Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!sri-spam!parcvax!rocksvax!sunybcs!ugsudy
From: ugsudy@sunybcs.UUCP (Sudhakar Bharadwaj)
Newsgroups: net.graphics
Subject: Ray Tracing
Message-ID: <418@sunybcs.UUCP>
Date: Wed, 16-Jul-86 00:50:55 EDT
Article-I.D.: sunybcs.418
Posted: Wed Jul 16 00:50:55 1986
Date-Received: Wed, 16-Jul-86 07:13:42 EDT
Organization: SUNY/Buffalo Computer Science
Lines: 25


    I saw a few articles on Ray Tracing.  I am trying to implement
Ray Tracing algorithms on 4 Mercury Zip 32016 Array Processors
connected to a Unix Machine with 5 MC68000's.  The intersection
algorithms will be run on the Mercurys and user I/O, Communications,
and display will be run on the 68000's.  Basically, I will do 1
scanline at a time, checking for intersections, calculations shading,
reflection and refraction values inside the mercurys.  I am planning
to have 1 mercury for preliminary calculations for the intersections,
another for the intersections (maybe even the third one also, or 
maybe the third for cleaning up - shading/reflection/refraction) and
the  fourth for any final calculations (if I can get around to it,
the fourth one will have a frame buffer attached to an auxliary
port of the mercury and the image will be displayed upon completion
of each scanline, that will save the time of trasnfering 1 scanline
of pixel information back to the 68000's).

	Vectorizing Ray Tracing is not that bad, once you know
that you are tracing 1024 x 1024 rays with 1 particular
object, so that the same intersection algorithm can be run.

	This subject does seem to provoke more interest then I
thought it would.  Does anyone out there have any benchmark
figures to show how much a reduction in time an algorithm using
an array processor yields?