Re: [Dev-C++] time - try this,

Guess what. I normally _always_ test things before sending out answers.
Why do you think I said that one machine stepped with 10ms granularity,
and another with 15.625ms? Anyway, the program always endied after >=
CLOCKS_PER_SEC ticks.

/Per W

On Thu, 30 Dec 2004, Eric wrote:

> XXXXXXXXX
>
> Sorry
>
>
> ----- Original Message -----
> From: "Eric" <eri...@cl...>
> To: "Per Westermark" <pw...@ia...>
> Cc: <dev...@li...>
> Sent: Thursday, December 30, 2004 3:47 PM
> Subject: Re: [Dev-C++] time - try this,
>
>
> CAN YOUI JOKERS STOP LOOKING AT THE ##### CODE ANF JUSTR TRY THE %$%%%t% THING, i DON'T ASJFOPE MUCG
>
>
> ----- Original Message -----
> From: "Per Westermark" <pw...@ia...>
> To: "Eric" <eri...@cl...>
> Cc: <dev...@li...>
> Sent: Thursday, December 30, 2004 12:39 PM
> Subject: Re: [Dev-C++] time - try this,
>
>
> If you do
>
> ((ticks2 - ticks1) / CLOCKS_PER_SEC) < 1
>
> (observe difference before division) there is no possibility that you can
> get an answer less than 1000. If you do two divisions, you might get into
> problems because of the round-off errors you introduce.
>
> While you have to divide the value from clock() with CLOCKS_PER_SEC to get
> seconds, there is no guarantee that the value returned by clock() can
> assume all possible values, i.e. if CLOCKS_PER_TICK is 1000, it is still
> possible that the only values you can get are 32, 64, 96 etc if the actual
> clock only counts in 32ms resolution. In real life, most of the time,
> clock() steps with a non-integer number of ticks because the underlying hw
> might tick on a frequency that is
>
> For Posix compliance, CLOCKS_PER_SEC should _always_ be 1000000 i.e.
> microsecond resolution, even of the real computer clock can't tick once
> every us.
>
> On one of my machines - WinXP, P4 HT - the possible steps are 15.625ms,
> i.e. I get the values 15,31,62,78,... for a total of 64 values/second.
>
> On another machine using Linux, clock() steps exactly exactly every 10ms,
> but since the resolution is in us, the possible values increases with
> 10000 at a time.
>
> /Per W