K 10
svn:author
V 5
markj
K 8
svn:date
V 27
2018-03-01T18:11:03.543325Z
K 7
svn:log
V 858
Revert changes to batch the insertion of pages into page queues.

It will be replaced by a more general mechanism in a future commit.
The approach in this change has a number of disadvantages:
- It bloats the per-domain structure quite a bit: we keep a batch
  queue per page lock per page queue, for a total of
  PQ_COUNT*PA_LOCK_COUNT queues per domain. We'd like to be able
  to increase PA_LOCK_COUNT without incurring bloat.
- It only improves scalability for enqueue operations; threads which
  wish to dequeue or requeue pages still must acquire the page queue
  lock. Thus, the page queue lock remains a bottleneck in certain
  workloads. Builds, for example, involve frequent removal of pages
  from PQ_ACTIVE as short-lived VM objects are destroyed.
- The page daemon still needs to acquire the page queue lock once
  per page during a queue scan.

END