From: Postmaster++at++casa.es
Date: 10/20/2001 02:15:43
Your message
Subject: info-performer Oct 19 2001
was not delivered to:
Eric.Engelsman++at++casa.eads.net
because:
Router: Unable to open mailbox file PROYECTOS1/EADSCASA mail.box: Server not responding
attached mail follows:
Welcome to the info-performer mailing list DIGEST for October 19 2001
List Archives, FAQ, FTP: http://www.sgi.com/software/performer/
Send Submissions to: info-performer++at++sgi.com
Add/Remove requests: info-performer-request++at++sgi.com
Message Subjects:
Probleme with .iv files
Double Precision & Way to search Archives
Question about mmap and shared arenas
gfx cluster linux multi-process glitches
Re: Problem with clipmaps and glPushAttrib
Re: gfx cluster linux multi-process glitches
******************************************************************************
From: Emmanuel Follin <follin++at++oktal.fr>
Date: Fri, 19 Oct 2001 10:37:52 +0200
Subject: Probleme with .iv files
I have a probleme while loading iv files. Those two files are
exactly the same, only texcoords and texcoordsindex fields are
differents.
The probleme is when I load them in Performer 2.4 with option
PFDBLDR_MESH_MAX_TRIS desable the two vertex list is not in the same
order.
I need to have the vertex list in the same order because I want to do a
multi-texture geometrie with those two files. Can someone explain me why
Performer re-order vertex and how avoid this ?
Thanks answering.
PS: with PFDBLDR_MESH_MAX_TRIS, it is worst because the vertex list
don't have the same size.
--
---------------------------------------------
Emmanuel FOLLIN Email: follin++at++oktal.fr
Ingenieur a OKTAL-Boulogne (F92)
******************************************************************************
From: "Martin, Tammy" <tammy.martin++at++baesystems.com>
Date: Fri, 19 Oct 2001 14:41:34 -0700
Subject: Double Precision & Way to search Archives
Hi all,
I have 2 questions:
1. Does Performer support double precision values? I am trying to implement
a geocentric database. I did see in the archives in 96 Performer did not
support double precision, does it now or do I need to implement the
geocentric database with an origin and an offset?
2. What is a good way to search the Performer mailing list archives in its
current state? Now is just an ftp site. Used to have some search engine to
search archives before....
Thanks for your help,
Tammy
******************************************************************************
From: Ken Sewell <sewell++at++donet.com>
Date: Fri, 19 Oct 2001 17:32:24 -0400
Subject: Question about mmap and shared arenas
From what I've read on the list, an Irix Performer program should be
able to have an arena size of up to 1.7G, is that correct?
My program under performer 2.2, on a machine with 1.6G physical, could
get about 900M for its shared arena. Now running under performer 2.4 on
a machine with 2.0G I can't grab more than 512M. My system limits on
memoryuse are set to 2G and so is the shmmax. I saw a post about one of
2.4's changes being that it mmaps the shared arena, is there a more
detailed explanation somewhere? How do I get past my 512M limit? Below
is the error that performer gives me.
1630 PF Info(2): PFSHAREDSIZE overriding size to
1953125.00 KBytes
1630 PF Info/SysErr(12): pfInitArenas() Could not mmap
/usr/tmp/pfShmemFile.a000PU for size 0x77359400 bytes at 0x0
1630 PF Info/SysErr: pfInitArenas() Could not create shared
arena of 1953125.00 KB; adjusting to 976562.50 KB
1630 PF Info/SysErr(12): pfInitArenas() Could not mmap
/usr/tmp/pfShmemFile.a000PU for size 0x3b9aca00 bytes at 0x0
1630 PF Info/SysErr: pfInitArenas() Could not create shared
arena of 976562.50 KB; adjusting to 488281.25 KB
1630 PF Debug/Resource: pfInitArenas() Allocated 488281KB:
0x422f8000 - 0x5ffce500
Thanks,
Ken.
--
Ken Sewell Defense Research Associates
Project Engineer 1430 Oak Ct., Suite 303
sewell++at++dramail.com Beavercreek, Ohio 45430
937-255-2811 x4377 (voice) 937-255-6663 (fax)
******************************************************************************
From: Ken Lindsay <kl++at++magnus.arc.nasa.gov>
Date: Thu, 18 Oct 2001 12:36:50 -0700
Subject: gfx cluster linux multi-process glitches
hi,
We just got an SGI graphics cluster (very cool machine)
but cannot get cfly to run unless it is set to -m 0
(PFMP_APPCULLDRAW), so no multiprocessing. This may be
the source of the frame glitches we're getting as well.
can some one give me insight into this? is multi
processing in linux not yet ready for prime time? is
there something I can tweak or tune to either split
the procs, or get rid of the frame glitches? If I
free run, the frame rate for performer town is about
60 hz, with frame miss counts running between zero
and 25-30. app time is typically 7-9 msec, draw is
typically 9-12 msec. cpu loads are about 50% of one
cpu (each node has two of course). If I run in frame
lock mode, the frame rate is typically slower, and
we still get a lot of frame misses.
It seems like this machine could run much faster and
smoother than I am seeing it do.
And, for those of you who don't need IRIX, I highly
recommend the cluster. so far.
thanks
ken
_____________________________
Ken Lindsay ) kl++at++email.arc.nasa.gov
NASA Ames Research Center ) QSS Group, Inc.
Bldg 269, Room 281 ) (650) 604 0461 (lab)
Mail Stop 269-1 ) (650) 604 3594 (fax)
Moffett Field, CA 94035 ) (650) 604 3181 (vox)
)______________________
When inlaws are outlawed, only outlaws will have inlaws
******************************************************************************
From: sewell++at++dramail.com
Date: Thu, 18 Oct 2001 13:19:04 GMT
Subject: Re: Problem with clipmaps and glPushAttrib
If I disable my second channel, and its OpenGL calls, then the clipmap works
correctly. The solution I've found for now is to not use the glPushAttrib.
I've gone through all of my OpenGL code and made sure that before I change
an attribute, I call glGet, save it, call the opengl, and then set the
setting to whatever I saved. Not exactly the most elegant solution, but its
working now. Thanks to everyone that replied.
Dorosky, Christopher G writes:
> I have had problems with this before. Down this road lies madness.......
>
> It seems that either performer or GL behaves COMPLETELY differently when a
> cliptexture is on one channel,
> verses when it is on two (or more).
>
> Try cutting down to one channel, and see what happens.
> If this temporarily fixes it, then this is not an easy problem to solve. I
> would investigate throwing out this
> approach and trying something different.
>
>
>
> Christopher Dorosky
> Lead Electronic Systems Engineer - Real Time Simulation
> Lockheed Martin Missiles and Fire Control - Dallas
> christopher.g.dorosky++at++lmco.com
> 972-603-2349
>
>
> -----Original Message-----
> From: Ken Sewell [mailto:sewell++at++donet.com]
> Sent: Wednesday, October 17, 2001 1:06 PM
> To: Performer Mailing list
> Subject: Problem with clipmaps and glPushAttrib
>
>
> I have an application that draws terrain w/clipmap
> in one channel and some OpenGL geometry in the other. In
> the OpenGL draw callback, I call glPushAttrib with the
> texture bit set. When I do this I get a strange effect on
> the clipmap. At the border of every cliplevel, there
> is a stripe of imagery than does not belong. The imagery
> is from elsewhere in the clipmap. After about a second the
> stripe is replaced by the correct imagery. About a second
> later it repeats. My best guess is that the clipcenter is
> being updated but its being drawn before the cliptexture is
> updated. If I don't call glPushAttrib, my terrain shows up,
> but no cliptexture.
> Right now I've stripped down the callback for the
> OpenGL channel, all it does is glPushAttrib, glClear, and
> glPopAttrib and I get this effect. Has anyone encountered
> this problem before? Attached is a screenshot showing the
> cliptexture problem. Thanks.
>
> Ken.
>
> --
> Ken Sewell Defense Research Associates
> Project Engineer 1430 Oak Ct., Suite 303
> sewell++at++dramail.com Beavercreek, Ohio 45430
> 937-255-2811 x4377 (voice) 937-255-6663 (fax)
******************************************************************************
From: Allan Schaffer <allan++at++sgi.com>
Date: Fri, 19 Oct 2001 19:45:13 -0700
Subject: Re: gfx cluster linux multi-process glitches
Hi Ken,
Ken Lindsay wrote:
> We just got an SGI graphics cluster (very cool machine)
> but cannot get cfly to run unless it is set to -m 0
> (PFMP_APPCULLDRAW), so no multiprocessing. This may be
> the source of the frame glitches we're getting as well.
We've been having a look into this from our side also (the support
folks let us know about it) but haven't identified the cause yet.
[Let's keep working on this with them]
Other cluster systems aren't having the problem you've seen.
As far as what to expect, multi-process operation is functional &
fast in Performer 2.4 (and 2.4.2, the version that ships with the
cluster), and there shouldn't be any unusual glitches (though
town.perfly is fill-heavy, certain views do stress the gfx card); we
think that once we figure out the underlying problem these symptoms
should go away.
> And, for those of you who don't need IRIX, I highly
> recommend the cluster. so far.
Thanks. And don't worry we'll keep it that way. :-)
Allan
--
Allan Schaffer allan++at++sgi.com
Engineering Manager, OpenGL Performer 1-650-933-2160
Silicon Graphics http://www.sgi.com
******************************************************************************
This archive was generated by hypermail 2b29 : Sun Oct 21 2001 - 02:36:29 PDT