blob: 047676d4c11efc50c39d2fc22a1b16a892af2c2c [file] [log] [blame]
//===- llvm/CodeGen/DwarfDebug.cpp - Dwarf Debug Framework ----------------===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// This file contains support for writing dwarf debug info into asm files.
#include "DwarfDebug.h"
#include "ByteStreamer.h"
#include "DIEHash.h"
#include "DwarfCompileUnit.h"
#include "DwarfExpression.h"
#include "DwarfUnit.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/Triple.h"
#include "llvm/ADT/Twine.h"
#include "llvm/CodeGen/AsmPrinter.h"
#include "llvm/CodeGen/DIE.h"
#include "llvm/CodeGen/LexicalScopes.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/DebugInfo/DWARF/DWARFExpression.h"
#include "llvm/DebugInfo/DWARF/DWARFDataExtractor.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/Module.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCSection.h"
#include "llvm/MC/MCStreamer.h"
#include "llvm/MC/MCSymbol.h"
#include "llvm/MC/MCTargetOptions.h"
#include "llvm/MC/MachineLocation.h"
#include "llvm/MC/SectionKind.h"
#include "llvm/Pass.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MD5.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/Timer.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetMachine.h"
#include <algorithm>
#include <cstddef>
#include <iterator>
#include <string>
using namespace llvm;
#define DEBUG_TYPE "dwarfdebug"
STATISTIC(NumCSParams, "Number of dbg call site params created");
static cl::opt<bool> UseDwarfRangesBaseAddressSpecifier(
"use-dwarf-ranges-base-address-specifier", cl::Hidden,
cl::desc("Use base address specifiers in debug_ranges"), cl::init(false));
static cl::opt<bool> GenerateARangeSection("generate-arange-section",
cl::desc("Generate dwarf aranges"),
static cl::opt<bool>
GenerateDwarfTypeUnits("generate-type-units", cl::Hidden,
cl::desc("Generate DWARF4 type units."),
static cl::opt<bool> SplitDwarfCrossCuReferences(
"split-dwarf-cross-cu-references", cl::Hidden,
cl::desc("Enable cross-cu references in DWO files"), cl::init(false));
enum DefaultOnOff { Default, Enable, Disable };
static cl::opt<DefaultOnOff> UnknownLocations(
"use-unknown-locations", cl::Hidden,
cl::desc("Make an absence of debug location information explicit."),
cl::values(clEnumVal(Default, "At top of block or after label"),
clEnumVal(Enable, "In all cases"), clEnumVal(Disable, "Never")),
static cl::opt<AccelTableKind> AccelTables(
"accel-tables", cl::Hidden, cl::desc("Output dwarf accelerator tables."),
cl::values(clEnumValN(AccelTableKind::Default, "Default",
"Default for platform"),
clEnumValN(AccelTableKind::None, "Disable", "Disabled."),
clEnumValN(AccelTableKind::Apple, "Apple", "Apple"),
clEnumValN(AccelTableKind::Dwarf, "Dwarf", "DWARF")),
static cl::opt<DefaultOnOff>
DwarfInlinedStrings("dwarf-inlined-strings", cl::Hidden,
cl::desc("Use inlined strings rather than string section."),
cl::values(clEnumVal(Default, "Default for platform"),
clEnumVal(Enable, "Enabled"),
clEnumVal(Disable, "Disabled")),
static cl::opt<bool>
NoDwarfRangesSection("no-dwarf-ranges-section", cl::Hidden,
cl::desc("Disable emission .debug_ranges section."),
static cl::opt<DefaultOnOff> DwarfSectionsAsReferences(
"dwarf-sections-as-references", cl::Hidden,
cl::desc("Use sections+offset as references rather than labels."),
cl::values(clEnumVal(Default, "Default for platform"),
clEnumVal(Enable, "Enabled"), clEnumVal(Disable, "Disabled")),
static cl::opt<bool>
UseGNUDebugMacro("use-gnu-debug-macro", cl::Hidden,
cl::desc("Emit the GNU .debug_macro format with DWARF <5"),
static cl::opt<DefaultOnOff> DwarfOpConvert(
"dwarf-op-convert", cl::Hidden,
cl::desc("Enable use of the DWARFv5 DW_OP_convert operator"),
cl::values(clEnumVal(Default, "Default for platform"),
clEnumVal(Enable, "Enabled"), clEnumVal(Disable, "Disabled")),
enum LinkageNameOption {
static cl::opt<LinkageNameOption>
DwarfLinkageNames("dwarf-linkage-names", cl::Hidden,
cl::desc("Which DWARF linkage-name attributes to emit."),
cl::values(clEnumValN(DefaultLinkageNames, "Default",
"Default for platform"),
clEnumValN(AllLinkageNames, "All", "All"),
clEnumValN(AbstractLinkageNames, "Abstract",
"Abstract subprograms")),
static cl::opt<DwarfDebug::MinimizeAddrInV5> MinimizeAddrInV5Option(
"minimize-addr-in-v5", cl::Hidden,
cl::desc("Always use DW_AT_ranges in DWARFv5 whenever it could allow more "
"address pool entry sharing to reduce relocations/object size"),
cl::values(clEnumValN(DwarfDebug::MinimizeAddrInV5::Default, "Default",
"Default address minimization strategy"),
clEnumValN(DwarfDebug::MinimizeAddrInV5::Ranges, "Ranges",
"Use rnglists for contiguous ranges if that allows "
"using a pre-existing base address"),
"Use exprloc addrx+offset expressions for any "
"address with a prior base address"),
clEnumValN(DwarfDebug::MinimizeAddrInV5::Form, "Form",
"Use addrx+offset extension form for any address "
"with a prior base address"),
clEnumValN(DwarfDebug::MinimizeAddrInV5::Disabled, "Disabled",
static constexpr unsigned ULEB128PadSize = 4;
void DebugLocDwarfExpression::emitOp(uint8_t Op, const char *Comment) {
Op, Comment ? Twine(Comment) + " " + dwarf::OperationEncodingString(Op)
: dwarf::OperationEncodingString(Op));
void DebugLocDwarfExpression::emitSigned(int64_t Value) {
getActiveStreamer().emitSLEB128(Value, Twine(Value));
void DebugLocDwarfExpression::emitUnsigned(uint64_t Value) {
getActiveStreamer().emitULEB128(Value, Twine(Value));
void DebugLocDwarfExpression::emitData1(uint8_t Value) {
getActiveStreamer().emitInt8(Value, Twine(Value));
void DebugLocDwarfExpression::emitBaseTypeRef(uint64_t Idx) {
assert(Idx < (1ULL << (ULEB128PadSize * 7)) && "Idx wont fit");
getActiveStreamer().emitULEB128(Idx, Twine(Idx), ULEB128PadSize);
bool DebugLocDwarfExpression::isFrameRegister(const TargetRegisterInfo &TRI,
llvm::Register MachineReg) {
// This information is not available while emitting .debug_loc entries.
return false;
void DebugLocDwarfExpression::enableTemporaryBuffer() {
assert(!IsBuffering && "Already buffering?");
if (!TmpBuf)
TmpBuf = std::make_unique<TempBuffer>(OutBS.GenerateComments);
IsBuffering = true;
void DebugLocDwarfExpression::disableTemporaryBuffer() { IsBuffering = false; }
unsigned DebugLocDwarfExpression::getTemporaryBufferSize() {
return TmpBuf ? TmpBuf->Bytes.size() : 0;
void DebugLocDwarfExpression::commitTemporaryBuffer() {
if (!TmpBuf)
for (auto Byte : enumerate(TmpBuf->Bytes)) {
const char *Comment = (Byte.index() < TmpBuf->Comments.size())
? TmpBuf->Comments[Byte.index()].c_str()
: "";
OutBS.emitInt8(Byte.value(), Comment);
const DIType *DbgVariable::getType() const {
return getVariable()->getType();
/// Get .debug_loc entry for the instruction range starting at MI.
static DbgValueLoc getDebugLocValue(const MachineInstr *MI) {
const DIExpression *Expr = MI->getDebugExpression();
const bool IsVariadic = MI->isDebugValueList();
assert(MI->getNumOperands() >= 3);
SmallVector<DbgValueLocEntry, 4> DbgValueLocEntries;
for (const MachineOperand &Op : MI->debug_operands()) {
if (Op.isReg()) {
MachineLocation MLoc(Op.getReg(),
MI->isNonListDebugValue() && MI->isDebugOffsetImm());
} else if (Op.isTargetIndex()) {
DbgValueLocEntry(TargetIndexLocation(Op.getIndex(), Op.getOffset())));
} else if (Op.isImm())
else if (Op.isFPImm())
else if (Op.isCImm())
llvm_unreachable("Unexpected debug operand in DBG_VALUE* instruction!");
return DbgValueLoc(Expr, DbgValueLocEntries, IsVariadic);
void DbgVariable::initializeDbgValue(const MachineInstr *DbgValue) {
assert(FrameIndexExprs.empty() && "Already initialized?");
assert(!ValueLoc.get() && "Already initialized?");
assert(getVariable() == DbgValue->getDebugVariable() && "Wrong variable");
assert(getInlinedAt() == DbgValue->getDebugLoc()->getInlinedAt() &&
"Wrong inlined-at");
ValueLoc = std::make_unique<DbgValueLoc>(getDebugLocValue(DbgValue));
if (auto *E = DbgValue->getDebugExpression())
if (E->getNumElements())
FrameIndexExprs.push_back({0, E});
ArrayRef<DbgVariable::FrameIndexExpr> DbgVariable::getFrameIndexExprs() const {
if (FrameIndexExprs.size() == 1)
return FrameIndexExprs;
[](const FrameIndexExpr &A) {
return A.Expr->isFragment();
}) &&
"multiple FI expressions without DW_OP_LLVM_fragment");
[](const FrameIndexExpr &A, const FrameIndexExpr &B) -> bool {
return A.Expr->getFragmentInfo()->OffsetInBits <
return FrameIndexExprs;
void DbgVariable::addMMIEntry(const DbgVariable &V) {
assert(DebugLocListIndex == ~0U && !ValueLoc.get() && "not an MMI entry");
assert(V.DebugLocListIndex == ~0U && !V.ValueLoc.get() && "not an MMI entry");
assert(V.getVariable() == getVariable() && "conflicting variable");
assert(V.getInlinedAt() == getInlinedAt() && "conflicting inlined-at location");
assert(!FrameIndexExprs.empty() && "Expected an MMI entry");
assert(!V.FrameIndexExprs.empty() && "Expected an MMI entry");
// FIXME: This logic should not be necessary anymore, as we now have proper
// deduplication. However, without it, we currently run into the assertion
// below, which means that we are likely dealing with broken input, i.e. two
// non-fragment entries for the same variable at different frame indices.
if (FrameIndexExprs.size()) {
auto *Expr = FrameIndexExprs.back().Expr;
if (!Expr || !Expr->isFragment())
for (const auto &FIE : V.FrameIndexExprs)
// Ignore duplicate entries.
if (llvm::none_of(FrameIndexExprs, [&](const FrameIndexExpr &Other) {
return FIE.FI == Other.FI && FIE.Expr == Other.Expr;
assert((FrameIndexExprs.size() == 1 ||
[](FrameIndexExpr &FIE) {
return FIE.Expr && FIE.Expr->isFragment();
})) &&
"conflicting locations for variable");
static AccelTableKind computeAccelTableKind(unsigned DwarfVersion,
bool GenerateTypeUnits,
DebuggerKind Tuning,
const Triple &TT) {
// Honor an explicit request.
if (AccelTables != AccelTableKind::Default)
return AccelTables;
// Accelerator tables with type units are currently not supported.
if (GenerateTypeUnits)
return AccelTableKind::None;
// Accelerator tables get emitted if targetting DWARF v5 or LLDB. DWARF v5
// always implies debug_names. For lower standard versions we use apple
// accelerator tables on apple platforms and debug_names elsewhere.
if (DwarfVersion >= 5)
return AccelTableKind::Dwarf;
if (Tuning == DebuggerKind::LLDB)
return TT.isOSBinFormatMachO() ? AccelTableKind::Apple
: AccelTableKind::Dwarf;
return AccelTableKind::None;
DwarfDebug::DwarfDebug(AsmPrinter *A)
: DebugHandlerBase(A), DebugLocs(A->OutStreamer->isVerboseAsm()),
InfoHolder(A, "info_string", DIEValueAllocator),
SkeletonHolder(A, "skel_string", DIEValueAllocator),
IsDarwin(A->TM.getTargetTriple().isOSDarwin()) {
const Triple &TT = Asm->TM.getTargetTriple();
// Make sure we know our "debugger tuning". The target option takes
// precedence; fall back to triple-based defaults.
if (Asm->TM.Options.DebuggerTuning != DebuggerKind::Default)
DebuggerTuning = Asm->TM.Options.DebuggerTuning;
else if (IsDarwin)
DebuggerTuning = DebuggerKind::LLDB;
else if (TT.isPS4CPU())
DebuggerTuning = DebuggerKind::SCE;
else if (TT.isOSAIX())
DebuggerTuning = DebuggerKind::DBX;
DebuggerTuning = DebuggerKind::GDB;
if (DwarfInlinedStrings == Default)
UseInlineStrings = TT.isNVPTX() || tuneForDBX();
UseInlineStrings = DwarfInlinedStrings == Enable;
UseLocSection = !TT.isNVPTX();
HasAppleExtensionAttributes = tuneForLLDB();
// Handle split DWARF.
HasSplitDwarf = !Asm->TM.Options.MCOptions.SplitDwarfFile.empty();
// SCE defaults to linkage names only for abstract subprograms.
if (DwarfLinkageNames == DefaultLinkageNames)
UseAllLinkageNames = !tuneForSCE();
UseAllLinkageNames = DwarfLinkageNames == AllLinkageNames;
unsigned DwarfVersionNumber = Asm->TM.Options.MCOptions.DwarfVersion;
unsigned DwarfVersion = DwarfVersionNumber ? DwarfVersionNumber
: MMI->getModule()->getDwarfVersion();
// Use dwarf 4 by default if nothing is requested. For NVPTX, use dwarf 2.
DwarfVersion =
TT.isNVPTX() ? 2 : (DwarfVersion ? DwarfVersion : dwarf::DWARF_VERSION);
bool Dwarf64 = DwarfVersion >= 3 && // DWARF64 was introduced in DWARFv3.
TT.isArch64Bit(); // DWARF64 requires 64-bit relocations.
// Support DWARF64
// 1: For ELF when requested.
// 2: For XCOFF64: the AIX assembler will fill in debug section lengths
// according to the DWARF64 format for 64-bit assembly, so we must use
// DWARF64 in the compiler too for 64-bit mode.
Dwarf64 &=
((Asm->TM.Options.MCOptions.Dwarf64 || MMI->getModule()->isDwarf64()) &&
TT.isOSBinFormatELF()) ||
if (!Dwarf64 && TT.isArch64Bit() && TT.isOSBinFormatXCOFF())
report_fatal_error("XCOFF requires DWARF64 for 64-bit mode!");
UseRangesSection = !NoDwarfRangesSection && !TT.isNVPTX();
// Use sections as references. Force for NVPTX.
if (DwarfSectionsAsReferences == Default)
UseSectionsAsReferences = TT.isNVPTX();
UseSectionsAsReferences = DwarfSectionsAsReferences == Enable;
// Don't generate type units for unsupported object file formats.
GenerateTypeUnits = (A->TM.getTargetTriple().isOSBinFormatELF() ||
A->TM.getTargetTriple().isOSBinFormatWasm()) &&
TheAccelTableKind = computeAccelTableKind(
DwarfVersion, GenerateTypeUnits, DebuggerTuning, A->TM.getTargetTriple());
// Work around a GDB bug. GDB doesn't support the standard opcode;
// SCE doesn't support GNU's; LLDB prefers the standard opcode, which
// is defined as of DWARF 3.
// See GDB bug 11616 - DW_OP_form_tls_address is unimplemented
UseGNUTLSOpcode = tuneForGDB() || DwarfVersion < 3;
// GDB does not fully support the DWARF 4 representation for bitfields.
UseDWARF2Bitfields = (DwarfVersion < 4) || tuneForGDB();
// The DWARF v5 string offsets table has - possibly shared - contributions
// from each compile and type unit each preceded by a header. The string
// offsets table used by the pre-DWARF v5 split-DWARF implementation uses
// a monolithic string offsets table without any header.
UseSegmentedStringOffsetsTable = DwarfVersion >= 5;
// Emit call-site-param debug info for GDB and LLDB, if the target supports
// the debug entry values feature. It can also be enabled explicitly.
EmitDebugEntryValues = Asm->TM.Options.ShouldEmitDebugEntryValues();
// It is unclear if the GCC .debug_macro extension is well-specified
// for split DWARF. For now, do not allow LLVM to emit it.
UseDebugMacroSection =
DwarfVersion >= 5 || (UseGNUDebugMacro && !useSplitDwarf());
if (DwarfOpConvert == Default)
EnableOpConvert = !((tuneForGDB() && useSplitDwarf()) || (tuneForLLDB() && !TT.isOSBinFormatMachO()));
EnableOpConvert = (DwarfOpConvert == Enable);
// Split DWARF would benefit object size significantly by trading reductions
// in address pool usage for slightly increased range list encodings.
if (DwarfVersion >= 5) {
MinimizeAddr = MinimizeAddrInV5Option;
// FIXME: In the future, enable this by default for Split DWARF where the
// tradeoff is more pronounced due to being able to offload the range
// lists to the dwo file and shrink object files/reduce relocations there.
if (MinimizeAddr == MinimizeAddrInV5::Default)
MinimizeAddr = MinimizeAddrInV5::Disabled;
Asm->OutStreamer->getContext().setDwarfFormat(Dwarf64 ? dwarf::DWARF64
: dwarf::DWARF32);
// Define out of line so we don't have to include DwarfUnit.h in DwarfDebug.h.
DwarfDebug::~DwarfDebug() = default;
static bool isObjCClass(StringRef Name) {
return Name.startswith("+") || Name.startswith("-");
static bool hasObjCCategory(StringRef Name) {
if (!isObjCClass(Name))
return false;
return Name.contains(") ");
static void getObjCClassCategory(StringRef In, StringRef &Class,
StringRef &Category) {
if (!hasObjCCategory(In)) {
Class = In.slice(In.find('[') + 1, In.find(' '));
Category = "";
Class = In.slice(In.find('[') + 1, In.find('('));
Category = In.slice(In.find('[') + 1, In.find(' '));
static StringRef getObjCMethodName(StringRef In) {
return In.slice(In.find(' ') + 1, In.find(']'));
// Add the various names to the Dwarf accelerator table names.
void DwarfDebug::addSubprogramNames(const DICompileUnit &CU,
const DISubprogram *SP, DIE &Die) {
if (getAccelTableKind() != AccelTableKind::Apple &&
CU.getNameTableKind() == DICompileUnit::DebugNameTableKind::None)
if (!SP->isDefinition())
if (SP->getName() != "")
addAccelName(CU, SP->getName(), Die);
// If the linkage name is different than the name, go ahead and output that as
// well into the name table. Only do that if we are going to actually emit
// that name.
if (SP->getLinkageName() != "" && SP->getName() != SP->getLinkageName() &&
(useAllLinkageNames() || InfoHolder.getAbstractSPDies().lookup(SP)))
addAccelName(CU, SP->getLinkageName(), Die);
// If this is an Objective-C selector name add it to the ObjC accelerator
// too.
if (isObjCClass(SP->getName())) {
StringRef Class, Category;
getObjCClassCategory(SP->getName(), Class, Category);
addAccelObjC(CU, Class, Die);
if (Category != "")
addAccelObjC(CU, Category, Die);
// Also add the base method name to the name table.
addAccelName(CU, getObjCMethodName(SP->getName()), Die);
/// Check whether we should create a DIE for the given Scope, return true
/// if we don't create a DIE (the corresponding DIE is null).
bool DwarfDebug::isLexicalScopeDIENull(LexicalScope *Scope) {
if (Scope->isAbstractScope())
return false;
// We don't create a DIE if there is no Range.
const SmallVectorImpl<InsnRange> &Ranges = Scope->getRanges();
if (Ranges.empty())
return true;
if (Ranges.size() > 1)
return false;
// We don't create a DIE if we have a single Range and the end label
// is null.
return !getLabelAfterInsn(Ranges.front().second);
template <typename Func> static void forBothCUs(DwarfCompileUnit &CU, Func F) {
if (auto *SkelCU = CU.getSkeleton())
if (CU.getCUNode()->getSplitDebugInlining())
bool DwarfDebug::shareAcrossDWOCUs() const {
return SplitDwarfCrossCuReferences;
void DwarfDebug::constructAbstractSubprogramScopeDIE(DwarfCompileUnit &SrcCU,
LexicalScope *Scope) {
assert(Scope && Scope->getScopeNode());
auto *SP = cast<DISubprogram>(Scope->getScopeNode());
// Find the subprogram's DwarfCompileUnit in the SPMap in case the subprogram
// was inlined from another compile unit.
if (useSplitDwarf() && !shareAcrossDWOCUs() && !SP->getUnit()->getSplitDebugInlining())
// Avoid building the original CU if it won't be used
else {
auto &CU = getOrCreateDwarfCompileUnit(SP->getUnit());
if (auto *SkelCU = CU.getSkeleton()) {
(shareAcrossDWOCUs() ? CU : SrcCU)
if (CU.getCUNode()->getSplitDebugInlining())
} else
/// Represents a parameter whose call site value can be described by applying a
/// debug expression to a register in the forwarded register worklist.
struct FwdRegParamInfo {
/// The described parameter register.
unsigned ParamReg;
/// Debug expression that has been built up when walking through the
/// instruction chain that produces the parameter's value.
const DIExpression *Expr;
/// Register worklist for finding call site values.
using FwdRegWorklist = MapVector<unsigned, SmallVector<FwdRegParamInfo, 2>>;
/// Append the expression \p Addition to \p Original and return the result.
static const DIExpression *combineDIExpressions(const DIExpression *Original,
const DIExpression *Addition) {
std::vector<uint64_t> Elts = Addition->getElements().vec();
// Avoid multiple DW_OP_stack_values.
if (Original->isImplicit() && Addition->isImplicit())
erase_value(Elts, dwarf::DW_OP_stack_value);
const DIExpression *CombinedExpr =
(Elts.size() > 0) ? DIExpression::append(Original, Elts) : Original;
return CombinedExpr;
/// Emit call site parameter entries that are described by the given value and
/// debug expression.
template <typename ValT>
static void finishCallSiteParams(ValT Val, const DIExpression *Expr,
ArrayRef<FwdRegParamInfo> DescribedParams,
ParamSet &Params) {
for (auto Param : DescribedParams) {
bool ShouldCombineExpressions = Expr && Param.Expr->getNumElements() > 0;
// TODO: Entry value operations can currently not be combined with any
// other expressions, so we can't emit call site entries in those cases.
if (ShouldCombineExpressions && Expr->isEntryValue())
// If a parameter's call site value is produced by a chain of
// instructions we may have already created an expression for the
// parameter when walking through the instructions. Append that to the
// base expression.
const DIExpression *CombinedExpr =
ShouldCombineExpressions ? combineDIExpressions(Expr, Param.Expr)
: Expr;
assert((!CombinedExpr || CombinedExpr->isValid()) &&
"Combined debug expression is invalid");
DbgValueLoc DbgLocVal(CombinedExpr, DbgValueLocEntry(Val));
DbgCallSiteParam CSParm(Param.ParamReg, DbgLocVal);
/// Add \p Reg to the worklist, if it's not already present, and mark that the
/// given parameter registers' values can (potentially) be described using
/// that register and an debug expression.
static void addToFwdRegWorklist(FwdRegWorklist &Worklist, unsigned Reg,
const DIExpression *Expr,
ArrayRef<FwdRegParamInfo> ParamsToAdd) {
auto I = Worklist.insert({Reg, {}});
auto &ParamsForFwdReg = I.first->second;
for (auto Param : ParamsToAdd) {
[Param](const FwdRegParamInfo &D) {
return D.ParamReg == Param.ParamReg;
}) &&
"Same parameter described twice by forwarding reg");
// If a parameter's call site value is produced by a chain of
// instructions we may have already created an expression for the
// parameter when walking through the instructions. Append that to the
// new expression.
const DIExpression *CombinedExpr = combineDIExpressions(Expr, Param.Expr);
ParamsForFwdReg.push_back({Param.ParamReg, CombinedExpr});
/// Interpret values loaded into registers by \p CurMI.
static void interpretValues(const MachineInstr *CurMI,
FwdRegWorklist &ForwardedRegWorklist,
ParamSet &Params) {
const MachineFunction *MF = CurMI->getMF();
const DIExpression *EmptyExpr =
DIExpression::get(MF->getFunction().getContext(), {});
const auto &TRI = *MF->getSubtarget().getRegisterInfo();
const auto &TII = *MF->getSubtarget().getInstrInfo();
const auto &TLI = *MF->getSubtarget().getTargetLowering();
// If an instruction defines more than one item in the worklist, we may run
// into situations where a worklist register's value is (potentially)
// described by the previous value of another register that is also defined
// by that instruction.
// This can for example occur in cases like this:
// $r1 = mov 123
// $r0, $r1 = mvrr $r1, 456
// call @foo, $r0, $r1
// When describing $r1's value for the mvrr instruction, we need to make sure
// that we don't finalize an entry value for $r0, as that is dependent on the
// previous value of $r1 (123 rather than 456).
// In order to not have to distinguish between those cases when finalizing
// entry values, we simply postpone adding new parameter registers to the
// worklist, by first keeping them in this temporary container until the
// instruction has been handled.
FwdRegWorklist TmpWorklistItems;
// If the MI is an instruction defining one or more parameters' forwarding
// registers, add those defines.
auto getForwardingRegsDefinedByMI = [&](const MachineInstr &MI,
SmallSetVector<unsigned, 4> &Defs) {
if (MI.isDebugInstr())
for (const MachineOperand &MO : MI.operands()) {
if (MO.isReg() && MO.isDef() &&
Register::isPhysicalRegister(MO.getReg())) {
for (auto &FwdReg : ForwardedRegWorklist)
if (TRI.regsOverlap(FwdReg.first, MO.getReg()))
// Set of worklist registers that are defined by this instruction.
SmallSetVector<unsigned, 4> FwdRegDefs;
getForwardingRegsDefinedByMI(*CurMI, FwdRegDefs);
if (FwdRegDefs.empty())
for (auto ParamFwdReg : FwdRegDefs) {
if (auto ParamValue = TII.describeLoadedValue(*CurMI, ParamFwdReg)) {
if (ParamValue->first.isImm()) {
int64_t Val = ParamValue->first.getImm();
finishCallSiteParams(Val, ParamValue->second,
ForwardedRegWorklist[ParamFwdReg], Params);
} else if (ParamValue->first.isReg()) {
Register RegLoc = ParamValue->first.getReg();
Register SP = TLI.getStackPointerRegisterToSaveRestore();
Register FP = TRI.getFrameRegister(*MF);
bool IsSPorFP = (RegLoc == SP) || (RegLoc == FP);
if (TRI.isCalleeSavedPhysReg(RegLoc, *MF) || IsSPorFP) {
MachineLocation MLoc(RegLoc, /*Indirect=*/IsSPorFP);
finishCallSiteParams(MLoc, ParamValue->second,
ForwardedRegWorklist[ParamFwdReg], Params);
} else {
// ParamFwdReg was described by the non-callee saved register
// RegLoc. Mark that the call site values for the parameters are
// dependent on that register instead of ParamFwdReg. Since RegLoc
// may be a register that will be handled in this iteration, we
// postpone adding the items to the worklist, and instead keep them
// in a temporary container.
addToFwdRegWorklist(TmpWorklistItems, RegLoc, ParamValue->second,
// Remove all registers that this instruction defines from the worklist.
for (auto ParamFwdReg : FwdRegDefs)
// Now that we are done handling this instruction, add items from the
// temporary worklist to the real one.
for (auto &New : TmpWorklistItems)
addToFwdRegWorklist(ForwardedRegWorklist, New.first, EmptyExpr, New.second);
static bool interpretNextInstr(const MachineInstr *CurMI,
FwdRegWorklist &ForwardedRegWorklist,
ParamSet &Params) {
// Skip bundle headers.
if (CurMI->isBundle())
return true;
// If the next instruction is a call we can not interpret parameter's
// forwarding registers or we finished the interpretation of all
// parameters.
if (CurMI->isCall())
return false;
if (ForwardedRegWorklist.empty())
return false;
// Avoid NOP description.
if (CurMI->getNumOperands() == 0)
return true;
interpretValues(CurMI, ForwardedRegWorklist, Params);
return true;
/// Try to interpret values loaded into registers that forward parameters
/// for \p CallMI. Store parameters with interpreted value into \p Params.
static void collectCallSiteParameters(const MachineInstr *CallMI,
ParamSet &Params) {
const MachineFunction *MF = CallMI->getMF();
const auto &CalleesMap = MF->getCallSitesInfo();
auto CallFwdRegsInfo = CalleesMap.find(CallMI);
// There is no information for the call instruction.
if (CallFwdRegsInfo == CalleesMap.end())
const MachineBasicBlock *MBB = CallMI->getParent();
// Skip the call instruction.
auto I = std::next(CallMI->getReverseIterator());
FwdRegWorklist ForwardedRegWorklist;
const DIExpression *EmptyExpr =
DIExpression::get(MF->getFunction().getContext(), {});
// Add all the forwarding registers into the ForwardedRegWorklist.
for (const auto &ArgReg : CallFwdRegsInfo->second) {
bool InsertedReg =
ForwardedRegWorklist.insert({ArgReg.Reg, {{ArgReg.Reg, EmptyExpr}}})
assert(InsertedReg && "Single register used to forward two arguments?");
// Do not emit CSInfo for undef forwarding registers.
for (auto &MO : CallMI->uses())
if (MO.isReg() && MO.isUndef())
// We erase, from the ForwardedRegWorklist, those forwarding registers for
// which we successfully describe a loaded value (by using
// the describeLoadedValue()). For those remaining arguments in the working
// list, for which we do not describe a loaded value by
// the describeLoadedValue(), we try to generate an entry value expression
// for their call site value description, if the call is within the entry MBB.
// TODO: Handle situations when call site parameter value can be described
// as the entry value within basic blocks other than the first one.
bool ShouldTryEmitEntryVals = MBB->getIterator() == MF->begin();
// Search for a loading value in forwarding registers inside call delay slot.
if (CallMI->hasDelaySlot()) {
auto Suc = std::next(CallMI->getIterator());
// Only one-instruction delay slot is supported.
auto BundleEnd = llvm::getBundleEnd(CallMI->getIterator());
assert(std::next(Suc) == BundleEnd &&
"More than one instruction in call delay slot");
// Try to interpret value loaded by instruction.
if (!interpretNextInstr(&*Suc, ForwardedRegWorklist, Params))
// Search for a loading value in forwarding registers.
for (; I != MBB->rend(); ++I) {
// Try to interpret values loaded by instruction.
if (!interpretNextInstr(&*I, ForwardedRegWorklist, Params))
// Emit the call site parameter's value as an entry value.
if (ShouldTryEmitEntryVals) {
// Create an expression where the register's entry value is used.
DIExpression *EntryExpr = DIExpression::get(
MF->getFunction().getContext(), {dwarf::DW_OP_LLVM_entry_value, 1});
for (auto &RegEntry : ForwardedRegWorklist) {
MachineLocation MLoc(RegEntry.first);
finishCallSiteParams(MLoc, EntryExpr, RegEntry.second, Params);
void DwarfDebug::constructCallSiteEntryDIEs(const DISubprogram &SP,
DwarfCompileUnit &CU, DIE &ScopeDIE,
const MachineFunction &MF) {
// Add a call site-related attribute (DWARF5, Sec. Do this only if
// the subprogram is required to have one.
if (!SP.areAllCallsDescribed() || !SP.isDefinition())
// Use DW_AT_call_all_calls to express that call site entries are present
// for both tail and non-tail calls. Don't use DW_AT_call_all_source_calls
// because one of its requirements is not met: call site entries for
// optimized-out calls are elided.
CU.addFlag(ScopeDIE, CU.getDwarf5OrGNUAttr(dwarf::DW_AT_call_all_calls));
const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
assert(TII && "TargetInstrInfo not found: cannot label tail calls");
// Delay slot support check.
auto delaySlotSupported = [&](const MachineInstr &MI) {
if (!MI.isBundledWithSucc())
return false;
auto Suc = std::next(MI.getIterator());
auto CallInstrBundle = getBundleStart(MI.getIterator());
auto DelaySlotBundle = getBundleStart(Suc);
// Ensure that label after call is following delay slot instruction.
assert(getLabelAfterInsn(&*CallInstrBundle) ==
getLabelAfterInsn(&*DelaySlotBundle) &&
"Call and its successor instruction don't have same label after.");
return true;
// Emit call site entries for each call or tail call in the function.
for (const MachineBasicBlock &MBB : MF) {
for (const MachineInstr &MI : MBB.instrs()) {
// Bundles with call in them will pass the isCall() test below but do not
// have callee operand information so skip them here. Iterator will
// eventually reach the call MI.
if (MI.isBundle())
// Skip instructions which aren't calls. Both calls and tail-calling jump
// instructions (e.g TAILJMPd64) are classified correctly here.
if (!MI.isCandidateForCallSiteEntry())
// Skip instructions marked as frame setup, as they are not interesting to
// the user.
if (MI.getFlag(MachineInstr::FrameSetup))
// Check if delay slot support is enabled.
if (MI.hasDelaySlot() && !delaySlotSupported(*&MI))
// If this is a direct call, find the callee's subprogram.
// In the case of an indirect call find the register that holds
// the callee.
const MachineOperand &CalleeOp = TII->getCalleeOperand(MI);
if (!CalleeOp.isGlobal() &&
(!CalleeOp.isReg() ||
unsigned CallReg = 0;
const DISubprogram *CalleeSP = nullptr;
const Function *CalleeDecl = nullptr;
if (CalleeOp.isReg()) {
CallReg = CalleeOp.getReg();
if (!CallReg)
} else {
CalleeDecl = dyn_cast<Function>(CalleeOp.getGlobal());
if (!CalleeDecl || !CalleeDecl->getSubprogram())
CalleeSP = CalleeDecl->getSubprogram();
// TODO: Omit call site entries for runtime calls (objc_msgSend, etc).
bool IsTail = TII->isTailCall(MI);
// If MI is in a bundle, the label was created after the bundle since
// EmitFunctionBody iterates over top-level MIs. Get that top-level MI
// to search for that label below.
const MachineInstr *TopLevelCallMI =
MI.isInsideBundle() ? &*getBundleStart(MI.getIterator()) : &MI;
// For non-tail calls, the return PC is needed to disambiguate paths in
// the call graph which could lead to some target function. For tail
// calls, no return PC information is needed, unless tuning for GDB in
// DWARF4 mode in which case we fake a return PC for compatibility.
const MCSymbol *PCAddr =
(!IsTail || CU.useGNUAnalogForDwarf5Feature())
? const_cast<MCSymbol *>(getLabelAfterInsn(TopLevelCallMI))
: nullptr;
// For tail calls, it's necessary to record the address of the branch
// instruction so that the debugger can show where the tail call occurred.
const MCSymbol *CallAddr =
IsTail ? getLabelBeforeInsn(TopLevelCallMI) : nullptr;
assert((IsTail || PCAddr) && "Non-tail call without return PC");
LLVM_DEBUG(dbgs() << "CallSiteEntry: " << MF.getName() << " -> "
<< (CalleeDecl ? CalleeDecl->getName()
: StringRef(MF.getSubtarget()
<< (IsTail ? " [IsTail]" : "") << "\n");
DIE &CallSiteDIE = CU.constructCallSiteEntryDIE(
ScopeDIE, CalleeSP, IsTail, PCAddr, CallAddr, CallReg);
// Optionally emit call-site-param debug info.
if (emitDebugEntryValues()) {
ParamSet Params;
// Try to interpret values of call site parameters.
collectCallSiteParameters(&MI, Params);
CU.constructCallSiteParmEntryDIEs(CallSiteDIE, Params);
void DwarfDebug::addGnuPubAttributes(DwarfCompileUnit &U, DIE &D) const {
if (!U.hasDwarfPubSections())
U.addFlag(D, dwarf::DW_AT_GNU_pubnames);
void DwarfDebug::finishUnitAttributes(const DICompileUnit *DIUnit,
DwarfCompileUnit &NewCU) {
DIE &Die = NewCU.getUnitDie();
StringRef FN = DIUnit->getFilename();
StringRef Producer = DIUnit->getProducer();
StringRef Flags = DIUnit->getFlags();
if (!Flags.empty() && !useAppleExtensionAttributes()) {
std::string ProducerWithFlags = Producer.str() + " " + Flags.str();
NewCU.addString(Die, dwarf::DW_AT_producer, ProducerWithFlags);
} else
NewCU.addString(Die, dwarf::DW_AT_producer, Producer);
NewCU.addUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data2,
NewCU.addString(Die, dwarf::DW_AT_name, FN);
StringRef SysRoot = DIUnit->getSysRoot();
if (!SysRoot.empty())
NewCU.addString(Die, dwarf::DW_AT_LLVM_sysroot, SysRoot);
StringRef SDK = DIUnit->getSDK();
if (!SDK.empty())
NewCU.addString(Die, dwarf::DW_AT_APPLE_sdk, SDK);
// Add DW_str_offsets_base to the unit DIE, except for split units.
if (useSegmentedStringOffsetsTable() && !useSplitDwarf())
if (!useSplitDwarf()) {
// If we're using split dwarf the compilation dir is going to be in the
// skeleton CU and so we don't need to duplicate it here.
if (!CompilationDir.empty())
NewCU.addString(Die, dwarf::DW_AT_comp_dir, CompilationDir);
addGnuPubAttributes(NewCU, Die);
if (useAppleExtensionAttributes()) {
if (DIUnit->isOptimized())
NewCU.addFlag(Die, dwarf::DW_AT_APPLE_optimized);
StringRef Flags = DIUnit->getFlags();
if (!Flags.empty())
NewCU.addString(Die, dwarf::DW_AT_APPLE_flags, Flags);
if (unsigned RVer = DIUnit->getRuntimeVersion())
NewCU.addUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
dwarf::DW_FORM_data1, RVer);
if (DIUnit->getDWOId()) {
// This CU is either a clang module DWO or a skeleton CU.
NewCU.addUInt(Die, dwarf::DW_AT_GNU_dwo_id, dwarf::DW_FORM_data8,
if (!DIUnit->getSplitDebugFilename().empty()) {
// This is a prefabricated skeleton CU.
dwarf::Attribute attrDWOName = getDwarfVersion() >= 5
? dwarf::DW_AT_dwo_name
: dwarf::DW_AT_GNU_dwo_name;
NewCU.addString(Die, attrDWOName, DIUnit->getSplitDebugFilename());
// Create new DwarfCompileUnit for the given metadata node with tag
// DW_TAG_compile_unit.
DwarfCompileUnit &
DwarfDebug::getOrCreateDwarfCompileUnit(const DICompileUnit *DIUnit) {
if (auto *CU = CUMap.lookup(DIUnit))
return *CU;
CompilationDir = DIUnit->getDirectory();
auto OwnedUnit = std::make_unique<DwarfCompileUnit>(
InfoHolder.getUnits().size(), DIUnit, Asm, this, &InfoHolder);
DwarfCompileUnit &NewCU = *OwnedUnit;
for (auto *IE : DIUnit->getImportedEntities())
// LTO with assembly output shares a single line table amongst multiple CUs.
// To avoid the compilation directory being ambiguous, let the line table
// explicitly describe the directory of all files, never relying on the
// compilation directory.
if (!Asm->OutStreamer->hasRawTextSupport() || SingleCU)
CompilationDir, DIUnit->getFilename(), getMD5AsBytes(DIUnit->getFile()),
DIUnit->getSource(), NewCU.getUniqueID());
if (useSplitDwarf()) {
} else {
finishUnitAttributes(DIUnit, NewCU);
CUMap.insert({DIUnit, &NewCU});
CUDieMap.insert({&NewCU.getUnitDie(), &NewCU});
return NewCU;
void DwarfDebug::constructAndAddImportedEntityDIE(DwarfCompileUnit &TheCU,
const DIImportedEntity *N) {
if (isa<DILocalScope>(N->getScope()))
if (DIE *D = TheCU.getOrCreateContextDIE(N->getScope()))
/// Sort and unique GVEs by comparing their fragment offset.
static SmallVectorImpl<DwarfCompileUnit::GlobalExpr> &
sortGlobalExprs(SmallVectorImpl<DwarfCompileUnit::GlobalExpr> &GVEs) {
GVEs, [](DwarfCompileUnit::GlobalExpr A, DwarfCompileUnit::GlobalExpr B) {
// Sort order: first null exprs, then exprs without fragment
// info, then sort by fragment offset in bits.
// FIXME: Come up with a more comprehensive comparator so
// the sorting isn't non-deterministic, and so the following
// std::unique call works correctly.
if (!A.Expr || !B.Expr)
return !!B.Expr;
auto FragmentA = A.Expr->getFragmentInfo();
auto FragmentB = B.Expr->getFragmentInfo();
if (!FragmentA || !FragmentB)
return !!FragmentB;
return FragmentA->OffsetInBits < FragmentB->OffsetInBits;
GVEs.erase(std::unique(GVEs.begin(), GVEs.end(),
[](DwarfCompileUnit::GlobalExpr A,
DwarfCompileUnit::GlobalExpr B) {
return A.Expr == B.Expr;
return GVEs;
// Emit all Dwarf sections that should come prior to the content. Create
// global DIEs and emit initial debug info sections. This is invoked by
// the target AsmPrinter.
void DwarfDebug::beginModule(Module *M) {
if (!Asm || !MMI->hasDebugInfo())
unsigned NumDebugCUs = std::distance(M->debug_compile_units_begin(),
assert(NumDebugCUs > 0 && "Asm unexpectedly initialized");
assert(MMI->hasDebugInfo() &&
"DebugInfoAvailabilty unexpectedly not initialized");
SingleCU = NumDebugCUs == 1;
DenseMap<DIGlobalVariable *, SmallVector<DwarfCompileUnit::GlobalExpr, 1>>
for (const GlobalVariable &Global : M->globals()) {
SmallVector<DIGlobalVariableExpression *, 1> GVs;
for (auto *GVE : GVs)
GVMap[GVE->getVariable()].push_back({&Global, GVE->getExpression()});
// Create the symbol that designates the start of the unit's contribution
// to the string offsets table. In a split DWARF scenario, only the skeleton
// unit has the DW_AT_str_offsets_base attribute (and hence needs the symbol).
if (useSegmentedStringOffsetsTable())
(useSplitDwarf() ? SkeletonHolder : InfoHolder)
// Create the symbols that designates the start of the DWARF v5 range list
// and locations list tables. They are located past the table headers.
if (getDwarfVersion() >= 5) {
DwarfFile &Holder = useSplitDwarf() ? SkeletonHolder : InfoHolder;
if (useSplitDwarf())
// Create the symbol that points to the first entry following the debug
// address table (.debug_addr) header.
for (DICompileUnit *CUNode : M->debug_compile_units()) {
// FIXME: Move local imported entities into a list attached to the
// subprogram, then this search won't be needed and a
// getImportedEntities().empty() test should go below with the rest.
bool HasNonLocalImportedEntities = llvm::any_of(
CUNode->getImportedEntities(), [](const DIImportedEntity *IE) {
return !isa<DILocalScope>(IE->getScope());
if (!HasNonLocalImportedEntities && CUNode->getEnumTypes().empty() &&
CUNode->getRetainedTypes().empty() &&
CUNode->getGlobalVariables().empty() && CUNode->getMacros().empty())
DwarfCompileUnit &CU = getOrCreateDwarfCompileUnit(CUNode);
// Global Variables.
for (auto *GVE : CUNode->getGlobalVariables()) {
// Don't bother adding DIGlobalVariableExpressions listed in the CU if we
// already know about the variable and it isn't adding a constant
// expression.
auto &GVMapEntry = GVMap[GVE->getVariable()];
auto *Expr = GVE->getExpression();
if (!GVMapEntry.size() || (Expr && Expr->isConstant()))
GVMapEntry.push_back({nullptr, Expr});
DenseSet<DIGlobalVariable *> Processed;
for (auto *GVE : CUNode->getGlobalVariables()) {
DIGlobalVariable *GV = GVE->getVariable();
if (Processed.insert(GV).second)
CU.getOrCreateGlobalVariableDIE(GV, sortGlobalExprs(GVMap[GV]));
for (auto *Ty : CUNode->getEnumTypes()) {
// The enum types array by design contains pointers to
// MDNodes rather than DIRefs. Unique them here.
for (auto *Ty : CUNode->getRetainedTypes()) {
// The retained types array by design contains pointers to
// MDNodes rather than DIRefs. Unique them here.
if (DIType *RT = dyn_cast<DIType>(Ty))
// There is no point in force-emitting a forward declaration.
// Emit imported_modules last so that the relevant context is already
// available.
for (auto *IE : CUNode->getImportedEntities())
constructAndAddImportedEntityDIE(CU, IE);
void DwarfDebug::finishEntityDefinitions() {
for (const auto &Entity : ConcreteEntities) {
DIE *Die = Entity->getDIE();
// FIXME: Consider the time-space tradeoff of just storing the unit pointer
// in the ConcreteEntities list, rather than looking it up again here.
// DIE::getUnit isn't simple - it walks parent pointers, etc.
DwarfCompileUnit *Unit = CUDieMap.lookup(Die->getUnitDie());
void DwarfDebug::finishSubprogramDefinitions() {
for (const DISubprogram *SP : ProcessedSPNodes) {
assert(SP->getUnit()->getEmissionKind() != DICompileUnit::NoDebug);
[&](DwarfCompileUnit &CU) { CU.finishSubprogramDefinition(SP); });
void DwarfDebug::finalizeModuleInfo() {
const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
// Include the DWO file name in the hash if there's more than one CU.
// This handles ThinLTO's situation where imported CUs may very easily be
// duplicate with the same CU partially imported into another ThinLTO unit.
StringRef DWOName;
if (CUMap.size() > 1)
DWOName = Asm->TM.Options.MCOptions.SplitDwarfFile;
// Handle anything that needs to be done on a per-unit basis after
// all other generation.
for (const auto &P : CUMap) {
auto &TheCU = *P.second;
if (TheCU.getCUNode()->isDebugDirectivesOnly())
// Emit DW_AT_containing_type attribute to connect types with their
// vtable holding type.
// Add CU specific attributes if we need to add any.
// If we're splitting the dwarf out now that we've got the entire
// CU then add the dwo id to it.
auto *SkCU = TheCU.getSkeleton();
bool HasSplitUnit = SkCU && !TheCU.getUnitDie().children().empty();
if (HasSplitUnit) {
dwarf::Attribute attrDWOName = getDwarfVersion() >= 5
? dwarf::DW_AT_dwo_name
: dwarf::DW_AT_GNU_dwo_name;
finishUnitAttributes(TheCU.getCUNode(), TheCU);
TheCU.addString(TheCU.getUnitDie(), attrDWOName,
SkCU->addString(SkCU->getUnitDie(), attrDWOName,
// Emit a unique identifier for this CU.
uint64_t ID =
DIEHash(Asm, &TheCU).computeCUSignature(DWOName, TheCU.getUnitDie());
if (getDwarfVersion() >= 5) {
} else {
TheCU.addUInt(TheCU.getUnitDie(), dwarf::DW_AT_GNU_dwo_id,
dwarf::DW_FORM_data8, ID);
SkCU->addUInt(SkCU->getUnitDie(), dwarf::DW_AT_GNU_dwo_id,
dwarf::DW_FORM_data8, ID);
if (getDwarfVersion() < 5 && !SkeletonHolder.getRangeLists().empty()) {
const MCSymbol *Sym = TLOF.getDwarfRangesSection()->getBeginSymbol();
SkCU->addSectionLabel(SkCU->getUnitDie(), dwarf::DW_AT_GNU_ranges_base,
Sym, Sym);
} else if (SkCU) {
finishUnitAttributes(SkCU->getCUNode(), *SkCU);
// If we have code split among multiple sections or non-contiguous
// ranges of code then emit a DW_AT_ranges attribute on the unit that will
// remain in the .o file, otherwise add a DW_AT_low_pc.
// FIXME: We should use ranges allow reordering of code ala
// .subsections_via_symbols in mach-o. This would mean turning on
// ranges for all subprogram DIEs for mach-o.
DwarfCompileUnit &U = SkCU ? *SkCU : TheCU;
if (unsigned NumRanges = TheCU.getRanges().size()) {
if (NumRanges > 1 && useRangesSection())
// A DW_AT_low_pc attribute may also be specified in combination with
// DW_AT_ranges to specify the default base address for use in
// location lists (see Section 2.6.2) and range lists (see Section
// 2.17.3).
U.addUInt(U.getUnitDie(), dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr, 0);
U.attachRangesOrLowHighPC(U.getUnitDie(), TheCU.takeRanges());
// We don't keep track of which addresses are used in which CU so this
// is a bit pessimistic under LTO.
if ((HasSplitUnit || getDwarfVersion() >= 5) && !AddrPool.isEmpty())
if (getDwarfVersion() >= 5) {
if (U.hasRangeLists())
if (!DebugLocs.getLists().empty()) {
if (!useSplitDwarf())
U.addSectionLabel(U.getUnitDie(), dwarf::DW_AT_loclists_base,
auto *CUNode = cast<DICompileUnit>(P.first);
// If compile Unit has macros, emit "DW_AT_macro_info/DW_AT_macros"
// attribute.
if (CUNode->getMacros()) {
if (UseDebugMacroSection) {
if (useSplitDwarf())
TheCU.getUnitDie(), dwarf::DW_AT_macros, U.getMacroLabelBegin(),
else {
dwarf::Attribute MacrosAttr = getDwarfVersion() >= 5
? dwarf::DW_AT_macros
: dwarf::DW_AT_GNU_macros;
U.addSectionLabel(U.getUnitDie(), MacrosAttr, U.getMacroLabelBegin(),
} else {
if (useSplitDwarf())
TheCU.getUnitDie(), dwarf::DW_AT_macro_info,
U.addSectionLabel(U.getUnitDie(), dwarf::DW_AT_macro_info,
// Emit all frontend-produced Skeleton CUs, i.e., Clang modules.
for (auto *CUNode : MMI->getModule()->debug_compile_units())
if (CUNode->getDWOId())
// Compute DIE offsets and sizes.
if (useSplitDwarf())
// Emit all Dwarf sections that should come after the content.
void DwarfDebug::endModule() {
// Terminate the pending line table.
if (PrevCU)
PrevCU = nullptr;
assert(CurFn == nullptr);
assert(CurMI == nullptr);
for (const auto &P : CUMap) {
auto &CU = *P.second;
// If we aren't actually generating debug info (check beginModule -
// conditionalized on the presence of the metadata node)
if (!Asm || !MMI->hasDebugInfo())
// Finalize the debug info for the module.
if (useSplitDwarf())
// Emit debug_loc.dwo/debug_loclists.dwo section.
// Emit debug_loc/debug_loclists section.
// Corresponding abbreviations into a abbrev section.
// Emit all the DIEs into a debug info section.
// Emit info into a debug aranges section.
if (GenerateARangeSection)
// Emit info into a debug ranges section.
if (useSplitDwarf())
// Emit info into a debug macinfo.dwo section.
// Emit info into a debug macinfo/macro section.
if (useSplitDwarf()) {
// Emit info into the dwarf accelerator table sections.
switch (getAccelTableKind()) {
case AccelTableKind::Apple:
case AccelTableKind::Dwarf:
case AccelTableKind::None:
case AccelTableKind::Default:
llvm_unreachable("Default should have already been resolved.");
// Emit the pubnames and pubtypes sections if requested.
// clean up.
// FIXME: AbstractVariables.clear();
void DwarfDebug::ensureAbstractEntityIsCreated(DwarfCompileUnit &CU,
const DINode *Node,
const MDNode *ScopeNode) {
if (CU.getExistingAbstractEntity(Node))
CU.createAbstractEntity(Node, LScopes.getOrCreateAbstractScope(
void DwarfDebug::ensureAbstractEntityIsCreatedIfScoped(DwarfCompileUnit &CU,
const DINode *Node, const MDNode *ScopeNode) {
if (CU.getExistingAbstractEntity(Node))
if (LexicalScope *Scope =
CU.createAbstractEntity(Node, Scope);
// Collect variable information from side table maintained by MF.
void DwarfDebug::collectVariableInfoFromMFTable(
DwarfCompileUnit &TheCU, DenseSet<InlinedEntity> &Processed) {
SmallDenseMap<InlinedEntity, DbgVariable *> MFVars;
LLVM_DEBUG(dbgs() << "DwarfDebug: collecting variables from MF side table\n");
for (const auto &VI : Asm->MF->getVariableDbgInfo()) {
if (!VI.Var)
assert(VI.Var->isValidLocationForIntrinsic(VI.Loc) &&
"Expected inlined-at fields to agree");
InlinedEntity Var(VI.Var, VI.Loc->getInlinedAt());
LexicalScope *Scope = LScopes.findLexicalScope(VI.Loc);
// If variable scope is not found then skip this variable.
if (!Scope) {
LLVM_DEBUG(dbgs() << "Dropping debug info for " << VI.Var->getName()
<< ", no variable scope found\n");
ensureAbstractEntityIsCreatedIfScoped(TheCU, Var.first, Scope->getScopeNode());
auto RegVar = std::make_unique<DbgVariable>(
cast<DILocalVariable>(Var.first), Var.second);
RegVar->initializeMMI(VI.Expr, VI.Slot);
LLVM_DEBUG(dbgs() << "Created DbgVariable for " << VI.Var->getName()
<< "\n");
if (DbgVariable *DbgVar = MFVars.lookup(Var))
else if (InfoHolder.addScopeVariable(Scope, RegVar.get())) {
MFVars.insert({Var, RegVar.get()});
/// Determine whether a *singular* DBG_VALUE is valid for the entirety of its
/// enclosing lexical scope. The check ensures there are no other instructions
/// in the same lexical scope preceding the DBG_VALUE and that its range is
/// either open or otherwise rolls off the end of the scope.
static bool validThroughout(LexicalScopes &LScopes,
const MachineInstr *DbgValue,
const MachineInstr *RangeEnd,
const InstructionOrdering &Ordering) {
assert(DbgValue->getDebugLoc() && "DBG_VALUE without a debug location");
auto MBB = DbgValue->getParent();
auto DL = DbgValue->getDebugLoc();
auto *LScope = LScopes.findLexicalScope(DL);
// Scope doesn't exist; this is a dead DBG_VALUE.
if (!LScope)
return false;
auto &LSRange = LScope->getRanges();
if (LSRange.size() == 0)
return false;
const MachineInstr *LScopeBegin = LSRange.front().first;
// If the scope starts before the DBG_VALUE then we may have a negative
// result. Otherwise the location is live coming into the scope and we
// can skip the following checks.
if (!Ordering.isBefore(DbgValue, LScopeBegin)) {
// Exit if the lexical scope begins outside of the current block.
if (LScopeBegin->getParent() != MBB)
return false;
MachineBasicBlock::const_reverse_iterator Pred(DbgValue);
for (++Pred; Pred != MBB->rend(); ++Pred) {
if (Pred->getFlag(MachineInstr::FrameSetup))
auto PredDL = Pred->getDebugLoc();
if (!PredDL || Pred->isMetaInstruction())
// Check whether the instruction preceding the DBG_VALUE is in the same
// (sub)scope as the DBG_VALUE.
if (DL->getScope() == PredDL->getScope())
return false;
auto *PredScope = LScopes.findLexicalScope(PredDL);
if (!PredScope || LScope->dominates(PredScope))
return false;
// If the range of the DBG_VALUE is open-ended, report success.
if (!RangeEnd)
return true;
// Single, constant DBG_VALUEs in the prologue are promoted to be live
// throughout the function. This is a hack, presumably for DWARF v2 and not
// necessarily correct. It would be much better to use a dbg.declare instead
// if we know the constant is live throughout the scope.
if (MBB->pred_empty() &&
[](const MachineOperand &Op) { return Op.isImm(); }))
return true;
// Test if the location terminates before the end of the scope.
const MachineInstr *LScopeEnd = LSRange.back().second;
if (Ordering.isBefore(RangeEnd, LScopeEnd))
return false;
// There's a single location which starts at the scope start, and ends at or
// after the scope end.
return true;
/// Build the location list for all DBG_VALUEs in the function that
/// describe the same variable. The resulting DebugLocEntries will have
/// strict monotonically increasing begin addresses and will never
/// overlap. If the resulting list has only one entry that is valid
/// throughout variable's scope return true.
// See the definition of DbgValueHistoryMap::Entry for an explanation of the
// different kinds of history map entries. One thing to be aware of is that if
// a debug value is ended by another entry (rather than being valid until the
// end of the function), that entry's instruction may or may not be included in
// the range, depending on if the entry is a clobbering entry (it has an
// instruction that clobbers one or more preceding locations), or if it is an
// (overlapping) debug value entry. This distinction can be seen in the example
// below. The first debug value is ended by the clobbering entry 2, and the
// second and third debug values are ended by the overlapping debug value entry
// 4.
// Input:
// History map entries [type, end index, mi]
// 0 | [DbgValue, 2, DBG_VALUE $reg0, [...] (fragment 0, 32)]
// 1 | | [DbgValue, 4, DBG_VALUE $reg1, [...] (fragment 32, 32)]
// 2 | | [Clobber, $reg0 = [...], -, -]
// 3 | | [DbgValue, 4, DBG_VALUE 123, [...] (fragment 64, 32)]
// 4 [DbgValue, ~0, DBG_VALUE @g, [...] (fragment 0, 96)]
// Output [start, end) [Value...]:
// [0-1) [(reg0, fragment 0, 32)]
// [1-3) [(reg0, fragment 0, 32), (reg1, fragment 32, 32)]
// [3-4) [(reg1, fragment 32, 32), (123, fragment 64, 32)]
// [4-) [(@g, fragment 0, 96)]
bool DwarfDebug::buildLocationList(SmallVectorImpl<DebugLocEntry> &DebugLoc,
const DbgValueHistoryMap::Entries &Entries) {
using OpenRange =
std::pair<DbgValueHistoryMap::EntryIndex, DbgValueLoc>;
SmallVector<OpenRange, 4> OpenRanges;
bool isSafeForSingleLocation = true;
const MachineInstr *StartDebugMI = nullptr;
const MachineInstr *EndMI = nullptr;
for (auto EB = Entries.begin(), EI = EB, EE = Entries.end(); EI != EE; ++EI) {
const MachineInstr *Instr = EI->getInstr();
// Remove all values that are no longer live.
size_t Index = std::distance(EB, EI);
erase_if(OpenRanges, [&](OpenRange &R) { return R.first <= Index; });
// If we are dealing with a clobbering entry, this iteration will result in
// a location list entry starting after the clobbering instruction.
const MCSymbol *StartLabel =
EI->isClobber() ? getLabelAfterInsn(Instr) : getLabelBeforeInsn(Instr);
assert(StartLabel &&
"Forgot label before/after instruction starting a range!");
const MCSymbol *EndLabel;
if (std::next(EI) == Entries.end()) {
const MachineBasicBlock &EndMBB = Asm->MF->back();
EndLabel = Asm->MBBSectionRanges[EndMBB.getSectionIDNum()].EndLabel;
if (EI->isClobber())
EndMI = EI->getInstr();
else if (std::next(EI)->isClobber())
EndLabel = getLabelAfterInsn(std::next(EI)->getInstr());
EndLabel = getLabelBeforeInsn(std::next(EI)->getInstr());
assert(EndLabel && "Forgot label after instruction ending a range!");
if (EI->isDbgValue())
LLVM_DEBUG(dbgs() << "DotDebugLoc: " << *Instr << "\n");
// If this history map entry has a debug value, add that to the list of
// open ranges and check if its location is valid for a single value
// location.
if (EI->isDbgValue()) {
// Do not add undef debug values, as they are redundant information in
// the location list entries. An undef debug results in an empty location
// description. If there are any non-undef fragments then padding pieces
// with empty location descriptions will automatically be inserted, and if
// all fragments are undef then the whole location list entry is
// redundant.
if (!Instr->isUndefDebugValue()) {
auto Value = getDebugLocValue(Instr);
OpenRanges.emplace_back(EI->getEndIndex(), Value);
// TODO: Add support for single value fragment locations.
if (Instr->getDebugExpression()->isFragment())
isSafeForSingleLocation = false;
if (!StartDebugMI)
StartDebugMI = Instr;
} else {
isSafeForSingleLocation = false;
// Location list entries with empty location descriptions are redundant
// information in DWARF, so do not emit those.
if (OpenRanges.empty())
// Omit entries with empty ranges as they do not have any effect in DWARF.
if (StartLabel == EndLabel) {
LLVM_DEBUG(dbgs() << "Omitting location list entry with empty range.\n");
SmallVector<DbgValueLoc, 4> Values;
for (auto &R : OpenRanges)
// With Basic block sections, it is posssible that the StartLabel and the
// Instr are not in the same section. This happens when the StartLabel is
// the function begin label and the dbg value appears in a basic block
// that is not the entry. In this case, the range needs to be split to
// span each individual section in the range from StartLabel to EndLabel.
if (Asm->MF->hasBBSections() && StartLabel == Asm->getFunctionBegin() &&
!Instr->getParent()->sameSection(&Asm->MF->front())) {
const MCSymbol *BeginSectionLabel = StartLabel;
for (const MachineBasicBlock &MBB : *Asm->MF) {
if (MBB.isBeginSection() && &MBB != &Asm->MF->front())
BeginSectionLabel = MBB.getSymbol();
if (MBB.sameSection(Instr->getParent())) {
DebugLoc.emplace_back(BeginSectionLabel, EndLabel, Values);
if (MBB.isEndSection())
DebugLoc.emplace_back(BeginSectionLabel, MBB.getEndSymbol(), Values);
} else {
DebugLoc.emplace_back(StartLabel, EndLabel, Values);
// Attempt to coalesce the ranges of two otherwise identical
// DebugLocEntries.
auto CurEntry = DebugLoc.rbegin();
dbgs() << CurEntry->getValues().size() << " Values:\n";
for (auto &Value : CurEntry->getValues())
dbgs() << "-----\n";
auto PrevEntry = std::next(CurEntry);
if (PrevEntry != DebugLoc.rend() && PrevEntry->MergeRanges(*CurEntry))
if (!isSafeForSingleLocation ||
!validThroughout(LScopes, StartDebugMI, EndMI, getInstOrdering()))
return false;
if (DebugLoc.size() == 1)
return true;
if (!Asm->MF->hasBBSections())
return false;
// Check here to see if loclist can be merged into a single range. If not,
// we must keep the split loclists per section. This does exactly what
// MergeRanges does without sections. We don't actually merge the ranges
// as the split ranges must be kept intact if this cannot be collapsed
// into a single range.
const MachineBasicBlock *RangeMBB = nullptr;
if (DebugLoc[0].getBeginSym() == Asm->getFunctionBegin())
RangeMBB = &Asm->MF->front();
RangeMBB = Entries.begin()->getInstr()->getParent();
auto *CurEntry = DebugLoc.begin();
auto *NextEntry = std::next(CurEntry);
while (NextEntry != DebugLoc.end()) {
// Get the last machine basic block of this section.
while (!RangeMBB->isEndSection())
RangeMBB = RangeMBB->getNextNode();
if (!RangeMBB->getNextNode())
return false;
// CurEntry should end the current section and NextEntry should start
// the next section and the Values must match for these two ranges to be
// merged.
if (CurEntry->getEndSym() != RangeMBB->getEndSymbol() ||
NextEntry->getBeginSym() != RangeMBB->getNextNode()->getSymbol() ||
CurEntry->getValues() != NextEntry->getValues())
return false;
RangeMBB = RangeMBB->getNextNode();
CurEntry = NextEntry;
NextEntry = std::next(CurEntry);
return true;
DbgEntity *DwarfDebug::createConcreteEntity(DwarfCompileUnit &TheCU,
LexicalScope &Scope,
const DINode *Node,
const DILocation *Location,
const MCSymbol *Sym) {
ensureAbstractEntityIsCreatedIfScoped(TheCU, Node, Scope.getScopeNode());
if (isa<const DILocalVariable>(Node)) {
std::make_unique<DbgVariable>(cast<const DILocalVariable>(Node),
} else if (isa<const DILabel>(Node)) {
std::make_unique<DbgLabel>(cast<const DILabel>(Node),
Location, Sym));
return ConcreteEntities.back().get();
// Find variables for each lexical scope.
void DwarfDebug::collectEntityInfo(DwarfCompileUnit &TheCU,
const DISubprogram *SP,
DenseSet<InlinedEntity> &Processed) {
// Grab the variable info that was squirreled away in the MMI side-table.
collectVariableInfoFromMFTable(TheCU, Processed);
for (const auto &I : DbgValues) {
InlinedEntity IV = I.first;
if (Processed.count(IV))
// Instruction ranges, specifying where IV is accessible.
const auto &HistoryMapEntries = I.second;
// Try to find any non-empty variable location. Do not create a concrete
// entity if there are no locations.
if (!DbgValues.hasNonEmptyLocation(HistoryMapEntries))
LexicalScope *Scope = nullptr;
const DILocalVariable *LocalVar = cast<DILocalVariable>(IV.first);
if (const DILocation *IA = IV.second)
Scope = LScopes.findInlinedScope(LocalVar->getScope(), IA);
Scope = LScopes.findLexicalScope(LocalVar->getScope());
// If variable scope is not found then skip this variable.
if (!Scope)
DbgVariable *RegVar = cast<DbgVariable>(createConcreteEntity(TheCU,
*Scope, LocalVar, IV.second));
const MachineInstr *MInsn = HistoryMapEntries.front().getInstr();
assert(MInsn->isDebugValue() && "History must begin with debug value");
// Check if there is a single DBG_VALUE, valid throughout the var's scope.
// If the history map contains a single debug value, there may be an
// additional entry which clobbers the debug value.
size_t HistSize = HistoryMapEntries.size();
bool SingleValueWithClobber =
HistSize == 2 && HistoryMapEntries[1].isClobber();
if (HistSize == 1 || SingleValueWithClobber) {
const auto *End =
SingleValueWithClobber ? HistoryMapEntries[1].getInstr() : nullptr;
if (validThroughout(LScopes, MInsn, End, getInstOrdering())) {
// Do not emit location lists if .debug_loc secton is disabled.
if (!useLocSection())
// Handle multiple DBG_VALUE instructions describing one variable.
DebugLocStream::ListBuilder List(DebugLocs, TheCU, *Asm, *RegVar, *MInsn);
// Build the location list for this variable.
SmallVector<DebugLocEntry, 8> Entries;
bool isValidSingleLocation = buildLocationList(Entries, HistoryMapEntries);
// Check whether buildLocationList managed to merge all locations to one
// that is valid throughout the variable's scope. If so, produce single
// value location.
if (isValidSingleLocation) {
// If the variable has a DIBasicType, extract it. Basic types cannot have
// unique identifiers, so don't bother resolving the type with the
// identifier map.
const DIBasicType *BT = dyn_cast<DIBasicType>(
static_cast<const Metadata *>(LocalVar->getType()));
// Finalize the entry by lowering it into a DWARF bytestream.
for (auto &Entry : Entries)
Entry.finalize(*Asm, List, BT, TheCU);
// For each InlinedEntity collected from DBG_LABEL instructions, convert to
// DWARF-related DbgLabel.
for (const auto &I : DbgLabels) {
InlinedEntity IL = I.first;
const MachineInstr *MI = I.second;
if (MI == nullptr)
LexicalScope *Scope = nullptr;
const DILabel *Label = cast<DILabel>(IL.first);
// The scope could have an extra lexical block file.
const DILocalScope *LocalScope =
// Get inlined DILocation if it is inlined label.
if (const DILocation *IA = IL.second)
Scope = LScopes.findInlinedScope(LocalScope, IA);
Scope = LScopes.findLexicalScope(LocalScope);
// If label scope is not found then skip this label.
if (!Scope)
/// At this point, the temporary label is created.
/// Save the temporary label to DbgLabel entity to get the
/// actually address when generating Dwarf DIE.
MCSymbol *Sym = getLabelBeforeInsn(MI);
createConcreteEntity(TheCU, *Scope, Label, IL.second, Sym);
// Collect info for variables/labels that were optimized out.
for (const DINode *DN : SP->getRetainedNodes()) {
if (!Processed.insert(InlinedEntity(DN, nullptr)).second)
LexicalScope *Scope = nullptr;
if (auto *DV = dyn_cast<DILocalVariable>(DN)) {
Scope = LScopes.findLexicalScope(DV->getScope());
} else if (auto *DL = dyn_cast<DILabel>(DN)) {
Scope = LScopes.findLexicalScope(DL->getScope());
if (Scope)
createConcreteEntity(TheCU, *Scope, DN, nullptr);
// Process beginning of an instruction.
void DwarfDebug::beginInstruction(const MachineInstr *MI) {
const MachineFunction &MF = *MI->getMF();
const auto *SP = MF.getFunction().getSubprogram();
bool NoDebug =
!SP || SP->getUnit()->getEmissionKind() == DICompileUnit::NoDebug;
// Delay slot support check.
auto delaySlotSupported = [](const MachineInstr &MI) {
if (!MI.isBundledWithSucc())
return false;
auto Suc = std::next(MI.getIterator());
// Ensure that delay slot instruction is successor of the call instruction.
assert(Suc->isBundledWithPred() &&
"Call bundle instructions are out of order");
return true;
// When describing calls, we need a label for the call instruction.
if (!NoDebug && SP->areAllCallsDescribed() &&
MI->isCandidateForCallSiteEntry(MachineInstr::AnyInBundle) &&
(!MI->hasDelaySlot() || delaySlotSupported(*MI))) {
const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
bool IsTail = TII->isTailCall(*MI);
// For tail calls, we need the address of the branch instruction for
// DW_AT_call_pc.
if (IsTail)
// For non-tail calls, we need the return address for the call for
// DW_AT_call_return_pc. Under GDB tuning, this information is needed for
// tail calls as well.
if (!CurMI)
if (NoDebug)
// Check if source location changes, but ignore DBG_VALUE and CFI locations.
// If the instruction is part of the function frame setup code, do not emit
// any line record, as there is no correspondence with any user code.
if (MI->isMetaInstruction() || MI->getFlag(MachineInstr::FrameSetup))
const DebugLoc &DL = MI->getDebugLoc();
// When we emit a line-0 record, we don't update PrevInstLoc; so look at
// the last line number actually emitted, to see if it was line 0.
unsigned LastAsmLine =
if (DL == PrevInstLoc) {
// If we have an ongoing unspecified location, nothing to do here.
if (!DL)
// We have an explicit location, same as the previous location.
// But we might be coming back to it after a line 0 record.
if (LastAsmLine == 0 && DL.getLine() != 0) {
// Reinstate the source location but not marked as a statement.
const MDNode *Scope = DL.getScope();
recordSourceLine(DL.getLine(), DL.getCol(), Scope, /*Flags=*/0);
if (!DL) {
// We have an unspecified location, which might want to be line 0.
// If we have already emitted a line-0 record, don't repeat it.
if (LastAsmLine == 0)
// If user said Don't Do That, don't do that.
if (UnknownLocations == Disable)
// See if we have a reason to emit a line-0 record now.
// Reasons to emit a line-0 record include:
// - User asked for it (UnknownLocations).
// - Instruction has a label, so it's referenced from somewhere else,
// possibly debug information; we want it to have a source location.
// - Instruction is at the top of a block; we don't want to inherit the
// location from the physically previous (maybe unrelated) block.
if (UnknownLocations == Enable || PrevLabel ||
(PrevInstBB && PrevInstBB != MI->getParent())) {
// Preserve the file and column numbers, if we can, to save space in
// the encoded line table.
// Do not update PrevInstLoc, it remembers the last non-0 line.
const MDNode *Scope = nullptr;
unsigned Column = 0;
if (PrevInstLoc) {
Scope = PrevInstLoc.getScope();
Column = PrevInstLoc.getCol();
recordSourceLine(/*Line=*/0, Column, Scope, /*Flags=*/0);
// We have an explicit location, different from the previous location.
// Don't repeat a line-0 record, but otherwise emit the new location.
// (The new location might be an explicit line 0, which we do emit.)
if (DL.getLine() == 0 && LastAsmLine == 0)
unsigned Flags = 0;
if (DL == PrologEndLoc) {
PrologEndLoc = DebugLoc();
// If the line changed, we call that a new statement; unless we went to
// line 0 and came back, in which case it is not a new statement.
unsigned OldLine = PrevInstLoc ? PrevInstLoc.getLine() : LastAsmLine;
if (DL.getLine() && DL.getLine() != OldLine)
const MDNode *Scope = DL.getScope();
recordSourceLine(DL.getLine(), DL.getCol(), Scope, Flags);
// If we're not at line 0, remember this location.
if (DL.getLine())
PrevInstLoc = DL;
static DebugLoc findPrologueEndLoc(const MachineFunction *MF) {
// First known non-DBG_VALUE and non-frame setup location marks
// the beginning of the function body.
DebugLoc LineZeroLoc;
for (const auto &MBB : *MF) {
for (const auto &MI : MBB) {
if (!MI.isMetaInstruction() && !MI.getFlag(MachineInstr::FrameSetup) &&
MI.getDebugLoc()) {
// Scan forward to try to find a non-zero line number. The prologue_end
// marks the first breakpoint in the function after the frame setup, and
// a compiler-generated line 0 location is not a meaningful breakpoint.
// If none is found, return the first location after the frame setup.
if (MI.getDebugLoc().getLine())
return MI.getDebugLoc();
LineZeroLoc = MI.getDebugLoc();
return LineZeroLoc;
/// Register a source line with debug info. Returns the unique label that was
/// emitted and which provides correspondence to the source line list.
static void recordSourceLine(AsmPrinter &Asm, unsigned Line, unsigned Col,
const MDNode *S, unsigned Flags, unsigned CUID,
uint16_t DwarfVersion,
ArrayRef<std::unique_ptr<DwarfCompileUnit>> DCUs) {
StringRef Fn;
unsigned FileNo = 1;
unsigned Discriminator = 0;
if (auto *Scope = cast_or_null<DIScope>(S)) {
Fn = Scope->getFilename();
if (Line != 0 && DwarfVersion >= 4)
if (auto *LBF = dyn_cast<DILexicalBlockFile>(Scope))
Discriminator = LBF->getDiscriminator();
FileNo = static_cast<DwarfCompileUnit &>(*DCUs[CUID])
Asm.OutStreamer->emitDwarfLocDirective(FileNo, Line, Col, Flags, 0,
Discriminator, Fn);
DebugLoc DwarfDebug::emitInitialLocDirective(const MachineFunction &MF,
unsigned CUID) {
// Get beginning of function.
if (DebugLoc PrologEndLoc = findPrologueEndLoc(&MF)) {
// Ensure the compile unit is created if the function is called before
// beginFunction().
// We'd like to list the prologue as "not statements" but GDB behaves
// poorly if we do that. Revisit this with caution/GDB (7.5+) testing.
const DISubprogram *SP = PrologEndLoc->getInlinedAtScope()->getSubprogram();
::recordSourceLine(*Asm, SP->getScopeLine(), 0, SP, DWARF2_FLAG_IS_STMT,
CUID, getDwarfVersion(), getUnits());
return PrologEndLoc;
return DebugLoc();
// Gather pre-function debug information. Assumes being called immediately
// after the function entry point has been emitted.
void DwarfDebug::beginFunctionImpl(const MachineFunction *MF) {
CurFn = MF;
auto *SP = MF->getFunction().getSubprogram();
assert(LScopes.empty() || SP == LScopes.getCurrentFunctionScope()->getScopeNode());
if (SP->getUnit()->getEmissionKind() == DICompileUnit::NoDebug)
DwarfCompileUnit &CU = getOrCreateDwarfCompileUnit(SP->getUnit());
// Record beginning of function.
PrologEndLoc = emitInitialLocDirective(
*MF, Asm->OutStreamer->getContext().getDwarfCompileUnitID());
DwarfDebug::getDwarfCompileUnitIDForLineTable(const DwarfCompileUnit &CU) {
// Set DwarfDwarfCompileUnitID in MCContext to the Compile Unit this function
// belongs to so that we add to the correct per-cu line table in the
// non-asm case.
if (Asm->OutStreamer->hasRawTextSupport())
// Use a single line table if we are generating assembly.
return 0;
return CU.getUniqueID();
void DwarfDebug::terminateLineTable(const DwarfCompileUnit *CU) {
const auto &CURanges = CU->getRanges();
auto &LineTable = Asm->OutStreamer->getContext().getMCDwarfLineTable(
// Add the last range label for the given CU.
const_cast<MCSymbol *>(CURanges.back().End));
void DwarfDebug::skippedNonDebugFunction() {
// If we don't have a subprogram for this function then there will be a hole
// in the range information. Keep note of this by setting the previously used
// section to nullptr.
// Terminate the pending line table.
if (PrevCU)
PrevCU = nullptr;
CurFn = nullptr;
// Gather and emit post-function debug information.
void DwarfDebug::endFunctionImpl(const MachineFunction *MF) {
const DISubprogram *SP = MF->getFunction().getSubprogram();
assert(CurFn == MF &&
"endFunction should be called with the same function as beginFunction");
// Set DwarfDwarfCompileUnitID in MCContext to default value.
LexicalScope *FnScope = LScopes.getCurrentFunctionScope();
assert(!FnScope || SP == FnScope->getScopeNode());
DwarfCompileUnit &TheCU = *CUMap.lookup(SP->getUnit());
if (TheCU.getCUNode()->isDebugDirectivesOnly()) {
PrevLabel = nullptr;
CurFn = nullptr;
DenseSet<InlinedEntity> Processed;
collectEntityInfo(TheCU, SP, Processed);
// Add the range of this function to the list of ranges for the CU.
// With basic block sections, add ranges for all basic block sections.
for (const auto &R : Asm->MBBSectionRanges)
TheCU.addRange({R.second.BeginLabel, R.second.EndLabel});
// Under -gmlt, skip building the subprogram if there are no inlined
// subroutines inside it. But with -fdebug-info-for-profiling, the subprogram
// is still needed as we need its source location.
if (!TheCU.getCUNode()->getDebugInfoForProfiling() &&
TheCU.getCUNode()->getEmissionKind() == DICompileUnit::LineTablesOnly &&
LScopes.getAbstractScopesList().empty() && !IsDarwin) {
PrevLabel = nullptr;
CurFn = nullptr;
#ifndef NDEBUG
size_t NumAbstractScopes = LScopes.getAbstractScopesList().size();
// Construct abstract scopes.
for (LexicalScope *AScope : LScopes.getAbstractScopesList()) {
auto *SP = cast<DISubprogram>(AScope->getScopeNode());
for (const DINode *DN : SP->getRetainedNodes()) {
if (!Processed.insert(InlinedEntity(DN, nullptr)).second)
const MDNode *Scope = nullptr;
if (auto *DV = dyn_cast<DILocalVariable>(DN))
Scope = DV->getScope();
else if (auto *DL = dyn_cast<DILabel>(DN))
Scope = DL->getScope();
llvm_unreachable("Unexpected DI type!");
// Collect info for variables/labels that were optimized out.
ensureAbstractEntityIsCreated(TheCU, DN, Scope);
assert(LScopes.getAbstractScopesList().size() == NumAbstractScopes
&& "ensureAbstractEntityIsCreated inserted abstract scopes");
constructAbstractSubprogramScopeDIE(TheCU, AScope);
DIE &ScopeDIE = TheCU.constructSubprogramScopeDIE(SP, FnScope);
if (auto *SkelCU = TheCU.getSkeleton())
if (!LScopes.getAbstractScopesList().empty() &&
SkelCU->constructSubprogramScopeDIE(SP, FnScope);
// Construct call site entries.
constructCallSiteEntryDIEs(*SP, TheCU, ScopeDIE, *MF);
// Clear debug info
// Ownership of DbgVariables is a bit subtle - ScopeVariables owns all the
// DbgVariables except those that are also in AbstractVariables (since they
// can be used cross-function)
PrevLabel = nullptr;
CurFn = nullptr;
// Register a source line with debug info. Returns the unique label that was
// emitted and which provides correspondence to the source line list.
void DwarfDebug::recordSourceLine(unsigned Line, unsigned Col, const MDNode *S,
unsigned Flags) {
::recordSourceLine(*Asm, Line, Col, S, Flags,
getDwarfVersion(), getUnits());
// Emit Methods
// Emit the debug info section.
void DwarfDebug::emitDebugInfo() {
DwarfFile &Holder = useSplitDwarf() ? SkeletonHolder : InfoHolder;
Holder.emitUnits(/* UseOffsets */ false);
// Emit the abbreviation section.
void DwarfDebug::emitAbbreviations() {
DwarfFile &Holder = useSplitDwarf() ? SkeletonHolder : InfoHolder;
void DwarfDebug::emitStringOffsetsTableHeader() {
DwarfFile &Holder = useSplitDwarf() ? SkeletonHolder : InfoHolder;
*Asm, Asm->getObjFileLowering().getDwarfStrOffSection(),
template <typename AccelTableT>
void DwarfDebug::emitAccel(AccelTableT &Accel, MCSection *Section,
StringRef TableName) {
// Emit the full data.
emitAppleAccelTable(Asm, Accel, TableName, Section->getBeginSymbol());
void DwarfDebug::emitAccelDebugNames() {
// Don't emit anything if we have no compilation units to index.
if (getUnits().empty())
emitDWARF5AccelTable(Asm, AccelDebugNames, *this, getUnits());
// Emit visible names into a hashed accelerator table section.
void DwarfDebug::emitAccelNames() {
emitAccel(AccelNames, Asm->getObjFileLowering().getDwarfAccelNamesSection(),
// Emit objective C classes and categories into a hashed accelerator table
// section.
void DwarfDebug::emitAccelObjC() {
emitAccel(AccelObjC, Asm->getObjFileLowering().getDwarfAccelObjCSection(),
// Emit namespace dies into a hashed accelerator table.
void DwarfDebug::emitAccelNamespaces() {
// Emit type dies into a hashed accelerator table.
void DwarfDebug::emitAccelTypes() {
emitAccel(AccelTypes, Asm->getObjFileLowering().getDwarfAccelTypesSection(),
// Public name handling.
// The format for the various pubnames:
// dwarf pubnames - offset/name pairs where the offset is the offset into the CU
// for the DIE that is named.
// gnu pubnames - offset/index value/name tuples where the offset is the offset
// into the CU and the index value is computed according to the type of value
// for the DIE that is named.
// For type units the offset is the offset of the skeleton DIE. For split dwarf
// it's the offset within the debug_info/debug_types dwo section, however, the
// reference in the pubname header doesn't change.
/// computeIndexValue - Compute the gdb index value for the DIE and CU.
static dwarf::PubIndexEntryDescriptor computeIndexValue(DwarfUnit *CU,
const DIE *Die) {
// Entities that ended up only in a Type Unit reference the CU instead (since
// the pub entry has offsets within the CU there's no real offset that can be
// provided anyway). As it happens all such entities (namespaces and types,
// types only in C++ at that) are rendered as TYPE+EXTERNAL. If this turns out
// not to be true it would be necessary to persist this information from the
// point at which the entry is added to the index data structure - since by
// the time the index is built from that, the original type/namespace DIE in a
// type unit has already been destroyed so it can't be queried for properties
// like tag, etc.
if (Die->getTag() == dwarf::DW_TAG_compile_unit)
return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_TYPE,
dwarf::GDBIndexEntryLinkage Linkage = dwarf::GIEL_STATIC;
// We could have a specification DIE that has our most of our knowledge,
// look for that now.
if (DIEValue SpecVal = Die->findAttribute(dwarf::DW_AT_specification)) {
DIE &SpecDIE = SpecVal.getDIEEntry().getEntry();
if (SpecDIE.findAttribute(dwarf::DW_AT_external))
Linkage = dwarf::GIEL_EXTERNAL;
} else if (Die->findAttribute(dwarf::DW_AT_external))
Linkage = dwarf::GIEL_EXTERNAL;
switch (Die->getTag()) {
case dwarf::DW_TAG_class_type:
case dwarf::DW_TAG_structure_type:
case dwarf::DW_TAG_union_type:
case dwarf::DW_TAG_enumeration_type:
return dwarf::PubIndexEntryDescriptor(
: dwarf::GIEL_STATIC);
case dwarf::DW_TAG_typedef:
case dwarf::DW_TAG_base_type:
case dwarf::DW_TAG_subrange_type:
return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_TYPE, dwarf::GIEL_STATIC);
case dwarf::DW_TAG_namespace:
return dwarf::GIEK_TYPE;
case dwarf::DW_TAG_subprogram:
return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_FUNCTION, Linkage);
case dwarf::DW_TAG_variable:
return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_VARIABLE, Linkage);
case dwarf::DW_TAG_enumerator:
return dwarf::PubIndexEntryDescriptor(dwarf::GIEK_VARIABLE,
return dwarf::GIEK_NONE;
/// emitDebugPubSections - Emit visible names and types into debug pubnames and
/// pubtypes sections.
void DwarfDebug::emitDebugPubSections() {
for (const auto &NU : CUMap) {
DwarfCompileUnit *TheU = NU.second;
if (!TheU->hasDwarfPubSections())
bool GnuStyle = TheU->getCUNode()->getNameTableKind() ==
GnuStyle ? Asm->getObjFileLowering().getDwarfGnuPubNamesSection()
: Asm->getObjFileLowering().getDwarfPubNamesSection());
emitDebugPubSection(GnuStyle, "Names", TheU, TheU->getGlobalNames());
GnuStyle ? Asm->getObjFileLowering().getDwarfGnuPubTypesSection()
: Asm->getObjFileLowering().getDwarfPubTypesSection());
emitDebugPubSection(GnuStyle, "Types", TheU, TheU->getGlobalTypes());
void DwarfDebug::emitSectionReference(const DwarfCompileUnit &CU) {
if (useSectionsAsReferences())
void DwarfDebug::emitDebugPubSection(bool GnuStyle, StringRef Name,
DwarfCompileUnit *TheU,
const StringMap<const DIE *> &Globals) {
if (auto *Skeleton = TheU->getSkeleton())
TheU = Skeleton;
// Emit the header.
MCSymbol *EndLabel = Asm->emitDwarfUnitLength(
"pub" + Name, "Length of Public " + Name + " Info");
Asm->OutStreamer->AddComment("DWARF Version");
Asm->OutStreamer->AddComment("Offset of Compilation Unit Info");
Asm->OutStreamer->AddComment("Compilation Unit Length");
// Emit the pubnames for this compilation unit.
for (const auto &GI : Globals) {
const char *Name = GI.getKeyData();
const DIE *Entity = GI.second;
Asm->OutStreamer->AddComment("DIE offset");
if (GnuStyle) {
dwarf::PubIndexEntryDescriptor Desc = computeIndexValue(TheU, Entity);
Twine("Attributes: ") + dwarf::GDBIndexEntryKindString(Desc.Kind) +
", " + dwarf::GDBIndexEntryLinkageString(Desc.Linkage));
Asm->OutStreamer->AddComment("External Name");
Asm->OutStreamer->emitBytes(StringRef(Name, GI.getKeyLength() + 1));
Asm->OutStreamer->AddComment("End Mark");
/// Emit null-terminated strings into a debug str section.
void DwarfDebug::emitDebugStr() {
MCSection *StringOffsetsSection = nullptr;
if (useSegmentedStringOffsetsTable()) {
StringOffsetsSection = Asm->getObjFileLowering().getDwarfStrOffSection();
DwarfFile &Holder = useSplitDwarf() ? SkeletonHolder : InfoHolder;
StringOffsetsSection, /* UseRelativeOffsets = */ true);
void DwarfDebug::emitDebugLocEntry(ByteStreamer &Streamer,
const DebugLocStream::Entry &Entry,
const DwarfCompileUnit *CU) {
auto &&Comments = DebugLocs.getComments(Entry);
auto Comment = Comments.begin();
auto End = Comments.end();
// The expressions are inserted into a byte stream rather early (see
// DwarfExpression::addExpression) so for those ops (e.g. DW_OP_convert) that
// need to reference a base_type DIE the offset of that DIE is not yet known.
// To deal with this we instead insert a placeholder early and then extract
// it here and replace it with the real reference.
unsigned PtrSize = Asm->MAI->getCodePointerSize();
DWARFDataExtractor Data(StringRef(DebugLocs.getBytes(Entry).data(),
Asm->getDataLayout().isLittleEndian(), PtrSize);
DWARFExpression Expr(Data, PtrSize, Asm->OutContext.getDwarfFormat());
using Encoding = DWARFExpression::Operation::Encoding;
uint64_t Offset = 0;
for (auto &Op : Expr) {
assert(Op.getCode() != dwarf::DW_OP_const_type &&
"3 operand ops not yet supported");
Streamer.emitInt8(Op.getCode(), Comment != End ? *(Comment++) : "");
for (unsigned I = 0; I < 2; ++I) {
if (Op.getDescription().Op[I] == Encoding::SizeNA)
if (Op.getDescription().Op[I] == Encoding::BaseTypeRef) {
uint64_t Offset =
assert(Offset < (1ULL << (ULEB128PadSize * 7)) && "Offset wont fit");
Streamer.emitULEB128(Offset, "", ULEB128PadSize);
// Make sure comments stay aligned.
for (unsigned J = 0; J < ULEB128PadSize; ++J)
if (Comment != End)
} else {
for (uint64_t J = Offset; J < Op.getOperandEndOffset(I); ++J)
Streamer.emitInt8(Data.getData()[J], Comment != End ? *(Comment++) : "");
Offset = Op.getOperandEndOffset(I);
assert(Offset == Op.getEndOffset());
void DwarfDebug::emitDebugLocValue(const AsmPrinter &AP, const DIBasicType *BT,
const DbgValueLoc &Value,
DwarfExpression &DwarfExpr) {
auto *DIExpr = Value.getExpression();
DIExpressionCursor ExprCursor(DIExpr);
// If the DIExpr is is an Entry Value, we want to follow the same code path
// regardless of whether the DBG_VALUE is variadic or not.
if (DIExpr && DIExpr->isEntryValue()) {
// Entry values can only be a single register with no additional DIExpr,
// so just add it directly.
assert(Value.getLocEntries().size() == 1);
MachineLocation Location = Value.getLocEntries()[0].getLoc();
DwarfExpr.setLocation(Location, DIExpr);
const TargetRegisterInfo &TRI = *AP.MF->getSubtarget().getRegisterInfo();
if (!DwarfExpr.addMachineRegExpression(TRI, ExprCursor, Location.getReg()))
return DwarfExpr.addExpression(std::move(ExprCursor));
// Regular entry.
auto EmitValueLocEntry = [&DwarfExpr, &BT,
&AP](const DbgValueLocEntry &Entry,
DIExpressionCursor &Cursor) -> bool {
if (Entry.isInt()) {
if (BT && (BT->getEncoding() == dwarf::DW_ATE_signed ||
BT->getEncoding() == dwarf::DW_ATE_signed_char))
} else if (Entry.isLocation()) {
MachineLocation Location = Entry.getLoc();
if (Location.isIndirect())
const TargetRegisterInfo &TRI = *AP.MF->getSubtarget().getRegisterInfo();
if (!DwarfExpr.addMachineRegExpression(TRI, Cursor, Location.getReg()))
return false;
} else if (Entry.isTargetIndexLocation()) {
TargetIndexLocation Loc = Entry.getTargetIndexLocation();
// TODO TargetIndexLocation is a target-independent. Currently only the
// WebAssembly-specific encoding is supported.
DwarfExpr.addWasmLocation(Loc.Index, static_cast<uint64_t>(Loc.Offset));
} else if (Entry.isConstantFP()) {
if (AP.getDwarfVersion() >= 4 && !AP.getDwarfDebug()->tuneForSCE() &&
!Cursor) {
DwarfExpr.addConstantFP(Entry.getConstantFP()->getValueAPF(), AP);
} else if (Entry.getConstantFP()
.getBitWidth() <= 64 /*bits*/) {
} else {
dbgs() << "Skipped DwarfExpression creation for ConstantFP of size"
<< Entry.getConstantFP()
<< " bits\n");
return false;
return true;
if (!Value.isVariadic()) {
if (!EmitValueLocEntry(Value.getLocEntries()[0], ExprCursor))
// If any of the location entries are registers with the value 0, then the
// location is undefined.
if (any_of(Value.getLocEntries(), [](const DbgValueLocEntry &Entry) {
return Entry.isLocation() && !Entry.getLoc().getReg();
[EmitValueLocEntry, &Value](unsigned Idx,
DIExpressionCursor &Cursor) -> bool {
return EmitValueLocEntry(Value.getLocEntries()[Idx], Cursor);
void DebugLocEntry::finalize(const AsmPrinter &AP,
DebugLocStream::ListBuilder &List,
const DIBasicType *BT,
DwarfCompileUnit &TheCU) {
assert(!Values.empty() &&
"location list entries without values are redundant");
assert(Begin != End && "unexpected location list entry with empty range");
DebugLocStream::EntryBuilder Entry(List, Begin, End);
BufferByteStreamer Streamer = Entry.getStreamer();
DebugLocDwarfExpression DwarfExpr(AP.getDwarfVersion(), Streamer, TheCU);
const DbgValueLoc &Value = Values[0];
if (Value.isFragment()) {
// Emit all fragments that belong to the same variable and range.
assert(llvm::all_of(Values, [](DbgValueLoc P) {
return P.isFragment();
}) && "all values are expected to be fragments");
assert(llvm::is_sorted(Values) && "fragments are expected to be sorted");
for (const auto &Fragment : Values)
DwarfDebug::emitDebugLocValue(AP, BT, Fragment, DwarfExpr);
} else {
assert(Values.size() == 1 && "only fragments may have >1 value");
DwarfDebug::emitDebugLocValue(AP, BT, Value, DwarfExpr);
if (DwarfExpr.TagOffset)
void DwarfDebug::emitDebugLocEntryLocation(const DebugLocStream::Entry &Entry,
const DwarfCompileUnit *CU) {
// Emit the size.
Asm->OutStreamer->AddComment("Loc expr size");
if (getDwarfVersion() >= 5)
else if (DebugLocs.getBytes(Entry).size() <= std::numeric_limits<uint16_t>::max())
else {
// The entry is too big to fit into 16 bit, drop it as there is nothing we
// can do.