Interesting. This basically implements a parallel algorithm for summation/counting, by calculating a partial sum in each slot, and then merging the results with

Interesting. This basically implements a parallel algorithm for summation/counting, by calculating a partial sum in each slot, and then merging the results with every read. This approach could be applied more generally to values in a commutative monoid, and there are probably other parallel algorithms that could be implemented in a database a similar way.