Hacker News Reader

Hacker News

Comment by memalign | Hacker News Reader

Full thread

memalign

I found this cool tool browsing the website[0] of another popular post on HN today[1].

Here are some cool examples:

Double Pendulum: https://viz.intelligence.rocks/#pendulum

Mandelbrot: https://viz.intelligence.rocks/#mandelbrot (This one is melting my brain)

0: https://jason.today/

1: https://news.ycombinator.com/item?id=31309616

View on HN

Having coded some of the original fractal math research for DeVanney and Mandelbrot, watching that fractal render so many frames per second is a mind blow. We used to have to wait several minutes for a single 256x256 image.

> Having coded some of the original fractal math research for DeVanney and Mandelbrot,

Really? I would love to read any story you have about this!

I've mentioned this on HackerNews before: it was spring of '86 when the fractal research group was finally getting around to playing with the software we'd been using to create our publication images. I was a sr graphics consultant at the Boston University 3D Graphics Lab, and due to my being staff at the graphics lab my user account had unlimited processing privileges. This was necessary to get the fractal renders in any reasonable time, running on the University's IBM 3090 mainframe which the entire university used for practically everything. Not really considering the consequences, I made a short program that would generate a 256x256x256 fractal cube of sin()+cos(), which is the specific formula contribution of mine for the "Beauty of Fractals" book. Well, I should have considered the consequences: I launched the program, and immediately everyone's terminals in the graphics lab froze. Suspicious it was due to my program, I figured it would finish in a minute and everything would be okay. Someone mentioned the larger terminal room downstairs with a few hundred terminals was also frozen. Then I heard someone in the hallways shouting "who the hell is bsln5!" (my username.) In a minute the graphics labs will full of the mainframe admins shouting red faced at me to kill my process. I killed it and was immediately escorted out of the building and told I'd be in a hearing for my malicious use of the mainframe. Well, once everything was figured out, I had my unlimited processing privilege revoked, and a new policy was put in place for high compute research projects. During the evaluation it was determined if let run to completion, the program would have run for 35 years - finishing just last year.

Manually sliding the slider in that example is particularly fun!

Fractint was one of the first programs I discovered with my 1990 high school graduation present, a 386sx16. It took forever. I eventually got the math coprocessor for it and was amazed at how much faster it could calculate. I made frame by frame animations with it and stitching the gif's together with the DOS version of Adobe Premiere, which was at version 1. I was lucky to have a 3.5" floppy, so it came on one less disk than the 5.25" floppy version.

Whoa, love the "hot re-loading" on edit. It's super fun but as there's no undo, kind of live coding with no safety net.

I use Inigo Quilez' Graph Toy as well. I figure combing the two into a pipeline, one to generate interesting looking (implicit surface) functions, and one to visualize them in a voxel system with monte carlo rendering is sort of the viz tool I'm evolving toward. Shader toy requires so much code for rapid prototyping because everyone has to constantly re-invent everything from scratch! It's a unified shading model sans standard lib ;)

I love Inigo’s work too!

Speaking of Graph Toy, I forked it and made a version which supports polar coordinates and which doesn’t require you to solve for y (for example, you can graph y^2+x^2=1):

https://memalign.github.io/m/formulagraph/index.html?f1(x,t)...

Is the physics on the double pendulum accurate. I understand with the conservation of angular momentum and shrinking the moment of inertia it could look like that intellectually, but my instinctive self is unsure.

It’s not. Increase the duration to 20,000 and you see a classic gaming bug of the momentum going haywire eventually. Unsure if it’s anything to do with the environment or the coding. On mobile so can’t really analyze it.

Well, even if the math is correct, the simulation will be lossy if one works with finite precision. And because of the chaotic nature of the double pendulum, small errors will eventually add up to big deviations.

Reply on news.ycombinator.com