Hacker News new | ask | show | jobs
by theocean154 3802 days ago
rearranged my repo, here is the scalable, pipelined way of doing this, it ends up being an insertion sort with the benefit of small fanout (for lower energy), better scalability (faster due to lower fanout and shorter critical path with pipelining), and maintains O(1) throughput (no fill/drain cycle).

https://github.com/theocean154/dau/blob/master/docs/part_sor...

1 comments

more explanation is in the paper in the comments below, this is just a simple way of visualizing how it works