Does anyone know of a hardware real-time MPEG encoder board that has
QNX drivers?
Thanks
Nick
I have looked about 6 month ago and found nothing.
“Nicolae P. Costescu” <nick@qrts.NOSPAM.com> wrote in message
news:398ae90a.182726326@news.concentric.net…
Does anyone know of a hardware real-time MPEG encoder board that has
QNX drivers?
Thanks
Nick
Do you have any preferred cards/platforms?
I’d write a driver for a nice PCI card if people
wanted it and I could get the specs.
Jim Boucher
Multimedia Team Leader
QNX
“Mario Charest” <mcharest@zinformatic.com> wrote in message
news:8mf6uk$giq$1@inn.qnx.com…
I have looked about 6 month ago and found nothing.
“Nicolae P. Costescu” <> nick@qrts.NOSPAM.com> > wrote in message
news:> 398ae90a.182726326@news.concentric.net> …
Does anyone know of a hardware real-time MPEG encoder board that has
QNX drivers?
Thanks
Nick
James Boucher wrote:
Do you have any preferred cards/platforms?
I’d write a driver for a nice PCI card if people
wanted it and I could get the specs.
Shouldn’t the fact that QNX is an RTOS mean that a decent S/W MPEG decoder (i.e.
as good as a H/W decoder) can be built. It would seem to me that a good S/W
decoder would be a lot more universal than a driver for a particular brand of
H/W. I find it difficult to imagine that a 400Mhz+ Pentium machine, can’t do
smooth decoding of MPEG’s with plenty of CPU to spare on a good RTOS like QNX.
–
A prototype will always be built, whether it is shipped, depends on how much
money is left in the budget.
On Mon, 07 Aug 2000 07:05:00 -0700, Rennie Allen
<rallen@computermotion.com> wrote:
Shouldn’t the fact that QNX is an RTOS mean that a decent S/W MPEG decoder (i.e.
as good as a H/W decoder) can be built. It would seem to me that a good S/W
decoder would be a lot more universal than a driver for a particular brand of
H/W. I find it difficult to imagine that a 400Mhz+ Pentium machine, can’t do
smooth decoding of MPEG’s with plenty of CPU to spare on a good RTOS like QNX.
True but my post asked for a hardware ENCODER 
Rennie Allen (rallen@computermotion.com) wrote:
: James Boucher wrote:
: >
: > Do you have any preferred cards/platforms?
: > I’d write a driver for a nice PCI card if people
: > wanted it and I could get the specs.
: >
: Shouldn’t the fact that QNX is an RTOS mean that a decent S/W MPEG decoder (i.e.
: as good as a H/W decoder) can be built. It would seem to me that a good S/W
: decoder would be a lot more universal than a driver for a particular brand of
: H/W. I find it difficult to imagine that a 400Mhz+ Pentium machine, can’t do
: smooth decoding of MPEG’s with plenty of CPU to spare on a good RTOS like QNX.
Huh? [let’s ignore the fact the he wanted an ENcoder ]
What does fast context switch time and short interrupt latency have to do with
complicated mathematical processing??? Maybe if we bumped the priority to 31
it would run faster???
And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG video decoding – been
there, done that; 350MHz at 320x200 can barely do 5 fps (s/w only, no h/w).
Granted, this is on a “standard” unoptimized distribution; perhaps with hand-crafted assembler
you might come close.
Heck, it can barely keep up with in-memory frame buffered data (without resorting
to hardware tricks) at 640x480 at 30 fps.
Also, why would people build hardware accellerators (apart from IPO and
venture capital considerations) to do something that s/w could do?
Cheers,
-RK
: –
: A prototype will always be built, whether it is shipped, depends on how much
: money is left in the budget.
–
Robert Krten, PARSE Software Devices; email my initials at parse dot com
Consulting, Systems Architecture / Design, Drivers, Training, QNX 4 & Neutrino
Check out our new QNX 4 and Neutrino books at http://www.parse.com/
+1 613 599 8316 or toll free +1 877 PARSE-SW (727-7379)
Robert Krten wrote:
Huh? [let’s ignore the fact the he wanted an ENcoder >
Yeah… can’t read subject lines 
What does fast context switch time and short interrupt latency have to do with
complicated mathematical processing??? Maybe if we bumped the priority to 31
it would run faster??? >
MPEG decoding (as opposed to encoding), is designed to be non CPU intensive. If
we bump the priority to 29 it won’t be pre-empted while it is decoding and
displaying a frame.
And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG video decoding – been
there, done that; 350MHz at 320x200 can barely do 5 fps (s/w only, no h/w).
Granted, this is on a “standard” unoptimized distribution; perhaps with hand-crafted assembler
you might come close.
Heck, it can barely keep up with in-memory frame buffered data (without resorting
to hardware tricks) at 640x480 at 30 fps.
Hmmm, let’s calculate this one out… 640x480x3x30 == 27 MB/sec. Off the shelf
66 Mhz PCI can do 256 MB/sec. Therefore you are only using 11% of the available
bus bandwidth (this is with PCI, not AGP); hardly seems that one would be
required to do hardware tricks to deliver this kind of throughput (and right
now, no one needs anything beyond 640x480 24bit for MPEG, since this is all most
sources can deliver).
Also, why would people build hardware accellerators (apart from IPO and
venture capital considerations) to do something that s/w could do? >
Simple, because the predominant desktop environment (Windows) is
non-deterministic, and while a 400Mhz Pentium has (on average) sufficient
horsepower to do the job, Windows is not capable of scheduling the CPU such that
it can deliver the cycles evenly enough to insure that each frame is decoded and
displayed in exactly 1/30th of a second.
–
A prototype will always be built, whether it is shipped, depends on how much
money is left in the budget.
Previously, Robert Krten wrote in comp.os.qnx:
And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG video decoding – been
there, done that; 350MHz at 320x200 can barely do 5 fps (s/w only, no h/w).
Interesting… what is NTO 2.1 / Photon 2.0 doing then, since Dan Dodge told the
crowd at QNX’2000 that QRtP could decode DVD (aka MPEG encoded video) on a minimum
of a Pentium II/III 350MHz machine?
Cheers,
Camz.
–
Martin Zimmerman camz@passageway.com
Camz Software Enterprises www.passageway.com/camz/qnx/
QNX Programming & Consulting
“Martin Zimmerman” <camz_AT_passageway_DOT_com@wooga.passageway.com> wrote
in message news:Voyager.000808090657.2544B@wooga.wooga.passageway.com…
Previously, Robert Krten wrote in comp.os.qnx:
And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG
video decoding – been
there, done that; 350MHz at 320x200 can barely do 5 fps (s/w only, no
h/w).
You’re wrong on that one Robert 
Interesting… what is NTO 2.1 / Photon 2.0 doing then, since Dan Dodge
told the
crowd at QNX’2000 that QRtP could decode DVD (aka MPEG encoded video) on a
minimum
of a Pentium II/III 350MHz machine?Not only did he told it, but he delivered >
Conference
and people have the CD had a change to try it as well.
I have a PII 400 and it’s very smooth and as far as I know the player
doesn’t
use special hardware features that some video card have (like the ATI).
Note that on QRTP that takes close to 100% of the CPU, while on
Windows it takes around 40% (because it uses ATI video card capability).
With the ATI Rage128 I have seen a Windows machine use only 20%
of the CPU to play a DVD (Pentium 400).
QRTP is still not capable of using specific features of every video card.
They are doing everything themselves, that represent a huge amount of
work!!!
Cheers,
Camz.–
Martin Zimmerman > camz@passageway.com
Camz Software Enterprises > www.passageway.com/camz/qnx/
QNX Programming & Consulting
On Tue, 8 Aug 2000, Rennie Allen wrote:
And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG video
decoding – been there, done that; 350MHz at 320x200 can barely do 5 fps
(s/w only, no h/w).
Granted, this is on a “standard” unoptimized distribution; perhaps with
hand-crafted assembler
you might come close.
Heck, it can barely keep up with in-memory frame buffered data
(without resorting
to hardware tricks) at 640x480 at 30 fps.
Hmm something must be wrong with your pc or you might have a wery old
gfx card. Had a k6-2 300(uppgraded to 500MHz now) and it could decode in
fullscreen full speed mpeg2/4. Even my old amiga 1200 with 68030 50/mhz
can decode in about 5fps.
On Tue, 8 Aug 2000, Mario Charest wrote:
With the ATI Rage128 I have seen a Windows machine use only 20%
of the CPU to play a DVD (Pentium 400).
Think the ATI Rage Fury and better even have full dvd support.
I only have the old ATI rage PRO and it has some dvd support
motion compensation or what it was called.
“Henrik Berglund SdU” <adb94hbd@mds.mdh.se> wrote in message
news:Pine.GSO.4.21L.0008082241100.21400-100000@legolas.mdh.se…
On Tue, 8 Aug 2000, Mario Charest wrote:
With the ATI Rage128 I have seen a Windows machine use only 20%
of the CPU to play a DVD (Pentium 400).
Think the ATI Rage Fury and better even have full dvd support.
I only have the old ATI rage PRO and it has some dvd support
motion compensation or what it was called.My point was that QRTP does make use of these features.
Both these card don’t have FULL dvd support, the CPU is
still need to do part of the job.
Rob, you have seen the dvdplayer plugin for phplay in RTP.
The QNX software only Mpeg II decoder/dvdplayer does a full
30 fps with PERFECT playback. Yes, this is the best dvdplayer
around for image quality and playback. The RTOS is to credit
for the smoothness of video playback as smoothing is very
predictable. The dvdplayer is a licensed component and not
in the free base package. That is why I have written the mpegsystem
player plugin for phplay. mpegsystem (and the streaming voyager
variant mpegsystemplugin) are in the base package of the current
baseline. This plugin has mpeg II support as of yesterday.
Plays at 75-80 % on my PII 450 but I haven’t optimized some
routines. Note the credit goes to the powerful Xing MMX routines.
This is software only, but if you want hardware dvd,
we will be supporting Sigma Designs Netstream 2000 shortly.
You can download the Sigma Designs package with the
mpegaudio_sd and mpegvideo_sd phplay plugins now, but
mpegsystem_sd and dvdplayer_sd are not ready yet. But
these will be the first hardware decoders. Which brings me
back to my my point.
This was an encoder question and only a nut would try software
mpeg encoding in realtime. This is because you would have to
cut corners on compression quality which defeats the reason
for compressing.
Later,
Jim
“Robert Krten” <nospam@parse.com> wrote in message
news:TlLj5.271702$7o1.7414971@news2.rdc1.on.home.com…
Rennie Allen (> rallen@computermotion.com> ) wrote:
: James Boucher wrote:
:
: > Do you have any preferred cards/platforms?
: > I’d write a driver for a nice PCI card if people
: > wanted it and I could get the specs.
:: Shouldn’t the fact that QNX is an RTOS mean that a decent S/W MPEG
decoder (i.e.
: as good as a H/W decoder) can be built. It would seem to me that a good
S/W
: decoder would be a lot more universal than a driver for a particular
brand of
: H/W. I find it difficult to imagine that a 400Mhz+ Pentium machine,
can’t do
: smooth decoding of MPEG’s with plenty of CPU to spare on a good RTOS
like QNX.Huh? [let’s ignore the fact the he wanted an ENcoder >
What does fast context switch time and short interrupt latency have to do
with
complicated mathematical processing??? Maybe if we bumped the priority to
31
it would run faster??? >And no, a 400 MHz+ Pentium machine can’t keep up with s/w-only MPEG video
decoding – been
there, done that; 350MHz at 320x200 can barely do 5 fps (s/w only, no
h/w).
Granted, this is on a “standard” unoptimized distribution; perhaps with
hand-crafted assembler
you might come close.
Heck, it can barely keep up with in-memory frame buffered data (without
resorting
to hardware tricks) at 640x480 at 30 fps.Also, why would people build hardware accellerators (apart from IPO and
venture capital considerations) to do something that s/w could do? >Cheers,
-RK: –
: A prototype will always be built, whether it is shipped, depends on how
much
: money is left in the budget.–
Robert Krten, PARSE Software Devices; email my initials at parse dot com
Consulting, Systems Architecture / Design, Drivers, Training, QNX 4 &
Neutrino
Check out our new QNX 4 and Neutrino books at > http://www.parse.com/
+1 613 599 8316 or toll free +1 877 PARSE-SW (727-7379)
“James Boucher” <jaboucher@home.com> wrote in message
news:Bzak5.23$c5.3675@news2.rdc1.on.home.com…
Rob, you have seen the dvdplayer plugin for phplay in RTP.
The QNX software only Mpeg II decoder/dvdplayer does a full
30 fps with PERFECT playback.Yes, this is the best dvdplayer
around for image quality and playback.
That’s a bold statement
This was an encoder question and only a nut would try software
mpeg encoding in realtime. This is because you would have to
cut corners on compression quality which defeats the reason
for compressing.
Are you calling ATI nuts They do offer MPEG-II compression
on PII 500Mzh and above. Review are saying that it’s missing
a frame once in a while, but that considering it’s a consumer
product it’s doing a good job.
On Sat, 05 Aug 2000 14:27:45 GMT, “James Boucher” <jaboucher@home.com>
wrote:
Do you have any preferred cards/platforms?
I’d write a driver for a nice PCI card if people
wanted it and I could get the specs.
James
Depends - if it’s for consumers the voodoo 3 3500 TV may be the thing
- it encodes MPEG1 plus it’s supported in Photon 1.14/2 and is the
best choice currently for RTP since it does hardware 3d.
Perhaps QSSL wants to do a driver for this board.
For industrial use, something more along the lines of the more
expensive (>$800 boards) might be more appropriate.
I personally would like to see a driver for the v3 3500 but the
question wasn’t drive by personal desires
“Mario Charest” <mcharest@zinformatic.com> wrote in message
news:8mqdfq$6be$1@inn.qnx.com…
“Henrik Berglund SdU” <> adb94hbd@mds.mdh.se> > wrote in message
news:> Pine.GSO.4.21L.0008082241100.21400-100000@legolas.mdh.se> …
On Tue, 8 Aug 2000, Mario Charest wrote:With the ATI Rage128 I have seen a Windows machine use only 20%
of the CPU to play a DVD (Pentium 400).
Think the ATI Rage Fury and better even have full dvd support.
I only have the old ATI rage PRO and it has some dvd support
motion compensation or what it was called.My point was that QRTP does make use of these features.
I meant does NOT
Both these card don’t have FULL dvd support, the CPU is
still need to do part of the job.
I have this card at home and run it with QNX RTP. If nobody else
does it shortly, I’ll try and do the FM tuner and TV tuner myself.
The card does not do any MPEG capture though, just YUV frames.
You have to do software encoding. This is why the quality of
the MPEG 2 video is so low. I have an iomega Buz, too (RIP)
a great device that has hardware motion JPEG capture. I capture
in avi motion JPEG and use DVMPEG to convert to MPEG 1 or 2
as I wish. I might take the YUV and convert it too MPEG I, I frames only on
the fly.
This could be faster but not space efficient, but might be the lesser of
evils.
Jim
“Nicolae P. Costescu” <nick@qrts.com> wrote in message
news:399180bd.67713586@news.concentric.net…
On Sat, 05 Aug 2000 14:27:45 GMT, “James Boucher” <> jaboucher@home.com
wrote:Do you have any preferred cards/platforms?
I’d write a driver for a nice PCI card if people
wanted it and I could get the specs.
James
Depends - if it’s for consumers the voodoo 3 3500 TV may be the thing
- it encodes MPEG1 plus it’s supported in Photon 1.14/2 and is the
best choice currently for RTP since it does hardware 3d.Perhaps QSSL wants to do a driver for this board.
For industrial use, something more along the lines of the more
expensive (>$800 boards) might be more appropriate.I personally would like to see a driver for the v3 3500 but the
question wasn’t drive by personal desires >
No, I have not used motion compensation, even though I think I could
get another 20% of CPU back. I guess, the issue is that it is
older technology. Motion Comp basically allows you to build B and P
frames in the frame buffer of video ram. You can pass motion vectors
into the hardware and it will build the frames in video ram. I would
rather tackle some new hardware for dvd playback. The first is the
Sigma Designs Netstream 2000. What a great card. Hardware
Mpeg 2, AC3, MPEG audio (not MP3 
) . All Dvd features in
hardware less navigator and MPEG II program stream demux.
The RTP has phplay plugins for video only and audio only now,
but should have Audio/video for first release. Maybe DVD by then.
Cheers,
Jim
“Mario Charest” <mcharest@zinformatic.com> wrote in message
news:8ms111$262$1@inn.qnx.com…
“Mario Charest” <> mcharest@zinformatic.com> > wrote in message
news:8mqdfq$6be$> 1@inn.qnx.com> …“Henrik Berglund SdU” <> adb94hbd@mds.mdh.se> > wrote in message
news:> Pine.GSO.4.21L.0008082241100.21400-100000@legolas.mdh.se> …
On Tue, 8 Aug 2000, Mario Charest wrote:With the ATI Rage128 I have seen a Windows machine use only 20%
of the CPU to play a DVD (Pentium 400).
Think the ATI Rage Fury and better even have full dvd support.
I only have the old ATI rage PRO and it has some dvd support
motion compensation or what it was called.My point was that QRTP does make use of these features.
I meant does NOT
Both these card don’t have FULL dvd support, the CPU is
still need to do part of the job.
\
On Wed, 09 Aug 2000 23:31:40 GMT, “James Boucher” <jaboucher@home.com>
wrote:
I have this card at home and run it with QNX RTP. If nobody else
does it shortly, I’ll try and do the FM tuner and TV tuner myself.
The card does not do any MPEG capture though, just YUV frames.
You have to do software encoding. This is why the quality of
the MPEG 2 video is so low. I have an iomega Buz, too (RIP)
Ahh I see. The files it creates look pretty good to me if you keep in
mind the 320x240 type capture size. Well it would be a boon to the
RTP I think if you were able to do what you say - since probably given
the choice of v. cards for RTP most ppl are getting 3dfx boads for the
3d hardware acc. This would be abother reason to go with them.
The Photon driver needs to be fixed to deal with 1600x1200 75Hz
properly though (now it gets a bright line at the top).
If that’s done then the 3dfx boards would be a good choice to
“standardize” on for new system purchases.
My MPEG 2’s 320x240 off the voodoo 3 have nice quality images, but they
are jerky when played. My Buzz MJPEG’s converted to MPEG 1 or 2 with
DVMPEG are smooth and clear. I have asked a graphics driver guru
about the mode question. I’ll let you know what I hear.
Anyway, an encoder is not high on the list right now, but
I’ll see what I can do.
Later,
Jim
“Nicolae P. Costescu” <nick@qrts.com> wrote in message
news:3992cdc5.3384807@news.concentric.net…
On Wed, 09 Aug 2000 23:31:40 GMT, “James Boucher” <> jaboucher@home.com
wrote:I have this card at home and run it with QNX RTP. If nobody else
does it shortly, I’ll try and do the FM tuner and TV tuner myself.
The card does not do any MPEG capture though, just YUV frames.
You have to do software encoding. This is why the quality of
the MPEG 2 video is so low. I have an iomega Buz, too (RIP)Ahh I see. The files it creates look pretty good to me if you keep in
mind the 320x240 type capture size. Well it would be a boon to the
RTP I think if you were able to do what you say - since probably given
the choice of v. cards for RTP most ppl are getting 3dfx boads for the
3d hardware acc. This would be abother reason to go with them.The Photon driver needs to be fixed to deal with 1600x1200 75Hz
properly though (now it gets a bright line at the top).If that’s done then the 3dfx boards would be a good choice to
“standardize” on for new system purchases.