Re: [PATCH 1/1] perf/x86/intel: Add perf core PMU support for Intel Knights Landing

[Harish Chegondi <harish.chegondi@intel.com>]

On 12/09/2015 03:37 PM, Peter Zijlstra wrote:
> On Wed, Dec 09, 2015 at 03:22:29PM -0800, Harish Chegondi wrote:
>
>> On 12/08/2015 12:37 AM, Peter Zijlstra wrote:
>>> On Mon, Dec 07, 2015 at 02:28:18PM -0800, Harish Chegondi wrote:
>>>> Knights Landing core is based on Silvermont core with several differences.
>>>> Like Silvermont, Knights Landing has 8 pairs of LBR MSRs. However, the
>>>> LBR MSRs addresses match those of the Xeon cores' first 8 pairs of LBR MSRs
>>>> +/* Knights Landing */
>>>> +void intel_pmu_lbr_init_knl(void)
>>>> +{
>>>> +	x86_pmu.lbr_nr	   = 8;
>>>> +	x86_pmu.lbr_tos    = MSR_LBR_TOS;
>>>> +	x86_pmu.lbr_from   = MSR_LBR_NHM_FROM;
>>>> +	x86_pmu.lbr_to     = MSR_LBR_NHM_TO;
>>>> +
>>>> +	x86_pmu.lbr_sel_mask = LBR_SEL_MASK;
>>>> +	x86_pmu.lbr_sel_map  = snb_lbr_sel_map;
>>> Also, unlike Silvermont, this thing seems to have hardware LBR filters.
>>> So would it not be more accurate to say the KNL has a big core LBR
>>> instead? (Note that this LBR setup isn't specific to Xeon's, all of the
>>> Core chips have this, including the client parts).
>> We cannot say that KNL has a big core LBR. This is because
>> architectural MSR IA32_PERF_CAPABILITIES[5:0] which indicates the
>> format of the address that is stored in the LBR stack is different for
>> KNL (IA32_PERF_CAPABILITIES[5:0] = 0x1) and big core (for example,
>> Haswell IA32_PERF_CAPABILITIES[5:0]=0x4). Haswell LBR stack has TSX
>> info which KNL LBR stack doesn't have.
> Fair enough I suppose. Applied the patch.
> .
>
Thank you Peter!