Precalculating the Mandelbrot set - because it's there

This is mostly a learning exercise for me.

Key things:
  • use of anonymous unions and NAN for efficient storage of the mandel_X.dat files (depite the fact that they are rather large still...) This was my original 'cool idea' / excuse for starting this.
  • use of mmap in a 64 bit address space for basically doing anything. Nice.
  • use of pthreads to utilise multi-core processors
And of course, the key thing:
  • using a computer for it's intended purpose: computing, taking all the resources it possibly can in the process :-)
For the future:
  • distributed parallelism
  • not only incremental in terms of iterations, but also incremental in spatial resolution (which will probably require a change in the file format, or something like a group of 4 mandel_X.dats which are postprocessed to create a mandel_(X+1).dat, by storing an offset of zero or 2^-(X+1) in each of the real and imaginary axes...
  • an actual command line interface, with options and everything

This is unlicensed software, see UNLICENSE for details. Created by Ben Bass 2010-2014 @Ben Bass