Mandelbrot

[Humus]

The daily demo spam

There's a new demo up on my site doing the maybe quite original idea of rendering mandelbrot sets in a fragment program.
http://humus2.campus.luth.se/~humus/

Enjoy.

 

[nAo]

I remember nvidia dev support did that with dx8 style hw, a fine tesselated mesh and a vertex shader. Quite obvious stuff anyway..

ciao,
Marco

 

[Rodéric]

You want to force to buy a RadeOn 9700 don't you ?

Congratulations about your web site.

I've to say your demos are nice, that's why you have so many hits, not taking into account they are cutting edges

 

[Thowllly]

Ah, mandelbrot fractals. I remember writing a mandelbrot viewer on my Amiga in Blitz basic 2. I drew a fullscreen fractal in 'only' 40 seconds But when I showed it to a friend he were like "That’s no good, it's not even in color!" I had selected a greyscale palette just like Humus. I quickly changed it to a color one, but it was too late to change his first impression. So you just might want to make it somewhat more colourful Humus.

 

[Bambers]

Will run on Radeon 9500/9700 only at this time.

Not suprising but anyway.

 

[OpenGL guy]

Ah, mandelbrot fractals. I remember writing a mandelbrot viewer on my Amiga in Blitz basic 2. I drew a fullscreen fractal in 'only' 40 seconds

Some friends and I had done our first Mandelbrots on our Apple IIe's back in high school (yes, that was a while ago )... a hi-res (280x192 or something) image only took something like a couple hours at a maximum of 20 iterations per pixel

We would change to lo-res (40x24, I think) so that we could find interesting places to zoom in less time.

When I started doing Mandelbrots in college on my Amiga 1000 with a 68020/68881 combo crammed inside, I was using a free Pascal compiler that generated assembly code. The assembly code was completely devoid of optimizations so I hacked up the main loop to use registers and got a rather large speed boost The great thing was that my Amiga 1000 was faster than the Mac II's that were in the lab, and it cost a heck of a lot less as well. Plus, AmigaDOS was multitasking friendly, but MacOS was, in those days, definitely not (Multifinder?? What a joke!).

 

[Humus]

I remember nvidia dev support did that with dx8 style hw, a fine tesselated mesh and a vertex shader. Quite obvious stuff anyway..

ciao,
Marco

Bah! is what I say to vertex shader mandelbrot
Anyway, what this thing really need is higher precision. I would also gain a whole lot from data dependent branching in fragment shaders. Guess I'll have to redo it with a R400 when they come around

 

[Humus]

You want to force to buy a RadeOn 9700 don't you ?

Congratulations about your web site.

I've to say your demos are nice, that's why you have so many hits, not taking into account they are cutting edges

Heh, well, I don't have any interest in boosting ATi's sales. I just enjoy exploring new tech
(3000+ visitors the last 24 hours )

 

[Humus]

Ah, mandelbrot fractals. I remember writing a mandelbrot viewer on my Amiga in Blitz basic 2. I drew a fullscreen fractal in 'only' 40 seconds But when I showed it to a friend he were like "That’s no good, it's not even in color!" I had selected a greyscale palette just like Humus. I quickly changed it to a color one, but it was too late to change his first impression. So you just might want to make it somewhat more colourful Humus.

I could of course do a dependent texture read into some kind of palette texture in the end. Will try that

 

[malcolm]

I have no idea what that is?

 

[Humus]

Normally you access textures in a linear fashion. You specify texture coordinates for each vertex and interpolate it in between for each fragment. A dependent texture read on the other hand means the texture is accessed with custom texture coordinates you calculate for instance in a pixel shader, or in more limited cases directly mapped from another texture (EMBM etc.). In this case I could instead of just outputting the final gray shade use it as a texture coordinate into a 1D texture would would then work like a palette.

 

[Captain Chickenpants]

Hehe very nice

Now what about an IFS fern?

 

[no_way]

er ... Humus, i think he's saying he doesnt know what a Mandelbrot set is.
http://aleph0.clarku.edu/~djoyce/julia/explorer.html

( Nice demo, BTW. are 4k demos coming back in form of PS proggies now ? 8) )

 

[nAo]

Anyway, what this thing really need is higher precision

Mr.David Kirk: Pinky, are you pondering what I'm pondering?

ciao,
Marco

 

[Joe DeFuria]

Mr.David Kirk: Pinky, are you pondering what I'm pondering?

NARF!

 

[no_way]

Humus, ever though of submitting some of your work to Cg contest ? thei have a NV30 based Shuttle PC as a first-place prize. You'll get a t-shirt too http://www.cgshaders.org/forums/viewtopic.php?t=465
( last time winner Ramires obviously hasnt got his NV30 yet )

 

[Humus]

Hehe very nice

Now what about an IFS fern?

Could be cool

 

[Humus]

Anyway, what this thing really need is higher precision

Mr.David Kirk: Pinky, are you pondering what I'm pondering?

ciao,
Marco

Heh, well I suppose Kirk will love to have something to prove how lame "half-measures like 96bit" is. But then I suppose they already have stuff running in their labs that will require more than 96bit already for the official launch.

 

[Humus]

Humus, ever though of submitting some of your work to Cg contest ? thei have a NV30 based Shuttle PC as a first-place prize. You'll get a t-shirt too http://www.cgshaders.org/forums/viewtopic.php?t=465
( last time winner Ramires obviously hasnt got his NV30 yet )

I suppose one need to use Cg then

 

[Rodéric]

Quick unrelated question :
Do the RadeOn 9700 have both a fixed function and programmable pipeline, or only a programmable pipeline ?