Hi, I see this output from xm info: # xm info ... release : 2.6.32.21-45.6xen version : #1 SMP Wed Feb 29 23:42:59 EST 2012 machine : x86_64 nr_cpus : 8 nr_nodes : 1 cores_per_socket : 4 threads_per_core : 1 cpu_mhz : 2826 I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_core But the above output is not. Could anyone explain it a little? Thanks, Zhigang
>>> Zhigang Wang <zhigang.x.wang@oracle.com> 04/06/12 2:45 PM >>> >nr_cpus : 8 >nr_nodes : 1 >cores_per_socket : 4 >threads_per_core : 1 > >I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_corenr_cpus = nr_nodes * sockets_per_node * cores_per_socket * threads_per_core Jan
On 04/06/2012 03:52 PM, Jan Beulich wrote:>>>> Zhigang Wang <zhigang.x.wang@oracle.com> 04/06/12 2:45 PM >>> >> nr_cpus : 8 >> nr_nodes : 1 >> cores_per_socket : 4 >> threads_per_core : 1 >> >> I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_core > nr_cpus = nr_nodes * sockets_per_node * cores_per_socket * threads_per_coreIt seems on xm/xl info we don''t show how many sockets_per_node. Can we do that in the future? It seems this machine has: sockets_per_node = 2. Thanks very much Jan. Zhigang
> From: Zhigang Wang > Sent: Friday, April 06, 2012 2:01 PM > To: Jan Beulich > Cc: xen-devel@lists.xen.org > Subject: Re: [Xen-devel] strange cpu number from xm info > > On 04/06/2012 03:52 PM, Jan Beulich wrote: > >>>> Zhigang Wang <zhigang.x.wang@oracle.com> 04/06/12 2:45 PM >>> > >> nr_cpus : 8 > >> nr_nodes : 1 > >> cores_per_socket : 4 > >> threads_per_core : 1 > >> > >> I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_core > > nr_cpus = nr_nodes * sockets_per_node * cores_per_socket * threads_per_core > It seems on xm/xl info we don''t show how many sockets_per_node. Can we do that > in the future? > > It seems this machine has: sockets_per_node = 2. > > Thanks very much Jan. > > ZhigangHi Zhigang -- Unfortunately, the world of CPUs is getting more complicated and can no longer be described so simply. See for example: http://lists.xen.org/archives/html/xen-devel/2009-11/msg00885.html On a Friday afternoon, this makes my head hurt... Dan
On Fri, 2012-04-06 at 21:01 +0100, Zhigang Wang wrote:> On 04/06/2012 03:52 PM, Jan Beulich wrote: > >>>> Zhigang Wang <zhigang.x.wang@oracle.com> 04/06/12 2:45 PM >>> > >> nr_cpus : 8 > >> nr_nodes : 1 > >> cores_per_socket : 4 > >> threads_per_core : 1 > >> > >> I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_core > > nr_cpus = nr_nodes * sockets_per_node * cores_per_socket * threads_per_core > It seems on xm/xl info we don''t show how many sockets_per_node. Can we do that > in the future?We can and should. Are you able to send a patch for xl at least? Thanks, Ian.> It seems this machine has: sockets_per_node = 2. > > Thanks very much Jan. > > Zhigang > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
On 04/07/2012 03:39 AM, Ian Campbell wrote:> On Fri, 2012-04-06 at 21:01 +0100, Zhigang Wang wrote: >> On 04/06/2012 03:52 PM, Jan Beulich wrote: >>>>>> Zhigang Wang <zhigang.x.wang@oracle.com> 04/06/12 2:45 PM >>> >>>> nr_cpus : 8 >>>> nr_nodes : 1 >>>> cores_per_socket : 4 >>>> threads_per_core : 1 >>>> >>>> I thought nr_cpus = nr_nodes * cores_per_socket * threads_per_core >>> nr_cpus = nr_nodes * sockets_per_node * cores_per_socket * threads_per_core >> It seems on xm/xl info we don''t show how many sockets_per_node. Can we do that >> in the future? > We can and should. Are you able to send a patch for xl at least? > >Here is another case: # xm info ... nr_cpus : 48 nr_nodes : 8 cores_per_socket : 12 threads_per_core : 1 This HP BL685 G7 (4 socket x 12 cores). Actually sockets_per_node = 0.5. Xen calculates it correctly. Should we should s 0.5 for sockets_per_node? Or just ignore sockets_per_node? Thanks, Zhigang
On Thu, 2012-05-31 at 09:15 -0400, Zhigang Wang wrote:> > We can and should. Are you able to send a patch for xl at least? > > > > > Here is another case: > > # xm info > ... > nr_cpus : 48 > nr_nodes : 8 > cores_per_socket : 12 > threads_per_core : 1 > > This HP BL685 G7 (4 socket x 12 cores). > > Actually sockets_per_node = 0.5. Xen calculates it correctly. > > Should we should s 0.5 for sockets_per_node? Or just ignore sockets_per_node? >I looked a bit into this and found out there has been some discussion already on why such a "counter" (sockets_per_node) has been removed and shouldn''t be reintroduced, and those archs yielding fractional numbers for it seem to be right that cause: http://old-list-archives.xen.org/xen-changelog/2010-04/msg00074.html http://old-list-archives.xen.org/archives/html/xen-devel/2010-02/msg00005.html Not sure what TheRightThing is here, just wanted to point these out... Regards, Dario -- <<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://retis.sssup.it/people/faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK) _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel