Coding: Peter Taraba (main contributor), Claude AI (since May 2, 2026)
Fixing Bugs: Claude AI (since May 2, 2026)
Help with installing OpenCL and making it work on Linux: Claude AI
Program Management: Peter Taraba
Unlike subdivision algorithm, this subdivides in all dimensions at the same time. So minor changes in comparison with Approximation of box dimension of attractors using the subdivision algorithm. Now with int64 support, you can theoretically go to ((r=20)+1)*(dim=3)=63<64, which is quite furher than the dimension paper.
Chaotic attractor
pepe@pepe-ms7c90:~/code/pepe78/subdivision_opencl$ python3 main.py
0 - chaotic
1 - rossler
2 - lorenz
3 - aizawa
4 - halvorsen
system = 0
r = 3 , Number of active boxes = 512
Number of testing points = 13824000
Number of active boxes for next step = 1066
r = 4 , Number of active boxes = 1066
Number of testing points = 28782000
Number of active boxes for next step = 2181
r = 5 , Number of active boxes = 2181
Number of testing points = 58887000
Number of active boxes for next step = 5782
r = 6 , Number of active boxes = 5782
Number of testing points = 156114000
Number of active boxes for next step = 15218
r = 7 , Number of active boxes = 15218
Number of testing points = 410886000
Number of active boxes for next step = 41431
r = 8 , Number of active boxes = 41431
Number of testing points = 1118637000 -> Yes, you are testing over one billion points on GPU
Number of active boxes for next step = 112544
(2^6)^3 = 262,144 boxes
(2^12)^3 = 68,719,476,736 boxes
Three Dimensional Black And White displaying of attractors on standard two dimensional colour monitors
This software is for eduction purposes and is free forever (high school + universities) [as well as for Linux community around the world].
Pictures are suitable even for younger kids and can be used in any book for free. No parental guidance needed.
If you are a Windows user, private company or goverment (if you engage in wars), please donate instead.