MI) 381 return false; 382 return MI->getOpcode() == TargetOpcode::G_FCONSTANT; 383 } 384 385 bool isAnyConstant(const MachineInstr *MI) { 386 if (!MI) 387 return false; 388 unsigned Opc = MI->getOpcode(); 389 return Opc == TargetOpcode::G_CONSTANT || Opc == TargetOpcode::G_FCONSTANT; 390 } 391 392 std::optional<APInt> getCImmAsAPInt(const MachineInstr *MI) { 393 const MachineOperand &CstVal = MI->getOperand(1); 394 if (CstVal.isCImm()) 395 return CstVal.getCImm()->getValue(); 396 return std::nullopt; 397 } 398 399 std::optional<APInt> getCImmOrFPImmAsAPInt(const MachineInstr *MI) { 400 const MachineOperand &CstVal = MI->getOperand(1); 401 if (CstVal.isCImm()) 402 return CstVal.getCImm()->getValue(); 403 if (CstVal.isFPImm()) 404 return CstVal.getFPImm()->getValueAPF().bitcastToAPInt(); 405 return std::nullopt; 406 } 407 408 } // end anonymous namespace 409 410 std::optional<ValueAndVReg> llvm::getIConstantVRegValWithLookThrough( 411 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs) { 412 return getConstantVRegValWithLookThrough(VReg, MRI, isIConstant, 413 getCImmAsAPInt, LookThroughInstrs); 414 } 415 416 std::optional<ValueAndVReg> llvm::getAnyConstantVRegValWithLookThrough( 417 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs, 418 bool LookThroughAnyExt) { 419 return getConstantVRegValWithLookThrough( 420 VReg, MRI, isAnyConstant, getCImmOrFPImmAsAPInt, LookThroughInstrs, 421 LookThroughAnyExt); 422 } 423 424 std::optional<FPValueAndVReg> llvm::getFConstantVRegValWithLookThrough( 425 Register VReg, const MachineRegisterInfo &MRI, bool LookThroughInstrs) { 426 auto Reg = getConstantVRegValWithLookThrough( 427 VReg, MRI, isFConstant, getCImmOrFPImmAsAPInt, LookThroughInstrs); 428 if (!Reg) 429 return std::nullopt; 430 return FPValueAndVReg{getConstantFPVRegVal(Reg->VReg, MRI)->getValueAPF(), 431 Reg->VReg}; 432 } 433 434 const ConstantFP * 435 llvm::getConstantFPVRegVal(Register VReg, const MachineRegisterInfo &MRI) { 436 MachineInstr *MI = MRI.getVRegDef(VReg); 437 if (TargetOpcode::G_FCONSTANT != MI->getOpcode()) 438 return nullptr; 439 return MI->getOperand(1).getFPImm(); 440 } 441 442 std::optional<DefinitionAndSourceRegister> 443 llvm::getDefSrcRegIgnoringCopies(Register Reg, const MachineRegisterInfo &MRI) { 444 Register DefSrcReg = Reg; 445 auto *DefMI = MRI.getVRegDef(Reg); 446 auto DstTy = MRI.getType(DefMI->getOperand(0).getReg()); 447 if (!DstTy.isValid()) 448 return std::nullopt; 449 unsigned Opc = DefMI->getOpcode(); 450 while (Opc == TargetOpcode::COPY || isPreISelGenericOptimizationHint(Opc)) { 451 Register SrcReg = DefMI->getOperand(1).getReg(); 452 auto SrcTy = MRI.getType(SrcReg); 453 if (!SrcTy.isValid()) 454 break; 455 DefMI = MRI.getVRegDef(SrcReg); 456 DefSrcReg = SrcReg; 457 Opc = DefMI->getOpcode(); 458 } 459 return DefinitionAndSourceRegister{DefMI, DefSrcReg}; 460 } 461 462 MachineInstr *llvm::getDefIgnoringCopies(Register Reg, 463 const MachineRegisterInfo &MRI) { 464 std::optional<DefinitionAndSourceRegister> DefSrcReg = 465 getDefSrcRegIgnoringCopies(Reg, MRI); 466 return DefSrcReg ? DefSrcReg->MI : nullptr; 467 } 468 469 Register llvm::getSrcRegIgnoringCopies(Register Reg, 470 const MachineRegisterInfo &MRI) { 471 std::optional<DefinitionAndSourceRegister> DefSrcReg = 472 getDefSrcRegIgnoringCopies(Reg, MRI); 473 return DefSrcReg ? DefSrcReg->Reg : Register(); 474 } 475 476 MachineInstr *llvm::getOpcodeDef(unsigned Opcode, Register Reg, 477 const MachineRegisterInfo &MRI) { 478 MachineInstr *DefMI = getDefIgnoringCopies(Reg, MRI); 479 return DefMI && DefMI->getOpcode() == Opcode ? DefMI : nullptr; 480 } 481 482 APFloat llvm::getAPFloatFromSize(double Val, unsigned Size) { 483 if (Size == 32) 484 return APFloat(float(Val)); 485 if (Size == 64) 486 return APFloat(Val); 487 if (Size != 16) 488 llvm_unreachable("Unsupported FPConstant size"); 489 bool Ignored; 490 APFloat APF(Val); 491 APF.convert(APFloat::IEEEhalf(), APFloat::rmNearestTiesToEven, &Ignored); 492 return APF; 493 } 494 495 std::optional<APInt> llvm::ConstantFoldBinOp(unsigned Opcode, 496 const Register Op1, 497 const Register Op2, 498 const MachineRegisterInfo &MRI) { 499 auto MaybeOp2Cst = getAnyConstantVRegValWithLookThrough(Op2, MRI, false); 500 if (!MaybeOp2Cst) 501 return std::nullopt; 502 503 auto MaybeOp1Cst = getAnyConstantVRegValWithLookThrough(Op1, MRI, false); 504 if (!MaybeOp1Cst) 505 return std::nullopt; 506 507 const APInt &C1 = MaybeOp1Cst->Value; 508 const APInt &C2 = MaybeOp2Cst->Value; 509 switch (Opcode) { 510 default: 511 break; 512 case TargetOpcode::G_ADD: 513 case TargetOpcode::G_PTR_ADD: 514 return C1 + C2; 515 case TargetOpcode::G_AND: 516 return C1 & C2; 517 case TargetOpcode::G_ASHR: 518 return C1.ashr(C2); 519 case TargetOpcode::G_LSHR: 520 return C1.lshr(C2); 521 case TargetOpcode::G_MUL: 522 return C1 * C2; 523 case TargetOpcode::G_OR: 524 return C1 | C2; 525 case TargetOpcode::G_SHL: 526 return C1 << C2; 527 case TargetOpcode::G_SUB: 528 return C1 - C2; 529 case TargetOpcode::G_XOR: 530 return C1 ^ C2; 531 case TargetOpcode::G_UDIV: 532 if (!C2.getBoolValue()) 533 break; 534 return C1.udiv(C2); 535 case TargetOpcode::G_SDIV: 536 if (!C2.getBoolValue()) 537 break; 538 return C1.sdiv(C2); 539 case TargetOpcode::G_UREM: 540 if (!C2.getBoolValue()) 541 break; 542 return C1.urem(C2); 543 case TargetOpcode::G_SREM: 544 if (!C2.getBoolValue()) 545 break; 546 return C1.srem(C2); 547 case TargetOpcode::G_SMIN: 548 return APIntOps::smin(C1, C2); 549 case TargetOpcode::G_SMAX: 550 return APIntOps::smax(C1, C2); 551 case TargetOpcode::G_UMIN: 552 return APIntOps::umin(C1, C2); 553 case TargetOpcode::G_UMAX: 554 return APIntOps::umax(C1, C2); 555 } 556 557 return std::nullopt; 558 } 559 560 std::optional<APFloat> 561 llvm::ConstantFoldFPBinOp(unsigned Opcode, const Register Op1, 562 const Register Op2, const MachineRegisterInfo &MRI) { 563 const ConstantFP *Op2Cst = getConstantFPVRegVal(Op2, MRI); 564 if (!Op2Cst) 565 return std::nullopt; 566 567 const ConstantFP *Op1Cst = getConstantFPVRegVal(Op1, MRI); 568 if (!Op1Cst) 569 return std::nullopt; 570 571 APFloat C1 = Op1Cst->getValueAPF(); 572 const APFloat &C2 = Op2Cst->getValueAPF(); 573 switch (Opcode) { 574 case TargetOpcode::G_FADD: 575 C1.add(C2, APFloat::rmNearestTiesToEven); 576 return C1; 577 case TargetOpcode::G_FSUB: 578 C1.subtract(C2, APFloat::rmNearestTiesToEven); 579 return C1; 580 case TargetOpcode::G_FMUL: 581 C1.multiply(C2, APFloat::rmNearestTiesToEven); 582 return C1; 583 case TargetOpcode::G_FDIV: 584 C1.divide(C2, APFloat::rmNearestTiesToEven); 585 return C1; 586 case TargetOpcode::G_FREM: 587 C1.mod(C2); 588 return C1; 589 case TargetOpcode::G_FCOPYSIGN: 590 C1.copySign(C2); 591 return C1; 592 case TargetOpcode::G_FMINNUM: 593 return minnum(C1, C2); 594 case TargetOpcode::G_FMAXNUM: 595 return maxnum(C1, C2); 596 case TargetOpcode::G_FMINIMUM: 597 return minimum(C1, C2); 598 case TargetOpcode::G_FMAXIMUM: 599 return maximum(C1, C2); 600 case TargetOpcode::G_FMINNUM_IEEE: 601 case TargetOpcode::G_FMAXNUM_IEEE: 602 // FIXME: These operations were unfortunately named. fminnum/fmaxnum do not 603 // follow the IEEE behavior for signaling nans and follow libm's fmin/fmax, 604 // and currently there isn't a nice wrapper in APFloat for the version with 605 // correct snan handling. 606 break; 607 default: 608 break; 609 } 610 611 return std::nullopt; 612 } 613 614 SmallVector<APInt> 615 llvm::ConstantFoldVectorBinop(unsigned Opcode, const Register Op1, 616 const Register Op2, 617 const MachineRegisterInfo &MRI) { 618 auto *SrcVec2 = getOpcodeDef<GBuildVector>(Op2, MRI); 619 if (!SrcVec2) 620 return SmallVector<APInt>(); 621 622 auto *SrcVec1 = getOpcodeDef<GBuildVector>(Op1, MRI); 623 if (!SrcVec1) 624 return SmallVector<APInt>(); 625 626 SmallVector<APInt> FoldedElements; 627 for (unsigned Idx = 0, E = SrcVec1->getNumSources(); Idx < E; ++Idx) { 628 auto MaybeCst = ConstantFoldBinOp(Opcode, SrcVec1->getSourceReg(Idx), 629 SrcVec2->getSourceReg(Idx), MRI); 630 if (!MaybeCst) 631 return SmallVector<APInt>(); 632 FoldedElements.push_back(*MaybeCst); 633 } 634 return FoldedElements; 635 } 636 637 bool llvm::isKnownNeverNaN(Register Val, const MachineRegisterInfo &MRI, 638 bool SNaN) { 639 const MachineInstr *DefMI = MRI.getVRegDef(Val); 640 if (!DefMI) 641 return false; 642 643 const TargetMachine& TM = DefMI->getMF()->getTarget(); 644 if (DefMI->getFlag(MachineInstr::FmNoNans) || TM.Options.NoNaNsFPMath) 645 return true; 646 647 // If the value is a constant, we can obviously see if it is a NaN or not. 648 if (const ConstantFP *FPVal = getConstantFPVRegVal(Val, MRI)) { 649 return !FPVal->getValueAPF().isNaN() || 650 (SNaN && !FPVal->getValueAPF().isSignaling()); 651 } 652 653 if (DefMI->getOpcode() == TargetOpcode::G_BUILD_VECTOR) { 654 for (const auto &Op : DefMI->uses()) 655 if (!isKnownNeverNaN(Op.getReg(), MRI, SNaN)) 656 return false; 657 return true; 658 } 659 660 switch (DefMI->getOpcode()) { 661 default: 662 break; 663 case TargetOpcode::G_FADD: 664 case TargetOpcode::G_FSUB: 665 case TargetOpcode::G_FMUL: 666 case TargetOpcode::G_FDIV: 667 case TargetOpcode::G_FREM: 668 case TargetOpcode::G_FSIN: 669 case TargetOpcode::G_FCOS: 670 case TargetOpcode::G_FMA: 671 case TargetOpcode::G_FMAD: 672 if (SNaN) 673 return true; 674 675 // TODO: Need isKnownNeverInfinity 676 return false; 677 case TargetOpcode::G_FMINNUM_IEEE: 678 case TargetOpcode::G_FMAXNUM_IEEE: { 679 if (SNaN) 680 return true; 681 // This can return a NaN if either operand is an sNaN, or if both operands 682 // are NaN. 683 return (isKnownNeverNaN(DefMI->getOperand(1).getReg(), MRI) && 684 isKnownNeverSNaN(DefMI->getOperand(2).getReg(), MRI)) || 685 (isKnownNeverSNaN(DefMI->getOperand(1).getReg(), MRI) && 686 isKnownNeverNaN(DefMI->getOperand(2).getReg(), MRI)); 687 } 688 case TargetOpcode::G_FMINNUM: 689 case TargetOpcode::G_FMAXNUM: { 690 // Only one needs to be known not-nan, since it will be returned if the 691 // other ends up being one. 692 return isKnownNeverNaN(DefMI->getOperand(1).getReg(), MRI, SNaN) || 693 isKnownNeverNaN(DefMI->getOperand(2).getReg(), MRI, SNaN); 694 } 695 } 696 697 if (SNaN) { 698 // FP operations quiet. For now, just handle the ones inserted during 699 // legalization. 700 switch (DefMI->getOpcode()) { 701 case TargetOpcode::G_FPEXT: 702 case TargetOpcode::G_FPTRUNC: 703 case TargetOpcode::G_FCANONICALIZE: 704 return true; 705 default: 706 return false; 707 } 708 } 709 710 return false; 711 } 712 713 Align llvm::inferAlignFromPtrInfo(MachineFunction &MF, 714 const MachinePointerInfo &MPO) { 715 auto PSV = MPO.V.dyn_cast<const PseudoSourceValue *>(); 716 if (auto FSPV = dyn_cast_or_null<FixedStackPseudoSourceValue>(PSV)) { 717 MachineFrameInfo &MFI = MF.getFrameInfo(); 718 return commonAlignment(MFI.getObjectAlign(FSPV->getFrameIndex()), 719 MPO.Offset); 720 } 721 722 if (const Value *V = MPO.V.dyn_cast<const Value *>()) { 723 const Module *M = MF.getFunction().getParent(); 724 return V->getPointerAlignment(M->getDataLayout()); 725 } 726 727 return Align(1); 728 } 729 730 Register llvm::getFunctionLiveInPhysReg(MachineFunction &MF, 731 const TargetInstrInfo &TII, 732 MCRegister PhysReg, 733 const TargetRegisterClass &RC, 734 const DebugLoc &DL, LLT RegTy) { 735 MachineBasicBlock &EntryMBB = MF.front(); 736 MachineRegisterInfo &MRI = MF.getRegInfo(); 737 Register LiveIn = MRI.getLiveInVirtReg(PhysReg); 738 if (LiveIn) { 739 MachineInstr *Def = MRI.getVRegDef(LiveIn); 740 if (Def) { 741 // FIXME: Should the verifier check this is in the entry block? 742 assert(Def->getParent() == &EntryMBB && "live-in copy not in entry block"); 743 return LiveIn; 744 } 745 746 // It's possible the incoming argument register and copy was added during 747 // lowering, but later deleted due to being/becoming dead. If this happens, 748 // re-insert the copy. 749 } else { 750 // The live in register was not present, so add it. 751 LiveIn = MF.addLiveIn(PhysReg, &RC); 752 if (RegTy.isValid()) 753 MRI.setType(LiveIn, RegTy); 754 } 755 756 BuildMI(EntryMBB, EntryMBB.begin(), DL, TII.get(TargetOpcode::COPY), LiveIn) 757 .addReg(PhysReg); 758 if (!EntryMBB.isLiveIn(PhysReg)) 759 EntryMBB.addLiveIn(PhysReg); 760 return LiveIn; 761 } 762 763 std::optional<APInt> llvm::ConstantFoldExtOp(unsigned Opcode, 764 const Register Op1, uint64_t Imm, 765 const MachineRegisterInfo &MRI) { 766 auto MaybeOp1Cst = getIConstantVRegVal(Op1, MRI); 767 if (MaybeOp1Cst) { 768 switch (Opcode) { 769 default: 770 break; 771 case TargetOpcode::G_SEXT_INREG: { 772 LLT Ty = MRI.getType(Op1); 773 return MaybeOp1Cst->trunc(Imm).sext(Ty.getScalarSizeInBits()); 774 } 775 } 776 } 777 return std::nullopt; 778 } 779 780 std::optional<APFloat> 781 llvm::ConstantFoldIntToFloat(unsigned Opcode, LLT DstTy, Register Src, 782 const MachineRegisterInfo &MRI) { 783 assert(Opcode == TargetOpcode::G_SITOFP || Opcode == TargetOpcode::G_UITOFP); 784 if (auto MaybeSrcVal = getIConstantVRegVal(Src, MRI)) { 785 APFloat DstVal(getFltSemanticForLLT(DstTy)); 786 DstVal.convertFromAPInt(*MaybeSrcVal, Opcode == TargetOpcode::G_SITOFP, 787 APFloat::rmNearestTiesToEven); 788 return DstVal; 789 } 790 return std::nullopt; 791 } 792 793 std::optional<SmallVector<unsigned>> 794 llvm::ConstantFoldCTLZ(Register Src, const MachineRegisterInfo &MRI) { 795 LLT Ty = MRI.getType(Src); 796 SmallVector<unsigned> FoldedCTLZs; 797 auto tryFoldScalar = [&](Register R) -> std::optional<unsigned> { 798 auto MaybeCst = getIConstantVRegVal(R, MRI); 799 if (!MaybeCst) 800 return std::nullopt; 801 return MaybeCst->countLeadingZeros(); 802 }; 803 if (Ty.isVector()) { 804 // Try to constant fold each element. 805 auto *BV = getOpcodeDef<GBuildVector>(Src, MRI); 806 if (!BV) 807 return std::nullopt; 808 for (unsigned SrcIdx = 0; SrcIdx < BV->getNumSources(); ++SrcIdx) { 809 if (auto MaybeFold = tryFoldScalar(BV->getSourceReg(SrcIdx))) { 810 FoldedCTLZs.emplace_back(*MaybeFold); 811 continue; 812 } 813 return std::nullopt; 814 } 815 return FoldedCTLZs; 816 } 817 if (auto MaybeCst = tryFoldScalar(Src)) { 818 FoldedCTLZs.emplace_back(*MaybeCst); 819 return FoldedCTLZs; 820 } 821 return std::nullopt; 822 } 823 824 bool llvm::isKnownToBeAPowerOfTwo(Register Reg, const MachineRegisterInfo &MRI, 825 GISelKnownBits *KB) { 826 std::optional<DefinitionAndSourceRegister> DefSrcReg = 827 getDefSrcRegIgnoringCopies(Reg, MRI); 828 if (!DefSrcReg) 829 return false; 830 831 const MachineInstr &MI = *DefSrcReg->MI; 832 const LLT Ty = MRI.getType(Reg); 833 834 switch (MI.getOpcode()) { 835 case TargetOpcode::G_CONSTANT: { 836 unsigned BitWidth = Ty.getScalarSizeInBits(); 837 const ConstantInt *CI = MI.getOperand(1).getCImm(); 838 return CI->getValue().zextOrTrunc(BitWidth).isPowerOf2(); 839 } 840 case TargetOpcode::G_SHL: { 841 // A left-shift of a constant one will have exactly one bit set because 842 // shifting the bit off the end is undefined. 843 844 // TODO: Constant splat 845 if (auto ConstLHS = getIConstantVRegVal(MI.getOperand(1).getReg(), MRI)) { 846 if (*ConstLHS == 1) 847 return true; 848 } 849 850 break; 851 } 852 case TargetOpcode::G_LSHR: { 853 if (auto ConstLHS = getIConstantVRegVal(MI.getOperand(1).getReg(), MRI)) { 854 if (ConstLHS->isSignMask()) 855 return true; 856 } 857 858 break; 859 } 860 case TargetOpcode::G_BUILD_VECTOR: { 861 // TODO: Probably should have a recursion depth guard since you could have 862 // bitcasted vector elements. 863 for (const MachineOperand &MO : llvm::drop_begin(MI.operands())) 864 if (!isKnownToBeAPowerOfTwo(MO.getReg(), MRI, KB)) 865 return false; 866 867 return true; 868 } 869 case TargetOpcode::G_BUILD_VECTOR_TRUNC: { 870 // Only handle constants since we would need to know if number of leading 871 // zeros is greater than the truncation amount. 872 const unsigned BitWidth = Ty.getScalarSizeInBits(); 873 for (const MachineOperand &MO : llvm::drop_begin(MI.operands())) { 874 auto Const = getIConstantVRegVal(MO.getReg(), MRI); 875 if (!Const || !Const->zextOrTrunc(BitWidth).isPowerOf2()) 876 return false; 877 } 878 879 return true; 880 } 881 default: 882 break; 883 } 884 885 if (!KB) 886 return false; 887 888 // More could be done here, though the above checks are enough 889 // to handle some common cases. 890 891 // Fall back to computeKnownBits to catch other known cases. 892 KnownBits Known = KB->getKnownBits(Reg); 893 return (Known.countMaxPopulation() == 1) && (Known.countMinPopulation() == 1); 894 } 895 896 void llvm::getSelectionDAGFallbackAnalysisUsage(AnalysisUsage &AU) { 897 AU.addPreserved<StackProtector>(); 898 } 899 900 LLT llvm::getLCMType(LLT OrigTy, LLT TargetTy) { 901 const unsigned OrigSize = OrigTy.getSizeInBits(); 902 const unsigned TargetSize = TargetTy.getSizeInBits(); 903 904 if (OrigSize == TargetSize) 905 return OrigTy; 906 907 if (OrigTy.isVector()) { 908 const LLT OrigElt = OrigTy.getElementType(); 909 910 if (TargetTy.isVector()) { 911 const LLT TargetElt = TargetTy.getElementType(); 912 913 if (OrigElt.getSizeInBits() == TargetElt.getSizeInBits()) { 914 int GCDElts = 915 std::gcd(OrigTy.getNumElements(), TargetTy.getNumElements()); 916 // Prefer the original element type. 917 ElementCount Mul = OrigTy.getElementCount() * TargetTy.getNumElements(); 918 return LLT::vector(Mul.divideCoefficientBy(GCDElts), 919 OrigTy.getElementType()); 920 } 921 } else { 922 if (OrigElt.getSizeInBits() == TargetSize) 923 return OrigTy; 924 } 925 926 unsigned LCMSize = std::lcm(OrigSize, TargetSize); 927 return LLT::fixed_vector(LCMSize / OrigElt.getSizeInBits(), OrigElt); 928 } 929 930 if (TargetTy.isVector()) { 931 unsigned LCMSize = std::lcm(OrigSize, TargetSize); 932 return LLT::fixed_vector(LCMSize / OrigSize, OrigTy); 933 } 934 935 unsigned LCMSize = std::lcm(OrigSize, TargetSize); 936 937 // Preserve pointer types. 938 if (LCMSize == OrigSize) 939 return OrigTy; 940 if (LCMSize == TargetSize) 941 return TargetTy; 942 943 return LLT::scalar(LCMSize); 944 } 945 946 LLT llvm::getCoverTy(LLT OrigTy, LLT TargetTy) { 947 if (!OrigTy.isVector() || !TargetTy.isVector() || OrigTy == TargetTy || 948 (OrigTy.getScalarSizeInBits() != TargetTy.getScalarSizeInBits())) 949 return getLCMType(OrigTy, TargetTy); 950 951 unsigned OrigTyNumElts = OrigTy.getNumElements(); 952 unsigned TargetTyNumElts = TargetTy.getNumElements(); 953 if (OrigTyNumElts % TargetTyNumElts == 0) 954 return OrigTy; 955 956 unsigned NumElts = alignTo(OrigTyNumElts, TargetTyNumElts); 957 return LLT::scalarOrVector(ElementCount::getFixed(NumElts), 958 OrigTy.getElementType()); 959 } 960 961 LLT llvm::getGCDType(LLT OrigTy, LLT TargetTy) { 962 const unsigned OrigSize = OrigTy.getSizeInBits(); 963 const unsigned TargetSize = TargetTy.getSizeInBits(); 964 965 if (OrigSize == TargetSize) 966 return OrigTy; 967 968 if (OrigTy.isVector()) { 969 LLT OrigElt = OrigTy.getElementType(); 970 if (TargetTy.isVector()) { 971 LLT TargetElt = TargetTy.getElementType(); 972 if (OrigElt.getSizeInBits() == TargetElt.getSizeInBits()) { 973 int GCD = std::gcd(OrigTy.getNumElements(), TargetTy.getNumElements()); 974 return LLT::scalarOrVector(ElementCount::getFixed(GCD), OrigElt); 975 } 976 } else { 977 // If the source is a vector of pointers, return a pointer element. 978 if (OrigElt.getSizeInBits() == TargetSize) 979 return OrigElt; 980 } 981 982 unsigned GCD = std::gcd(OrigSize, TargetSize); 983 if (GCD == OrigElt.getSizeInBits()) 984 return OrigElt; 985 986 // If we can't produce the original element type, we have to use a smaller 987 // scalar. 988 if (GCD < OrigElt.getSizeInBits()) 989 return LLT::scalar(GCD); 990 return LLT::fixed_vector(GCD / OrigElt.getSizeInBits(), OrigElt); 991 } 992 993 if (TargetTy.isVector()) { 994 // Try to preserve the original element type. 995 LLT TargetElt = TargetTy.getElementType(); 996 if (TargetElt.getSizeInBits() == OrigSize) 997 return OrigTy; 998 } 999 1000 unsigned GCD = std::gcd(OrigSize, TargetSize); 1001 return LLT::scalar(GCD); 1002 } 1003 1004 std::optional<int> llvm::getSplatIndex(MachineInstr &MI) { 1005 assert(MI.getOpcode() == TargetOpcode::G_SHUFFLE_VECTOR && 1006 "Only G_SHUFFLE_VECTOR can have a splat index!"); 1007 ArrayRef<int> Mask = MI.getOperand(3).getShuffleMask(); 1008 auto FirstDefinedIdx = find_if(Mask, [](int Elt) { return Elt >= 0; }); 1009 1010 // If all elements are undefined, this shuffle can be considered a splat. 1011 // Return 0 for better potential for callers to simplify. 1012 if (FirstDefinedIdx == Mask.end()) 1013 return 0; 1014 1015 // Make sure all remaining elements are either undef or the same 1016 // as the first non-undef value. 1017 int SplatValue = *FirstDefinedIdx; 1018 if (any_of(make_range(std::next(FirstDefinedIdx), Mask.end()), 1019 [&SplatValue](int Elt) { return Elt >= 0 && Elt != SplatValue; })) 1020 return std::nullopt; 1021 1022 return SplatValue; 1023 } 1024 1025 static bool isBuildVectorOp(unsigned Opcode) { 1026 return Opcode == TargetOpcode::G_BUILD_VECTOR || 1027 Opcode == TargetOpcode::G_BUILD_VECTOR_TRUNC; 1028 } 1029 1030 namespace { 1031 1032 std::optional<ValueAndVReg> getAnyConstantSplat(Register VReg, 1033 const MachineRegisterInfo &MRI, 1034 bool AllowUndef) { 1035 MachineInstr *MI = getDefIgnoringCopies(VReg, MRI); 1036 if (!MI) 1037 return std::nullopt; 1038 1039 if (!isBuildVectorOp(MI->getOpcode())) 1040 return std::nullopt; 1041 1042 std::optional<ValueAndVReg> SplatValAndReg; 1043 for (MachineOperand &Op : MI->uses()) { 1044 Register Element = Op.getReg(); 1045 auto ElementValAndReg = 1046 getAnyConstantVRegValWithLookThrough(Element, MRI, true, true); 1047 1048 // If AllowUndef, treat undef as value that will result in a constant splat. 1049 if (!ElementValAndReg) { 1050 if (AllowUndef && isa<GImplicitDef>(MRI.getVRegDef(Element))) 1051 continue; 1052 return std::nullopt; 1053 } 1054 1055 // Record splat value 1056 if (!SplatValAndReg) 1057 SplatValAndReg = ElementValAndReg; 1058 1059 // Different constant then the one already recorded, not a constant splat. 1060 if (SplatValAndReg->Value != ElementValAndReg->Value) 1061 return std::nullopt; 1062 } 1063 1064 return SplatValAndReg; 1065 } 1066 1067 } // end anonymous namespace 1068 1069 bool llvm::isBuildVectorConstantSplat(const Register Reg, 1070 const MachineRegisterInfo &MRI, 1071 int64_t SplatValue, bool AllowUndef) { 1072 if (auto SplatValAndReg = getAnyConstantSplat(Reg, MRI, AllowUndef)) 1073 return mi_match(SplatValAndReg->VReg, MRI, m_SpecificICst(SplatValue)); 1074 return false; 1075 } 1076 1077 bool llvm::isBuildVectorConstantSplat(const MachineInstr &MI, 1078 const MachineRegisterInfo &MRI, 1079 int64_t SplatValue, bool AllowUndef) { 1080 return isBuildVectorConstantSplat(MI.getOperand(0).getReg(), MRI, SplatValue, 1081 AllowUndef); 1082 } 1083 1084 std::optional<APInt> 1085 llvm::getIConstantSplatVal(const Register Reg, const MachineRegisterInfo &MRI) { 1086 if (auto SplatValAndReg = 1087 getAnyConstantSplat(Reg, MRI, /* AllowUndef */ false)) { 1088 std::optional<ValueAndVReg> ValAndVReg = 1089 getIConstantVRegValWithLookThrough(SplatValAndReg->VReg, MRI); 1090 return ValAndVReg->Value; 1091 } 1092 1093 return std::nullopt; 1094 } 1095 1096 std::optional<APInt> 1097 llvm::getIConstantSplatVal(const MachineInstr &MI, 1098 const MachineRegisterInfo &MRI) { 1099 return getIConstantSplatVal(MI.getOperand(0).getReg(), MRI); 1100 } 1101 1102 std::optional<int64_t> 1103 llvm::getIConstantSplatSExtVal(const Register Reg, 1104 const MachineRegisterInfo &MRI) { 1105 if (auto SplatValAndReg = 1106 getAnyConstantSplat(Reg, MRI, /* AllowUndef */ false)) 1107 return getIConstantVRegSExtVal(SplatValAndReg->VReg, MRI); 1108 return std::nullopt; 1109 } 1110 1111 std::optional<int64_t> 1112 llvm::getIConstantSplatSExtVal(const MachineInstr &MI, 1113 const MachineRegisterInfo &MRI) { 1114 return getIConstantSplatSExtVal(MI.getOperand(0).getReg(), MRI); 1115 } 1116 1117 std::optional<FPValueAndVReg> 1118 llvm::getFConstantSplat(Register VReg, const MachineRegisterInfo &MRI, 1119 bool AllowUndef) { 1120 if (auto SplatValAndReg = getAnyConstantSplat(VReg, MRI, AllowUndef)) 1121 return getFConstantVRegValWithLookThrough(SplatValAndReg->VReg, MRI); 1122 return std::nullopt; 1123 } 1124 1125 bool llvm::isBuildVectorAllZeros(const MachineInstr &MI, 1126 const MachineRegisterInfo &MRI, 1127 bool AllowUndef) { 1128 return isBuildVectorConstantSplat(MI, MRI, 0, AllowUndef); 1129 } 1130 1131 bool llvm::isBuildVectorAllOnes(const MachineInstr &MI, 1132 const MachineRegisterInfo &MRI, 1133 bool AllowUndef) { 1134 return isBuildVectorConstantSplat(MI, MRI, -1, AllowUndef); 1135 } 1136 1137 std::optional<RegOrConstant> 1138 llvm::getVectorSplat(const MachineInstr &MI, const MachineRegisterInfo &MRI) { 1139 unsigned Opc = MI.getOpcode(); 1140 if (!isBuildVectorOp(Opc)) 1141 return std::nullopt; 1142 if (auto Splat = getIConstantSplatSExtVal(MI, MRI)) 1143 return RegOrConstant(*Splat); 1144 auto Reg = MI.getOperand(1).getReg(); 1145 if (any_of(make_range(MI.operands_begin() + 2, MI.operands_end()), 1146 [&Reg](const MachineOperand &Op) { return Op.getReg() != Reg; })) 1147 return std::nullopt; 1148 return RegOrConstant(Reg); 1149 } 1150 1151 static bool isConstantScalar(const MachineInstr &MI, 1152 const MachineRegisterInfo &MRI, 1153 bool AllowFP = true, 1154 bool AllowOpaqueConstants = true) { 1155 switch (MI.getOpcode()) { 1156 case TargetOpcode::G_CONSTANT: 1157 case TargetOpcode::G_IMPLICIT_DEF: 1158 return true; 1159 case TargetOpcode::G_FCONSTANT: 1160 return AllowFP; 1161 case TargetOpcode::G_GLOBAL_VALUE: 1162 case TargetOpcode::G_FRAME_INDEX: 1163 case TargetOpcode::G_BLOCK_ADDR: 1164 case TargetOpcode::G_JUMP_TABLE: 1165 return AllowOpaqueConstants; 1166 default: 1167 return false; 1168 } 1169 } 1170 1171 bool llvm::isConstantOrConstantVector(MachineInstr &MI, 1172 const MachineRegisterInfo &MRI) { 1173 Register Def = MI.getOperand(0).getReg(); 1174 if (auto C = getIConstantVRegValWithLookThrough(Def, MRI)) 1175 return true; 1176 GBuildVector *BV = dyn_cast<GBuildVector>(&MI); 1177 if (!BV) 1178 return false; 1179 for (unsigned SrcIdx = 0; SrcIdx < BV->getNumSources(); ++SrcIdx) { 1180 if (getIConstantVRegValWithLookThrough(BV->getSourceReg(SrcIdx), MRI) || 1181 getOpcodeDef<GImplicitDef>(BV->getSourceReg(SrcIdx), MRI)) 1182 continue; 1183 return false; 1184 } 1185 return true; 1186 } 1187 1188 bool llvm::isConstantOrConstantVector(const MachineInstr &MI, 1189 const MachineRegisterInfo &MRI, 1190 bool AllowFP, bool AllowOpaqueConstants) { 1191 if (isConstantScalar(MI, MRI, AllowFP, AllowOpaqueConstants)) 1192 return true; 1193 1194 if (!isBuildVectorOp(MI.getOpcode())) 1195 return false; 1196 1197 const unsigned NumOps = MI.getNumOperands(); 1198 for (unsigned I = 1; I != NumOps; ++I) { 1199 const MachineInstr *ElementDef = MRI.getVRegDef(MI.getOperand(I).getReg()); 1200 if (!isConstantScalar(*ElementDef, MRI, AllowFP, AllowOpaqueConstants)) 1201 return false; 1202 } 1203 1204 return true; 1205 } 1206 1207 std::optional<APInt> 1208 llvm::isConstantOrConstantSplatVector(MachineInstr &MI, 1209 const MachineRegisterInfo &MRI) { 1210 Register Def = MI.getOperand(0).getReg(); 1211 if (auto C = getIConstantVRegValWithLookThrough(Def, MRI)) 1212 return C->Value; 1213 auto MaybeCst = getIConstantSplatSExtVal(MI, MRI); 1214 if (!MaybeCst) 1215 return std::nullopt; 1216 const unsigned ScalarSize = MRI.getType(Def).getScalarSizeInBits(); 1217 return APInt(ScalarSize, *MaybeCst, true); 1218 } 1219 1220 bool llvm::isNullOrNullSplat(const MachineInstr &MI, 1221 const MachineRegisterInfo &MRI, bool AllowUndefs) { 1222 switch (MI.getOpcode()) { 1223 case TargetOpcode::G_IMPLICIT_DEF: 1224 return AllowUndefs; 1225 case TargetOpcode::G_CONSTANT: 1226 return MI.getOperand(1).getCImm()->isNullValue(); 1227 case TargetOpcode::G_FCONSTANT: { 1228 const ConstantFP *FPImm = MI.getOperand(1).getFPImm(); 1229 return FPImm->isZero() && !FPImm->isNegative(); 1230 } 1231 default: 1232 if (!AllowUndefs) // TODO: isBuildVectorAllZeros assumes undef is OK already 1233 return false; 1234 return isBuildVectorAllZeros(MI, MRI); 1235 } 1236 } 1237 1238 bool llvm::isAllOnesOrAllOnesSplat(const MachineInstr &MI, 1239 const MachineRegisterInfo &MRI, 1240 bool AllowUndefs) { 1241 switch (MI.getOpcode()) { 1242 case TargetOpcode::G_IMPLICIT_DEF: 1243 return AllowUndefs; 1244 case TargetOpcode::G_CONSTANT: 1245 return MI.getOperand(1).getCImm()->isAllOnesValue(); 1246 default: 1247 if (!AllowUndefs) // TODO: isBuildVectorAllOnes assumes undef is OK already 1248 return false; 1249 return isBuildVectorAllOnes(MI, MRI); 1250 } 1251 } 1252 1253 bool llvm::matchUnaryPredicate( 1254 const MachineRegisterInfo &MRI, Register Reg, 1255 std::function<bool(const Constant *ConstVal)> Match, bool AllowUndefs) { 1256 1257 const MachineInstr *Def = getDefIgnoringCopies(Reg, MRI); 1258 if (AllowUndefs && Def->getOpcode() == TargetOpcode::G_IMPLICIT_DEF) 1259 return Match(nullptr); 1260 1261 // TODO: Also handle fconstant 1262 if (Def->getOpcode() == TargetOpcode::G_CONSTANT) 1263 return Match(Def->getOperand(1).getCImm()); 1264 1265 if (Def->getOpcode() != TargetOpcode::G_BUILD_VECTOR) 1266 return false; 1267 1268 for (unsigned I = 1, E = Def->getNumOperands(); I != E; ++I) { 1269 Register SrcElt = Def->getOperand(I).getReg(); 1270 const MachineInstr *SrcDef = getDefIgnoringCopies(SrcElt, MRI); 1271 if (AllowUndefs && SrcDef->getOpcode() == TargetOpcode::G_IMPLICIT_DEF) { 1272 if (!Match(nullptr)) 1273 return false; 1274 continue; 1275 } 1276 1277 if (SrcDef->getOpcode() != TargetOpcode::G_CONSTANT || 1278 !Match(SrcDef->getOperand(1).getCImm())) 1279 return false; 1280 } 1281 1282 return true; 1283 } 1284 1285 bool llvm::isConstTrueVal(const TargetLowering &TLI, int64_t Val, bool IsVector, 1286 bool IsFP) { 1287 switch (TLI.getBooleanContents(IsVector, IsFP)) { 1288 case TargetLowering::UndefinedBooleanContent: 1289 return Val & 0x1; 1290 case TargetLowering::ZeroOrOneBooleanContent: 1291 return Val == 1; 1292 case TargetLowering::ZeroOrNegativeOneBooleanContent: 1293 return Val == -1; 1294 } 1295 llvm_unreachable("Invalid boolean contents"); 1296 } 1297 1298 bool llvm::isConstFalseVal(const TargetLowering &TLI, int64_t Val, 1299 bool IsVector, bool IsFP) { 1300 switch (TLI.getBooleanContents(IsVector, IsFP)) { 1301 case TargetLowering::UndefinedBooleanContent: 1302 return ~Val & 0x1; 1303 case TargetLowering::ZeroOrOneBooleanContent: 1304 case TargetLowering::ZeroOrNegativeOneBooleanContent: 1305 return Val == 0; 1306 } 1307 llvm_unreachable("Invalid boolean contents"); 1308 } 1309 1310 int64_t llvm::getICmpTrueVal(const TargetLowering &TLI, bool IsVector, 1311 bool IsFP) { 1312 switch (TLI.getBooleanContents(IsVector, IsFP)) { 1313 case TargetLowering::UndefinedBooleanContent: 1314 case TargetLowering::ZeroOrOneBooleanContent: 1315 return 1; 1316 case TargetLowering::ZeroOrNegativeOneBooleanContent: 1317 return -1; 1318 } 1319 llvm_unreachable("Invalid boolean contents"); 1320 } 1321 1322 bool llvm::shouldOptForSize(const MachineBasicBlock &MBB, 1323 ProfileSummaryInfo *PSI, BlockFrequencyInfo *BFI) { 1324 const auto &F = MBB.getParent()->getFunction(); 1325 return F.hasOptSize() || F.hasMinSize() || 1326 llvm::shouldOptimizeForSize(MBB.getBasicBlock(), PSI, BFI); 1327 } 1328 1329 void llvm::saveUsesAndErase(MachineInstr &MI, MachineRegisterInfo &MRI, 1330 LostDebugLocObserver *LocObserver, 1331 SmallInstListTy &DeadInstChain) { 1332 for (MachineOperand &Op : MI.uses()) { 1333 if (Op.isReg() && Op.getReg().isVirtual()) 1334 DeadInstChain.insert(MRI.getVRegDef(Op.getReg())); 1335 } 1336 LLVM_DEBUG(dbgs() << MI << "Is dead; erasing.\n"); 1337 DeadInstChain.remove(&MI); 1338 MI.eraseFromParent(); 1339 if (LocObserver) 1340 LocObserver->checkpoint(false); 1341 } 1342 1343 void llvm::eraseInstrs(ArrayRef<MachineInstr *> DeadInstrs, 1344 MachineRegisterInfo &MRI, 1345 LostDebugLocObserver *LocObserver) { 1346 SmallInstListTy DeadInstChain; 1347 for (MachineInstr *MI : DeadInstrs) 1348 saveUsesAndErase(*MI, MRI, LocObserver, DeadInstChain); 1349 1350 while (!DeadInstChain.empty()) { 1351 MachineInstr *Inst = DeadInstChain.pop_back_val(); 1352 if (!isTriviallyDead(*Inst, MRI)) 1353 continue; 1354 saveUsesAndErase(*Inst, MRI, LocObserver, DeadInstChain); 1355 } 1356 } 1357 1358 void llvm::eraseInstr(MachineInstr &MI, MachineRegisterInfo &MRI, 1359 LostDebugLocObserver *LocObserver) { 1360 return eraseInstrs({&MI}, MRI, LocObserver); 1361 } 1362 1363 void llvm::salvageDebugInfo(const MachineRegisterInfo &MRI, MachineInstr &MI) { 1364 for (auto &Def : MI.defs()) { 1365 assert(Def.isReg() && "Must be a reg"); 1366 1367 SmallVector<MachineOperand *, 16> DbgUsers; 1368 for (auto &MOUse : MRI.use_operands(Def.getReg())) { 1369 MachineInstr *DbgValue = MOUse.getParent(); 1370 // Ignore partially formed DBG_VALUEs. 1371 if (DbgValue->isNonListDebugValue() && DbgValue->getNumOperands() == 4) { 1372 DbgUsers.push_back(&MOUse); 1373 } 1374 } 1375 1376 if (!DbgUsers.empty()) { 1377 salvageDebugInfoForDbgValue(MRI, MI, DbgUsers); 1378 } 1379 } 1380 } 1381