[boost] [chrono] now() performance on several clocks

Subject: [boost] [chrono] now() performance on several clocks
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2011-10-03 11:07:37

Le 28/09/11 13:29, Vicente J. Botet Escriba a écrit :
> Le 28/09/11 13:12, Stewart, Robert a écrit :
>> Vicente J. Botet Escriba wrote:
>>> Le 27/09/11 17:42, Stewart, Robert a écrit :
>>>
>>>>>    - There's no information on Boost.Timers overhead relative
>>>>>      to timing the target code.
>>>>>
>>>> Are you asking for the overhead of the time computations or
>>>> the reporting?
>>> The former.
>>>
>>>
> I will add some performance tests in Boost.Chrono to measure the time 
> spent by the now functions and the difference between two time_points 
> for each one of the clocks, so we can be fixed.
>
>
Hi,

I have measured the cost of calling Clock::now for some clocks and 
platforms. All the test are run with variant=release.

The measured code stores on a vector the result of Clock::now() 10000000 
times.
///

   for (int i=size-1; i>=0; --i)
   {
     vec[i]=Clock::now();
     //vec[i]=typename Clock::time_point(typename 
Clock::time_point::duration(i));
   }
///

On iMac 3.06 GHz Inter Core i3.

clang2.9

system_clock  around 44 nanoseconds.
steady_clock/high_resolution_clock  around 24 nanoseconds.
process_real_cpu_clock more than 600 nanoseconds.
process_user_cpu_clock more than 1000 nanoseconds.

Currently process_real_cpu_clock uses clock() and process_user_cpu_clock 
uses times().

On MinGw Windows

msvc.10

system_clock  around 32 nanoseconds.
steady_clock/high_resolution_clock  more that 2400 nanoseconds.
process_real_cpu_clock around 84 nanoseconds.

gcc-4.6

system_clock  around 38 nanoseconds.
steady_clock/high_resolution_clock  more that 2500 nanoseconds.
process_real_cpu_clock around 110 nanoseconds.

Clearly I need to make more measures for the Windows based steady_clock 
code.

The program is available on the trunk under the libs/chrono/perf directory.

Could someone run the program on a posix platform?

Please let me know if you think there is a better way to make these 
measures.

Best,
Vicente