pouët.net

topic says it all. when was it made and by whom? was this an allready known technique in the games industry, was it written papers about it before it was used in a demo. im just curious...

This is widely cited as the first paper, I think, but stand to be corrected...

Sphere tracing: A geometric method for the antialiased ray tracing of implicit surfaces
John C. Hart
The Visual Computer 12 (10), Dec. 1996, pp. 527-545

that's interesting

but in realtime and on the demoscene, i can think of.. kinderplomber by Rgba? (easter 2007)
good question really :)

Sphere tracing (aka raymarching) has also been described also as a technique for relief mapping in GPU Gems 2 (though using a 3d texture is really an overkill for a bump mapping related tech, just my 2 cents).

Chapter 8. Per-Pixel Displacement Mapping with Distance Functions

> Second printing, April 2005

nystep: polygons, and fallty is older anyway.
However, that's for gpu marching on pc.. then we have Aura for Laura for PS2, and most likely before that some slow and ugly software rendered stuff I can't remember ;)

Sounds like my first attempt at 3D metaballs around the time of Gamma :)

gonna check out Fallty.

ive never coded a raymarcher before. it seems to me that raymarching is specifically directed towards implicit surface-rendering, but is there a global overhead-saving on scenes that doesnt necessary have implicit surfaces?

p01: marching cubes != raymarching

if you consider most techniques used to accelerate raytracing as raymarching then this was already used long time ago.

Yeah, the lines are pretty blurred -- imho if you traverse a regular grid you're marching along a ray.

But I think rudi means specifically sphere tracing..

rudi: Thank you captain Obvious. I know the difference. My first attempt at 3D metaballs was brute force, fixed step marching of the volume. No marching cube whatsoever.

the first paper i know of using raymarching for distance fields is from 1989, used to render julia sets: Ray Tracing Deterministic 3-D Fractals [John Hart et al, 1989]. Ironically, it's the same guy that years later would mess up with the name of the algorithm and add the unfortunate "sphere" term.

As for the demoscene, I can think of Lattice [2001], which raymarches a sin/cos field (quite like Falty)

--------

So, 1989 offline rendering, 2001 demoscene, as far as i know. Can we beat these numbers?

and with proper bbcode:

offline, 1989: Ray Tracing Deterministic 3-D Fractals

for the demoscene, 2001: Lattice

Ahh yep ofc, that 89 one was the one I was thinking of, honest ;)

thanks for the replies!

p01: ok :D misinterpretation of sentences is (sometimes) my middle name. hehe

iq: cool. i couldnt believe it could be done in 256b.

as a matter of fact by that previous intro, i remember this one too, which is raymarching, but from 2009.

so far some kind of listing of demoscene prods that use raym.:

well, this wasnt about making any listing. continue topic :P

It's been done several times in 256b, and at least twice in 128b.

Yawn, what a boring topic. Let's discuss what prod first used HypnoGlow instead!

just states that GPU sucks, and raymarchin gis nothing new!
yeah, hypnoglow is like blur!!. raymarching are not like used everywhere to impress people today. its so boring. :D

I think 2D constant-step raymarching (for heightfields) is pretty oldschool. 1992? 1994?

yeh, if voxel landscape stuff counts, then surely Panic already had it.

Hypnoglow!

Some of my first raytracers were actually raymarchers, from 1990 or so (nonrealtime, unpublished, doesn't count, blabla..)
Proper intersections with integer math was just slightly too difficult :)

Real men use fixed-point math to do raymarching... ;)