diff options
author | Dimitry Andric <dim@FreeBSD.org> | 2017-04-26 19:45:00 +0000 |
---|---|---|
committer | Dimitry Andric <dim@FreeBSD.org> | 2017-04-26 19:45:00 +0000 |
commit | 12f3ca4cdb95b193af905a00e722a4dcb40b3de3 (patch) | |
tree | ae1a7fcfc24a8d4b23206c57121c3f361d4b7f84 /lib/CodeGen | |
parent | d99dafe2e4a385dd2a6c76da6d8258deb100657b (diff) |
Diffstat (limited to 'lib/CodeGen')
36 files changed, 380 insertions, 316 deletions
diff --git a/lib/CodeGen/AggressiveAntiDepBreaker.cpp b/lib/CodeGen/AggressiveAntiDepBreaker.cpp index 955524c2a676e..3a57772cc7f55 100644 --- a/lib/CodeGen/AggressiveAntiDepBreaker.cpp +++ b/lib/CodeGen/AggressiveAntiDepBreaker.cpp @@ -964,10 +964,8 @@ unsigned AggressiveAntiDepBreaker::BreakAntiDependencies( // sure to update that as well. const SUnit *SU = MISUnitMap[Q.second.Operand->getParent()]; if (!SU) continue; - for (DbgValueVector::iterator DVI = DbgValues.begin(), - DVE = DbgValues.end(); DVI != DVE; ++DVI) - if (DVI->second == Q.second.Operand->getParent()) - UpdateDbgValue(*DVI->first, AntiDepReg, NewReg); + UpdateDbgValues(DbgValues, Q.second.Operand->getParent(), + AntiDepReg, NewReg); } // We just went back in time and modified history; the diff --git a/lib/CodeGen/AntiDepBreaker.h b/lib/CodeGen/AntiDepBreaker.h index 04f7f419f5eac..d14d93100adbf 100644 --- a/lib/CodeGen/AntiDepBreaker.h +++ b/lib/CodeGen/AntiDepBreaker.h @@ -60,6 +60,25 @@ public: if (MI.getOperand(0).isReg() && MI.getOperand(0).getReg() == OldReg) MI.getOperand(0).setReg(NewReg); } + + /// Update all DBG_VALUE instructions that may be affected by the dependency + /// breaker's update of ParentMI to use NewReg. + void UpdateDbgValues(const DbgValueVector &DbgValues, MachineInstr *ParentMI, + unsigned OldReg, unsigned NewReg) { + // The following code is dependent on the order in which the DbgValues are + // constructed in ScheduleDAGInstrs::buildSchedGraph. + MachineInstr *PrevDbgMI = nullptr; + for (const auto &DV : make_range(DbgValues.crbegin(), DbgValues.crend())) { + MachineInstr *PrevMI = DV.second; + if ((PrevMI == ParentMI) || (PrevMI == PrevDbgMI)) { + MachineInstr *DbgMI = DV.first; + UpdateDbgValue(*DbgMI, OldReg, NewReg); + PrevDbgMI = DbgMI; + } else if (PrevDbgMI) { + break; // If no match and already found a DBG_VALUE, we're done. + } + } + } }; } diff --git a/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/lib/CodeGen/AsmPrinter/AsmPrinter.cpp index 028c79f3ab6d2..d99065b1b67a0 100644 --- a/lib/CodeGen/AsmPrinter/AsmPrinter.cpp +++ b/lib/CodeGen/AsmPrinter/AsmPrinter.cpp @@ -825,41 +825,25 @@ static bool emitDebugValueComment(const MachineInstr *MI, AsmPrinter &AP) { OS << Name << ":"; } OS << V->getName(); - - const DIExpression *Expr = MI->getDebugExpression(); - auto Fragment = Expr->getFragmentInfo(); - if (Fragment) - OS << " [fragment offset=" << Fragment->OffsetInBits - << " size=" << Fragment->SizeInBits << "]"; OS << " <- "; // The second operand is only an offset if it's an immediate. - bool Deref = false; bool MemLoc = MI->getOperand(0).isReg() && MI->getOperand(1).isImm(); int64_t Offset = MemLoc ? MI->getOperand(1).getImm() : 0; - for (unsigned i = 0; i < Expr->getNumElements(); ++i) { - uint64_t Op = Expr->getElement(i); - if (Op == dwarf::DW_OP_LLVM_fragment) { - // There can't be any operands after this in a valid expression - break; - } else if (Deref) { - // We currently don't support extra Offsets or derefs after the first - // one. Bail out early instead of emitting an incorrect comment. - OS << " [complex expression]"; - AP.OutStreamer->emitRawComment(OS.str()); - return true; - } else if (Op == dwarf::DW_OP_deref) { - Deref = true; - continue; - } - - uint64_t ExtraOffset = Expr->getElement(i++); - if (Op == dwarf::DW_OP_plus) - Offset += ExtraOffset; - else { - assert(Op == dwarf::DW_OP_minus); - Offset -= ExtraOffset; + const DIExpression *Expr = MI->getDebugExpression(); + if (Expr->getNumElements()) { + OS << '['; + bool NeedSep = false; + for (auto Op : Expr->expr_ops()) { + if (NeedSep) + OS << ", "; + else + NeedSep = true; + OS << dwarf::OperationEncodingString(Op.getOp()); + for (unsigned I = 0; I < Op.getNumArgs(); ++I) + OS << ' ' << Op.getArg(I); } + OS << "] "; } // Register or immediate value. Register 0 means undef. @@ -890,7 +874,7 @@ static bool emitDebugValueComment(const MachineInstr *MI, AsmPrinter &AP) { const TargetFrameLowering *TFI = AP.MF->getSubtarget().getFrameLowering(); Offset += TFI->getFrameIndexReference(*AP.MF, MI->getOperand(0).getIndex(), Reg); - Deref = true; + MemLoc = true; } if (Reg == 0) { // Suppress offset, it is not meaningful here. @@ -899,12 +883,12 @@ static bool emitDebugValueComment(const MachineInstr *MI, AsmPrinter &AP) { AP.OutStreamer->emitRawComment(OS.str()); return true; } - if (MemLoc || Deref) + if (MemLoc) OS << '['; OS << PrintReg(Reg, AP.MF->getSubtarget().getRegisterInfo()); } - if (MemLoc || Deref) + if (MemLoc) OS << '+' << Offset << ']'; // NOTE: Want this comment at start of line, don't emit with AddComment. @@ -936,6 +920,16 @@ void AsmPrinter::emitCFIInstruction(const MachineInstr &MI) { if (needsCFIMoves() == CFI_M_None) return; + // If there is no "real" instruction following this CFI instruction, skip + // emitting it; it would be beyond the end of the function's FDE range. + auto *MBB = MI.getParent(); + auto I = std::next(MI.getIterator()); + while (I != MBB->end() && I->isTransient()) + ++I; + if (I == MBB->instr_end() && + MBB->getReverseIterator() == MBB->getParent()->rbegin()) + return; + const std::vector<MCCFIInstruction> &Instrs = MF->getFrameInstructions(); unsigned CFIIndex = MI.getOperand(0).getCFIIndex(); const MCCFIInstruction &CFI = Instrs[CFIIndex]; @@ -1046,15 +1040,23 @@ void AsmPrinter::EmitFunctionBody() { // If the function is empty and the object file uses .subsections_via_symbols, // then we need to emit *something* to the function body to prevent the // labels from collapsing together. Just emit a noop. - if ((MAI->hasSubsectionsViaSymbols() && !HasAnyRealCode)) { + // Similarly, don't emit empty functions on Windows either. It can lead to + // duplicate entries (two functions with the same RVA) in the Guard CF Table + // after linking, causing the kernel not to load the binary: + // https://developercommunity.visualstudio.com/content/problem/45366/vc-linker-creates-invalid-dll-with-clang-cl.html + // FIXME: Hide this behind some API in e.g. MCAsmInfo or MCTargetStreamer. + const Triple &TT = TM.getTargetTriple(); + if (!HasAnyRealCode && (MAI->hasSubsectionsViaSymbols() || + (TT.isOSWindows() && TT.isOSBinFormatCOFF()))) { MCInst Noop; - MF->getSubtarget().getInstrInfo()->getNoopForMachoTarget(Noop); - OutStreamer->AddComment("avoids zero-length function"); + MF->getSubtarget().getInstrInfo()->getNoop(Noop); // Targets can opt-out of emitting the noop here by leaving the opcode // unspecified. - if (Noop.getOpcode()) + if (Noop.getOpcode()) { + OutStreamer->AddComment("avoids zero-length function"); OutStreamer->EmitInstruction(Noop, getSubtargetInfo()); + } } const Function *F = MF->getFunction(); diff --git a/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp b/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp index 683e622e3d537..a0bf1632dff39 100644 --- a/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp +++ b/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp @@ -144,6 +144,9 @@ void AsmPrinter::EmitInlineAsm(StringRef Str, const MCSubtargetInfo &STI, " we don't have an asm parser for this target\n"); Parser->setAssemblerDialect(Dialect); Parser->setTargetParser(*TAP.get()); + if (Dialect == InlineAsm::AD_Intel) + // We need this flag to be able to parse numbers like "0bH" + Parser->setParsingInlineAsm(true); if (MF) { const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo(); TAP->SetFrameRegister(TRI->getFrameRegister(*MF)); diff --git a/lib/CodeGen/AsmPrinter/DIE.cpp b/lib/CodeGen/AsmPrinter/DIE.cpp index 31c2b3b5e752f..30bfd7c94e68b 100644 --- a/lib/CodeGen/AsmPrinter/DIE.cpp +++ b/lib/CodeGen/AsmPrinter/DIE.cpp @@ -655,20 +655,12 @@ void DIEEntry::EmitValue(const AsmPrinter *AP, dwarf::Form Form) const { case dwarf::DW_FORM_ref_addr: { // Get the absolute offset for this DIE within the debug info/types section. unsigned Addr = Entry->getDebugSectionOffset(); - if (AP->MAI->doesDwarfUseRelocationsAcrossSections()) { - const DwarfDebug *DD = AP->getDwarfDebug(); - if (DD) - assert(!DD->useSplitDwarf() && - "TODO: dwo files can't have relocations."); - const DIEUnit *Unit = Entry->getUnit(); - assert(Unit && "CUDie should belong to a CU."); - MCSection *Section = Unit->getSection(); - if (Section) { - const MCSymbol *SectionSym = Section->getBeginSymbol(); - AP->EmitLabelPlusOffset(SectionSym, Addr, SizeOf(AP, Form), true); - return; - } + if (const MCSymbol *SectionSym = + Entry->getUnit()->getCrossSectionRelativeBaseAddress()) { + AP->EmitLabelPlusOffset(SectionSym, Addr, SizeOf(AP, Form), true); + return; } + AP->OutStreamer->EmitIntValue(Addr, SizeOf(AP, Form)); return; } diff --git a/lib/CodeGen/AsmPrinter/DwarfCompileUnit.h b/lib/CodeGen/AsmPrinter/DwarfCompileUnit.h index 9a64b4b76b06e..20a415150b4d7 100644 --- a/lib/CodeGen/AsmPrinter/DwarfCompileUnit.h +++ b/lib/CodeGen/AsmPrinter/DwarfCompileUnit.h @@ -28,7 +28,7 @@ class DwarfFile; class MCSymbol; class LexicalScope; -class DwarfCompileUnit : public DwarfUnit { +class DwarfCompileUnit final : public DwarfUnit { /// A numeric ID unique among all CUs in the module unsigned UniqueID; diff --git a/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/lib/CodeGen/AsmPrinter/DwarfDebug.cpp index d72656bcc58d4..6f442f5c3172d 100644 --- a/lib/CodeGen/AsmPrinter/DwarfDebug.cpp +++ b/lib/CodeGen/AsmPrinter/DwarfDebug.cpp @@ -91,14 +91,6 @@ DwarfAccelTables("dwarf-accel-tables", cl::Hidden, cl::init(Default)); static cl::opt<DefaultOnOff> -SplitDwarf("split-dwarf", cl::Hidden, - cl::desc("Output DWARF5 split debug info."), - cl::values(clEnumVal(Default, "Default for platform"), - clEnumVal(Enable, "Enabled"), - clEnumVal(Disable, "Disabled")), - cl::init(Default)); - -static cl::opt<DefaultOnOff> DwarfPubSections("generate-dwarf-pub-sections", cl::Hidden, cl::desc("Generate DWARF pubnames and pubtypes sections"), cl::values(clEnumVal(Default, "Default for platform"), @@ -253,11 +245,8 @@ DwarfDebug::DwarfDebug(AsmPrinter *A, Module *M) HasAppleExtensionAttributes = tuneForLLDB(); - // Handle split DWARF. Off by default for now. - if (SplitDwarf == Default) - HasSplitDwarf = false; - else - HasSplitDwarf = SplitDwarf == Enable; + // Handle split DWARF. + HasSplitDwarf = !Asm->TM.Options.MCOptions.SplitDwarfFile.empty(); // Pubnames/pubtypes on by default for GDB. if (DwarfPubSections == Default) @@ -412,7 +401,7 @@ DwarfDebug::constructDwarfCompileUnit(const DICompileUnit *DIUnit) { if (useSplitDwarf()) { NewCU.setSkeleton(constructSkeletonCU(NewCU)); NewCU.addString(Die, dwarf::DW_AT_GNU_dwo_name, - DIUnit->getSplitDebugFilename()); + Asm->TM.Options.MCOptions.SplitDwarfFile); } // LTO with assembly output shares a single line table amongst multiple CUs. @@ -1885,7 +1874,7 @@ void DwarfDebug::emitDebugMacinfo() { void DwarfDebug::initSkeletonUnit(const DwarfUnit &U, DIE &Die, std::unique_ptr<DwarfCompileUnit> NewU) { NewU->addString(Die, dwarf::DW_AT_GNU_dwo_name, - U.getCUNode()->getSplitDebugFilename()); + Asm->TM.Options.MCOptions.SplitDwarfFile); if (!CompilationDir.empty()) NewU->addString(Die, dwarf::DW_AT_comp_dir, CompilationDir); diff --git a/lib/CodeGen/AsmPrinter/DwarfExpression.cpp b/lib/CodeGen/AsmPrinter/DwarfExpression.cpp index f65dc151f3019..ccd326917bfd3 100644 --- a/lib/CodeGen/AsmPrinter/DwarfExpression.cpp +++ b/lib/CodeGen/AsmPrinter/DwarfExpression.cpp @@ -117,8 +117,9 @@ bool DwarfExpression::addMachineReg(const TargetRegisterInfo &TRI, // Otherwise, attempt to find a covering set of sub-register numbers. // For example, Q0 on ARM is a composition of D0+D1. unsigned CurPos = 0; - // The size of the register in bits, assuming 8 bits per byte. - unsigned RegSize = TRI.getMinimalPhysRegClass(MachineReg)->getSize() * 8; + // The size of the register in bits. + const TargetRegisterClass *RC = TRI.getMinimalPhysRegClass(MachineReg); + unsigned RegSize = TRI.getRegSizeInBits(*RC); // Keep track of the bits in the register we already emitted, so we // can avoid emitting redundant aliasing subregs. SmallBitVector Coverage(RegSize, false); @@ -198,8 +199,10 @@ bool DwarfExpression::addMachineRegExpression(const TargetRegisterInfo &TRI, unsigned MachineReg, unsigned FragmentOffsetInBits) { auto Fragment = ExprCursor.getFragmentInfo(); - if (!addMachineReg(TRI, MachineReg, Fragment ? Fragment->SizeInBits : ~1U)) + if (!addMachineReg(TRI, MachineReg, Fragment ? Fragment->SizeInBits : ~1U)) { + LocationKind = Unknown; return false; + } bool HasComplexExpression = false; auto Op = ExprCursor.peek(); @@ -212,6 +215,7 @@ bool DwarfExpression::addMachineRegExpression(const TargetRegisterInfo &TRI, // operation to multiple DW_OP_pieces. if (HasComplexExpression && DwarfRegs.size() > 1) { DwarfRegs.clear(); + LocationKind = Unknown; return false; } @@ -233,6 +237,7 @@ bool DwarfExpression::addMachineRegExpression(const TargetRegisterInfo &TRI, return Op.getOp() == dwarf::DW_OP_stack_value; })) { DwarfRegs.clear(); + LocationKind = Unknown; return false; } @@ -343,7 +348,6 @@ void DwarfExpression::addExpression(DIExpressionCursor &&ExprCursor, emitUnsigned(Op->getArg(0)); break; case dwarf::DW_OP_stack_value: - assert(LocationKind == Unknown || LocationKind == Implicit); LocationKind = Implicit; break; case dwarf::DW_OP_swap: diff --git a/lib/CodeGen/AsmPrinter/DwarfUnit.cpp b/lib/CodeGen/AsmPrinter/DwarfUnit.cpp index bac0c204d04fd..16fb20dd7e20d 100644 --- a/lib/CodeGen/AsmPrinter/DwarfUnit.cpp +++ b/lib/CodeGen/AsmPrinter/DwarfUnit.cpp @@ -1595,3 +1595,11 @@ void DwarfTypeUnit::addGlobalType(const DIType *Ty, const DIE &Die, const DIScope *Context) { getCU().addGlobalTypeUnitType(Ty, Context); } + +const MCSymbol *DwarfUnit::getCrossSectionRelativeBaseAddress() const { + if (!Asm->MAI->doesDwarfUseRelocationsAcrossSections()) + return nullptr; + if (isDwoUnit()) + return nullptr; + return getSection()->getBeginSymbol(); +} diff --git a/lib/CodeGen/AsmPrinter/DwarfUnit.h b/lib/CodeGen/AsmPrinter/DwarfUnit.h index d626ef920f956..e84df46508827 100644 --- a/lib/CodeGen/AsmPrinter/DwarfUnit.h +++ b/lib/CodeGen/AsmPrinter/DwarfUnit.h @@ -104,8 +104,6 @@ protected: bool applySubprogramDefinitionAttributes(const DISubprogram *SP, DIE &SPDie); public: - virtual ~DwarfUnit(); - // Accessors. AsmPrinter* getAsmPrinter() const { return Asm; } uint16_t getLanguage() const { return CUNode->getSourceLanguage(); } @@ -289,6 +287,8 @@ public: void constructTypeDIE(DIE &Buffer, const DICompositeType *CTy); protected: + ~DwarfUnit(); + /// Create new static data member DIE. DIE *getOrCreateStaticMemberDIE(const DIDerivedType *DT); @@ -335,9 +335,10 @@ private: void setIndexTyDie(DIE *D) { IndexTyDie = D; } virtual bool isDwoUnit() const = 0; + const MCSymbol *getCrossSectionRelativeBaseAddress() const override; }; -class DwarfTypeUnit : public DwarfUnit { +class DwarfTypeUnit final : public DwarfUnit { uint64_t TypeSignature; const DIE *Ty; DwarfCompileUnit &CU; diff --git a/lib/CodeGen/CriticalAntiDepBreaker.cpp b/lib/CodeGen/CriticalAntiDepBreaker.cpp index e1eeddf0816c1..b2d6652b075e7 100644 --- a/lib/CodeGen/CriticalAntiDepBreaker.cpp +++ b/lib/CodeGen/CriticalAntiDepBreaker.cpp @@ -648,10 +648,8 @@ BreakAntiDependencies(const std::vector<SUnit>& SUnits, // as well. const SUnit *SU = MISUnitMap[Q->second->getParent()]; if (!SU) continue; - for (DbgValueVector::iterator DVI = DbgValues.begin(), - DVE = DbgValues.end(); DVI != DVE; ++DVI) - if (DVI->second == Q->second->getParent()) - UpdateDbgValue(*DVI->first, AntiDepReg, NewReg); + UpdateDbgValues(DbgValues, Q->second->getParent(), + AntiDepReg, NewReg); } // We just went back in time and modified history; the diff --git a/lib/CodeGen/GlobalISel/InstructionSelect.cpp b/lib/CodeGen/GlobalISel/InstructionSelect.cpp index 26454c1ef00f9..cf97c635e79ab 100644 --- a/lib/CodeGen/GlobalISel/InstructionSelect.cpp +++ b/lib/CodeGen/GlobalISel/InstructionSelect.cpp @@ -145,6 +145,8 @@ bool InstructionSelect::runOnMachineFunction(MachineFunction &MF) { } } + const TargetRegisterInfo &TRI = *MF.getSubtarget().getRegisterInfo(); + // Now that selection is complete, there are no more generic vregs. Verify // that the size of the now-constrained vreg is unchanged and that it has a // register class. @@ -165,7 +167,7 @@ bool InstructionSelect::runOnMachineFunction(MachineFunction &MF) { continue; if (VRegToType.second.isValid() && - VRegToType.second.getSizeInBits() > (RC->getSize() * 8)) { + VRegToType.second.getSizeInBits() > TRI.getRegSizeInBits(*RC)) { reportGISelFailure(MF, TPC, MORE, "gisel-select", "VReg has explicit size different from class size", *MI); diff --git a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp index 58778077bc0e7..ef5818dabe232 100644 --- a/lib/CodeGen/GlobalISel/LegalizerHelper.cpp +++ b/lib/CodeGen/GlobalISel/LegalizerHelper.cpp @@ -76,6 +76,12 @@ void LegalizerHelper::extractParts(unsigned Reg, LLT Ty, int NumParts, static RTLIB::Libcall getRTLibDesc(unsigned Opcode, unsigned Size) { switch (Opcode) { + case TargetOpcode::G_SDIV: + assert(Size == 32 && "Unsupported size"); + return RTLIB::SDIV_I32; + case TargetOpcode::G_UDIV: + assert(Size == 32 && "Unsupported size"); + return RTLIB::UDIV_I32; case TargetOpcode::G_FADD: assert((Size == 32 || Size == 64) && "Unsupported size"); return Size == 64 ? RTLIB::ADD_F64 : RTLIB::ADD_F32; @@ -87,31 +93,43 @@ static RTLIB::Libcall getRTLibDesc(unsigned Opcode, unsigned Size) { llvm_unreachable("Unknown libcall function"); } +static LegalizerHelper::LegalizeResult +simpleLibcall(MachineInstr &MI, MachineIRBuilder &MIRBuilder, unsigned Size, + Type *OpType) { + auto &CLI = *MIRBuilder.getMF().getSubtarget().getCallLowering(); + auto &TLI = *MIRBuilder.getMF().getSubtarget().getTargetLowering(); + auto Libcall = getRTLibDesc(MI.getOpcode(), Size); + const char *Name = TLI.getLibcallName(Libcall); + MIRBuilder.getMF().getFrameInfo().setHasCalls(true); + CLI.lowerCall(MIRBuilder, TLI.getLibcallCallingConv(Libcall), + MachineOperand::CreateES(Name), + {MI.getOperand(0).getReg(), OpType}, + {{MI.getOperand(1).getReg(), OpType}, + {MI.getOperand(2).getReg(), OpType}}); + MI.eraseFromParent(); + return LegalizerHelper::Legalized; +} + LegalizerHelper::LegalizeResult LegalizerHelper::libcall(MachineInstr &MI) { LLT Ty = MRI.getType(MI.getOperand(0).getReg()); unsigned Size = Ty.getSizeInBits(); + auto &Ctx = MIRBuilder.getMF().getFunction()->getContext(); MIRBuilder.setInstr(MI); switch (MI.getOpcode()) { default: return UnableToLegalize; + case TargetOpcode::G_SDIV: + case TargetOpcode::G_UDIV: { + Type *Ty = Type::getInt32Ty(Ctx); + return simpleLibcall(MI, MIRBuilder, Size, Ty); + } case TargetOpcode::G_FADD: case TargetOpcode::G_FPOW: case TargetOpcode::G_FREM: { - auto &Ctx = MIRBuilder.getMF().getFunction()->getContext(); Type *Ty = Size == 64 ? Type::getDoubleTy(Ctx) : Type::getFloatTy(Ctx); - auto &CLI = *MIRBuilder.getMF().getSubtarget().getCallLowering(); - auto &TLI = *MIRBuilder.getMF().getSubtarget().getTargetLowering(); - auto Libcall = getRTLibDesc(MI.getOpcode(), Size); - const char *Name = TLI.getLibcallName(Libcall); - MIRBuilder.getMF().getFrameInfo().setHasCalls(true); - CLI.lowerCall( - MIRBuilder, TLI.getLibcallCallingConv(Libcall), - MachineOperand::CreateES(Name), {MI.getOperand(0).getReg(), Ty}, - {{MI.getOperand(1).getReg(), Ty}, {MI.getOperand(2).getReg(), Ty}}); - MI.eraseFromParent(); - return Legalized; + return simpleLibcall(MI, MIRBuilder, Size, Ty); } } } diff --git a/lib/CodeGen/GlobalISel/RegisterBank.cpp b/lib/CodeGen/GlobalISel/RegisterBank.cpp index 940957d021524..83b21e6370971 100644 --- a/lib/CodeGen/GlobalISel/RegisterBank.cpp +++ b/lib/CodeGen/GlobalISel/RegisterBank.cpp @@ -48,7 +48,7 @@ bool RegisterBank::verify(const TargetRegisterInfo &TRI) const { // Verify that the Size of the register bank is big enough to cover // all the register classes it covers. - assert((getSize() >= SubRC.getSize() * 8) && + assert(getSize() >= TRI.getRegSizeInBits(SubRC) && "Size is not big enough for all the subclasses!"); assert(covers(SubRC) && "Not all subclasses are covered"); } diff --git a/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp b/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp index b2df2f1596769..d5ae9a6776a4e 100644 --- a/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp +++ b/lib/CodeGen/GlobalISel/RegisterBankInfo.cpp @@ -421,7 +421,7 @@ unsigned RegisterBankInfo::getSizeInBits(unsigned Reg, RC = MRI.getRegClass(Reg); } assert(RC && "Unable to deduce the register class"); - return RC->getSize() * 8; + return TRI.getRegSizeInBits(*RC); } //------------------------------------------------------------------------------ diff --git a/lib/CodeGen/MIRPrinter.cpp b/lib/CodeGen/MIRPrinter.cpp index 6da174a536666..b6624b88fe231 100644 --- a/lib/CodeGen/MIRPrinter.cpp +++ b/lib/CodeGen/MIRPrinter.cpp @@ -925,9 +925,6 @@ void MIPrinter::print(const MachineOperand &Op, const TargetRegisterInfo *TRI, << CmpInst::getPredicateName(Pred) << ')'; break; } - case MachineOperand::MO_Placeholder: - OS << "<placeholder>"; - break; } } diff --git a/lib/CodeGen/MachineInstr.cpp b/lib/CodeGen/MachineInstr.cpp index 4bd5fbfe38e6e..1faf6292a9c1a 100644 --- a/lib/CodeGen/MachineInstr.cpp +++ b/lib/CodeGen/MachineInstr.cpp @@ -287,8 +287,6 @@ bool MachineOperand::isIdenticalTo(const MachineOperand &Other) const { return getIntrinsicID() == Other.getIntrinsicID(); case MachineOperand::MO_Predicate: return getPredicate() == Other.getPredicate(); - case MachineOperand::MO_Placeholder: - return true; } llvm_unreachable("Invalid machine operand type"); } @@ -337,8 +335,6 @@ hash_code llvm::hash_value(const MachineOperand &MO) { return hash_combine(MO.getType(), MO.getTargetFlags(), MO.getIntrinsicID()); case MachineOperand::MO_Predicate: return hash_combine(MO.getType(), MO.getTargetFlags(), MO.getPredicate()); - case MachineOperand::MO_Placeholder: - return hash_combine(); } llvm_unreachable("Invalid machine operand type"); } @@ -515,9 +511,6 @@ void MachineOperand::print(raw_ostream &OS, ModuleSlotTracker &MST, << CmpInst::getPredicateName(Pred) << '>'; break; } - case MachineOperand::MO_Placeholder: - OS << "<placeholder>"; - break; } if (unsigned TF = getTargetFlags()) OS << "[TF=" << TF << ']'; diff --git a/lib/CodeGen/MachineLICM.cpp b/lib/CodeGen/MachineLICM.cpp index b3d18435985e8..7eb991744f01c 100644 --- a/lib/CodeGen/MachineLICM.cpp +++ b/lib/CodeGen/MachineLICM.cpp @@ -330,7 +330,7 @@ bool MachineLICM::runOnMachineFunction(MachineFunction &MF) { /// Return true if instruction stores to the specified frame. static bool InstructionStoresToFI(const MachineInstr *MI, int FI) { // If we lost memory operands, conservatively assume that the instruction - // writes to all slots. + // writes to all slots. if (MI->memoperands_empty()) return true; for (const MachineMemOperand *MemOp : MI->memoperands()) { @@ -708,7 +708,7 @@ void MachineLICM::SinkIntoLoop() { for (MachineBasicBlock::instr_iterator I = Preheader->instr_begin(); I != Preheader->instr_end(); ++I) { // We need to ensure that we can safely move this instruction into the loop. - // As such, it must not have side-effects, e.g. such as a call has. + // As such, it must not have side-effects, e.g. such as a call has. if (IsLoopInvariantInst(*I) && !HasLoopPHIUse(&*I)) Candidates.push_back(&*I); } @@ -837,9 +837,9 @@ MachineLICM::calcRegisterCost(const MachineInstr *MI, bool ConsiderSeen, /// constant pool. static bool mayLoadFromGOTOrConstantPool(MachineInstr &MI) { assert (MI.mayLoad() && "Expected MI that loads!"); - + // If we lost memory operands, conservatively assume that the instruction - // reads from everything.. + // reads from everything.. if (MI.memoperands_empty()) return true; @@ -1337,7 +1337,7 @@ bool MachineLICM::Hoist(MachineInstr *MI, MachineBasicBlock *Preheader) { Preheader->splice(Preheader->getFirstTerminator(),MI->getParent(),MI); // Since we are moving the instruction out of its basic block, we do not - // retain its debug location. Doing so would degrade the debugging + // retain its debug location. Doing so would degrade the debugging // experience and adversely affect the accuracy of profiling information. MI->setDebugLoc(DebugLoc()); diff --git a/lib/CodeGen/PrologEpilogInserter.cpp b/lib/CodeGen/PrologEpilogInserter.cpp index 1354009794cb4..570a0cd0ba901 100644 --- a/lib/CodeGen/PrologEpilogInserter.cpp +++ b/lib/CodeGen/PrologEpilogInserter.cpp @@ -373,22 +373,22 @@ static void assignCalleeSavedSpillSlots(MachineFunction &F, FixedSlot->Reg != Reg) ++FixedSlot; + unsigned Size = RegInfo->getSpillSize(*RC); if (FixedSlot == FixedSpillSlots + NumFixedSpillSlots) { // Nope, just spill it anywhere convenient. - unsigned Align = RC->getAlignment(); + unsigned Align = RegInfo->getSpillAlignment(*RC); unsigned StackAlign = TFI->getStackAlignment(); // We may not be able to satisfy the desired alignment specification of // the TargetRegisterClass if the stack alignment is smaller. Use the // min. Align = std::min(Align, StackAlign); - FrameIdx = MFI.CreateStackObject(RC->getSize(), Align, true); + FrameIdx = MFI.CreateStackObject(Size, Align, true); if ((unsigned)FrameIdx < MinCSFrameIndex) MinCSFrameIndex = FrameIdx; if ((unsigned)FrameIdx > MaxCSFrameIndex) MaxCSFrameIndex = FrameIdx; } else { // Spill it to the stack where we must. - FrameIdx = - MFI.CreateFixedSpillStackObject(RC->getSize(), FixedSlot->Offset); + FrameIdx = MFI.CreateFixedSpillStackObject(Size, FixedSlot->Offset); } CS.setFrameIdx(FrameIdx); diff --git a/lib/CodeGen/RegAllocFast.cpp b/lib/CodeGen/RegAllocFast.cpp index 283d84629f8ee..c606b7b833104 100644 --- a/lib/CodeGen/RegAllocFast.cpp +++ b/lib/CodeGen/RegAllocFast.cpp @@ -212,8 +212,9 @@ int RAFast::getStackSpaceFor(unsigned VirtReg, const TargetRegisterClass *RC) { return SS; // Already has space allocated? // Allocate a new stack object for this spill location... - int FrameIdx = MF->getFrameInfo().CreateSpillStackObject(RC->getSize(), - RC->getAlignment()); + unsigned Size = TRI->getSpillSize(*RC); + unsigned Align = TRI->getSpillAlignment(*RC); + int FrameIdx = MF->getFrameInfo().CreateSpillStackObject(Size, Align); // Assign the slot. StackSlotForVirtReg[VirtReg] = FrameIdx; diff --git a/lib/CodeGen/RegisterScavenging.cpp b/lib/CodeGen/RegisterScavenging.cpp index 6392136fa2909..35db30f899768 100644 --- a/lib/CodeGen/RegisterScavenging.cpp +++ b/lib/CodeGen/RegisterScavenging.cpp @@ -395,8 +395,8 @@ unsigned RegScavenger::scavengeRegister(const TargetRegisterClass *RC, // Find an available scavenging slot with size and alignment matching // the requirements of the class RC. const MachineFrameInfo &MFI = MF.getFrameInfo(); - unsigned NeedSize = RC->getSize(); - unsigned NeedAlign = RC->getAlignment(); + unsigned NeedSize = TRI->getSpillSize(*RC); + unsigned NeedAlign = TRI->getSpillAlignment(*RC); unsigned SI = Scavenged.size(), Diff = std::numeric_limits<unsigned>::max(); int FIB = MFI.getObjectIndexBegin(), FIE = MFI.getObjectIndexEnd(); diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp index 4702d63cb617a..1251ae6262b89 100644 --- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp +++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -3878,27 +3878,29 @@ static bool isBSwapHWordElement(SDValue N, MutableArrayRef<SDNode *> Parts) { if (Opc != ISD::AND && Opc != ISD::SHL && Opc != ISD::SRL) return false; + SDValue N0 = N.getOperand(0); + unsigned Opc0 = N0.getOpcode(); + ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N.getOperand(1)); if (!N1C) return false; - unsigned Num; + unsigned MaskByteOffset; switch (N1C->getZExtValue()) { default: return false; - case 0xFF: Num = 0; break; - case 0xFF00: Num = 1; break; - case 0xFF0000: Num = 2; break; - case 0xFF000000: Num = 3; break; + case 0xFF: MaskByteOffset = 0; break; + case 0xFF00: MaskByteOffset = 1; break; + case 0xFF0000: MaskByteOffset = 2; break; + case 0xFF000000: MaskByteOffset = 3; break; } // Look for (x & 0xff) << 8 as well as ((x << 8) & 0xff00). - SDValue N0 = N.getOperand(0); if (Opc == ISD::AND) { - if (Num == 0 || Num == 2) { + if (MaskByteOffset == 0 || MaskByteOffset == 2) { // (x >> 8) & 0xff // (x >> 8) & 0xff0000 - if (N0.getOpcode() != ISD::SRL) + if (Opc0 != ISD::SRL) return false; ConstantSDNode *C = dyn_cast<ConstantSDNode>(N0.getOperand(1)); if (!C || C->getZExtValue() != 8) @@ -3906,7 +3908,7 @@ static bool isBSwapHWordElement(SDValue N, MutableArrayRef<SDNode *> Parts) { } else { // (x << 8) & 0xff00 // (x << 8) & 0xff000000 - if (N0.getOpcode() != ISD::SHL) + if (Opc0 != ISD::SHL) return false; ConstantSDNode *C = dyn_cast<ConstantSDNode>(N0.getOperand(1)); if (!C || C->getZExtValue() != 8) @@ -3915,7 +3917,7 @@ static bool isBSwapHWordElement(SDValue N, MutableArrayRef<SDNode *> Parts) { } else if (Opc == ISD::SHL) { // (x & 0xff) << 8 // (x & 0xff0000) << 8 - if (Num != 0 && Num != 2) + if (MaskByteOffset != 0 && MaskByteOffset != 2) return false; ConstantSDNode *C = dyn_cast<ConstantSDNode>(N.getOperand(1)); if (!C || C->getZExtValue() != 8) @@ -3923,17 +3925,17 @@ static bool isBSwapHWordElement(SDValue N, MutableArrayRef<SDNode *> Parts) { } else { // Opc == ISD::SRL // (x & 0xff00) >> 8 // (x & 0xff000000) >> 8 - if (Num != 1 && Num != 3) + if (MaskByteOffset != 1 && MaskByteOffset != 3) return false; ConstantSDNode *C = dyn_cast<ConstantSDNode>(N.getOperand(1)); if (!C || C->getZExtValue() != 8) return false; } - if (Parts[Num]) + if (Parts[MaskByteOffset]) return false; - Parts[Num] = N0.getOperand(0).getNode(); + Parts[MaskByteOffset] = N0.getOperand(0).getNode(); return true; } @@ -4198,20 +4200,22 @@ SDValue DAGCombiner::visitOR(SDNode *N) { // reassociate or if (SDValue ROR = ReassociateOps(ISD::OR, SDLoc(N), N0, N1)) return ROR; + // Canonicalize (or (and X, c1), c2) -> (and (or X, c2), c1|c2) // iff (c1 & c2) != 0. - if (N1C && N0.getOpcode() == ISD::AND && N0.getNode()->hasOneUse() && - isa<ConstantSDNode>(N0.getOperand(1))) { - ConstantSDNode *C1 = cast<ConstantSDNode>(N0.getOperand(1)); - if ((C1->getAPIntValue() & N1C->getAPIntValue()) != 0) { - if (SDValue COR = DAG.FoldConstantArithmetic(ISD::OR, SDLoc(N1), VT, - N1C, C1)) - return DAG.getNode( - ISD::AND, SDLoc(N), VT, - DAG.getNode(ISD::OR, SDLoc(N0), VT, N0.getOperand(0), N1), COR); - return SDValue(); + if (N1C && N0.getOpcode() == ISD::AND && N0.getNode()->hasOneUse()) { + if (ConstantSDNode *C1 = dyn_cast<ConstantSDNode>(N0.getOperand(1))) { + if (C1->getAPIntValue().intersects(N1C->getAPIntValue())) { + if (SDValue COR = + DAG.FoldConstantArithmetic(ISD::OR, SDLoc(N1), VT, N1C, C1)) + return DAG.getNode( + ISD::AND, SDLoc(N), VT, + DAG.getNode(ISD::OR, SDLoc(N0), VT, N0.getOperand(0), N1), COR); + return SDValue(); + } } } + // Simplify: (or (op x...), (op y...)) -> (op (or x, y)) if (N0.getOpcode() == N1.getOpcode()) if (SDValue Tmp = SimplifyBinOpWithSameOpcodeHands(N)) @@ -5611,24 +5615,24 @@ SDValue DAGCombiner::visitSRL(SDNode *N) { // fold (srl (trunc (srl x, c1)), c2) -> 0 or (trunc (srl x, (add c1, c2))) if (N1C && N0.getOpcode() == ISD::TRUNCATE && - N0.getOperand(0).getOpcode() == ISD::SRL && - isa<ConstantSDNode>(N0.getOperand(0)->getOperand(1))) { - uint64_t c1 = - cast<ConstantSDNode>(N0.getOperand(0)->getOperand(1))->getZExtValue(); - uint64_t c2 = N1C->getZExtValue(); - EVT InnerShiftVT = N0.getOperand(0).getValueType(); - EVT ShiftCountVT = N0.getOperand(0)->getOperand(1).getValueType(); - uint64_t InnerShiftSize = InnerShiftVT.getScalarSizeInBits(); - // This is only valid if the OpSizeInBits + c1 = size of inner shift. - if (c1 + OpSizeInBits == InnerShiftSize) { - SDLoc DL(N0); - if (c1 + c2 >= InnerShiftSize) - return DAG.getConstant(0, DL, VT); - return DAG.getNode(ISD::TRUNCATE, DL, VT, - DAG.getNode(ISD::SRL, DL, InnerShiftVT, - N0.getOperand(0)->getOperand(0), - DAG.getConstant(c1 + c2, DL, - ShiftCountVT))); + N0.getOperand(0).getOpcode() == ISD::SRL) { + if (auto N001C = isConstOrConstSplat(N0.getOperand(0).getOperand(1))) { + uint64_t c1 = N001C->getZExtValue(); + uint64_t c2 = N1C->getZExtValue(); + EVT InnerShiftVT = N0.getOperand(0).getValueType(); + EVT ShiftCountVT = N0.getOperand(0).getOperand(1).getValueType(); + uint64_t InnerShiftSize = InnerShiftVT.getScalarSizeInBits(); + // This is only valid if the OpSizeInBits + c1 = size of inner shift. + if (c1 + OpSizeInBits == InnerShiftSize) { + SDLoc DL(N0); + if (c1 + c2 >= InnerShiftSize) + return DAG.getConstant(0, DL, VT); + return DAG.getNode(ISD::TRUNCATE, DL, VT, + DAG.getNode(ISD::SRL, DL, InnerShiftVT, + N0.getOperand(0).getOperand(0), + DAG.getConstant(c1 + c2, DL, + ShiftCountVT))); + } } } @@ -11641,7 +11645,7 @@ bool DAGCombiner::SliceUpLoad(SDNode *N) { // Check if this is a trunc(lshr). if (User->getOpcode() == ISD::SRL && User->hasOneUse() && isa<ConstantSDNode>(User->getOperand(1))) { - Shift = cast<ConstantSDNode>(User->getOperand(1))->getZExtValue(); + Shift = User->getConstantOperandVal(1); User = *User->use_begin(); } @@ -13120,8 +13124,7 @@ SDValue DAGCombiner::visitINSERT_VECTOR_ELT(SDNode *N) { // do this only if indices are both constants and Idx1 < Idx0. if (InVec.getOpcode() == ISD::INSERT_VECTOR_ELT && InVec.hasOneUse() && isa<ConstantSDNode>(InVec.getOperand(2))) { - unsigned OtherElt = - cast<ConstantSDNode>(InVec.getOperand(2))->getZExtValue(); + unsigned OtherElt = InVec.getConstantOperandVal(2); if (Elt < OtherElt) { // Swap nodes. SDValue NewOp = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, VT, @@ -14065,7 +14068,7 @@ static SDValue combineConcatVectorOfExtracts(SDNode *N, SelectionDAG &DAG) { if (!isa<ConstantSDNode>(Op.getOperand(1))) return SDValue(); - int ExtIdx = cast<ConstantSDNode>(Op.getOperand(1))->getZExtValue(); + int ExtIdx = Op.getConstantOperandVal(1); // Ensure that we are extracting a subvector from a vector the same // size as the result. @@ -15049,7 +15052,7 @@ SDValue DAGCombiner::visitINSERT_SUBVECTOR(SDNode *N) { if (N0.getOpcode() == ISD::INSERT_SUBVECTOR && N0.hasOneUse() && N1.getValueType() == N0.getOperand(1).getValueType() && isa<ConstantSDNode>(N0.getOperand(2))) { - unsigned OtherIdx = cast<ConstantSDNode>(N0.getOperand(2))->getZExtValue(); + unsigned OtherIdx = N0.getConstantOperandVal(2); if (InsIdx < OtherIdx) { // Swap nodes. SDValue NewOp = DAG.getNode(ISD::INSERT_SUBVECTOR, SDLoc(N), VT, @@ -16088,6 +16091,19 @@ bool DAGCombiner::isAlias(LSBaseSDNode *Op0, LSBaseSDNode *Op1) const { if (Op1->isInvariant() && Op0->writeMem()) return false; + unsigned NumBytes0 = Op0->getMemoryVT().getSizeInBits() >> 3; + unsigned NumBytes1 = Op1->getMemoryVT().getSizeInBits() >> 3; + + // Check for BaseIndexOffset matching. + BaseIndexOffset BasePtr0 = BaseIndexOffset::match(Op0->getBasePtr(), DAG); + BaseIndexOffset BasePtr1 = BaseIndexOffset::match(Op1->getBasePtr(), DAG); + if (BasePtr0.equalBaseIndex(BasePtr1)) + return !((BasePtr0.Offset + NumBytes0 <= BasePtr1.Offset) || + (BasePtr1.Offset + NumBytes1 <= BasePtr0.Offset)); + + // FIXME: findBaseOffset and ConstantValue/GlobalValue/FrameIndex analysis + // modified to use BaseIndexOffset. + // Gather base node and offset information. SDValue Base0, Base1; int64_t Offset0, Offset1; @@ -16099,8 +16115,6 @@ bool DAGCombiner::isAlias(LSBaseSDNode *Op0, LSBaseSDNode *Op1) const { Base1, Offset1, GV1, CV1); // If they have the same base address, then check to see if they overlap. - unsigned NumBytes0 = Op0->getMemoryVT().getSizeInBits() >> 3; - unsigned NumBytes1 = Op1->getMemoryVT().getSizeInBits() >> 3; if (Base0 == Base1 || (GV0 && (GV0 == GV1)) || (CV0 && (CV0 == CV1))) return !((Offset0 + NumBytes0) <= Offset1 || (Offset1 + NumBytes1) <= Offset0); diff --git a/lib/CodeGen/SelectionDAG/InstrEmitter.cpp b/lib/CodeGen/SelectionDAG/InstrEmitter.cpp index e85d1951e3aed..b235e19aaab29 100644 --- a/lib/CodeGen/SelectionDAG/InstrEmitter.cpp +++ b/lib/CodeGen/SelectionDAG/InstrEmitter.cpp @@ -161,7 +161,8 @@ EmitCopyFromReg(SDNode *Node, unsigned ResNo, bool IsClone, bool IsCloned, if (VRBase) { DstRC = MRI->getRegClass(VRBase); } else if (UseRC) { - assert(UseRC->hasType(VT) && "Incompatible phys register def and uses!"); + assert(TRI->isTypeLegalForClass(*UseRC, VT) && + "Incompatible phys register def and uses!"); DstRC = UseRC; } else { DstRC = TLI->getRegClassFor(VT); diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp index 3bae3bf9ab7cf..fdebb8bd00dbd 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp @@ -3497,11 +3497,11 @@ bool SelectionDAGLegalize::ExpandNode(SDNode *Node) { // part. unsigned LoSize = VT.getSizeInBits(); SDValue HiLHS = - DAG.getNode(ISD::SRA, dl, VT, RHS, + DAG.getNode(ISD::SRA, dl, VT, LHS, DAG.getConstant(LoSize - 1, dl, TLI.getPointerTy(DAG.getDataLayout()))); SDValue HiRHS = - DAG.getNode(ISD::SRA, dl, VT, LHS, + DAG.getNode(ISD::SRA, dl, VT, RHS, DAG.getConstant(LoSize - 1, dl, TLI.getPointerTy(DAG.getDataLayout()))); diff --git a/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp b/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp index 85068e890756b..9ed70c9b4db97 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp @@ -3251,7 +3251,7 @@ SDValue DAGTypeLegalizer::PromoteIntRes_EXTRACT_SUBVECTOR(SDNode *N) { Ops.push_back(Op); } - return DAG.getNode(ISD::BUILD_VECTOR, dl, NOutVT, Ops); + return DAG.getBuildVector(NOutVT, dl, Ops); } @@ -3294,7 +3294,7 @@ SDValue DAGTypeLegalizer::PromoteIntRes_BUILD_VECTOR(SDNode *N) { Ops.push_back(Op); } - return DAG.getNode(ISD::BUILD_VECTOR, dl, NOutVT, Ops); + return DAG.getBuildVector(NOutVT, dl, Ops); } SDValue DAGTypeLegalizer::PromoteIntRes_SCALAR_TO_VECTOR(SDNode *N) { @@ -3342,7 +3342,7 @@ SDValue DAGTypeLegalizer::PromoteIntRes_CONCAT_VECTORS(SDNode *N) { } } - return DAG.getNode(ISD::BUILD_VECTOR, dl, NOutVT, Ops); + return DAG.getBuildVector(NOutVT, dl, Ops); } SDValue DAGTypeLegalizer::PromoteIntRes_EXTEND_VECTOR_INREG(SDNode *N) { @@ -3445,5 +3445,5 @@ SDValue DAGTypeLegalizer::PromoteIntOp_CONCAT_VECTORS(SDNode *N) { } } - return DAG.getNode(ISD::BUILD_VECTOR, dl, N->getValueType(0), NewOps); + return DAG.getBuildVector(N->getValueType(0), dl, NewOps); } diff --git a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp index c02b8960b36cb..aa69e0e2adfce 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeTypesGeneric.cpp @@ -362,8 +362,8 @@ SDValue DAGTypeLegalizer::ExpandOp_BITCAST(SDNode *N) { SmallVector<SDValue, 8> Ops; IntegerToVector(N->getOperand(0), NumElts, Ops, NVT.getVectorElementType()); - SDValue Vec = DAG.getNode(ISD::BUILD_VECTOR, dl, NVT, - makeArrayRef(Ops.data(), NumElts)); + SDValue Vec = + DAG.getBuildVector(NVT, dl, makeArrayRef(Ops.data(), NumElts)); return DAG.getNode(ISD::BITCAST, dl, N->getValueType(0), Vec); } @@ -396,10 +396,8 @@ SDValue DAGTypeLegalizer::ExpandOp_BUILD_VECTOR(SDNode *N) { NewElts.push_back(Hi); } - SDValue NewVec = DAG.getNode(ISD::BUILD_VECTOR, dl, - EVT::getVectorVT(*DAG.getContext(), - NewVT, NewElts.size()), - NewElts); + EVT NewVecVT = EVT::getVectorVT(*DAG.getContext(), NewVT, NewElts.size()); + SDValue NewVec = DAG.getBuildVector(NewVecVT, dl, NewElts); // Convert the new vector to the old vector type. return DAG.getNode(ISD::BITCAST, dl, VecVT, NewVec); @@ -458,7 +456,7 @@ SDValue DAGTypeLegalizer::ExpandOp_SCALAR_TO_VECTOR(SDNode *N) { SDValue UndefVal = DAG.getUNDEF(Ops[0].getValueType()); for (unsigned i = 1; i < NumElts; ++i) Ops[i] = UndefVal; - return DAG.getNode(ISD::BUILD_VECTOR, dl, VT, Ops); + return DAG.getBuildVector(VT, dl, Ops); } SDValue DAGTypeLegalizer::ExpandOp_NormalStore(SDNode *N, unsigned OpNo) { diff --git a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp index 1a7d7b7af5fa1..4a3160297d647 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp @@ -512,7 +512,7 @@ SDValue DAGTypeLegalizer::ScalarizeVecOp_UnaryOp(SDNode *N) { N->getValueType(0).getScalarType(), Elt); // Revectorize the result so the types line up with what the uses of this // expression expect. - return DAG.getNode(ISD::BUILD_VECTOR, SDLoc(N), N->getValueType(0), Op); + return DAG.getBuildVector(N->getValueType(0), SDLoc(N), Op); } /// The vectors to concatenate have length one - use a BUILD_VECTOR instead. @@ -523,16 +523,16 @@ SDValue DAGTypeLegalizer::ScalarizeVecOp_CONCAT_VECTORS(SDNode *N) { return DAG.getBuildVector(N->getValueType(0), SDLoc(N), Ops); } -/// If the input is a vector that needs to be scalarized, it must be <1 x ty>, -/// so just return the element, ignoring the index. -SDValue DAGTypeLegalizer::ScalarizeVecOp_EXTRACT_VECTOR_ELT(SDNode *N) { - SDValue Res = GetScalarizedVector(N->getOperand(0)); - if (Res.getValueType() != N->getValueType(0)) - Res = DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), N->getValueType(0), - Res); - return Res; -} - +/// If the input is a vector that needs to be scalarized, it must be <1 x ty>,
+/// so just return the element, ignoring the index.
+SDValue DAGTypeLegalizer::ScalarizeVecOp_EXTRACT_VECTOR_ELT(SDNode *N) {
+ EVT VT = N->getValueType(0);
+ SDValue Res = GetScalarizedVector(N->getOperand(0));
+ if (Res.getValueType() != VT)
+ Res = DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), VT, Res);
+ return Res;
+}
+
/// If the input condition is a vector that needs to be scalarized, it must be /// <1 x i1>, so just convert to a normal ISD::SELECT @@ -2631,7 +2631,7 @@ SDValue DAGTypeLegalizer::WidenVecRes_BITCAST(SDNode *N) { if (InVT.isVector()) NewVec = DAG.getNode(ISD::CONCAT_VECTORS, dl, NewInVT, Ops); else - NewVec = DAG.getNode(ISD::BUILD_VECTOR, dl, NewInVT, Ops); + NewVec = DAG.getBuildVector(NewInVT, dl, Ops); return DAG.getNode(ISD::BITCAST, dl, WidenVT, NewVec); } } diff --git a/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp b/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp index e923e30e50377..69b76fbe57d27 100644 --- a/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp +++ b/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp @@ -1320,6 +1320,18 @@ DelayForLiveRegsBottomUp(SUnit *SU, SmallVectorImpl<unsigned> &LRegs) { RegAdded, LRegs); const MCInstrDesc &MCID = TII->get(Node->getMachineOpcode()); + if (MCID.hasOptionalDef()) { + // Most ARM instructions have an OptionalDef for CPSR, to model the S-bit. + // This operand can be either a def of CPSR, if the S bit is set; or a use + // of %noreg. When the OptionalDef is set to a valid register, we need to + // handle it in the same way as an ImplicitDef. + for (unsigned i = 0; i < MCID.getNumDefs(); ++i) + if (MCID.OpInfo[i].isOptionalDef()) { + const SDValue &OptionalDef = Node->getOperand(i - Node->getNumValues()); + unsigned Reg = cast<RegisterSDNode>(OptionalDef)->getReg(); + CheckForLiveRegDef(SU, Reg, LiveRegDefs.get(), RegAdded, LRegs, TRI); + } + } if (!MCID.ImplicitDefs) continue; for (const MCPhysReg *Reg = MCID.getImplicitDefs(); *Reg; ++Reg) diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp index 523f409e6b2cb..439f67f1e155d 100644 --- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp +++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp @@ -36,6 +36,7 @@ #include "llvm/IR/Intrinsics.h" #include "llvm/Support/Debug.h" #include "llvm/Support/ErrorHandling.h" +#include "llvm/Support/KnownBits.h" #include "llvm/Support/ManagedStatic.h" #include "llvm/Support/MathExtras.h" #include "llvm/Support/Mutex.h" @@ -2868,7 +2869,7 @@ bool SelectionDAG::isKnownToBeAPowerOfTwo(SDValue Val) const { // A left-shift of a constant one will have exactly one bit set because // shifting the bit off the end is undefined. if (Val.getOpcode() == ISD::SHL) { - auto *C = dyn_cast<ConstantSDNode>(Val.getOperand(0)); + auto *C = isConstOrConstSplat(Val.getOperand(0)); if (C && C->getAPIntValue() == 1) return true; } @@ -2876,7 +2877,7 @@ bool SelectionDAG::isKnownToBeAPowerOfTwo(SDValue Val) const { // Similarly, a logical right-shift of a constant sign-bit will have exactly // one bit set. if (Val.getOpcode() == ISD::SRL) { - auto *C = dyn_cast<ConstantSDNode>(Val.getOperand(0)); + auto *C = isConstOrConstSplat(Val.getOperand(0)); if (C && C->getAPIntValue().isSignMask()) return true; } @@ -7539,10 +7540,10 @@ unsigned SelectionDAG::InferPtrAlignment(SDValue Ptr) const { int64_t GVOffset = 0; if (TLI->isGAPlusOffset(Ptr.getNode(), GV, GVOffset)) { unsigned PtrWidth = getDataLayout().getPointerTypeSizeInBits(GV->getType()); - APInt KnownZero(PtrWidth, 0), KnownOne(PtrWidth, 0); - llvm::computeKnownBits(const_cast<GlobalValue *>(GV), KnownZero, KnownOne, + KnownBits Known(PtrWidth); + llvm::computeKnownBits(const_cast<GlobalValue *>(GV), Known, getDataLayout()); - unsigned AlignBits = KnownZero.countTrailingOnes(); + unsigned AlignBits = Known.Zero.countTrailingOnes(); unsigned Align = AlignBits ? 1 << std::min(31U, AlignBits) : 0; if (Align) return MinAlign(Align, GVOffset); @@ -7629,52 +7630,52 @@ Type *ConstantPoolSDNode::getType() const { return Val.ConstVal->getType(); } -bool BuildVectorSDNode::isConstantSplat(APInt &SplatValue, - APInt &SplatUndef, +bool BuildVectorSDNode::isConstantSplat(APInt &SplatValue, APInt &SplatUndef, unsigned &SplatBitSize, bool &HasAnyUndefs, unsigned MinSplatBits, - bool isBigEndian) const { + bool IsBigEndian) const { EVT VT = getValueType(0); assert(VT.isVector() && "Expected a vector type"); - unsigned sz = VT.getSizeInBits(); - if (MinSplatBits > sz) + unsigned VecWidth = VT.getSizeInBits(); + if (MinSplatBits > VecWidth) return false; - SplatValue = APInt(sz, 0); - SplatUndef = APInt(sz, 0); + // FIXME: The widths are based on this node's type, but build vectors can + // truncate their operands. + SplatValue = APInt(VecWidth, 0); + SplatUndef = APInt(VecWidth, 0); - // Get the bits. Bits with undefined values (when the corresponding element + // Get the bits. Bits with undefined values (when the corresponding element // of the vector is an ISD::UNDEF value) are set in SplatUndef and cleared - // in SplatValue. If any of the values are not constant, give up and return + // in SplatValue. If any of the values are not constant, give up and return // false. - unsigned int nOps = getNumOperands(); - assert(nOps > 0 && "isConstantSplat has 0-size build vector"); - unsigned EltBitSize = VT.getScalarSizeInBits(); + unsigned int NumOps = getNumOperands(); + assert(NumOps > 0 && "isConstantSplat has 0-size build vector"); + unsigned EltWidth = VT.getScalarSizeInBits(); - for (unsigned j = 0; j < nOps; ++j) { - unsigned i = isBigEndian ? nOps-1-j : j; + for (unsigned j = 0; j < NumOps; ++j) { + unsigned i = IsBigEndian ? NumOps - 1 - j : j; SDValue OpVal = getOperand(i); - unsigned BitPos = j * EltBitSize; + unsigned BitPos = j * EltWidth; if (OpVal.isUndef()) - SplatUndef.setBits(BitPos, BitPos + EltBitSize); - else if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(OpVal)) - SplatValue.insertBits(CN->getAPIntValue().zextOrTrunc(EltBitSize), - BitPos); - else if (ConstantFPSDNode *CN = dyn_cast<ConstantFPSDNode>(OpVal)) + SplatUndef.setBits(BitPos, BitPos + EltWidth); + else if (auto *CN = dyn_cast<ConstantSDNode>(OpVal)) + SplatValue.insertBits(CN->getAPIntValue().zextOrTrunc(EltWidth), BitPos); + else if (auto *CN = dyn_cast<ConstantFPSDNode>(OpVal)) SplatValue.insertBits(CN->getValueAPF().bitcastToAPInt(), BitPos); else return false; } - // The build_vector is all constants or undefs. Find the smallest element + // The build_vector is all constants or undefs. Find the smallest element // size that splats the vector. - HasAnyUndefs = (SplatUndef != 0); - while (sz > 8) { - unsigned HalfSize = sz / 2; + // FIXME: This does not work for vectors with elements less than 8 bits. + while (VecWidth > 8) { + unsigned HalfSize = VecWidth / 2; APInt HighValue = SplatValue.lshr(HalfSize).trunc(HalfSize); APInt LowValue = SplatValue.trunc(HalfSize); APInt HighUndef = SplatUndef.lshr(HalfSize).trunc(HalfSize); @@ -7688,10 +7689,10 @@ bool BuildVectorSDNode::isConstantSplat(APInt &SplatValue, SplatValue = HighValue | LowValue; SplatUndef = HighUndef & LowUndef; - sz = HalfSize; + VecWidth = HalfSize; } - SplatBitSize = sz; + SplatBitSize = VecWidth; return true; } diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp index 2c58953ee9089..6a737ed84ea43 100644 --- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp +++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp @@ -362,11 +362,11 @@ static SDValue getCopyFromPartsVector(SelectionDAG &DAG, const SDLoc &DL, return DAG.getUNDEF(ValueVT); } - if (ValueVT.getVectorNumElements() == 1 && - ValueVT.getVectorElementType() != PartEVT) - Val = DAG.getAnyExtOrTrunc(Val, DL, ValueVT.getScalarType()); + EVT ValueSVT = ValueVT.getVectorElementType(); + if (ValueVT.getVectorNumElements() == 1 && ValueSVT != PartEVT) + Val = DAG.getAnyExtOrTrunc(Val, DL, ValueSVT); - return DAG.getNode(ISD::BUILD_VECTOR, DL, ValueVT, Val); + return DAG.getBuildVector(ValueVT, DL, Val); } static void getCopyToPartsVector(SelectionDAG &DAG, const SDLoc &dl, @@ -537,7 +537,7 @@ static void getCopyToPartsVector(SelectionDAG &DAG, const SDLoc &DL, e = PartVT.getVectorNumElements(); i != e; ++i) Ops.push_back(DAG.getUNDEF(ElementVT)); - Val = DAG.getNode(ISD::BUILD_VECTOR, DL, PartVT, Ops); + Val = DAG.getBuildVector(PartVT, DL, Ops); // FIXME: Use CONCAT for 2x -> 4x. @@ -1088,8 +1088,7 @@ SDValue SelectionDAGBuilder::getValueImpl(const Value *V) { if (isa<ArrayType>(CDS->getType())) return DAG.getMergeValues(Ops, getCurSDLoc()); - return NodeMap[V] = DAG.getNode(ISD::BUILD_VECTOR, getCurSDLoc(), - VT, Ops); + return NodeMap[V] = DAG.getBuildVector(VT, getCurSDLoc(), Ops); } if (C->getType()->isStructTy() || C->getType()->isArrayTy()) { @@ -1141,7 +1140,7 @@ SDValue SelectionDAGBuilder::getValueImpl(const Value *V) { } // Create a BUILD_VECTOR node. - return NodeMap[V] = DAG.getNode(ISD::BUILD_VECTOR, getCurSDLoc(), VT, Ops); + return NodeMap[V] = DAG.getBuildVector(VT, getCurSDLoc(), Ops); } // If this is a static alloca, generate it as the frameindex instead of @@ -3147,7 +3146,7 @@ void SelectionDAGBuilder::visitShuffleVector(const User &I) { Ops.push_back(Res); } - setValue(&I, DAG.getNode(ISD::BUILD_VECTOR, DL, VT, Ops)); + setValue(&I, DAG.getBuildVector(VT, DL, Ops)); } void SelectionDAGBuilder::visitInsertValue(const InsertValueInst &I) { @@ -3969,9 +3968,9 @@ void SelectionDAGBuilder::visitFence(const FenceInst &I) { SDValue Ops[3]; Ops[0] = getRoot(); Ops[1] = DAG.getConstant((unsigned)I.getOrdering(), dl, - TLI.getPointerTy(DAG.getDataLayout())); + TLI.getFenceOperandTy(DAG.getDataLayout())); Ops[2] = DAG.getConstant(I.getSynchScope(), dl, - TLI.getPointerTy(DAG.getDataLayout())); + TLI.getFenceOperandTy(DAG.getDataLayout())); DAG.setRoot(DAG.getNode(ISD::ATOMIC_FENCE, dl, MVT::Other, Ops)); } @@ -4896,11 +4895,11 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) { Entry.Node = Src; Args.push_back(Entry); - + Entry.Ty = I.getArgOperand(2)->getType(); Entry.Node = NumElements; Args.push_back(Entry); - + Entry.Ty = Type::getInt32Ty(*DAG.getContext()); Entry.Node = ElementSize; Args.push_back(Entry); @@ -5183,7 +5182,7 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) { SDValue ShOps[2]; ShOps[0] = ShAmt; ShOps[1] = DAG.getConstant(0, sdl, MVT::i32); - ShAmt = DAG.getNode(ISD::BUILD_VECTOR, sdl, ShAmtVT, ShOps); + ShAmt = DAG.getBuildVector(ShAmtVT, sdl, ShOps); EVT DestVT = TLI.getValueType(DAG.getDataLayout(), I.getType()); ShAmt = DAG.getNode(ISD::BITCAST, sdl, DestVT, ShAmt); Res = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, sdl, DestVT, @@ -5743,7 +5742,7 @@ void SelectionDAGBuilder::visitConstrainedFPIntrinsic(const CallInst &I, unsigned Opcode; switch (Intrinsic) { default: llvm_unreachable("Impossible intrinsic"); // Can't reach here. - case Intrinsic::experimental_constrained_fadd: + case Intrinsic::experimental_constrained_fadd: Opcode = ISD::STRICT_FADD; break; case Intrinsic::experimental_constrained_fsub: @@ -6653,12 +6652,12 @@ static void GetRegistersForValue(SelectionDAG &DAG, const TargetLowering &TLI, MachineFunction &MF = DAG.getMachineFunction(); SmallVector<unsigned, 4> Regs; + const TargetRegisterInfo &TRI = *MF.getSubtarget().getRegisterInfo(); // If this is a constraint for a single physreg, or a constraint for a // register class, find it. std::pair<unsigned, const TargetRegisterClass *> PhysReg = - TLI.getRegForInlineAsmConstraint(MF.getSubtarget().getRegisterInfo(), - OpInfo.ConstraintCode, + TLI.getRegForInlineAsmConstraint(&TRI, OpInfo.ConstraintCode, OpInfo.ConstraintVT); unsigned NumRegs = 1; @@ -6666,12 +6665,12 @@ static void GetRegistersForValue(SelectionDAG &DAG, const TargetLowering &TLI, // If this is a FP input in an integer register (or visa versa) insert a bit // cast of the input value. More generally, handle any case where the input // value disagrees with the register class we plan to stick this in. - if (OpInfo.Type == InlineAsm::isInput && - PhysReg.second && !PhysReg.second->hasType(OpInfo.ConstraintVT)) { + if (OpInfo.Type == InlineAsm::isInput && PhysReg.second && + !TRI.isTypeLegalForClass(*PhysReg.second, OpInfo.ConstraintVT)) { // Try to convert to the first EVT that the reg class contains. If the // types are identical size, use a bitcast to convert (e.g. two differing // vector types). - MVT RegVT = *PhysReg.second->vt_begin(); + MVT RegVT = *TRI.legalclasstypes_begin(*PhysReg.second); if (RegVT.getSizeInBits() == OpInfo.CallOperand.getValueSizeInBits()) { OpInfo.CallOperand = DAG.getNode(ISD::BITCAST, DL, RegVT, OpInfo.CallOperand); @@ -6699,12 +6698,12 @@ static void GetRegistersForValue(SelectionDAG &DAG, const TargetLowering &TLI, if (unsigned AssignedReg = PhysReg.first) { const TargetRegisterClass *RC = PhysReg.second; if (OpInfo.ConstraintVT == MVT::Other) - ValueVT = *RC->vt_begin(); + ValueVT = *TRI.legalclasstypes_begin(*RC); // Get the actual register value type. This is important, because the user // may have asked for (e.g.) the AX register in i32 type. We need to // remember that AX is actually i16 to get the right extension. - RegVT = *RC->vt_begin(); + RegVT = *TRI.legalclasstypes_begin(*RC); // This is a explicit reference to a physical register. Regs.push_back(AssignedReg); @@ -6730,7 +6729,7 @@ static void GetRegistersForValue(SelectionDAG &DAG, const TargetLowering &TLI, // Otherwise, if this was a reference to an LLVM register class, create vregs // for this reference. if (const TargetRegisterClass *RC = PhysReg.second) { - RegVT = *RC->vt_begin(); + RegVT = *TRI.legalclasstypes_begin(*RC); if (OpInfo.ConstraintVT == MVT::Other) ValueVT = RegVT; diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp index 93c6738f650d4..136dec873cb8f 100644 --- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp +++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp @@ -342,11 +342,16 @@ TargetLowering::isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const { /// If the specified instruction has a constant integer operand and there are /// bits set in that constant that are not demanded, then clear those bits and /// return true. -bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant( - SDValue Op, const APInt &Demanded) { +bool TargetLowering::ShrinkDemandedConstant(SDValue Op, const APInt &Demanded, + TargetLoweringOpt &TLO) const { + SelectionDAG &DAG = TLO.DAG; SDLoc DL(Op); unsigned Opcode = Op.getOpcode(); + // Do target-specific constant optimization. + if (targetShrinkDemandedConstant(Op, Demanded, TLO)) + return TLO.New.getNode(); + // FIXME: ISD::SELECT, ISD::SELECT_CC switch (Opcode) { default: @@ -367,7 +372,7 @@ bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant( EVT VT = Op.getValueType(); SDValue NewC = DAG.getConstant(Demanded & C, DL, VT); SDValue NewOp = DAG.getNode(Opcode, DL, VT, Op.getOperand(0), NewC); - return CombineTo(Op, NewOp); + return TLO.CombineTo(Op, NewOp); } break; @@ -380,15 +385,17 @@ bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant( /// Convert x+y to (VT)((SmallVT)x+(SmallVT)y) if the casts are free. /// This uses isZExtFree and ZERO_EXTEND for the widening cast, but it could be /// generalized for targets with other types of implicit widening casts. -bool TargetLowering::TargetLoweringOpt::ShrinkDemandedOp(SDValue Op, - unsigned BitWidth, - const APInt &Demanded, - const SDLoc &dl) { +bool TargetLowering::ShrinkDemandedOp(SDValue Op, unsigned BitWidth, + const APInt &Demanded, + TargetLoweringOpt &TLO) const { assert(Op.getNumOperands() == 2 && "ShrinkDemandedOp only supports binary operators!"); assert(Op.getNode()->getNumValues() == 1 && "ShrinkDemandedOp only supports nodes with one result!"); + SelectionDAG &DAG = TLO.DAG; + SDLoc dl(Op); + // Early return, as this function cannot handle vector types. if (Op.getValueType().isVector()) return false; @@ -418,23 +425,22 @@ bool TargetLowering::TargetLoweringOpt::ShrinkDemandedOp(SDValue Op, bool NeedZext = DemandedSize > SmallVTBits; SDValue Z = DAG.getNode(NeedZext ? ISD::ZERO_EXTEND : ISD::ANY_EXTEND, dl, Op.getValueType(), X); - return CombineTo(Op, Z); + return TLO.CombineTo(Op, Z); } } return false; } bool -TargetLowering::TargetLoweringOpt::SimplifyDemandedBits(SDNode *User, - unsigned OpIdx, - const APInt &Demanded, - DAGCombinerInfo &DCI) { - const TargetLowering &TLI = DAG.getTargetLoweringInfo(); +TargetLowering::SimplifyDemandedBits(SDNode *User, unsigned OpIdx, + const APInt &Demanded, + DAGCombinerInfo &DCI, + TargetLoweringOpt &TLO) const { SDValue Op = User->getOperand(OpIdx); APInt KnownZero, KnownOne; - if (!TLI.SimplifyDemandedBits(Op, Demanded, KnownZero, KnownOne, - *this, 0, true)) + if (!SimplifyDemandedBits(Op, Demanded, KnownZero, KnownOne, + TLO, 0, true)) return false; @@ -446,9 +452,9 @@ TargetLowering::TargetLoweringOpt::SimplifyDemandedBits(SDNode *User, // with the value 'x', which will give us: // Old = i32 and x, 0xffffff // New = x - if (Old.hasOneUse()) { + if (TLO.Old.hasOneUse()) { // For the one use case, we just commit the change. - DCI.CommitTargetLoweringOpt(*this); + DCI.CommitTargetLoweringOpt(TLO); return true; } @@ -456,17 +462,17 @@ TargetLowering::TargetLoweringOpt::SimplifyDemandedBits(SDNode *User, // AssumeSingleUse flag is not propogated to recursive calls of // SimplifyDemanded bits, so the only node with multiple use that // it will attempt to combine will be opt. - assert(Old == Op); + assert(TLO.Old == Op); SmallVector <SDValue, 4> NewOps; for (unsigned i = 0, e = User->getNumOperands(); i != e; ++i) { if (i == OpIdx) { - NewOps.push_back(New); + NewOps.push_back(TLO.New); continue; } NewOps.push_back(User->getOperand(i)); } - DAG.UpdateNodeOperands(User, NewOps); + TLO.DAG.UpdateNodeOperands(User, NewOps); // Op has less users now, so we may be able to perform additional combines // with it. DCI.AddToWorklist(Op.getNode()); @@ -585,7 +591,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, // If any of the set bits in the RHS are known zero on the LHS, shrink // the constant. - if (TLO.ShrinkDemandedConstant(Op, ~LHSZero & NewMask)) + if (ShrinkDemandedConstant(Op, ~LHSZero & NewMask, TLO)) return true; // Bitwise-not (xor X, -1) is a special case: we don't usually shrink its @@ -620,10 +626,10 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, if ((NewMask & (KnownZero|KnownZero2)) == NewMask) return TLO.CombineTo(Op, TLO.DAG.getConstant(0, dl, Op.getValueType())); // If the RHS is a constant, see if we can simplify it. - if (TLO.ShrinkDemandedConstant(Op, ~KnownZero2 & NewMask)) + if (ShrinkDemandedConstant(Op, ~KnownZero2 & NewMask, TLO)) return true; // If the operation can be done in a smaller type, do so. - if (TLO.ShrinkDemandedOp(Op, BitWidth, NewMask, dl)) + if (ShrinkDemandedOp(Op, BitWidth, NewMask, TLO)) return true; // Output known-1 bits are only known if set in both the LHS & RHS. @@ -654,10 +660,10 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, if ((NewMask & ~KnownZero2 & KnownOne) == (~KnownZero2 & NewMask)) return TLO.CombineTo(Op, Op.getOperand(1)); // If the RHS is a constant, see if we can simplify it. - if (TLO.ShrinkDemandedConstant(Op, NewMask)) + if (ShrinkDemandedConstant(Op, NewMask, TLO)) return true; // If the operation can be done in a smaller type, do so. - if (TLO.ShrinkDemandedOp(Op, BitWidth, NewMask, dl)) + if (ShrinkDemandedOp(Op, BitWidth, NewMask, TLO)) return true; // Output known-0 bits are only known if clear in both the LHS & RHS. @@ -682,7 +688,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, if ((KnownZero2 & NewMask) == NewMask) return TLO.CombineTo(Op, Op.getOperand(1)); // If the operation can be done in a smaller type, do so. - if (TLO.ShrinkDemandedOp(Op, BitWidth, NewMask, dl)) + if (ShrinkDemandedOp(Op, BitWidth, NewMask, TLO)) return true; // If all of the unknown bits are known to be zero on one side or the other @@ -727,7 +733,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, } // If it already has all the bits set, nothing to change // but don't shrink either! - } else if (TLO.ShrinkDemandedConstant(Op, NewMask)) { + } else if (ShrinkDemandedConstant(Op, NewMask, TLO)) { return true; } } @@ -746,7 +752,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If the operands are constants, see if we can simplify them. - if (TLO.ShrinkDemandedConstant(Op, NewMask)) + if (ShrinkDemandedConstant(Op, NewMask, TLO)) return true; // Only known if known in both the LHS and RHS. @@ -764,7 +770,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If the operands are constants, see if we can simplify them. - if (TLO.ShrinkDemandedConstant(Op, NewMask)) + if (ShrinkDemandedConstant(Op, NewMask, TLO)) return true; // Only known if known in both the LHS and RHS. @@ -1284,7 +1290,7 @@ bool TargetLowering::SimplifyDemandedBits(SDValue Op, SimplifyDemandedBits(Op.getOperand(1), LoMask, KnownZero2, KnownOne2, TLO, Depth+1) || // See if the operation should be performed at a smaller bit width. - TLO.ShrinkDemandedOp(Op, BitWidth, NewMask, dl)) { + ShrinkDemandedOp(Op, BitWidth, NewMask, TLO)) { const SDNodeFlags *Flags = Op.getNode()->getFlags(); if (Flags->hasNoSignedWrap() || Flags->hasNoUnsignedWrap()) { // Disable the nsw and nuw flags. We can no longer guarantee that we @@ -1358,31 +1364,38 @@ unsigned TargetLowering::ComputeNumSignBitsForTargetNode(SDValue Op, return 1; } +// FIXME: Ideally, this would use ISD::isConstantSplatVector(), but that must +// work with truncating build vectors and vectors with elements of less than +// 8 bits. bool TargetLowering::isConstTrueVal(const SDNode *N) const { if (!N) return false; - const ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N); - if (!CN) { - const BuildVectorSDNode *BV = dyn_cast<BuildVectorSDNode>(N); - if (!BV) - return false; - - // Only interested in constant splats, we don't care about undef - // elements in identifying boolean constants and getConstantSplatNode - // returns NULL if all ops are undef; - CN = BV->getConstantSplatNode(); + APInt CVal; + if (auto *CN = dyn_cast<ConstantSDNode>(N)) { + CVal = CN->getAPIntValue(); + } else if (auto *BV = dyn_cast<BuildVectorSDNode>(N)) { + auto *CN = BV->getConstantSplatNode(); if (!CN) return false; + + // If this is a truncating build vector, truncate the splat value. + // Otherwise, we may fail to match the expected values below. + unsigned BVEltWidth = BV->getValueType(0).getScalarSizeInBits(); + CVal = CN->getAPIntValue(); + if (BVEltWidth < CVal.getBitWidth()) + CVal = CVal.trunc(BVEltWidth); + } else { + return false; } switch (getBooleanContents(N->getValueType(0))) { case UndefinedBooleanContent: - return CN->getAPIntValue()[0]; + return CVal[0]; case ZeroOrOneBooleanContent: - return CN->isOne(); + return CVal == 1; case ZeroOrNegativeOneBooleanContent: - return CN->isAllOnesValue(); + return CVal.isAllOnesValue(); } llvm_unreachable("Invalid boolean contents"); @@ -2535,7 +2548,7 @@ TargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *RI, for (const TargetRegisterClass *RC : RI->regclasses()) { // If none of the value types for this register class are valid, we // can't use it. For example, 64-bit reg classes on 32-bit targets. - if (!isLegalRC(RC)) + if (!isLegalRC(*RI, *RC)) continue; for (TargetRegisterClass::iterator I = RC->begin(), E = RC->end(); @@ -2547,9 +2560,9 @@ TargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *RI, // If this register class has the requested value type, return it, // otherwise keep searching and return the first class found // if no other is found which explicitly has the requested type. - if (RC->hasType(VT)) + if (RI->isTypeLegalForClass(*RC, VT)) return S; - else if (!R.second) + if (!R.second) R = S; } } diff --git a/lib/CodeGen/StackMaps.cpp b/lib/CodeGen/StackMaps.cpp index 1a8ec5bff3229..315b059c5ac91 100644 --- a/lib/CodeGen/StackMaps.cpp +++ b/lib/CodeGen/StackMaps.cpp @@ -161,7 +161,8 @@ StackMaps::parseOperand(MachineInstr::const_mop_iterator MOI, if (SubRegIdx) Offset = TRI->getSubRegIdxOffset(SubRegIdx); - Locs.emplace_back(Location::Register, RC->getSize(), DwarfRegNum, Offset); + Locs.emplace_back(Location::Register, TRI->getSpillSize(*RC), + DwarfRegNum, Offset); return ++MOI; } @@ -245,7 +246,7 @@ void StackMaps::print(raw_ostream &OS) { StackMaps::LiveOutReg StackMaps::createLiveOutReg(unsigned Reg, const TargetRegisterInfo *TRI) const { unsigned DwarfRegNum = getDwarfRegNum(Reg, TRI); - unsigned Size = TRI->getMinimalPhysRegClass(Reg)->getSize(); + unsigned Size = TRI->getSpillSize(*TRI->getMinimalPhysRegClass(Reg)); return LiveOutReg(Reg, DwarfRegNum, Size); } diff --git a/lib/CodeGen/TargetInstrInfo.cpp b/lib/CodeGen/TargetInstrInfo.cpp index 711144a347430..14c5adc0d898b 100644 --- a/lib/CodeGen/TargetInstrInfo.cpp +++ b/lib/CodeGen/TargetInstrInfo.cpp @@ -345,12 +345,12 @@ bool TargetInstrInfo::getStackSlotRange(const TargetRegisterClass *RC, unsigned SubIdx, unsigned &Size, unsigned &Offset, const MachineFunction &MF) const { + const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo(); if (!SubIdx) { - Size = RC->getSize(); + Size = TRI->getSpillSize(*RC); Offset = 0; return true; } - const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo(); unsigned BitSize = TRI->getSubRegIdxSize(SubIdx); // Convert bit size to byte size to be consistent with // MCRegisterClass::getSize(). @@ -364,10 +364,10 @@ bool TargetInstrInfo::getStackSlotRange(const TargetRegisterClass *RC, Size = BitSize /= 8; Offset = (unsigned)BitOffset / 8; - assert(RC->getSize() >= (Offset + Size) && "bad subregister range"); + assert(TRI->getSpillSize(*RC) >= (Offset + Size) && "bad subregister range"); if (!MF.getDataLayout().isLittleEndian()) { - Offset = RC->getSize() - (Offset + Size); + Offset = TRI->getSpillSize(*RC) - (Offset + Size); } return true; } @@ -428,8 +428,8 @@ static const TargetRegisterClass *canFoldCopy(const MachineInstr &MI, return nullptr; } -void TargetInstrInfo::getNoopForMachoTarget(MCInst &NopInst) const { - llvm_unreachable("Not a MachO target"); +void TargetInstrInfo::getNoop(MCInst &NopInst) const { + llvm_unreachable("Not implemented"); } static MachineInstr *foldPatchpoint(MachineFunction &MF, MachineInstr &MI, diff --git a/lib/CodeGen/TargetLoweringBase.cpp b/lib/CodeGen/TargetLoweringBase.cpp index 27630a3055cb3..e579922bb69e7 100644 --- a/lib/CodeGen/TargetLoweringBase.cpp +++ b/lib/CodeGen/TargetLoweringBase.cpp @@ -1184,12 +1184,11 @@ static unsigned getVectorTypeBreakdownMVT(MVT VT, MVT &IntermediateVT, /// isLegalRC - Return true if the value types that can be represented by the /// specified register class are all legal. -bool TargetLoweringBase::isLegalRC(const TargetRegisterClass *RC) const { - for (TargetRegisterClass::vt_iterator I = RC->vt_begin(), E = RC->vt_end(); - I != E; ++I) { +bool TargetLoweringBase::isLegalRC(const TargetRegisterInfo &TRI, + const TargetRegisterClass &RC) const { + for (auto I = TRI.legalclasstypes_begin(RC); *I != MVT::Other; ++I) if (isTypeLegal(*I)) return true; - } return false; } @@ -1299,9 +1298,9 @@ TargetLoweringBase::findRepresentativeClass(const TargetRegisterInfo *TRI, for (int i = SuperRegRC.find_first(); i >= 0; i = SuperRegRC.find_next(i)) { const TargetRegisterClass *SuperRC = TRI->getRegClass(i); // We want the largest possible spill size. - if (SuperRC->getSize() <= BestRC->getSize()) + if (TRI->getSpillSize(*SuperRC) <= TRI->getSpillSize(*BestRC)) continue; - if (!isLegalRC(SuperRC)) + if (!isLegalRC(*TRI, *SuperRC)) continue; BestRC = SuperRC; } diff --git a/lib/CodeGen/TargetRegisterInfo.cpp b/lib/CodeGen/TargetRegisterInfo.cpp index 66cdad278e8da..f6e4c17d514cd 100644 --- a/lib/CodeGen/TargetRegisterInfo.cpp +++ b/lib/CodeGen/TargetRegisterInfo.cpp @@ -156,8 +156,8 @@ TargetRegisterInfo::getMinimalPhysRegClass(unsigned reg, MVT VT) const { // this physreg. const TargetRegisterClass* BestRC = nullptr; for (const TargetRegisterClass* RC : regclasses()) { - if ((VT == MVT::Other || RC->hasType(VT)) && RC->contains(reg) && - (!BestRC || BestRC->hasSubClass(RC))) + if ((VT == MVT::Other || isTypeLegalForClass(*RC, VT)) && + RC->contains(reg) && (!BestRC || BestRC->hasSubClass(RC))) BestRC = RC; } @@ -207,7 +207,7 @@ const TargetRegisterClass *firstCommonClass(const uint32_t *A, if (unsigned Common = *A++ & *B++) { const TargetRegisterClass *RC = TRI->getRegClass(I + countTrailingZeros(Common)); - if (SVT == MVT::SimpleValueType::Any || RC->hasType(VT)) + if (SVT == MVT::SimpleValueType::Any || TRI->isTypeLegalForClass(*RC, VT)) return RC; } return nullptr; @@ -265,7 +265,7 @@ getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA, const TargetRegisterClass *BestRC = nullptr; unsigned *BestPreA = &PreA; unsigned *BestPreB = &PreB; - if (RCA->getSize() < RCB->getSize()) { + if (getRegSizeInBits(*RCA) < getRegSizeInBits(*RCB)) { std::swap(RCA, RCB); std::swap(SubA, SubB); std::swap(BestPreA, BestPreB); @@ -273,7 +273,7 @@ getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA, // Also terminate the search one we have found a register class as small as // RCA. - unsigned MinSize = RCA->getSize(); + unsigned MinSize = getRegSizeInBits(*RCA); for (SuperRegClassIterator IA(RCA, this, true); IA.isValid(); ++IA) { unsigned FinalA = composeSubRegIndices(IA.getSubReg(), SubA); @@ -281,7 +281,7 @@ getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA, // Check if a common super-register class exists for this index pair. const TargetRegisterClass *RC = firstCommonClass(IA.getMask(), IB.getMask(), this); - if (!RC || RC->getSize() < MinSize) + if (!RC || getRegSizeInBits(*RC) < MinSize) continue; // The indexes must compose identically: PreA+SubA == PreB+SubB. @@ -290,7 +290,7 @@ getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA, continue; // Is RC a better candidate than BestRC? - if (BestRC && RC->getSize() >= BestRC->getSize()) + if (BestRC && getRegSizeInBits(*RC) >= getRegSizeInBits(*BestRC)) continue; // Yes, RC is the smallest super-register seen so far. @@ -299,7 +299,7 @@ getCommonSuperRegClass(const TargetRegisterClass *RCA, unsigned SubA, *BestPreB = IB.getSubReg(); // Bail early if we reached MinSize. We won't find a better candidate. - if (BestRC->getSize() == MinSize) + if (getRegSizeInBits(*BestRC) == MinSize) return BestRC; } } diff --git a/lib/CodeGen/VirtRegMap.cpp b/lib/CodeGen/VirtRegMap.cpp index c8946010e9d15..d10ca1a7ff918 100644 --- a/lib/CodeGen/VirtRegMap.cpp +++ b/lib/CodeGen/VirtRegMap.cpp @@ -73,8 +73,9 @@ void VirtRegMap::grow() { } unsigned VirtRegMap::createSpillSlot(const TargetRegisterClass *RC) { - int SS = MF->getFrameInfo().CreateSpillStackObject(RC->getSize(), - RC->getAlignment()); + unsigned Size = TRI->getSpillSize(*RC); + unsigned Align = TRI->getSpillAlignment(*RC); + int SS = MF->getFrameInfo().CreateSpillStackObject(Size, Align); ++NumSpillSlots; return SS; } |