Ray Tracing News

"Light Makes Right"

November 18, 1991

Volume 4, Number 3

Compiled by Eric Haines [email protected] . Opinions expressed are mine.

All contents are copyright (c) 1991, all rights reserved by the individual authors

Archive locations: anonymous FTP at ftp://ftp-graphics.stanford.edu/pub/Graphics/RTNews/,
wuarchive.wustl.edu:/graphics/graphics/RTNews, and many others.

You may also want to check out the Ray Tracing News issue guide and the ray tracing FAQ.


Contents:


Introduction

Well, it's been awhile - RealWork (TM) has been getting in the way of putting out an issue of the Ray Tracing News. So, rewind your brains back to August...

SIGGRAPH was interesting, as usual. Las Vegas is an amusing place; now that I've seen it once, I don't ever need to go back. To my surprise, there was quite a turnout for the ray tracing roundtable get together at SIGGRAPH. The roundtable is a nice excuse for people to get in a room and put faces to names, and I finally got to meet some people who had been just authors with email addresses before this.

Some papers of note at SIGGRAPH which directly affect ray tracing were Kirk & Arvo's paper on unbiased sampling techniques and Mitchell's on optimal sampling for ray tracing. The first warns that re-using initial samples results in bias when adaptively supersampling; the last talks of image sampling strategies. Other papers of interest include those on new procedural texturing methods, which all look fairly easy to implement in their simpler forms.

Chen et al presented "A Progressive Multi-Pass Method for Global Illumination", which does about every trick in the book to attempt to achieve maximum realism. Xiao He et al presented "A Comprehensive Physical Model for Light Reflection", which is just that; it seems about the most realistic shading model I've seen, with some very serious mathematics behind it. Another paper from Cornell, "A Global Illumination Solution for General Reflectance Distributions" by Sillion et al, gives an interesting method of storing reflectance functions by using spherical harmonics.

The most theoretically significant radiosity paper was done by Hanrahan et al, who presented a method of limiting the amount of computation by use of hierarchy and error limits. This method opens up interesting new lines of thought and research in radiosity.

I did not spend a lot of time on the floor, but did run across an interesting demo at the Intergraph booth. They had a cute ray tracing program that implemented parameterized ray tracing (Sequin & Smyrl, SIGGRAPH '89), where you essentially store the shading equation parameters for each pixel. Changing colors, applying textures, etc then becomes pleasantly fast, as all you have to do is substitute the proper parameter values and reevaluate, getting a new full ray traced image in seconds.

Other new ray tracing products I noticed were from Ray Dream and Strata. Ray Dream has a ray tracer for the Mac, with the program LightForge for modeling surfaces and SceneBuilder for scene description. They have also added a distributed computing feature to poll Macs on a network for idle CPU time and uses it for rendering. Strata offers StrataVision 3d, again for the Mac. They claim ray tracing and radiosity rendering and gave us a demo disk - the radiosity images are no great shakes, but it's interesting to see the word "radiosity" making its way into the microcomputer market.

AT&T Pixel Machines has been adding radiosity capabilities to their rendering library set. Silicon Graphics is still demoing radiosity, though no product seems in the offing. They did have a good tutorial film showing the ideas behind the progressive radiosity algorithm, and Baum et al had a worthwhile paper in the Proceedings on making radiosity usable. This paper is indispensable for anyone designing a robust radiosity system for general use (i.e. you plan on rendering more that a few axis aligned boxes in a room). HP demoed their radiosity rendering product (ARTCore) with a room designer demonstration, and had a movie in the film show (positive adjectives avoided, since I worked on both projects).

One of the more clever tricks I learnt from the room designer was how to get reasonable wallpaper, floor covering, and other such textures scanned in using a flatbed scanner. In the past I went to building supply places and borrowed or bought samples ("Yes, I want to see how this will look in my kitchen", not mentioning that the kitchen existed only in the computer). However, with a flatbed scanner you can get stuck: the samples can be bigger than the scanning surface. Even if small enough, repetition of the texture can lead to unrealistic effects (for example, a brick pattern is obviously tiled if the brick colors keep repeating in a too regular fashion). I've also tried photographing large areas of a surface (e.g. a brick wall), but then variations in the scene's lighting often appear and make for patterning or odd shading artifacts.

Tamar Cohen, who developed the room designer, realized that there was an excellent solution to these problems: dollhouse supplies! Dollhouse wallpaper and floor coverings easily fit on a flatbed scanner, and all the repetition and lighting problems go away.

For those of you who are deeply into texturing, you should consider looking into the Khoros image processing system (ftp from pprg.eece.unm.edu [129.24.24.10]: /pub/khoros - check release first). It's a huge (~100 Meg) system, but from my minimal exposure seems extremely powerful and easy to use. It has a visual programming language, so you can interactively attach various function boxes together to perform operations. This makes the system easy to quickly start using for simple manipulations, though I think I'm going to have to break down and read the documentation at this point. The system is X based and has been ported to most major workstations on up, and the group at the University of New Mexico are enthusiastic and willing to help. Recommended.

I've also finally scratched the surface of Greg Ward et al's Radiance package. I was impressed first off by the portability: one of his displayers was the first serious X program I've ever compiled and linked without having to diddle around with something to make it go. In fact, I didn't even know it was an X program until I ran it and a window popped up on the screen! If you want physically based rendering, this is the only package I know that even attempts it. It also seems to be a fine renderer, and I enjoy the progressive ray tracing feature (the image refines while you watch it).

As far as speed goes, Rich Marisa at the Cornell Theory Center kindly gave me an explanation and demonstration of their Ray Casting Engine. Duke and Cornell have been developing this piece of hardware for some time, and it embodies an interesting approach: represent the CSG model as a network of processors, then, given a direction of view, convert the model into sets of spans. These spans can then be used for analysis, rendering, etc. For more information, see the Feb. 1991 issue of Mechanical Engineering, or Kedem and Ellis' article in _Parallel Processing of Computer Vision and Display_ (ed. by Dew, Heywood & Earnshaw).

back to contents


New People, Address Changes, etc

# Andy Newton - physical radiance modelling, natural scenes, rays with solid
#       angle
# Remote Sensing Research, University College London
# Photogrammetry,
# UCL, Gower Street
# London, ENGLAND
# +44 71 387 7050 x2742
alias andy_newton       [email protected]
alias andy_newton       [email protected]

Although the graphics is more fun than the Remote Sensing this is what I'm supposed to be doing ...

Applying ray tracing to the understanding of remote sensed images of the natural world, mainly satellite imagery. Much more interested in physical accuracy than efficiency. Also how to correctly model, and sample, very large and non-uniform light sources (the sky!) in ray tracing. How to relate the point sampling paradigm of the infinitesimal ray to light energy transport. Physical reflectance models like BRDF. Doing distance attenuation and variable light source sampling properly (probably) using solid angle.

I'd be really interested in any references anyone has to ray tracing for physical process simulations or radiance calculations using solid angle as a ray property.

On offer: a realistic sky radiance model based on atmospheric scattering.

________

# Denise Blakeley
# 1455 Runaway Bay Dr. #2B
# Columbus, OH 43204
# (614) 487-8442
[email protected]

Ray-tracing interests: general

What I'm doing these days: I'm trying to finish my MS in Computer Science (concentrating in graphics) here at Ohio State December '91. I'd like to finish the program with at least one fairly complete project to show for it, so I'm trying to expand my basic ray-tracer into a more complete rendering system. Nothing ground-breaking; I'm just trying to learn as much as I can at this point, and have fun doing it!

________

# Rick Turner - weird primitives, non-Euclidean raytracing, textures # IBM UK Science Centre # Athelstan House, St. Clement Street, Winchester SO23 9DR, England. [email protected]

I'm a scientist at UKSC, working in the area of remote sensing and the application of image and visualisation techniques to earth science problems. Raytracing is a spare time activity. I've written one raytracer, as well as a substantial part of a second. These use all the common CSG primitives, and for the large tracer (called RT), I've added support for bicubic spline patches (bezier, b-spline, ..., continuous beta-splines) and implicit functions as well. Currently I'm playing around with texture and image mapping and volume objects.

________

Matthew Williams
501 Chapel Drive #1417
Tallahassee, Fl  32304
(904)681-0873
[email protected]

Interests: Anything and everything

At the moment I am a student at Florida State University majoring in Russian Language with minors in Math, Physics, and Computer Science. All the time that I am not in class (including some times when I should be in class) I am on my PC playing around with DKBTrace (or should I say PVRay). One of my larger projects that I want to attempt is translating the C source for DKB to assembly and hopefully gain some speed. I would also like to add a fractal section to it so I can have vines and stuff growing on different objects.

back to contents


ElectroGig Free Software Offer

[I don't know much about GIG, except that they have a CSG ray tracer. Sounds like quite a deal, though! - EAH]

From Communications of the ACM, Nov. 1991:

In an effort to enhance computer graphics education on a national level, GIG USA is offering a limited number of complete 3D graphics packages free of charge to accredited universities, colleges and schools throughout the U.S. The ElectroGIG system, which lists for $30,000, includes retracing [sic - should be ray-tracing] and animation applications and runs on Silicon Graphics and DEC 5000 workstations. Written requests must be mailed (phone calls or faxes will not be accepted) on official school letterhead by staff or faculty members only to GIG USA, Inc., 7380 Sand Lake Rd., Suite 390, Orlando, FL 32819, attention: GIG Educational Software Program.

back to contents


Spectrum: A Proposed Image Synthesis Architecture, by Andrew Glassner ([email protected])

Andrew Glassner is currently working on a proposal called "Spectrum", which is a new ray tracer architecture. The document outlining this design was made available in the "Frontiers in Rendering" course notes. The idea is to make a flexible public domain ray tracer available among researchers and educators.

back to contents


Spline Intersection, Texture Mapping, and Whatnot, by Rick Turner ([email protected]) and Eric Haines

The code that I developed is based essentially on algorithms developed by Kajiya and extended by Marini et al (the paper is in one of the Eurographics procs; I can dig out a reference for you if you're interested). Basically, we model the ray as a pair of orthogonal planes. Each of these is intersected with the spline surface, giving a pair of space curves. You intersect the space curves giving the ray/surface intersection.

Intersecting curves is a manageable problem, so it works quite well. The same basic method is employed for all types of bicubic spline surface. It actually turns out that splines having a constant basis matrix (eg b-spline, power splines, hermites, Catmull-Rom splines etc) are cheaper to compute if you first do a basis transformation to bezier splines. Beta splines and so on that have a variable basis matrix require special treatment, which is quite complex.

I run the code on an i860 based accelerator card to get it to work in reasonable time; a 1024*1024 image with a few hundred primitives takes 5-10 minutes to compute. Spline surfaces can increase this considerably.

The raytracer in question was originally developed by IBM Poughkeepsie as part of a system called GDP (Geometric Design Processor). This was essentially a CSG system and was used to design parts of IBM mainframes. It ran on a mainframe and was written in PL/I and Assembler. The mainframe code was hacked out as a standalone module some years ago, and was then re-written in 'C' in peoples spare time. Most of the basics were done in Burlington, Va., though extensions were done all over the place.

I'm currently working with mapping images and textures onto objects. Yes, I know this is a largely solved problem nowadays, but there are some interesting 'gotchas'. I'm particularly interested in singular mappings, where you don't have a one-to-one correspondence. This overlaps in some ways with my 'real' work, which often involves rendering some earth science dataset. I'm currently fooling around with Magellan data from JPL, rendering combinations of terrain and image data in various ways.

________

My reply:

I know the two-plane intersection method you refer to, in fact I coded it up once long ago (though I don't know the Marini paper - maybe he solves the problem of sometimes converging on the farther intersections instead of the closest? Seems like I remember that guaranteeing the right root is found was a headache, though I recall Kajiya's solution was something like "use Laguerre's method and find them all" or somesuch - I'm probably mixing this up, as I haven't looked at these numerical methods in years...)

As far as texture mapping, that's something I'm still playing with here, too. Solved? Well, how does adaptive sampling work along with textures and mipmaps and so on (mipmaps sample area, but what do you do if more sample rays are shot in a pixel?) - there was a paper on this topic in Eurographics '91, so it's of interest. Also, specifying parameterizations for sets of primitives is easy enough in theory (e.g. define some projection (spherical, conical, plane) in space and use this to determine xyz -> uv), but this kind of thing can look really bad in some cases. I've been playing with other parameterization functions, with some interesting results (my VW bug covered with straw weave is certain to become a style trend soon, I'm sure). Have you run across any interesting parameterization papers/techniques lately?

________

Reply from Rick:

In my code, the patches are subdivided, though not by very much. Each 'mini-patch' has its own bounding box, and both are built into a tree structure. The more curvature the patch has, the greater the subdivision that will be used; this reduces the chances of having a local maximum or minimum in the middle of the patch go outside the bounding box. It also allows you to fit the bounding boxes much more tightly to the surface, so cuts down the number of false intersections where the ray intersects the bounding box but not the surface. One interesting side effect of the subdivision is that by making the bounding box smaller than the surface, you get disconnected pieces of surface floating in space. A nice example of this is on the cover of the Bartels/Beatty/Barsky book on splines. I've done the teapot in a similar way, and it looks rather neat. I went further and mapped the baboon onto each disconnected patch, and it's quite eye catching.

Basically, the convergence method is a hybrid of multi-dimensional conjugate gradient and quasi-Newton techniques. This offers speed plus reasonable stability (although you can always find a pathological case to defeat the algorithms). One thing about it is that the iterations happen in (u,v) space rather than in (x,y,z) space; this ensures that the iterations will always converge to a valid solution; if either u or v go outside the range valid for the particular mini-patch that you're testing against you can immediately reject the solution, as there will be no intersection with that patch. Typically false intersections such as this are rejected on the first (or sometimes the second) iteration, which improves the performance a bit.

Ray tracing splines is tedious though, so I've given my code an option to render the bounding boxes for the mini-patches. This gives a pretty good first look at what you're going to get out, and takes a couple of orders of magnitude less time.

Texture mapping people may want to look through the cartographic literature. Map makers have similar problems in projective geometry when it comes to changing map (or image) coordinate systems - eg, from Lat/Long to UTM for example. Much effort has been devoted to solving the mapping (in the mathematical sense) problem; less on resampling. Almost always the resampling used are standard bi-linear or bi-cubics with the attendant problems. On the whole though, digital cartography can be a useful source on information that is often overlooked by graphics people. A good reference to start with is USGS Professional Paper 1395, 'Map Projections, a Working Manual' by John Snyder. In the US you can get a copy from any Government bookstore - when I got mine I think that it cost me about 25 dollars.

back to contents


Satellite Image Interpretation, by Andy Newton ([email protected])

I work in a satellite image interpretation research group. Our interest in ray tracing is for simulating all parts of the process of the formation of satellite images - optics, camera motion, atmosphere, surface scattering and global (as in hemispherical!) light source illumination. We do work at two scales - where the basic scene is a DEM (height grid) and for very complicated 3-d scenes for plant canopy reflectance simulation.

So ray tracing is a really powerful tool to allow us to simulate as many parts of these complex processes as we can model but what we need to do is not quite computer graphics. What I mean by this is that what we need out of our models are accurate, truthful, energies (in real units) not measures of visual brightness. We need physical results. One example of this is wavelength sampling by importance sampling a spectral response curve as opposed to treating light as an RGB 'colour'. (Though I can't recall any references to doing exactly this it is proposed in Cook's original stochastic sampling papers and my implementation is very similar to his for reflected ray direction).

Our main problems are (i) illumination from such a big light source (the sky hemisphere) with 'diffuse' reflectors (ii) ensuring that we model energies correctly and (iii) using physical directional reflectances. I may be going out on a bit of a limb here so please excuse me if I do - I'm not as familiar as I ought to be with general CG practice - but I'll try to explain what I mean.

I've spent the better part of 3 years supporting and trying to add to a tracer I inherited when I came to work here. Now that has turned out not to be particularly smart as its been a lot harder to modify and bug fix than if it was my own work. Anyway the point is that through out most of that time we had no good model for the directional and spectral variation of that thing outside the window - the sky. As our main interest is _supposed_ to be the satellite work, not the graphics, I felt we couldn't come to grand conclusions from our results if the illumination function was simply not representative of the real world. Now I've managed to solve this problem by implementing a model of atmospheric scattering processes due to Zibordi and Voss so its time to make the physics correct and BTW write a fresh tracer. I have seen some work on CG models of the sky which are functional and not physical. This model is quite fast enough to create a sky radiance LUT at any resolution required on a per scene basis so if you know of anyone out there who needs a model of the sky's irradiance maybe I could help.

The thing about any such illumination model is, of course, that the energy results are per steradian of solid angle of illuminating source. With the point sampling infinitesimally thin ray what solid angle does a ray reaching the sky have? If it's a primary ray then OK it can have some solid angle associated with the pixel but how should ray solid angles be transformed by reflection etc? The only work I've seen that is remotely like this is Amantide's cones. However that doesn't use the geometric concept of a solid angle. One plus point of considering solid angle is of course that the effects of distance attenuation by divergence are implicitly included. So I'm very interested in how much solid angle is used as yet another ray parameter in more general CG work. Is this really common, or unheard of?

There's a reflectance concept in remote sensing called Bi-Directional Reflectance Distribution Function (BRDF), which may also be use in CG or have a parallel, that defines directional reflectance as a 5-d array of pairwise directional spectral reflectance coefficients. For each wavelength (quantized, of course), for each incident direction (over the 2 Pi hemisphere), for each emergent direction, define a reflectance coefficient. Such things can be used as reflectance LUTs or integrated, subdivided and importance sampled. Are similar things done for interesting material reflectances in the main stream?

back to contents


Material Properties, by Ken Turkowski ([email protected])

[I edited Ken's notes into a coherent whole. - EAH]

>Does anyone know where I can pick up a list of material properties for
>different metals and other objects?
>
>I need to know refractive index, diffuse component, specular component and
>specular exponent.

Purdue has a catalog of transmissive, reflective, absorptive and emissive spectral data for conductors, dielectrics, pigments, emulsions, and light sources. From this you can calculate the refractive index and diffuse spectrum. It's called _Thermophysical Properties of Matter_ by the Purdue University Thermophysical properties research center. There are multiple volumes. We have found volumes 6, 7, and 8 most useful. These contain data for dielectrics, conductors, and surface coatings.

Unfortunately, this book is out of print. We got our copy by photocopying one that Purdue had.

Specular data is a function of the finish (i.e. rough, smooth), and can be calculated by the method of He (SIGGRAPH '91) given surface statistics.

Roy Hall's book, _Illumination and Color in Computer Generated Imagery_, points to some other sources of reflective data. The reflective spectrum *is* the diffuse "color". Specular properties are a function on the finish, not the material.

back to contents


New Library of 3D Objects Available via FTP, by Steve Worley ([email protected])

On the ftp site cs.uoregon.edu (128.223.4.13), I have assembled a set of over 150 3D objects in a binary format called TDDD in the directory /incoming/TDDDobjs. These objects range from human figures to airplanes, from semi-trucks to lampposts. These objects are all freely distributable, and most have READMEs that describe them. There are over six megabytes of these binary objects.

In order to convert these objects to a human-readable format, a file with the specification of TDDD is included in the directory with the objects. There is also a shareware utility called TDDDconv that will convert the binary objects into either OFF, NFF, Rayshade, or vort objects. This utility is also found on cs.uoregon.edu, in the file /incoming/TDDDconv.tar.Z.

[There are some interesting things here. You might have to diddle a bit, and I noticed that some databases don't translate, but good stuff for the price! One very cute thing in the package is "tddd2ps", which "converts" a TDDD file to a printable set of four orthogonal views - nice touch! - EAH]

back to contents


Object Oriented Ray Tracing Book

> I am looking for a book named "Object-oriented ray_tracing" by Melcher,
> and published by Wiley 91.

This is, in fact, an article by Karl Melcher and G. Scott Owen, more fully entitled "Object Oriented Ray Tracing: A Comparison of C++ Versus C Implementations" which will appear in a Wiley book early in 1992 (and which was in the Wiley booth at SIGGRAPH '91). The title of the book will be "Computer Graphics Using Object-Oriented Programming" and the editors are Cunningham, Craighill, Fong, and Brown.

back to contents


New and Updated Ray Tracing and Radiosity Bibliographies

At weedeater (see the header of this issue) via anonymous FTP are a number of new or updated ray tracing and radiosity bibliographies. I've updated the ray tracing bibliography (pub/Papers/RayBib.09.91.Z) and radiosity bib (RadBib.09.91) from last year's version. Rick Speer has provided a postscript (only) version of his extensively cross-referenced ray tracing bibliography (speer.raytrace.bib.ps). Tom Wilson's fine ray tracing abstract collection is also available, with June 1 being the latest version (rtabs.6.91.shar.Z). Also, the file "NetPapers" lists a number of worthwhile articles and theses available on the net and where to get them.

back to contents


DKBTrace 2.12 Port to Mac, by Thomas Okken ([email protected])

The public-domain raytracer DKBTrace, which runs on FPU-equipped Macs, has been made available for anonymous ftp from "alfred.ccs.carleton.ca", files /pub/dkbtrace/dkb2.12/other_ports/MacPort1.0.2.*.

back to contents


Graphics Gems II Source Code

FTP from:

weedeater.math.yale.edu [130.132.23.17]
gondwana.ecr.mu.oz.au [128.250.1.63]

file: pub/GraphicsGems/GemsII/GGemsII.tar.Z

back to contents


Radiance Digest Archive, by Greg Ward ([email protected])

I have just made back issues of the Radiance Digest available from anonymous ftp at hobbes.lbl.gov (128.3.12.38) in the pub/digest directory. Those of you who have limited network access can still ask me to send back issues to you directly.

back to contents


Model Generation Software, by Paul D. Bourke ([email protected])

[Paul has a facet based modeller for the Mac called VISION-3D, which can be used to generate models directly in the RayShade and Radiance file formats. He wrote telling me of other programs that might be of interest - EAH]

I have some other "niche" model generators that also export to Radiance and RayShade.

A brief description of some of them follows:

FracHill - generates the old fractal landscapes using the spatial subdivision technique (ie: not the fourier method) It has all the usual settings for roughness, sea level, seed, land/sea colour, etc

3D LSystems - allows the user to generate 3D LSystems (0L). Uses all the standard symbols from the literature, an extension of my 2D LSystem which I wrote years ago.

Triangulate - takes a set of randomly distributed samples on a surface and generates either a triangulated (Delaunay) of gridded mesh representing the surface. We use it for our landscape Architecture course. Note: surfaces (functions) only, not solids!

Anyway, these and other applications can be obtained from my FTP directory [email protected] (130.216.1.5) located in the architec directory. Because we pay for FTP to the US people should be asked to FTP the README file in the above directory, it will inform them of alternative sites in the US.

back to contents


Rayshade 4.0 Release, Patches 1 & 2, and DOS Port, by Craig Kolb and Rod Bogart ([email protected])

Rayshade 4.0 is now available. This version is extremely different from 3.0, and is very different from 4.0beta.

Rayshade 4.0 features include:
        + Eleven primitives (blob, box, cone, cylinder, height field,
                plane, polygon, sphere, torus, flat- and Phong-shaded triangle)
        + Aggregate objects
        + Constructive solid geometry
        + Point, directional, extended, spot, and quadrilateral light sources
        + Solid procedural texturing, bump mapping, and
                2D "image" texture mapping
        + Antialiasing through variable-rate "jittered" sampling
        + Arbitrary linear transformations on objects and texture/bump maps.
        + Use of uniform spatial subdivision or hierarchy of bounding
                volumes to speed rendering
        + Options to facilitate rendering of stereo pairs
        + Rudimentary animation support and motion blur
        + Numerous bug fixes and syntax changes

Apologies to all the folks who felt that their Rayshade 4.0beta questions were not handled in a timely fashion. Both Rod and Craig have had to deal with Real Life and did not have as much time for Rayshade as we had hoped. We still feel that Rayshade is the best Un*x raytracing package for the price.

Rayshade 4.0 is available via anonymous ftp from weedeater.math.yale.edu (130.132.23.17) in pub/rayshade.4.0. The shar files will be posted to alt.sources and submitted to comp.sources.misc.

________

Patches 1 and 2 to rayshade 4.0 are now available through anonymous ftp from weedeater.math.yale.edu (130.132.23.17) as pub/rayshade.4.0/patches/patch{1,2}. The patches have also been posted to comp.sources.misc.

________

Tom Hite managed to port rayshade to DOS, and was kind enough to send me a set of diffs, a couple of configuration files, and a short note describing what one needs to do in order to coax rayshade into running on PCs.

I haven't had the courage yet to find myself a PC and to verify that Tom's instructions are idiot-proof. In addition, Tom's diffs were for rayshade 4.0 patchlevel 0, and the new patches will undoubtedly cause some minor problems when it comes to applying Tom's diffs.

The files are available from weedeater.math.yale.edu (130.132.23.17) as pub/rayshade.4.0/raydiffs.dos.

back to contents


RayShade Timings, by Craig Kolb

Below for your amusement(?) are timings for the latest version of rayshade running on an HP-730.

I suspect that rayshade is a good bit less efficient than it used to be, but I have yet to actually put this suspicion to the test.

Rayshade v4.0, patchlevel 1, on an HP-730 running HPUX-8.05, ?? MB memory. Wed Oct 9 13:50:29 EDT 1991

         Setup       Total     |  Polygon   Sphere    Cyl/Cone    Bounding
             (seconds)         |   Tests     Tests      Tests    Box Tests
--------------------------------------------------------------------------
balls     3.18       116.24    |    356K     1564K        0        2763K
gears    15.79       705.25    |   8345K       0          0       11260K
mount     6.13       165.03    |   1035K     2096K        0        2991K
rings     5.44       235.37    |    103K      206K      4883K      5536K
teapot   13.49       126.43    |   1677K       0          0        2761K
tetra     2.96        31.93    |    578K       0          0         694K
tree      4.94       103.28    |    716K       16K       366K      1467K
All timings are sum of user and system time. Setup includes time to read the database and initialize all appropriate data structures. Total time is setup time plus rendering time. Test figures are rounded upwards.

Uniform spatial subdivision, employing 22^3 voxels, is used to accelerate rendering. In the balls, gears, and tree databases, the ground plane is moved outside of the 22^3 grid in an attempt to generate a more uniform object distribution within the grid.

back to contents


RayShade vs. DKBtrace Timings, by Iain Dick Sinclair ([email protected])

> In your posting on RayShade vs. DKBtrace you mention doing timings
>on them using the SPD. Any chance you have the timings sitting around?

A friend here did the comparison, but it was by no means thorough -- it only used one benchmark (the balls?). In any event, the results of his quick experiment seem to have been discarded (apparently it was a slight pain to translate NFF -> DKB's verbose input format). I seem to remember him saying that Rayshade completed the scene about 20% quicker. Unfortunately, SPD wasn't used exhaustively, though it may be in the near future.

back to contents


PVRay Beta Release, by David Buck ([email protected])

The freely distributable raytracer PVRay (Persistence of Vision) is available for BETA testing from alfred.ccs.carleton.ca (134.117.1.1) in the directory pub/dkbtrace/pvraybeta. This program has been developed by the "Persistence of Vision" group on CompuServe and is built on top of DKBTrace version 2.12 with my permission and blessing.

Please note that this is a BETA release, so it may exhibit some bugs or portability problems to different platforms. Please refer any problems to Drew Wells at [email protected]. Also, this version does not contain some of the ports which were previously available for DKBTrace. This situation is being rectified. However, you should find that the product-specific modules developed for DKBTrace should be easily adaptable to PVRay.

Program Synopsis:
-----------------

PVRay is a Freely Distributable raytracer built on top of DKBTrace. New additions include:

   - Bezier surfaces
   - Height Fields (GIF only at this time)
   - Bump maps
   - Improved Quartic surface support
   - Input language can now optionally accept lowercase keywords
   - New textures: Onion, Leopard
   - C-style comments accepted (  /* */ and //)
   - Algebraic surfaces
   - Sphere, Cylinder, and Torus image mappings

Please direct inquiries to Drew Wells at [email protected].

back to contents


Vort 2.1 Release, by Eric H. Echidna

        gondwana.ecr.mu.oz.au [128.250.1.63] pub/vort.tar.Z
        munnari.oz.au [128.250.1.21] pub/graphics/vort.tar.Z
        uunet.uu.net [192.48.96.2] graphics/vogle/vort.tar.Z
         (uucp access as well ~ftp/graphics/vogle/vort.tar.Z)

Australian ACSnet sites may use fetchfile:
        fetchfile -dgondwana.ecr.mu.oz pub/vort.tar.Z

The major changes are to the ray-tracer which now allows orthographic projections, lights to be in composite objects, provides a transform operator, a few other odds and sods plus the usual set of bug fixes. There are also a couple of utilities for starting art up via inetd to help simplify the generation of animations across networks.

It runs on IBM PC's, VMS, and a variety of UNIX boxes.

Contributed scene files for the ray-tracer can be found in contrib/artscenes on gondwana. Apart from the scene files the tar files in this directory also includes some useful tile patterns and geometry files.

Anyone with anything they'd like to add is welcome to put it in gondwana's incoming directory and send us mail.

Includes [among much else]:

art - a ray tracer for doing algebraic surfaces and CSG models.

back to contents


BRL-CAD 4.0 Release, by Michael J. Muuss and Glenn M. Gillis

The U. S. Army Ballistic Research Laboratory (BRL) is proud to announce the availability of Release 4.0 of the BRL-CAD Package.

The BRL-CAD Package is a powerful Constructive Solid Geometry (CSG) based solid modeling system. BRL-CAD includes an interactive geometry editor, a ray tracing library, two ray-tracing based lighting models, a generic framebuffer library, a network-distributed image-processing and signal-processing capability, and a large collection of related tools and utilities. Release 4.0 is the latest version of software which has been undergoing continuous development since 1979.

The most significant new feature for Release 4.0 is the addition of n-Manifold Geometry (NMG) support based on the work of Kevin Weiler. The NMG software converts CSG solid models into approximate polygonalized boundary representations, suitable for processing by subsequent applications and high-speed hardware display.

BRL-CAD is used at over 800 sites located throughout the world. It is provided in source code form only, and totals more than 280,000 lines of "C" code.

BRL-CAD supports a great variety of geometric representations, including an extensive set of traditional CSG primitive solids such as blocks, cones and tori, solids made from closed collections of Uniform B-Spline Surfaces as well as Non-Uniform Rational B-Spline (NURBS) Surfaces, purely faceted geometry, and n-Manifold Geometry (NMG). All geometric objects may be combined using boolean set-theoretic operations such as union, intersection, and subtraction.

Material properties and other attribute properties can be associated with geometry objects. This combining of material properties with geometry is a critical part of the link to applications codes. BRL-CAD supports a rich object-oriented set of extensible interfaces by means of which geometry and attribute data are passed to applications.

A few of the applications linked to BRL-CAD include:

*) Optical Image Generation (including specular/diffuse reflection,
        refraction, multiple light sources, and articulated animation)
*) An array of military vehicle design and evaluation V/L Codes
*) Bistatic laser analysis
*) Predictive Synthetic Aperture Radar Codes (including codes due to ERIM)
*) High-Energy Laser Damage
*) High-Power Microwave Damage
*) Weights and Moments-of-Inertia
*) Neutron Transport Code
*) PATRAN [TM] and hence to ADINA, EPIC-2, NASTRAN, etc.
        for structural/stress analysis
*) X-Ray image calculation

BRL-CAD requires the UNIX operating system and is supported on more than a dozen product lines from workstations to supercomputers, including: Alliant FX/8 and FX/80, Alliant FX/2800, Apple Macintosh II, Convex C1, Cray-1, Cray X-MP, Cray Y-MP, Cray-2, Digital Equipment VAX, Gould/Encore PN 6000/9000, IBM RS/6000, Pyramid 9820, Silicon Graphics 3030, Silicon Graphics 4D ``Iris'', Sun Microsystems Sun-3, and the Sun Microsystems Sun-4 ``SparcStation''. Porting to other UNIX systems is very easy, and generally only takes a day or two.

You may obtain a copy of the BRL-CAD Package distribution materials in one of two ways:

1. FREE distribution with no support privileges: Those users with online access to the DARPA InterNet may obtain the BRL-CAD Package via FTP file transfer, at no cost, after completing and returning a signed copy of the printed distribution agreement. A blank agreement form is available only via anonymous FTP from host ftp.brl.mil (address 128.63.16.158) from file "brl-cad/agreement". There are encrypted FTP-able files in several countries around the world. Directions on how to obtain and decrypt the files will be sent to you upon receipt of your signed agreement. One printed set of BRL-CAD documentation will be mailed to you at no cost. Note that installation assistance or telephone support are available only with full service distributions.

2. FULL SERVICE distribution: The Survivability/Vulnerability Information Analysis Center (SURVIAC) administers the supported BRL-CAD distributions and information exchange programs for BRL. Full service distributions cost US$500, and include a copy of the full distribution materials on your choice of magnetic tape media. You may elect to obtain your copy via network FTP. One printed set of BRL-CAD documentation will be mailed to you. BRL-CAD maintenance releases and errata sheets will be provided at no additional charge, and you will have access to full technical assistance by phone, FAX, letter, or E-mail. Agencies of the U.S. Federal Government may acquire the full service distribution with a simple MIPR or OGA funds transfer.

For further details, call Mr. Glenn Gillis at USA (410)-273-7794, send E-mail to ([email protected]), FAX your letter to USA (410)-272-7413, or write to:

        BRL-CAD Distribution
        SURVIAC Aberdeen Satellite Office
        1003 Old Philadelphia Road
        Suite 103
        Aberdeen MD  21001  USA

Note that USA area code 410 will not go into effect until 1-Nov-91. Prior to that date, please use area code 301.

Those sites selecting the free distribution may upgrade to full service status at any time. All users have access to the BRL-CAD Symposia, workshops, user's group, and E-mail mailing list.

Sincerely,

Michael J. Muuss                        Glenn M. Gillis
Advanced Computer Systems               SURVIAC
Ballistic Research Lab                  Aberdeen Satellite Office

back to contents


Eric Haines / [email protected]