Received: from watserv1.uwaterloo.ca (watserv1.waterloo.edu) by karazm.math.UH.EDU with SMTP id AA04863 (5.65c/IDA-1.4.4 for ); Sun, 20 Oct 1991 00:47:03 -0500 Received: by watserv1.uwaterloo.ca id ; Sun, 20 Oct 91 01:43:00 -0400 Date: Sun, 20 Oct 91 01:43:00 -0400 From: Dave Stampe-Psy+Eng Message-Id: <9110200543.AA02928@watserv1.uwaterloo.ca> To: glove-list@karazm.math.uh.edu I'd like to try something on for _size_ here, having to do with the number of polys in a low-end VR database versus rendering capacity. This means that mail and postings are solicited, on opinions, facts, or whatever. I'm trying to get a feel for the whole subject, as the VR graphics project seems to be of interest right now. First, I'm looking at the followup to Fundamentals of Interactive Computer Graphics, which is titled Computer Graphics: Principles and Practice by James Foley et al. It seems pretty decent, if you avoid the inevitable IRIS bias. It has good stuff on transformations and a whole chapter on various visible-surface determination algorithms. I think I see a fast way way to do the scan-line algorith that would take less than 25% of the rendering time, and would result in each screen pixel being written to once (i.e 10 frames per second, 320x200x256 colors on a 386/25 IBM PC). Have to check further, though. Now, the main event. I am proposing a top-down clipping sequence for a 10,000 polygon database (world, that is), that goes as follows: database: 10,000 polys, arranged as objects with some extra fields clipping 1: 4,000 polys, by known hidden objects (i.e behind user) clipping 2: 2,000 polys, by quick plane tests to see if in viewport backside: 1,000 polys, by removing polys on back of objects. All this clipping can be done BEFORE converting the polys to viewport (X,Y,depth) coordinate system. Most of this is easily done during the transversal of the database (i.e. movong the world model to the renderer). 1000 polys is a lot to have to draw, but it's within a factor of 3 of what my proposed renderer *may* be able to do at 10 frames/sec. SO... Any comments? Is the 10,000 poly database unrealisticly big? Is my clipping estimate overoptimistic? If so, how many polys SHOULD I count on getting thru to the renderer? All suggestions welcome. - Dave Stampe