Bug: performance falls off a cliff as iterations increase moderately

[mjbaldwin]

The given code examples use 50 iterations, which executes performantly on the GPU, but is only useful for "toy" fractal renderings.

With just a little bit of zoom, you'll quickly be wanting 1,000 or 10,000 iterations (or far more).

Unfortunately GPU compilation of the function then takes minutes and/or runs out of memory completely, due to "loop unrolling". I've commented on the code Fractal Jax uses as a model (which suffers the same issue), and a solution is given here.

I'm not expecting Fractal Jax to be updated, but I just want to leave this issue open as a warning in case anyone else struggles with Fractal Jax performance, and these links explain how to fix it.