Messages in this thread | | | Date | Thu, 18 Sep 2014 09:32:24 +0200 | From | Borislav Petkov <> | Subject | Re: [PATCH] x86: Consider multiple nodes in a single socket to be "sane" |
| |
On Wed, Sep 17, 2014 at 02:55:26PM +0200, Borislav Petkov wrote: > This sounds misleading to me. If I would have to explain how I > understand physical_package_id, I'd say it is the physical piece of > silicon containing the core. Which is consistent with what Peter says > that using it to identify NUMA nodes is wrong. > > Btw, I'm trying to get on an AMD MCM box to dump those fields but it is > kinda hard currently. Will report back once I have something...
Ok, so Brice sent me some AMD MCM data from a 4-socket box. physical_package_id there really denotes the physical package, i.e. silicon, i.e. physical socket which contains a core.
So on a 4-socket, 16 cores on each machine, you have:
physical_package_id cpu0-15 : 0 cpu16-31: 1 cpu32-47: 2 cpu48-63: 3
which all looks nicely regular and clean.
core_siblings mirrors exactly that too, so you don't see the internal nodes from that either, i.e:
cpu0/topology/core_siblings_list:0-15 ... cpu16/topology/core_siblings_list:16-31 ... cpu32/topology/core_siblings_list:32-47 ... cpu50/topology/core_siblings_list:48-63
-- Regards/Gruss, Boris. --
| |