Newsgroups: comp.windows.ms.programmer
Path: funic!news.funet.fi!sunic!psinntp!psinntp!rpi!usc!sol.ctr.columbia.edu!ira.uka.de!fauern!lrz-muenchen.de!sun!walter
From: walter@lkn.e-technik.tu-muenchen.de (Walter Loepsinger)
Subject: Re: Real time Windows programming
Message-ID: <walter.730037883@sun>
Sender: news@news.lrz-muenchen.de (Mr. News)
Organization: Lehrstuhl fuer Kommunikationsnetze, TU Muenchen (Germany)
References: <C2Hxq8.F8y@dcs.ed.ac.uk>
Date: Thu, 18 Feb 1993 12:18:03 GMT
Lines: 26

ibb@dcs.ed.ac.uk (Ian Brockbank) writes:

>I have the following situation.  I am recieving co-ordinate information
>in a continuous stream (dummy, dummy, x,y,x,y,x,y,x,y,x,y,x,y,...) which

>I have to process _exactly_ one (x,y) pair every tenth of a second (although
>this can be as simple as storing the value for later use).  My 
>question is: is this possible using Borland C++ 3.1 and Windows 3.1?  If
>so can you give me any pointers as to where to look.  Would I have to
>write a device driver for this, or can I get sufficiently snappy response
>from the basic Windows?  I stress that the timing is absolutely vital.

Do _not_ do it! is my first impression. As i have just finished a
bigger application which depends on real time data: the problems
from Windows can always kill the real time conditions.
If you only need to process _ten_ events in _one_ second or even
_hundred_ events in _ten_ seconds, it would work. But inside the
one second, Windows often takes the time away from you and 
decides to, for example, draw a window, taking half a second or 
sometimes even more. When some other applications rise up, these
conditions take place more often. You could only keep up with
your events _in average_ , never in _every_ case.

hopeithelps,
Walter.

