Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.javascript > #34056

Re: pre-alpha vector field DLA, with point-and-click attractors...

Show all headers | View raw

On 9/26/2017 9:05 AM, Julio Di Egidio wrote:
> On Monday, September 25, 2017 at 9:31:33 PM UTC+2, Chris M. Thomasson wrote:
>> On 9/24/2017 10:54 PM, Julio Di Egidio wrote:
>>> On Monday, September 25, 2017 at 7:45:00 AM UTC+2, Chris M. Thomasson wrote:
>>>> On 9/24/2017 10:36 PM, Julio Di Egidio wrote:
>>>>> On Monday, September 25, 2017 at 6:56:52 AM UTC+2, Chris M. Thomasson wrote:
>>> <snip>
>>>>>> Can anybody get it to work? Thanks.
>>>>>
>>>>> What do you mean?  Doesn't it work already?
>>>>
>>>> I have not tried it out in all of the browsers yet. Still need to try
>>>> SeaMonkey. It works in FireFox, so it should work there. Also, need to
>>>> try it in Opera. Afaict, it should work fine.
>>>
>>> Well, it's hard to fix or even improve if you do not give the exact
>>> mathematical formula (unless you already have and I am missing it).
>>
>> Working on that. I need to convert the following core logic into a
>> mathematical formula:
>>
>> http://funwithfractals.atspace.cc/ct_fdla_anime_dynamic_test/ct_field.js
>>
>> Think I can get it down to one or two iterative Sigma functions.
>>
>> http://funwithfractals.atspace.cc/ct_fdla_anime_dynamic_test/ct_fdla.js
>>
>> This one is going to be more difficult to turn into pure mathematical
>> formula.
> 
> Sorry, I was a bit quick, but I actually meant that we need a specification
> to work with.  That is not usually as simple as just a mathematical formula,
> anyway I didn't meant to make the conversation too heavy, either.
> 
> That said, I will look into the code when I find some time: might be
> already correct mathematically, but it is very heavy on the CPU.

Still thinking about how to word the specification. It comes in two 
layers, the vector field and the logic for the interactions between 
particles and attractors in said field. The logic layer shows how a 
particle gets cast from a point and computes its vector field line, 
segment by segment, until it hits an attractor or runs out of 
iterations. If it hits, the segment from attractor to previous point in 
the field line is drawn, and the previous point becomes a new attracting 
agent. Well, this can get very computationally intensive! It will burn CPU.

I need to code it a different way wrt:

https://groups.google.com/d/topic/comp.lang.javascript/dgakiMzhgv0/discussion

The garbage collector is going crazy here.


>>> Anyway, it's running in my (up to date) Google Chrome, the only weird thing
>>> I have noticed (running it for few minutes only) is that it seems to get
>>> stuck at the points created manually, i.e. it seems to almost stop growing
>>> as soon as I add points: but maybe I should have waited for longer?  Again,
>>> I just cannot say what exact outcome is expected.
>>
>> Can you please let it run for around a minute then click a couple of
>> points in the square, let it run for around 10 more minutes, take a
>> simple screenshot and post the image? The points you clicked should have
>> clusters by then. This would really help me out!
> 
> Sorry again, it seems to work fine, actually.  Here is the screenshot of a
> run, where I have let it run for 1 minute, then I have added 4 points and
> let it run for 5 more minutes (then stopped as my CPU was over-heating:-):
> <https://drive.google.com/file/d/0BxPmJcnnbWOTZVdKYlJ5NEh0em8/view>

Very nice: Your screenshot is excellent information for me. Thank you so 
much for giving it more time. I really do appreciate it Julio.


>> One more point, your computer is probably faster than mine, so one
>> minute of growth might be different. ;^)
> 
> I am on a high-end laptop with an AMD Radeon on board, nothing more than
> that.  The browser is not taking advantage of the AMD GPU anyway, but
> right away I wouldn't know how to improve on that (or maybe force the
> canvas to use the GPU?).  To be investigated...

Well, unfortunately, if I do not add some method to allow attractors to 
"die off" during the simulation, it will eventually overwhelm a GPU 
and/or CPU.

Wrt, implementing on GPU, I was thinking of a limited form using/abusing 
Uniform variables where the GPU can perform the critical vector field 
summation function. Need to think about how to implement the DLA hit 
logic in GPU. I am not sure its going to like it...

Back to comp.lang.javascript | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-24 21:56 -0700
  Re: pre-alpha vector field DLA, with point-and-click attractors... Julio Di Egidio <julio@diegidio.name> - 2017-09-24 22:36 -0700
    Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-24 22:44 -0700
      Re: pre-alpha vector field DLA, with point-and-click attractors... Julio Di Egidio <julio@diegidio.name> - 2017-09-24 22:54 -0700
        Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-25 12:31 -0700
          Re: pre-alpha vector field DLA, with point-and-click attractors... Julio Di Egidio <julio@diegidio.name> - 2017-09-26 09:05 -0700
            Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-28 21:24 -0700
              Re: pre-alpha vector field DLA, with point-and-click attractors... Julio Di Egidio <julio@diegidio.name> - 2017-09-29 06:47 -0700
        Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-25 12:39 -0700
  Re: pre-alpha vector field DLA, with point-and-click attractors... "Evertjan." <exxjxw.hannivoort@inter.nl.net> - 2017-09-25 22:00 +0200
    Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-25 13:25 -0700
      Re: pre-alpha vector field DLA, with point-and-click attractors... "Evertjan." <exxjxw.hannivoort@inter.nl.net> - 2017-09-25 22:36 +0200
        Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-27 13:38 -0700
  Re: pre-alpha vector field DLA, with point-and-click attractors... Une Bévue <unbewusst.sein@fai.invalid> - 2017-09-26 05:50 +0200
    Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-09-27 13:19 -0700
      Re: pre-alpha vector field DLA, with point-and-click attractors... "Michael Haufe (TNO)" <tno@thenewobjective.com> - 2017-09-30 08:41 -0700
        Re: pre-alpha vector field DLA, with point-and-click attractors... "Chris M. Thomasson" <invalid_chris_thomasson@invalid.invalid> - 2017-10-04 20:48 -0700
          Re: pre-alpha vector field DLA, with point-and-click attractors... "Michael Haufe (TNO)" <tno@thenewobjective.com> - 2017-10-05 08:25 -0700

csiph-web