From: Jürgen Gotschlich (juergen.gotschlich++at++dlr.de)
Date: 03/25/2003 05:38:27
Goncalo,
well, I guess you're right in implementing this threading model. So I'll
have to look forward in this way.
But I read in some older articles that performer internally is based on
sproc threading and therefore can get into seriously troubles when
mixing with pthreads. Did they change their internal coding with respect
to pthreads ?
Again a question about shared memory: Infact, using 2 communicating
processes will need to implement an interprocess communication based on
shared memory. But aren't 2 threads (spawned from 1 process) actually
using the same portion of memory ?
I posted this as well directly to info-performer++at++sgi.com to read for
other interested users.
Thanks for your help
Juergen
-------------------------------------------
Goncalo Carvalho wrote:
>
> Hey,
>
> >
> > What are you using to spawn this ?
> >
>
> Pthreads, but you can fork if you like.
>
> > > to run a loop to process the input
> > > and put it in shared memory. The main loop or some other object fetches
> > > as necessary.
> >
> > What do you mean by this ?
> > You have continuously running a loop to read data via network in this separate process ?
>
> Yes. What I have is a way to swap between continuous incoming data and a
> request based
> approach.
>
> > But you need some scheduling for this process ? Maybe pausing for 10ms after each read request.
>
> It uses an event based approach. If no data is available
> no cpu cycles are used. I use a mutex to synchronize write/read access
> from the shared data.
>
> >
> > In my case the network source is sending motion-data every 20ms and I need this very properly
> > for a smooth display-update of the terrain-scenery. The Performer-App. is running with 30Hz/33ms.
> > If the Performer-App. is doing it's own motion control via mouse input, I don't have any delay,
> > but reading from the network in each frame (hopefully in the APP-process ) causes delays of up to 60ms.
> >
> > What is the advantage of using a separate process and shared memory ?
> >
>
> Marc answered this. Shared memory is a convenient way of making sure the
> data is available for all processes.
>
> Goncalo
>
> --
> Goncalo Carvalho
> Digital Design Studio - Glasgow School of Art
> House for an Art Lover - Bellahouston Park
> 10 Dumbreck Road
> Glasgow G41 5BW
> United Kingdom
>
> Tel: +44 (0) 141 353 4616
> Fax: +44 (0) 141 353 4422
> Mobile: +44 (0) 7766 295867
> mailto:g.carvalho++at++gsa.ac.uk
-- ________________________________________________________________________Jürgen Gotschlich email: juergen.gotschlich++at++dlr.de __/|_ Phone: +49(531)295-3248/2778 /_/_/_/ Fax : +49(531)295-2640 |/ DLR
Deutsches Zentrum für Luft-und Raumfahrt | German Aerospace Center in der Helmholtz-Gemeinschaft | Institut für Flugsystemtechnik | Inst. of Flight Research Abt. Flugsimulation | Flight Simulation Dep. Lilienthalplatz 7 | Lilienthalplatz 7 D-38108 Braunschweig | D-38108 Braunschweig, Germany ________________________________________________________________________
This archive was generated by hypermail 2b29 : Tue Mar 25 2003 - 05:36:34 PST