- Add an introductory sentence for the multi-threaded pagedaemon entry [1]
- Fix up some minor mistakes in the Intel IOMMU entry - Fix spelling of pkg(8) in the PC-BSD entry Requested by: theraven [1] Discussed with: kib [1]
This commit is contained in:
parent
90b02e2860
commit
e14491e8f8
Notes:
svn2git
2020-12-08 03:00:23 +00:00
svn path=/head/; revision=42081
1 changed files with 10 additions and 9 deletions
|
@ -63,7 +63,7 @@
|
|||
<body>
|
||||
<p>Progress on moving PC-BSD & TrueOS to a "rolling release"
|
||||
is happening quickly. We have implemented our own package
|
||||
repository, fully based on PKGNG, which is updated twice
|
||||
repository, fully based on <tt>pkg(8)</tt>, which is updated twice
|
||||
monthly, and are now hosting dedicated
|
||||
<tt>freebsd-update(8)</tt> systems. In addition to the
|
||||
<tt>9.1-RELEASE</tt> ISO images, we have begun to create a
|
||||
|
@ -161,12 +161,12 @@
|
|||
</body>
|
||||
|
||||
<help>
|
||||
<task>Implements workarounds for chipset errata.</task>
|
||||
<task>Implement workarounds for chipset errata.</task>
|
||||
|
||||
<task>Commit to HEAD after additional testing.</task>
|
||||
|
||||
<task>Rebalance MSI/MSI-X using interrupt remapping unit, also
|
||||
required for x2APIC use on big machines</task>
|
||||
required for x2APIC use on big machines.</task>
|
||||
|
||||
<task>Integrate with the Intel GPU MMU and handle Ironlake and
|
||||
SandyBridge errata for the GFXVTd unit.</task>
|
||||
|
@ -195,12 +195,13 @@
|
|||
</links>
|
||||
|
||||
<body>
|
||||
<p>Based on a prototype change from Jeff Roberson, per-domain page
|
||||
queues and per-domain pagedaemon working threads have been
|
||||
implemented. At the moment, the domains coincide with the NUMA
|
||||
proximity domains, but this is not neccessary and could be
|
||||
improved with further separation to allow more parallelism in
|
||||
the pagedaemon.</p>
|
||||
<p>This project aims to improve scalability of the virtual memory
|
||||
subsystem. Based on a prototype change from Jeff Roberson,
|
||||
per-domain page queues and per-domain pagedaemon working threads
|
||||
have been implemented to enable this. At the moment, the
|
||||
domains coincide with the NUMA proximity domains, but this is
|
||||
not neccessary and could be improved with further separation to
|
||||
allow more parallelism in the pagedaemon.</p>
|
||||
|
||||
<p>The patch is relatively simple, with the most delicate parts
|
||||
being the page laundry and OOM logic, which requires coordination
|
||||
|
|
Loading…
Reference in a new issue