Comment by repiret

I discovered this algorithm a month ago or so, and have been telling everyone I can about it. I think it’s so damn cool.

My present Rust implementation can keep a sliding window sum of f32 at 300M/s.