X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FMC%2FMCAssembler.cpp;h=18687cdd6a2b17768463f95c560acbcafb8e52bf;hp=457abd2a0aaa47aa991848ae442e5c84647fd7c0;hb=19696daa2159e059b9a15cda222e97d2b91b279d;hpb=5c10f509f45820d1198bfb975840e93a782745ac diff --git a/lib/MC/MCAssembler.cpp b/lib/MC/MCAssembler.cpp index 457abd2a0aa..18687cdd6a2 100644 --- a/lib/MC/MCAssembler.cpp +++ b/lib/MC/MCAssembler.cpp @@ -7,12 +7,12 @@ // //===----------------------------------------------------------------------===// -#define DEBUG_TYPE "assembler" #include "llvm/MC/MCAssembler.h" #include "llvm/ADT/Statistic.h" #include "llvm/ADT/StringExtras.h" #include "llvm/ADT/Twine.h" #include "llvm/MC/MCAsmBackend.h" +#include "llvm/MC/MCAsmInfo.h" #include "llvm/MC/MCAsmLayout.h" #include "llvm/MC/MCCodeEmitter.h" #include "llvm/MC/MCContext.h" @@ -21,6 +21,7 @@ #include "llvm/MC/MCFixupKindInfo.h" #include "llvm/MC/MCObjectWriter.h" #include "llvm/MC/MCSection.h" +#include "llvm/MC/MCSectionELF.h" #include "llvm/MC/MCSymbol.h" #include "llvm/MC/MCValue.h" #include "llvm/Support/Debug.h" @@ -28,12 +29,26 @@ #include "llvm/Support/LEB128.h" #include "llvm/Support/TargetRegistry.h" #include "llvm/Support/raw_ostream.h" - +#include using namespace llvm; +#define DEBUG_TYPE "assembler" + namespace { namespace stats { -STATISTIC(EmittedFragments, "Number of emitted assembler fragments"); +STATISTIC(EmittedFragments, "Number of emitted assembler fragments - total"); +STATISTIC(EmittedRelaxableFragments, + "Number of emitted assembler fragments - relaxable"); +STATISTIC(EmittedDataFragments, + "Number of emitted assembler fragments - data"); +STATISTIC(EmittedCompactEncodedInstFragments, + "Number of emitted assembler fragments - compact encoded inst"); +STATISTIC(EmittedAlignFragments, + "Number of emitted assembler fragments - align"); +STATISTIC(EmittedFillFragments, + "Number of emitted assembler fragments - fill"); +STATISTIC(EmittedOrgFragments, + "Number of emitted assembler fragments - org"); STATISTIC(evaluateFixup, "Number of evaluated fixups"); STATISTIC(FragmentLayouts, "Number of fragment layouts"); STATISTIC(ObjectBytes, "Number of emitted object file bytes"); @@ -61,7 +76,7 @@ MCAsmLayout::MCAsmLayout(MCAssembler &Asm) SectionOrder.push_back(&*it); } -bool MCAsmLayout::isFragmentUpToDate(const MCFragment *F) const { +bool MCAsmLayout::isFragmentValid(const MCFragment *F) const { const MCSectionData &SD = *F->getParent(); const MCFragment *LastValid = LastValidFragment.lookup(&SD); if (!LastValid) @@ -70,17 +85,18 @@ bool MCAsmLayout::isFragmentUpToDate(const MCFragment *F) const { return F->getLayoutOrder() <= LastValid->getLayoutOrder(); } -void MCAsmLayout::Invalidate(MCFragment *F) { - // If this fragment wasn't already up-to-date, we don't need to do anything. - if (!isFragmentUpToDate(F)) +void MCAsmLayout::invalidateFragmentsFrom(MCFragment *F) { + // If this fragment wasn't already valid, we don't need to do anything. + if (!isFragmentValid(F)) return; - // Otherwise, reset the last valid fragment to this fragment. + // Otherwise, reset the last valid fragment to the previous fragment + // (if this is the first fragment, it will be NULL). const MCSectionData &SD = *F->getParent(); - LastValidFragment[&SD] = F; + LastValidFragment[&SD] = F->getPrevNode(); } -void MCAsmLayout::EnsureValid(const MCFragment *F) const { +void MCAsmLayout::ensureValid(const MCFragment *F) const { MCSectionData &SD = *F->getParent(); MCFragment *Cur = LastValidFragment[&SD]; @@ -89,49 +105,113 @@ void MCAsmLayout::EnsureValid(const MCFragment *F) const { else Cur = Cur->getNextNode(); - // Advance the layout position until the fragment is up-to-date. - while (!isFragmentUpToDate(F)) { - const_cast(this)->LayoutFragment(Cur); + // Advance the layout position until the fragment is valid. + while (!isFragmentValid(F)) { + assert(Cur && "Layout bookkeeping error"); + const_cast(this)->layoutFragment(Cur); Cur = Cur->getNextNode(); } } uint64_t MCAsmLayout::getFragmentOffset(const MCFragment *F) const { - EnsureValid(F); + ensureValid(F); assert(F->Offset != ~UINT64_C(0) && "Address not set!"); return F->Offset; } -uint64_t MCAsmLayout::getSymbolOffset(const MCSymbolData *SD) const { +// Simple getSymbolOffset helper for the non-varibale case. +static bool getLabelOffset(const MCAsmLayout &Layout, const MCSymbolData &SD, + bool ReportError, uint64_t &Val) { + if (!SD.getFragment()) { + if (ReportError) + report_fatal_error("unable to evaluate offset to undefined symbol '" + + SD.getSymbol().getName() + "'"); + return false; + } + Val = Layout.getFragmentOffset(SD.getFragment()) + SD.getOffset(); + return true; +} + +static bool getSymbolOffsetImpl(const MCAsmLayout &Layout, + const MCSymbolData *SD, bool ReportError, + uint64_t &Val) { const MCSymbol &S = SD->getSymbol(); - // If this is a variable, then recursively evaluate now. - if (S.isVariable()) { - MCValue Target; - if (!S.getVariableValue()->EvaluateAsRelocatable(Target, *this)) - report_fatal_error("unable to evaluate offset for variable '" + - S.getName() + "'"); + if (!S.isVariable()) + return getLabelOffset(Layout, *SD, ReportError, Val); - // Verify that any used symbols are defined. - if (Target.getSymA() && Target.getSymA()->getSymbol().isUndefined()) - report_fatal_error("unable to evaluate offset to undefined symbol '" + - Target.getSymA()->getSymbol().getName() + "'"); - if (Target.getSymB() && Target.getSymB()->getSymbol().isUndefined()) - report_fatal_error("unable to evaluate offset to undefined symbol '" + - Target.getSymB()->getSymbol().getName() + "'"); + // If SD is a variable, evaluate it. + MCValue Target; + if (!S.getVariableValue()->EvaluateAsRelocatable(Target, &Layout, nullptr)) + report_fatal_error("unable to evaluate offset for variable '" + + S.getName() + "'"); + + uint64_t Offset = Target.getConstant(); + + const MCAssembler &Asm = Layout.getAssembler(); + + const MCSymbolRefExpr *A = Target.getSymA(); + if (A) { + uint64_t ValA; + if (!getLabelOffset(Layout, Asm.getSymbolData(A->getSymbol()), ReportError, + ValA)) + return false; + Offset += ValA; + } - uint64_t Offset = Target.getConstant(); - if (Target.getSymA()) - Offset += getSymbolOffset(&Assembler.getSymbolData( - Target.getSymA()->getSymbol())); - if (Target.getSymB()) - Offset -= getSymbolOffset(&Assembler.getSymbolData( - Target.getSymB()->getSymbol())); - return Offset; + const MCSymbolRefExpr *B = Target.getSymB(); + if (B) { + uint64_t ValB; + if (!getLabelOffset(Layout, Asm.getSymbolData(B->getSymbol()), ReportError, + ValB)) + return false; + Offset -= ValB; } - assert(SD->getFragment() && "Invalid getOffset() on undefined symbol!"); - return getFragmentOffset(SD->getFragment()) + SD->getOffset(); + Val = Offset; + return true; +} + +bool MCAsmLayout::getSymbolOffset(const MCSymbolData *SD, uint64_t &Val) const { + return getSymbolOffsetImpl(*this, SD, false, Val); +} + +uint64_t MCAsmLayout::getSymbolOffset(const MCSymbolData *SD) const { + uint64_t Val; + getSymbolOffsetImpl(*this, SD, true, Val); + return Val; +} + +const MCSymbol *MCAsmLayout::getBaseSymbol(const MCSymbol &Symbol) const { + if (!Symbol.isVariable()) + return &Symbol; + + const MCExpr *Expr = Symbol.getVariableValue(); + MCValue Value; + if (!Expr->evaluateAsValue(Value, *this)) + llvm_unreachable("Invalid Expression"); + + const MCSymbolRefExpr *RefB = Value.getSymB(); + if (RefB) + Assembler.getContext().reportFatalError( + SMLoc(), Twine("symbol '") + RefB->getSymbol().getName() + + "' could not be evaluated in a subtraction expression"); + + const MCSymbolRefExpr *A = Value.getSymA(); + if (!A) + return nullptr; + + const MCSymbol &ASym = A->getSymbol(); + const MCAssembler &Asm = getAssembler(); + const MCSymbolData &ASD = Asm.getSymbolData(ASym); + if (ASD.isCommon()) { + // FIXME: we should probably add a SMLoc to MCExpr. + Asm.getContext().reportFatalError(SMLoc(), + "Common symbol " + ASym.getName() + + " cannot be used in assignment expr"); + } + + return &ASym; } uint64_t MCAsmLayout::getSectionAddressSize(const MCSectionData *SD) const { @@ -149,6 +229,47 @@ uint64_t MCAsmLayout::getSectionFileSize(const MCSectionData *SD) const { return getSectionAddressSize(SD); } +uint64_t llvm::computeBundlePadding(const MCAssembler &Assembler, + const MCFragment *F, + uint64_t FOffset, uint64_t FSize) { + uint64_t BundleSize = Assembler.getBundleAlignSize(); + assert(BundleSize > 0 && + "computeBundlePadding should only be called if bundling is enabled"); + uint64_t BundleMask = BundleSize - 1; + uint64_t OffsetInBundle = FOffset & BundleMask; + uint64_t EndOfFragment = OffsetInBundle + FSize; + + // There are two kinds of bundling restrictions: + // + // 1) For alignToBundleEnd(), add padding to ensure that the fragment will + // *end* on a bundle boundary. + // 2) Otherwise, check if the fragment would cross a bundle boundary. If it + // would, add padding until the end of the bundle so that the fragment + // will start in a new one. + if (F->alignToBundleEnd()) { + // Three possibilities here: + // + // A) The fragment just happens to end at a bundle boundary, so we're good. + // B) The fragment ends before the current bundle boundary: pad it just + // enough to reach the boundary. + // C) The fragment ends after the current bundle boundary: pad it until it + // reaches the end of the next bundle boundary. + // + // Note: this code could be made shorter with some modulo trickery, but it's + // intentionally kept in its more explicit form for simplicity. + if (EndOfFragment == BundleSize) + return 0; + else if (EndOfFragment < BundleSize) + return BundleSize - EndOfFragment; + else { // EndOfFragment > BundleSize + return 2 * BundleSize - EndOfFragment; + } + } else if (EndOfFragment > BundleSize) + return BundleSize - OffsetInBundle; + else + return 0; +} + /* *** */ MCFragment::MCFragment() : Kind(FragmentType(~0)) { @@ -157,40 +278,99 @@ MCFragment::MCFragment() : Kind(FragmentType(~0)) { MCFragment::~MCFragment() { } -MCFragment::MCFragment(FragmentType _Kind, MCSectionData *_Parent) - : Kind(_Kind), Parent(_Parent), Atom(0), Offset(~UINT64_C(0)) -{ +MCFragment::MCFragment(FragmentType Kind, MCSectionData *Parent) + : Kind(Kind), Parent(Parent), Atom(nullptr), Offset(~UINT64_C(0)) { if (Parent) Parent->getFragmentList().push_back(this); } /* *** */ -MCSectionData::MCSectionData() : Section(0) {} +MCEncodedFragment::~MCEncodedFragment() { +} + +/* *** */ + +MCEncodedFragmentWithFixups::~MCEncodedFragmentWithFixups() { +} -MCSectionData::MCSectionData(const MCSection &_Section, MCAssembler *A) - : Section(&_Section), - Ordinal(~UINT32_C(0)), - Alignment(1), - HasInstructions(false) -{ +/* *** */ + +MCSectionData::MCSectionData() : Section(nullptr) {} + +MCSectionData::MCSectionData(const MCSection &Section, MCAssembler *A) + : Section(&Section), Ordinal(~UINT32_C(0)), Alignment(1), + BundleLockState(NotBundleLocked), BundleLockNestingDepth(0), + BundleGroupBeforeFirstInst(false), HasInstructions(false) { if (A) A->getSectionList().push_back(this); } +MCSectionData::iterator +MCSectionData::getSubsectionInsertionPoint(unsigned Subsection) { + if (Subsection == 0 && SubsectionFragmentMap.empty()) + return end(); + + SmallVectorImpl >::iterator MI = + std::lower_bound(SubsectionFragmentMap.begin(), SubsectionFragmentMap.end(), + std::make_pair(Subsection, (MCFragment *)nullptr)); + bool ExactMatch = false; + if (MI != SubsectionFragmentMap.end()) { + ExactMatch = MI->first == Subsection; + if (ExactMatch) + ++MI; + } + iterator IP; + if (MI == SubsectionFragmentMap.end()) + IP = end(); + else + IP = MI->second; + if (!ExactMatch && Subsection != 0) { + // The GNU as documentation claims that subsections have an alignment of 4, + // although this appears not to be the case. + MCFragment *F = new MCDataFragment(); + SubsectionFragmentMap.insert(MI, std::make_pair(Subsection, F)); + getFragmentList().insert(IP, F); + F->setParent(this); + } + + return IP; +} + +void MCSectionData::setBundleLockState(BundleLockStateType NewState) { + if (NewState == NotBundleLocked) { + if (BundleLockNestingDepth == 0) { + report_fatal_error("Mismatched bundle_lock/unlock directives"); + } + if (--BundleLockNestingDepth == 0) { + BundleLockState = NotBundleLocked; + } + return; + } + + // If any of the directives is an align_to_end directive, the whole nested + // group is align_to_end. So don't downgrade from align_to_end to just locked. + if (BundleLockState != BundleLockedAlignToEnd) { + BundleLockState = NewState; + } + ++BundleLockNestingDepth; +} + /* *** */ -MCSymbolData::MCSymbolData() : Symbol(0) {} +MCSymbolData::MCSymbolData() : Symbol(nullptr) {} -MCSymbolData::MCSymbolData(const MCSymbol &_Symbol, MCFragment *_Fragment, - uint64_t _Offset, MCAssembler *A) - : Symbol(&_Symbol), Fragment(_Fragment), Offset(_Offset), - IsExternal(false), IsPrivateExtern(false), - CommonSize(0), SymbolSize(0), CommonAlign(0), - Flags(0), Index(0) -{ - if (A) - A->getSymbolList().push_back(this); +void MCSymbolData::initialize(const MCSymbol &Symbol, MCFragment *Fragment, + uint64_t Offset) { + assert(!isInitialized() && "Expected uninitialized symbol"); + + this->Symbol = &Symbol; + this->Fragment.setPointer(Fragment); + this->Offset = Offset; + this->SymbolSize = nullptr; + this->CommonAlign = -1U; + this->Flags = 0; + this->Index = 0; } /* *** */ @@ -198,13 +378,73 @@ MCSymbolData::MCSymbolData(const MCSymbol &_Symbol, MCFragment *_Fragment, MCAssembler::MCAssembler(MCContext &Context_, MCAsmBackend &Backend_, MCCodeEmitter &Emitter_, MCObjectWriter &Writer_, raw_ostream &OS_) - : Context(Context_), Backend(Backend_), Emitter(Emitter_), Writer(Writer_), - OS(OS_), RelaxAll(false), NoExecStack(false), SubsectionsViaSymbols(false) { + : Context(Context_), Backend(Backend_), Emitter(Emitter_), Writer(Writer_), + OS(OS_), BundleAlignSize(0), RelaxAll(false), + SubsectionsViaSymbols(false), ELFHeaderEFlags(0) { + VersionMinInfo.Major = 0; // Major version == 0 for "none specified" } MCAssembler::~MCAssembler() { } +void MCAssembler::reset() { + Sections.clear(); + Symbols.clear(); + SectionMap.clear(); + IndirectSymbols.clear(); + DataRegions.clear(); + LinkerOptions.clear(); + FileNames.clear(); + ThumbFuncs.clear(); + BundleAlignSize = 0; + RelaxAll = false; + SubsectionsViaSymbols = false; + ELFHeaderEFlags = 0; + LOHContainer.reset(); + VersionMinInfo.Major = 0; + + // reset objects owned by us + getBackend().reset(); + getEmitter().reset(); + getWriter().reset(); + getLOHContainer().reset(); +} + +bool MCAssembler::isThumbFunc(const MCSymbol *Symbol) const { + if (ThumbFuncs.count(Symbol)) + return true; + + if (!Symbol->isVariable()) + return false; + + // FIXME: It looks like gas supports some cases of the form "foo + 2". It + // is not clear if that is a bug or a feature. + const MCExpr *Expr = Symbol->getVariableValue(); + const MCSymbolRefExpr *Ref = dyn_cast(Expr); + if (!Ref) + return false; + + if (Ref->getKind() != MCSymbolRefExpr::VK_None) + return false; + + const MCSymbol &Sym = Ref->getSymbol(); + if (!isThumbFunc(&Sym)) + return false; + + ThumbFuncs.insert(Symbol); // Cache it. + return true; +} + +void MCAssembler::addLocalUsedInReloc(const MCSymbol &Sym) { + assert(Sym.isTemporary()); + LocalsUsedInReloc.insert(&Sym); +} + +bool MCAssembler::isLocalUsedInReloc(const MCSymbol &Sym) const { + assert(Sym.isTemporary()); + return LocalsUsedInReloc.count(&Sym); +} + bool MCAssembler::isSymbolLinkerVisible(const MCSymbol &Symbol) const { // Non-temporary labels should always be visible to the linker. if (!Symbol.isTemporary()) @@ -214,24 +454,26 @@ bool MCAssembler::isSymbolLinkerVisible(const MCSymbol &Symbol) const { if (!Symbol.isInSection()) return false; - // Otherwise, check if the section requires symbols even for temporary labels. - return getBackend().doesSectionRequireSymbols(Symbol.getSection()); + if (isLocalUsedInReloc(Symbol)) + return true; + + return false; } -const MCSymbolData *MCAssembler::getAtom(const MCSymbolData *SD) const { +const MCSymbol *MCAssembler::getAtom(const MCSymbolData *SD) const { // Linker visible symbols define atoms. if (isSymbolLinkerVisible(SD->getSymbol())) - return SD; + return &SD->getSymbol(); // Absolute and undefined symbols have no defining atom. if (!SD->getFragment()) - return 0; + return nullptr; // Non-linker visible symbols in sections which can't be atomized have no // defining atom. - if (!getBackend().isSectionAtomizable( - SD->getFragment()->getParent()->getSection())) - return 0; + if (!getContext().getAsmInfo()->isSectionAtomizableBySymbols( + SD->getFragment()->getParent()->getSection())) + return nullptr; // Otherwise, return the atom for the containing fragment. return SD->getFragment()->getAtom(); @@ -242,8 +484,12 @@ bool MCAssembler::evaluateFixup(const MCAsmLayout &Layout, MCValue &Target, uint64_t &Value) const { ++stats::evaluateFixup; - if (!Fixup.getValue()->EvaluateAsRelocatable(Target, Layout)) - getContext().FatalError(Fixup.getLoc(), "expected relocatable expression"); + // FIXME: This code has some duplication with RecordRelocation. We should + // probably merge the two into a single callback that tries to evaluate a + // fixup and records a relocation if one is needed. + const MCExpr *Expr = Fixup.getValue(); + if (!Expr->EvaluateAsRelocatable(Target, &Layout, &Fixup)) + getContext().reportFatalError(Fixup.getLoc(), "expected relocatable expression"); bool IsPCRel = Backend.getFixupKindInfo( Fixup.getKind()).Flags & MCFixupKindInfo::FKF_IsPCRel; @@ -257,14 +503,11 @@ bool MCAssembler::evaluateFixup(const MCAsmLayout &Layout, } else { const MCSymbolRefExpr *A = Target.getSymA(); const MCSymbol &SA = A->getSymbol(); - if (A->getKind() != MCSymbolRefExpr::VK_None || - SA.AliasedSymbol().isUndefined()) { + if (A->getKind() != MCSymbolRefExpr::VK_None || SA.isUndefined()) { IsResolved = false; } else { - const MCSymbolData &DataA = getSymbolData(SA); - IsResolved = - getWriter().IsSymbolRefDifferenceFullyResolvedImpl(*this, DataA, - *DF, false, true); + IsResolved = getWriter().IsSymbolRefDifferenceFullyResolvedImpl( + *this, SA, *DF, false, true); } } } else { @@ -274,12 +517,12 @@ bool MCAssembler::evaluateFixup(const MCAsmLayout &Layout, Value = Target.getConstant(); if (const MCSymbolRefExpr *A = Target.getSymA()) { - const MCSymbol &Sym = A->getSymbol().AliasedSymbol(); + const MCSymbol &Sym = A->getSymbol(); if (Sym.isDefined()) Value += Layout.getSymbolOffset(&getSymbolData(Sym)); } if (const MCSymbolRefExpr *B = Target.getSymB()) { - const MCSymbol &Sym = B->getSymbol().AliasedSymbol(); + const MCSymbol &Sym = B->getSymbol(); if (Sym.isDefined()) Value -= Layout.getSymbolOffset(&getSymbolData(Sym)); } @@ -311,11 +554,11 @@ uint64_t MCAssembler::computeFragmentSize(const MCAsmLayout &Layout, const MCFragment &F) const { switch (F.getKind()) { case MCFragment::FT_Data: - return cast(F).getContents().size(); + case MCFragment::FT_Relaxable: + case MCFragment::FT_CompactEncodedInst: + return cast(F).getContents().size(); case MCFragment::FT_Fill: return cast(F).getSize(); - case MCFragment::FT_Inst: - return cast(F).getInstSize(); case MCFragment::FT_LEB: return cast(F).getContents().size(); @@ -336,7 +579,7 @@ uint64_t MCAssembler::computeFragmentSize(const MCAsmLayout &Layout, } case MCFragment::FT_Org: { - MCOrgFragment &OF = cast(F); + const MCOrgFragment &OF = cast(F); int64_t TargetLocation; if (!OF.getOffset().EvaluateAsAbsolute(TargetLocation, Layout)) report_fatal_error("expected assembly-time absolute expression"); @@ -359,45 +602,127 @@ uint64_t MCAssembler::computeFragmentSize(const MCAsmLayout &Layout, llvm_unreachable("invalid fragment kind"); } -void MCAsmLayout::LayoutFragment(MCFragment *F) { +void MCAsmLayout::layoutFragment(MCFragment *F) { MCFragment *Prev = F->getPrevNode(); - // We should never try to recompute something which is up-to-date. - assert(!isFragmentUpToDate(F) && "Attempt to recompute up-to-date fragment!"); - // We should never try to compute the fragment layout if it's predecessor - // isn't up-to-date. - assert((!Prev || isFragmentUpToDate(Prev)) && - "Attempt to compute fragment before it's predecessor!"); + // We should never try to recompute something which is valid. + assert(!isFragmentValid(F) && "Attempt to recompute a valid fragment!"); + // We should never try to compute the fragment layout if its predecessor + // isn't valid. + assert((!Prev || isFragmentValid(Prev)) && + "Attempt to compute fragment before its predecessor!"); ++stats::FragmentLayouts; // Compute fragment offset and size. - uint64_t Offset = 0; if (Prev) - Offset += Prev->Offset + getAssembler().computeFragmentSize(*this, *Prev); - - F->Offset = Offset; + F->Offset = Prev->Offset + getAssembler().computeFragmentSize(*this, *Prev); + else + F->Offset = 0; LastValidFragment[F->getParent()] = F; + + // If bundling is enabled and this fragment has instructions in it, it has to + // obey the bundling restrictions. With padding, we'll have: + // + // + // BundlePadding + // ||| + // ------------------------------------- + // Prev |##########| F | + // ------------------------------------- + // ^ + // | + // F->Offset + // + // The fragment's offset will point to after the padding, and its computed + // size won't include the padding. + // + // When the -mc-relax-all flag is used, we optimize bundling by writting the + // bundle padding directly into fragments when the instructions are emitted + // inside the streamer. + // + if (Assembler.isBundlingEnabled() && !Assembler.getRelaxAll() && + F->hasInstructions()) { + assert(isa(F) && + "Only MCEncodedFragment implementations have instructions"); + uint64_t FSize = Assembler.computeFragmentSize(*this, *F); + + if (FSize > Assembler.getBundleAlignSize()) + report_fatal_error("Fragment can't be larger than a bundle size"); + + uint64_t RequiredBundlePadding = computeBundlePadding(Assembler, F, + F->Offset, FSize); + if (RequiredBundlePadding > UINT8_MAX) + report_fatal_error("Padding cannot exceed 255 bytes"); + F->setBundlePadding(static_cast(RequiredBundlePadding)); + F->Offset += RequiredBundlePadding; + } +} + +/// \brief Write the contents of a fragment to the given object writer. Expects +/// a MCEncodedFragment. +static void writeFragmentContents(const MCFragment &F, MCObjectWriter *OW) { + const MCEncodedFragment &EF = cast(F); + OW->WriteBytes(EF.getContents()); } -/// WriteFragmentData - Write the \p F data to the output file. -static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, - const MCFragment &F) { +void MCAssembler::writeFragmentPadding(const MCFragment &F, uint64_t FSize, + MCObjectWriter *OW) const { + // Should NOP padding be written out before this fragment? + unsigned BundlePadding = F.getBundlePadding(); + if (BundlePadding > 0) { + assert(isBundlingEnabled() && + "Writing bundle padding with disabled bundling"); + assert(F.hasInstructions() && + "Writing bundle padding for a fragment without instructions"); + + unsigned TotalLength = BundlePadding + static_cast(FSize); + if (F.alignToBundleEnd() && TotalLength > getBundleAlignSize()) { + // If the padding itself crosses a bundle boundary, it must be emitted + // in 2 pieces, since even nop instructions must not cross boundaries. + // v--------------v <- BundleAlignSize + // v---------v <- BundlePadding + // ---------------------------- + // | Prev |####|####| F | + // ---------------------------- + // ^-------------------^ <- TotalLength + unsigned DistanceToBoundary = TotalLength - getBundleAlignSize(); + if (!getBackend().writeNopData(DistanceToBoundary, OW)) + report_fatal_error("unable to write NOP sequence of " + + Twine(DistanceToBoundary) + " bytes"); + BundlePadding -= DistanceToBoundary; + } + if (!getBackend().writeNopData(BundlePadding, OW)) + report_fatal_error("unable to write NOP sequence of " + + Twine(BundlePadding) + " bytes"); + } +} + +/// \brief Write the fragment \p F to the output file. +static void writeFragment(const MCAssembler &Asm, const MCAsmLayout &Layout, + const MCFragment &F) { MCObjectWriter *OW = &Asm.getWriter(); + + // FIXME: Embed in fragments instead? + uint64_t FragmentSize = Asm.computeFragmentSize(Layout, F); + + Asm.writeFragmentPadding(F, FragmentSize, OW); + + // This variable (and its dummy usage) is to participate in the assert at + // the end of the function. uint64_t Start = OW->getStream().tell(); (void) Start; ++stats::EmittedFragments; - // FIXME: Embed in fragments instead? - uint64_t FragmentSize = Asm.computeFragmentSize(Layout, F); switch (F.getKind()) { case MCFragment::FT_Align: { - MCAlignFragment &AF = cast(F); - uint64_t Count = FragmentSize / AF.getValueSize(); - + ++stats::EmittedAlignFragments; + const MCAlignFragment &AF = cast(F); assert(AF.getValueSize() && "Invalid virtual align in concrete fragment!"); + uint64_t Count = FragmentSize / AF.getValueSize(); + // FIXME: This error shouldn't actually occur (the front end should emit // multiple .align directives to enforce the semantics it wants), but is // severe enough that we want to report it. How to handle this? @@ -431,15 +756,24 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, break; } - case MCFragment::FT_Data: { - MCDataFragment &DF = cast(F); - assert(FragmentSize == DF.getContents().size() && "Invalid size!"); - OW->WriteBytes(DF.getContents().str()); + case MCFragment::FT_Data: + ++stats::EmittedDataFragments; + writeFragmentContents(F, OW); + break; + + case MCFragment::FT_Relaxable: + ++stats::EmittedRelaxableFragments; + writeFragmentContents(F, OW); + break; + + case MCFragment::FT_CompactEncodedInst: + ++stats::EmittedCompactEncodedInstFragments; + writeFragmentContents(F, OW); break; - } case MCFragment::FT_Fill: { - MCFillFragment &FF = cast(F); + ++stats::EmittedFillFragments; + const MCFillFragment &FF = cast(F); assert(FF.getValueSize() && "Invalid virtual align in concrete fragment!"); @@ -455,20 +789,15 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, break; } - case MCFragment::FT_Inst: { - MCInstFragment &IF = cast(F); - OW->WriteBytes(StringRef(IF.getCode().begin(), IF.getCode().size())); - break; - } - case MCFragment::FT_LEB: { - MCLEBFragment &LF = cast(F); - OW->WriteBytes(LF.getContents().str()); + const MCLEBFragment &LF = cast(F); + OW->WriteBytes(LF.getContents()); break; } case MCFragment::FT_Org: { - MCOrgFragment &OF = cast(F); + ++stats::EmittedOrgFragments; + const MCOrgFragment &OF = cast(F); for (uint64_t i = 0, e = FragmentSize; i != e; ++i) OW->Write8(uint8_t(OF.getValue())); @@ -478,17 +807,18 @@ static void WriteFragmentData(const MCAssembler &Asm, const MCAsmLayout &Layout, case MCFragment::FT_Dwarf: { const MCDwarfLineAddrFragment &OF = cast(F); - OW->WriteBytes(OF.getContents().str()); + OW->WriteBytes(OF.getContents()); break; } case MCFragment::FT_DwarfFrame: { const MCDwarfCallFrameFragment &CF = cast(F); - OW->WriteBytes(CF.getContents().str()); + OW->WriteBytes(CF.getContents()); break; } } - assert(OW->getStream().tell() - Start == FragmentSize); + assert(OW->getStream().tell() - Start == FragmentSize && + "The stream should advance by fragment size"); } void MCAssembler::writeSectionData(const MCSectionData *SD, @@ -506,23 +836,29 @@ void MCAssembler::writeSectionData(const MCSectionData *SD, // Check that we aren't trying to write a non-zero contents (or fixups) // into a virtual section. This is to support clients which use standard // directives to fill the contents of virtual sections. - MCDataFragment &DF = cast(*it); + const MCDataFragment &DF = cast(*it); assert(DF.fixup_begin() == DF.fixup_end() && "Cannot have fixups in virtual section!"); for (unsigned i = 0, e = DF.getContents().size(); i != e; ++i) - assert(DF.getContents()[i] == 0 && - "Invalid data value for virtual section!"); + if (DF.getContents()[i]) { + if (auto *ELFSec = dyn_cast(&SD->getSection())) + report_fatal_error("non-zero initializer found in section '" + + ELFSec->getSectionName() + "'"); + else + report_fatal_error("non-zero initializer found in virtual section"); + } break; } case MCFragment::FT_Align: // Check that we aren't trying to write a non-zero value into a virtual // section. - assert((!cast(it)->getValueSize() || - !cast(it)->getValue()) && + assert((cast(it)->getValueSize() == 0 || + cast(it)->getValue() == 0) && "Invalid align in virtual section!"); break; case MCFragment::FT_Fill: - assert(!cast(it)->getValueSize() && + assert((cast(it)->getValueSize() == 0 || + cast(it)->getValue() == 0) && "Invalid fill in virtual section!"); break; } @@ -534,29 +870,31 @@ void MCAssembler::writeSectionData(const MCSectionData *SD, uint64_t Start = getWriter().getStream().tell(); (void)Start; - for (MCSectionData::const_iterator it = SD->begin(), - ie = SD->end(); it != ie; ++it) - WriteFragmentData(*this, Layout, *it); + for (MCSectionData::const_iterator it = SD->begin(), ie = SD->end(); + it != ie; ++it) + writeFragment(*this, Layout, *it); assert(getWriter().getStream().tell() - Start == Layout.getSectionAddressSize(SD)); } - -uint64_t MCAssembler::handleFixup(const MCAsmLayout &Layout, - MCFragment &F, - const MCFixup &Fixup) { - // Evaluate the fixup. - MCValue Target; - uint64_t FixedValue; - if (!evaluateFixup(Layout, Fixup, &F, Target, FixedValue)) { - // The fixup was unresolved, we need a relocation. Inform the object - // writer of the relocation, and give it an opportunity to adjust the - // fixup value if need be. - getWriter().RecordRelocation(*this, Layout, &F, Fixup, Target, FixedValue); - } - return FixedValue; - } +std::pair MCAssembler::handleFixup(const MCAsmLayout &Layout, + MCFragment &F, + const MCFixup &Fixup) { + // Evaluate the fixup. + MCValue Target; + uint64_t FixedValue; + bool IsPCRel = Backend.getFixupKindInfo(Fixup.getKind()).Flags & + MCFixupKindInfo::FKF_IsPCRel; + if (!evaluateFixup(Layout, Fixup, &F, Target, FixedValue)) { + // The fixup was unresolved, we need a relocation. Inform the object + // writer of the relocation, and give it an opportunity to adjust the + // fixup value if need be. + getWriter().RecordRelocation(*this, Layout, &F, Fixup, Target, IsPCRel, + FixedValue); + } + return std::make_pair(FixedValue, IsPCRel); +} void MCAssembler::Finish() { DEBUG_WITH_TYPE("mc-dump", { @@ -583,9 +921,9 @@ void MCAssembler::Finish() { SD->setLayoutOrder(i); unsigned FragmentIndex = 0; - for (MCSectionData::iterator it2 = SD->begin(), - ie2 = SD->end(); it2 != ie2; ++it2) - it2->setLayoutOrder(FragmentIndex++); + for (MCSectionData::iterator iFrag = SD->begin(), iFragEnd = SD->end(); + iFrag != iFragEnd; ++iFrag) + iFrag->setLayoutOrder(FragmentIndex++); } // Layout until everything fits. @@ -613,24 +951,17 @@ void MCAssembler::Finish() { for (MCAssembler::iterator it = begin(), ie = end(); it != ie; ++it) { for (MCSectionData::iterator it2 = it->begin(), ie2 = it->end(); it2 != ie2; ++it2) { - MCDataFragment *DF = dyn_cast(it2); - if (DF) { - for (MCDataFragment::fixup_iterator it3 = DF->fixup_begin(), - ie3 = DF->fixup_end(); it3 != ie3; ++it3) { - MCFixup &Fixup = *it3; - uint64_t FixedValue = handleFixup(Layout, *DF, Fixup); - getBackend().applyFixup(Fixup, DF->getContents().data(), - DF->getContents().size(), FixedValue); - } - } - MCInstFragment *IF = dyn_cast(it2); - if (IF) { - for (MCInstFragment::fixup_iterator it3 = IF->fixup_begin(), - ie3 = IF->fixup_end(); it3 != ie3; ++it3) { + MCEncodedFragmentWithFixups *F = + dyn_cast(it2); + if (F) { + for (MCEncodedFragmentWithFixups::fixup_iterator it3 = F->fixup_begin(), + ie3 = F->fixup_end(); it3 != ie3; ++it3) { MCFixup &Fixup = *it3; - uint64_t FixedValue = handleFixup(Layout, *IF, Fixup); - getBackend().applyFixup(Fixup, IF->getCode().data(), - IF->getCode().size(), FixedValue); + uint64_t FixedValue; + bool IsPCRel; + std::tie(FixedValue, IsPCRel) = handleFixup(Layout, *F, Fixup); + getBackend().applyFixup(Fixup, F->getContents().data(), + F->getContents().size(), FixedValue, IsPCRel); } } } @@ -643,11 +974,8 @@ void MCAssembler::Finish() { } bool MCAssembler::fixupNeedsRelaxation(const MCFixup &Fixup, - const MCInstFragment *DF, + const MCRelaxableFragment *DF, const MCAsmLayout &Layout) const { - if (getRelaxAll()) - return true; - // If we cannot resolve the fixup value, it requires relaxation. MCValue Target; uint64_t Value; @@ -657,25 +985,25 @@ bool MCAssembler::fixupNeedsRelaxation(const MCFixup &Fixup, return getBackend().fixupNeedsRelaxation(Fixup, Value, DF, Layout); } -bool MCAssembler::fragmentNeedsRelaxation(const MCInstFragment *IF, +bool MCAssembler::fragmentNeedsRelaxation(const MCRelaxableFragment *F, const MCAsmLayout &Layout) const { // If this inst doesn't ever need relaxation, ignore it. This occurs when we // are intentionally pushing out inst fragments, or because we relaxed a // previous instruction to one that doesn't need relaxation. - if (!getBackend().mayNeedRelaxation(IF->getInst())) + if (!getBackend().mayNeedRelaxation(F->getInst())) return false; - for (MCInstFragment::const_fixup_iterator it = IF->fixup_begin(), - ie = IF->fixup_end(); it != ie; ++it) - if (fixupNeedsRelaxation(*it, IF, Layout)) + for (MCRelaxableFragment::const_fixup_iterator it = F->fixup_begin(), + ie = F->fixup_end(); it != ie; ++it) + if (fixupNeedsRelaxation(*it, F, Layout)) return true; return false; } bool MCAssembler::relaxInstruction(MCAsmLayout &Layout, - MCInstFragment &IF) { - if (!fragmentNeedsRelaxation(&IF, Layout)) + MCRelaxableFragment &F) { + if (!fragmentNeedsRelaxation(&F, Layout)) return false; ++stats::RelaxedInstructions; @@ -686,7 +1014,7 @@ bool MCAssembler::relaxInstruction(MCAsmLayout &Layout, // Relax the fragment. MCInst Relaxed; - getBackend().relaxInstruction(IF.getInst(), Relaxed); + getBackend().relaxInstruction(F.getInst(), Relaxed); // Encode the new instruction. // @@ -695,26 +1023,23 @@ bool MCAssembler::relaxInstruction(MCAsmLayout &Layout, SmallVector Fixups; SmallString<256> Code; raw_svector_ostream VecOS(Code); - getEmitter().EncodeInstruction(Relaxed, VecOS, Fixups); + getEmitter().encodeInstruction(Relaxed, VecOS, Fixups, F.getSubtargetInfo()); VecOS.flush(); - // Update the instruction fragment. - IF.setInst(Relaxed); - IF.getCode() = Code; - IF.getFixups().clear(); - // FIXME: Eliminate copy. - for (unsigned i = 0, e = Fixups.size(); i != e; ++i) - IF.getFixups().push_back(Fixups[i]); + // Update the fragment. + F.setInst(Relaxed); + F.getContents() = Code; + F.getFixups() = Fixups; return true; } bool MCAssembler::relaxLEB(MCAsmLayout &Layout, MCLEBFragment &LF) { - int64_t Value = 0; uint64_t OldSize = LF.getContents().size(); - bool IsAbs = LF.getValue().EvaluateAsAbsolute(Value, Layout); - (void)IsAbs; - assert(IsAbs); + int64_t Value; + bool Abs = LF.getValue().evaluateKnownAbsolute(Value, Layout); + if (!Abs) + report_fatal_error("sleb128 and uleb128 expressions must be absolute"); SmallString<8> &Data = LF.getContents(); Data.clear(); raw_svector_ostream OSE(Data); @@ -728,69 +1053,75 @@ bool MCAssembler::relaxLEB(MCAsmLayout &Layout, MCLEBFragment &LF) { bool MCAssembler::relaxDwarfLineAddr(MCAsmLayout &Layout, MCDwarfLineAddrFragment &DF) { - int64_t AddrDelta = 0; + MCContext &Context = Layout.getAssembler().getContext(); uint64_t OldSize = DF.getContents().size(); - bool IsAbs = DF.getAddrDelta().EvaluateAsAbsolute(AddrDelta, Layout); - (void)IsAbs; - assert(IsAbs); + int64_t AddrDelta; + bool Abs = DF.getAddrDelta().evaluateKnownAbsolute(AddrDelta, Layout); + assert(Abs && "We created a line delta with an invalid expression"); + (void) Abs; int64_t LineDelta; LineDelta = DF.getLineDelta(); SmallString<8> &Data = DF.getContents(); Data.clear(); raw_svector_ostream OSE(Data); - MCDwarfLineAddr::Encode(LineDelta, AddrDelta, OSE); + MCDwarfLineAddr::Encode(Context, LineDelta, AddrDelta, OSE); OSE.flush(); return OldSize != Data.size(); } bool MCAssembler::relaxDwarfCallFrameFragment(MCAsmLayout &Layout, MCDwarfCallFrameFragment &DF) { - int64_t AddrDelta = 0; + MCContext &Context = Layout.getAssembler().getContext(); uint64_t OldSize = DF.getContents().size(); - bool IsAbs = DF.getAddrDelta().EvaluateAsAbsolute(AddrDelta, Layout); - (void)IsAbs; - assert(IsAbs); + int64_t AddrDelta; + bool Abs = DF.getAddrDelta().evaluateKnownAbsolute(AddrDelta, Layout); + assert(Abs && "We created call frame with an invalid expression"); + (void) Abs; SmallString<8> &Data = DF.getContents(); Data.clear(); raw_svector_ostream OSE(Data); - MCDwarfFrameEmitter::EncodeAdvanceLoc(AddrDelta, OSE); + MCDwarfFrameEmitter::EncodeAdvanceLoc(Context, AddrDelta, OSE); OSE.flush(); return OldSize != Data.size(); } -bool MCAssembler::layoutSectionOnce(MCAsmLayout &Layout, - MCSectionData &SD) { - MCFragment *FirstInvalidFragment = NULL; - // Scan for fragments that need relaxation. - for (MCSectionData::iterator it2 = SD.begin(), - ie2 = SD.end(); it2 != ie2; ++it2) { - // Check if this is an fragment that needs relaxation. - bool relaxedFrag = false; - switch(it2->getKind()) { +bool MCAssembler::layoutSectionOnce(MCAsmLayout &Layout, MCSectionData &SD) { + // Holds the first fragment which needed relaxing during this layout. It will + // remain NULL if none were relaxed. + // When a fragment is relaxed, all the fragments following it should get + // invalidated because their offset is going to change. + MCFragment *FirstRelaxedFragment = nullptr; + + // Attempt to relax all the fragments in the section. + for (MCSectionData::iterator I = SD.begin(), IE = SD.end(); I != IE; ++I) { + // Check if this is a fragment that needs relaxation. + bool RelaxedFrag = false; + switch(I->getKind()) { default: - break; - case MCFragment::FT_Inst: - relaxedFrag = relaxInstruction(Layout, *cast(it2)); + break; + case MCFragment::FT_Relaxable: + assert(!getRelaxAll() && + "Did not expect a MCRelaxableFragment in RelaxAll mode"); + RelaxedFrag = relaxInstruction(Layout, *cast(I)); break; case MCFragment::FT_Dwarf: - relaxedFrag = relaxDwarfLineAddr(Layout, - *cast(it2)); + RelaxedFrag = relaxDwarfLineAddr(Layout, + *cast(I)); break; case MCFragment::FT_DwarfFrame: - relaxedFrag = + RelaxedFrag = relaxDwarfCallFrameFragment(Layout, - *cast(it2)); + *cast(I)); break; case MCFragment::FT_LEB: - relaxedFrag = relaxLEB(Layout, *cast(it2)); + RelaxedFrag = relaxLEB(Layout, *cast(I)); break; } - // Update the layout, and remember that we relaxed. - if (relaxedFrag && !FirstInvalidFragment) - FirstInvalidFragment = it2; + if (RelaxedFrag && !FirstRelaxedFragment) + FirstRelaxedFragment = I; } - if (FirstInvalidFragment) { - Layout.Invalidate(FirstInvalidFragment); + if (FirstRelaxedFragment) { + Layout.invalidateFragmentsFrom(FirstRelaxedFragment); return true; } return false; @@ -802,7 +1133,7 @@ bool MCAssembler::layoutOnce(MCAsmLayout &Layout) { bool WasRelaxed = false; for (iterator it = begin(), ie = end(); it != ie; ++it) { MCSectionData &SD = *it; - while(layoutSectionOnce(Layout, SD)) + while (layoutSectionOnce(Layout, SD)) WasRelaxed = true; } @@ -837,8 +1168,10 @@ void MCFragment::dump() { switch (getKind()) { case MCFragment::FT_Align: OS << "MCAlignFragment"; break; case MCFragment::FT_Data: OS << "MCDataFragment"; break; + case MCFragment::FT_CompactEncodedInst: + OS << "MCCompactEncodedInstFragment"; break; case MCFragment::FT_Fill: OS << "MCFillFragment"; break; - case MCFragment::FT_Inst: OS << "MCInstFragment"; break; + case MCFragment::FT_Relaxable: OS << "MCRelaxableFragment"; break; case MCFragment::FT_Org: OS << "MCOrgFragment"; break; case MCFragment::FT_Dwarf: OS << "MCDwarfFragment"; break; case MCFragment::FT_DwarfFrame: OS << "MCDwarfCallFrameFragment"; break; @@ -846,7 +1179,9 @@ void MCFragment::dump() { } OS << ""; + << " Offset:" << Offset + << " HasInstructions:" << hasInstructions() + << " BundlePadding:" << static_cast(getBundlePadding()) << ">"; switch (getKind()) { case MCFragment::FT_Align: { @@ -882,17 +1217,30 @@ void MCFragment::dump() { } break; } + case MCFragment::FT_CompactEncodedInst: { + const MCCompactEncodedInstFragment *CEIF = + cast(this); + OS << "\n "; + OS << " Contents:["; + const SmallVectorImpl &Contents = CEIF->getContents(); + for (unsigned i = 0, e = Contents.size(); i != e; ++i) { + if (i) OS << ","; + OS << hexdigit((Contents[i] >> 4) & 0xF) << hexdigit(Contents[i] & 0xF); + } + OS << "] (" << Contents.size() << " bytes)"; + break; + } case MCFragment::FT_Fill: { const MCFillFragment *FF = cast(this); OS << " Value:" << FF->getValue() << " ValueSize:" << FF->getValueSize() << " Size:" << FF->getSize(); break; } - case MCFragment::FT_Inst: { - const MCInstFragment *IF = cast(this); + case MCFragment::FT_Relaxable: { + const MCRelaxableFragment *F = cast(this); OS << "\n "; OS << " Inst:"; - IF->getInst().dump_pretty(OS); + F->getInst().dump_pretty(OS); break; } case MCFragment::FT_Org: { @@ -928,7 +1276,8 @@ void MCSectionData::dump() { raw_ostream &OS = llvm::errs(); OS << "dump(); @@ -936,12 +1285,14 @@ void MCSectionData::dump() { OS << "]>"; } -void MCSymbolData::dump() { +void MCSymbolData::dump() const { raw_ostream &OS = llvm::errs(); OS << "