Provably Squeak under DOS can be used
 and link the VM to a function
 that wait until video register signal the
 endOfFrame.
Ale.

----- Original Message -----
From: "Timothy Rowledge" <tim at sumeru.stanford.edu>
To: "The general-purpose Squeak developers list"
<squeak-dev at lists.squeakfoundation.org>
Sent: Friday, February 13, 2004 8:20 PM
Subject: Re: real-time, time to display, accuracy of event timestamps


> Most OSs can provide microsecond timing values (as Andreas said, the
> accuracy and precision are probably very variable) and in fact I wrote
> a minimally usable microsecond time returning prim for the exobox
> project (it merely provided a SmallInteger microsecond count with no
> attempt to worry about wrapping etc).
>
> The display framerate is certainly an issue; the only OS I've ever come
> across that attempted to address this is RISC OS, which has an OS level
> event that tells you when the frame flyback occurs and in its BASIC has
> a way to make display wait until the next flyback. What you could try
> is modifying the ioForceDisplay to post an event to the VM with some
> suitable new type and the information is just the timestamp. That way
> you can track the time of each display and you might just possibly be
> able to work out a relationship with the frame rate. Might there be (in
> windows for example) some API to track the display framerate for games
> purposes?
>
> Another issue that might affect you is the timestamp put on the events;
> it's added _at the time the event is queued in the VM_ not the time the
> OS detects it - except in the case of some events in Windows, which
> appears to extract a timestamp from the OS event.
>
> If you have enough time to spare to investigate all this you might well
> get something very interesting working.
>
> tim
>