1BRC-zig

This is based on the challenge here: https://github.com/gunnarmorling/1brc/.

The One Billion Row Challenge (1BRC) is a fun exploration of how far modern Java can be pushed for aggregating one billion rows from a text file. Grab all your (virtual) threads, reach out to SIMD, optimize your GC, or pull any other trick, and create the fastest implementation for solving this task!

I decided to implement this in Zig as a learning experience.

Run

First, you'll need a Zig binary from here: https://ziglang.org/download/. I'm using the latest master version: zig-0.12.0-dev.2059.

Then run zig build -Doptimize=ReleaseFast to build.

If you don't already have a measurements.txt file available, run ./zig-out/bin/run-create-sample 1000000000 to build your ~12GB input file. This can take several minutes.

Now run time ./zig-out/bin/1brc-zig measurements.txt.

With a warm cache (as allowed by the challenge rules), I get this time on my AMD Ryzen 7 5800H:

Executed in    5.76 secs    fish           external
   usr time   37.81 secs  129.00 micros   37.81 secs
   sys time   10.11 secs  526.00 micros   10.11 secs