Since file reading speed is crucial in this task 2 versions of solution is included.
Difference in code is tiny-tiny, one uses mmap that maps the file into memory and another uses plain read (default solution).
In my tests read works faster on the first run (multicore machines), but mmap is faster on subsequent runs as it utilizes RAM and doesn't even need to copy from cache.
I think, if benchmarking is done properly (without giving the app a chance to use a cache), the default reading method should outperform the other one.
Progress chart, just for fun:
