----- Original Message ----- From: "Bruno Lalande" <bruno.lalande@gmail.com> To: <boost-users@lists.boost.org> Sent: Thursday, July 24, 2008 10:51 AM Subject: Re: [Boost-users] Boost.Date_Time - How to measure time elapsed?

Hello,

...

I would suggest using boost::timer instead. It's quite simple:

Quite simple but quite limited, for portability reasons. I personally no longer use it since I've discovered it can't run over 35 minutes on my platform. It's also far from having the microsecond resolution requested here.

The request was mili seconds (msec). Vicente

Hi,

Quite simple but quite limited, for portability reasons. I personally no longer use it since I've discovered it can't run over 35 minutes on my platform.

What platform is that? Cheers, Andrej __________________________________________________________ Not happy with your email address?. Get the one you really want - millions of new email addresses available now at Yahoo! http://uk.docs.yahoo.com/ymail/new.html

On Fri, Jul 25, 2008 at 5:29 PM, Bruno Lalande <bruno.lalande@gmail.com> wrote:

...

What platform is that?

A very common one: Linux 2.6 with GCC 4.1 or 4.2. The distribution is Kubuntu.

timer().elapsed_max() gives 2147.48 on this platform and indeed, elapsed() returns negative numbers above that time (before returning to 0).

I had thought about the possibility of making a policy-based version of boost::timer, with something like a "heart policy" that would define the way in which the time would be obtained and stored internally. There would be a portable one (the one of boost::timer currently), another one based on boost::date_time, and why not some platform-specific ones (QueryPerformanceCounter on windows for instance). But I never had the time to dig more :-(

I don't mean to hijack the thread but please let me share another way to use timer, with C, not C++ or boots. In the code, not only the time of one test, but average time, and its standard derivation are calculated so we will know how "close" is the average time. My question: Is there any similar implementation in boost? The code is as follows: int count = 30; long long times[count]; for (i = 0; i < count; i++) ｛ time start; gettimeofday (&start, NULL); // do something gettimeofday (&end, NULL); long long time = ((end.tv_sec - start.tv_sec) * 1000000 + (end.tv_usec - start.tv_usec)); times[i] = time; } // and use it printf("mean: %f, stddev: %f\n", mean(times, 20), stddev(times, 20); double variance (long long *x, int n) { double m = x[0]; double s = 0; int i; for (i = 0; i < n; i++) { double m_new = m + (x[i] - m) / (i + 1); s = s + (x[i] - m) * (x[i] - m_new); m = m_new; } return s / (n - 1); } double mean (long long *x, int n) { double m = x[0]; int i; for (i = 0; i < n; i++) { m = m + (x[i] - m) / (i + 1); } return m; } double stddev (long long *x, int n) { return sqrt (variance (x, n)); } -- Best Regards, Nguyen Hung Vu ( Nguyễn Vũ Hưng ) vuhung16plus{remove}@gmail.dot.com <vuhung16plus%7Bremove%7D@gmail.dot.com>, YIM: vuhung16 , Skype: vuhung16dg Japan through an eye of a gaijin: http://www.flickr.com/photos/vuhung/tags/fav/

I don't know it this helps, but I think it is getting system time to a reasonable precision on Windows and Linux platforms, I use two doubles representing the proleptic Julian day Number and the fraction of the day: #ifdef _MSC_VER struct __timeb64 timebuffer; struct tm *sysUTC = new tm; #if _MSC_VER < 1400 // Visual Studio 7.1 and earlier. _ftime64( &timebuffer ); sysUTC = _gmtime64( &timebuffer.time ); #else _ftime64_s( &timebuffer ); _gmtime64_s( sysUTC, &timebuffer.time ); #endif sec = sysUTC->tm_sec + double(timebuffer.millitm) / 1000.0; double sysDate = JulianDayNumber(sysUTC->tm_year+1900,(char)sysUTC->tm_mon+1,(char)sysUTC->tm_mday); double sysFoD = ToFoD(sysUTC->tm_hour,(char)sysUTC->tm_min,sec); delete sysUTC; #else struct timeval ltime; gettimeofday(<ime,NULL); // Obtain UTC from the computer. struct tm sysUTC; gmtime_r( <ime.tv_sec, &sysUTC ); sec = sysUTC.tm_sec+ltime.tv_usec/1000000.0; double sysDate = JulianDayNumber(sysUTC.tm_year+1900,(char)sysUTC.tm_mon+1,(char)sysUTC.tm_mday); double sysFoD = ToFoD(sysUTC.tm_hour,(char)sysUTC.tm_min,sec); #endif Andrew On Fri, Jul 25, 2008 at 6:57 PM, Nguyen Vu Hung <vuhung16plus@gmail.com> wrote:

On Fri, Jul 25, 2008 at 5:29 PM, Bruno Lalande <bruno.lalande@gmail.com> wrote:

...

...

What platform is that?

A very common one: Linux 2.6 with GCC 4.1 or 4.2. The distribution is Kubuntu.

timer().elapsed_max() gives 2147.48 on this platform and indeed, elapsed() returns negative numbers above that time (before returning to 0).

I had thought about the possibility of making a policy-based version of boost::timer, with something like a "heart policy" that would define the way in which the time would be obtained and stored internally. There would be a portable one (the one of boost::timer currently), another one based on boost::date_time, and why not some platform-specific ones (QueryPerformanceCounter on windows for instance). But I never had the time to dig more :-(

I don't mean to hijack the thread but please let me share another way to use timer, with C, not C++ or boots. In the code, not only the time of one test, but average time, and its standard derivation are calculated so we will know how "close" is the average time.

My question: Is there any similar implementation in boost?

The code is as follows:

int count = 30; long long times[count];

for (i = 0; i < count; i++)　｛ time start; gettimeofday (&start, NULL); // do something gettimeofday (&end, NULL); long long time = ((end.tv_sec - start.tv_sec) * 1000000 + (end.tv_usec - start.tv_usec)); times[i] = time; } // and use it printf("mean: %f, stddev: %f\n", mean(times, 20), stddev(times, 20);

double variance (long long *x, int n) { double m = x[0]; double s = 0; int i; for (i = 0; i < n; i++) { double m_new = m + (x[i] - m) / (i + 1); s = s + (x[i] - m) * (x[i] - m_new); m = m_new; } return s / (n - 1); }

double mean (long long *x, int n) { double m = x[0]; int i; for (i = 0; i < n; i++) { m = m + (x[i] - m) / (i + 1); } return m; }

double stddev (long long *x, int n) { return sqrt (variance (x, n)); }

-- Best Regards, Nguyen Hung Vu ( Nguyễn Vũ Hưng ) vuhung16plus{remove}@gmail.dot.com , YIM: vuhung16 , Skype: vuhung16dg Japan through an eye of a gaijin: http://www.flickr.com/photos/vuhung/tags/fav/ _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

-- ___________________________________________ Andrew J. P. Maclean Centre for Autonomous Systems The Rose Street Building J04 The University of Sydney 2006 NSW AUSTRALIA Ph: +61 2 9351 3283 Fax: +61 2 9351 7474 URL: http://www.acfr.usyd.edu.au/ ___________________________________________