This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


[Xen-devel] Re: [RFC] [PATCH] Memory allocation from the nearest node.

To: Yuji Shimada <shimada-yxb@xxxxxxxxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxx>
Subject: [Xen-devel] Re: [RFC] [PATCH] Memory allocation from the nearest node.
From: Keir Fraser <keir.fraser@xxxxxxxxxxxxx>
Date: Tue, 07 Apr 2009 08:50:20 +0100
Delivery-date: Tue, 07 Apr 2009 00:51:16 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <20090407160245.9D90.27C06F64@xxxxxxxxxxxxxxx>
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
Thread-index: Acm3U0VriiId/ThCRY6kf9WQG3av6QAAjpTV
Thread-topic: [RFC] [PATCH] Memory allocation from the nearest node.
User-agent: Microsoft-Entourage/
On 07/04/2009 08:34, "Yuji Shimada" <shimada-yxb@xxxxxxxxxxxxxxx> wrote:

> BIOS sets distance between NUMA node to SLIT table of ACPI. This patch
> uses the value of SLIT table to get distance between NUMA node.
> If ACPI does not have SLIT table, the value of "10" is defined as the
> same node and "20" is defined as the different node.
> I would like ask any comments of developers.
> I will make the modified patch after Xen 3.4 is released.

I'd personally rather initialise a N*N array and then search
next_node[first_node][i], 0<=i<N. Otherwise we turned the O(N) node search
into at least O(N^2), and it looks like node_distance() could hide further
costs. And we'll typically be running that search *per page* allocated. It
could easily add up to significant overhead. If N is big then we could
allocate rows of the array on demand.

 -- Keir

Xen-devel mailing list

<Prev in Thread] Current Thread [Next in Thread>