blob: 71afebe441527ad89a206f49955cc73bbed7de18 [file] [log] [blame]
/*
* Copyright (C) 2006-2016 Apple Inc. All rights reserved.
* Copyright (C) 2008 Google Inc. All rights reserved.
* Copyright (C) 2007-2009 Torch Mobile, Inc.
* Copyright (C) 2008 Cameron Zwarich <cwzwarich@uwaterloo.ca>
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Google Inc. nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#include "CurrentTime.h"
#include "Condition.h"
#include "Lock.h"
#if OS(DARWIN)
#include <mach/mach.h>
#include <mach/mach_time.h>
#include <mutex>
#include <sys/time.h>
#elif OS(WINDOWS)
// Windows is first since we want to use hires timers, despite USE(CF)
// being defined.
// If defined, WIN32_LEAN_AND_MEAN disables timeBeginPeriod/timeEndPeriod.
#undef WIN32_LEAN_AND_MEAN
#include <windows.h>
#include <math.h>
#include <stdint.h>
#include <time.h>
#else
#include <sys/time.h>
#endif
#if USE(GLIB)
#include <glib.h>
#endif
namespace WTF {
#if OS(WINDOWS)
// Number of 100 nanosecond between January 1, 1601 and January 1, 1970.
static const ULONGLONG epochBias = 116444736000000000ULL;
static const double hundredsOfNanosecondsPerMillisecond = 10000;
static double lowResUTCTime()
{
FILETIME fileTime;
GetSystemTimeAsFileTime(&fileTime);
// As per Windows documentation for FILETIME, copy the resulting FILETIME structure to a
// ULARGE_INTEGER structure using memcpy (using memcpy instead of direct assignment can
// prevent alignment faults on 64-bit Windows).
ULARGE_INTEGER dateTime;
memcpy(&dateTime, &fileTime, sizeof(dateTime));
// Windows file times are in 100s of nanoseconds.
return (dateTime.QuadPart - epochBias) / hundredsOfNanosecondsPerMillisecond;
}
#if USE(QUERY_PERFORMANCE_COUNTER)
static LARGE_INTEGER qpcFrequency;
static bool syncedTime;
static double highResUpTime()
{
// We use QPC, but only after sanity checking its result, due to bugs:
// http://support.microsoft.com/kb/274323
// http://support.microsoft.com/kb/895980
// http://msdn.microsoft.com/en-us/library/ms644904.aspx ("...you can get different results on different processors due to bugs in the basic input/output system (BIOS) or the hardware abstraction layer (HAL)."
static LARGE_INTEGER qpcLast;
static DWORD tickCountLast;
static bool inited;
LARGE_INTEGER qpc;
QueryPerformanceCounter(&qpc);
#if defined(_M_IX86) || defined(__i386__)
DWORD tickCount = GetTickCount();
#else
ULONGLONG tickCount = GetTickCount64();
#endif
if (inited) {
__int64 qpcElapsed = ((qpc.QuadPart - qpcLast.QuadPart) * 1000) / qpcFrequency.QuadPart;
__int64 tickCountElapsed;
if (tickCount >= tickCountLast)
tickCountElapsed = (tickCount - tickCountLast);
else {
#if COMPILER(MINGW)
__int64 tickCountLarge = tickCount + 0x100000000ULL;
#else
__int64 tickCountLarge = tickCount + 0x100000000I64;
#endif
tickCountElapsed = tickCountLarge - tickCountLast;
}
// force a re-sync if QueryPerformanceCounter differs from GetTickCount by more than 500ms.
// (500ms value is from http://support.microsoft.com/kb/274323)
__int64 diff = tickCountElapsed - qpcElapsed;
if (diff > 500 || diff < -500)
syncedTime = false;
} else
inited = true;
qpcLast = qpc;
tickCountLast = tickCount;
return (1000.0 * qpc.QuadPart) / static_cast<double>(qpcFrequency.QuadPart);
}
static bool qpcAvailable()
{
static bool available;
static bool checked;
if (checked)
return available;
available = QueryPerformanceFrequency(&qpcFrequency);
checked = true;
return available;
}
double currentTime()
{
// Use a combination of ftime and QueryPerformanceCounter.
// ftime returns the information we want, but doesn't have sufficient resolution.
// QueryPerformanceCounter has high resolution, but is only usable to measure time intervals.
// To combine them, we call ftime and QueryPerformanceCounter initially. Later calls will use QueryPerformanceCounter
// by itself, adding the delta to the saved ftime. We periodically re-sync to correct for drift.
static double syncLowResUTCTime;
static double syncHighResUpTime;
static double lastUTCTime;
double lowResTime = lowResUTCTime();
if (!qpcAvailable())
return lowResTime / 1000.0;
double highResTime = highResUpTime();
if (!syncedTime) {
timeBeginPeriod(1); // increase time resolution around low-res time getter
syncLowResUTCTime = lowResTime = lowResUTCTime();
timeEndPeriod(1); // restore time resolution
syncHighResUpTime = highResTime;
syncedTime = true;
}
double highResElapsed = highResTime - syncHighResUpTime;
double utc = syncLowResUTCTime + highResElapsed;
// force a clock re-sync if we've drifted
double lowResElapsed = lowResTime - syncLowResUTCTime;
const double maximumAllowedDriftMsec = 15.625 * 2.0; // 2x the typical low-res accuracy
if (fabs(highResElapsed - lowResElapsed) > maximumAllowedDriftMsec)
syncedTime = false;
// make sure time doesn't run backwards (only correct if difference is < 2 seconds, since DST or clock changes could occur)
const double backwardTimeLimit = 2000.0;
if (utc < lastUTCTime && (lastUTCTime - utc) < backwardTimeLimit)
return lastUTCTime / 1000.0;
lastUTCTime = utc;
return utc / 1000.0;
}
#else
double currentTime()
{
static bool init = false;
static double lastTime;
static DWORD lastTickCount;
if (!init) {
lastTime = lowResUTCTime();
lastTickCount = GetTickCount();
init = true;
return lastTime;
}
DWORD tickCountNow = GetTickCount();
DWORD elapsed = tickCountNow - lastTickCount;
double timeNow = lastTime + (double)elapsed / 1000.;
if (elapsed >= 0x7FFFFFFF) {
lastTime = timeNow;
lastTickCount = tickCountNow;
}
return timeNow;
}
#endif // USE(QUERY_PERFORMANCE_COUNTER)
#elif USE(GLIB)
// Note: GTK on Windows will pick up the PLATFORM(WIN) implementation above which provides
// better accuracy compared with Windows implementation of g_get_current_time:
// (http://www.google.com/codesearch/p?hl=en#HHnNRjks1t0/glib-2.5.2/glib/gmain.c&q=g_get_current_time).
// Non-Windows GTK builds could use gettimeofday() directly but for the sake of consistency lets use GTK function.
double currentTime()
{
GTimeVal now;
g_get_current_time(&now);
return static_cast<double>(now.tv_sec) + static_cast<double>(now.tv_usec / 1000000.0);
}
#else
double currentTime()
{
struct timeval now;
gettimeofday(&now, 0);
return now.tv_sec + now.tv_usec / 1000000.0;
}
#endif
#if USE(GLIB)
double monotonicallyIncreasingTime()
{
return static_cast<double>(g_get_monotonic_time() / 1000000.0);
}
#elif OS(DARWIN)
double monotonicallyIncreasingTime()
{
// Based on listing #2 from Apple QA 1398, but modified to be thread-safe.
static mach_timebase_info_data_t timebaseInfo;
static std::once_flag initializeTimerOnceFlag;
std::call_once(initializeTimerOnceFlag, [] {
kern_return_t kr = mach_timebase_info(&timebaseInfo);
ASSERT_UNUSED(kr, kr == KERN_SUCCESS);
ASSERT(timebaseInfo.denom);
});
return (mach_absolute_time() * timebaseInfo.numer) / (1.0e9 * timebaseInfo.denom);
}
#else
double monotonicallyIncreasingTime()
{
static double lastTime = 0;
double currentTimeNow = currentTime();
if (currentTimeNow < lastTime)
return lastTime;
lastTime = currentTimeNow;
return currentTimeNow;
}
#endif
std::chrono::microseconds currentCPUTime()
{
#if OS(DARWIN)
mach_msg_type_number_t infoCount = THREAD_BASIC_INFO_COUNT;
thread_basic_info_data_t info;
// Get thread information
mach_port_t threadPort = mach_thread_self();
thread_info(threadPort, THREAD_BASIC_INFO, reinterpret_cast<thread_info_t>(&info), &infoCount);
mach_port_deallocate(mach_task_self(), threadPort);
return std::chrono::seconds(info.user_time.seconds + info.system_time.seconds) + std::chrono::microseconds(info.user_time.microseconds + info.system_time.microseconds);
#elif OS(WINDOWS)
union {
FILETIME fileTime;
unsigned long long fileTimeAsLong;
} userTime, kernelTime;
// GetThreadTimes won't accept null arguments so we pass these even though
// they're not used.
FILETIME creationTime, exitTime;
GetThreadTimes(GetCurrentThread(), &creationTime, &exitTime, &kernelTime.fileTime, &userTime.fileTime);
return std::chrono::microseconds((userTime.fileTimeAsLong + kernelTime.fileTimeAsLong) / 10);
#else
// FIXME: We should return the time the current thread has spent executing.
static auto firstTime = std::chrono::steady_clock::now();
return std::chrono::duration_cast<std::chrono::microseconds>(std::chrono::steady_clock::now() - firstTime);
#endif
}
void sleep(double value)
{
// It's very challenging to find portable ways of sleeping for less than a second. On UNIX, you want to
// use usleep() but it's hard to #include it in a portable way (you'd think it's in unistd.h, but then
// you'd be wrong on some OSX SDKs). Also, usleep() won't save you on Windows. Hence, bottoming out in
// lock code, which already solves the sleeping problem, is probably for the best.
Lock fakeLock;
Condition fakeCondition;
LockHolder fakeLocker(fakeLock);
fakeCondition.waitFor(fakeLock, Seconds(value));
}
} // namespace WTF