...one of the most highly
regarded and expertly designed C++ library projects in the
world.
— Herb Sutter and Andrei
Alexandrescu, C++
Coding Standards
Timer Home CPU timers Original timers |
Knowing how long a program takes to execute is useful in both test and production environments. It may also be helpful if such timing information is broken down into wall clock time, CPU time spent by the user, and CPU time spent by the operating system servicing user requests.
Class cpu_timer
measures
wall clock time, user CPU process time, and system CPU process time. Class
auto_cpu_timer
is a refinement of
cpu_timer
that automatically reports the elapsed times when an
auto_cpu_timer
object is destroyed.
Boost.Timer is implemented as a separately compiled library, so you must install binaries in a location that can be found by your linker. If you followed the Boost Getting Started instructions, that's already done for you.
auto_cpu_timer
The simplest and most common use is to add the two lines highlighted below
to a scope you want to time. See
auto_cpu_timer_example.cpp
for the source code.
#include <boost/timer/timer.hpp> #include <cmath> int main() { boost::timer::auto_cpu_timer t; for (long i = 0; i < 100000000; ++i) std::sqrt(123.456L); // burn some time return 0; }
When the auto_cpu_timer
object is created, it starts timing. When
it is destroyed at the end of the scope, its destructor stops the timer and
displays timing information on the default output stream, std::cout
.
The output of this program will look something like this:
5.713010s wall, 5.709637s user + 0.000000s system =
5.709637s CPU (99.9%)
In other words, this program ran in 5.713010
seconds as would be measured by a
clock on the wall, the operating system charged it for 5.709637
seconds of user CPU
time and 0 seconds of system CPU time, the total of these two was 5.709637
, and that
represented 99.9
percent of the wall clock time.
The output stream, number of decimal places reported, and reporting format
can be controlled by auto_cpu_timer
constructor arguments. Here is
what the output from the above program would look like for several different
sets of constructor arguments:
Construction | Output |
t |
5.713010s wall, 5.709637s user + 0.000000s system = 5.709637s
CPU (99.9%) |
t(std::cerr, 2) |
5.71s wall, 5.70s user + 0.00s system = 5.70s CPU (99.9%) |
t(1) |
5.7s wall, 5.7s user + 0.0s system = 5.7s CPU (99.9%) |
t(3, "%w seconds\n") |
5.713 seconds |
t("%t sec CPU, %w sec real")
|
5.709637 sec CPU, 5.713010 sec real |
The processing of the format string is described here.
cpu_timer
The following code creates a checkpoint every 20 CPU seconds:
using boost::timer::cpu_timer; using boost::timer::cpu_times; using boost::timer::nanosecond_type; ... nanosecond_type const twenty_seconds(20 * 1000000000LL); nanosecond_type last(0); cpu_timer timer; while (more_transactions) { process_a_transaction(); cpu_times const elapsed_times(timer.elapsed()); nanosecond_type const elapsed(elapsed_times.system + elapsed_times.user); if (elapsed >= twenty_seconds) { ... create a checkpoint ... last = elapsed; } }
How accurate are these timers?
The resolution of a clock, and thus timers built on that clock,
is the minimum period time that can be measured. The program
cpu_timer_info.cpp
measures
the resolution of cpu_timer
.
O/S | Processor | Wall-clock | CPU | ||
Resolution | Comments | User Resolution |
System Resolution |
||
Mac OS X Lion | Intel circa 2007 | 2100ns 2200ns |
Some variation within a range. | 10000000ns | 10000000ns |
Ubuntu Linux 11.4 | Intel circa 2005 | 516ns | Very little variation, typically less than 5ns | 10000000ns | 10000000ns |
Windows 7 | Intel Core i7 860 @ 2.9 GHz | 366ns | Some variation, usually in multiples of 366ns | 15600100ns | 15600100ns |
Windows 7 | Intel Mobile T7200 @ 2.0 GHz | 2050ns | Much variation. Resolution degrades when processor slows, probably due to known chipset errata. | 15600100ns | 15600100ns |
Windows XP | Intel Atom N2800 @ 1.0 GHz | 1437ns | Some variation. | 15625000ns | 15625000ns |
Wall-clock timings are subject to many outside influences, such as the impact of other processes.
cpu_timer
andauto_cpu_timer
obtain Wall-clock timings from Boost.Chrono'shigh_resolution_clock
. On Intel compatible CPU's running Windows, Linux, and Mac OS X, this is a "steady clock" [C++11 20.11.3], but may not be steady on other platforms.cpu_timer_info.cpp
reports whether or not thehigh_resolution_clock
is steady on a particular platform.Steady clocks are defined by the C++11 standard as clocks for which values never decrease as physical time advances and for which values advance at a steady rate relative to real time. That is, the clock may not be adjusted. Clocks that are steady never run backwards, even when the operating system's clock is reset backwards such as during a daylight saving time transition.
Timings of debug builds are often several times slower than release builds, because compiler optimization is turned off and because libraries often supply very expensive error checks on debug builds.
Synthetic benchmark code may be optimized way, particularly if NDEBUG is defined. It may be necessary to inspect generated code to verify this isn't happening.
Think about what is important to your application. For a production process, the wall clock time may be what is most important. To study the efficiency of code, total CPU time (user + system) is often a much better measure.
A useful recommendation is to never trust timings unless they are (1) at least 100 times longer than the CPU time resolution, (2) run multiple times, and (3) run on release builds. And results that are too good to be true need to be should be investigated skeptically.
Shared libraries (DLLs and .so's) may incur extra time delays, including expensive disk accesses, the first time a timer or other function is called. If that would be misleading, static linking should be considered.
Specifications are given in the style of the C++ standard library (C++11, 17.5.1.4 [structure.specifications]). An additional Overview element may be provided to aid understanding. Overview elements are only informative - actual semantics are given by the other detailed specification elements.
Functions not specified as noexcept
will throw
std::bad_alloc
exceptions if a memory allocation error occurs. Other
errors are reported by time values of -1. [Note: Modern hardware and
operating systems have robust clock subsystems, so such errors are unusual if
even possible at all. -- end note]
The Timer library meets the same data race avoidance requirements as the C++11 standard library (17.6.5.9 [res.on.data.races]). Shared objects of Timer library types risk undefined behavior unless the user supplies a locking mechanism. See C++11, 17.6.4.10 [res.on.objects], Shared objects and the library.
<boost/timer/timer.hpp>
synopsis
namespace boost { namespace timer { class cpu_timer; // wall clock, user, and system timer class auto_cpu_timer; // automatic report() on destruction typedef boost::int_least64_t nanosecond_type; struct cpu_times { nanosecond_type wall; nanosecond_type user; nanosecond_type system; void clear(); }; const int default_places = 6; std::string format(const cpu_times& times, short places, const std::string& format); std::string format(const cpu_times& times, short places = default_places); } // namespace timer } // namespace boost |
The default format is " %ws wall, %us user + %ss system = %ts CPU (%p%)\n".
nanosecond_type
The typedef nanosecond_type
provides an implementation defined type capable
of representing nanoseconds. For POSIX and Windows systems,
nanoseconds_type
is boost::int_least64_t
.
The underlying type is not based on the Boost Date-Time or Chrono library to avoid a dependency on a large library. This design choice may change at some future date.
Although nanosecond_type
is capable of representing one
nanosecond, the actual resolution of common operating system timers may be
much lower. For wall clock time on desktop systems circa 2010, resolution is
often no better than than one microsecond. For user and system time, typical
resolution is 15 milliseconds on Windows and 10 milliseconds on
POSIX.
cpu_times
Struct cpu_times
packages the elapsed wall clock time, user
process CPU time, and system process CPU time. See
Current time values for definitions of the
source of these elapsed times.
void clear();
Effects:
wall = user = system = 0LL
.
std::string format(const cpu_times& times, short places, const std::string& format); std::string format(const cpu_times& times, short places = default_places);
Overview: Converts
times
's values to strings representing seconds toplaces
decimal places, and inserts them into the return string as controlled byformat
.Remarks: For the overload without the
format
argument, the default format is used asformat
.Returns: A string that is a copy of
format
, except that any instances of the sequences shown below are replaced by the indicated value. Times are reported in seconds, shown tostd::max(0, std::min(default_places, 9))
decimal places. Percentage is reported to one decimal place. [Note: percentage may exceed 100% due to differences in how operating systems measure various times. --end note]
Sequence Replacement value %w
times.wall
%u
times.user
%s
times.system
%t
times.user + times.system
%p
The percentage of times.wall
represented bytimes.user + times.system
cpu_timer
cpu_timer
objects measure wall clock elapsed time and process elapsed
time charged to the user and system.
Current time values are the current wall clock time, user process time, and system process time as provided by the operating system:
cpu_timer
synopsis
class cpu_timer { public: // constructor cpu_timer() noexcept; // compiler generated; shown for exposition only ~cpu_timer() noexcept = default; cpu_timer(const cpu_timer&) noexcept = default; cpu_timer& operator=(const cpu_timer&) noexcept = default; // observers bool is_stopped() const noexcept; cpu_times elapsed() const noexcept; std::string format(int places, const std::string& format) const; std::string format(int places = default_places) const; // actions void start() noexcept; void stop() noexcept; void resume() noexcept; }; |
cpu_timer
constructorcpu_timer() noexcept;
Effects: Constructs an object of type
cpu_timer
. Callsstart()
.
cpu_timer
observersbool is_stopped() const noexcept;
Returns:
true
if stop() was the most recent action function called, otherwisefalse
.
cpu_times elapsed() const noexcept;
Returns: If
is_stopped()
, the accumulated elapsed times as of the previous stop(). Otherwise, the elapsed times accumulated between the most recent call to start() or resume() and the current time values.
std::string format(int places, const std::string& format) const; std::string format(int places = default_places) const;
Overview: Returns a string for the current elapsed time as formatted by the format non-member function.
Returns:
boost::timer::format(elapsed(), places[, format])
.
cpu_timer
actionsvoid start() noexcept;
Effects: Begins accumulating elapsed time as of the current time values.
Postconditions:
!is_stopped()
.
void stop() noexcept;
Effects: If
!is_stopped()
, stops accumulating elapsed time as of the current time values.[Note: This is observable via
elapsed()
. -- end note]Postconditions:
is_stopped()
.
void resume() noexcept;
Overview: Restarts the timer, accumulating additional elapsed time.
Effects: If
is_stopped()
, resumes accumulating additional elapsed time, as of the current time values. Otherwise, no effect.
auto_cpu_timer
Class auto_cpu_timer
adds a report()
function to class cpu_timer
, and automatically calls report()
on destruction.
auto_cpu_timer
synopsis
class auto_cpu_timer : public cpu_timer { public: explicit auto_cpu_timer(short places = default_places); auto_cpu_timer(short places, const std::string& format); explicit auto_cpu_timer(const std::string& format); auto_cpu_timer(std::ostream& os, short places, const std::string& format); explicit auto_cpu_timer(std::ostream& os, short places = default_places); auto_cpu_timer(std::ostream& os, const std::string& format); ~auto_cpu_timer() noexcept; // compiler generated; shown for exposition only auto_cpu_timer(const auto_cpu_timer&) = default; auto_cpu_timer& operator=(const auto_cpu_timer&) = default; // observers std::ostream& ostream() const noexcept; short places() const noexcept; const std::string& format_string() const noexcept; // actions void report(); }; |
[Note: Constructors without a std::ostream&
argument argument imply
std::cout
. An argument default is avoided as it would require including <iostream>
,
with its high costs, even when the standard streams are not used. --end note]
auto_cpu_timer
constructorsexplicit auto_cpu_timer(short places = default_places); auto_cpu_timer(short places, const std::string& format); explicit auto_cpu_timer(const std::string& format); auto_cpu_timer(std::ostream& os, short places, const std::string& format);
explicit auto_cpu_timer(std::ostream& os, short places = default_places);
auto_cpu_timer(std::ostream& os, const std::string& format);
Effects: Constructs an object of type
auto_cpu_timer
and stores the ostream, places, and format string data needed to establish the postconditions.Postconditions:
- For overloads with an
os
argument,ostream() == os
. Otherwiseostream() == std::cout
.places() == places
.- For overloads with a
format
argument,format_string() == format
. Otherwiseformat_string() == std::cout
auto_cpu_timer
destructor~auto_cpu_timer() noexcept;
Effects: If
!is_stopped()
, stop(), report().[Note: Because the function is
noexcept
, implementation must ensure no exception escapes. --end note]
The observers allow testing of constructor postconditions and specification of other functionality without resorting to "for exposition only" private members.
std::ostream& ostream() const noexcept;
Returns: The ostream stored by construction or subsequent copy assignment.
short places() const noexcept;
Returns: The places stored by construction or subsequent copy assignment.
const std::string& format_string() const noexcept;
Returns: The format string stored by construction or subsequent copy assignment.
auto_cpu_timer
actionsvoid report();
Effects: As if:
ostream() << timer::format(elapsed(), places(), format_string());[Note: It may be desirable to call
stop()
before callingreport()
because doing I/O while the timer is running might produce misleading results.resume()
may be called afterwards to continue timing. --end note]
Beman Dawes and Rob Stewart developed version 2 of the library.
Beman did the initial development. Rob contributed many corrections, comments, and suggestions. In
particular, he suggested the resume()
and format()
functions, resulting in improved ease-of-use for several use cases.
Comments and suggestions came from Greg Rubino, Dave Abrahams, Vicente Botet, and John Maddock.
Revised: 08 October 2011
© Copyright Beman Dawes, 2006
© Copyright Beman Dawes and Robert Stewart, 2011
Distributed under the Boost Software License, Version 1.0. See www.boost.org/ LICENSE_1_0.txt