This was 16 years ago. Did they complete the stuff under Future Work or are there still gains to be had there?
Yep, I believe that most of that work is now complete, although it took a long time. For example, VFS giant lock compatibility was still in the 9 tree, ten years later.
For those that missed them, the FreeBSD 5.x releases (the first with SMPng) were painful…
Not sure about all the details, but one big thing that happened soon after was Matthew Dillon’s Dragonfly fork:
Dragonfly forked off FreeBSD 4 rather than the 5 branch in which the SMPng work took place, so it’s not entirely relevant for this.
The fork took place two years after this paper. For many reasons, but a large part of it was the difficulty realizing the course plotted here. So very relevant, IMO. DFly went back and forked from 4 because 5 was floundering.
Correct, reading my comment again I realize I was very unclear. With “not entirely relevant” I was referring to the code in Dragonfly not being based on the ideas/realizations in the paper but on a prevision version, so anyone wanting to read/trace code shouldn’t expect to find SMPng in Dragonfly.