svn commit: r205311 -
releng/7.3/release/doc/en_US.ISO8859-1/relnotes
John Baldwin
jhb at freebsd.org
Fri Mar 19 13:19:27 UTC 2010
On Friday 19 March 2010 5:20:06 am Hiroki Sato wrote:
> --- releng/7.3/release/doc/en_US.ISO8859-1/relnotes/article.sgml Fri Mar 19
05:40:47 2010 (r205310)
> +++ releng/7.3/release/doc/en_US.ISO8859-1/relnotes/article.sgml Fri Mar 19
09:20:06 2010 (r205311)
Just a couple of notes:
> @@ -151,13 +151,35 @@
> <sect2 id="kernel">
> <title>Kernel Changes</title>
>
> + <para>The &man.acpi.4; subsystem now supports parsing SRAT
> + (System Resource Affinity Table used to describe affinity
> + relationships between CPUs and memory.</para>
This is probably a bit of an overstatement. We only support examining it via
acpidump(8). We do not support using it in the kernel to provide NUMA support
(I have some WIP for this, but it's not even in HEAD yet).
> + <para arch="amd64,i386">The &os; virtual memory
> + subsystem now supports fully transparent use of
> + <application>superpages</application> for application memory;
> + application memory pages are dynamically promoted to or
> + demoted from superpages without any modification to
> + application code. This change offers the benefit of large
> + page sizes such as improved virtual memory efficiency and
> + reduced TLB (translation lookaside buffer) misses without
> + downsides like application changes and virtual memory
> + inflexibility. This can be enabled by setting a loader tunable
> + <varname>vm.pmap.pg_ps_enabled</varname> to
> + <literal>1</literal> and is enabled by default on
> + &arch.amd64;.</para>
Hmm, I thought superpages was actually present in 7.2 as well, the only change
in 7.3 was to enable it by default (which was subsequently reverted). Note
that it is not enabled by default for 7.3, so that part should be updated,
though this entire paragraph is perhaps not relevant for 7.3.
Also, I think this paragraph isn't quite right:
<para>The &man.pci.4; subsystem now supports proxying of PCI
Express MSI/MSI-X (Message Signaled Interrupt) requests and
bus interrupt requests for child devices. This allows child
devices to use MSI/MSI-X interrupts.</para>
My guess is that this is actually describing changes made to the vgapci(4)
driver so that child devices such as drm(4) can use MSI? Also, MSI is not
PCI-e specific, so I would just say 'PCI MSI/MSI-X (...) requests'.
--
John Baldwin
More information about the svn-src-all
mailing list