A study of efficient and precise fractal rendering https://bitgloo.com/projects/happy-fractal
You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Clyne 5e14f88f39 optimize r128 kernel, square function 1 year ago
opencl optimize r128 kernel, square function 1 year ago
LICENSE initial upload 1 year ago
Makefile initial upload 1 year ago
README.md initial upload 1 year ago
main.cpp optimize r128 kernel, square function 1 year ago
r128.h trim down r128 1 year ago



This is a casual project investigating Mandelbrot rendering. It began with the task of rendering the fractal, then achieving a smooth framerate (60+ FPS); now, I am substituting custom data types to allow for deeper zooms.

This project is written in modern C++, using SDL2 for rendering. OpenCL has been adopted for GPU-accelerated rendering.

The r128 library was modified to provide a Q4.124 fixed-point data type. Rendering is slow, but more precise than native double. With Q4.124, we can get very close to the precision of XaoS, which I believe uses 80-bit extended-precision floating-point.

In the future, this may either see optimization for faster and smoother Q4.124 rendering, or a further increase in precision.

Controls, notes

Point your mouse to where you would like to zoom from. Left click to zoom in, right click to zoom out.

The scroll wheel adjusts the zoom speed.

The source code has a BENCHMARK flag, which times an automated zoom to a given point.

To switch the source code between double and Q4.124, change the Float type accordingly and set the appropriate OpenCL kernel (in main(), add or remove the _r128 suffix).