Re: [PATCH v5 00/24] Intel vendor events and TMA 5.01 metrics

["Falcon, Thomas" <thomas.falcon@intel.com>]

On Wed, 2025-02-05 at 12:47 -0800, Ian Rogers wrote:
> On Wed, Feb 5, 2025 at 11:44 AM Falcon, Thomas
> <thomas.falcon@intel.com> wrote:
> > 
> > On Wed, 2025-02-05 at 09:31 -0800, Ian Rogers wrote:
> > > Update the Intel vendor events to the latest.
> > > Update the metrics to TMA 5.01.
> > > Add Arrowlake and Clearwaterforest support.
> > > Add metrics for LNL and GNR.
> > > Address IIO uncore issue spotted on EMR, GRR, GNR, SPR and SRF.
> > > 
> > > The perf json was generated using the script:
> > > https://github.com/intel/perfmon/blob/main/scripts/create_perf_json.py
> > > with the generated json being in:
> > > https://github.com/intel/perfmon/tree/main/scripts/perf
> > > 
> > > Thanks to Perry Taylor <perry.taylor@intel.com>, Caleb Biggers
> > > <caleb.biggers@intel.com>, Edward Baker <edward.baker@intel.com>,
> > > Weilin Wang <weilin.wang@intel.com>, Kan Liang
> > > <kan.liang@linux.intel.com> and Thomas Falcon
> > > <thomas.falcon@intel.com> for helping get this patch series
> > > together.
> > > 
> > > v5: Remove Valkyrie metrics mistakenly added on atom. Ensure
> > > "perf
> > > all
> > >     metrics test" doesn't skip for missing events.
> > > v4: Fix TSC events on hybrid mistakenly specifying the core PMU
> > >     inhibiting the use of the msr PMU.
> > > v3: Fixes for hybrid metrics that were missing PMU. Update to the
> > >     latest events.
> > > v2: Fix hybrid and Co-authored-by tag issues reported by
> > >     Arnaldo. Updates to Lunarlake and Meteorlake events. Addition
> > > of
> > >     Clearwaterforest.
> > > 
> > 
> > Thanks, I'm seeing some errors during the perf all metrics test:
> > 
> > Metric 'tma_info_system_cpus_utilized' not printed in:
> > event syntax error: '{cpu_core/TSC,metric-
> > id=cpu_core!3TSC!3/,cpu_core/CPU_CL..'
> >                       \___ Bad event or PMU
> > 
> > Unable to find PMU or event on a PMU of 'cpu_core'
> > 
> > event syntax error: '{cpu_core/TSC,metric-
> > id=cpu_core!3TSC!3/,cpu_core/CPU_CL..'
> >                                \___ unknown term 'TSC' for pmu
> > 'cpu_core'
> > 
> > valid terms:
> > event,pc,edge,offcore_rsp,ldlat,inv,umask,frontend,cmask,config,con
> > fig1
> > ,config2,config3,name,period,percore,metric-id
> > 
> > event syntax error: '{cpu_core/TSC,metric-
> > id=cpu_core!3TSC!3/,cpu_core/CPU_CL..'
> >                                \___ unknown term 'TSC' for pmu
> > 'cpu_core'
> > 
> > valid terms:
> > event,pc,edge,offcore_rsp,ldlat,inv,umask,frontend,cmask,config,con
> > fig1
> > ,config2,config3,name,period,percore,metric-id
> > 
> > There are all some errors like this,
> > 
> > Testing tma_cisc
> > Metric contains missing events
> > Cannot resolve IDs for tma_cisc: cpu_atom@TOPDOWN_FE_BOUND.CISC@ /
> > (5 *
> > cpu_atom@CPU_CLK_UNHALTED.CORE@)
> > 
> 
> Thanks, I forgot to reapply the TSC fix. You can do it with
> $ find tools/perf/pmu-events/arch/x86 -type f -exec sed -i
> 's/cpu_core@TSC@/TSC/g' {} \;
> I can't eyeball this later missing events problem. What model type
> does it happen on?

Thanks, I was testing on Alder Lake, which seems ok now other than the
"Cannot resolve IDs..." errors after your TSC change. I am seeing
similar errors on an Arrow Lake though as well as these errors:

Metric 'C7_Pkg_Residency' not printed in:
event syntax error: '{cstate_pkg/c7-residency,metric-
id=cstate_pkg!3c7!1resid..'
                      \___ Bad event or PMU

Unable to find PMU or event on a PMU of 'cstate_pkg'

event syntax error: '{cstate_pkg/c7-residency,metric-
id=cstate_pkg!3c7!1resid..'
                                 \___ unknown term 'c7-residency' for
pmu 'cstate_pkg'

valid terms:
event,config,config1,config2,config3,name,period,percore,metric-id

event syntax error: '{cstate_pkg/c7-residency,metric-
id=cstate_pkg!3c7!1resid..'
                                 \___ unknown term 'c7-residency' for
pmu 'cstate_pkg'

valid terms:
event,config,config1,config2,config3,name,period,percore,metric-id
Testing C8_Pkg_Residency
Testing C9_Pkg_Residency
Metric 'C9_Pkg_Residency' not printed in:
event syntax error: '{cstate_pkg/c9-residency,metric-
id=cstate_pkg!3c9!1resid..'
                      \___ Bad event or PMU

Unable to find PMU or event on a PMU of 'cstate_pkg'

event syntax error: '{cstate_pkg/c9-residency,metric-
id=cstate_pkg!3c9!1resid..'
                                 \___ unknown term 'c9-residency' for
pmu 'cstate_pkg'

valid terms:
event,config,config1,config2,config3,name,period,percore,metric-id

event syntax error: '{cstate_pkg/c9-residency,metric-
id=cstate_pkg!3c9!1resid..'
                                 \___ unknown term 'c9-residency' for
pmu 'cstate_pkg'

valid terms:
event,config,config1,config2,config3,name,period,percore,metric-id
Testing tma_info_core_epc
Testing tma_info_system_core_frequency
Testing tma_info_system_power
Testing tma_info_system_turbo_utilization
Testing tma_info_inst_mix_ipswpf
Testing tma_info_memory_prefetches_useless_hwpf
Testing tma_info_thread_ipc
Testing tma_heavy_operations
Testing tma_light_operations
Testing UNCORE_FREQ
Metric 'UNCORE_FREQ' not printed in:
event syntax error: '{UNC_CLOCK.SOCKET/metric-
id=UNC_CLOCK.SOCKET/}:W,duratio..'
                      \___ Bad event or PMU

Unable to find PMU or event on a PMU of 'UNC_CLOCK.SOCKET'
Testing tma_info_system_socket_clks
Metric 'tma_info_system_socket_clks' not printed in:
event syntax error: '{UNC_CLOCK.SOCKET/metric-id=UNC_CLOCK.SOCKET/}:W'
                      \___ Bad event or PMU

Unable to find PMU or event on a PMU of 'UNC_CLOCK.SOCKET'

I submitted a PR to perfmon for the c7/c9-residency errors.

Tom

> Thanks,
> Ian
>