I was saying "every number can be explained" while giving the performance training but this one may require more time (and 10032 10033 traces) to explain it: sorting 43MB in 68 passes and only 131MB read and write 🤔
Conversation
Polyphase merge has unbalanced run sizes, which was important back when multi-pass sorts were the norm. Hardly comes up with modern hardware, in part due to a quadratic scaling effect. You can still perform very fast sorting with only a tiny amount of memory - which is cool.
That's amazing. Thanks. When I see how much memory Google Chrome requires to show a simple page... we should have missed a step in IT evolution ;)
1