IR: Add COMDATs to the IR

This new IR facility allows us to represent the object-file semantic of
a COMDAT group.

COMDATs allow us to tie together sections and make the inclusion of one
dependent on another. This is required to implement features like MS
ABI VFTables and optimizing away certain kinds of initialization in C++.

This functionality is only representable in COFF and ELF, Mach-O has no
similar mechanism.

Differential Revision: http://reviews.llvm.org/D4178

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211920 91177308-0d34-0410-b5e6-96231b3b80d8

docs/LangRef.rst

@@ -562,6 +562,8 @@ is zero. The address space qualifier must precede any other attributes.
 
 LLVM allows an explicit section to be specified for globals. If the
 target supports it, it will emit globals to the section specified.
+Additionally, the global can placed in a comdat if the target has the necessary
+support.
 
 By default, global initializers are optimized by assuming that global
 variables defined within the module are not modified from their
@@ -627,8 +629,9 @@ an optional ``unnamed_addr`` attribute, a return type, an optional
 :ref:`parameter attribute <paramattrs>` for the return type, a function
 name, a (possibly empty) argument list (each with optional :ref:`parameter
 attributes <paramattrs>`), optional :ref:`function attributes <fnattrs>`,
-an optional section, an optional alignment, an optional :ref:`garbage
-collector name <gc>`, an optional :ref:`prefix <prefixdata>`, an opening
+an optional section, an optional alignment,
+an optional :ref:`comdat <langref_comdats>`,
+an optional :ref:`garbage collector name <gc>`, an optional :ref:`prefix <prefixdata>`, an opening
 curly brace, a list of basic blocks, and a closing curly brace.
 
 LLVM function declarations consist of the "``declare``" keyword, an
@@ -658,6 +661,7 @@ predecessors, it also cannot have any :ref:`PHI nodes <i_phi>`.
 
 LLVM allows an explicit section to be specified for functions. If the
 target supports it, it will emit functions to the section specified.
+Additionally, the function can placed in a COMDAT.
 
 An explicit alignment may be specified for a function. If not present,
 or if the alignment is set to zero, the alignment of the function is set
@@ -673,8 +677,8 @@ Syntax::
     define [linkage] [visibility] [DLLStorageClass]
            [cconv] [ret attrs]
            <ResultType> @<FunctionName> ([argument list])
-           [unnamed_addr] [fn Attrs] [section "name"] [align N]
-           [gc] [prefix Constant] { ... }
+           [unnamed_addr] [fn Attrs] [section "name"] [comdat $<ComdatName>]
+           [align N] [gc] [prefix Constant] { ... }
 
 .. _langref_aliases:
 
@@ -716,6 +720,89 @@ some can only be checked when producing an object file:
 * No global value in the expression can be a declaration, since that
   would require a relocation, which is not possible.
 
+.. _langref_comdats:
+
+Comdats
+-------
+
+Comdat IR provides access to COFF and ELF object file COMDAT functionality.
+
+Comdats have a name which represents the COMDAT key.  All global objects which
+specify this key will only end up in the final object file if the linker chooses
+that key over some other key.  Aliases are placed in the same COMDAT that their
+aliasee computes to, if any.
+
+Comdats have a selection kind to provide input on how the linker should
+choose between keys in two different object files.
+
+Syntax::
+
+    $<Name> = comdat SelectionKind
+
+The selection kind must be one of the following:
+
+``any``
+    The linker may choose any COMDAT key, the choice is arbitrary.
+``exactmatch``
+    The linker may choose any COMDAT key but the sections must contain the
+    same data.
+``largest``
+    The linker will choose the section containing the largest COMDAT key.
+``noduplicates``
+    The linker requires that only section with this COMDAT key exist.
+``samesize``
+    The linker may choose any COMDAT key but the sections must contain the
+    same amount of data.
+
+Note that the Mach-O platform doesn't support COMDATs and ELF only supports
+``any`` as a selection kind.
+
+Here is an example of a COMDAT group where a function will only be selected if
+the COMDAT key's section is the largest:
+
+.. code-block:: llvm
+
+   $foo = comdat largest
+   @foo = global i32 2, comdat $foo
+
+   define void @bar() comdat $foo {
+     ret void
+   }
+
+In a COFF object file, this will create a COMDAT section with selection kind
+``IMAGE_COMDAT_SELECT_LARGEST`` containing the contents of the ``@foo`` symbol
+and another COMDAT section with selection kind
+``IMAGE_COMDAT_SELECT_ASSOCIATIVE`` which is associated with the first COMDAT
+section and contains the contents of the ``@baz`` symbol.
+
+There are some restrictions on the properties of the global object.
+It, or an alias to it, must have the same name as the COMDAT group when
+targeting COFF.
+The contents and size of this object may be used during link-time to determine
+which COMDAT groups get selected depending on the selection kind.
+Because the name of the object must match the name of the COMDAT group, the
+linkage of the global object must not be local; local symbols can get renamed
+if a collision occurs in the symbol table.
+
+The combined use of COMDATS and section attributes may yield surprising results.
+For example:
+
+.. code-block:: llvm
+
+   $foo = comdat any
+   $bar = comdat any
+   @g1 = global i32 42, section "sec", comdat $foo
+   @g2 = global i32 42, section "sec", comdat $bar
+
+From the object file perspective, this requires the creation of two sections
+with the same name.  This is necessary because both globals belong to different
+COMDAT groups and COMDATs, at the object file level, are represented by
+sections.
+
+Note that certain IR constructs like global variables and functions may create
+COMDATs in the object file in addition to any which are specified using COMDAT
+IR.  This arises, for example, when a global variable has linkonce_odr linkage.
+
 .. _namedmetadatastructure:
 
 Named Metadata

include/llvm/ADT/UniqueVector.h

@@ -22,13 +22,18 @@ namespace llvm {
 /// class should have an implementation of operator== and of operator<.
 /// Entries can be fetched using operator[] with the entry ID.
 template<class T> class UniqueVector {
+public:
+  typedef typename std::vector<T> VectorType;
+  typedef typename VectorType::iterator iterator;
+  typedef typename VectorType::const_iterator const_iterator;
+
 private:
   // Map - Used to handle the correspondence of entry to ID.
   std::map<T, unsigned> Map;
 
   // Vector - ID ordered vector of entries. Entries can be indexed by ID - 1.
   //
-  std::vector<T> Vector;
+  VectorType Vector;
 
 public:
   /// insert - Append entry to the vector if it doesn't already exist.  Returns
@@ -68,6 +73,18 @@ public:
     return Vector[ID - 1];
   }
 
+  /// \brief Return an iterator to the start of the vector.
+  iterator begin() { return Vector.begin(); }
+
+  /// \brief Return an iterator to the start of the vector.
+  const_iterator begin() const { return Vector.begin(); }
+
+  /// \brief Return an iterator to the end of the vector.
+  iterator end() { return Vector.end(); }
+
+  /// \brief Return an iterator to the end of the vector.
+  const_iterator end() const { return Vector.end(); }
+
   /// size - Returns the number of entries in the vector.
   ///
   size_t size() const { return Vector.size(); }

include/llvm/Bitcode/LLVMBitCodes.h

@@ -71,7 +71,8 @@ namespace bitc {
     // MODULE_CODE_PURGEVALS: [numvals]
     MODULE_CODE_PURGEVALS   = 10,
 
-    MODULE_CODE_GCNAME      = 11   // GCNAME: [strchr x N]
+    MODULE_CODE_GCNAME      = 11,  // GCNAME: [strchr x N]
+    MODULE_CODE_COMDAT      = 12,  // COMDAT: [selection_kind, name]
   };
 
   /// PARAMATTR blocks have code for defining a parameter attribute set.
@@ -376,6 +377,14 @@ namespace bitc {
     ATTR_KIND_JUMP_TABLE = 40
   };
 
+  enum ComdatSelectionKindCodes {
+    COMDAT_SELECTION_KIND_ANY = 1,
+    COMDAT_SELECTION_KIND_EXACT_MATCH = 2,
+    COMDAT_SELECTION_KIND_LARGEST = 3,
+    COMDAT_SELECTION_KIND_NO_DUPLICATES = 4,
+    COMDAT_SELECTION_KIND_SAME_SIZE = 5,
+  };
+
 } // End bitc namespace
 } // End llvm namespace
 

include/llvm/IR/Comdat.h

@@ -0,0 +1,66 @@
+//===-- llvm/IR/Comdat.h - Comdat definitions -------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+/// @file
+/// This file contains the declaration of the Comdat class, which represents a
+/// single COMDAT in LLVM.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_IR_COMDAT_H
+#define LLVM_IR_COMDAT_H
+
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/Compiler.h"
+
+namespace llvm {
+
+class raw_ostream;
+template <typename ValueTy> class StringMapEntry;
+
+// This is a Name X SelectionKind pair. The reason for having this be an
+// independent object instead of just adding the name and the SelectionKind
+// to a GlobalObject is that it is invalid to have two Comdats with the same
+// name but different SelectionKind. This structure makes that unrepresentable.
+class Comdat {
+public:
+  enum SelectionKind {
+    Any,          ///< The linker may choose any COMDAT.
+    ExactMatch,   ///< The data referenced by the COMDAT must be the same.
+    Largest,      ///< The linker will choose the largest COMDAT.
+    NoDuplicates, ///< No other Module may specify this COMDAT.
+    SameSize,     ///< The data referenced by the COMDAT must be the same size.
+  };
+
+  Comdat(Comdat &&C);
+  SelectionKind getSelectionKind() const { return SK; }
+  void setSelectionKind(SelectionKind Val) { SK = Val; }
+  StringRef getName() const;
+  void print(raw_ostream &OS) const;
+  void dump() const;
+
+private:
+  friend class Module;
+  Comdat();
+  Comdat(SelectionKind SK, StringMapEntry<Comdat> *Name);
+  Comdat(const Comdat &) LLVM_DELETED_FUNCTION;
+
+  // Points to the map in Module.
+  StringMapEntry<Comdat> *Name;
+  SelectionKind SK;
+};
+
+inline raw_ostream &operator<<(raw_ostream &OS, const Comdat &C) {
+  C.print(OS);
+  return OS;
+}
+
+} // end llvm namespace
+
+#endif

include/llvm/IR/GlobalAlias.h

@@ -87,6 +87,21 @@ public:
     return getOperand(0);
   }
 
+  const GlobalObject *getBaseObject() const {
+    return const_cast<GlobalAlias *>(this)->getBaseObject();
+  }
+  GlobalObject *getBaseObject() {
+    return dyn_cast<GlobalObject>(getAliasee()->stripInBoundsOffsets());
+  }
+
+  const GlobalObject *getBaseObject(const DataLayout &DL, APInt &Offset) const {
+    return const_cast<GlobalAlias *>(this)->getBaseObject(DL, Offset);
+  }
+  GlobalObject *getBaseObject(const DataLayout &DL, APInt &Offset) {
+    return dyn_cast<GlobalObject>(
+        getAliasee()->stripAndAccumulateInBoundsConstantOffsets(DL, Offset));
+  }
+
   static bool isValidLinkage(LinkageTypes L) {
     return isExternalLinkage(L) || isLocalLinkage(L) ||
       isWeakLinkage(L) || isLinkOnceLinkage(L);

include/llvm/IR/GlobalObject.h

@@ -20,7 +20,7 @@
 #include "llvm/IR/GlobalValue.h"
 
 namespace llvm {
-
+class Comdat;
 class Module;
 
 class GlobalObject : public GlobalValue {
@@ -29,11 +29,12 @@ class GlobalObject : public GlobalValue {
 protected:
   GlobalObject(Type *Ty, ValueTy VTy, Use *Ops, unsigned NumOps,
                LinkageTypes Linkage, const Twine &Name)
-      : GlobalValue(Ty, VTy, Ops, NumOps, Linkage, Name) {
+      : GlobalValue(Ty, VTy, Ops, NumOps, Linkage, Name), ObjComdat(nullptr) {
     setGlobalValueSubClassData(0);
   }
 
   std::string Section;     // Section to emit this into, empty means default
+  Comdat *ObjComdat;
 public:
   unsigned getAlignment() const {
     return (1u << getGlobalValueSubClassData()) >> 1;
@@ -44,6 +45,11 @@ public:
   const char *getSection() const { return Section.c_str(); }
   void setSection(StringRef S);
 
+  bool hasComdat() const { return getComdat() != nullptr; }
+  const Comdat *getComdat() const { return ObjComdat; }
+  Comdat *getComdat() { return ObjComdat; }
+  void setComdat(Comdat *C) { ObjComdat = C; }
+
   void copyAttributesFrom(const GlobalValue *Src) override;
 
   // Methods for support type inquiry through isa, cast, and dyn_cast:

include/llvm/IR/GlobalValue.h

@@ -23,6 +23,7 @@
 
 namespace llvm {
 
+class Comdat;
 class PointerType;
 class Module;
 
@@ -110,6 +111,12 @@ public:
   bool hasUnnamedAddr() const { return UnnamedAddr; }
   void setUnnamedAddr(bool Val) { UnnamedAddr = Val; }
 
+  bool hasComdat() const { return getComdat() != nullptr; }
+  Comdat *getComdat();
+  const Comdat *getComdat() const {
+    return const_cast<GlobalValue *>(this)->getComdat();
+  }
+
   VisibilityTypes getVisibility() const { return VisibilityTypes(Visibility); }
   bool hasDefaultVisibility() const { return Visibility == DefaultVisibility; }
   bool hasHiddenVisibility() const { return Visibility == HiddenVisibility; }

include/llvm/IR/Module.h

@@ -16,6 +16,7 @@
 #define LLVM_IR_MODULE_H
 
 #include "llvm/ADT/iterator_range.h"
+#include "llvm/IR/Comdat.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalAlias.h"
@@ -123,6 +124,8 @@ public:
   typedef iplist<GlobalAlias> AliasListType;
   /// The type for the list of named metadata.
   typedef ilist<NamedMDNode> NamedMDListType;
+  /// The type of the comdat "symbol" table.
+  typedef StringMap<Comdat> ComdatSymTabType;
 
   /// The Global Variable iterator.
   typedef GlobalListType::iterator                      global_iterator;
@@ -197,6 +200,7 @@ private:
   NamedMDListType NamedMDList;    ///< The named metadata in the module
   std::string GlobalScopeAsm;     ///< Inline Asm at global scope.
   ValueSymbolTable *ValSymTab;    ///< Symbol table for values
+  ComdatSymTabType ComdatSymTab;  ///< Symbol table for COMDATs
   std::unique_ptr<GVMaterializer>
   Materializer;                   ///< Used to materialize GlobalValues
   std::string ModuleID;           ///< Human readable identifier for the module
@@ -403,6 +407,14 @@ public:
   /// Remove the given NamedMDNode from this module and delete it.
   void eraseNamedMetadata(NamedMDNode *NMD);
 
+/// @}
+/// @name Comdat Accessors
+/// @{
+
+  /// Return the Comdat in the module with the specified name. It is created
+  /// if it didn't already exist.
+  Comdat *getOrInsertComdat(StringRef Name);
+
 /// @}
 /// @name Module Flags Accessors
 /// @{
@@ -504,6 +516,10 @@ public:
   const ValueSymbolTable &getValueSymbolTable() const { return *ValSymTab; }
   /// Get the Module's symbol table of global variable and function identifiers.
   ValueSymbolTable       &getValueSymbolTable()       { return *ValSymTab; }
+  /// Get the Module's symbol table for COMDATs (constant).
+  const ComdatSymTabType &getComdatSymbolTable() const { return ComdatSymTab; }
+  /// Get the Module's symbol table for COMDATs.
+  ComdatSymTabType &getComdatSymbolTable() { return ComdatSymTab; }
 
 /// @}
 /// @name Global Variable Iteration

include/llvm/Linker/Linker.h

@@ -15,6 +15,8 @@
 
 namespace llvm {
 
+class Comdat;
+class GlobalValue;
 class Module;
 class StringRef;
 class StructType;

lib/AsmParser/LLLexer.cpp

@@ -209,6 +209,7 @@ lltok::Kind LLLexer::LexToken() {
     return LexToken();
   case '+': return LexPositive();
   case '@': return LexAt();
+  case '$': return LexDollar();
   case '%': return LexPercent();
   case '"': return LexQuote();
   case '.':
@@ -222,13 +223,6 @@ lltok::Kind LLLexer::LexToken() {
       return lltok::dotdotdot;
     }
     return lltok::Error;
-  case '$':
-    if (const char *Ptr = isLabelTail(CurPtr)) {
-      CurPtr = Ptr;
-      StrVal.assign(TokStart, CurPtr-1);
-      return lltok::LabelStr;
-    }
-    return lltok::Error;
   case ';':
     SkipLineComment();
     return LexToken();
@@ -307,6 +301,43 @@ lltok::Kind LLLexer::LexAt() {
   return lltok::Error;
 }
 
+lltok::Kind LLLexer::LexDollar() {
+  if (const char *Ptr = isLabelTail(TokStart)) {
+    CurPtr = Ptr;
+    StrVal.assign(TokStart, CurPtr - 1);
+    return lltok::LabelStr;
+  }
+
+  // Handle DollarStringConstant: $\"[^\"]*\"
+  if (CurPtr[0] == '"') {
+    ++CurPtr;
+
+    while (1) {
+      int CurChar = getNextChar();
+
+      if (CurChar == EOF) {
+        Error("end of file in COMDAT variable name");
+        return lltok::Error;
+      }
+      if (CurChar == '"') {
+        StrVal.assign(TokStart + 2, CurPtr - 1);
+        UnEscapeLexed(StrVal);
+        if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
+          Error("Null bytes are not allowed in names");
+          return lltok::Error;
+        }
+        return lltok::ComdatVar;
+      }
+    }
+  }
+
+  // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
+  if (ReadVarName())
+    return lltok::ComdatVar;
+
+  return lltok::Error;
+}
+
 /// ReadString - Read a string until the closing quote.
 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
   const char *Start = CurPtr;
@@ -618,6 +649,15 @@ lltok::Kind LLLexer::LexIdentifier() {
   KEYWORD(type);
   KEYWORD(opaque);
 
+  KEYWORD(comdat);
+
+  // Comdat types
+  KEYWORD(any);
+  KEYWORD(exactmatch);
+  KEYWORD(largest);
+  KEYWORD(noduplicates);
+  KEYWORD(samesize);
+
   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);

lib/AsmParser/LLLexer.h

@@ -81,6 +81,7 @@ namespace llvm {
     lltok::Kind LexDigitOrNegative();
     lltok::Kind LexPositive();
     lltok::Kind LexAt();
+    lltok::Kind LexDollar();
     lltok::Kind LexExclaim();
     lltok::Kind LexPercent();
     lltok::Kind LexQuote();

lib/AsmParser/LLParser.cpp

@@ -163,6 +163,11 @@ bool LLParser::ValidateEndOfModule() {
       return Error(I->second.second,
                    "use of undefined type named '" + I->getKey() + "'");
 
+  if (!ForwardRefComdats.empty())
+    return Error(ForwardRefComdats.begin()->second,
+                 "use of undefined comdat '$" +
+                     ForwardRefComdats.begin()->first + "'");
+
   if (!ForwardRefVals.empty())
     return Error(ForwardRefVals.begin()->second.second,
                  "use of undefined value '@" + ForwardRefVals.begin()->first +
@@ -238,6 +243,7 @@ bool LLParser::ParseTopLevelEntities() {
     case lltok::LocalVar:   if (ParseNamedType()) return true; break;
     case lltok::GlobalID:   if (ParseUnnamedGlobal()) return true; break;
     case lltok::GlobalVar:  if (ParseNamedGlobal()) return true; break;
+    case lltok::ComdatVar:  if (parseComdat()) return true; break;
     case lltok::exclaim:    if (ParseStandaloneMetadata()) return true; break;
     case lltok::MetadataVar:if (ParseNamedMetadata()) return true; break;
 
@@ -513,6 +519,56 @@ bool LLParser::ParseNamedGlobal() {
                     UnnamedAddr);
 }
 
+bool LLParser::parseComdat() {
+  assert(Lex.getKind() == lltok::ComdatVar);
+  std::string Name = Lex.getStrVal();
+  LocTy NameLoc = Lex.getLoc();
+  Lex.Lex();
+
+  if (ParseToken(lltok::equal, "expected '=' here"))
+    return true;
+
+  if (ParseToken(lltok::kw_comdat, "expected comdat keyword"))
+    return TokError("expected comdat type");
+
+  Comdat::SelectionKind SK;
+  switch (Lex.getKind()) {
+  default:
+    return TokError("unknown selection kind");
+  case lltok::kw_any:
+    SK = Comdat::Any;
+    break;
+  case lltok::kw_exactmatch:
+    SK = Comdat::ExactMatch;
+    break;
+  case lltok::kw_largest:
+    SK = Comdat::Largest;
+    break;
+  case lltok::kw_noduplicates:
+    SK = Comdat::NoDuplicates;
+    break;
+  case lltok::kw_samesize:
+    SK = Comdat::SameSize;
+    break;
+  }
+  Lex.Lex();
+
+  // See if the comdat was forward referenced, if so, use the comdat.
+  Module::ComdatSymTabType &ComdatSymTab = M->getComdatSymbolTable();
+  Module::ComdatSymTabType::iterator I = ComdatSymTab.find(Name);
+  if (I != ComdatSymTab.end() && !ForwardRefComdats.erase(Name))
+    return Error(NameLoc, "redefinition of comdat '$" + Name + "'");
+
+  Comdat *C;
+  if (I != ComdatSymTab.end())
+    C = &I->second;
+  else
+    C = M->getOrInsertComdat(Name);
+  C->setSelectionKind(SK);
+
+  return false;
+}
+
 // MDString:
 //   ::= '!' STRINGCONSTANT
 bool LLParser::ParseMDString(MDString *&Result) {
@@ -838,7 +894,13 @@ bool LLParser::ParseGlobal(const std::string &Name, LocTy NameLoc,
       if (ParseOptionalAlignment(Alignment)) return true;
       GV->setAlignment(Alignment);
     } else {
-      TokError("unknown global variable property!");
+      Comdat *C;
+      if (parseOptionalComdat(C))
+        return true;
+      if (C)
+        GV->setComdat(C);
+      else
+        return TokError("unknown global variable property!");
     }
   }
 
@@ -1096,6 +1158,24 @@ GlobalValue *LLParser::GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc) {
 }
 
 
+//===----------------------------------------------------------------------===//
+// Comdat Reference/Resolution Routines.
+//===----------------------------------------------------------------------===//
+
+Comdat *LLParser::getComdat(const std::string &Name, LocTy Loc) {
+  // Look this name up in the comdat symbol table.
+  Module::ComdatSymTabType &ComdatSymTab = M->getComdatSymbolTable();
+  Module::ComdatSymTabType::iterator I = ComdatSymTab.find(Name);
+  if (I != ComdatSymTab.end())
+    return &I->second;
+
+  // Otherwise, create a new forward reference for this value and remember it.
+  Comdat *C = M->getOrInsertComdat(Name);
+  ForwardRefComdats[Name] = Loc;
+  return C;
+}
+
+
 //===----------------------------------------------------------------------===//
 // Helper Routines.
 //===----------------------------------------------------------------------===//
@@ -2790,6 +2870,19 @@ bool LLParser::ParseGlobalTypeAndValue(Constant *&V) {
          ParseGlobalValue(Ty, V);
 }
 
+bool LLParser::parseOptionalComdat(Comdat *&C) {
+  C = nullptr;
+  if (!EatIfPresent(lltok::kw_comdat))
+    return false;
+  if (Lex.getKind() != lltok::ComdatVar)
+    return TokError("expected comdat variable");
+  LocTy Loc = Lex.getLoc();
+  StringRef Name = Lex.getStrVal();
+  C = getComdat(Name, Loc);
+  Lex.Lex();
+  return false;
+}
+
 /// ParseGlobalValueVector
 ///   ::= /*empty*/
 ///   ::= TypeAndValue (',' TypeAndValue)*
@@ -3090,6 +3183,7 @@ bool LLParser::ParseFunctionHeader(Function *&Fn, bool isDefine) {
   bool UnnamedAddr;
   LocTy UnnamedAddrLoc;
   Constant *Prefix = nullptr;
+  Comdat *C;
 
   if (ParseArgumentList(ArgList, isVarArg) ||
       ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr,
@@ -3098,6 +3192,7 @@ bool LLParser::ParseFunctionHeader(Function *&Fn, bool isDefine) {
                                  BuiltinLoc) ||
       (EatIfPresent(lltok::kw_section) &&
        ParseStringConstant(Section)) ||
+      parseOptionalComdat(C) ||
       ParseOptionalAlignment(Alignment) ||
       (EatIfPresent(lltok::kw_gc) &&
        ParseStringConstant(GC)) ||
@@ -3200,6 +3295,7 @@ bool LLParser::ParseFunctionHeader(Function *&Fn, bool isDefine) {
   Fn->setUnnamedAddr(UnnamedAddr);
   Fn->setAlignment(Alignment);
   Fn->setSection(Section);
+  Fn->setComdat(C);
   if (!GC.empty()) Fn->setGC(GC.c_str());
   Fn->setPrefixData(Prefix);
   ForwardRefAttrGroups[Fn] = FwdRefAttrGrps;

lib/AsmParser/LLParser.h

@@ -34,6 +34,7 @@ namespace llvm {
   class Instruction;
   class Constant;
   class GlobalValue;
+  class Comdat;
   class MDString;
   class MDNode;
   class StructType;
@@ -122,6 +123,9 @@ namespace llvm {
     std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
     std::vector<GlobalValue*> NumberedVals;
 
+    // Comdat forward reference information.
+    std::map<std::string, LocTy> ForwardRefComdats;
+
     // References to blockaddress.  The key is the function ValID, the value is
     // a list of references to blocks in that function.
     std::map<ValID, std::vector<std::pair<ValID, GlobalValue*> > >
@@ -154,6 +158,10 @@ namespace llvm {
     GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
     GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
 
+    /// Get a Comdat with the specified name, creating a forward reference
+    /// record if needed.
+    Comdat *getComdat(const std::string &N, LocTy Loc);
+
     // Helper Routines.
     bool ParseToken(lltok::Kind T, const char *ErrMsg);
     bool EatIfPresent(lltok::Kind T) {
@@ -247,6 +255,7 @@ namespace llvm {
     bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility,
                     unsigned DLLStorageClass,
                     GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
+    bool parseComdat();
     bool ParseStandaloneMetadata();
     bool ParseNamedMetadata();
     bool ParseMDString(MDString *&Result);
@@ -358,6 +367,7 @@ namespace llvm {
     bool ParseGlobalValue(Type *Ty, Constant *&V);
     bool ParseGlobalTypeAndValue(Constant *&V);
     bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
+    bool parseOptionalComdat(Comdat *&C);
     bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
     bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
     bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);

lib/AsmParser/LLToken.h

@@ -142,6 +142,15 @@ namespace lltok {
     kw_type,
     kw_opaque,
 
+    kw_comdat,
+
+    // Comdat types
+    kw_any,
+    kw_exactmatch,
+    kw_largest,
+    kw_noduplicates,
+    kw_samesize,
+
     kw_eq, kw_ne, kw_slt, kw_sgt, kw_sle, kw_sge, kw_ult, kw_ugt, kw_ule,
     kw_uge, kw_oeq, kw_one, kw_olt, kw_ogt, kw_ole, kw_oge, kw_ord, kw_uno,
     kw_ueq, kw_une,
@@ -180,6 +189,7 @@ namespace lltok {
     // String valued tokens (StrVal).
     LabelStr,          // foo:
     GlobalVar,         // @foo @"foo"
+    ComdatVar,         // $foo
     LocalVar,          // %foo %"foo"
     MetadataVar,       // !foo
     StringConstant,    // "foo"

lib/Bitcode/Reader/BitcodeReader.cpp

@@ -43,6 +43,7 @@ void BitcodeReader::FreeState() {
   std::vector<Type*>().swap(TypeList);
   ValueList.clear();
   MDValueList.clear();
+  std::vector<Comdat *>().swap(ComdatList);
 
   std::vector<AttributeSet>().swap(MAttributes);
   std::vector<BasicBlock*>().swap(FunctionBBs);
@@ -203,6 +204,22 @@ static SynchronizationScope GetDecodedSynchScope(unsigned Val) {
   }
 }
 
+static Comdat::SelectionKind getDecodedComdatSelectionKind(unsigned Val) {
+  switch (Val) {
+  default: // Map unknown selection kinds to any.
+  case bitc::COMDAT_SELECTION_KIND_ANY:
+    return Comdat::Any;
+  case bitc::COMDAT_SELECTION_KIND_EXACT_MATCH:
+    return Comdat::ExactMatch;
+  case bitc::COMDAT_SELECTION_KIND_LARGEST:
+    return Comdat::Largest;
+  case bitc::COMDAT_SELECTION_KIND_NO_DUPLICATES:
+    return Comdat::NoDuplicates;
+  case bitc::COMDAT_SELECTION_KIND_SAME_SIZE:
+    return Comdat::SameSize;
+  }
+}
+
 static void UpgradeDLLImportExportLinkage(llvm::GlobalValue *GV, unsigned Val) {
   switch (Val) {
   case 5: GV->setDLLStorageClass(GlobalValue::DLLImportStorageClass); break;
@@ -1838,6 +1855,20 @@ std::error_code BitcodeReader::ParseModule(bool Resume) {
       GCTable.push_back(S);
       break;
     }
+    case bitc::MODULE_CODE_COMDAT: { // COMDAT: [selection_kind, name]
+      if (Record.size() < 2)
+        return Error(InvalidRecord);
+      Comdat::SelectionKind SK = getDecodedComdatSelectionKind(Record[0]);
+      unsigned ComdatNameSize = Record[1];
+      std::string ComdatName;
+      ComdatName.reserve(ComdatNameSize);
+      for (unsigned i = 0; i != ComdatNameSize; ++i)
+        ComdatName += (char)Record[2 + i];
+      Comdat *C = TheModule->getOrInsertComdat(ComdatName);
+      C->setSelectionKind(SK);
+      ComdatList.push_back(C);
+      break;
+    }
     // GLOBALVAR: [pointer type, isconst, initid,
     //             linkage, alignment, section, visibility, threadlocal,
     //             unnamed_addr, dllstorageclass]
@@ -1898,6 +1929,12 @@ std::error_code BitcodeReader::ParseModule(bool Resume) {
       // Remember which value to use for the global initializer.
       if (unsigned InitID = Record[2])
         GlobalInits.push_back(std::make_pair(NewGV, InitID-1));
+
+      if (Record.size() > 11)
+        if (unsigned ComdatID = Record[11]) {
+          assert(ComdatID <= ComdatList.size());
+          NewGV->setComdat(ComdatList[ComdatID - 1]);
+        }
       break;
     }
     // FUNCTION:  [type, callingconv, isproto, linkage, paramattr,
@@ -1951,6 +1988,12 @@ std::error_code BitcodeReader::ParseModule(bool Resume) {
       else
         UpgradeDLLImportExportLinkage(Func, Record[3]);
 
+      if (Record.size() > 12)
+        if (unsigned ComdatID = Record[12]) {
+          assert(ComdatID <= ComdatList.size());
+          Func->setComdat(ComdatList[ComdatID - 1]);
+        }
+
       ValueList.push_back(Func);
 
       // If this is a function with a body, remember the prototype we are

lib/Bitcode/Reader/BitcodeReader.h

@@ -26,6 +26,7 @@
 #include <vector>
 
 namespace llvm {
+  class Comdat;
   class MemoryBuffer;
   class LLVMContext;
 
@@ -135,6 +136,7 @@ class BitcodeReader : public GVMaterializer {
   std::vector<Type*> TypeList;
   BitcodeReaderValueList ValueList;
   BitcodeReaderMDValueList MDValueList;
+  std::vector<Comdat *> ComdatList;
   SmallVector<Instruction *, 64> InstructionList;
   SmallVector<SmallVector<uint64_t, 64>, 64> UseListRecords;
 

lib/Bitcode/Writer/BitcodeWriter.cpp

@@ -524,6 +524,35 @@ static unsigned getEncodedThreadLocalMode(const GlobalValue &GV) {
   llvm_unreachable("Invalid TLS model");
 }
 
+static unsigned getEncodedComdatSelectionKind(const Comdat &C) {
+  switch (C.getSelectionKind()) {
+  case Comdat::Any:
+    return bitc::COMDAT_SELECTION_KIND_ANY;
+  case Comdat::ExactMatch:
+    return bitc::COMDAT_SELECTION_KIND_EXACT_MATCH;
+  case Comdat::Largest:
+    return bitc::COMDAT_SELECTION_KIND_LARGEST;
+  case Comdat::NoDuplicates:
+    return bitc::COMDAT_SELECTION_KIND_NO_DUPLICATES;
+  case Comdat::SameSize:
+    return bitc::COMDAT_SELECTION_KIND_SAME_SIZE;
+  }
+  llvm_unreachable("Invalid selection kind");
+}
+
+static void writeComdats(const ValueEnumerator &VE, BitstreamWriter &Stream) {
+  SmallVector<uint8_t, 64> Vals;
+  for (const Comdat *C : VE.getComdats()) {
+    // COMDAT: [selection_kind, name]
+    Vals.push_back(getEncodedComdatSelectionKind(*C));
+    Vals.push_back(C->getName().size());
+    for (char Chr : C->getName())
+      Vals.push_back((unsigned char)Chr);
+    Stream.EmitRecord(bitc::MODULE_CODE_COMDAT, Vals, /*AbbrevToUse=*/0);
+    Vals.clear();
+  }
+}
+
 // Emit top-level description of module, including target triple, inline asm,
 // descriptors for global variables, and function prototype info.
 static void WriteModuleInfo(const Module *M, const ValueEnumerator &VE,
@@ -625,12 +654,14 @@ static void WriteModuleInfo(const Module *M, const ValueEnumerator &VE,
     if (GV.isThreadLocal() ||
         GV.getVisibility() != GlobalValue::DefaultVisibility ||
         GV.hasUnnamedAddr() || GV.isExternallyInitialized() ||
-        GV.getDLLStorageClass() != GlobalValue::DefaultStorageClass) {
+        GV.getDLLStorageClass() != GlobalValue::DefaultStorageClass ||
+        GV.hasComdat()) {
       Vals.push_back(getEncodedVisibility(GV));
       Vals.push_back(getEncodedThreadLocalMode(GV));
       Vals.push_back(GV.hasUnnamedAddr());
       Vals.push_back(GV.isExternallyInitialized());
       Vals.push_back(getEncodedDLLStorageClass(GV));
+      Vals.push_back(GV.hasComdat() ? VE.getComdatID(GV.getComdat()) : 0);
     } else {
       AbbrevToUse = SimpleGVarAbbrev;
     }
@@ -656,6 +687,7 @@ static void WriteModuleInfo(const Module *M, const ValueEnumerator &VE,
     Vals.push_back(F.hasPrefixData() ? (VE.getValueID(F.getPrefixData()) + 1)
                                       : 0);
     Vals.push_back(getEncodedDLLStorageClass(F));
+    Vals.push_back(F.hasComdat() ? VE.getComdatID(F.getComdat()) : 0);
 
     unsigned AbbrevToUse = 0;
     Stream.EmitRecord(bitc::MODULE_CODE_FUNCTION, Vals, AbbrevToUse);
@@ -1915,6 +1947,8 @@ static void WriteModule(const Module *M, BitstreamWriter &Stream) {
   // Emit information describing all of the types in the module.
   WriteTypeTable(VE, Stream);
 
+  writeComdats(VE, Stream);
+
   // Emit top-level description of module, including target triple, inline asm,
   // descriptors for global variables, and function prototype info.
   WriteModuleInfo(M, VE, Stream);

lib/Bitcode/Writer/ValueEnumerator.cpp

@@ -117,6 +117,12 @@ unsigned ValueEnumerator::getInstructionID(const Instruction *Inst) const {
   return I->second;
 }
 
+unsigned ValueEnumerator::getComdatID(const Comdat *C) const {
+  unsigned ComdatID = Comdats.idFor(C);
+  assert(ComdatID && "Comdat not found!");
+  return ComdatID;
+}
+
 void ValueEnumerator::setInstructionID(const Instruction *I) {
   InstructionMap[I] = InstructionCount++;
 }
@@ -307,6 +313,10 @@ void ValueEnumerator::EnumerateValue(const Value *V) {
     return;
   }
 
+  if (auto *GO = dyn_cast<GlobalObject>(V))
+    if (const Comdat *C = GO->getComdat())
+      Comdats.insert(C);
+
   // Enumerate the type of this value.
   EnumerateType(V->getType());
 

lib/Bitcode/Writer/ValueEnumerator.h

@@ -16,6 +16,7 @@
 
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/UniqueVector.h"
 #include "llvm/IR/Attributes.h"
 #include <vector>
 
@@ -25,6 +26,7 @@ class Type;
 class Value;
 class Instruction;
 class BasicBlock;
+class Comdat;
 class Function;
 class Module;
 class MDNode;
@@ -48,6 +50,10 @@ private:
   typedef DenseMap<const Value*, unsigned> ValueMapType;
   ValueMapType ValueMap;
   ValueList Values;
+
+  typedef UniqueVector<const Comdat *> ComdatSetType;
+  ComdatSetType Comdats;
+
   ValueList MDValues;
   SmallVector<const MDNode *, 8> FunctionLocalMDs;
   ValueMapType MDValueMap;
@@ -139,6 +145,9 @@ public:
     return AttributeGroups;
   }
 
+  const ComdatSetType &getComdats() const { return Comdats; }
+  unsigned getComdatID(const Comdat *C) const;
+
   /// getGlobalBasicBlockID - This returns the function-specific ID for the
   /// specified basic block.  This is relatively expensive information, so it
   /// should only be used by rare constructs such as address-of-label.

lib/CodeGen/TargetLoweringObjectFileImpl.cpp

@@ -192,6 +192,18 @@ getELFSectionFlags(SectionKind K) {
   return Flags;
 }
 
+static const Comdat *getELFComdat(const GlobalValue *GV) {
+  const Comdat *C = GV->getComdat();
+  if (!C)
+    return nullptr;
+
+  if (C->getSelectionKind() != Comdat::Any)
+    report_fatal_error("ELF COMDATs only support SelectionKind::Any, '" +
+                       C->getName() + "' cannot be lowered.");
+
+  return C;
+}
+
 const MCSection *TargetLoweringObjectFileELF::getExplicitSectionGlobal(
     const GlobalValue *GV, SectionKind Kind, Mangler &Mang,
     const TargetMachine &TM) const {
@@ -200,9 +212,15 @@ const MCSection *TargetLoweringObjectFileELF::getExplicitSectionGlobal(
   // Infer section flags from the section name if we can.
   Kind = getELFKindForNamedSection(SectionName, Kind);
 
+  StringRef Group = "";
+  unsigned Flags = getELFSectionFlags(Kind);
+  if (const Comdat *C = getELFComdat(GV)) {
+    Group = C->getName();
+    Flags |= ELF::SHF_GROUP;
+  }
   return getContext().getELFSection(SectionName,
-                                    getELFSectionType(SectionName, Kind),
-                                    getELFSectionFlags(Kind), Kind);
+                                    getELFSectionType(SectionName, Kind), Flags,
+                                    Kind, /*EntrySize=*/0, Group);
 }
 
 /// getSectionPrefixForGlobal - Return the section prefix name used by options
@@ -224,7 +242,6 @@ static StringRef getSectionPrefixForGlobal(SectionKind Kind) {
   return ".data.rel.ro.";
 }
 
-
 const MCSection *TargetLoweringObjectFileELF::
 SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
                        Mangler &Mang, const TargetMachine &TM) const {
@@ -238,7 +255,7 @@ SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
 
   // If this global is linkonce/weak and the target handles this by emitting it
   // into a 'uniqued' section name, create and return the section now.
-  if ((GV->isWeakForLinker() || EmitUniquedSection) &&
+  if ((GV->isWeakForLinker() || EmitUniquedSection || GV->hasComdat()) &&
       !Kind.isCommon()) {
     StringRef Prefix = getSectionPrefixForGlobal(Kind);
 
@@ -247,8 +264,11 @@ SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
 
     StringRef Group = "";
     unsigned Flags = getELFSectionFlags(Kind);
-    if (GV->isWeakForLinker()) {
-      Group = Name.substr(Prefix.size());
+    if (GV->isWeakForLinker() || GV->hasComdat()) {
+      if (const Comdat *C = getELFComdat(GV))
+        Group = C->getName();
+      else
+        Group = Name.substr(Prefix.size());
       Flags |= ELF::SHF_GROUP;
     }
 
@@ -482,6 +502,15 @@ emitModuleFlags(MCStreamer &Streamer,
   Streamer.AddBlankLine();
 }
 
+static void checkMachOComdat(const GlobalValue *GV) {
+  const Comdat *C = GV->getComdat();
+  if (!C)
+    return;
+
+  report_fatal_error("MachO doesn't support COMDATs, '" + C->getName() +
+                     "' cannot be lowered.");
+}
+
 const MCSection *TargetLoweringObjectFileMachO::getExplicitSectionGlobal(
     const GlobalValue *GV, SectionKind Kind, Mangler &Mang,
     const TargetMachine &TM) const {
@@ -489,6 +518,9 @@ const MCSection *TargetLoweringObjectFileMachO::getExplicitSectionGlobal(
   StringRef Segment, Section;
   unsigned TAA = 0, StubSize = 0;
   bool TAAParsed;
+
+  checkMachOComdat(GV);
+
   std::string ErrorCode =
     MCSectionMachO::ParseSectionSpecifier(GV->getSection(), Segment, Section,
                                           TAA, TAAParsed, StubSize);
@@ -559,6 +591,7 @@ bool TargetLoweringObjectFileMachO::isSectionAtomizableBySymbols(
 const MCSection *TargetLoweringObjectFileMachO::
 SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
                        Mangler &Mang, const TargetMachine &TM) const {
+  checkMachOComdat(GV);
 
   // Handle thread local data.
   if (Kind.isThreadBSS()) return TLSBSSSection;
@@ -727,6 +760,50 @@ getCOFFSectionFlags(SectionKind K) {
   return Flags;
 }
 
+const GlobalValue *getComdatGVForCOFF(const GlobalValue *GV) {
+  const Comdat *C = GV->getComdat();
+  assert(C && "expected GV to have a Comdat!");
+
+  StringRef ComdatGVName = C->getName();
+  const GlobalValue *ComdatGV = GV->getParent()->getNamedValue(ComdatGVName);
+  if (!ComdatGV)
+    report_fatal_error("Associative COMDAT symbol '" + ComdatGVName +
+                       "' does not exist.");
+
+  if (ComdatGV->getComdat() != C)
+    report_fatal_error("Associative COMDAT symbol '" + ComdatGVName +
+                       "' is not a key for it's COMDAT.");
+
+  return ComdatGV;
+}
+
+static int getSelectionForCOFF(const GlobalValue *GV) {
+  if (const Comdat *C = GV->getComdat()) {
+    const GlobalValue *ComdatKey = getComdatGVForCOFF(GV);
+    if (const auto *GA = dyn_cast<GlobalAlias>(ComdatKey))
+      ComdatKey = GA->getBaseObject();
+    if (ComdatKey == GV) {
+      switch (C->getSelectionKind()) {
+      case Comdat::Any:
+        return COFF::IMAGE_COMDAT_SELECT_ANY;
+      case Comdat::ExactMatch:
+        return COFF::IMAGE_COMDAT_SELECT_EXACT_MATCH;
+      case Comdat::Largest:
+        return COFF::IMAGE_COMDAT_SELECT_LARGEST;
+      case Comdat::NoDuplicates:
+        return COFF::IMAGE_COMDAT_SELECT_NODUPLICATES;
+      case Comdat::SameSize:
+        return COFF::IMAGE_COMDAT_SELECT_SAME_SIZE;
+      }
+    } else {
+      return COFF::IMAGE_COMDAT_SELECT_ASSOCIATIVE;
+    }
+  } else if (GV->isWeakForLinker()) {
+    return COFF::IMAGE_COMDAT_SELECT_ANY;
+  }
+  return 0;
+}
+
 const MCSection *TargetLoweringObjectFileCOFF::getExplicitSectionGlobal(
     const GlobalValue *GV, SectionKind Kind, Mangler &Mang,
     const TargetMachine &TM) const {
@@ -734,11 +811,21 @@ const MCSection *TargetLoweringObjectFileCOFF::getExplicitSectionGlobal(
   unsigned Characteristics = getCOFFSectionFlags(Kind);
   StringRef Name = GV->getSection();
   StringRef COMDATSymName = "";
-  if (GV->isWeakForLinker()) {
-    Selection = COFF::IMAGE_COMDAT_SELECT_ANY;
-    Characteristics |= COFF::IMAGE_SCN_LNK_COMDAT;
-    MCSymbol *Sym = TM.getSymbol(GV, Mang);
-    COMDATSymName = Sym->getName();
+  if ((GV->isWeakForLinker() || GV->hasComdat()) && !Kind.isCommon()) {
+    Selection = getSelectionForCOFF(GV);
+    const GlobalValue *ComdatGV;
+    if (Selection == COFF::IMAGE_COMDAT_SELECT_ASSOCIATIVE)
+      ComdatGV = getComdatGVForCOFF(GV);
+    else
+      ComdatGV = GV;
+
+    if (!ComdatGV->hasPrivateLinkage()) {
+      MCSymbol *Sym = TM.getSymbol(ComdatGV, Mang);
+      COMDATSymName = Sym->getName();
+      Characteristics |= COFF::IMAGE_SCN_LNK_COMDAT;
+    } else {
+      Selection = 0;
+    }
   }
   return getContext().getCOFFSection(Name,
                                      Characteristics,
@@ -775,17 +862,27 @@ SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
   // into a 'uniqued' section name, create and return the section now.
   // Section names depend on the name of the symbol which is not feasible if the
   // symbol has private linkage.
-  if ((GV->isWeakForLinker() || EmitUniquedSection) &&
-      !GV->hasPrivateLinkage() && !Kind.isCommon()) {
+  if ((GV->isWeakForLinker() || EmitUniquedSection || GV->hasComdat()) &&
+      !Kind.isCommon()) {
     const char *Name = getCOFFSectionNameForUniqueGlobal(Kind);
     unsigned Characteristics = getCOFFSectionFlags(Kind);
 
     Characteristics |= COFF::IMAGE_SCN_LNK_COMDAT;
-    MCSymbol *Sym = TM.getSymbol(GV, Mang);
-    return getContext().getCOFFSection(
-        Name, Characteristics, Kind, Sym->getName(),
-        GV->isWeakForLinker() ? COFF::IMAGE_COMDAT_SELECT_ANY
-                              : COFF::IMAGE_COMDAT_SELECT_NODUPLICATES);
+    int Selection = getSelectionForCOFF(GV);
+    if (!Selection)
+      Selection = COFF::IMAGE_COMDAT_SELECT_NODUPLICATES;
+    const GlobalValue *ComdatGV;
+    if (GV->hasComdat())
+      ComdatGV = getComdatGVForCOFF(GV);
+    else
+      ComdatGV = GV;
+
+    if (!ComdatGV->hasPrivateLinkage()) {
+      MCSymbol *Sym = TM.getSymbol(ComdatGV, Mang);
+      StringRef COMDATSymName = Sym->getName();
+      return getContext().getCOFFSection(Name, Characteristics, Kind,
+                                         COMDATSymName, Selection);
+    }
   }
 
   if (Kind.isText())

lib/IR/AsmWriter.cpp

@@ -106,6 +106,7 @@ static void PrintEscapedString(StringRef Name, raw_ostream &Out) {
 
 enum PrefixType {
   GlobalPrefix,
+  ComdatPrefix,
   LabelPrefix,
   LocalPrefix,
   NoPrefix
@@ -119,6 +120,7 @@ static void PrintLLVMName(raw_ostream &OS, StringRef Name, PrefixType Prefix) {
   switch (Prefix) {
   case NoPrefix: break;
   case GlobalPrefix: OS << '@'; break;
+  case ComdatPrefix: OS << '$'; break;
   case LabelPrefix:  break;
   case LocalPrefix:  OS << '%'; break;
   }
@@ -1165,8 +1167,15 @@ static void WriteAsOperandInternal(raw_ostream &Out, const Value *V,
 }
 
 void AssemblyWriter::init() {
-  if (TheModule)
-    TypePrinter.incorporateTypes(*TheModule);
+  if (!TheModule)
+    return;
+  TypePrinter.incorporateTypes(*TheModule);
+  for (const Function &F : *TheModule)
+    if (const Comdat *C = F.getComdat())
+      Comdats.insert(C);
+  for (const GlobalVariable &GV : TheModule->globals())
+    if (const Comdat *C = GV.getComdat())
+      Comdats.insert(C);
 }
 
 
@@ -1308,6 +1317,15 @@ void AssemblyWriter::printModule(const Module *M) {
 
   printTypeIdentities();
 
+  // Output all comdats.
+  if (!Comdats.empty())
+    Out << '\n';
+  for (const Comdat *C : Comdats) {
+    printComdat(C);
+    if (C != Comdats.back())
+      Out << '\n';
+  }
+
   // Output all globals.
   if (!M->global_empty()) Out << '\n';
   for (Module::const_global_iterator I = M->global_begin(), E = M->global_end();
@@ -1470,6 +1488,10 @@ void AssemblyWriter::printGlobal(const GlobalVariable *GV) {
     PrintEscapedString(GV->getSection(), Out);
     Out << '"';
   }
+  if (GV->hasComdat()) {
+    Out << ", comdat ";
+    PrintLLVMName(Out, GV->getComdat()->getName(), ComdatPrefix);
+  }
   if (GV->getAlignment())
     Out << ", align " << GV->getAlignment();
 
@@ -1506,10 +1528,19 @@ void AssemblyWriter::printAlias(const GlobalAlias *GA) {
     writeOperand(Aliasee, !isa<ConstantExpr>(Aliasee));
   }
 
+  if (GA->hasComdat()) {
+    Out << ", comdat ";
+    PrintLLVMName(Out, GA->getComdat()->getName(), ComdatPrefix);
+  }
+
   printInfoComment(*GA);
   Out << '\n';
 }
 
+void AssemblyWriter::printComdat(const Comdat *C) {
+  C->print(Out);
+}
+
 void AssemblyWriter::printTypeIdentities() {
   if (TypePrinter.NumberedTypes.empty() &&
       TypePrinter.NamedTypes.empty())
@@ -1647,6 +1678,10 @@ void AssemblyWriter::printFunction(const Function *F) {
     PrintEscapedString(F->getSection(), Out);
     Out << '"';
   }
+  if (F->hasComdat()) {
+    Out << " comdat ";
+    PrintLLVMName(Out, F->getComdat()->getName(), ComdatPrefix);
+  }
   if (F->getAlignment())
     Out << " align " << F->getAlignment();
   if (F->hasGC())
@@ -2158,6 +2193,31 @@ void NamedMDNode::print(raw_ostream &ROS) const {
   W.printNamedMDNode(this);
 }
 
+void Comdat::print(raw_ostream &ROS) const {
+  PrintLLVMName(ROS, getName(), ComdatPrefix);
+  ROS << " = comdat ";
+
+  switch (getSelectionKind()) {
+  case Comdat::Any:
+    ROS << "any";
+    break;
+  case Comdat::ExactMatch:
+    ROS << "exactmatch";
+    break;
+  case Comdat::Largest:
+    ROS << "largest";
+    break;
+  case Comdat::NoDuplicates:
+    ROS << "noduplicates";
+    break;
+  case Comdat::SameSize:
+    ROS << "samesize";
+    break;
+  }
+
+  ROS << '\n';
+}
+
 void Type::print(raw_ostream &OS) const {
   TypePrinting TP;
   TP.print(const_cast<Type*>(this), OS);
@@ -2241,5 +2301,8 @@ void Type::dump() const { print(dbgs()); }
 // Module::dump() - Allow printing of Modules from the debugger.
 void Module::dump() const { print(dbgs(), nullptr); }
 
+// \brief Allow printing of Comdats from the debugger.
+void Comdat::dump() const { print(dbgs()); }
+
 // NamedMDNode::dump() - Allow printing of NamedMDNodes from the debugger.
 void NamedMDNode::dump() const { print(dbgs()); }

lib/IR/AsmWriter.h

@@ -16,6 +16,7 @@
 #define LLVM_IR_ASSEMBLYWRITER_H
 
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SetVector.h"
 #include "llvm/IR/Attributes.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/TypeFinder.h"
@@ -26,6 +27,7 @@ namespace llvm {
 class BasicBlock;
 class Function;
 class GlobalValue;
+class Comdat;
 class Module;
 class NamedMDNode;
 class Value;
@@ -70,6 +72,7 @@ private:
   SlotTracker &Machine;
   TypePrinting TypePrinter;
   AssemblyAnnotationWriter *AnnotationWriter;
+  SetVector<const Comdat *> Comdats;
 
 public:
   /// Construct an AssemblyWriter with an external SlotTracker
@@ -101,6 +104,7 @@ public:
   void printTypeIdentities();
   void printGlobal(const GlobalVariable *GV);
   void printAlias(const GlobalAlias *GV);
+  void printComdat(const Comdat *C);
   void printFunction(const Function *F);
   void printArgument(const Argument *FA, AttributeSet Attrs, unsigned Idx);
   void printBasicBlock(const BasicBlock *BB);

lib/IR/CMakeLists.txt

@@ -3,6 +3,7 @@ add_llvm_library(LLVMCore
   Attributes.cpp
   AutoUpgrade.cpp
   BasicBlock.cpp
+  Comdat.cpp
   ConstantFold.cpp
   ConstantRange.cpp
   Constants.cpp

lib/IR/Comdat.cpp

@@ -0,0 +1,25 @@
+//===-- Comdat.cpp - Implement Metadata classes --------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the Comdat class.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/IR/Comdat.h"
+#include "llvm/ADT/StringMap.h"
+using namespace llvm;
+
+Comdat::Comdat(SelectionKind SK, StringMapEntry<Comdat> *Name)
+    : Name(Name), SK(SK) {}
+
+Comdat::Comdat(Comdat &&C) : Name(C.Name), SK(C.SK) {}
+
+Comdat::Comdat() : Name(nullptr), SK(Comdat::Any) {}
+
+StringRef Comdat::getName() const { return Name->first(); }

lib/IR/Globals.cpp

@@ -59,15 +59,10 @@ void GlobalValue::copyAttributesFrom(const GlobalValue *Src) {
   setDLLStorageClass(Src->getDLLStorageClass());
 }
 
-static const GlobalObject *getBaseObject(const Constant &C) {
-  // FIXME: We should probably return a base + offset pair for non-zero GEPs.
-  return dyn_cast<GlobalObject>(C.stripPointerCasts());
-}
-
 unsigned GlobalValue::getAlignment() const {
   if (auto *GA = dyn_cast<GlobalAlias>(this)) {
     // In general we cannot compute this at the IR level, but we try.
-    if (const GlobalObject *GO = getBaseObject(*GA->getAliasee()))
+    if (const GlobalObject *GO = GA->getBaseObject())
       return GO->getAlignment();
 
     // FIXME: we should also be able to handle:
@@ -96,13 +91,23 @@ void GlobalObject::copyAttributesFrom(const GlobalValue *Src) {
 const char *GlobalValue::getSection() const {
   if (auto *GA = dyn_cast<GlobalAlias>(this)) {
     // In general we cannot compute this at the IR level, but we try.
-    if (const GlobalObject *GO = getBaseObject(*GA->getAliasee()))
+    if (const GlobalObject *GO = GA->getBaseObject())
       return GO->getSection();
     return "";
   }
   return cast<GlobalObject>(this)->getSection();
 }
 
+Comdat *GlobalValue::getComdat() {
+  if (auto *GA = dyn_cast<GlobalAlias>(this)) {
+    // In general we cannot compute this at the IR level, but we try.
+    if (const GlobalObject *GO = GA->getBaseObject())
+      return const_cast<GlobalObject *>(GO)->getComdat();
+    return nullptr;
+  }
+  return cast<GlobalObject>(this)->getComdat();
+}
+
 void GlobalObject::setSection(StringRef S) { Section = S; }
 
 bool GlobalValue::isDeclaration() const {

lib/IR/Module.cpp

@@ -453,3 +453,11 @@ unsigned Module::getDwarfVersion() const {
     return dwarf::DWARF_VERSION;
   return cast<ConstantInt>(Val)->getZExtValue();
 }
+
+Comdat *Module::getOrInsertComdat(StringRef Name) {
+  Comdat C;
+  StringMapEntry<Comdat> &Entry =
+      ComdatSymTab.GetOrCreateValue(Name, std::move(C));
+  Entry.second.Name = &Entry;
+  return &Entry.second;
+}

lib/IR/Verifier.cpp

@@ -107,6 +107,12 @@ struct VerifierSupport {
     OS << ' ' << *T;
   }
 
+  void WriteComdat(const Comdat *C) {
+    if (!C)
+      return;
+    OS << *C;
+  }
+
   // CheckFailed - A check failed, so print out the condition and the message
   // that failed.  This provides a nice place to put a breakpoint if you want
   // to see why something is not correct.
@@ -138,6 +144,12 @@ struct VerifierSupport {
     WriteType(T3);
     Broken = true;
   }
+
+  void CheckFailed(const Twine &Message, const Comdat *C) {
+    OS << Message.str() << "\n";
+    WriteComdat(C);
+    Broken = true;
+  }
 };
 class Verifier : public InstVisitor<Verifier>, VerifierSupport {
   friend class InstVisitor<Verifier>;
@@ -230,6 +242,9 @@ public:
          I != E; ++I)
       visitNamedMDNode(*I);
 
+    for (const StringMapEntry<Comdat> &SMEC : M.getComdatSymbolTable())
+      visitComdat(SMEC.getValue());
+
     visitModuleFlags(M);
     visitModuleIdents(M);
 
@@ -246,6 +261,7 @@ private:
                            const GlobalAlias &A, const Constant &C);
   void visitNamedMDNode(const NamedMDNode &NMD);
   void visitMDNode(MDNode &MD, Function *F);
+  void visitComdat(const Comdat &C);
   void visitModuleIdents(const Module &M);
   void visitModuleFlags(const Module &M);
   void visitModuleFlag(const MDNode *Op,
@@ -387,6 +403,7 @@ void Verifier::visitGlobalVariable(const GlobalVariable &GV) {
               "'common' global must have a zero initializer!", &GV);
       Assert1(!GV.isConstant(), "'common' global may not be marked constant!",
               &GV);
+      Assert1(!GV.hasComdat(), "'common' global may not be in a Comdat!", &GV);
     }
   } else {
     Assert1(GV.hasExternalLinkage() || GV.hasExternalWeakLinkage(),
@@ -578,6 +595,22 @@ void Verifier::visitMDNode(MDNode &MD, Function *F) {
   }
 }
 
+void Verifier::visitComdat(const Comdat &C) {
+  // All Comdat::SelectionKind values other than Comdat::Any require a
+  // GlobalValue with the same name as the Comdat.
+  const GlobalValue *GV = M->getNamedValue(C.getName());
+  if (C.getSelectionKind() != Comdat::Any)
+    Assert1(GV,
+            "comdat selection kind requires a global value with the same name",
+            &C);
+  // The Module is invalid if the GlobalValue has local linkage.  Allowing
+  // otherwise opens us up to seeing the underling global value get renamed if
+  // collisions occur.
+  if (GV)
+    Assert1(!GV->hasLocalLinkage(), "comdat global value has local linkage",
+            GV);
+}
+
 void Verifier::visitModuleIdents(const Module &M) {
   const NamedMDNode *Idents = M.getNamedMetadata("llvm.ident");
   if (!Idents) 

lib/Linker/LinkModules.cpp

@@ -426,6 +426,18 @@ namespace {
       return true;
     }
 
+    bool getComdatLeader(Module *M, StringRef ComdatName,
+                         const GlobalVariable *&GVar);
+    bool computeResultingSelectionKind(StringRef ComdatName,
+                                       Comdat::SelectionKind Src,
+                                       Comdat::SelectionKind Dst,
+                                       Comdat::SelectionKind &Result,
+                                       bool &LinkFromSrc);
+    std::map<const Comdat *, std::pair<Comdat::SelectionKind, bool>>
+        ComdatsChosen;
+    bool getComdatResult(const Comdat *SrcC, Comdat::SelectionKind &SK,
+                         bool &LinkFromSrc);
+
     /// getLinkageResult - This analyzes the two global values and determines
     /// what the result will look like in the destination module.
     bool getLinkageResult(GlobalValue *Dest, const GlobalValue *Src,
@@ -534,6 +546,115 @@ Value *ValueMaterializerTy::materializeValueFor(Value *V) {
   return DF;
 }
 
+bool ModuleLinker::getComdatLeader(Module *M, StringRef ComdatName,
+                                   const GlobalVariable *&GVar) {
+  const GlobalValue *GVal = M->getNamedValue(ComdatName);
+  if (const auto *GA = dyn_cast_or_null<GlobalAlias>(GVal)) {
+    GVal = GA->getBaseObject();
+    if (!GVal)
+      // We cannot resolve the size of the aliasee yet.
+      return emitError("Linking COMDATs named '" + ComdatName +
+                       "': COMDAT key involves incomputable alias size.");
+  }
+
+  GVar = dyn_cast_or_null<GlobalVariable>(GVal);
+  if (!GVar)
+    return emitError(
+        "Linking COMDATs named '" + ComdatName +
+        "': GlobalVariable required for data dependent selection!");
+
+  return false;
+}
+
+bool ModuleLinker::computeResultingSelectionKind(StringRef ComdatName,
+                                                 Comdat::SelectionKind Src,
+                                                 Comdat::SelectionKind Dst,
+                                                 Comdat::SelectionKind &Result,
+                                                 bool &LinkFromSrc) {
+  // The ability to mix Comdat::SelectionKind::Any with
+  // Comdat::SelectionKind::Largest is a behavior that comes from COFF.
+  bool DstAnyOrLargest = Dst == Comdat::SelectionKind::Any ||
+                         Dst == Comdat::SelectionKind::Largest;
+  bool SrcAnyOrLargest = Src == Comdat::SelectionKind::Any ||
+                         Src == Comdat::SelectionKind::Largest;
+  if (DstAnyOrLargest && SrcAnyOrLargest) {
+    if (Dst == Comdat::SelectionKind::Largest ||
+        Src == Comdat::SelectionKind::Largest)
+      Result = Comdat::SelectionKind::Largest;
+    else
+      Result = Comdat::SelectionKind::Any;
+  } else if (Src == Dst) {
+    Result = Dst;
+  } else {
+    return emitError("Linking COMDATs named '" + ComdatName +
+                     "': invalid selection kinds!");
+  }
+
+  switch (Result) {
+  case Comdat::SelectionKind::Any:
+    // Go with Dst.
+    LinkFromSrc = false;
+    break;
+  case Comdat::SelectionKind::NoDuplicates:
+    return emitError("Linking COMDATs named '" + ComdatName +
+                     "': noduplicates has been violated!");
+  case Comdat::SelectionKind::ExactMatch:
+  case Comdat::SelectionKind::Largest:
+  case Comdat::SelectionKind::SameSize: {
+    const GlobalVariable *DstGV;
+    const GlobalVariable *SrcGV;
+    if (getComdatLeader(DstM, ComdatName, DstGV) ||
+        getComdatLeader(SrcM, ComdatName, SrcGV))
+      return true;
+
+    const DataLayout *DstDL = DstM->getDataLayout();
+    const DataLayout *SrcDL = SrcM->getDataLayout();
+    if (!DstDL || !SrcDL) {
+      return emitError(
+          "Linking COMDATs named '" + ComdatName +
+          "': can't do size dependent selection without DataLayout!");
+    }
+    uint64_t DstSize =
+        DstDL->getTypeAllocSize(DstGV->getType()->getPointerElementType());
+    uint64_t SrcSize =
+        SrcDL->getTypeAllocSize(SrcGV->getType()->getPointerElementType());
+    if (Result == Comdat::SelectionKind::ExactMatch) {
+      if (SrcGV->getInitializer() != DstGV->getInitializer())
+        return emitError("Linking COMDATs named '" + ComdatName +
+                         "': ExactMatch violated!");
+      LinkFromSrc = false;
+    } else if (Result == Comdat::SelectionKind::Largest) {
+      LinkFromSrc = SrcSize > DstSize;
+    } else if (Result == Comdat::SelectionKind::SameSize) {
+      if (SrcSize != DstSize)
+        return emitError("Linking COMDATs named '" + ComdatName +
+                         "': SameSize violated!");
+      LinkFromSrc = false;
+    } else {
+      llvm_unreachable("unknown selection kind");
+    }
+    break;
+  }
+  }
+
+  return false;
+}
+
+bool ModuleLinker::getComdatResult(const Comdat *SrcC,
+                                   Comdat::SelectionKind &Result,
+                                   bool &LinkFromSrc) {
+  StringRef ComdatName = SrcC->getName();
+  Module::ComdatSymTabType &ComdatSymTab = DstM->getComdatSymbolTable();
+  Module::ComdatSymTabType::iterator DstCI = ComdatSymTab.find(ComdatName);
+  if (DstCI != ComdatSymTab.end()) {
+    const Comdat *DstC = &DstCI->second;
+    Comdat::SelectionKind SSK = SrcC->getSelectionKind();
+    Comdat::SelectionKind DSK = DstC->getSelectionKind();
+    if (computeResultingSelectionKind(ComdatName, SSK, DSK, Result, LinkFromSrc))
+      return true;
+  }
+  return false;
+}
 
 /// getLinkageResult - This analyzes the two global values and determines what
 /// the result will look like in the destination module.  In particular, it
@@ -764,34 +885,47 @@ bool ModuleLinker::linkGlobalProto(GlobalVariable *SGV) {
   llvm::Optional<GlobalValue::VisibilityTypes> NewVisibility;
   bool HasUnnamedAddr = SGV->hasUnnamedAddr();
 
+  bool LinkFromSrc = false;
+  Comdat *DC = nullptr;
+  if (const Comdat *SC = SGV->getComdat()) {
+    Comdat::SelectionKind SK;
+    std::tie(SK, LinkFromSrc) = ComdatsChosen[SC];
+    DC = DstM->getOrInsertComdat(SC->getName());
+    DC->setSelectionKind(SK);
+  }
+
   if (DGV) {
-    // Concatenation of appending linkage variables is magic and handled later.
-    if (DGV->hasAppendingLinkage() || SGV->hasAppendingLinkage())
-      return linkAppendingVarProto(cast<GlobalVariable>(DGV), SGV);
-
-    // Determine whether linkage of these two globals follows the source
-    // module's definition or the destination module's definition.
-    GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
-    GlobalValue::VisibilityTypes NV;
-    bool LinkFromSrc = false;
-    if (getLinkageResult(DGV, SGV, NewLinkage, NV, LinkFromSrc))
-      return true;
-    NewVisibility = NV;
-    HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+    if (!DC) {
+      // Concatenation of appending linkage variables is magic and handled later.
+      if (DGV->hasAppendingLinkage() || SGV->hasAppendingLinkage())
+        return linkAppendingVarProto(cast<GlobalVariable>(DGV), SGV);
+
+      // Determine whether linkage of these two globals follows the source
+      // module's definition or the destination module's definition.
+      GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
+      GlobalValue::VisibilityTypes NV;
+      if (getLinkageResult(DGV, SGV, NewLinkage, NV, LinkFromSrc))
+        return true;
+      NewVisibility = NV;
+      HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+
+      // If we're not linking from the source, then keep the definition that we
+      // have.
+      if (!LinkFromSrc) {
+        // Special case for const propagation.
+        if (GlobalVariable *DGVar = dyn_cast<GlobalVariable>(DGV))
+          if (DGVar->isDeclaration() && SGV->isConstant() &&
+              !DGVar->isConstant())
+            DGVar->setConstant(true);
+
+        // Set calculated linkage, visibility and unnamed_addr.
+        DGV->setLinkage(NewLinkage);
+        DGV->setVisibility(*NewVisibility);
+        DGV->setUnnamedAddr(HasUnnamedAddr);
+      }
+    }
 
-    // If we're not linking from the source, then keep the definition that we
-    // have.
     if (!LinkFromSrc) {
-      // Special case for const propagation.
-      if (GlobalVariable *DGVar = dyn_cast<GlobalVariable>(DGV))
-        if (DGVar->isDeclaration() && SGV->isConstant() && !DGVar->isConstant())
-          DGVar->setConstant(true);
-
-      // Set calculated linkage, visibility and unnamed_addr.
-      DGV->setLinkage(NewLinkage);
-      DGV->setVisibility(*NewVisibility);
-      DGV->setUnnamedAddr(HasUnnamedAddr);
-
       // Make sure to remember this mapping.
       ValueMap[SGV] = ConstantExpr::getBitCast(DGV,TypeMap.get(SGV->getType()));
 
@@ -803,6 +937,12 @@ bool ModuleLinker::linkGlobalProto(GlobalVariable *SGV) {
     }
   }
 
+  // If the Comdat this variable was inside of wasn't selected, skip it.
+  if (DC && !DGV && !LinkFromSrc) {
+    DoNotLinkFromSource.insert(SGV);
+    return false;
+  }
+
   // No linking to be performed or linking from the source: simply create an
   // identical version of the symbol over in the dest module... the
   // initializer will be filled in later by LinkGlobalInits.
@@ -818,6 +958,9 @@ bool ModuleLinker::linkGlobalProto(GlobalVariable *SGV) {
     NewDGV->setVisibility(*NewVisibility);
   NewDGV->setUnnamedAddr(HasUnnamedAddr);
 
+  if (DC)
+    NewDGV->setComdat(DC);
+
   if (DGV) {
     DGV->replaceAllUsesWith(ConstantExpr::getBitCast(NewDGV, DGV->getType()));
     DGV->eraseFromParent();
@@ -835,21 +978,33 @@ bool ModuleLinker::linkFunctionProto(Function *SF) {
   llvm::Optional<GlobalValue::VisibilityTypes> NewVisibility;
   bool HasUnnamedAddr = SF->hasUnnamedAddr();
 
+  bool LinkFromSrc = false;
+  Comdat *DC = nullptr;
+  if (const Comdat *SC = SF->getComdat()) {
+    Comdat::SelectionKind SK;
+    std::tie(SK, LinkFromSrc) = ComdatsChosen[SC];
+    DC = DstM->getOrInsertComdat(SC->getName());
+    DC->setSelectionKind(SK);
+  }
+
   if (DGV) {
-    GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
-    bool LinkFromSrc = false;
-    GlobalValue::VisibilityTypes NV;
-    if (getLinkageResult(DGV, SF, NewLinkage, NV, LinkFromSrc))
-      return true;
-    NewVisibility = NV;
-    HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+    if (!DC) {
+      GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
+      GlobalValue::VisibilityTypes NV;
+      if (getLinkageResult(DGV, SF, NewLinkage, NV, LinkFromSrc))
+        return true;
+      NewVisibility = NV;
+      HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+
+      if (!LinkFromSrc) {
+        // Set calculated linkage
+        DGV->setLinkage(NewLinkage);
+        DGV->setVisibility(*NewVisibility);
+        DGV->setUnnamedAddr(HasUnnamedAddr);
+      }
+    }
 
     if (!LinkFromSrc) {
-      // Set calculated linkage
-      DGV->setLinkage(NewLinkage);
-      DGV->setVisibility(*NewVisibility);
-      DGV->setUnnamedAddr(HasUnnamedAddr);
-
       // Make sure to remember this mapping.
       ValueMap[SF] = ConstantExpr::getBitCast(DGV, TypeMap.get(SF->getType()));
 
@@ -869,6 +1024,12 @@ bool ModuleLinker::linkFunctionProto(Function *SF) {
     return false;
   }
 
+  // If the Comdat this function was inside of wasn't selected, skip it.
+  if (DC && !DGV && !LinkFromSrc) {
+    DoNotLinkFromSource.insert(SF);
+    return false;
+  }
+
   // If there is no linkage to be performed or we are linking from the source,
   // bring SF over.
   Function *NewDF = Function::Create(TypeMap.get(SF->getFunctionType()),
@@ -878,6 +1039,9 @@ bool ModuleLinker::linkFunctionProto(Function *SF) {
     NewDF->setVisibility(*NewVisibility);
   NewDF->setUnnamedAddr(HasUnnamedAddr);
 
+  if (DC)
+    NewDF->setComdat(DC);
+
   if (DGV) {
     // Any uses of DF need to change to NewDF, with cast.
     DGV->replaceAllUsesWith(ConstantExpr::getBitCast(NewDF, DGV->getType()));
@@ -895,21 +1059,33 @@ bool ModuleLinker::linkAliasProto(GlobalAlias *SGA) {
   llvm::Optional<GlobalValue::VisibilityTypes> NewVisibility;
   bool HasUnnamedAddr = SGA->hasUnnamedAddr();
 
+  bool LinkFromSrc = false;
+  Comdat *DC = nullptr;
+  if (const Comdat *SC = SGA->getComdat()) {
+    Comdat::SelectionKind SK;
+    std::tie(SK, LinkFromSrc) = ComdatsChosen[SC];
+    DC = DstM->getOrInsertComdat(SC->getName());
+    DC->setSelectionKind(SK);
+  }
+
   if (DGV) {
-    GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
-    GlobalValue::VisibilityTypes NV;
-    bool LinkFromSrc = false;
-    if (getLinkageResult(DGV, SGA, NewLinkage, NV, LinkFromSrc))
-      return true;
-    NewVisibility = NV;
-    HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+    if (!DC) {
+      GlobalValue::LinkageTypes NewLinkage = GlobalValue::InternalLinkage;
+      GlobalValue::VisibilityTypes NV;
+      if (getLinkageResult(DGV, SGA, NewLinkage, NV, LinkFromSrc))
+        return true;
+      NewVisibility = NV;
+      HasUnnamedAddr = HasUnnamedAddr && DGV->hasUnnamedAddr();
+
+      if (!LinkFromSrc) {
+        // Set calculated linkage.
+        DGV->setLinkage(NewLinkage);
+        DGV->setVisibility(*NewVisibility);
+        DGV->setUnnamedAddr(HasUnnamedAddr);
+      }
+    }
 
     if (!LinkFromSrc) {
-      // Set calculated linkage.
-      DGV->setLinkage(NewLinkage);
-      DGV->setVisibility(*NewVisibility);
-      DGV->setUnnamedAddr(HasUnnamedAddr);
-
       // Make sure to remember this mapping.
       ValueMap[SGA] = ConstantExpr::getBitCast(DGV,TypeMap.get(SGA->getType()));
 
@@ -920,6 +1096,12 @@ bool ModuleLinker::linkAliasProto(GlobalAlias *SGA) {
     }
   }
 
+  // If the Comdat this alias was inside of wasn't selected, skip it.
+  if (DC && !DGV && !LinkFromSrc) {
+    DoNotLinkFromSource.insert(SGA);
+    return false;
+  }
+
   // If there is no linkage to be performed or we're linking from the source,
   // bring over SGA.
   auto *PTy = cast<PointerType>(TypeMap.get(SGA->getType()));
@@ -1254,6 +1436,18 @@ bool ModuleLinker::run() {
   // Loop over all of the linked values to compute type mappings.
   computeTypeMapping();
 
+  ComdatsChosen.clear();
+  for (const StringMapEntry<llvm::Comdat> &SMEC : SrcM->getComdatSymbolTable()) {
+    const Comdat &C = SMEC.getValue();
+    if (ComdatsChosen.count(&C))
+      continue;
+    Comdat::SelectionKind SK;
+    bool LinkFromSrc;
+    if (getComdatResult(&C, SK, LinkFromSrc))
+      return true;
+    ComdatsChosen[&C] = std::make_pair(SK, LinkFromSrc);
+  }
+
   // Insert all of the globals in src into the DstM module... without linking
   // initializers (which could refer to functions not yet mapped over).
   for (Module::global_iterator I = SrcM->global_begin(),

lib/Transforms/IPO/GlobalDCE.cpp

@@ -77,13 +77,19 @@ bool GlobalDCE::runOnModule(Module &M) {
   // Remove empty functions from the global ctors list.
   Changed |= optimizeGlobalCtorsList(M, isEmptyFunction);
 
+  typedef std::multimap<const Comdat *, GlobalValue *> ComdatGVPairsTy;
+  ComdatGVPairsTy ComdatGVPairs;
+
   // Loop over the module, adding globals which are obviously necessary.
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
     Changed |= RemoveUnusedGlobalValue(*I);
     // Functions with external linkage are needed if they have a body
-    if (!I->isDiscardableIfUnused() &&
-        !I->isDeclaration() && !I->hasAvailableExternallyLinkage())
-      GlobalIsNeeded(I);
+    if (!I->isDeclaration() && !I->hasAvailableExternallyLinkage()) {
+      if (!I->isDiscardableIfUnused())
+        GlobalIsNeeded(I);
+      else if (const Comdat *C = I->getComdat())
+        ComdatGVPairs.insert(std::make_pair(C, I));
+    }
   }
 
   for (Module::global_iterator I = M.global_begin(), E = M.global_end();
@@ -91,17 +97,38 @@ bool GlobalDCE::runOnModule(Module &M) {
     Changed |= RemoveUnusedGlobalValue(*I);
     // Externally visible & appending globals are needed, if they have an
     // initializer.
-    if (!I->isDiscardableIfUnused() &&
-        !I->isDeclaration() && !I->hasAvailableExternallyLinkage())
-      GlobalIsNeeded(I);
+    if (!I->isDeclaration() && !I->hasAvailableExternallyLinkage()) {
+      if (!I->isDiscardableIfUnused())
+        GlobalIsNeeded(I);
+      else if (const Comdat *C = I->getComdat())
+        ComdatGVPairs.insert(std::make_pair(C, I));
+    }
   }
 
   for (Module::alias_iterator I = M.alias_begin(), E = M.alias_end();
        I != E; ++I) {
     Changed |= RemoveUnusedGlobalValue(*I);
     // Externally visible aliases are needed.
-    if (!I->isDiscardableIfUnused())
+    if (!I->isDiscardableIfUnused()) {
       GlobalIsNeeded(I);
+    } else if (const Comdat *C = I->getComdat()) {
+      ComdatGVPairs.insert(std::make_pair(C, I));
+    }
+  }
+
+  for (ComdatGVPairsTy::iterator I = ComdatGVPairs.begin(),
+                                 E = ComdatGVPairs.end();
+       I != E;) {
+    ComdatGVPairsTy::iterator UB = ComdatGVPairs.upper_bound(I->first);
+    bool CanDiscard = std::all_of(I, UB, [](ComdatGVPairsTy::value_type Pair) {
+      return Pair.second->isDiscardableIfUnused();
+    });
+    if (!CanDiscard) {
+      std::for_each(I, UB, [this](ComdatGVPairsTy::value_type Pair) {
+        GlobalIsNeeded(Pair.second);
+      });
+    }
+    I = UB;
   }
 
   // Now that all globals which are needed are in the AliveGlobals set, we loop

lib/Transforms/IPO/GlobalOpt.cpp

@@ -17,6 +17,7 @@
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/ConstantFolding.h"
@@ -1699,9 +1700,6 @@ static bool TryToShrinkGlobalToBoolean(GlobalVariable *GV, Constant *OtherVal) {
 /// possible.  If we make a change, return true.
 bool GlobalOpt::ProcessGlobal(GlobalVariable *GV,
                               Module::global_iterator &GVI) {
-  if (!GV->isDiscardableIfUnused())
-    return false;
-
   // Do more involved optimizations if the global is internal.
   GV->removeDeadConstantUsers();
 
@@ -1944,6 +1942,13 @@ bool GlobalOpt::OptimizeFunctions(Module &M) {
 
 bool GlobalOpt::OptimizeGlobalVars(Module &M) {
   bool Changed = false;
+
+  SmallSet<const Comdat *, 8> NotDiscardableComdats;
+  for (const GlobalVariable &GV : M.globals())
+    if (const Comdat *C = GV.getComdat())
+      if (!GV.isDiscardableIfUnused())
+        NotDiscardableComdats.insert(C);
+
   for (Module::global_iterator GVI = M.global_begin(), E = M.global_end();
        GVI != E; ) {
     GlobalVariable *GV = GVI++;
@@ -1958,7 +1963,12 @@ bool GlobalOpt::OptimizeGlobalVars(Module &M) {
           GV->setInitializer(New);
       }
 
-    Changed |= ProcessGlobal(GV, GVI);
+    if (GV->isDiscardableIfUnused()) {
+      if (const Comdat *C = GV->getComdat())
+        if (NotDiscardableComdats.count(C))
+          continue;
+      Changed |= ProcessGlobal(GV, GVI);
+    }
   }
   return Changed;
 }

test/Assembler/invalid-comdat.ll

@@ -0,0 +1,4 @@
+; RUN: not llvm-as < %s 2>&1 | FileCheck %s
+
+@v = global i32 0, comdat $v
+; CHECK: use of undefined comdat '$v'

test/Assembler/invalid-comdat2.ll

@@ -0,0 +1,5 @@
+; RUN: not llvm-as < %s 2>&1 | FileCheck %s
+
+$v = comdat any
+$v = comdat any
+; CHECK: redefinition of comdat '$v'

test/CodeGen/X86/coff-comdat.ll

@@ -0,0 +1,92 @@
+; RUN: llc -mtriple i386-pc-win32 < %s | FileCheck %s
+
+$f1 = comdat any
+@v1 = global i32 0, comdat $f1
+define void @f1() comdat $f1 {
+  ret void
+}
+
+$f2 = comdat exactmatch
+@v2 = global i32 0, comdat $f2
+define void @f2() comdat $f2 {
+  ret void
+}
+
+$f3 = comdat largest
+@v3 = global i32 0, comdat $f3
+define void @f3() comdat $f3 {
+  ret void
+}
+
+$f4 = comdat noduplicates
+@v4 = global i32 0, comdat $f4
+define void @f4() comdat $f4 {
+  ret void
+}
+
+$f5 = comdat samesize
+@v5 = global i32 0, comdat $f5
+define void @f5() comdat $f5 {
+  ret void
+}
+
+$f6 = comdat samesize
+@v6 = global i32 0, comdat $f6
+@f6 = global i32 0, comdat $f6
+
+$"\01@f7@0" = comdat any
+define x86_fastcallcc void @"\01@v7@0"() comdat $"\01@f7@0" {
+  ret void
+}
+define x86_fastcallcc void @"\01@f7@0"() comdat $"\01@f7@0" {
+  ret void
+}
+
+$f8 = comdat any
+define x86_fastcallcc void @v8() comdat $f8 {
+  ret void
+}
+define x86_fastcallcc void @f8() comdat $f8 {
+  ret void
+}
+
+$vftable = comdat largest
+
+@some_name = private unnamed_addr constant [2 x i8*] zeroinitializer, comdat $vftable
+@vftable = alias getelementptr([2 x i8*]* @some_name, i32 0, i32 1)
+
+; CHECK: .section        .text,"xr",discard,_f1
+; CHECK: .globl  _f1
+; CHECK: .section        .text,"xr",same_contents,_f2
+; CHECK: .globl  _f2
+; CHECK: .section        .text,"xr",largest,_f3
+; CHECK: .globl  _f3
+; CHECK: .section        .text,"xr",one_only,_f4
+; CHECK: .globl  _f4
+; CHECK: .section        .text,"xr",same_size,_f5
+; CHECK: .globl  _f5
+; CHECK: .section        .text,"xr",associative,@f7@0
+; CHECK: .globl  @v7@0
+; CHECK: .section        .text,"xr",discard,@f7@0
+; CHECK: .globl  @f7@0
+; CHECK: .section        .text,"xr",associative,@f8@0
+; CHECK: .globl  @v8@0
+; CHECK: .section        .text,"xr",discard,@f8@0
+; CHECK: .globl  @f8@0
+; CHECK: .section        .bss,"bw",associative,_f1
+; CHECK: .globl  _v1
+; CHECK: .section        .bss,"bw",associative,_f2
+; CHECK: .globl  _v2
+; CHECK: .section        .bss,"bw",associative,_f3
+; CHECK: .globl  _v3
+; CHECK: .section        .bss,"bw",associative,_f4
+; CHECK: .globl  _v4
+; CHECK: .section        .bss,"bw",associative,_f5
+; CHECK: .globl  _v5
+; CHECK: .section        .bss,"bw",associative,_f6
+; CHECK: .globl  _v6
+; CHECK: .section        .bss,"bw",same_size,_f6
+; CHECK: .globl  _f6
+; CHECK: .section        .rdata,"rd",largest,_vftable
+; CHECK: .globl  _vftable
+; CHECK: _vftable = L_some_name+4

test/CodeGen/X86/coff-comdat.s

@@ -0,0 +1,156 @@
+	.text
+	.def	 @feat.00;
+	.scl	3;
+	.type	0;
+	.endef
+	.globl	@feat.00
+@feat.00 = 1
+	.def	 _f1;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",discard,_f1
+	.globl	_f1
+	.align	16, 0x90
+_f1:                                    # @f1
+# BB#0:
+	retl
+
+	.def	 _f2;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",same_contents,_f2
+	.globl	_f2
+	.align	16, 0x90
+_f2:                                    # @f2
+# BB#0:
+	retl
+
+	.def	 _f3;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",largest,_f3
+	.globl	_f3
+	.align	16, 0x90
+_f3:                                    # @f3
+# BB#0:
+	retl
+
+	.def	 _f4;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",one_only,_f4
+	.globl	_f4
+	.align	16, 0x90
+_f4:                                    # @f4
+# BB#0:
+	retl
+
+	.def	 _f5;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",same_size,_f5
+	.globl	_f5
+	.align	16, 0x90
+_f5:                                    # @f5
+# BB#0:
+	retl
+
+	.def	 @v7@0;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",associative,@f7@0
+	.globl	@v7@0
+	.align	16, 0x90
+@v7@0:                                  # @"\01@v7@0"
+# BB#0:
+	retl
+
+	.def	 @f7@0;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",discard,@f7@0
+	.globl	@f7@0
+	.align	16, 0x90
+@f7@0:                                  # @"\01@f7@0"
+# BB#0:
+	retl
+
+	.def	 @v8@0;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",associative,@f8@0
+	.globl	@v8@0
+	.align	16, 0x90
+@v8@0:                                  # @v8
+# BB#0:
+	retl
+
+	.def	 @f8@0;
+	.scl	2;
+	.type	32;
+	.endef
+	.section	.text,"xr",discard,@f8@0
+	.globl	@f8@0
+	.align	16, 0x90
+@f8@0:                                  # @f8
+# BB#0:
+	retl
+
+	.section	.bss,"bw",associative,_f1
+	.globl	_v1                     # @v1
+	.align	4
+_v1:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",associative,_f2
+	.globl	_v2                     # @v2
+	.align	4
+_v2:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",associative,_f3
+	.globl	_v3                     # @v3
+	.align	4
+_v3:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",associative,_f4
+	.globl	_v4                     # @v4
+	.align	4
+_v4:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",associative,_f5
+	.globl	_v5                     # @v5
+	.align	4
+_v5:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",associative,_f6
+	.globl	_v6                     # @v6
+	.align	4
+_v6:
+	.long	0                       # 0x0
+
+	.section	.bss,"bw",same_size,_f6
+	.globl	_f6                     # @f6
+	.align	4
+_f6:
+	.long	0                       # 0x0
+
+	.section	.rdata,"rd"
+	.align	4                       # @some_name
+L_some_name:
+	.zero	8
+
+
+	.globl	_vftable
+_vftable = L_some_name+4

test/CodeGen/X86/coff-comdat2.ll

@@ -0,0 +1,9 @@
+; RUN: not llc %s -o /dev/null 2>&1 | FileCheck %s
+
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat largest
+@foo = global i32 0
+@bar = global i32 0, comdat $foo
+; CHECK: Associative COMDAT symbol 'foo' is not a key for it's COMDAT.

test/CodeGen/X86/coff-comdat3.ll

@@ -0,0 +1,8 @@
+; RUN: not llc %s -o /dev/null 2>&1 | FileCheck %s
+
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat largest
+@bar = global i32 0, comdat $foo
+; CHECK: Associative COMDAT symbol 'foo' does not exist.

test/CodeGen/X86/elf-comdat.ll

@@ -0,0 +1,11 @@
+; RUN: llc -mtriple x86_64-pc-linux-gnu < %s | FileCheck %s
+
+$f = comdat any
+@v = global i32 0, comdat $f
+define void @f() comdat $f {
+  ret void
+}
+; CHECK: .section        .text.f,"axG",@progbits,f,comdat
+; CHECK: .globl  f
+; CHECK: .section        .bss.v,"aGw",@nobits,f,comdat
+; CHECK: .globl  v

test/CodeGen/X86/elf-comdat2.ll

@@ -0,0 +1,12 @@
+; RUN: llc -mtriple x86_64-pc-linux-gnu < %s | FileCheck %s
+
+$foo = comdat any
+@bar = global i32 42, comdat $foo
+@foo = global i32 42
+
+; CHECK:      .type   bar,@object
+; CHECK-NEXT: .section        .data.bar,"aGw",@progbits,foo,comdat
+; CHECK-NEXT: .globl  bar
+; CHECK:      .type   foo,@object
+; CHECK-NEXT: .data
+; CHECK-NEXT: .globl  foo

test/CodeGen/X86/macho-comdat.ll

@@ -0,0 +1,6 @@
+; RUN: not llc -mtriple x86_64-apple-darwin < %s 2> %t
+; RUN: FileCheck < %t %s
+
+$f = comdat any
+@v = global i32 0, comdat $f
+; CHECK: LLVM ERROR: MachO doesn't support COMDATs, 'f' cannot be lowered.

test/Feature/comdat.ll

@@ -0,0 +1,15 @@
+; RUN: llvm-as < %s | llvm-dis | FileCheck %s
+
+$f = comdat any
+; CHECK: $f = comdat any
+
+$f2 = comdat any
+; CHECK-NOT: f2
+
+@v = global i32 0, comdat $f
+; CHECK: @v = global i32 0, comdat $f
+
+define void @f() comdat $f {
+  ret void
+}
+; CHECK: define void @f() comdat $f

test/Instrumentation/AddressSanitizer/do-not-instrument-llvm-metadata.ll

@@ -5,7 +5,7 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
 @.str_noinst = private unnamed_addr constant [4 x i8] c"aaa\00", section "llvm.metadata"
-@.str_inst = private unnamed_addr constant [4 x i8] c"aaa\00",
+@.str_inst = private unnamed_addr constant [4 x i8] c"aaa\00"
 
 ; CHECK-NOT: {{asan_gen.*str_noinst}}
 ; CHECK: {{asan_gen.*str_inst}}

test/Linker/Inputs/comdat.ll

@@ -0,0 +1,20 @@
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat largest
+@foo = global i64 43, comdat $foo
+
+define i32 @bar() comdat $foo {
+  ret i32 43
+}
+
+$qux = comdat largest
+@qux = global i32 13, comdat $qux
+@in_unselected_group = global i32 13, comdat $qux
+
+define i32 @baz() comdat $qux {
+  ret i32 13
+}
+
+$any = comdat any
+@any = global i64 7, comdat $any

test/Linker/Inputs/comdat2.ll

@@ -0,0 +1,2 @@
+$foo = comdat largest
+@foo = global i64 43, comdat $foo

test/Linker/Inputs/comdat3.ll

@@ -0,0 +1,2 @@
+$foo = comdat noduplicates
+@foo = global i64 43, comdat $foo

test/Linker/Inputs/comdat4.ll

@@ -0,0 +1,5 @@
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat samesize
+@foo = global i64 42, comdat $foo

test/Linker/Inputs/comdat5.ll

@@ -0,0 +1,15 @@
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+%MSRTTICompleteObjectLocator = type { i32, i32, i32, i8*, %MSRTTIClassHierarchyDescriptor* }
+%MSRTTIClassHierarchyDescriptor = type { i32, i32, i32, %MSRTTIBaseClassDescriptor** }
+%MSRTTIBaseClassDescriptor = type { i8*, i32, i32, i32, i32, i32, %MSRTTIClassHierarchyDescriptor* }
+%struct.S = type { i32 (...)** }
+
+$"\01??_7S@@6B@" = comdat largest
+
+@"\01??_R4S@@6B@" = external constant %MSRTTICompleteObjectLocator
+@some_name = private unnamed_addr constant [2 x i8*] [i8* bitcast (%MSRTTICompleteObjectLocator* @"\01??_R4S@@6B@" to i8*), i8* bitcast (void (%struct.S*, i32)* @"\01??_GS@@UAEPAXI@Z" to i8*)], comdat $"\01??_7S@@6B@"
+@"\01??_7S@@6B@" = alias getelementptr([2 x i8*]* @some_name, i32 0, i32 1)
+
+declare x86_thiscallcc void @"\01??_GS@@UAEPAXI@Z"(%struct.S*, i32) unnamed_addr

test/Linker/comdat.ll

@@ -0,0 +1,32 @@
+; RUN: llvm-link %s %p/Inputs/comdat.ll -S -o - | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat largest
+@foo = global i32 42, comdat $foo
+
+define i32 @bar() comdat $foo {
+  ret i32 42
+}
+
+$qux = comdat largest
+@qux = global i64 12, comdat $qux
+
+define i32 @baz() comdat $qux {
+  ret i32 12
+}
+
+$any = comdat any
+@any = global i64 6, comdat $any
+
+; CHECK: $qux = comdat largest
+; CHECK: $foo = comdat largest
+; CHECK: $any = comdat any
+
+; CHECK: @qux = global i64 12, comdat $qux
+; CHECK: @any = global i64 6, comdat $any
+; CHECK: @foo = global i64 43, comdat $foo
+; CHECK-NOT: @in_unselected_group = global i32 13, comdat $qux
+
+; CHECK: define i32 @baz() comdat $qux
+; CHECK: define i32 @bar() comdat $foo

test/Linker/comdat2.ll

@@ -0,0 +1,7 @@
+; RUN: not llvm-link %s %p/Inputs/comdat.ll -S -o - 2>&1 | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat samesize
+@foo = global i32 42, comdat $foo
+; CHECK: Linking COMDATs named 'foo': invalid selection kinds!

test/Linker/comdat3.ll

@@ -0,0 +1,5 @@
+; RUN: not llvm-link %s %p/Inputs/comdat2.ll -S -o - 2>&1 | FileCheck %s
+
+$foo = comdat largest
+@foo = global i32 43, comdat $foo
+; CHECK: Linking COMDATs named 'foo': can't do size dependent selection without DataLayout!

test/Linker/comdat4.ll

@@ -0,0 +1,5 @@
+; RUN: not llvm-link %s %p/Inputs/comdat3.ll -S -o - 2>&1 | FileCheck %s
+
+$foo = comdat noduplicates
+@foo = global i64 43, comdat $foo
+; CHECK: Linking COMDATs named 'foo': noduplicates has been violated!

test/Linker/comdat5.ll

@@ -0,0 +1,7 @@
+; RUN: not llvm-link %s %p/Inputs/comdat4.ll -S -o - 2>&1 | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$foo = comdat samesize
+@foo = global i32 42, comdat $foo
+; CHECK: Linking COMDATs named 'foo': SameSize violated!

test/Linker/comdat6.ll

@@ -0,0 +1,13 @@
+; RUN: llvm-link %s %p/Inputs/comdat5.ll -S -o - 2>&1 | FileCheck %s
+; RUN: llvm-link %p/Inputs/comdat5.ll %s -S -o - 2>&1 | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+%struct.S = type { i32 (...)** }
+
+$"\01??_7S@@6B@" = comdat largest
+@"\01??_7S@@6B@" = linkonce_odr unnamed_addr constant [1 x i8*] [i8* bitcast (void (%struct.S*, i32)* @"\01??_GS@@UAEPAXI@Z" to i8*)], comdat $"\01??_7S@@6B@"
+
+; CHECK: @"\01??_7S@@6B@" = alias getelementptr inbounds ([2 x i8*]* @some_name, i32 0, i32 1)
+
+declare x86_thiscallcc void @"\01??_GS@@UAEPAXI@Z"(%struct.S*, i32) unnamed_addr

test/Linker/comdat7.ll

@@ -0,0 +1,9 @@
+; RUN: not llvm-link %s %p/Inputs/comdat5.ll -S -o - 2>&1 | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$"\01??_7S@@6B@" = comdat largest
+define void @"\01??_7S@@6B@"() {
+  ret void
+}
+; CHECK: GlobalVariable required for data dependent selection!

test/Linker/comdat8.ll

@@ -0,0 +1,10 @@
+; RUN: not llvm-link %s %p/Inputs/comdat5.ll -S -o - 2>&1 | FileCheck %s
+target datalayout = "e-m:w-p:32:32-i64:64-f80:32-n8:16:32-S32"
+target triple = "i686-pc-windows-msvc"
+
+$"\01??_7S@@6B@" = comdat largest
+define void @some_name() {
+  ret void
+}
+@"\01??_7S@@6B@" = alias i8* inttoptr (i32 ptrtoint (void ()* @some_name to i32) to i8*)
+; CHECK: COMDAT key involves incomputable alias size.

test/Verifier/comdat.ll

@@ -0,0 +1,5 @@
+; RUN: not llvm-as %s -o /dev/null 2>&1 | FileCheck %s
+
+$v = comdat any
+@v = common global i32 0, comdat $v
+; CHECK: 'common' global may not be in a Comdat!

test/Verifier/comdat2.ll

@@ -0,0 +1,5 @@
+; RUN: not llvm-as %s -o /dev/null 2>&1 | FileCheck %s
+
+$v = comdat any
+@v = private global i32 0, comdat $v
+; CHECK: comdat global value has local linkage