chipper 0.4.2

data/ext/util/arena.h

@@ -0,0 +1,103 @@
+// Copyright 2000 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Sometimes it is necessary to allocate a large number of small
+// objects.  Doing this the usual way (malloc, new) is slow,
+// especially for multithreaded programs.  An UnsafeArena provides a
+// mark/release method of memory management: it asks for a large chunk
+// from the operating system and doles it out bit by bit as required.
+// Then you free all the memory at once by calling UnsafeArena::Reset().
+// The "Unsafe" refers to the fact that UnsafeArena is not safe to
+// call from multiple threads.
+//
+// The global operator new that can be used as follows:
+//
+//   #include "lib/arena-inl.h"
+//
+//   UnsafeArena arena(1000);
+//   Foo* foo = new (AllocateInArena, &arena) Foo;
+//
+
+#ifndef RE2_UTIL_ARENA_H_
+#define RE2_UTIL_ARENA_H_
+
+namespace re2 {
+
+// This class is thread-compatible.
+class UnsafeArena {
+ public:
+  UnsafeArena(const size_t block_size);
+  virtual ~UnsafeArena();
+
+  void Reset();
+
+  // This should be the worst-case alignment for any type.  This is
+  // good for IA-32, SPARC version 7 (the last one I know), and
+  // supposedly Alpha.  i386 would be more time-efficient with a
+  // default alignment of 8, but ::operator new() uses alignment of 4,
+  // and an assertion will fail below after the call to MakeNewBlock()
+  // if you try to use a larger alignment.
+#ifdef __i386__
+  static const int kDefaultAlignment = 4;
+#else
+  static const int kDefaultAlignment = 8;
+#endif
+
+ private:
+  void* GetMemoryFallback(const size_t size, const int align);
+
+ public:
+  void* GetMemory(const size_t size, const int align) {
+    if ( size > 0 && size < remaining_ && align == 1 ) {       // common case
+      last_alloc_ = freestart_;
+      freestart_ += size;
+      remaining_ -= size;
+      return reinterpret_cast<void*>(last_alloc_);
+    }
+    return GetMemoryFallback(size, align);
+  }
+
+ private:
+  struct AllocatedBlock {
+    char *mem;
+    size_t size;
+  };
+
+  // The returned AllocatedBlock* is valid until the next call to AllocNewBlock
+  // or Reset (i.e. anything that might affect overflow_blocks_).
+  AllocatedBlock *AllocNewBlock(const size_t block_size);
+
+  const AllocatedBlock *IndexToBlock(int index) const;
+
+  const size_t block_size_;
+  char* freestart_;         // beginning of the free space in most recent block
+  char* freestart_when_empty_;  // beginning of the free space when we're empty
+  char* last_alloc_;         // used to make sure ReturnBytes() is safe
+  size_t remaining_;
+  // STL vector isn't as efficient as it could be, so we use an array at first
+  int blocks_alloced_;       // how many of the first_blocks_ have been alloced
+  AllocatedBlock first_blocks_[16];   // the length of this array is arbitrary
+  // if the first_blocks_ aren't enough, expand into overflow_blocks_.
+  vector<AllocatedBlock>* overflow_blocks_;
+
+  void FreeBlocks();         // Frees all except first block
+
+  DISALLOW_EVIL_CONSTRUCTORS(UnsafeArena);
+};
+
+// Operators for allocation on the arena
+// Syntax: new (AllocateInArena, arena) MyClass;
+// STL containers, etc.
+enum AllocateInArenaType { AllocateInArena };
+
+}  // namespace re2
+
+inline void* operator new(size_t size,
+                          re2::AllocateInArenaType /* unused */,
+                          re2::UnsafeArena *arena) {
+  return reinterpret_cast<char*>(arena->GetMemory(size, 1));
+}
+
+#endif  // RE2_UTIL_ARENA_H_
+

data/ext/util/atomicops.h

@@ -0,0 +1,79 @@
+// Copyright 2006-2008 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#ifndef RE2_UTIL_ATOMICOPS_H__
+#define RE2_UTIL_ATOMICOPS_H__
+
+#if defined(__i386__)
+
+static inline void WriteMemoryBarrier() {
+  int x;
+  __asm__ __volatile__("xchgl (%0),%0"  // The lock prefix is implicit for xchg.
+                       :: "r" (&x));
+}
+
+#elif defined(__x86_64__)
+
+// 64-bit implementations of memory barrier can be simpler, because
+// "sfence" is guaranteed to exist.
+static inline void WriteMemoryBarrier() {
+  __asm__ __volatile__("sfence" : : : "memory");
+}
+
+#elif defined(__ppc__)
+
+static inline void WriteMemoryBarrier() {
+  __asm__ __volatile__("eieio" : : : "memory");
+}
+
+#elif defined(__alpha__)
+
+static inline void WriteMemoryBarrier() {
+  __asm__ __volatile__("wmb" : : : "memory");
+}
+
+#else
+
+#include "util/mutex.h"
+
+static inline void WriteMemoryBarrier() {
+  // Slight overkill, but good enough:
+  // any mutex implementation must have
+  // a read barrier after the lock operation and
+  // a write barrier before the unlock operation.
+  //
+  // It may be worthwhile to write architecture-specific
+  // barriers for the common platforms, as above, but
+  // this is a correct fallback.
+  re2::Mutex mu;
+  re2::MutexLock l(&mu);
+}
+
+/*
+#error Need WriteMemoryBarrier for architecture.
+
+// Windows
+inline void WriteMemoryBarrier() {
+  LONG x;
+  ::InterlockedExchange(&x, 0);
+}
+*/
+
+#endif
+
+// Alpha has very weak memory ordering. If relying on WriteBarriers, must one
+// use read barriers for the readers too.
+#if defined(__alpha__)
+
+static inline void MaybeReadMemoryBarrier() {
+  __asm__ __volatile__("rmb" : : : "memory");
+}
+
+#else
+
+static inline void MaybeReadMemoryBarrier() {}
+
+#endif // __alpha__
+
+#endif  // RE2_UTIL_ATOMICOPS_H__

data/ext/util/benchmark.h

@@ -0,0 +1,41 @@
+// Copyright 2009 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#ifndef RE2_UTIL_BENCHMARK_H__
+#define RE2_UTIL_BENCHMARK_H__
+
+namespace testing {
+struct Benchmark {
+  const char* name;
+  void (*fn)(int);
+  void (*fnr)(int, int);
+  int lo;
+  int hi;
+  int threadlo;
+  int threadhi;
+  
+  void Register();
+  Benchmark(const char* name, void (*f)(int)) { Clear(name); fn = f; Register(); }
+  Benchmark(const char* name, void (*f)(int, int), int l, int h) { Clear(name); fnr = f; lo = l; hi = h; Register(); }
+  void Clear(const char* n) { name = n; fn = 0; fnr = 0; lo = 0; hi = 0; threadlo = 0; threadhi = 0; }
+  Benchmark* ThreadRange(int lo, int hi) { threadlo = lo; threadhi = hi; return this; }
+};
+}  // namespace testing
+
+void SetBenchmarkBytesProcessed(long long);
+void StopBenchmarkTiming();
+void StartBenchmarkTiming();
+void BenchmarkMemoryUsage();
+void SetBenchmarkItemsProcessed(int);
+
+int NumCPUs();
+
+#define BENCHMARK(f) \
+	::testing::Benchmark* _benchmark_##f = (new ::testing::Benchmark(#f, f))
+
+#define BENCHMARK_RANGE(f, lo, hi) \
+	::testing::Benchmark* _benchmark_##f = \
+	(new ::testing::Benchmark(#f, f, lo, hi))
+
+#endif  // RE2_UTIL_BENCHMARK_H__

data/ext/util/flags.h

@@ -0,0 +1,27 @@
+// Copyright 2009 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Simplified version of Google's command line flags.
+// Does not support parsing the command line.
+// If you want to do that, see
+// http://code.google.com/p/google-gflags
+
+#ifndef RE2_UTIL_FLAGS_H__
+#define RE2_UTIL_FLAGS_H__
+
+#define DEFINE_flag(type, name, deflt, desc) \
+	namespace re2 { type FLAGS_##name = deflt; }
+
+#define DECLARE_flag(type, name) \
+	namespace re2 { extern type FLAGS_##name; }
+
+#define DEFINE_bool(name, deflt, desc) DEFINE_flag(bool, name, deflt, desc)
+#define DEFINE_int32(name, deflt, desc) DEFINE_flag(int32, name, deflt, desc)
+#define DEFINE_string(name, deflt, desc) DEFINE_flag(string, name, deflt, desc)
+
+#define DECLARE_bool(name) DECLARE_flag(bool, name)
+#define DECLARE_int32(name) DECLARE_flag(int32, name)
+#define DECLARE_string(name) DECLARE_flag(string, name)
+
+#endif  // RE2_UTIL_FLAGS_H__

data/ext/util/logging.h

@@ -0,0 +1,78 @@
+// Copyright 2009 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Simplified version of Google's logging.
+
+#ifndef RE2_UTIL_LOGGING_H__
+#define RE2_UTIL_LOGGING_H__
+
+#include <unistd.h>  /* for write */
+#include <sstream>
+
+// Debug-only checking.
+#define DCHECK(condition) assert(condition)
+#define DCHECK_EQ(val1, val2) assert((val1) == (val2))
+#define DCHECK_NE(val1, val2) assert((val1) != (val2))
+#define DCHECK_LE(val1, val2) assert((val1) <= (val2))
+#define DCHECK_LT(val1, val2) assert((val1) < (val2))
+#define DCHECK_GE(val1, val2) assert((val1) >= (val2))
+#define DCHECK_GT(val1, val2) assert((val1) > (val2))
+
+// Always-on checking
+#define CHECK(x)	if(x){}else LogMessageFatal(__FILE__, __LINE__).stream() << "Check failed: " #x
+#define CHECK_LT(x, y)	CHECK((x) < (y))
+#define CHECK_GT(x, y)	CHECK((x) > (y))
+#define CHECK_LE(x, y)	CHECK((x) <= (y))
+#define CHECK_GE(x, y)	CHECK((x) >= (y))
+#define CHECK_EQ(x, y)	CHECK((x) == (y))
+#define CHECK_NE(x, y)	CHECK((x) != (y))
+
+#define LOG_INFO LogMessage(__FILE__, __LINE__)
+#define LOG_ERROR LOG_INFO
+#define LOG_WARNING LOG_INFO
+#define LOG_FATAL LogMessageFatal(__FILE__, __LINE__)
+#define LOG_QFATAL LOG_FATAL
+
+#define VLOG(x) if((x)>0){}else LOG_INFO.stream()
+
+#ifdef NDEBUG
+#define DEBUG_MODE 0
+#define LOG_DFATAL LOG_ERROR
+#else
+#define DEBUG_MODE 1
+#define LOG_DFATAL LOG_FATAL
+#endif
+
+#define LOG(severity) LOG_ ## severity.stream()
+
+class LogMessage {
+ public:
+  LogMessage(const char* file, int line) {
+    stream() << file << ":" << line << ": ";
+  }
+  ~LogMessage() {
+    stream() << "\n";
+    string s = str_.str();
+    if(write(2, s.data(), s.size()) < 0) {}  // shut up gcc
+  }
+  ostream& stream() { return str_; }
+ 
+ private:
+  std::ostringstream str_;
+  DISALLOW_EVIL_CONSTRUCTORS(LogMessage);
+};
+
+class LogMessageFatal : public LogMessage {
+ public:
+  LogMessageFatal(const char* file, int line)
+    : LogMessage(file, line) { }
+  ~LogMessageFatal() {
+    std::cerr << "\n";
+    abort();
+  }
+ private:
+  DISALLOW_EVIL_CONSTRUCTORS(LogMessageFatal);
+};
+
+#endif  // RE2_UTIL_LOGGING_H__

data/ext/util/mutex.h

@@ -0,0 +1,190 @@
+// Copyright 2007 The RE2 Authors.  All Rights Reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+/*
+ * A simple mutex wrapper, supporting locks and read-write locks.
+ * You should assume the locks are *not* re-entrant.
+ */
+
+#ifndef RE2_UTIL_MUTEX_H_
+#define RE2_UTIL_MUTEX_H_
+
+namespace re2 {
+
+#define HAVE_PTHREAD 1
+#define HAVE_RWLOCK 1
+
+#if defined(NO_THREADS)
+  typedef int MutexType;      // to keep a lock-count
+#elif defined(HAVE_PTHREAD) && defined(HAVE_RWLOCK)
+  // Needed for pthread_rwlock_*.  If it causes problems, you could take it
+  // out, but then you'd have to unset HAVE_RWLOCK (at least on linux -- it
+  // *does* cause problems for FreeBSD, or MacOSX, but isn't needed
+  // for locking there.)
+# ifdef __linux__
+#   undef _XOPEN_SOURCE
+#   define _XOPEN_SOURCE 500  // may be needed to get the rwlock calls
+# endif
+# include <pthread.h>
+  typedef pthread_rwlock_t MutexType;
+#elif defined(HAVE_PTHREAD)
+# include <pthread.h>
+  typedef pthread_mutex_t MutexType;
+#elif defined(WIN32)
+# define WIN32_LEAN_AND_MEAN  // We only need minimal includes
+# ifdef GMUTEX_TRYLOCK
+  // We need Windows NT or later for TryEnterCriticalSection().  If you
+  // don't need that functionality, you can remove these _WIN32_WINNT
+  // lines, and change TryLock() to assert(0) or something.
+#   ifndef _WIN32_WINNT
+#     define _WIN32_WINNT 0x0400
+#   endif
+# endif
+# include <windows.h>
+  typedef CRITICAL_SECTION MutexType;
+#else
+# error Need to implement mutex.h for your architecture, or #define NO_THREADS
+#endif
+
+class Mutex {
+ public:
+  // Create a Mutex that is not held by anybody.
+  inline Mutex();
+
+  // Destructor
+  inline ~Mutex();
+
+  inline void Lock();    // Block if needed until free then acquire exclusively
+  inline void Unlock();  // Release a lock acquired via Lock()
+  inline bool TryLock(); // If free, Lock() and return true, else return false
+  // Note that on systems that don't support read-write locks, these may
+  // be implemented as synonyms to Lock() and Unlock().  So you can use
+  // these for efficiency, but don't use them anyplace where being able
+  // to do shared reads is necessary to avoid deadlock.
+  inline void ReaderLock();   // Block until free or shared then acquire a share
+  inline void ReaderUnlock(); // Release a read share of this Mutex
+  inline void WriterLock() { Lock(); }     // Acquire an exclusive lock
+  inline void WriterUnlock() { Unlock(); } // Release a lock from WriterLock()
+  inline void AssertHeld() { }
+
+ private:
+  MutexType mutex_;
+
+  // Catch the error of writing Mutex when intending MutexLock.
+  Mutex(Mutex *ignored) {}
+  // Disallow "evil" constructors
+  Mutex(const Mutex&);
+  void operator=(const Mutex&);
+};
+
+// Now the implementation of Mutex for various systems
+#if defined(NO_THREADS)
+
+// When we don't have threads, we can be either reading or writing,
+// but not both.  We can have lots of readers at once (in no-threads
+// mode, that's most likely to happen in recursive function calls),
+// but only one writer.  We represent this by having mutex_ be -1 when
+// writing and a number > 0 when reading (and 0 when no lock is held).
+//
+// In debug mode, we assert these invariants, while in non-debug mode
+// we do nothing, for efficiency.  That's why everything is in an
+// assert.
+#include <assert.h>
+
+Mutex::Mutex() : mutex_(0) { }
+Mutex::~Mutex()            { assert(mutex_ == 0); }
+void Mutex::Lock()         { assert(--mutex_ == -1); }
+void Mutex::Unlock()       { assert(mutex_++ == -1); }
+bool Mutex::TryLock()      { if (mutex_) return false; Lock(); return true; }
+void Mutex::ReaderLock()   { assert(++mutex_ > 0); }
+void Mutex::ReaderUnlock() { assert(mutex_-- > 0); }
+
+#elif defined(HAVE_PTHREAD) && defined(HAVE_RWLOCK)
+
+#include <stdlib.h>      // for abort()
+#define SAFE_PTHREAD(fncall)  do { if ((fncall) != 0) abort(); } while (0)
+
+Mutex::Mutex()             { SAFE_PTHREAD(pthread_rwlock_init(&mutex_, NULL)); }
+Mutex::~Mutex()            { SAFE_PTHREAD(pthread_rwlock_destroy(&mutex_)); }
+void Mutex::Lock()         { SAFE_PTHREAD(pthread_rwlock_wrlock(&mutex_)); }
+void Mutex::Unlock()       { SAFE_PTHREAD(pthread_rwlock_unlock(&mutex_)); }
+bool Mutex::TryLock()      { return pthread_rwlock_trywrlock(&mutex_) == 0; }
+void Mutex::ReaderLock()   { SAFE_PTHREAD(pthread_rwlock_rdlock(&mutex_)); }
+void Mutex::ReaderUnlock() { SAFE_PTHREAD(pthread_rwlock_unlock(&mutex_)); }
+
+#undef SAFE_PTHREAD
+
+#elif defined(HAVE_PTHREAD)
+
+#include <stdlib.h>      // for abort()
+#define SAFE_PTHREAD(fncall)  do { if ((fncall) != 0) abort(); } while (0)
+
+Mutex::Mutex()             { SAFE_PTHREAD(pthread_mutex_init(&mutex_, NULL)); }
+Mutex::~Mutex()            { SAFE_PTHREAD(pthread_mutex_destroy(&mutex_)); }
+void Mutex::Lock()         { SAFE_PTHREAD(pthread_mutex_lock(&mutex_)); }
+void Mutex::Unlock()       { SAFE_PTHREAD(pthread_mutex_unlock(&mutex_)); }
+bool Mutex::TryLock()      { return pthread_mutex_trylock(&mutex_) == 0; }
+void Mutex::ReaderLock()   { Lock(); }      // we don't have read-write locks
+void Mutex::ReaderUnlock() { Unlock(); }
+#undef SAFE_PTHREAD
+
+#elif defined(WIN32)
+
+Mutex::Mutex()             { InitializeCriticalSection(&mutex_); }
+Mutex::~Mutex()            { DeleteCriticalSection(&mutex_); }
+void Mutex::Lock()         { EnterCriticalSection(&mutex_); }
+void Mutex::Unlock()       { LeaveCriticalSection(&mutex_); }
+bool Mutex::TryLock()      { return TryEnterCriticalSection(&mutex_) != 0; }
+void Mutex::ReaderLock()   { Lock(); }      // we don't have read-write locks
+void Mutex::ReaderUnlock() { Unlock(); }
+
+#endif
+
+
+// --------------------------------------------------------------------------
+// Some helper classes
+
+// MutexLock(mu) acquires mu when constructed and releases it when destroyed.
+class MutexLock {
+ public:
+  explicit MutexLock(Mutex *mu) : mu_(mu) { mu_->Lock(); }
+  ~MutexLock() { mu_->Unlock(); }
+ private:
+  Mutex * const mu_;
+  // Disallow "evil" constructors
+  MutexLock(const MutexLock&);
+  void operator=(const MutexLock&);
+};
+
+// ReaderMutexLock and WriterMutexLock do the same, for rwlocks
+class ReaderMutexLock {
+ public:
+  explicit ReaderMutexLock(Mutex *mu) : mu_(mu) { mu_->ReaderLock(); }
+  ~ReaderMutexLock() { mu_->ReaderUnlock(); }
+ private:
+  Mutex * const mu_;
+  // Disallow "evil" constructors
+  ReaderMutexLock(const ReaderMutexLock&);
+  void operator=(const ReaderMutexLock&);
+};
+
+class WriterMutexLock {
+ public:
+  explicit WriterMutexLock(Mutex *mu) : mu_(mu) { mu_->WriterLock(); }
+  ~WriterMutexLock() { mu_->WriterUnlock(); }
+ private:
+  Mutex * const mu_;
+  // Disallow "evil" constructors
+  WriterMutexLock(const WriterMutexLock&);
+  void operator=(const WriterMutexLock&);
+};
+
+// Catch bug where variable name is omitted, e.g. MutexLock (&mu);
+#define MutexLock(x) COMPILE_ASSERT(0, mutex_lock_decl_missing_var_name)
+#define ReaderMutexLock(x) COMPILE_ASSERT(0, rmutex_lock_decl_missing_var_name)
+#define WriterMutexLock(x) COMPILE_ASSERT(0, wmutex_lock_decl_missing_var_name)
+
+}  // namespace re2
+
+#endif  /* #define RE2_UTIL_MUTEX_H_ */