[DebugInfo] Add DILabel metadata and intrinsic llvm.dbg.label.

In order to set breakpoints on labels and list source code around
labels, we need collect debug information for labels, i.e., label
name, the function label belong, line number in the file, and the
address label located. In order to keep these information in LLVM
IR and to allow backend to generate debug information correctly.
We create a new kind of metadata for labels, DILabel. The format
of DILabel is

!DILabel(scope: !1, name: "foo", file: !2, line: 3)

We hope to keep debug information as much as possible even the
code is optimized. So, we create a new kind of intrinsic for label
metadata to avoid the metadata is eliminated with basic block.
The intrinsic will keep existing if we keep it from optimized out.
The format of the intrinsic is

llvm.dbg.label(metadata !1)

It has only one argument, that is the DILabel metadata. The
intrinsic will follow the label immediately. Backend could get the
label metadata through the intrinsic's parameter.

We also create DIBuilder API for labels to be used by Frontend.
Frontend could use createLabel() to allocate DILabel objects, and use
insertLabel() to insert llvm.dbg.label intrinsic in LLVM IR.

Differential Revision: https://reviews.llvm.org/D45024

Patch by Hsiangkai Wang.

llvm-svn: 331841

6 files changed, 174 insertions, 23 deletions

diff --git a/llvm/lib/IR/AsmWriter.cpp b/llvm/lib/IR/AsmWriter.cpp
index 08a14934b57..624350404ff 100644
--- a/llvm/lib/IR/AsmWriter.cpp
+++ b/llvm/lib/IR/AsmWriter.cpp
@@ -1831,7 +1831,7 @@ static void writeDISubprogram(raw_ostream &Out, const DISubprogram *N,
   Printer.printMetadata("unit", N->getRawUnit());
   Printer.printMetadata("templateParams", N->getRawTemplateParams());
   Printer.printMetadata("declaration", N->getRawDeclaration());
-  Printer.printMetadata("variables", N->getRawVariables());
+  Printer.printMetadata("retainedNodes", N->getRawRetainedNodes());
   Printer.printMetadata("thrownTypes", N->getRawThrownTypes());
   Out << ")";
 }
@@ -1971,6 +1971,18 @@ static void writeDILocalVariable(raw_ostream &Out, const DILocalVariable *N,
   Out << ")";
 }
 
+static void writeDILabel(raw_ostream &Out, const DILabel *N,
+                         TypePrinting *TypePrinter,
+                         SlotTracker *Machine, const Module *Context) {
+  Out << "!DILabel(";
+  MDFieldPrinter Printer(Out, TypePrinter, Machine, Context);
+  Printer.printMetadata("scope", N->getRawScope(), /* ShouldSkipNull */ false);
+  Printer.printString("name", N->getName());
+  Printer.printMetadata("file", N->getRawFile());
+  Printer.printInt("line", N->getLine());
+  Out << ")";
+}
+
 static void writeDIExpression(raw_ostream &Out, const DIExpression *N,
                               TypePrinting *TypePrinter, SlotTracker *Machine,
                               const Module *Context) {
diff --git a/llvm/lib/IR/DIBuilder.cpp b/llvm/lib/IR/DIBuilder.cpp
index 3e5ecdc2091..8596ebd9527 100644
--- a/llvm/lib/IR/DIBuilder.cpp
+++ b/llvm/lib/IR/DIBuilder.cpp
@@ -33,7 +33,7 @@ cl::opt<bool>
 
 DIBuilder::DIBuilder(Module &m, bool AllowUnresolvedNodes, DICompileUnit *CU)
   : M(m), VMContext(M.getContext()), CUNode(CU),
-      DeclareFn(nullptr), ValueFn(nullptr),
+      DeclareFn(nullptr), ValueFn(nullptr), LabelFn(nullptr),
       AllowUnresolvedNodes(AllowUnresolvedNodes) {}
 
 void DIBuilder::trackIfUnresolved(MDNode *N) {
@@ -47,18 +47,23 @@ void DIBuilder::trackIfUnresolved(MDNode *N) {
 }
 
 void DIBuilder::finalizeSubprogram(DISubprogram *SP) {
-  MDTuple *Temp = SP->getVariables().get();
+  MDTuple *Temp = SP->getRetainedNodes().get();
   if (!Temp || !Temp->isTemporary())
     return;
 
-  SmallVector<Metadata *, 4> Variables;
+  SmallVector<Metadata *, 16> RetainedNodes;
 
   auto PV = PreservedVariables.find(SP);
   if (PV != PreservedVariables.end())
-    Variables.append(PV->second.begin(), PV->second.end());
+    RetainedNodes.append(PV->second.begin(), PV->second.end());
 
-  DINodeArray AV = getOrCreateArray(Variables);
-  TempMDTuple(Temp)->replaceAllUsesWith(AV.get());
+  auto PL = PreservedLabels.find(SP);
+  if (PL != PreservedLabels.end())
+    RetainedNodes.append(PL->second.begin(), PL->second.end());
+
+  DINodeArray Node = getOrCreateArray(RetainedNodes);
+
+  TempMDTuple(Temp)->replaceAllUsesWith(Node.get());
 }
 
 void DIBuilder::finalize() {
@@ -699,6 +704,26 @@ DILocalVariable *DIBuilder::createParameterVariable(
                              /* AlignInBits */0);
 }
 
+DILabel *DIBuilder::createLabel(
+    DIScope *Scope, StringRef Name, DIFile *File,
+    unsigned LineNo, bool AlwaysPreserve) {
+  DIScope *Context = getNonCompileUnitScope(Scope);
+
+  auto *Node =
+      DILabel::get(VMContext, cast_or_null<DILocalScope>(Context), Name,
+                   File, LineNo);
+
+  if (AlwaysPreserve) {
+    /// The optimizer may remove labels. If there is an interest
+    /// to preserve label info in such situation then append it to
+    /// the list of retained nodes of the DISubprogram.
+    DISubprogram *Fn = getDISubprogram(Scope);
+    assert(Fn && "Missing subprogram for label");
+    PreservedLabels[Fn].emplace_back(Node);
+  }
+  return Node;
+}
+
 DIExpression *DIBuilder::createExpression(ArrayRef<uint64_t> Addr) {
   return DIExpression::get(VMContext, Addr);
 }
@@ -821,6 +846,18 @@ Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
   return insertDeclare(Storage, VarInfo, Expr, DL, InsertAtEnd, InsertBefore);
 }
 
+Instruction *DIBuilder::insertLabel(DILabel *LabelInfo, const DILocation *DL,
+                                    Instruction *InsertBefore) {
+  return insertLabel(
+      LabelInfo, DL, InsertBefore ? InsertBefore->getParent() : nullptr,
+      InsertBefore);
+}
+
+Instruction *DIBuilder::insertLabel(DILabel *LabelInfo, const DILocation *DL,
+                                    BasicBlock *InsertAtEnd) {
+  return insertLabel(LabelInfo, DL, InsertAtEnd, nullptr);
+}
+
 Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V,
                                                 DILocalVariable *VarInfo,
                                                 DIExpression *Expr,
@@ -906,6 +943,24 @@ Instruction *DIBuilder::insertDbgValueIntrinsic(
   return B.CreateCall(ValueFn, Args);
 }
 
+Instruction *DIBuilder::insertLabel(
+    DILabel *LabelInfo, const DILocation *DL,
+    BasicBlock *InsertBB, Instruction *InsertBefore) {
+  assert(LabelInfo && "empty or invalid DILabel* passed to dbg.label");
+  assert(DL && "Expected debug loc");
+  assert(DL->getScope()->getSubprogram() ==
+             LabelInfo->getScope()->getSubprogram() &&
+         "Expected matching subprograms");
+  if (!LabelFn)
+    LabelFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_label);
+
+  trackIfUnresolved(LabelInfo);
+  Value *Args[] = {MetadataAsValue::get(VMContext, LabelInfo)};
+
+  IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
+  return B.CreateCall(LabelFn, Args);
+}
+
 void DIBuilder::replaceVTableHolder(DICompositeType *&T,
                                     DIType *VTableHolder) {
   {
diff --git a/llvm/lib/IR/DebugInfo.cpp b/llvm/lib/IR/DebugInfo.cpp
index 1e771d03028..8cf38e74135 100644
--- a/llvm/lib/IR/DebugInfo.cpp
+++ b/llvm/lib/IR/DebugInfo.cpp
@@ -568,7 +568,7 @@ void DebugTypeInfoRemoval::traverse(MDNode *N) {
   // parts of the graph.
   auto prune = [](MDNode *Parent, MDNode *Child) {
     if (auto *MDS = dyn_cast<DISubprogram>(Parent))
-      return Child == MDS->getVariables().get();
+      return Child == MDS->getRetainedNodes().get();
     return false;
   };
 
diff --git a/llvm/lib/IR/DebugInfoMetadata.cpp b/llvm/lib/IR/DebugInfoMetadata.cpp
index bb497e458c0..a9f96e2549d 100644
--- a/llvm/lib/IR/DebugInfoMetadata.cpp
+++ b/llvm/lib/IR/DebugInfoMetadata.cpp
@@ -489,7 +489,7 @@ DISubprogram *DISubprogram::getImpl(
     bool IsLocalToUnit, bool IsDefinition, unsigned ScopeLine,
     Metadata *ContainingType, unsigned Virtuality, unsigned VirtualIndex,
     int ThisAdjustment, DIFlags Flags, bool IsOptimized, Metadata *Unit,
-    Metadata *TemplateParams, Metadata *Declaration, Metadata *Variables,
+    Metadata *TemplateParams, Metadata *Declaration, Metadata *RetainedNodes,
     Metadata *ThrownTypes, StorageType Storage, bool ShouldCreate) {
   assert(isCanonical(Name) && "Expected canonical MDString");
   assert(isCanonical(LinkageName) && "Expected canonical MDString");
@@ -497,10 +497,10 @@ DISubprogram *DISubprogram::getImpl(
       DISubprogram, (Scope, Name, LinkageName, File, Line, Type, IsLocalToUnit,
                      IsDefinition, ScopeLine, ContainingType, Virtuality,
                      VirtualIndex, ThisAdjustment, Flags, IsOptimized, Unit,
-                     TemplateParams, Declaration, Variables, ThrownTypes));
+                     TemplateParams, Declaration, RetainedNodes, ThrownTypes));
   SmallVector<Metadata *, 11> Ops = {
-      File,        Scope,     Name,           LinkageName,    Type,       Unit,
-      Declaration, Variables, ContainingType, TemplateParams, ThrownTypes};
+      File,        Scope,         Name,           LinkageName,    Type,       Unit,
+      Declaration, RetainedNodes, ContainingType, TemplateParams, ThrownTypes};
   if (!ThrownTypes) {
     Ops.pop_back();
     if (!TemplateParams) {
@@ -653,6 +653,18 @@ Optional<uint64_t> DIVariable::getSizeInBits() const {
   return None;
 }
 
+DILabel *DILabel::getImpl(LLVMContext &Context, Metadata *Scope,
+                          MDString *Name, Metadata *File, unsigned Line,
+                          StorageType Storage,
+                          bool ShouldCreate) {
+  assert(Scope && "Expected scope");
+  assert(isCanonical(Name) && "Expected canonical MDString");
+  DEFINE_GETIMPL_LOOKUP(DILabel,
+                        (Scope, Name, File, Line));
+  Metadata *Ops[] = {Scope, Name, File};
+  DEFINE_GETIMPL_STORE(DILabel, (Line), Ops);
+}
+
 DIExpression *DIExpression::getImpl(LLVMContext &Context,
                                     ArrayRef<uint64_t> Elements,
                                     StorageType Storage, bool ShouldCreate) {
diff --git a/llvm/lib/IR/LLVMContextImpl.h b/llvm/lib/IR/LLVMContextImpl.h
index 13579dcab08..7ceeca6e347 100644
--- a/llvm/lib/IR/LLVMContextImpl.h
+++ b/llvm/lib/IR/LLVMContextImpl.h
@@ -619,7 +619,7 @@ template <> struct MDNodeKeyImpl<DISubprogram> {
   Metadata *Unit;
   Metadata *TemplateParams;
   Metadata *Declaration;
-  Metadata *Variables;
+  Metadata *RetainedNodes;
   Metadata *ThrownTypes;
 
   MDNodeKeyImpl(Metadata *Scope, MDString *Name, MDString *LinkageName,
@@ -628,7 +628,7 @@ template <> struct MDNodeKeyImpl<DISubprogram> {
                 Metadata *ContainingType, unsigned Virtuality,
                 unsigned VirtualIndex, int ThisAdjustment, unsigned Flags,
                 bool IsOptimized, Metadata *Unit, Metadata *TemplateParams,
-                Metadata *Declaration, Metadata *Variables,
+                Metadata *Declaration, Metadata *RetainedNodes,
                 Metadata *ThrownTypes)
       : Scope(Scope), Name(Name), LinkageName(LinkageName), File(File),
         Line(Line), Type(Type), IsLocalToUnit(IsLocalToUnit),
@@ -637,7 +637,7 @@ template <> struct MDNodeKeyImpl<DISubprogram> {
         VirtualIndex(VirtualIndex), ThisAdjustment(ThisAdjustment),
         Flags(Flags), IsOptimized(IsOptimized), Unit(Unit),
         TemplateParams(TemplateParams), Declaration(Declaration),
-        Variables(Variables), ThrownTypes(ThrownTypes) {}
+        RetainedNodes(RetainedNodes), ThrownTypes(ThrownTypes) {}
   MDNodeKeyImpl(const DISubprogram *N)
       : Scope(N->getRawScope()), Name(N->getRawName()),
         LinkageName(N->getRawLinkageName()), File(N->getRawFile()),
@@ -648,7 +648,7 @@ template <> struct MDNodeKeyImpl<DISubprogram> {
         ThisAdjustment(N->getThisAdjustment()), Flags(N->getFlags()),
         IsOptimized(N->isOptimized()), Unit(N->getRawUnit()),
         TemplateParams(N->getRawTemplateParams()),
-        Declaration(N->getRawDeclaration()), Variables(N->getRawVariables()),
+        Declaration(N->getRawDeclaration()), RetainedNodes(N->getRawRetainedNodes()),
         ThrownTypes(N->getRawThrownTypes()) {}
 
   bool isKeyOf(const DISubprogram *RHS) const {
@@ -666,7 +666,7 @@ template <> struct MDNodeKeyImpl<DISubprogram> {
            Unit == RHS->getUnit() &&
            TemplateParams == RHS->getRawTemplateParams() &&
            Declaration == RHS->getRawDeclaration() &&
-           Variables == RHS->getRawVariables() &&
+           RetainedNodes == RHS->getRawRetainedNodes() &&
            ThrownTypes == RHS->getRawThrownTypes();
   }
 
@@ -948,6 +948,29 @@ template <> struct MDNodeKeyImpl<DILocalVariable> {
   }
 };
 
+template <> struct MDNodeKeyImpl<DILabel> {
+  Metadata *Scope;
+  MDString *Name;
+  Metadata *File;
+  unsigned Line;
+
+  MDNodeKeyImpl(Metadata *Scope, MDString *Name, Metadata *File, unsigned Line)
+      : Scope(Scope), Name(Name), File(File), Line(Line) {}
+  MDNodeKeyImpl(const DILabel *N)
+      : Scope(N->getRawScope()), Name(N->getRawName()), File(N->getRawFile()),
+        Line(N->getLine()) {}
+
+  bool isKeyOf(const DILabel *RHS) const {
+    return Scope == RHS->getRawScope() && Name == RHS->getRawName() &&
+           File == RHS->getRawFile() && Line == RHS->getLine();
+  }
+
+  /// Using name and line to get hash value. It should already be mostly unique.
+  unsigned getHashValue() const {
+    return hash_combine(Scope, Name, Line);
+  }
+};
+
 template <> struct MDNodeKeyImpl<DIExpression> {
   ArrayRef<uint64_t> Elements;
 
diff --git a/llvm/lib/IR/Verifier.cpp b/llvm/lib/IR/Verifier.cpp
index 770d21dfb7b..dca9d9b64ff 100644
--- a/llvm/lib/IR/Verifier.cpp
+++ b/llvm/lib/IR/Verifier.cpp
@@ -467,6 +467,7 @@ private:
   void visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS);
   void visitConstrainedFPIntrinsic(ConstrainedFPIntrinsic &FPI);
   void visitDbgIntrinsic(StringRef Kind, DbgInfoIntrinsic &DII);
+  void visitDbgLabelIntrinsic(StringRef Kind, DbgLabelInst &DLI);
   void visitAtomicCmpXchgInst(AtomicCmpXchgInst &CXI);
   void visitAtomicRMWInst(AtomicRMWInst &RMWI);
   void visitFenceInst(FenceInst &FI);
@@ -1085,12 +1086,13 @@ void Verifier::visitDISubprogram(const DISubprogram &N) {
   if (auto *S = N.getRawDeclaration())
     AssertDI(isa<DISubprogram>(S) && !cast<DISubprogram>(S)->isDefinition(),
              "invalid subprogram declaration", &N, S);
-  if (auto *RawVars = N.getRawVariables()) {
-    auto *Vars = dyn_cast<MDTuple>(RawVars);
-    AssertDI(Vars, "invalid variable list", &N, RawVars);
-    for (Metadata *Op : Vars->operands()) {
-      AssertDI(Op && isa<DILocalVariable>(Op), "invalid local variable", &N,
-               Vars, Op);
+  if (auto *RawNode = N.getRawRetainedNodes()) {
+    auto *Node = dyn_cast<MDTuple>(RawNode);
+    AssertDI(Node, "invalid retained nodes list", &N, RawNode);
+    for (Metadata *Op : Node->operands()) {
+      AssertDI(Op && (isa<DILocalVariable>(Op) || isa<DILabel>(Op)),
+               "invalid retained nodes, expected DILocalVariable or DILabel",
+               &N, Node, Op);
     }
   }
   AssertDI(!hasConflictingReferenceFlags(N.getFlags()),
@@ -1222,6 +1224,17 @@ void Verifier::visitDILocalVariable(const DILocalVariable &N) {
            "local variable requires a valid scope", &N, N.getRawScope());
 }
 
+void Verifier::visitDILabel(const DILabel &N) {
+  if (auto *S = N.getRawScope())
+    AssertDI(isa<DIScope>(S), "invalid scope", &N, S);
+  if (auto *F = N.getRawFile())
+    AssertDI(isa<DIFile>(F), "invalid file", &N, F);
+
+  AssertDI(N.getTag() == dwarf::DW_TAG_label, "invalid tag", &N);
+  AssertDI(N.getRawScope() && isa<DILocalScope>(N.getRawScope()),
+           "label requires a valid scope", &N, N.getRawScope());
+}
+
 void Verifier::visitDIExpression(const DIExpression &N) {
   AssertDI(N.isValid(), "invalid expression", &N);
 }
@@ -4065,6 +4078,9 @@ void Verifier::visitIntrinsicCallSite(Intrinsic::ID ID, CallSite CS) {
   case Intrinsic::dbg_value: // llvm.dbg.value
     visitDbgIntrinsic("value", cast<DbgInfoIntrinsic>(*CS.getInstruction()));
     break;
+  case Intrinsic::dbg_label: // llvm.dbg.label
+    visitDbgLabelIntrinsic("label", cast<DbgLabelInst>(*CS.getInstruction()));
+    break;
   case Intrinsic::memcpy:
   case Intrinsic::memmove:
   case Intrinsic::memset: {
@@ -4494,7 +4510,40 @@ void Verifier::visitDbgIntrinsic(StringRef Kind, DbgInfoIntrinsic &DII) {
   verifyFnArgs(DII);
 }
 
+void Verifier::visitDbgLabelIntrinsic(StringRef Kind, DbgLabelInst &DLI) {
+  AssertDI(isa<DILabel>(DLI.getRawVariable()),
+         "invalid llvm.dbg." + Kind + " intrinsic variable", &DLI,
+         DLI.getRawVariable());
+
+  // Ignore broken !dbg attachments; they're checked elsewhere.
+  if (MDNode *N = DLI.getDebugLoc().getAsMDNode())
+    if (!isa<DILocation>(N))
+      return;
+
+  BasicBlock *BB = DLI.getParent();
+  Function *F = BB ? BB->getParent() : nullptr;
+
+  // The scopes for variables and !dbg attachments must agree.
+  DILabel *Label = DLI.getLabel();
+  DILocation *Loc = DLI.getDebugLoc();
+  Assert(Loc, "llvm.dbg." + Kind + " intrinsic requires a !dbg attachment",
+         &DLI, BB, F);
+
+  DISubprogram *LabelSP = getSubprogram(Label->getRawScope());
+  DISubprogram *LocSP = getSubprogram(Loc->getRawScope());
+  if (!LabelSP || !LocSP)
+    return;
+
+  AssertDI(LabelSP == LocSP, "mismatched subprogram between llvm.dbg." + Kind +
+                             " label and !dbg attachment",
+           &DLI, BB, F, Label, Label->getScope()->getSubprogram(), Loc,
+           Loc->getScope()->getSubprogram());
+}
+
 void Verifier::verifyFragmentExpression(const DbgInfoIntrinsic &I) {
+  if (dyn_cast<DbgLabelInst>(&I))
+    return;
+
   DILocalVariable *V = dyn_cast_or_null<DILocalVariable>(I.getRawVariable());
   DIExpression *E = dyn_cast_or_null<DIExpression>(I.getRawExpression());