Hi.
We are trying to set up a patch for non-linear control of various video snippets via midi. (NN, please hold your breath - we can't afford a G4 running Nato ;-)
Machine is an Athlon 700, 128MB PC100 and an AGPx2 Asus Geforce with 32MB DDR RAM. Running Win98/Win2k.
The patch is based on pix_movie and the size of the AVIs is under 5MB and with 160-320 pixels width.
The videos are mapped on squares.
Performance is quite disappointing. One video runs quite smoothly. By adding a second one the …
[View More]framerate already drops into very ugly ranges.
Any traps we're running into with this concept? What are the performance limiters for actions like this?
There's nothing else running and memory should be enough for at least 5-10 AVIs...
Can't imagine that the AGP already chokes on stuff like this, but maybe I'm wrong.
Also we have the feeling that OpenGL performance is a bit low on our machine, even with simple objects.
I read something in the FAQ about copying the opengl32.dll into the PD directory to let GEM know about the driver.
What exactly is GEM/PD looking for? The Nvidia OpenGL-driver has a different name. Should we copy it anyway?
Is there any documentation available for the pix_movie object? Alpha blending, chroma-keying?
Any helpful input would be greatly appreciated.
.loop.
[View Less]
I am not certain why you are having performance problems. I have used 5
pix_movie's on a 450PII with a TNT1 on WinNT with a minimal hit. Memory
really isn't the issue, because the data is just streamed in from the drive.
You could convert the movie to a series of images and use pix_multiimage to
cycle through them. This uses a different mechanism for dealing with the
texture data.
As far as the OpenGL performance, just make sure that you have the newest
drivers from nVidia's web site. If …
[View More]you are just using the ones from the CD,
then they are slow and buggy. You do not need to copy any dll's around.
You should have screaming performance with the GeForce card. To make sure
that you have the OpenGL drivers installed correctly, send a print message
to a gemwin. If the vendor reads Microsoft or the renderer is OpenGL
generic (or software something) then you are doing software OpenGL
rendering...which is where your performance problems are.
As far as pix processing, etc, pix_movie automatically uploads the texture
to the video card, so you can't do things like pix_2grey or pix_threshold.
However, you can set the color and alpha with a color object.
If you have an example patch with data, then I can test it out on my
machine and see what is going on.
Later, Mark
============================
= mdanks(a)stormfront.com
= http://www.danks.org/mark
============================
-----Original Message-----
From: Thomas Loop [mailto:thomas.loop@unibas.ch]
Sent: Wednesday, April 12, 2000 8:59 AM
To: pd-list(a)iem.mhsg.ac.at
Subject: pix_movie performance
Hi.
We are trying to set up a patch for non-linear control of various video
snippets via midi. (NN, please hold your breath - we can't afford a G4
running Nato ;-)
Machine is an Athlon 700, 128MB PC100 and an AGPx2 Asus Geforce with 32MB
DDR RAM. Running Win98/Win2k.
The patch is based on pix_movie and the size of the AVIs is under 5MB and
with 160-320 pixels width.
The videos are mapped on squares.
Performance is quite disappointing. One video runs quite smoothly. By adding
a second one the framerate already drops into very ugly ranges.
Any traps we're running into with this concept? What are the performance
limiters for actions like this?
There's nothing else running and memory should be enough for at least 5-10
AVIs...
Can't imagine that the AGP already chokes on stuff like this, but maybe I'm
wrong.
Also we have the feeling that OpenGL performance is a bit low on our
machine, even with simple objects.
I read something in the FAQ about copying the opengl32.dll into the PD
directory to let GEM know about the driver.
What exactly is GEM/PD looking for? The Nvidia OpenGL-driver has a different
name. Should we copy it anyway?
Is there any documentation available for the pix_movie object? Alpha
blending, chroma-keying?
Any helpful input would be greatly appreciated.
.loop.
[View Less]
>It's normal forthe "obj" directory to be empty; if you recompile Pd it
fills
>with ".o" files.
>
>When Pd quits with a strange error message, I normally just search for the
>message in the source code to find out where things went wrong. If it
isn't
>immediately obvious what happend (or didn't happen) you could just report
>the exact error message to the list and I can probably psychoanalyze it
>right away.
>
>cheers
>Miller
Well, I am not experienced in …
[View More]reading source code, but I simply searched the
words pd tells me, I didn't find anything useful (at least, nothing useful
to me).
Unfortunately, I forgot the slip of paper on which I wrote down the error
msg again. But I remember more or less the exact output:
"Error in loading shared libraries"
then something about cant find __bzero
and then "Broken pipe (32)"
all that in about 4 lines. I am not sure if pd causes the problem. Maybe my
soundcard installation (or sth else) isn't right. [Btw, Should I use ALSA or
the creative opensource emu10k1 one ? Differences?] But should the soundcard
matter at all if I start pd with -nosound? All this about shared libs makes
me think that the problem lies somewhere else, but as I am not experienced
in linux (I only started the whole thing to be able to use jmax and pd), I
cannot locate it.
[View Less]
As Miller mentioned, you need tcl/tk to get started:
http://www.be.com/software/beware/languages/tk.html
OpenGL works fine under Be.
I suspect that Pd/GEM would work really well under Be, if someone spent
the time to do the port... Every time I talked with Intel multi-media
engineers, they wondered why we hadn't ported it yet, since that is what a
lot of them are running.
Considering that Pd/GEM have been ported to a few platforms already, the
hard part is done. You would just need …
[View More]to implement the system specific
functions. Ask if you need more info.
Later, Mark
============================
= mdanks(a)stormfront.com
= http://www.danks.org/mark
============================
> -----Original Message-----
> From: Benjamin Israel [mailto:ibenz@pacbell.net]
> Sent: Sunday, April 09, 2000 2:23 AM
> To: pd list
> Subject: BeOS
>
>
> Miller, Guenter, Mark, . . . anyone. . . anyone:
>
> What would be involved with porting pd to BeOS? It seems to
> me that Be
> is theoretically ideal for an app like pd. Has anyone explored this
> possibility?
>
> Ben,
>
>
>
[View Less]
How does one make the open flag on NT work. The path that gets passed
appears to have '\' replaced by '/', which means it won't work too hot
on Windows machines. I vaguely remember reading something about this in
the past, but heck if I can remember or find where I read it in the
first place. I'm running pd 3.0 if it matters. Thanks in advance.
Hi there !
I have assembled a new ggext test version called
"ggext-0.14pre9"
New features are:
- bugfixes in old objects like envgen, readsf, ...
- some objects to generate filter coefficients
- a complete rewrite of the serialin object, this is the part
that is not really finished (e.g the mouse only works if
it is initialized with gpm)
- other things I really can't remember what they could be good for.
I'm announcing this, because I don't have an idea when I will have to
time to …
[View More]really clean it up. So use at your own risk :)
Ah, yes and I haven't tried to compile it on NT for years,
so please don't shot me, but tell me what's missing.
<ftp://epy.co.at/pub/gige/ggext/ggext-0.14pre9.tgz>
<ftp://iem.mhsg.ac.at/pub/gige/ggext/ggext-0.14pre9.tgz>
.. ah, yes did I already mention that gem-0.83 for linux is
there too ?
Guenter
[View Less]
I have put a fixed version of the SBLive driver on
"ftp://epy.co.at/pub/gige/pd/"
I have not tested midiout yet (hope nothing got broken there).
Attached is are two files midi.h and midi.c, which should replace
the respective files in the emu10k1 directory.
(I you have a recent driver, this might be faster than downloading
the entire driver)
Don't forget to build and insmod the driver :)
.. Now I have to figure out how to submit the fix to creative.
Guenter
HI all,
In light of all the trouble people are having with "resync" messages,
I'm thinking about making "-noresync" the default, at least when running
stereo. The "resync" checking is useful when it works, but it seems not to
work a lot of the time...
I've tried to recreate Guenter's problem with abstractions and "netreceive"...
I couldn't find any problems yet. I tried both "netreceive" and an abstraction
in the same patch, and also a "netreceive" inside an abstraction, no problems.
The …
[View More]next step would be for Guenter to send me the patch he's running that
fails...
Finally, I tried for a while to get MIDI input with an emu10k1 card
(SB Live; although I expect to see the same behavior for the cheaper
"PCI 512". Note that "PCI 128 has a different, older chip and does not have
this problem. It seems that when you're using the emu10k1 driver, select()
doesn't report available incoming MIDI bytes, so Wini's s_linux_midi_io
code never realizes that there's data to read.
This is all under OSS. I think the next step would be to make a bug report
but I can't figure out whether to send to Creative or to whomever provides
the OSS audio drivers in teh Linux kernel.
cheers
Miller
[View Less]
Miller, Guenter, Mark, . . . anyone. . . anyone:
What would be involved with porting pd to BeOS? It seems to me that Be
is theoretically ideal for an app like pd. Has anyone explored this
possibility?
Ben,
Hi,
even for me using the -noresync opt launchin' pd is the only way not
to get the continuous "resync" output line...
I work with w98 first_ed
LørdZøster