Inspired by this article, this PR removes most if not all reduce...spread usage.
It would be nice to have performance tests, to test how much the performance improved, but it's late, it's an order of magnitude faster (O(n^2) to O(n)), and I don't wanna code that RN.
Inspired by this article, this PR removes most if not all
reduce...spread
usage.It would be nice to have performance tests, to test how much the performance improved, but it's late, it's an order of magnitude faster (
O(n^2)
toO(n)
), and I don't wanna code that RN.