foo@titan.rice.edu (Mark Hall) (08/29/89)
I just received the April 1989 issue of Computer Graphics.
It has a letter on page 181 which obviously came from
comp.graphics, written by Loren Carpenter. In it, Loren
corrects someone who was putting out misinformation about
the "Genesis Effect" sequence.
The problem is that the original text has been reformatted,
causing much confusion. It appears that Loren is saying the
incorrect info, and then saying the opposite.
I saw the original article here, or I would not be able
to figure out what was going on. The typesetters really need
to preserve the spacing of the original articles if comp.graphics
articles are going to be included in Computer Graphics.
(Kelly, could you make sure the guilty parties are made aware
of the problem?)
- mark
Here is the original article:
Article 2119 of comp.graphics:
>From: loren@pixar.UUCP (Loren Carpenter)
Newsgroups: rec.arts.startrek,comp.graphics
Subject: Re: Star Trek II graphics: "The Genesis Effect"
Keywords: star trek graphics genesis
Message-ID: <2534@pixar.UUCP>
Date: 10 Oct 88 01:02:28 GMT
References: <277@pte.UUCP> <523@epicb.UUCP> <12672@oberon.USC.EDU>
Reply-To: loren@pixar.UUCP (Loren Carpenter)
Organization: Pixar -- Marin County, California
Lines: 52
This message is in need of some commentary...
In article <12672@oberon.USC.EDU> ahoffman@skat.usc.edu (Alan M. Hoffman) writes:
>An interesting sidebar to the "Genesis Effect" article is that all of those
>funky "particle" sequences--the random "flames" and the wave effect--were
>generated using fractical math.
The particle system effects (volcanoes & flames) didn't use any fractal math
at all. They were, however, stochastically modelled (deterministically
repeatable random processes). The fractal code was used on the mountains
and the cooling/molten surface after the flame passed by.
>Fractals, as you know, are the hot new computer technique that "randomly"
>produces realistic-looking mountains, vegetation, coastlines, and
>other real-world pictures.
The "Genesis Demo" was computed in the Winter of 1981-1982. The fractal
programs were an anti-aliased version of those I used in 1979 to make
"Vol Libre".
>The reason each frame takes so long is that EVERY POINT in the array need
>to be calculated in relation to surrounding points, and then a ray traced
>back to the "observer" which determines whether it can be seen, what color
>it is, and what kind of surface it has. Last I heard, they were using a
>version of the Cray 2 called an XMP, designed especially for graphics work,
>to do the number crunching. The final image is printed on a film recorder.
The frames took about 20-40 minutes each on a VAX 11/780 (with fpa). There was
absolutely no "ray tracing" anywhere in there. Rendering was done with the
A-buffer hidden surface method. Frames were computed at 512x480 32-bits
(8 each rgb, plus 8 bits alpha matte). Completed frames (stars + flames, etc)
were shot off a Barco monitor with a high quality 35mm movie camera. The
director liked the "video" effect. Now we use a film recorder.
We borrowed time on a couple of Cray XMP's to compute the character parts of
"The Adventures of Andre and Wally B.". Neither we, nor Lucasfilm, have,
or ever had, a Cray of any flavor. Not that we didn't want one. ;-)
>Wish I could do that on my XT!
You can. It might take a few hours per frame. All you need is a C compiler
and a bunch of SIGGRAPH proceedings. That's how we did it.
>----------------------------------- OOO OOO OOOOO OOOOO ---------------
> Alan M. Hoffman OOO OOO OOO OOO
> ARPA: AHOFFMAN@skat.usc.edu OOO OOO OOO OOO "The Rose Bowl 1989!
>--------------------------------- OOOOOO OOOOO OOOOO -----------------
Loren Carpenter
...{ucbvax,sun}!pixar!lorenksbooth@watcgl.waterloo.edu (Kelly Booth) (08/30/89)
In article <951@brazos.Rice.edu> foo@titan.rice.edu (Mark Hall) writes: > (Kelly, could you make sure the guilty parties are made aware > of the problem?) Done. [This is the first time this was tried. The copy editors aren't comp.graphics readers. Someone will have to develop a set of guidlines for transliterating between net-style and more standard publishing style (the SIGGRAPH newsletter has shorter lines because of the two-column format). I'm sure it can be done right once we get the hang of it.] P.S. A set of "pseudo-domain" addresses is being set up for SIGGRAPH with a number of different userids for various SIGGRAPH functions. Details will be posted to comp.graphics once things are in operation.