[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Calling vm_object_deactivate_pages for each vm_object_deallocate kil
|
From: |
Sergio López |
|
Subject: |
Re: Calling vm_object_deactivate_pages for each vm_object_deallocate kills performance. |
|
Date: |
Sat, 8 Oct 2011 02:06:04 +0200 |
2011/10/7 Samuel Thibault <samuel.thibault@gnu.org>:
> Perhaps vm_object_deactivate_pages should at least keep the
> vm_page_queue_lock for more than just one iteration. XNU does it for 64
> pages.
>
> Otherwise, it looks to me like a simple and sensible optimization.
Reducing lock contention would surely help. I'm also considering if
the additional work of maintaining secondary queue_chain_t for active
pages in every object could be compensated by the reduction of loops
in vm_object_deactivate_pages.