Messages in this thread | | | Date | Thu, 21 Jun 2012 16:43:06 +0800 | From | "Yan, Zheng" <> | Subject | Re: [PATCH V6 0/13] perf: Intel uncore pmu counting support |
| |
On 06/21/2012 04:10 PM, Stephane Eranian wrote: > On Thu, Jun 21, 2012 at 4:34 AM, Yan, Zheng <zheng.z.yan@intel.com> wrote: >> On 06/21/2012 12:01 AM, Peter Zijlstra wrote: >>> >>> ok, so with this stuff something like: >>> >>> perf stat -ae uncore/event=clockticks/ foo >>> >>> will register a counter per cpu, which is somewhat silly since we only >>> need one per node. What would be the best way to 'fix' this? >>> >>> We could of course create another variant of -a which iterates nodes >>> instead of cpus, -N or so. >>> >>> Alternatively we could try and describe this in sysfs in some way, one >>> possibility would be to include a link to /sys/devices/system/{cpu,node} >>> or somesuch and use that link to iterate the correct space. >>> >>> Any other suggestions? >>> >> How about treat the 'cpu' parameter for uncore event as socket id instead >> of cpu id? >> > But that does not address the use case of Peter, i.e., no cpu parameter passed. > Looks like sysfs might be the only way to do this in a portable manner. > It does. For example, on a dual socket system, perf can only register uncore counter with 'cpu' parameter is equal to 0 or 1. This method is hacky, but it requires minimal change for the kernel and perf tool.
Regards Yan, Zheng
| |