PTH: Use Token::setLiteralData() to directly store a pointer to cached spelling data in the PTH file. This removes a ton of code for looking up spellings using sourcelocations in the PTH file. This simplifies both PTH-generation and reading.

Performance impact for -fsyntax-only on Cocoa.h (with Cocoa.h in the PTH file):
- PTH generation time improves by 5%
- PTH reading improves by 0.3%.


git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@63072 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 11 insertions, 60 deletions

diff --git a/include/clang/Lex/PTHLexer.h b/include/clang/Lex/PTHLexer.h
index c6837d2a8b..369b818a1f 100644
--- a/include/clang/Lex/PTHLexer.h
+++ b/include/clang/Lex/PTHLexer.h
@@ -45,10 +45,6 @@ class PTHLexer : public PreprocessorLexer {
   ///  to process when doing quick skipping of preprocessor blocks.
   const unsigned char* CurPPCondPtr;
 
-  /// MySpellingMgr - Reference to the spelling manager used to get spellings
-  ///  for the source file indicated by \c FileID.
-  PTHSpellingSearch& MySpellingSrch;  
-
   PTHLexer(const PTHLexer&);  // DO NOT IMPLEMENT
   void operator=(const PTHLexer&); // DO NOT IMPLEMENT
   
@@ -65,8 +61,7 @@ protected:
 
   /// Create a PTHLexer for the specified token stream.
   PTHLexer(Preprocessor& pp, FileID FID, const unsigned char *D,
-           const unsigned char* ppcond,
-           PTHSpellingSearch& mySpellingSrch, PTHManager &PM);
+           const unsigned char* ppcond, PTHManager &PM);
 public:  
 
   ~PTHLexer() {}
@@ -95,16 +90,6 @@ public:
   /// IndirectLex - An indirect call to 'Lex' that can be invoked via
   ///  the PreprocessorLexer interface.
   void IndirectLex(Token &Result) { Lex(Result); }
-
-  /// Returns the cached spelling of a token.
-  /// \param[in] sloc The SourceLocation of the token.
-  /// \param[out] Buffer If a token's spelling is found in the PTH file then
-  ///   upon exit from this method \c Buffer will be set to the address of
-  ///   the character array representing that spelling.  No characters
-  ///   are copied.
-  /// \returns The number of characters for the spelling of the token.  This
-  ///   value is 0 if the spelling could not be found in the PTH file.
-  unsigned getSpelling(SourceLocation sloc, const char *&Buffer);
   
   /// getSourceLocation - Return a source location for the token in
   /// the current file.
diff --git a/include/clang/Lex/PTHManager.h b/include/clang/Lex/PTHManager.h
index 31cf78f036..b77cda1f0b 100644
--- a/include/clang/Lex/PTHManager.h
+++ b/include/clang/Lex/PTHManager.h
@@ -29,42 +29,13 @@ namespace clang {
 
 class FileEntry;
 class PTHLexer;
-class PTHManager;
-
-class PTHSpellingSearch {
-  PTHManager& PTHMgr;
-  
-  const unsigned char* const TableBeg;
-  const unsigned char* const TableEnd;
-  
-  const unsigned NumSpellings;
-  const unsigned char* LinearItr;
-  
-public:
-  enum { SpellingEntrySize = 4*2 };
-  
-  unsigned getSpellingBinarySearch(unsigned fpos, const char *&Buffer);
-  unsigned getSpellingLinearSearch(unsigned fpos, const char *&Buffer);
-  
-  PTHSpellingSearch(PTHManager& pm, unsigned numSpellings,
-                    const unsigned char* tableBeg)
-    : PTHMgr(pm),
-      TableBeg(tableBeg),
-      TableEnd(tableBeg + numSpellings*SpellingEntrySize),
-      NumSpellings(numSpellings),
-      LinearItr(tableBeg) {}
-};  
   
 class PTHManager : public IdentifierInfoLookup {
   friend class PTHLexer;
-  friend class PTHSpellingSearch;
   
   /// The memory mapped PTH file.
   const llvm::MemoryBuffer* Buf;
-  
-  /// A map from FileIDs to SpellingSearch objects.
-  llvm::DenseMap<FileID, PTHSpellingSearch*> SpellingMap;
-  
+
   /// Alloc - Allocator used for IdentifierInfo objects.
   llvm::BumpPtrAllocator Alloc;
   
@@ -84,7 +55,7 @@ class PTHManager : public IdentifierInfoLookup {
   /// SortedIdTable - Array ordering persistent identifier IDs by the lexical
   ///  order of their corresponding strings.  This is used by get().
   const unsigned char* const SortedIdTable;
-  
+
   /// NumIds - The number of identifiers in the PTH file.
   const unsigned NumIds;
 
@@ -92,11 +63,16 @@ class PTHManager : public IdentifierInfoLookup {
   ///  PTHLexer objects.
   Preprocessor* PP;
   
+  /// SpellingBase - The base offset within the PTH memory buffer that 
+  ///  contains the cached spellings for literals.
+  const unsigned char* const SpellingBase;
+  
   /// This constructor is intended to only be called by the static 'Create'
   /// method.
   PTHManager(const llvm::MemoryBuffer* buf, void* fileLookup,
              const unsigned char* idDataTable, IdentifierInfo** perIDCache,
-             const unsigned char* sortedIdTable, unsigned numIds);
+             const unsigned char* sortedIdTable, unsigned numIds,
+             const unsigned char* spellingBase);
 
   // Do not implement.
   PTHManager();
@@ -119,7 +95,7 @@ class PTHManager : public IdentifierInfoLookup {
   
 public:
   // The current PTH version.
-  enum { Version = 0 };
+  enum { Version = 1 };
 
   ~PTHManager();
   
@@ -138,11 +114,7 @@ public:
   /// CreateLexer - Return a PTHLexer that "lexes" the cached tokens for the
   ///  specified file.  This method returns NULL if no cached tokens exist.
   ///  It is the responsibility of the caller to 'delete' the returned object.
-  PTHLexer *CreateLexer(FileID FID);
-  
-  unsigned getSpelling(SourceLocation Loc, const char *&Buffer);  
-private:
-  unsigned getSpelling(FileID FID, unsigned fpos, const char *& Buffer);  
+  PTHLexer *CreateLexer(FileID FID);  
 };
   
 }  // end namespace clang
diff --git a/include/clang/Lex/Preprocessor.h b/include/clang/Lex/Preprocessor.h
index 095c1db1de..4e57069385 100644
--- a/include/clang/Lex/Preprocessor.h
+++ b/include/clang/Lex/Preprocessor.h
@@ -461,12 +461,6 @@ public:
     // If the token is carrying a literal data pointer, just use it.
     if (const char *D = Tok.getLiteralData())
       return *D;
-    
-    if (PTH) {
-      const char *Data;
-      if (PTH->getSpelling(Tok.getLocation(), Data))
-        return *Data;
-    }
 
     // Otherwise, fall back on getCharacterData, which is slower, but always
     // works.