// Copyright (C) 2016 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only #include "qelapsedtimer.h" QT_BEGIN_NAMESPACE /*! \class QElapsedTimer \inmodule QtCore \brief The QElapsedTimer class provides a fast way to calculate elapsed times. \since 4.7 \reentrant \ingroup tools The QElapsedTimer class is usually used to quickly calculate how much time has elapsed between two events. Its API is similar to that of QTime, so code that was using that can be ported quickly to the new class. However, unlike QTime, QElapsedTimer tries to use monotonic clocks if possible. This means it's not possible to convert QElapsedTimer objects to a human-readable time. The typical use-case for the class is to determine how much time was spent in a slow operation. The simplest example of such a case is for debugging purposes, as in the following example: \snippet qelapsedtimer/main.cpp 0 In this example, the timer is started by a call to start() and the elapsed time is calculated by the elapsed() function. The time elapsed can also be used to recalculate the time available for another operation, after the first one is complete. This is useful when the execution must complete within a certain time period, but several steps are needed. The \tt{waitFor}-type functions in QIODevice and its subclasses are good examples of such need. In that case, the code could be as follows: \snippet qelapsedtimer/main.cpp 1 Another use-case is to execute a certain operation for a specific timeslice. For this, QElapsedTimer provides the hasExpired() convenience function, which can be used to determine if a certain number of milliseconds has already elapsed: \snippet qelapsedtimer/main.cpp 2 It is often more convenient to use \l{QDeadlineTimer} in this case, which counts towards a timeout in the future instead of tracking elapsed time. \section1 Reference Clocks QElapsedTimer will use the platform's monotonic reference clock in all platforms that support it (see QElapsedTimer::isMonotonic()). This has the added benefit that QElapsedTimer is immune to time adjustments, such as the user correcting the time. Also unlike QTime, QElapsedTimer is immune to changes in the timezone settings, such as daylight-saving periods. On the other hand, this means QElapsedTimer values can only be compared with other values that use the same reference. This is especially true if the time since the reference is extracted from the QElapsedTimer object (QElapsedTimer::msecsSinceReference()) and serialised. These values should never be exchanged across the network or saved to disk, since there's no telling whether the computer node receiving the data is the same as the one originating it or if it has rebooted since. It is, however, possible to exchange the value with other processes running on the same machine, provided that they also use the same reference clock. QElapsedTimer will always use the same clock, so it's safe to compare with the value coming from another process in the same machine. If comparing to values produced by other APIs, you should check that the clock used is the same as QElapsedTimer (see QElapsedTimer::clockType()). \sa QTime, QChronoTimer, QDeadlineTimer */ /*! \enum QElapsedTimer::ClockType This enum contains the different clock types that QElapsedTimer may use. QElapsedTimer will always use the same clock type in a particular machine, so this value will not change during the lifetime of a program. It is provided so that QElapsedTimer can be used with other non-Qt implementations, to guarantee that the same reference clock is being used. \value SystemTime The human-readable system time. This clock is not monotonic. \value MonotonicClock The system's monotonic clock, usually found in Unix systems. This clock is monotonic. \value TickCounter Not used anymore. \value MachAbsoluteTime The Mach kernel's absolute time (\macos and iOS). This clock is monotonic. \value PerformanceCounter The performance counter provided by Windows. This clock is monotonic. \section2 SystemTime The system time clock is purely the real time, expressed in milliseconds since Jan 1, 1970 at 0:00 UTC. It's equivalent to the value returned by the C and POSIX \tt{time} function, with the milliseconds added. This clock type is currently only used on Unix systems that do not support monotonic clocks (see below). This is the only non-monotonic clock that QElapsedTimer may use. \section2 MonotonicClock This is the system's monotonic clock, expressed in milliseconds since an arbitrary point in the past. This clock type is used on Unix systems which support POSIX monotonic clocks (\tt{_POSIX_MONOTONIC_CLOCK}). \section2 MachAbsoluteTime This clock type is based on the absolute time presented by Mach kernels, such as that found on \macos. This clock type is presented separately from MonotonicClock since \macos and iOS are also Unix systems and may support a POSIX monotonic clock with values differing from the Mach absolute time. This clock is monotonic. \section2 PerformanceCounter This clock uses the Windows functions \tt{QueryPerformanceCounter} and \tt{QueryPerformanceFrequency} to access the system's performance counter. This clock is monotonic. \sa clockType(), isMonotonic() */ /*! \fn QElapsedTimer::QElapsedTimer() \since 5.4 Constructs an invalid QElapsedTimer. A timer becomes valid once it has been started. \sa isValid(), start() */ /*! \fn bool QElapsedTimer::operator==(const QElapsedTimer &lhs, const QElapsedTimer &rhs) noexcept Returns \c true if \a lhs and \a rhs contain the same time, false otherwise. */ /*! \fn bool QElapsedTimer::operator!=(const QElapsedTimer &lhs, const QElapsedTimer &rhs) noexcept Returns \c true if \a lhs and \a rhs contain different times, false otherwise. */ /*! \fn bool operator<(const QElapsedTimer &lhs, const QElapsedTimer &rhs) noexcept \relates QElapsedTimer Returns \c true if \a lhs was started before \a rhs, false otherwise. The returned value is undefined if one of the two parameters is invalid and the other isn't. However, two invalid timers are equal and thus this function will return false. */ /*! \fn QElapsedTimer::clockType() noexcept Returns the clock type that this QElapsedTimer implementation uses. Since Qt 6.6, QElapsedTimer uses \c{std::chrono::steady_clock}, so the clock type is always \l MonotonicClock. \sa isMonotonic() */ QElapsedTimer::ClockType QElapsedTimer::clockType() noexcept { // we use std::chrono::steady_clock return MonotonicClock; } /*! \fn QElapsedTimer::isMonotonic() noexcept Returns \c true if this is a monotonic clock, false otherwise. See the information on the different clock types to understand which ones are monotonic. Since Qt 6.6, QElapsedTimer uses \c{std::chrono::steady_clock}, so this function now always returns true. \sa clockType(), QElapsedTimer::ClockType */ bool QElapsedTimer::isMonotonic() noexcept { // We trust std::chrono::steady_clock to be steady (monotonic); if the // Standard Library is lying to us, users must complain to their vendor. return true; } /*! \typealias QElapsedTimer::Duration Synonym for \c std::chrono::nanoseconds. */ /*! \typealias QElapsedTimer::TimePoint Synonym for \c {std::chrono::time_point}. */ /*! Starts this timer. Once started, a timer value can be checked with elapsed() or msecsSinceReference(). Normally, a timer is started just before a lengthy operation, such as: \snippet qelapsedtimer/main.cpp 0 Also, starting a timer makes it valid again. \sa restart(), invalidate(), elapsed() */ void QElapsedTimer::start() noexcept { static_assert(sizeof(t1) == sizeof(Duration::rep)); // This assignment will work so long as TimePoint uses the same time // duration or one of finer granularity than steady_clock::time_point. That // means it will work until the first steady_clock using picoseconds. TimePoint now = std::chrono::steady_clock::now(); t1 = now.time_since_epoch().count(); QT6_ONLY(t2 = 0); } /*! Restarts the timer and returns the number of milliseconds elapsed since the previous start. This function is equivalent to obtaining the elapsed time with elapsed() and then starting the timer again with start(), but it does so in one single operation, avoiding the need to obtain the clock value twice. Calling this function on a QElapsedTimer that is invalid results in undefined behavior. The following example illustrates how to use this function to calibrate a parameter to a slow operation (for example, an iteration count) so that this operation takes at least 250 milliseconds: \snippet qelapsedtimer/main.cpp 3 \sa start(), invalidate(), elapsed(), isValid() */ qint64 QElapsedTimer::restart() noexcept { QElapsedTimer old = *this; start(); return old.msecsTo(*this); } /*! \since 6.6 Returns a \c{std::chrono::nanoseconds} with the time since this QElapsedTimer was last started. Calling this function on a QElapsedTimer that is invalid results in undefined behavior. On platforms that do not provide nanosecond resolution, the value returned will be the best estimate available. \sa start(), restart(), hasExpired(), invalidate() */ auto QElapsedTimer::durationElapsed() const noexcept -> Duration { TimePoint then{Duration(t1)}; return std::chrono::steady_clock::now() - then; } /*! \since 4.8 Returns the number of nanoseconds since this QElapsedTimer was last started. Calling this function on a QElapsedTimer that is invalid results in undefined behavior. On platforms that do not provide nanosecond resolution, the value returned will be the best estimate available. \sa start(), restart(), hasExpired(), invalidate() */ qint64 QElapsedTimer::nsecsElapsed() const noexcept { return durationElapsed().count(); } /*! Returns the number of milliseconds since this QElapsedTimer was last started. Calling this function on a QElapsedTimer that is invalid results in undefined behavior. \sa start(), restart(), hasExpired(), isValid(), invalidate() */ qint64 QElapsedTimer::elapsed() const noexcept { using namespace std::chrono; return duration_cast(durationElapsed()).count(); } /*! Returns the number of milliseconds between last time this QElapsedTimer object was started and its reference clock's start. This number is usually arbitrary for all clocks except the QElapsedTimer::SystemTime clock. For that clock type, this number is the number of milliseconds since January 1st, 1970 at 0:00 UTC (that is, it is the Unix time expressed in milliseconds). On Linux, Windows and Apple platforms, this value is usually the time since the system boot, though it usually does not include the time the system has spent in sleep states. \sa clockType(), elapsed() */ qint64 QElapsedTimer::msecsSinceReference() const noexcept { using namespace std::chrono; return duration_cast(Duration(t1)).count(); } /*! \since 6.6 Returns the time difference between this QElapsedTimer and \a other as a \c{std::chrono::nanoseconds}. If \a other was started before this object, the returned value will be negative. If it was started later, the returned value will be positive. The return value is undefined if this object or \a other were invalidated. \sa secsTo(), elapsed() */ auto QElapsedTimer::durationTo(const QElapsedTimer &other) const noexcept -> Duration { Duration d1(t1); Duration d2(other.t1); return d2 - d1; } /*! Returns the number of milliseconds between this QElapsedTimer and \a other. If \a other was started before this object, the returned value will be negative. If it was started later, the returned value will be positive. The return value is undefined if this object or \a other were invalidated. \sa secsTo(), elapsed() */ qint64 QElapsedTimer::msecsTo(const QElapsedTimer &other) const noexcept { using namespace std::chrono; return duration_cast(durationTo(other)).count(); } /*! Returns the number of seconds between this QElapsedTimer and \a other. If \a other was started before this object, the returned value will be negative. If it was started later, the returned value will be positive. Calling this function on or with a QElapsedTimer that is invalid results in undefined behavior. \sa msecsTo(), elapsed() */ qint64 QElapsedTimer::secsTo(const QElapsedTimer &other) const noexcept { using namespace std::chrono; return duration_cast(durationTo(other)).count(); } static const qint64 invalidData = Q_INT64_C(0x8000000000000000); /*! \fn QElapsedTimer::invalidate() noexcept Marks this QElapsedTimer object as invalid. An invalid object can be checked with isValid(). Calculations of timer elapsed since invalid data are undefined and will likely produce bizarre results. \sa isValid(), start(), restart() */ void QElapsedTimer::invalidate() noexcept { t1 = t2 = invalidData; } /*! Returns \c false if the timer has never been started or invalidated by a call to invalidate(). \sa invalidate(), start(), restart() */ bool QElapsedTimer::isValid() const noexcept { return t1 != invalidData && t2 != invalidData; } /*! Returns \c true if elapsed() exceeds the given \a timeout, otherwise \c false. A negative \a timeout is interpreted as infinite, so \c false is returned in this case. Otherwise, this is equivalent to \c {elapsed() > timeout}. You can do the same for a duration by comparing durationElapsed() to a duration timeout. \sa elapsed(), QDeadlineTimer */ bool QElapsedTimer::hasExpired(qint64 timeout) const noexcept { // if timeout is -1, quint64(timeout) is LLINT_MAX, so this will be // considered as never expired return quint64(elapsed()) > quint64(timeout); } bool operator<(const QElapsedTimer &lhs, const QElapsedTimer &rhs) noexcept { return lhs.t1 < rhs.t1; } QT_END_NAMESPACE