Realtime buddha-like fractal generator

---

Table of contents

1. Gallery
2. Requirements
3. How to use
4. How it works

---

Gallery

Anti-buddha-burningship fractal

Anti-buddha-tricorn fractal

Anti-buddhabrot fractal

Buddha-burningship-power-3 fractal

Buddha-tricorn fractal

Buddha-tricorn-power-3 fractal

Buddhabrot fractal

Requirements

• Linux (tested on Ubuntu 24 LTS)

• Python 3.12.3

pip3 install -r requirements.txt

How to use

The program uses OpenGL to render variety of buddhabrot fractals. There are two main programs:

• generate_once.py

  It generates the fractal once. To choose which fractal you want to render change the PROGRAM variable at the top. It is recommended to use high iteraction count inside shader when using generate_once to get higher quality. To change this go to main() inside the compute shader and modify the iterationCount variable. To get higher quality use numbers > 10_000.

  There is an option to generate transparent images by uncommenting a Transparent image section in generate_once.py script.

• explorer.py

  It to choose the fractal follow the same steps as generate_once. To acheive interactivity it is recommended to use low iterationCount like ~10. explorer was mainly written for buddha_juliaset or antubuddha_juliaset fractals. Use ARROWS on keyboard to change the c uniform which changes the fractal apperance. Current c is printed in the console. The drag & pan features are also available by using mouse.

• cpu.py

  Script mainly used for debugging.

There a different options for buddhabrot.glsl fractal like tricorn fractal or burning ship fractal varians. To change it pick corrent formula in .glsl file.

// Mandelbrot: z = power(z, vec2(2, 0.0)) + c;
// Tricorn: z = power(conjugate(z), vec2(2, 0.0)) + c;
// BurningShip: z = power(cabs(z), vec2(2, 0.0)) + c;

How it works

The program uses compute shaders to generate buddha-like sets nearly realtime. All calculations including sampling the random points take place on GPU. I use gl_GlobalInvocationID variable to generate a unique seed for each GPU worker which is later used in custom random function to generate a pseudo-random number <0, 1>.

To increase the number of samples I use additional loop inside each worker which is controlled by iterationCount variable which controls how many samples each worker will generate.

Feel free to to experiment with layout (local_size_x = 32, local_size_y = 32) in; inside shaders or glDispatchCompute(50, 50, 1) in python scripts to increase the number of samples for your machine.