[chmike]

I would have tried parallelism just by curiosity. Split and spread the computation over multiple cores. If you have n cores you could get close to a factor n increase minus the cost of spreading the data and combining the results. That's an easy optimization right out of the box with go (no assembly required).

[camdencheek]

Oh, we do also heavily parallelize. This blog post is just focusing on the single-core perf.