One billion row challenge

see https://github.com/gunnarmorling/1brc for an explanation. tl;dr process 1 billion rows as fast as possible.

Disclaimer

I did use 3rd party crates as I only did this for myself to learn some rust and not to actually participate in the challenge

cargo run --bin sequentiell

This a naive sequential take on the challenge without any thoughts wasted on optimization.

cargo run --bin parallel

This is my somewhat optimized take and is obviously much faster.

Some optimizations i did:

split workload in chunks and run in 1 thread per core (hence the bin name)
dont parse to floats, parse to integer and devide by 10 at the end
use &[u8] instead of &str and only parse to &str at the end.
use FxHashMap in threads and combine into BtreeMap to automatically sort it.
use rfind instead of find for finding the index of ; because the city names are most likely longer than the temperature part.
inline functions
activate lto and some other release improvements (see cargo.toml)

runs in around 2.5s to 3s including the file read compared to originally 600s