Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!wuarchive!hsdndev!dartvax!eleazar.dartmouth.edu!npw
From: npw@eleazar.dartmouth.edu (Nicholas Wilt)
Newsgroups: comp.graphics
Subject: Re: Question (silly?)
Keywords: raytrace, etc.
Message-ID: <1991Jan16.200924.6784@dartvax.dartmouth.edu>
Date: 16 Jan 91 20:09:24 GMT
References: <79641@unix.cis.pitt.edu> <5427@idunno.Princeton.EDU>
Sender: news@dartvax.dartmouth.edu (The News Manager)
Organization: Dartmouth College, Hanover, NH
Lines: 21

>	We have 480000 pixels, each one of which has a single ray.  According
>	to my book, the fastest sphere intersector I have is around 16 adds,
>	13 multiplies, and one square root, for a total of 30 floating point
>	ops.  Since we are not using a hierarchy, we have 6 * 30 floating
>	point operations, or 180 per ray, per bounce.  Since we trace
>	rays to level k, we have 2^k * as many rays per pixel, or 180 * 2^k.
>	Since we wanted real time (lets assume the video rate of 30fps),
>	then we have 
>		800*600*30*180*2^k = 2,592,000,000 * 2^k floating point ops
>			per second.
>	And this doesn't even take into account shading.

Or color quantization; the problem stated 256 colors.  So you'd have to
quantize on the fly, as well, or figure out a clever way to do it
incrementally.

A PC would have a heck of a time doing the real time display even if the 
frames and color mappings were precomputed and compressed.

--Nicholas Wilt
  npw@eleazar.dartmouth.edu