Implement support for globally associating abbrevs with block IDs, which

relieves us from having to emit the abbrevs into each instance of the block.
This shrinks kc.bit from 3368K to 3333K, but will be a more significant win
once instructions are abbreviated.

The VST went from:

  Block ID #14 (VALUE_SYMTAB):
      Num Instances: 2345
         Total Size: 1.29508e+07b/1.61885e+06B/404713W
       Average Size: 5522.73b/690.342B/172.585W
          % of file: 48.0645
  Tot/Avg SubBlocks: 0/0
    Tot/Avg Abbrevs: 7035/3
    Tot/Avg Records: 120924/51.5667
      % Abbrev Recs: 100

to:

  Block ID #14 (VALUE_SYMTAB):
      Num Instances: 2345
         Total Size: 1.26713e+07b/1.58391e+06B/395978W
       Average Size: 5403.53b/675.442B/168.86W
          % of file: 47.5198
  Tot/Avg SubBlocks: 0/0
    Tot/Avg Abbrevs: 0/0
    Tot/Avg Records: 120924/51.5667
      % Abbrev Recs: 100

because we didn't emit the same 3 abbrevs 2345 times :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@36767 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/Bitcode/BitstreamReader.h

@@ -48,6 +48,14 @@ class BitstreamReader {
   /// BlockScope - This tracks the codesize of parent blocks.
   SmallVector<Block, 8> BlockScope;
 
+  /// BlockInfo - This contains information emitted to BLOCKINFO_BLOCK blocks.
+  /// These describe abbreviations that all blocks of the specified ID inherit.
+  struct BlockInfo {
+    unsigned BlockID;
+    std::vector<BitCodeAbbrev*> Abbrevs;
+  };
+  std::vector<BlockInfo> BlockInfoRecords;
+  
   /// FirstChar - This remembers the first byte of the stream.
   const unsigned char *FirstChar;
 public:
@@ -82,6 +90,15 @@ public:
       for (unsigned i = 0, e = Abbrevs.size(); i != e; ++i)
         Abbrevs[i]->dropRef();
     }
+    
+    // Free the BlockInfoRecords.
+    while (!BlockInfoRecords.empty()) {
+      BlockInfo &Info = BlockInfoRecords.back();
+      // Free blockinfo abbrev info.
+      for (unsigned i = 0, e = Info.Abbrevs.size(); i != e; ++i)
+        Info.Abbrevs[i]->dropRef();
+      BlockInfoRecords.pop_back();
+    }
   }
   
   bool AtEndOfStream() const { return NextChar == LastChar; }
@@ -206,6 +223,22 @@ public:
   // Block Manipulation
   //===--------------------------------------------------------------------===//
   
+private:
+  /// getBlockInfo - If there is block info for the specified ID, return it,
+  /// otherwise return null.
+  BlockInfo *getBlockInfo(unsigned BlockID) {
+    // Common case, the most recent entry matches BlockID.
+    if (!BlockInfoRecords.empty() && BlockInfoRecords.back().BlockID == BlockID)
+      return &BlockInfoRecords.back();
+    
+    for (unsigned i = 0, e = BlockInfoRecords.size(); i != e; ++i)
+      if (BlockInfoRecords[i].BlockID == BlockID)
+        return &BlockInfoRecords[i];
+    return 0;
+  }
+public:
+  
+  
   // Block header:
   //    [ENTER_SUBBLOCK, blockid, newcodelen, <align4bytes>, blocklen]
 
@@ -236,10 +269,19 @@ public:
   
   /// EnterSubBlock - Having read the ENTER_SUBBLOCK abbrevid, read and enter
   /// the block, returning the BlockID of the block we just entered.
-  bool EnterSubBlock(unsigned *NumWordsP = 0) {
+  bool EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = 0) {
+    // Save the current block's state on BlockScope.
     BlockScope.push_back(Block(CurCodeSize));
     BlockScope.back().PrevAbbrevs.swap(CurAbbrevs);
     
+    // Add the abbrevs specific to this block to the CurAbbrevs list.
+    if (BlockInfo *Info = getBlockInfo(BlockID)) {
+      for (unsigned i = 0, e = Info->Abbrevs.size(); i != e; ++i) {
+        CurAbbrevs.push_back(Info->Abbrevs[i]);
+        CurAbbrevs.back()->addRef();
+      }
+    }
+    
     // Get the codesize of this block.
     CurCodeSize = ReadVBR(bitc::CodeLenWidth);
     SkipToWord();
@@ -352,6 +394,64 @@ public:
     }
     CurAbbrevs.push_back(Abbv);
   }
+  
+  //===--------------------------------------------------------------------===//
+  // BlockInfo Block Reading
+  //===--------------------------------------------------------------------===//
+  
+private:  
+  BlockInfo &getOrCreateBlockInfo(unsigned BlockID) {
+    if (BlockInfo *BI = getBlockInfo(BlockID))
+      return *BI;
+    
+    // Otherwise, add a new record.
+    BlockInfoRecords.push_back(BlockInfo());
+    BlockInfoRecords.back().BlockID = BlockID;
+    return BlockInfoRecords.back();
+  }
+  
+public:
+    
+  bool ReadBlockInfoBlock() {
+    if (EnterSubBlock(bitc::BLOCKINFO_BLOCK_ID)) return true;
+
+    SmallVector<uint64_t, 64> Record;
+    BlockInfo *CurBlockInfo = 0;
+    
+    // Read all the records for this module.
+    while (1) {
+      unsigned Code = ReadCode();
+      if (Code == bitc::END_BLOCK)
+        return ReadBlockEnd();
+      if (Code == bitc::ENTER_SUBBLOCK) {
+        ReadSubBlockID();
+        if (SkipBlock()) return true;
+        continue;
+      }
+
+      // Read abbrev records, associate them with CurBID.
+      if (Code == bitc::DEFINE_ABBREV) {
+        if (!CurBlockInfo) return true;
+        ReadAbbrevRecord();
+        
+        // ReadAbbrevRecord installs the abbrev in CurAbbrevs.  Move it to the
+        // appropriate BlockInfo.
+        BitCodeAbbrev *Abbv = CurAbbrevs.back();
+        CurAbbrevs.pop_back();
+        CurBlockInfo->Abbrevs.push_back(Abbv);
+        continue;
+      }
+
+      // Read a record.
+      switch (ReadRecord(Code, Record)) {
+      default: break;  // Default behavior, ignore unknown content.
+      case bitc::BLOCKINFO_CODE_SETBID:
+        if (Record.size() < 1) return true;
+        CurBlockInfo = &getOrCreateBlockInfo(Record[0]);
+        break;
+      }
+    }      
+  }
 };
 
 } // End llvm namespace