diff mbox series

perf vendor events amd: Fix Zen 4 cache latency events

Message ID 20240301084431.646221-1-sandipan.das@amd.com
State New
Headers
Series perf vendor events amd: Fix Zen 4 cache latency events |

Commit Message

Sandipan Das March 1, 2024, 8:44 a.m. UTC 
  L3PMCx0AC and L3PMCx0AD, used in l3_xi_sampled_latency* events, have a
quirk that requires them to be programmed with SliceId set to 0x3.
Without this, the events do not count at all and affects dependent
metrics such as l3_read_miss_latency.

If ThreadMask is not specified, the amd-uncore driver internally sets
ThreadMask to 0x3, EnAllCores to 0x1 and EnAllSlices to 0x1 but does
not set SliceId. Since SliceId must also be set to 0x3 in this case,
specify all the other fields explicitly.

E.g.

  $ sudo perf stat -e l3_xi_sampled_latency.all,l3_xi_sampled_latency_requests.all -a sleep 1

Before:

   Performance counter stats for 'system wide':

                   0      l3_xi_sampled_latency.all
                   0      l3_xi_sampled_latency_requests.all

         1.005155399 seconds time elapsed

After:

   Performance counter stats for 'system wide':

             921,446      l3_xi_sampled_latency.all
              54,210      l3_xi_sampled_latency_requests.all

         1.005664472 seconds time elapsed

Fixes: 5b2ca349c313 ("perf vendor events amd: Add Zen 4 uncore events")
Signed-off-by: Sandipan Das <sandipan.das@amd.com>
---
 .../pmu-events/arch/x86/amdzen4/cache.json    | 56 +++++++++++++++++++
 tools/perf/pmu-events/jevents.py              |  4 ++
 2 files changed, 60 insertions(+)

Comments

Ian Rogers March 1, 2024, 4:39 p.m. UTC | #1 
On Fri, Mar 1, 2024 at 12:44 AM Sandipan Das <sandipan.das@amd.com> wrote:
>
> L3PMCx0AC and L3PMCx0AD, used in l3_xi_sampled_latency* events, have a
> quirk that requires them to be programmed with SliceId set to 0x3.
> Without this, the events do not count at all and affects dependent
> metrics such as l3_read_miss_latency.
>
> If ThreadMask is not specified, the amd-uncore driver internally sets
> ThreadMask to 0x3, EnAllCores to 0x1 and EnAllSlices to 0x1 but does
> not set SliceId. Since SliceId must also be set to 0x3 in this case,
> specify all the other fields explicitly.
>
> E.g.
>
>   $ sudo perf stat -e l3_xi_sampled_latency.all,l3_xi_sampled_latency_requests.all -a sleep 1
>
> Before:
>
>    Performance counter stats for 'system wide':
>
>                    0      l3_xi_sampled_latency.all
>                    0      l3_xi_sampled_latency_requests.all
>
>          1.005155399 seconds time elapsed
>
> After:
>
>    Performance counter stats for 'system wide':
>
>              921,446      l3_xi_sampled_latency.all
>               54,210      l3_xi_sampled_latency_requests.all
>
>          1.005664472 seconds time elapsed
>
> Fixes: 5b2ca349c313 ("perf vendor events amd: Add Zen 4 uncore events")
> Signed-off-by: Sandipan Das <sandipan.das@amd.com>

Reviewed-by: Ian Rogers <irogers@google.com>

Thanks,
Ian
diff mbox series

Patch

diff --git a/tools/perf/pmu-events/arch/x86/amdzen4/cache.json b/tools/perf/pmu-events/arch/x86/amdzen4/cache.json
index ecbe9660b2b3..e6d710cf3ce2 100644
--- a/tools/perf/pmu-events/arch/x86/amdzen4/cache.json
+++ b/tools/perf/pmu-events/arch/x86/amdzen4/cache.json
@@ -676,6 +676,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from DRAM in the same NUMA node.",
     "UMask": "0x01",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -683,6 +687,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from DRAM in a different NUMA node.",
     "UMask": "0x02",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -690,6 +698,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from another CCX's cache when the address was in the same NUMA node.",
     "UMask": "0x04",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -697,6 +709,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from another CCX's cache when the address was in a different NUMA node.",
     "UMask": "0x08",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -704,6 +720,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from extension memory (CXL) in the same NUMA node.",
     "UMask": "0x10",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -711,6 +731,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency when data is sourced from extension memory (CXL) in a different NUMA node.",
     "UMask": "0x20",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -718,6 +742,10 @@ 
     "EventCode": "0xac",
     "BriefDescription": "Average sampled latency from all data sources.",
     "UMask": "0x3f",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -725,6 +753,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from DRAM in the same NUMA node.",
     "UMask": "0x01",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -732,6 +764,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from DRAM in a different NUMA node.",
     "UMask": "0x02",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -739,6 +775,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from another CCX's cache when the address was in the same NUMA node.",
     "UMask": "0x04",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -746,6 +786,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from another CCX's cache when the address was in a different NUMA node.",
     "UMask": "0x08",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -753,6 +797,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from extension memory (CXL) in the same NUMA node.",
     "UMask": "0x10",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -760,6 +808,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from extension memory (CXL) in a different NUMA node.",
     "UMask": "0x20",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   },
   {
@@ -767,6 +819,10 @@ 
     "EventCode": "0xad",
     "BriefDescription": "L3 cache fill requests sourced from all data sources.",
     "UMask": "0x3f",
+    "EnAllCores": "0x1",
+    "EnAllSlices": "0x1",
+    "SliceId": "0x3",
+    "ThreadMask": "0x3",
     "Unit": "L3PMC"
   }
 ]
diff --git a/tools/perf/pmu-events/jevents.py b/tools/perf/pmu-events/jevents.py
index 2c7e5d61ce92..e42efc16723e 100755
--- a/tools/perf/pmu-events/jevents.py
+++ b/tools/perf/pmu-events/jevents.py
@@ -373,6 +373,10 @@  class JsonEvent:
         ('UMask', 'umask='),
         ('NodeType', 'type='),
         ('RdWrMask', 'rdwrmask='),
+        ('EnAllCores', 'enallcores='),
+        ('EnAllSlices', 'enallslices='),
+        ('SliceId', 'sliceid='),
+        ('ThreadMask', 'threadmask='),
     ]
     for key, value in event_fields:
       if key in jd and not is_zero(jd[key]):