[serialization] no transitive decl change (#92083)

Following of #86912

The motivation of the patch series is that, for a module interface unit
`X`, when the dependent modules of `X` changes, if the changes is not
relevant with `X`, we hope the BMI of `X` won't change. For the specific
patch, we hope if the changes was about irrelevant declaration changes,
we hope the BMI of `X` won't change. **However**, I found the patch
itself is not very useful in practice, since the adding or removing
declarations, will change the state of identifiers and types in most
cases.

That said, for the most simple example,

```
// partA.cppm
export module m:partA;

// partA.v1.cppm
export module m:partA;
export void a() {}

// partB.cppm
export module m:partB;
export void b() {}

// m.cppm
export module m;
export import :partA;
export import :partB;

// onlyUseB;
export module onlyUseB;
import m;
export inline void onluUseB() {
    b();
}
```

the BMI of `onlyUseB` will change after we change the implementation of
`partA.cppm` to `partA.v1.cppm`. Since `partA.v1.cppm` introduces new
identifiers and types (the function prototype).

So in this patch, we have to write the tests as:

```
// partA.cppm
export module m:partA;
export int getA() { ... }
export int getA2(int) { ... }

// partA.v1.cppm
export module m:partA;
export int getA() { ... }
export int getA(int) { ... }
export int getA2(int) { ... }

// partB.cppm
export module m:partB;
export void b() {}

// m.cppm
export module m;
export import :partA;
export import :partB;

// onlyUseB;
export module onlyUseB;
import m;
export inline void onluUseB() {
    b();
}
```

so that the new introduced declaration `int getA(int)` doesn't introduce
new identifiers and types, then the BMI of `onlyUseB` can keep
unchanged.

While it looks not so great, the patch should be the base of the patch
to erase the transitive change for identifiers and types since I don't
know how can we introduce new types and identifiers without introducing
new declarations. Given how tightly the relationship between
declarations, types and identifiers, I think we can only reach the ideal
state after we made the series for all of the three entties.

The design of the patch is similar to
#86912, which extends the
32-bit DeclID to 64-bit and use the higher bits to store the module file
index and the lower bits to store the Local Decl ID.

A slight difference is that we only use 48 bits to store the new DeclID
since we try to use the higher 16 bits to store the module ID in the
prefix of Decl class. Previously, we use 32 bits to store the module ID
and 32 bits to store the DeclID. I don't want to allocate additional
space so I tried to make the additional space the same as 64 bits. An
potential interesting thing here is about the relationship between the
module ID and the module file index. I feel we can get the module file
index by the module ID. But I didn't prove it or implement it. Since I
want to make the patch itself as small as possible. We can make it in
the future if we want.

Another change in the patch is the new concept Decl Index, which means
the index of the very big array `DeclsLoaded` in ASTReader. Previously,
the index of a loaded declaration is simply the Decl ID minus
PREDEFINED_DECL_NUMs. So there are some places they got used
ambiguously. But this patch tried to split these two concepts.

As #86912 did, the change will
increase the on-disk PCM file sizes. As the declaration ID may be the
most IDs in the PCM file, this can have the biggest impact on the size.
In my experiments, this change will bring 6.6% increase of the on-disk
PCM size. No compile-time performance regression observed. Given the
benefits in the motivation example, I think the cost is worthwhile.

Author: ChuanqiXu9

clang/include/clang/AST/DeclBase.h

@@ -701,10 +701,7 @@ class alignas(8) Decl {
 
   /// Set the owning module ID.  This may only be called for
   /// deserialized Decls.
-  void setOwningModuleID(unsigned ID) {
-    assert(isFromASTFile() && "Only works on a deserialized declaration");
-    *((unsigned*)this - 2) = ID;
-  }
+  void setOwningModuleID(unsigned ID);
 
 public:
   /// Determine the availability of the given declaration.
@@ -777,19 +774,11 @@ class alignas(8) Decl {
 
   /// Retrieve the global declaration ID associated with this
   /// declaration, which specifies where this Decl was loaded from.
-  GlobalDeclID getGlobalID() const {
-    if (isFromASTFile())
-      return (*((const GlobalDeclID *)this - 1));
-    return GlobalDeclID();
-  }
+  GlobalDeclID getGlobalID() const;
 
   /// Retrieve the global ID of the module that owns this particular
   /// declaration.
-  unsigned getOwningModuleID() const {
-    if (isFromASTFile())
-      return *((const unsigned*)this - 2);
-    return 0;
-  }
+  unsigned getOwningModuleID() const;
 
 private:
   Module *getOwningModuleSlow() const;

clang/include/clang/AST/DeclID.h

@@ -19,6 +19,8 @@
 #include "llvm/ADT/DenseMapInfo.h"
 #include "llvm/ADT/iterator.h"
 
+#include <climits>
+
 namespace clang {
 
 /// Predefined declaration IDs.
@@ -107,12 +109,16 @@ class DeclIDBase {
   ///
   /// DeclID should only be used directly in serialization. All other users
   /// should use LocalDeclID or GlobalDeclID.
-  using DeclID = uint32_t;
+  using DeclID = uint64_t;
 
 protected:
   DeclIDBase() : ID(PREDEF_DECL_NULL_ID) {}
   explicit DeclIDBase(DeclID ID) : ID(ID) {}
 
+  explicit DeclIDBase(unsigned LocalID, unsigned ModuleFileIndex) {
+    ID = (DeclID)LocalID | ((DeclID)ModuleFileIndex << 32);
+  }
+
 public:
   DeclID get() const { return ID; }
 
@@ -124,6 +130,10 @@ class DeclIDBase {
 
   bool isInvalid() const { return ID == PREDEF_DECL_NULL_ID; }
 
+  unsigned getModuleFileIndex() const { return ID >> 32; }
+
+  unsigned getLocalDeclIndex() const;
+
   friend bool operator==(const DeclIDBase &LHS, const DeclIDBase &RHS) {
     return LHS.ID == RHS.ID;
   }
@@ -156,6 +166,9 @@ class LocalDeclID : public DeclIDBase {
   LocalDeclID(PredefinedDeclIDs ID) : Base(ID) {}
   explicit LocalDeclID(DeclID ID) : Base(ID) {}
 
+  explicit LocalDeclID(unsigned LocalID, unsigned ModuleFileIndex)
+      : Base(LocalID, ModuleFileIndex) {}
+
   LocalDeclID &operator++() {
     ++ID;
     return *this;
@@ -175,6 +188,9 @@ class GlobalDeclID : public DeclIDBase {
   GlobalDeclID() : Base() {}
   explicit GlobalDeclID(DeclID ID) : Base(ID) {}
 
+  explicit GlobalDeclID(unsigned LocalID, unsigned ModuleFileIndex)
+      : Base(LocalID, ModuleFileIndex) {}
+
   // For DeclIDIterator<GlobalDeclID> to be able to convert a GlobalDeclID
   // to a LocalDeclID.
   explicit operator LocalDeclID() const { return LocalDeclID(this->ID); }

clang/include/clang/Serialization/ASTBitCodes.h

@@ -255,6 +255,12 @@ class DeclOffset {
   }
 };
 
+// The unaligned decl ID used in the Blobs of bistreams.
+using unaligned_decl_id_t =
+    llvm::support::detail::packed_endian_specific_integral<
+        serialization::DeclID, llvm::endianness::native,
+        llvm::support::unaligned>;
+
 /// The number of predefined preprocessed entity IDs.
 const unsigned int NUM_PREDEF_PP_ENTITY_IDS = 1;
 
@@ -1979,33 +1985,46 @@ enum CleanupObjectKind { COK_Block, COK_CompoundLiteral };
 
 /// Describes the categories of an Objective-C class.
 struct ObjCCategoriesInfo {
-  // The ID of the definition
-  LocalDeclID DefinitionID;
+  // The ID of the definition. Use unaligned_decl_id_t to keep
+  // ObjCCategoriesInfo 32-bit aligned.
+  unaligned_decl_id_t DefinitionID;
 
   // Offset into the array of category lists.
   unsigned Offset;
 
+  ObjCCategoriesInfo() = default;
+  ObjCCategoriesInfo(LocalDeclID ID, unsigned Offset)
+    : DefinitionID(ID.get()), Offset(Offset) {}
+
+  LocalDeclID getDefinitionID() const {
+    return LocalDeclID(DefinitionID);
+  }
+
   friend bool operator<(const ObjCCategoriesInfo &X,
                         const ObjCCategoriesInfo &Y) {
-    return X.DefinitionID < Y.DefinitionID;
+    return X.getDefinitionID() < Y.getDefinitionID();
   }
 
   friend bool operator>(const ObjCCategoriesInfo &X,
                         const ObjCCategoriesInfo &Y) {
-    return X.DefinitionID > Y.DefinitionID;
+    return X.getDefinitionID() > Y.getDefinitionID();
   }
 
   friend bool operator<=(const ObjCCategoriesInfo &X,
                          const ObjCCategoriesInfo &Y) {
-    return X.DefinitionID <= Y.DefinitionID;
+    return X.getDefinitionID() <= Y.getDefinitionID();
   }
 
   friend bool operator>=(const ObjCCategoriesInfo &X,
                          const ObjCCategoriesInfo &Y) {
-    return X.DefinitionID >= Y.DefinitionID;
+    return X.getDefinitionID() >= Y.getDefinitionID();
   }
 };
 
+static_assert(alignof(ObjCCategoriesInfo) <= 4);
+static_assert(std::is_standard_layout_v<ObjCCategoriesInfo> &&
+              std::is_trivial_v<ObjCCategoriesInfo>);
+
 /// A key used when looking up entities by \ref DeclarationName.
 ///
 /// Different \ref DeclarationNames are mapped to different keys, but the

clang/include/clang/Serialization/ASTReader.h

@@ -504,12 +504,6 @@ class ASTReader
   /// = I + 1 has already been loaded.
   llvm::PagedVector<Decl *> DeclsLoaded;
 
-  using GlobalDeclMapType = ContinuousRangeMap<GlobalDeclID, ModuleFile *, 4>;
-
-  /// Mapping from global declaration IDs to the module in which the
-  /// declaration resides.
-  GlobalDeclMapType GlobalDeclMap;
-
   using FileOffset = std::pair<ModuleFile *, uint64_t>;
   using FileOffsetsTy = SmallVector<FileOffset, 2>;
   using DeclUpdateOffsetsMap = llvm::DenseMap<GlobalDeclID, FileOffsetsTy>;
@@ -592,10 +586,11 @@ class ASTReader
 
   struct FileDeclsInfo {
     ModuleFile *Mod = nullptr;
-    ArrayRef<LocalDeclID> Decls;
+    ArrayRef<serialization::unaligned_decl_id_t> Decls;
 
     FileDeclsInfo() = default;
-    FileDeclsInfo(ModuleFile *Mod, ArrayRef<LocalDeclID> Decls)
+    FileDeclsInfo(ModuleFile *Mod,
+                  ArrayRef<serialization::unaligned_decl_id_t> Decls)
         : Mod(Mod), Decls(Decls) {}
   };
 
@@ -604,11 +599,7 @@ class ASTReader
 
   /// An array of lexical contents of a declaration context, as a sequence of
   /// Decl::Kind, DeclID pairs.
-  using unaligned_decl_id_t =
-      llvm::support::detail::packed_endian_specific_integral<
-          serialization::DeclID, llvm::endianness::native,
-          llvm::support::unaligned>;
-  using LexicalContents = ArrayRef<unaligned_decl_id_t>;
+  using LexicalContents = ArrayRef<serialization::unaligned_decl_id_t>;
 
   /// Map from a DeclContext to its lexical contents.
   llvm::DenseMap<const DeclContext*, std::pair<ModuleFile*, LexicalContents>>
@@ -1489,22 +1480,23 @@ class ASTReader
                                unsigned ClientLoadCapabilities);
 
 public:
-  class ModuleDeclIterator : public llvm::iterator_adaptor_base<
-                                 ModuleDeclIterator, const LocalDeclID *,
-                                 std::random_access_iterator_tag, const Decl *,
-                                 ptrdiff_t, const Decl *, const Decl *> {
+  class ModuleDeclIterator
+      : public llvm::iterator_adaptor_base<
+            ModuleDeclIterator, const serialization::unaligned_decl_id_t *,
+            std::random_access_iterator_tag, const Decl *, ptrdiff_t,
+            const Decl *, const Decl *> {
     ASTReader *Reader = nullptr;
     ModuleFile *Mod = nullptr;
 
   public:
     ModuleDeclIterator() : iterator_adaptor_base(nullptr) {}
 
     ModuleDeclIterator(ASTReader *Reader, ModuleFile *Mod,
-                       const LocalDeclID *Pos)
+                       const serialization::unaligned_decl_id_t *Pos)
         : iterator_adaptor_base(Pos), Reader(Reader), Mod(Mod) {}
 
     value_type operator*() const {
-      return Reader->GetDecl(Reader->getGlobalDeclID(*Mod, *I));
+      return Reader->GetDecl(Reader->getGlobalDeclID(*Mod, (LocalDeclID)*I));
     }
 
     value_type operator->() const { return **this; }
@@ -1544,6 +1536,9 @@ class ASTReader
              StringRef Arg2 = StringRef(), StringRef Arg3 = StringRef()) const;
   void Error(llvm::Error &&Err) const;
 
+  /// Translate a \param GlobalDeclID to the index of DeclsLoaded array.
+  unsigned translateGlobalDeclIDToIndex(GlobalDeclID ID) const;
+
 public:
   /// Load the AST file and validate its contents against the given
   /// Preprocessor.
@@ -1915,7 +1910,8 @@ class ASTReader
 
   /// Retrieve the module file that owns the given declaration, or NULL
   /// if the declaration is not from a module file.
-  ModuleFile *getOwningModuleFile(const Decl *D);
+  ModuleFile *getOwningModuleFile(const Decl *D) const;
+  ModuleFile *getOwningModuleFile(GlobalDeclID ID) const;
 
   /// Returns the source location for the decl \p ID.
   SourceLocation getSourceLocationForDeclID(GlobalDeclID ID);

clang/include/clang/Serialization/ModuleFile.h

@@ -454,23 +454,11 @@ class ModuleFile {
   /// by the declaration ID (-1).
   const DeclOffset *DeclOffsets = nullptr;
 
-  /// Base declaration ID for declarations local to this module.
-  serialization::DeclID BaseDeclID = 0;
-
-  /// Remapping table for declaration IDs in this module.
-  ContinuousRangeMap<serialization::DeclID, int, 2> DeclRemap;
-
-  /// Mapping from the module files that this module file depends on
-  /// to the base declaration ID for that module as it is understood within this
-  /// module.
-  ///
-  /// This is effectively a reverse global-to-local mapping for declaration
-  /// IDs, so that we can interpret a true global ID (for this translation unit)
-  /// as a local ID (for this module file).
-  llvm::DenseMap<ModuleFile *, serialization::DeclID> GlobalToLocalDeclIDs;
+  /// Base declaration index in ASTReader for declarations local to this module.
+  unsigned BaseDeclIndex = 0;
 
   /// Array of file-level DeclIDs sorted by file.
-  const LocalDeclID *FileSortedDecls = nullptr;
+  const serialization::unaligned_decl_id_t *FileSortedDecls = nullptr;
   unsigned NumFileSortedDecls = 0;
 
   /// Array of category list location information within this

clang/include/clang/Serialization/ModuleManager.h

@@ -45,7 +45,7 @@ namespace serialization {
 /// Manages the set of modules loaded by an AST reader.
 class ModuleManager {
   /// The chain of AST files, in the order in which we started to load
-  /// them (this order isn't really useful for anything).
+  /// them.
   SmallVector<std::unique_ptr<ModuleFile>, 2> Chain;
 
   /// The chain of non-module PCH files. The first entry is the one named

clang/lib/AST/DeclBase.cpp

@@ -74,18 +74,17 @@ void *Decl::operator new(std::size_t Size, const ASTContext &Context,
                          GlobalDeclID ID, std::size_t Extra) {
   // Allocate an extra 8 bytes worth of storage, which ensures that the
   // resulting pointer will still be 8-byte aligned.
-  static_assert(sizeof(unsigned) * 2 >= alignof(Decl),
-                "Decl won't be misaligned");
+  static_assert(sizeof(uint64_t) >= alignof(Decl), "Decl won't be misaligned");
   void *Start = Context.Allocate(Size + Extra + 8);
   void *Result = (char*)Start + 8;
 
-  unsigned *PrefixPtr = (unsigned *)Result - 2;
+  uint64_t *PrefixPtr = (uint64_t *)Result - 1;
 
-  // Zero out the first 4 bytes; this is used to store the owning module ID.
-  PrefixPtr[0] = 0;
+  *PrefixPtr = ID.get();
 
-  // Store the global declaration ID in the second 4 bytes.
-  PrefixPtr[1] = ID.get();
+  // We leave the upper 16 bits to store the module IDs. 48 bits should be
+  // sufficient to store a declaration ID.
+  assert(*PrefixPtr < llvm::maskTrailingOnes<uint64_t>(48));
 
   return Result;
 }
@@ -111,6 +110,28 @@ void *Decl::operator new(std::size_t Size, const ASTContext &Ctx,
   return ::operator new(Size + Extra, Ctx);
 }
 
+GlobalDeclID Decl::getGlobalID() const {
+  if (!isFromASTFile())
+    return GlobalDeclID();
+  // See the comments in `Decl::operator new` for details.
+  uint64_t ID = *((const uint64_t *)this - 1);
+  return GlobalDeclID(ID & llvm::maskTrailingOnes<uint64_t>(48));
+}
+
+unsigned Decl::getOwningModuleID() const {
+  if (!isFromASTFile())
+    return 0;
+
+  uint64_t ID = *((const uint64_t *)this - 1);
+  return ID >> 48;
+}
+
+void Decl::setOwningModuleID(unsigned ID) {
+  assert(isFromASTFile() && "Only works on a deserialized declaration");
+  uint64_t *IDAddress = (uint64_t *)this - 1;
+  *IDAddress |= (uint64_t)ID << 48;
+}
+
 Module *Decl::getOwningModuleSlow() const {
   assert(isFromASTFile() && "Not from AST file?");
   return getASTContext().getExternalSource()->getModule(getOwningModuleID());
@@ -2164,3 +2185,7 @@ DependentDiagnostic *DependentDiagnostic::Create(ASTContext &C,
 
   return DD;
 }
+
+unsigned DeclIDBase::getLocalDeclIndex() const {
+  return ID & llvm::maskTrailingOnes<DeclID>(32);
+}