Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!apple!ames!vsi1!altos86!dtynan
From: jmberkley@watnext.waterloo.edu (J. Michael Berkley)
Newsgroups: comp.unix
Subject: Re: QNX
Message-ID: <3482@altos86.Altos.COM>
Date: 6 Jul 89 19:06:20 GMT
References: <1046@altos86.UUCP> <1096@altos86.UUCP> <1226@altos86.UUCP>
Sender: dtynan@altos86.Altos.COM
Reply-To: uunet!RELAY.CS.NET!jmberkley%watnext.waterloo.edu
Organization: University of Waterloo, Waterloo, Ontario, Canada
Lines: 19
Approved: dtynan@zorba.Tynan.COM
In-Reply-To: peter@ficc.uu.net's message of 12 Jun 89 07:43:50 GMT


In article <1226@altos86.UUCP> peter@ficc.uu.net writes:
> I was also under the impression that QNX does NOT implement the fork()
> system call. This is just fine for realtime work (fork() is a pig), but
> it's not adequate in a development environment where you're going to want
> to run your favorite UNIX tools.

No, QNX does have fork(), and it works very efficiently.  I once wrote
a terminal emulator for QNX that used fork() to start two copies of
itself, one task for each of the screen, keyboard and serial port.
QNX also shares code, so the only memory overhead was the data.
Inter-task messages under QNX are so efficient that using the three
tasks with inter-task messaging made the emulator faster than one
monolithic task polling the keyboard and serial port.

 Mike Berkley, University of Waterloo
 PAMI Lab
 jmberkley@watnext.waterloo.edu
 {utai,uunet}!watmath!watnext!jmberkley