Re: [PATCH v3 2/7] perf metric: Event "Compat" value supports matching multiple identifiers

[Jing Zhang]

在 2023/6/19 下午3:07, John Garry 写道:
> On 19/06/2023 03:58, Jing Zhang wrote:
>>> +++ b/tools/perf/pmu-events/arch/arm64/arm/cmn/sys/metrics.json
>>> @@ -0,0 +1,74 @@
>>> +[
>>> +    {
>>> +        "MetricName": "slc_miss_rate",
>>> +        "BriefDescription": "The system level cache miss rate include.",
>>> +        "MetricGroup": "arm_cmn",
>>> +        "MetricExpr": "hnf_cache_miss / hnf_slc_sf_cache_access",
>>>
>>> So this expression uses event aliases hnf_cache_miss and hnf_slc_sf_cache_access - where are they defined in a JSON?
>>>
>> Hi John,
>>
>> I defined the aliases for these events in the JSON file during the RFC version. However, I later removed the alias
>> definitions for these events in subsequent versions due to the possibility of non-uniqueness and difficulty in defining
>> their EventCode. But this does not affect their usage in metrics. In other words, metrics can use the aliases without
>> defining event aliases in the JSON file.
> 
> Really? So how can we resolve the event aliases when we try to run the metric?
> 
> Please verify running these metrics with 'perf stat', like 'perf stat -v -M slc_miss_rate'
> 

Ok, it shows:
#./perf stat -v -M slc_miss_rate sleep 1

metric expr hnf_cache_miss / hnf_slc_sf_cache_access for slc_miss_rate
found event duration_time
found event hnf_slc_sf_cache_access
found event hnf_cache_miss
Parsing metric events '{hnf_slc_sf_cache_access/metric-id=hnf_slc_sf_cache_access/,hnf_cache_miss/metric-id=hnf_cache_miss/}:W,duration_time'
hnf_slc_sf_cache_access -> arm_cmn_0/type=0x5,eventid=0x2/
hnf_slc_sf_cache_access -> arm_cmn_1/type=0x5,eventid=0x2/
hnf_cache_miss -> arm_cmn_0/type=0x5,eventid=0x1/
hnf_cache_miss -> arm_cmn_1/type=0x5,eventid=0x1/
Control descriptor is not initialized
hnf_slc_sf_cache_access: 127615 1001344900 1001344900
hnf_cache_miss: 36829 1001344900 1001344900
hnf_slc_sf_cache_access: 131526 1001343540 1001343540
hnf_cache_miss: 40587 1001343540 1001343540
duration_time: 1001381687 1001381687 1001381687

 Performance counter stats for 'system wide':

           259,141      hnf_slc_sf_cache_access   #     29.9 %  slc_miss_rate
            77,416      hnf_cache_miss
     1,001,381,687 ns   duration_time

       1.001381687 seconds time elapsed



#./perf list
...
 arm_cmn_0/hnf_cache_miss/                          [Kernel PMU event]
 arm_cmn_0/hnf_slc_sf_cache_access/                 [Kernel PMU event]
...
 arm_cmn_1/hnf_cache_miss/                          [Kernel PMU event]
 arm_cmn_1/hnf_slc_sf_cache_access/                 [Kernel PMU event]
...

>>
>> In the past, I always thought that the function of the alias was to explain the meaning of these events in the perf list.
>> Or maybe I'm missing something?
> 
> Event aliases do give the ability to describe the event in perf list. But we can also run them for 'perf stat', like:
> 
> ./perf list uncore
> List of pre-defined events (to be used in -e or -M):
> 
>   uncore_cbox_0/clockticks/                          [Kernel PMU event]
>   uncore_cbox_1/clockticks/                          [Kernel PMU event]
>   uncore_imc/data_reads/                             [Kernel PMU event]
>   uncore_imc/data_writes/                            [Kernel PMU event]
>   uncore_imc/gt_requests/                            [Kernel PMU event]
>   uncore_imc/ia_requests/                            [Kernel PMU event]
>   uncore_imc/io_requests/                            [Kernel PMU event]
> 
> uncore cache:
>   unc_cbo_cache_lookup.any_es
>        [L3 Lookup any request that access cache and found line in E or S-state. Unit: uncore_cbox]
> ...
> 
> sudo ./perf stat -v -e unc_cbo_cache_lookup.any_es
> Using CPUID GenuineIntel-6-3D-4
> unc_cbo_cache_lookup.any_es -> uncore_cbox_0/event=0x34,umask=0x86/
> unc_cbo_cache_lookup.any_es -> uncore_cbox_1/event=0x34,umask=0x86/
> Control descriptor is not initialized
> ^Cunc_cbo_cache_lookup.any_es: 14361103 1853372468 1853372468
> unc_cbo_cache_lookup.any_es: 14322188 1853360415 1853360415
> 
>  Performance counter stats for 'system wide':
> 
>         14,361,103      unc_cbo_cache_lookup.any_es
>         14,322,188      unc_cbo_cache_lookup.any_es
> 
>        1.853388227 seconds time elapsed
> 

Ok, thanks. If I use events without a prefix, such as perf stat -e clockticks sleep 1, will this also work?

Thanks,
Jing


> 
> Thanks,
> John