As an exercise for learning Go, I decided to start writing a simple program to generate the Mandelbrot set. While it still needs quite a lot of refactoring, the program still keeps all the methods I have used to generate the set:
- First, simply sequentially computing the pixels and storing everything in a RGBA image.
- Then, receiving a number of routines to launch, split the image in several vertical sections and make each of the goroutines compute a band of the image. Once the corresponding section is drawn, signal the main routine using a channel.
- Eventually, the best results were obtained by creating a channel that works as a job pool and serves each routine an individual line to work on by demand. That way, the workload is distributed equally among all routines.
To run it, use the flag -n to define the number of routines to be launched:
for example: fractal-go -n 8 will create 8 goroutines to query jobs from the
channel. This will generate an 8K image of the fractal, which in my laptop
takes around 30 seconds to compute.