Re: [PATCH v2 03/10] perf vendor events intel: Update emeraldrapids events to v1.20

[Arnaldo Carvalho de Melo <acme@kernel.org>]

On Mon, Sep 29, 2025 at 11:45:20PM +0000, Falcon, Thomas wrote:
> On Thu, 2025-09-25 at 10:27 -0700, Ian Rogers wrote:
> > Update emeraldrapids events to v1.20 released in:
> > https://github.com/intel/perfmon/commit/868b433955f3e94126420ee9374b9e0a6ce2d83e
> > https://github.com/intel/perfmon/commit/43681e2817a960d06c5b8870cc6d3e5b7b6feeb9
> > 
> > Also adds cpu_cstate_c0 and cpu_cstate_c6 metrics.
> > 
> > Event json automatically generated by:
> > https://github.com/intel/perfmon/blob/main/scripts/create_perf_json.py
> > 
> > Signed-off-by: Ian Rogers <irogers@google.com>
> 
> I found an Emerald Rapids to test this on. All metrics tests passed.

I'll take this as a Tested-by for this specific patch, ok?

- Arnaldo
 
> Thanks,
> Tom
> 
> > ---
> >  .../arch/x86/emeraldrapids/cache.json         | 63
> > +++++++++++++++++++
> >  .../arch/x86/emeraldrapids/emr-metrics.json   | 12 ++++
> >  .../arch/x86/emeraldrapids/uncore-cache.json  | 11 ++++
> >  .../arch/x86/emeraldrapids/uncore-memory.json | 22 +++++++
> >  .../arch/x86/emeraldrapids/uncore-power.json  |  2 -
> >  tools/perf/pmu-events/arch/x86/mapfile.csv    |  2 +-
> >  6 files changed, 109 insertions(+), 3 deletions(-)
> > 
> > diff --git a/tools/perf/pmu-events/arch/x86/emeraldrapids/cache.json
> > b/tools/perf/pmu-events/arch/x86/emeraldrapids/cache.json
> > index e96f938587bb..26568e4b77f7 100644
> > --- a/tools/perf/pmu-events/arch/x86/emeraldrapids/cache.json
> > +++ b/tools/perf/pmu-events/arch/x86/emeraldrapids/cache.json
> > @@ -1,4 +1,67 @@
> >  [
> > +    {
> > +        "BriefDescription": "Hit snoop reply with data, line
> > invalidated.",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.I_FWD_FE",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line will now be (I)nvalidated: removed from this core's cache,
> > after the data is forwarded back to the requestor and indicating the
> > data was found unmodified in the (FE) Forward or Exclusive State in
> > this cores caches cache.  A single snoop response from the core
> > counts on all hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x20"
> > +    },
> > +    {
> > +        "BriefDescription": "HitM snoop reply with data, line
> > invalidated.",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.I_FWD_M",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line will now be (I)nvalidated: removed from this core's caches,
> > after the data is forwarded back to the requestor, and indicating the
> > data was found modified(M) in this cores caches cache (aka HitM
> > response).  A single snoop response from the core counts on all
> > hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x10"
> > +    },
> > +    {
> > +        "BriefDescription": "Hit snoop reply without sending the
> > data, line invalidated.",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.I_HIT_FSE",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line will now be (I)nvalidated in this core's caches without
> > forwarded back to the requestor. The line was in Forward, Shared or
> > Exclusive (FSE) state in this cores caches.  A single snoop response
> > from the core counts on all hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x2"
> > +    },
> > +    {
> > +        "BriefDescription": "Line not found snoop reply",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.MISS",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > that the data was not found (IHitI) in this core's caches. A single
> > snoop response from the core counts on all hyperthreads of the
> > Core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x1"
> > +    },
> > +    {
> > +        "BriefDescription": "Hit snoop reply with data, line kept in
> > Shared state.",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.S_FWD_FE",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line may be kept on this core in the (S)hared state, after the
> > data is forwarded back to the requestor, initially the data was found
> > in the cache in the (FS) Forward or Shared state.  A single snoop
> > response from the core counts on all hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x40"
> > +    },
> > +    {
> > +        "BriefDescription": "HitM snoop reply with data, line kept
> > in Shared state",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.S_FWD_M",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line may be kept on this core in the (S)hared state, after the
> > data is forwarded back to the requestor, initially the data was found
> > in the cache in the (M)odified state.  A single snoop response from
> > the core counts on all hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x8"
> > +    },
> > +    {
> > +        "BriefDescription": "Hit snoop reply without sending the
> > data, line kept in Shared state.",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x27",
> > +        "EventName": "CORE_SNOOP_RESPONSE.S_HIT_FSE",
> > +        "PublicDescription": "Counts responses to snoops indicating
> > the line was kept on this core in the (S)hared state, and that the
> > data was found unmodified but not forwarded back to the requestor,
> > initially the data was found in the cache in the (FSE) Forward,
> > Shared state or Exclusive state.  A single snoop response from the
> > core counts on all hyperthreads of the core.",
> > +        "SampleAfterValue": "1000003",
> > +        "UMask": "0x4"
> > +    },
> >      {
> >          "BriefDescription": "L1D.HWPF_MISS",
> >          "Counter": "0,1,2,3",
> > diff --git a/tools/perf/pmu-events/arch/x86/emeraldrapids/emr-
> > metrics.json b/tools/perf/pmu-events/arch/x86/emeraldrapids/emr-
> > metrics.json
> > index af0a7dd81e93..433ae5f50704 100644
> > --- a/tools/perf/pmu-events/arch/x86/emeraldrapids/emr-metrics.json
> > +++ b/tools/perf/pmu-events/arch/x86/emeraldrapids/emr-metrics.json
> > @@ -39,6 +39,18 @@
> >          "MetricName": "cpi",
> >          "ScaleUnit": "1per_instr"
> >      },
> > +    {
> > +        "BriefDescription": "The average number of cores that are in
> > cstate C0 as observed by the power control unit (PCU)",
> > +        "MetricExpr": "UNC_P_POWER_STATE_OCCUPANCY_CORES_C0 /
> > UNC_P_CLOCKTICKS * #num_packages",
> > +        "MetricGroup": "cpu_cstate",
> > +        "MetricName": "cpu_cstate_c0"
> > +    },
> > +    {
> > +        "BriefDescription": "The average number of cores are in
> > cstate C6 as observed by the power control unit (PCU)",
> > +        "MetricExpr": "UNC_P_POWER_STATE_OCCUPANCY_CORES_C6 /
> > UNC_P_CLOCKTICKS * #num_packages",
> > +        "MetricGroup": "cpu_cstate",
> > +        "MetricName": "cpu_cstate_c6"
> > +    },
> >      {
> >          "BriefDescription": "CPU operating frequency (in GHz)",
> >          "MetricExpr": "CPU_CLK_UNHALTED.THREAD /
> > CPU_CLK_UNHALTED.REF_TSC * #SYSTEM_TSC_FREQ / 1e9",
> > diff --git a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > cache.json b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > cache.json
> > index f453202d80c2..92cf47967f0b 100644
> > --- a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-cache.json
> > +++ b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-cache.json
> > @@ -311,6 +311,17 @@
> >          "UMask": "0x2",
> >          "Unit": "CHA"
> >      },
> > +    {
> > +        "BriefDescription": "Distress signal asserted : DPT Remote",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0xaf",
> > +        "EventName": "UNC_CHA_DISTRESS_ASSERTED.DPT_NONLOCAL",
> > +        "Experimental": "1",
> > +        "PerPkg": "1",
> > +        "PublicDescription": "Distress signal asserted : DPT Remote
> > : Counts the number of cycles either the local or incoming distress
> > signals are asserted. : Dynamic Prefetch Throttle received by this
> > tile",
> > +        "UMask": "0x8",
> > +        "Unit": "CHA"
> > +    },
> >      {
> >          "BriefDescription": "Egress Blocking due to Ordering
> > requirements : Down",
> >          "Counter": "0,1,2,3",
> > diff --git a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > memory.json b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > memory.json
> > index 90f61c9511fc..30044177ccf8 100644
> > --- a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-memory.json
> > +++ b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-memory.json
> > @@ -3129,6 +3129,28 @@
> >          "PublicDescription": "Clock-Enabled Self-Refresh : Counts
> > the number of cycles when the iMC is in self-refresh and the iMC
> > still has a clock.  This happens in some package C-states.  For
> > example, the PCU may ask the iMC to enter self-refresh even though
> > some of the cores are still processing.  One use of this is for
> > Monroe technology.  Self-refresh is required during package C3 and
> > C6, but there is no clock in the iMC at this time, so it is not
> > possible to count these cases.",
> >          "Unit": "iMC"
> >      },
> > +    {
> > +        "BriefDescription": "Throttle Cycles for Rank 0",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x46",
> > +        "EventName": "UNC_M_POWER_THROTTLE_CYCLES.SLOT0",
> > +        "Experimental": "1",
> > +        "PerPkg": "1",
> > +        "PublicDescription": "Throttle Cycles for Rank 0 : Counts
> > the number of cycles while the iMC is being throttled by either
> > thermal constraints or by the PCU throttling.  It is not possible to
> > distinguish between the two.  This can be filtered by rank.  If
> > multiple ranks are selected and are being throttled at the same time,
> > the counter will only increment by 1. : Thermal throttling is
> > performed per DIMM.  We support 3 DIMMs per channel.  This ID allows
> > us to filter by ID.",
> > +        "UMask": "0x1",
> > +        "Unit": "iMC"
> > +    },
> > +    {
> > +        "BriefDescription": "Throttle Cycles for Rank 0",
> > +        "Counter": "0,1,2,3",
> > +        "EventCode": "0x46",
> > +        "EventName": "UNC_M_POWER_THROTTLE_CYCLES.SLOT1",
> > +        "Experimental": "1",
> > +        "PerPkg": "1",
> > +        "PublicDescription": "Throttle Cycles for Rank 0 : Counts
> > the number of cycles while the iMC is being throttled by either
> > thermal constraints or by the PCU throttling.  It is not possible to
> > distinguish between the two.  This can be filtered by rank.  If
> > multiple ranks are selected and are being throttled at the same time,
> > the counter will only increment by 1.",
> > +        "UMask": "0x2",
> > +        "Unit": "iMC"
> > +    },
> >      {
> >          "BriefDescription": "Precharge due to read, write,
> > underfill, or PGT.",
> >          "Counter": "0,1,2,3",
> > diff --git a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > power.json b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-
> > power.json
> > index 9482ddaea4d1..71c35b165a3e 100644
> > --- a/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-power.json
> > +++ b/tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-power.json
> > @@ -178,7 +178,6 @@
> >          "Counter": "0,1,2,3",
> >          "EventCode": "0x35",
> >          "EventName": "UNC_P_POWER_STATE_OCCUPANCY_CORES_C0",
> > -        "Experimental": "1",
> >          "PerPkg": "1",
> >          "PublicDescription": "Number of cores in C0 : This is an
> > occupancy event that tracks the number of cores that are in the
> > chosen C-State.  It can be used by itself to get the average number
> > of cores in that C-state with thresholding to generate histograms, or
> > with other PCU events and occupancy triggering to capture other
> > details.",
> >          "Unit": "PCU"
> > @@ -198,7 +197,6 @@
> >          "Counter": "0,1,2,3",
> >          "EventCode": "0x37",
> >          "EventName": "UNC_P_POWER_STATE_OCCUPANCY_CORES_C6",
> > -        "Experimental": "1",
> >          "PerPkg": "1",
> >          "PublicDescription": "Number of cores in C6 : This is an
> > occupancy event that tracks the number of cores that are in the
> > chosen C-State.  It can be used by itself to get the average number
> > of cores in that C-state with thresholding to generate histograms, or
> > with other PCU events and occupancy triggering to capture other
> > details.",
> >          "Unit": "PCU"
> > diff --git a/tools/perf/pmu-events/arch/x86/mapfile.csv
> > b/tools/perf/pmu-events/arch/x86/mapfile.csv
> > index 8daaa8f40b66..dec7bdd770cf 100644
> > --- a/tools/perf/pmu-events/arch/x86/mapfile.csv
> > +++ b/tools/perf/pmu-events/arch/x86/mapfile.csv
> > @@ -9,7 +9,7 @@ GenuineIntel-6-4F,v23,broadwellx,core
> >  GenuineIntel-6-55-[56789ABCDEF],v1.25,cascadelakex,core
> >  GenuineIntel-6-DD,v1.00,clearwaterforest,core
> >  GenuineIntel-6-9[6C],v1.05,elkhartlake,core
> > -GenuineIntel-6-CF,v1.16,emeraldrapids,core
> > +GenuineIntel-6-CF,v1.20,emeraldrapids,core
> >  GenuineIntel-6-5[CF],v13,goldmont,core
> >  GenuineIntel-6-7A,v1.01,goldmontplus,core
> >  GenuineIntel-6-B6,v1.09,grandridge,core
>