pouët.net

"I tought everybody knows that" - apparently some people didn't, because ive definately heard a few people (/zealots) say something along the lines of "tbl is so amazing! amigaaaaaaaaa rules! they can do stuff that looks better than pc! pc coders suck, they need some superhighend pixelshaders to make realtime shadows like that and all we need is an amigaaaaaaaaaa!" and so on. :)
but hey, it's good for a laugh, isn't it?

Make that mid-90s, for that effect you need either perspective correct texturing or at least a decently tesselated cube ;)

Kebby: Many of the early AGA demo texturemappers were perspective-correct. They used tricks and stuff to pull it off. :)

Smash: But they can make stuff that looks better than most PC stuff. It just has nothing to do with code.

doom: believe me, if you wanted to render that many animated skyboxes at a res acceptable on a pc, the democompo sizelimit's going to have to go a lot higher than 64mb. :)

smash: it's like that talkshow about wrestling and this guy in the audience crying and stuttering 'It's still real to me dammit!'

A Youtube movie of it to back up my somewhat unclear ranting :D http://www.youtube.com/watch?v=BvTNyKIGXiI

I'm with okkie on this one.

@influence device: yes, it's all about low resolution, a decent atmosphere and a blowing-your-mind-soundtrack... but it's fuckin' ace when watching it for the first time.

and of course there is another type of animation: http://www.pouet.net/prod.php?which=30237

duh...

okkie: :D

ok that sounds interesting didnt they need "a" division per pixel back then if there is a trick for fast perspective texturemapping except this biquadratic/bicubic stuff i really want to know.. that must be really cool.. or was it a hardware trickery?

Seriously, what year are you posting from? Most Amiga demos that you people frequently jerk off to are trying to shoehorn what demos on modern hardware do into a way less "advanced" platform. Don't confuse that with Amiga bashing, I still think that's pretty neat to do, but "better" is hardly the term.

Several Amiga demos use these skybox techniques and at first I couldn't clearly understand what is realtime and what's not and seemed preety impressive at first. Also some Amiga demos they create such an atmosphere that I haven't seen on a PC, I don't why. Even though people bragging that Amiga can do more technically than PC or anything, just make me laugh. Haha, I remember also some demos comparing the C64 with the PC just because C64 demos had smoother scroller than most Windows demos.

mad: one div per pixel ? have you ever heard of Quake ?

mad: Amiga demos today don't do a division per pixel. You typically map every 16th pixel or so correctly and then do linear interpolation between them. You can do the same interpolation between rows (although that's problematic for other reasons), and you're down to one division every 256 pixels. Also, as Kebby suggested you can (dynamically) subdivide faces and just do affine mapping, that's a more elegant way of achieving the same thing, but it's not always faster. You could also precalculate a list of screen+texture coordinates for every frame, which is especially easy with a convex object like a cube. And you can use division tables.

Shifter: It's better because it's more interesting and fun to watch. All those people who voted for Starstruck agree. Starstruck might not have won over Lifeforce and The Popular Demo, but it would still stand out. It's all about good, hand-crafted graphics and design, strong colours and strict moderation in the use of "psychedelic" elements. TBL are very good at that, and most people who make PC demos aren't.

I remeber reading this some (many) years ago:

http://www.exaflop.org/docs/fatmap/

thumbs up for mri.

Good stuff. Affine texturemapping is the kind that isn't perspective-correct, though.

inf: i know the span technique(16 pixel) from quake it really made sense on pc (fpu "parallelization") but you told EARLY amiga demos used perspective correct mapping.. your point with precalculating truly is a point, but i wouldnt call this perspective correct texturemapping. i thought there could be some trick i can use further.. i do not think early cubes were tesselated. but anyway ..

Exactly, it's the people, not the platform that makes those demos so good.

but... but.. but... AMIIIGAAAA! ... ??

AMIIIIGAAAAH indeed!

AMIGAAAAAAAA ?

kus ma!

the_Ye-Ti: Ok, GBAAAAAAAAH!

Integer divisions are slow as hell on 68k processors. It's not about parallelization, just about reducing the number of divisions.

I think whatever method you use to convincingly approximate perspective-correctness will be "perspective-correct" to as high a degree as any other method of approximation. If your affine mapper is sub-pixel and sub-texel accurate, adaptive subdivision can potentially give you the best results, though I doubt it was used much (if at all) in early AGA demos.

GAMEPAAAAHHHH!