Re: [boost] [timer]Behavior under MS Windows and documentation

From: Yuriy Koblents-Mishke (yurakm_at_[hidden])
Date: 2007-02-01 19:23:27

On 2/1/07, Philippe Vaucher <philippe.vaucher_at_[hidden]> wrote:
>
> Hello,
>
> I'm currently creating a new version of the boost timer, you can grab it
> there :
>
> http://www.boost-consulting.com/vault/index.php?directory=date_time
> (get timer_2007_01_03.zip)
>
> [snip]

Dear Philippe:

I glanced on your library, but am not ready yet for discussing it.

You can include into your library a couple of additional timers (see the
attached files).

The POSIX timer was not really much better than a stock clock() function for
my applications, except that on a two-processor Apple Mac it sometimes
prints a bit higher timings (in third sign): with kernel and user running on
two different processors, it happens that cpu times are a bit higher than
elapsed times.

Windows timer with GetThreadTimes is really useful for my applications, but
I did not test it as thoroughly as it is necessary for submission. I would
suggest to test it using your harness.

It is easy to write a similar timer with GetThreadTimes replaced by
GetProcessTimes. I would expect that GetProcessTimes will have somewhat
higher overheads. Of course, for multithreaded codes GetThreadTimes and
GetProcessTimes will measure different things.

Sincerely,

   Yura