Support use of third party time function for about:profiler

This uses TCMalloc to provide a time function, allowing
us to see how much memory was allocated on a single thread
during the running of a task.  The alternate time function
is put in place only when a specific environment variable
is detected during TCMalloc startup.

This change currently is activated only in Windows/Linux,
as it is based on changes TCMalloc (not used on Mac).

We also create an infrastructure for using any alternate
timer, to replace the "wall clock time," on a per-thread
basis, in the about:profiler infrastructure.  That interface
may be used on other platforms, including scenarios where
we have a per-thread-CPU-time function to replace the
wall-clock timer.  In all cases, when this alternate timer
is activated, we lose the ability to calculated queueing
time.  Queueing time is based on a time snapshot taken on
a second thread, and hence is not comparable to the
alternate timer (when the alternate timer is engaged).


r=rtenneti
BUG=103321
Review URL: https://chromiumcodereview.appspot.com/9212025

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@122180 0039d316-1c4b-4281-b951-d872f2087c98

12 files changed, 183 insertions, 4 deletions

diff --git a/base/allocator/allocator.gyp b/base/allocator/allocator.gyp
index b72f6f7..5aef0a6 100644
--- a/base/allocator/allocator.gyp
+++ b/base/allocator/allocator.gyp
@@ -416,6 +416,8 @@
       ],
       'sources': [
         'allocator_unittests.cc',
+        '../profiler/alternate_timer.cc',
+        '../profiler/alternate_timer.h',
       ],
     },
   ],
diff --git a/base/allocator/allocator_shim.cc b/base/allocator/allocator_shim.cc
index 097fff2..d61a9a4 100644
--- a/base/allocator/allocator_shim.cc
+++ b/base/allocator/allocator_shim.cc
@@ -5,6 +5,7 @@
 #include "base/allocator/allocator_shim.h"
 
 #include <config.h>
+#include "base/profiler/alternate_timer.h"
 #include "base/sysinfo.h"
 
 // When defined, different heap allocators can be used via an environment
@@ -48,8 +49,8 @@ static Allocator allocator = TCMALLOC;
 // selection of the allocator.  The primary may be used to control overall
 // allocator selection, and the secondary can be used to specify an allocator
 // to use in sub-processes.
-static const char* primary_name = "CHROME_ALLOCATOR";
-static const char* secondary_name = "CHROME_ALLOCATOR_2";
+static const char primary_name[] = "CHROME_ALLOCATOR";
+static const char secondary_name[] = "CHROME_ALLOCATOR_2";
 
 // We include tcmalloc and the win_allocator to get as much inlining as
 // possible.
@@ -261,6 +262,17 @@ extern "C" int _heap_init() {
   // lifetime.  Trying to teardown at _heap_term() is so late that
   // you can't do anything useful anyway.
   new TCMallocGuard();
+
+  // Provide optional hook for monitoring allocation quantities on a per-thread
+  // basis.  Only set the hook if the environment indicates this needs to be
+  // enabled.
+  const char* profiling =
+      GetenvBeforeMain(tracked_objects::kAlternateProfilerTime);
+  if (profiling && *profiling == '1') {
+    tracked_objects::SetAlternateTimeSource(
+        tcmalloc::ThreadCache::GetBytesAllocatedOnCurrentThread);
+  }
+
   return 1;
 }
 
@@ -302,5 +314,5 @@ void SetupSubprocessAllocator() {
 #endif  // ENABLE_DYNAMIC_ALLOCATOR_SWITCHING
 }
 
-}  // namespace base.
 }  // namespace allocator.
+}  // namespace base.
diff --git a/base/allocator/allocator_shim.h b/base/allocator/allocator_shim.h
index 342710f..b16f6ce 100644
--- a/base/allocator/allocator_shim.h
+++ b/base/allocator/allocator_shim.h
@@ -14,7 +14,7 @@ namespace allocator {
 // then a default value (typically set to TCMALLOC).
 void SetupSubprocessAllocator();
 
-}  // namespace base.
 }  // namespace allocator.
+}  // namespace base.
 
 #endif   // BASE_ALLOCATOR_ALLOCATOR_SHIM_H_
diff --git a/base/base.gypi b/base/base.gypi
index a5e819b..9d9d3f5 100644
--- a/base/base.gypi
+++ b/base/base.gypi
@@ -238,6 +238,8 @@
           'process_win.cc',
           'profiler/scoped_profile.cc',
           'profiler/scoped_profile.h',
+          'profiler/alternate_timer.cc',
+          'profiler/alternate_timer.h',
           'profiler/tracked_time.cc',
           'profiler/tracked_time.h',
           'property_bag.cc',
diff --git a/base/profiler/alternate_timer.cc b/base/profiler/alternate_timer.cc
new file mode 100644
index 0000000..05a983c
--- /dev/null
+++ b/base/profiler/alternate_timer.cc
@@ -0,0 +1,25 @@
+// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/profiler/alternate_timer.h"
+
+#include "base/logging.h"
+
+namespace tracked_objects {
+
+static NowFunction* g_time_function = NULL;
+
+const char kAlternateProfilerTime[] = "CHROME_PROFILER_TIME";
+
+// Set an alternate timer function to replace the OS time function when
+// profiling.
+void SetAlternateTimeSource(NowFunction* now_function) {
+  DCHECK_EQ(g_time_function, reinterpret_cast<NowFunction*>(NULL));
+  g_time_function = now_function;
+}
+
+extern NowFunction* GetAlternateTimeSource() {
+  return g_time_function;
+}
+}  // tracked_objects
diff --git a/base/profiler/alternate_timer.h b/base/profiler/alternate_timer.h
new file mode 100644
index 0000000..883b24f
--- /dev/null
+++ b/base/profiler/alternate_timer.h
@@ -0,0 +1,36 @@
+// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+// This is a glue file, which allows third party code to call into our profiler
+// without having to include most any functions from base.
+
+
+#ifndef BASE_PROFILER_ALTERNATE_TIMER_H_
+#define BASE_PROFILER_ALTERNATE_TIMER_H_
+
+namespace tracked_objects {
+
+// Provide type for an alternate timer function.
+typedef unsigned int NowFunction();
+
+// Set an alternate timer function to replace the OS time function when
+// profiling.  Typically this is called by an allocator that is providing a
+// function that indicates how much memory has been allocated on any given
+// thread.
+extern void SetAlternateTimeSource(NowFunction* now_function);
+
+// Gets the pointer to a function that was set via SetAlternateTimeSource().
+// Returns NULL if no set was done prior to calling GetAlternateTimeSource.
+extern NowFunction* GetAlternateTimeSource();
+
+// Environment variable name that is used to activate alternate timer profiling
+// (such as using TCMalloc allocations to provide a pseudo-timer) for tasks
+// instead of wall clock profiling.
+extern const char kAlternateProfilerTime[];
+
+
+
+}  // tracked_objects
+
+#endif  // BASE_PROFILER_ALTERNATE_TIMER_H_
diff --git a/base/profiler/tracked_time.h b/base/profiler/tracked_time.h
index 6f4bc7e..5493e20 100644
--- a/base/profiler/tracked_time.h
+++ b/base/profiler/tracked_time.h
@@ -62,6 +62,8 @@ class BASE_EXPORT TrackedTime {  // Similar to base::TimeTicks.
   TrackedTime operator+(const Duration& other) const;
   bool is_null() const;
 
+  static TrackedTime FromMilliseconds(int32 ms) { return TrackedTime(ms); }
+
  private:
   friend class Duration;
   explicit TrackedTime(int32 ms);
diff --git a/base/tracked_objects.cc b/base/tracked_objects.cc
index 87af8dc..c837b9a 100644
--- a/base/tracked_objects.cc
+++ b/base/tracked_objects.cc
@@ -8,6 +8,7 @@
 
 #include "base/format_macros.h"
 #include "base/message_loop.h"
+#include "base/profiler/alternate_timer.h"
 #include "base/stringprintf.h"
 #include "base/third_party/valgrind/memcheck.h"
 #include "base/threading/thread_restrictions.h"
@@ -35,6 +36,14 @@ const bool kTrackParentChildLinks = false;
 const ThreadData::Status kInitialStartupState =
     ThreadData::PROFILING_CHILDREN_ACTIVE;
 
+// Control whether an alternate time source (Now() function) is supported by
+// the ThreadData class.  This compile time flag should be set to true if we
+// want other modules (such as a memory allocator, or a thread-specific CPU time
+// clock) to be able to provide a thread-specific Now() function.  Without this
+// compile-time flag, the code will only support the wall-clock time.  This flag
+// can be flipped to efficiently disable this path (if there is a performance
+// problem with its presence).
+static const bool kAllowAlternateTimeSourceHandling = true;
 }  // namespace
 
 //------------------------------------------------------------------------------
@@ -176,6 +185,9 @@ void Births::Clear() { birth_count_ = 0; }
 // optimize layout so that we benefit from locality of reference during accesses
 // to them.
 
+// static
+NowFunction* ThreadData::now_function_ = NULL;
+
 // A TLS slot which points to the ThreadData instance for the current thread. We
 // do a fake initialization here (zeroing out data), and then the real in-place
 // construction happens when we call tls_index_.Initialize().
@@ -370,6 +382,12 @@ void ThreadData::TallyADeath(const Births& birth,
   // An address is going to have some randomness to it as well ;-).
   random_number_ ^= static_cast<int32>(&birth - reinterpret_cast<Births*>(0));
 
+  // We don't have queue durations without OS timer. OS timer is automatically
+  // used for task-post-timing, so the use of an alternate timer implies all
+  // queue times are invalid.
+  if (kAllowAlternateTimeSourceHandling && now_function_)
+    queue_duration = 0;
+
   DeathMap::iterator it = death_map_.find(&birth);
   DeathData* death_data;
   if (it != death_map_.end()) {
@@ -579,6 +597,24 @@ void ThreadData::Reset() {
     it->second->Clear();
 }
 
+static void OptionallyInitializeAlternateTimer() {
+  char* alternate_selector = getenv(kAlternateProfilerTime);
+  if (!alternate_selector)
+    return;
+  switch (*alternate_selector) {
+    case '0':  // This is the default value, and uses the wall clock time.
+      break;
+    case '1':  {
+      // Use the TCMalloc allocations-on-thread as a pseudo-time.
+      ThreadData::SetAlternateTimeSource(GetAlternateTimeSource());
+      break;
+      }
+    default:
+      NOTREACHED();
+      break;
+  }
+}
+
 bool ThreadData::Initialize() {
   if (!kTrackAllTaskObjects)
     return false;  // Not compiled in.
@@ -594,6 +630,13 @@ bool ThreadData::Initialize() {
   if (status_ >= DEACTIVATED)
     return true;  // Someone raced in here and beat us.
 
+  // Put an alternate timer in place if the environment calls for it, such as
+  // for tracking TCMalloc allocations.  This insertion is idempotent, so we
+  // don't mind if there is a race, and we'd prefer not to be in a lock while
+  // doing this work.
+  if (kAllowAlternateTimeSourceHandling)
+    OptionallyInitializeAlternateTimer();
+
   // Perform the "real" TLS initialization now, and leave it intact through
   // process termination.
   if (!tls_index_.initialized()) {  // Testing may have initialized this.
@@ -666,7 +709,16 @@ TrackedTime ThreadData::NowForEndOfRun() {
 }
 
 // static
+void ThreadData::SetAlternateTimeSource(NowFunction* now_function) {
+  DCHECK(now_function);
+  if (kAllowAlternateTimeSourceHandling)
+    now_function_ = now_function;
+}
+
+// static
 TrackedTime ThreadData::Now() {
+  if (kAllowAlternateTimeSourceHandling && now_function_)
+    return TrackedTime::FromMilliseconds((*now_function_)());
   if (kTrackAllTaskObjects && TrackingStatus())
     return TrackedTime::Now();
   return TrackedTime();  // Super fast when disabled, or not compiled.
diff --git a/base/tracked_objects.h b/base/tracked_objects.h
index 96814f6..cd50594 100644
--- a/base/tracked_objects.h
+++ b/base/tracked_objects.h
@@ -17,6 +17,7 @@
 #include "base/gtest_prod_util.h"
 #include "base/lazy_instance.h"
 #include "base/location.h"
+#include "base/profiler/alternate_timer.h"
 #include "base/profiler/tracked_time.h"
 #include "base/time.h"
 #include "base/synchronization/lock.h"
@@ -467,6 +468,12 @@ class BASE_EXPORT ThreadData {
   // the code).
   static TrackedTime Now();
 
+  // Use the function |now| to provide current times, instead of calling the
+  // TrackedTime::Now() function.  Since this alternate function is being used,
+  // the other time arguments (used for calculating queueing delay) will be
+  // ignored.
+  static void SetAlternateTimeSource(NowFunction* now);
+
   // This function can be called at process termination to validate that thread
   // cleanup routines have been called for at least some number of named
   // threads.
@@ -543,6 +550,10 @@ class BASE_EXPORT ThreadData {
   // ThreadData instances.
   static void ShutdownSingleThreadedCleanup(bool leak);
 
+  // When non-null, this specifies an external function that supplies monotone
+  // increasing time functcion.
+  static NowFunction* now_function_;
+
   // We use thread local store to identify which ThreadData to interact with.
   static base::ThreadLocalStorage::StaticSlot tls_index_;
 
diff --git a/third_party/tcmalloc/chromium/src/tcmalloc.cc b/third_party/tcmalloc/chromium/src/tcmalloc.cc
index 51fe5b3..9381aaf 100644
--- a/third_party/tcmalloc/chromium/src/tcmalloc.cc
+++ b/third_party/tcmalloc/chromium/src/tcmalloc.cc
@@ -1130,6 +1130,8 @@ inline void* do_malloc_pages(ThreadCache* heap, size_t size) {
   Length num_pages = tcmalloc::pages(size);
   size = num_pages << kPageShift;
 
+  heap->AddToByteAllocatedTotal(size);  // Chromium profiling.
+
   if ((FLAGS_tcmalloc_sample_parameter > 0) && heap->SampleAllocation(size)) {
     result = DoSampledAllocation(size);
 
@@ -1159,6 +1161,12 @@ inline void* do_malloc(size_t size) {
     size_t cl = Static::sizemap()->SizeClass(size);
     size = Static::sizemap()->class_to_size(cl);
 
+    // TODO(jar): If this has any detectable performance impact, it can be
+    // optimized by only tallying sizes if the profiler was activated to recall
+    // these tallies.  I don't think this is performance critical, but we really
+    // should measure it.
+    heap->AddToByteAllocatedTotal(size);  // Chromium profiling.
+
     if ((FLAGS_tcmalloc_sample_parameter > 0) && heap->SampleAllocation(size)) {
       ret = DoSampledAllocation(size);
       MarkAllocatedRegion(ret);
diff --git a/third_party/tcmalloc/chromium/src/thread_cache.cc b/third_party/tcmalloc/chromium/src/thread_cache.cc
index a951f77..1c189f3 100644
--- a/third_party/tcmalloc/chromium/src/thread_cache.cc
+++ b/third_party/tcmalloc/chromium/src/thread_cache.cc
@@ -103,6 +103,7 @@ bool kernel_supports_tls = false;      // be conservative
 
 void ThreadCache::Init(pthread_t tid) {
   size_ = 0;
+  total_bytes_allocated_ = 0;
 
   max_size_ = 0;
   IncreaseCacheLimitLocked();
@@ -303,6 +304,11 @@ int ThreadCache::GetSamplePeriod() {
   return sampler_.GetSamplePeriod();
 }
 
+// static
+unsigned int ThreadCache::GetBytesAllocatedOnCurrentThread() {
+  return ThreadCache::GetThreadHeap()->GetTotalBytesAllocated();
+}
+
 void ThreadCache::InitModule() {
   SpinLockHolder h(Static::pageheap_lock());
   if (!phinited) {
diff --git a/third_party/tcmalloc/chromium/src/thread_cache.h b/third_party/tcmalloc/chromium/src/thread_cache.h
index 9220aab..d631f45 100644
--- a/third_party/tcmalloc/chromium/src/thread_cache.h
+++ b/third_party/tcmalloc/chromium/src/thread_cache.h
@@ -96,6 +96,17 @@ class ThreadCache {
   // should be sampled
   bool SampleAllocation(size_t k);
 
+  // Record additional bytes allocated.
+  void AddToByteAllocatedTotal(size_t k) { total_bytes_allocated_ += k; }
+
+  // Return the total number of bytes allocated from this heap.  The value will
+  // wrap when there is an overflow, and so only the differences between two
+  // values should be relied on (and even then, modulo 2^32).
+  uint32 GetTotalBytesAllocated() const;
+
+  // On the current thread, return GetTotalBytesAllocated().
+  static uint32 GetBytesAllocatedOnCurrentThread();
+
   static void         InitModule();
   static void         InitTSD();
   static ThreadCache* GetThreadHeap();
@@ -291,6 +302,14 @@ class ThreadCache {
   size_t        size_;                  // Combined size of data
   size_t        max_size_;              // size_ > max_size_ --> Scavenge()
 
+  // The following is the tally of bytes allocated on a thread as a response to
+  // any flavor of malloc() call.  The aggegated amount includes all padding to
+  // the smallest class that can hold the request, or to the nearest whole page
+  // when a large allocation is made without using a class.  This sum is
+  // currently used for Chromium profiling, where tallies are kept of the amount
+  // of memory allocated during the running of each task on each thread.
+  uint32        total_bytes_allocated_;  // Total, modulo 2^32.
+
   // We sample allocations, biased by the size of the allocation
   Sampler       sampler_;               // A sampler
 
@@ -327,6 +346,10 @@ inline bool ThreadCache::SampleAllocation(size_t k) {
   return sampler_.SampleAllocation(k);
 }
 
+inline uint32 ThreadCache::GetTotalBytesAllocated() const {
+  return total_bytes_allocated_;
+}
+
 inline void* ThreadCache::Allocate(size_t size, size_t cl) {
   ASSERT(size <= kMaxSize);
   ASSERT(size == Static::sizemap()->ByteSizeForClass(cl));