Crypto collab

  • 8 Replies
  • 212 Views

0 Members and 1 Guest are viewing this topic.

Offline E8Guy

  • *
  • Fractal Friend
  • **
  • Posts: 14
« on: August 08, 2018, 02:59:56 PM »
Hi, I was thinking that it would be cool to render a zoom with GNT over a Recursive InterNetwork Architecture implementation. Stupid? Or do you want to be on the email to GNT & RINA folks?

Offline Fraktalist

  • *
  • Administrator
  • *******
  • Strange Attractor
  • Posts: 916
« Reply #1 on: August 09, 2018, 08:07:27 PM »
Not sure if I understand you correctly..
Are you talking about using a cryptocurrency network to render fractals?
I thought about that too, but I'm kind of convinced that the nubmer of interested users will be far too low to make sense.

Do you have any experience with setting up such a network?
Interested, I'm deep into crypto and totally believe it will have major impact in the next years.

Offline E8Guy

  • *
  • Fractal Friend
  • **
  • Posts: 14
« Reply #2 on: August 10, 2018, 02:14:49 AM »
Yeah, GNT=Golem does renderings of Blender. I just searched FF & saw ppl use Blender so that's cool, I haven't used it. My understanding of Golem is that you borrow computing power from their network, not that you'd have to build a network. I was thinking there could be some expediency by using RINA bc ("I think") you could cycle computing elements like a fartlek or Indian run and then zoom as far as you want. But if you and they both use Blender maybe we're done. Seems like it'd be easy to measure & compare each method how much it would cost to get to say a certain zoom level? & then compare say a farm of tethered smartphones for speed, 90000 ppl?

Offline Fraktalist

  • *
  • Administrator
  • *******
  • Strange Attractor
  • Posts: 916
« Reply #3 on: August 10, 2018, 03:23:49 PM »
I don't think many here use blender, and if they do it's probably for post-editing, not rendering the fractals themselves.
I guess you'd have to integrate at least mandelbulb3d, mandelbulber or kalles fraktaler which are probably the most commonly used.
And I know that ultrafractal, a commercial software allows network rendering spread over several machines.

Sorry to say, I have little background knowledge and won't be able to help in any technical way.

But I do like the idea a lot, and I'd definitely be willing to pay for quick bigger renders as well as mining to render others images when my machine is idle.

Offline E8Guy

  • *
  • Fractal Friend
  • **
  • Posts: 14
« Reply #4 on: August 10, 2018, 06:07:34 PM »
I talked to a RINA guy who said the hard part of distributed computing is 1) timing or bounds on the time when a certain computational outcome will be known, and 2) knowing when computational elements/blocks are equivalent (for scheduling purposes). He described the block equivalence problem as "bleeding edge". I feel like mandel stuff is a great use case though because there will be few different kinds of blocks needed (just one?) so the block scheduler would only need to determine if x block in device X is the one kind of block we need or not - easier.
Ultimately I think we will be able to continuously (&recursively) render the network itself at the point of computation, for practically unlimited zoom. I need to learn more about how the actual computation is done..
Sorry if too technical - your last post is super helpful.

Offline Fraktalist

  • *
  • Administrator
  • *******
  • Strange Attractor
  • Posts: 916
« Reply #5 on: August 10, 2018, 10:32:30 PM »
I read about how a lot of small specialized jobs might in the future get their own cryptocurrency network. It's starting to make sense.

I've had this idea on my mind:
It would make so much sense to somehow store all that computational power that went into millions of images of different locations in the mandelbrot set rendered over the times.
(especially regarding the buddhabrot-set method but thats too far offtopic too soon ;))

Right now all that data is simply discarded, or just stored on single machines. Even if you open a parameter file you already calculated, you'll have to calculate it again.
It feels like such a waste to throw this all away constantly.
It's like using an expensive mathematical telescope for many years and then just discard all the collected data and possible images.

why not  store all these points in the always growing ledger and build a decentralized database of the mandelbrot set. That might be a great "sideffect" of a fractal cryptocurrency used to outsource render power.

the m-set seems indeed ideal for this kind of calculation.
as each coordinate set will always reproduce the exact same result.
lets say you have an image of 1920*1080 pixels. Whatever single pixel you choose, it doesn't matter when or by whom it is calculated.
All you do is insert the coordinates into z->z²+c and see if and when it escapes to infinite, and then send that raw data to the ledger where it is pieced together.
how the one who submitted the task to the network uses these datapoints is up to him.
If you don't know: you can change a lot in the finished render, by fine tuning some parameters. It's where 50% of the artistic side happens imho.
The program of choice probably wouldn't even matter too much, it shouldn't be too hard to use a common language.
AFter all, we're speaking of nothing butthis information:
2 coordinates + the result in how many iterations it took to escape to infinity(or not).

to me as non-technical-non-software guy this looks simple and manageable.

do you have any programming skills and/or intention to actually start such a project?
I spend a third my day with crypto, the second third with fractals and the third with making music for a living (and combine the music with fractals, because that's the best)
If there's way to fuse these together, count me in.  :thumbs:
If you have any fractal questions I'll be happy to answer them and participate in brainstorming.

« Last Edit: August 10, 2018, 11:11:19 PM by Fraktalist »

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 1401
« Reply #6 on: August 10, 2018, 11:27:08 PM »
AFter all, we're speaking of nothing butthis information:
2 coordinates + the result in how many iterations it took to escape to infinity(or not).
I think there are some details to consider before starting the implementation.

It take infinitely many iterations to escape to infinity, but you of course mean "how many iters to go to certain escape radius". So you need also to store the escape radius used. This would allow only the most primitive coloring, you probably also want to store |z| (just before or right after escape, depending on convention) to make the smooth iteration count.  If you want to do distance rendering you also need to store |z'|. This would allow DE and iteration rendering but no orbit traps and related methods, for that you'd need to store the entire orbit. Then you have to store c, but at which precision?

Offline E8Guy

  • *
  • Fractal Friend
  • **
  • Posts: 14
« Reply #7 on: August 11, 2018, 05:14:45 AM »
I read about how a lot of small specialized jobs might in the future get their own cryptocurrency network. It's starting to make sense.

I've had this idea on my mind:
It would make so much sense to somehow store all that computational power that went into millions of images of different locations in the mandelbrot set rendered over the times.

Yeah so instead of dumb preimage hashing you would have "miners" iterating off known "big M" Mandelbrot set-belonging images, trying to find connected areas, prob looks a lot like this(https://duckduckgo.com/?q=model+lightning+mandelbrot+site%3Ariousset.com&t=fpas&ia=web) &just like lightning redirects current when possible branches are known not to lead to ground, a (token) scheduler creates current-cy by redirecting computing blocks (coins) that are synched/scheduled by the token. The network builds itself in the same order that lightning searches for a place to discharge. The difference is once the network unloads we just move the ground or whatever. That's the "IPC" part of RINA & it's the reason you can read the network diagram off the calculation ordering. But yeah you would not have to calculate again  8)

Offline E8Guy

  • *
  • Fractal Friend
  • **
  • Posts: 14
« Reply #8 on: August 13, 2018, 06:05:37 PM »
Already a lot to digest here but https://cardanodocs.com/introduction/