[pdb] Add a new command for analyzing hash collisions.

This introduces the `analyze` subcommand.  For now there is only
one option, to analyze hash collisions in the type streams.  In
the future, however, we could add many more things here, such
as performing size analyses, compacting, and statistics about
the type of records etc.

llvm-svn: 293795

include/llvm/DebugInfo/CodeView/TypeDatabase.h

@@ -34,6 +34,8 @@ public:
 
   StringRef getTypeName(TypeIndex Index) const;
 
+  const CVType &getTypeRecord(TypeIndex Index) const;
+
   bool containsTypeIndex(TypeIndex Index) const;
 
   uint32_t size() const;

lib/DebugInfo/CodeView/TypeDatabase.cpp

@@ -106,6 +106,10 @@ StringRef TypeDatabase::getTypeName(TypeIndex Index) const {
   return "<unknown UDT>";
 }
 
+const CVType &TypeDatabase::getTypeRecord(TypeIndex Index) const {
+  return TypeRecords[Index.getIndex() - TypeIndex::FirstNonSimpleIndex];
+}
+
 bool TypeDatabase::containsTypeIndex(TypeIndex Index) const {
   uint32_t I = Index.getIndex() - TypeIndex::FirstNonSimpleIndex;
   return I < CVUDTNames.size();

tools/llvm-pdbdump/Analyze.cpp

@@ -0,0 +1,164 @@
+//===- Analyze.cpp - PDB analysis functions ---------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Analyze.h"
+
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/DebugInfo/CodeView/CVTypeVisitor.h"
+#include "llvm/DebugInfo/CodeView/TypeDatabase.h"
+#include "llvm/DebugInfo/CodeView/TypeDatabaseVisitor.h"
+#include "llvm/DebugInfo/CodeView/TypeDeserializer.h"
+#include "llvm/DebugInfo/CodeView/TypeRecord.h"
+#include "llvm/DebugInfo/CodeView/TypeVisitorCallbackPipeline.h"
+#include "llvm/DebugInfo/CodeView/TypeVisitorCallbacks.h"
+#include "llvm/DebugInfo/PDB/Native/PDBFile.h"
+#include "llvm/DebugInfo/PDB/Native/RawError.h"
+#include "llvm/DebugInfo/PDB/Native/TpiStream.h"
+
+#include "llvm/Support/FormatVariadic.h"
+#include "llvm/Support/raw_ostream.h"
+
+#include <list>
+
+using namespace llvm;
+using namespace llvm::codeview;
+using namespace llvm::pdb;
+
+static StringRef getLeafTypeName(TypeLeafKind LT) {
+  switch (LT) {
+#define TYPE_RECORD(ename, value, name)                                        \
+  case ename:                                                                  \
+    return #name;
+#include "llvm/DebugInfo/CodeView/TypeRecords.def"
+  default:
+    break;
+  }
+  return "UnknownLeaf";
+}
+
+namespace {
+struct HashLookupVisitor : public TypeVisitorCallbacks {
+  struct Entry {
+    TypeIndex TI;
+    CVType Record;
+  };
+
+  explicit HashLookupVisitor(TpiStream &Tpi) : Tpi(Tpi) {}
+
+  Error visitTypeBegin(CVType &Record) override {
+    uint32_t H = Tpi.getHashValues()[I];
+    Record.Hash = H;
+    TypeIndex TI(I + TypeIndex::FirstNonSimpleIndex);
+    Lookup[H].push_back(Entry{TI, Record});
+    ++I;
+    return Error::success();
+  }
+
+  uint32_t I = 0;
+  DenseMap<uint32_t, std::list<Entry>> Lookup;
+  TpiStream &Tpi;
+};
+}
+
+AnalysisStyle::AnalysisStyle(PDBFile &File) : File(File) {}
+
+Error AnalysisStyle::dump() {
+  auto Tpi = File.getPDBTpiStream();
+  if (!Tpi)
+    return Tpi.takeError();
+
+  TypeDatabase TypeDB;
+  TypeDatabaseVisitor DBV(TypeDB);
+  TypeDeserializer Deserializer;
+  TypeVisitorCallbackPipeline Pipeline;
+  HashLookupVisitor Hasher(*Tpi);
+  // Deserialize the types
+  Pipeline.addCallbackToPipeline(Deserializer);
+  // Add them to the database
+  Pipeline.addCallbackToPipeline(DBV);
+  // Store their hash values
+  Pipeline.addCallbackToPipeline(Hasher);
+
+  CVTypeVisitor Visitor(Pipeline);
+
+  bool Error = false;
+  for (auto Item : Tpi->types(&Error)) {
+    if (auto EC = Visitor.visitTypeRecord(Item))
+      return EC;
+  }
+  if (Error)
+    return make_error<RawError>(raw_error_code::corrupt_file,
+                                "TPI stream contained corrupt record");
+
+  auto &Adjusters = Tpi->getHashAdjusters();
+  DenseSet<uint32_t> AdjusterSet;
+  for (const auto &Adj : Adjusters) {
+    assert(AdjusterSet.find(Adj.second) == AdjusterSet.end());
+    AdjusterSet.insert(Adj.second);
+  }
+
+  uint32_t Count = 0;
+  outs() << "Searching for hash collisions\n";
+  for (const auto &H : Hasher.Lookup) {
+    if (H.second.size() <= 1)
+      continue;
+    ++Count;
+    outs() << formatv("Hash: {0}, Count: {1} records\n", H.first,
+                      H.second.size());
+    for (const auto &R : H.second) {
+      auto Iter = AdjusterSet.find(R.TI.getIndex());
+      StringRef Prefix;
+      if (Iter != AdjusterSet.end()) {
+        Prefix = "[HEAD]";
+        AdjusterSet.erase(Iter);
+      }
+      StringRef LeafName = getLeafTypeName(R.Record.Type);
+      uint32_t TI = R.TI.getIndex();
+      StringRef TypeName = TypeDB.getTypeName(R.TI);
+      outs() << formatv("{0,-6} {1} ({2:x}) {3}\n", Prefix, LeafName, TI,
+                        TypeName);
+    }
+  }
+
+  outs() << "\n";
+  outs() << "Dumping hash adjustment chains\n";
+  for (const auto &A : Tpi->getHashAdjusters()) {
+    TypeIndex TI(A.second);
+    StringRef TypeName = TypeDB.getTypeName(TI);
+    const CVType &HeadRecord = TypeDB.getTypeRecord(TI);
+    assert(HeadRecord.Hash.hasValue());
+
+    auto CollisionsIter = Hasher.Lookup.find(*HeadRecord.Hash);
+    if (CollisionsIter == Hasher.Lookup.end())
+      continue;
+
+    const auto &Collisions = CollisionsIter->second;
+    outs() << TypeName << "\n";
+    outs() << formatv("    [HEAD] {0:x} {1} {2}\n", A.second,
+                      getLeafTypeName(HeadRecord.Type), TypeName);
+    for (const auto &Chain : Collisions) {
+      if (Chain.TI == TI)
+        continue;
+      const CVType &TailRecord = TypeDB.getTypeRecord(Chain.TI);
+      outs() << formatv("           {0:x} {1} {2}\n", Chain.TI.getIndex(),
+                        getLeafTypeName(TailRecord.Type),
+                        TypeDB.getTypeName(Chain.TI));
+    }
+  }
+  outs() << formatv("There are {0} orphaned hash adjusters\n",
+                    AdjusterSet.size());
+  for (const auto &Adj : AdjusterSet) {
+    outs() << formatv("    {0}\n", Adj);
+  }
+
+  uint32_t DistinctHashValues = Hasher.Lookup.size();
+  outs() << formatv("{0}/{1} hash collisions", Count, DistinctHashValues);
+  return Error::success();
+}

tools/llvm-pdbdump/Analyze.h

@@ -0,0 +1,30 @@
+//===- Analyze.h - PDB analysis functions -----------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TOOLS_LLVMPDBDUMP_ANALYSIS_H
+#define LLVM_TOOLS_LLVMPDBDUMP_ANALYSIS_H
+
+#include "OutputStyle.h"
+
+namespace llvm {
+namespace pdb {
+class PDBFile;
+class AnalysisStyle : public OutputStyle {
+public:
+  explicit AnalysisStyle(PDBFile &File);
+
+  Error dump() override;
+
+private:
+  PDBFile &File;
+};
+}
+}
+
+#endif

tools/llvm-pdbdump/CMakeLists.txt

@@ -7,6 +7,7 @@ set(LLVM_LINK_COMPONENTS
   )
 
 add_llvm_tool(llvm-pdbdump
+  Analyze.cpp
   CompactTypeDumpVisitor.cpp
   llvm-pdbdump.cpp
   YamlSymbolDumper.cpp

tools/llvm-pdbdump/llvm-pdbdump.cpp

@@ -14,6 +14,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm-pdbdump.h"
+
+#include "Analyze.h"
 #include "LLVMOutputStyle.h"
 #include "LinePrinter.h"
 #include "OutputStyle.h"
@@ -86,6 +88,10 @@ cl::SubCommand
     PdbToYamlSubcommand("pdb2yaml",
                         "Generate a detailed YAML description of a PDB File");
 
+cl::SubCommand
+    AnalyzeSubcommand("analyze",
+                      "Analyze various aspects of a PDB's structure");
+
 cl::OptionCategory TypeCategory("Symbol Type Options");
 cl::OptionCategory FilterCategory("Filtering Options");
 cl::OptionCategory OtherOptions("Other Options");
@@ -318,6 +324,14 @@ cl::list<std::string> InputFilename(cl::Positional,
                                     cl::desc("<input PDB file>"), cl::Required,
                                     cl::sub(PdbToYamlSubcommand));
 }
+
+namespace analyze {
+cl::opt<bool> StringTable("hash-collisions", cl::desc("Find hash collisions"),
+                          cl::sub(AnalyzeSubcommand), cl::init(false));
+cl::list<std::string> InputFilename(cl::Positional,
+                                    cl::desc("<input PDB file>"), cl::Required,
+                                    cl::sub(AnalyzeSubcommand));
+}
 }
 
 static ExitOnError ExitOnErr;
@@ -423,6 +437,17 @@ static void dumpRaw(StringRef Path) {
   ExitOnErr(O->dump());
 }
 
+static void dumpAnalysis(StringRef Path) {
+  std::unique_ptr<IPDBSession> Session;
+  ExitOnErr(loadDataForPDB(PDB_ReaderType::Native, Path, Session));
+
+  NativeSession *NS = static_cast<NativeSession *>(Session.get());
+  PDBFile &File = NS->getPDBFile();
+  auto O = llvm::make_unique<AnalysisStyle>(File);
+
+  ExitOnErr(O->dump());
+}
+
 static void dumpPretty(StringRef Path) {
   std::unique_ptr<IPDBSession> Session;
 
@@ -608,6 +633,8 @@ int main(int argc_, const char *argv_[]) {
     pdb2Yaml(opts::pdb2yaml::InputFilename.front());
   } else if (opts::YamlToPdbSubcommand) {
     yamlToPdb(opts::yaml2pdb::InputFilename.front());
+  } else if (opts::AnalyzeSubcommand) {
+    dumpAnalysis(opts::analyze::InputFilename.front());
   } else if (opts::PrettySubcommand) {
     if (opts::pretty::Lines)
       opts::pretty::Compilands = true;