From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <10303101646.AA24079@clio.rice.edu>
Subject: Re: Win2K/XP uclock() - using rdtsc
To: acottrel AT ihug DOT com DOT au (Andrew Cottrell)
Date: Mon, 10 Mar 2003 10:46:24 -0600 (CST)
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <002501c2e6cb$a2ec2610$0100a8c0@acp42g> from "Andrew Cottrell" at Mar 10, 2003 05:09:39 PM
X-Mailer: ELM [version 2.5 PL2]
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> 2) If I try uclock() over say a few hours is it still approx rounded up 2%
> or is it less? I expect it to go to zero as the elapsed time between calls
> increases.

It doesn't improve - the scaling is based on the initial calibration loop.
If you are unlucky and get a 5% error on the calibration, then the long
term times will be off 5% - which could make it less accurate for large
times than the 18.2 tics/sec clock.  

We could check this and re-calibrate for the longer times, so it would
essentially re-calibrate each time you called uclock(). But I wasn't
sure it was worth it.

> 3) Looking at the code example quickly in the other email is it possible to
> only perform the calibration once on the first call to uclock() instead of
> on every call?

The calibration is only done on the first call, and it waits a worse case
of 6 Tics (0.32 seconds).  Essentially it's trying to figure out what
your CPU frequency is.

> 4) BTW remember the libc docs need to be updated to include notes / info on
> the 2K/XP delays

I know, but I don't want to document something if we are going to change it.
The current code is more proof of concept/prototype/discussion.

> 5) In the libc docs there is reference to 'UCLOCKS_PER_SEC' as per:-
>     " This function returns the number of uclock ticks since an arbitrary
>     time, actually, since the first call to `uclock', which itself returns
>     zero.  The number of tics per second is `UCLOCKS_PER_SEC' (declared in
>     the `time.h' header file."
> 
>    My query is what would this be and how is it updated depending on what OS
> you run the app on?

We create a scaling factor internally in uclock() which converts the 
return value from rdtsc() to a value which matches UCLOCKS_PER_SEC.  So there
would be no need to update this.