Date:   Thu, 29 Sep 2022 16:12:51 +0800
From:   Xing Zhengjun <zhengjun.xing@...ux.intel.com>
To:     Ian Rogers <irogers@...gle.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>
Cc:     Kan Liang <kan.liang@...ux.intel.com>,
        Andi Kleen <ak@...ux.intel.com>, perry.taylor@...el.com,
        caleb.biggers@...el.com, kshipra.bopardikar@...el.com,
        samantha.alt@...el.com, ahmad.yasin@...el.com,
        Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Mark Rutland <mark.rutland@....com>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Jiri Olsa <jolsa@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        John Garry <john.garry@...wei.com>,
        James Clark <james.clark@....com>,
        Kajol Jain <kjain@...ux.ibm.com>,
        Thomas Richter <tmricht@...ux.ibm.com>,
        Miaoqian Lin <linmq006@...il.com>,
        Florian Fischer <florian.fischer@...q.space>,
        linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org,
        Stephane Eranian <eranian@...gle.com>
Subject: Re: [PATCH v1 05/22] perf vendor events: Update Intel alderlake



On 9/29/2022 10:23 AM, Ian Rogers wrote:
> On Wed, Sep 28, 2022 at 7:08 AM Arnaldo Carvalho de Melo
> <acme@...nel.org> wrote:
>>
>> Em Wed, Sep 28, 2022 at 04:22:53PM +0800, Xing Zhengjun escreveu:
>>> On 9/28/2022 3:21 PM, Ian Rogers wrote:
>>>> diff --git a/tools/perf/pmu-events/arch/x86/mapfile.csv b/tools/perf/pmu-events/arch/x86/mapfile.csv
>>>> index 7f2d777fd97f..594c6e96f0ce 100644
>>>> --- a/tools/perf/pmu-events/arch/x86/mapfile.csv
>>>> +++ b/tools/perf/pmu-events/arch/x86/mapfile.csv
>>>> @@ -1,5 +1,5 @@
>>>>    Family-model,Version,Filename,EventType
>>>> -GenuineIntel-6-9[7A],v1.13,alderlake,core
>>>> +GenuineIntel-6-(97|9A|B7|BA|BF),v1.15,alderlake,core
>>>
>>> The commit description should mention this change "Add more CPUID support
>>> for ADL"
>>
>> I added this note, with that can I have your Reviewed-by or Acked-by?
>>
>> - Arnaldo
> 
> I think I should send a v2. I can add this then. Other things for v2 are:
> - this week there will likely be an update to [1] which is implicitly
> integrated here.
> - the topdown metrics are all percentages but are currently values
> zero to 1. It is straightforward to add a ScaleUnit of 100% to them
> all.
> - eyeballing topdown event metrics like:
> 
>      {
>         "BriefDescription": "This category represents fraction of slots
> utilized by useful work i.e. issued uops that eventually get retired",
>         "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound +
> topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound)",
>         "MetricGroup": "TopdownL1;tma_L1_group",
>         "MetricName": "tma_retiring",
> ...
> it looks like the group will fail due to the missing slots event. This
> can be fixed with:
>         "MetricExpr": "topdown\\-retiring / (topdown\\-fe\\-bound +
> topdown\\-bad\\-spec + topdown\\-retiring + topdown\\-be\\-bound) + 0
> * TOPDOWN.SLOTS",
> 
> Thanks,
> Ian

I just test all the events for SPR/ADL, no error find.
But there are a lots of metrics fail due to slots missed.

|0929_005708.420|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_branch_mispredicts" -a sleep 0.1 failed. Return code is 255|
|0929_005709.221|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_backend_bound" -a sleep 0.1 failed. Return code is 255|
|0929_005709.544|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_memory_bound" -a sleep 0.1 failed. Return code is 255|
|0929_005723.757|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_core_bound" -a sleep 0.1 failed. Return code is 255|
|0929_005724.531|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_ports_utilization" -a sleep 0.1 failed. Return code is 255|
|0929_005731.541|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_retiring" -a sleep 0.1 failed. Return code is 255|
|0929_005731.860|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_light_operations" -a sleep 0.1 failed. Return code is 255|
|0929_005732.811|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_x87_use" -a sleep 0.1 failed. Return code is 255|
|0929_005741.026|ERROR| common.sh:123:do_cmd() - perf stat -M 
"tma_heavy_operations" -a sleep 0.1 failed. Return code is 255|
|0929_005746.165|ERROR| common.sh:123:do_cmd() - perf stat -M 
"Memory_Latency" -a sleep 0.1 failed. Return code is 255|

# perf stat -v  -M "tma_branch_mispredicts" -a sleep 1
Using CPUID GenuineIntel-6-8F-3
metric expr topdown\-br\-mispredict / (topdown\-fe\-bound + 
topdown\-bad\-spec + topdown\-retiring + topdown\-be\-bound) for 
tma_branch_mispredicts
found event topdown-retiring
found event topdown-fe-bound
found event topdown-be-bound
found event topdown-br-mispredict
found event topdown-bad-spec
Parsing metric events 
'{topdown-retiring/metric-id=topdown!1retiring/,topdown-fe-bound/metric-id=topdown!1fe!1bound/,topdown-be-bound/metric-id=topdown!1be!1bound/,topdown-br-mispredict/metric-id=topdown!1br!1mispredict/,topdown-bad-spec/metric-id=topdown!1bad!1spec/}:W'
topdown-retiring -> cpu/event=0,umask=0x80/
topdown-fe-bound -> cpu/event=0,umask=0x82/
topdown-be-bound -> cpu/event=0,umask=0x83/
topdown-br-mispredict -> cpu/event=0,umask=0x85/
topdown-bad-spec -> cpu/event=0,umask=0x81/
Control descriptor is not initialized
Warning:
topdown-retiring event is not supported by the kernel.
Error:
The sys_perf_event_open() syscall returned with 22 (Invalid argument) 
for event (topdown-retiring).
/bin/dmesg | grep -i perf may provide additional information.


> 
> [1] https://github.com/intel/perfmon-metrics
> 
>>>>    GenuineIntel-6-(1C|26|27|35|36),v4,bonnell,core
>>>>    GenuineIntel-6-(3D|47),v26,broadwell,core
>>>>    GenuineIntel-6-56,v23,broadwellde,core
>>>
>>> --
>>> Zhengjun Xing
>>
>> --
>>
>> - Arnaldo

-- 
Zhengjun Xing

Powered by blists - more mailing lists