Re: Antialiasing and other RE stuff

New Message Reply Date view Thread view Subject view Author view

Lance R. Marrou (marrou++at++vsl.ist.ucf.edu)
Thu, 24 Nov 1994 02:21:46 -0500 (EST)


On Wed, 23 Nov 1994, Allan Schaffer wrote:
...
> This could be a few things. If you only have one RM board in the
> onyx, then you do not have enough framebuffer memory to allocate
> space for multisampling with the standard 1280x1024 VOF.

The Onyx has 4 CPUs and 4 RM boards. I am only using one pipe with
one channel.

> Another possibility has to do with a 5.2 bug where the various
> administration tools will gobble(*) up your framebuffer memory. So
> be sure that 'clogin' and 'chost' are not running at the same time
> you run your Performer application.

Not sure what these are, but they are not running. :)

> And of course there could be some compiled-on-Indigo-running-on-Onyx
> bug that I'm not aware of.
>
> Does antialiasing work in 'perfly' on the Onyx?

Yes, it does. So it must be some problem in either my code or the way it
was linked. Short or recompiling in 5.2 I have tried everything (I think),
so I must hvae missed something.

> On the compiling-with-COFF-under-5.2 topic, there are some IRIX4
> compatible (COFF) libraries that you can load from the IDO disc.
> Once loading the irix4_* subsystems you can set an environment
> variable (SGI_IRIX4) and compile a COFF executable. Check your
> relnotes for the details ('grelnotes IRIX').

Ok, I heard about this, though haven't tried it yet. However, if those
livraries are the same as what I use in IRIX 4.0.5H, it won't be any different.

> Allan
> (*) turkey-day reference for our US subscribers

On Wed, 23 Nov 1994, Jim Helman wrote:

> Two issues:
>
> First, statically linked Performer 1.0 binaries are hardware specific.
> This was necessary since some 4.0.5 platforms completely lacked the
> new RealityEngine GL routines. So a Performer 1.0 binary linked on a
> 4.0.5 Indigo2 will not use any RE features like multisampling.

Well, I am using Performer 1.2 on both machines.

> Second, when running a 4.0.5 binary under 5.X, the multiple processes
> are restricted to a single CPU by IRIX, so you won't get any
> multiprocessing benefit.

This I do not believe. I set the multiprocessing mode explicitly to
APP_CULL_DRAW and it works fine (creating all the separate processes).
So, unless the OS is nailing the other 3 CPUs somehow, it seems to run on
more than one processor because I see all 4 being used. I do not know how
to determine which processors are being used by my program, however. Is
there a sysmp() call which would show this accurately? I'll have to check,
I find this quite interesting. However, if you are right (which you undoubtedly
are :) I guess I will just have to forego the IRIX 4 OS and recompile.

>
> Performer 1.2 is the only version of Performer really supported on 5.2.
> The proper solution is to get Performer 1.2 for the Onyx.

Right-o! :)
>
> rgds,
>
> -jim helman

Something I forgot to mention before which may help pinpoint my problem,
is that not only does antialiasing not work, but the channel stats (via the
pfDrawChanStats() call) do not get displayed correctly. The framerate,
et al, on the top of the rectangular graph do not get displayed. Also, the
machine SLOWS down to about half speed (from a cool 60Hz to not sure what).
Also, points do not get drawn from a normal GL call. I do nothing to set
up the ms buffer, I allow pfInitGfx to do that for me.

Thanks again for your help and have a great Turkey day (and a happy Nov. 24th
to you listeners outside the US :)

_______________________________________________________
IST __ E-mail: marrou++at++vsl.ist.ucf.edu
Visual / / ______ /\____ ______ ______
Systems / / / _ / / __ // ____// ____/
Lab / /__ / /_/ / / / / // /___ / __/_ R. Marrou
________/____//____/\\/_/ /_//_____//_____/____________


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:50:41 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.