From: Jeff Minor (jeffm++at++orion-sim.com)
Date: 01/25/2000 06:38:38
Oops, hit the wrong reply button. Here's the message I mistakenly sent
just to Mark along with his reply.
"Wilson, Mark A" wrote:
>
> According to the address list, I don't think your reply made it to the
> whole list. You may want to re-send it with the following.
>
> Thanks for the catch on CATI. I ought not work from memory.
>
> I suggested reflective memory for two reasons: it is blindingly fast
> therefore data will be received by all three channels in the same video
> frame, and delivery is guaranteed. UDP may indeed work in a local net
> as was the case for the XIG demo. There is some transport dely to
> content with but, it isn't as bad as the delay using TCP-IP.
>
> Mark A. Wilson
> Advanced Simulation Lab
> mark.a.wilson++at++lmco.com
> 407.356.6387
>
> -----Original Message-----
> From: Jeff Minor [mailto:jeffm++at++orion-sim.com]
> Sent: Tuesday, January 25, 2000 8:41 AM
> To: Wilson, Mark A
> Subject: Re: panoramic on Performer linux?
>
> CATI (not CACI) uses UDP to synchronize data in XIG. As Mark said, their
> demo looked good. The reflective memory route may not be justified
> unless you're sending a lot more data than the handful of entity
> positions transfered in the demo.
>
> Jeff
>
> "Wilson, Mark A" wrote:
> >
> > The couple between IG channels is too tight to use UDP to
> > sync, especially since delivery is not guaranteed. DIS
> > employs time tags and dead reckoning in an attempt to overcome that
> > limitation resulting in well known discontinuities.
> >
> > I'm fairly certain that simply tying sync signals together would
> > result in an unviewable image. Sync is sourced from the graphics card.
> > Imagine a monitor getting 3 unevenly spaced vertical syncs in a frame.
> > Most monitors wouldn't be able to display anything.
> >
> > My first inclination is three PCs, each driving their own channel, and
> > a separate simulation host. Use a genlockable graphics cards to sync
> > vertical on all the IG systems. Use reflective memory (ie SCRAMNet) to
> > transfer data to and from the simulation host. This is not a low cost
> > solution (about $7500 per channel) but, it has a high probability of
> > success.
> >
> > MaK has a demo of PC Stealth (not Linux Performer and not a general
> > purpose image generator) running three channels on a single PC.
> > If memory serves frame rate was in the 20Hz range. I believe they
> > were using 200SBs from Quantum3D.
> >
> > Another, non-Linux Performer, solution might be XIG from CACI. They
> > had a pretty nice multi-channel demo is IITSEC this past December.
> >
> > Mark A. Wilson
> > Advanced Simulation Lab
> > mark.a.wilson++at++lmco.com
> > 407.356.6387
>
> --
> ----------------------------------------------------------------------
> Jeff Minor 5455 Corporate Dr., Suite 116
> OASIS, Inc. Troy, MI 48098
> jeffm++at++orion-sim.com voice: (248)952-1490, ext 203 fax: (248)952-1493
-- ---------------------------------------------------------------------- Jeff Minor 5455 Corporate Dr., Suite 116 OASIS, Inc. Troy, MI 48098 jeffm++at++orion-sim.com voice: (248)952-1490, ext 203 fax: (248)952-1493
This archive was generated by hypermail 2b29 : Tue Jan 25 2000 - 06:39:39 PST