Re: passive references

[Taylor R Campbell <riastradh%NetBSD.org@localhost>]

   Date: Fri, 11 Mar 2016 11:07:01 +0100
   From: Joerg Sonnenberger <joerg%britannica.bec.de@localhost>

   On Sun, Jan 24, 2016 at 07:10:42PM +0000, Taylor R Campbell wrote:
   > The idea is that acquiring a reference puts an entry on a CPU-local
   > list, which can be done inside a pserialize read section.

   Is there a specific reason why you are using a list for the per-CPU
   storage and not just an array? Given the expected short list size, I
   wonder if the pointer chasing is really better than just searching a
   short array?

Reader: Single pointer dereference (likely into stack in a recent so
likely cached stack frame) vs searching for a slot in an array.

Writer: Broadcast cross-call + pointer-chasing list search vs
broadcast cross-call + array search.

Engineer: No need to think up front about a fixed limit on the size of
array.

I don't expect a significant difference for reader or writer.